The Yellowstone volcanic field is one of the largest and best-studied centres of rhyolitic volcanism on Earth, yet it still contains little-studied periods of activity. Such an example is the Island Park–Mount Jackson series, which erupted between the Mesa Falls and Lava Creek caldera-forming events as a series of rhyolitic domes and lavas. Here we present the first detailed characterisation of these lavas and use our findings to provide a framework for rhyolite generation in Yellowstone between 1·3 and 0·6 Ma, as well as to assess whether magmatic evolution hints at a forthcoming super-eruption. These porphyritic (15–40% crystals) lavas contain mostly sanidine and quartz with lesser amounts of plagioclase (consistent with equilibrium magmatic modelling via rhyolite-MELTS) and a complex assemblage of mafic minerals. Mineral compositions vary significantly between crystals in each unit, with larger ranges than expected from a single homogeneous population in equilibrium with its host melt. Oxygen isotopes in quartz and sanidine indicate slight depletions (δ18Omagma of 5·0–6·1‰), suggesting some contribution by localised remelting of hydrothermally altered material in the area of the previous Mesa Falls Tuff-related caldera collapse. The preservation of variable O isotopic compositions in quartz requires crystal entrainment less than a few thousand years prior to eruption. Late entrainment of rhyolitic material is supported by the occurrence of subtly older sanidines dated by single-grain 40Ar/39Ar geochronology. The eruption ages of the lavas show discrete clusters illustrating that extended quiescence (>100 kyr) in magmatic activity may be a recurring feature in Yellowstone volcanism. Ubiquitous crystal aggregates, dominated by plagioclase, pyroxene and Fe–Ti oxides, are interpreted as cumulates co-erupted with their extracted liquid. Identical crystal aggregates are found in both normal-δ18O and low-δ18O rocks from Yellowstone, indicating that common petrogenetic processes characterise both volcanic suites, including the late-stage extraction of melt from an incrementally built upper crustal mush zone.

INTRODUCTION

The largest silicic eruptions on Earth are produced from caldera-forming explosive volcanism. Although events on the scale of the 5 km3 eruption of Pinatubo in 1991 can result in measurable global climatic effects (Minnis et al., 1993), the geological record illustrates that the largest silicic eruptions can be as much as three orders of magnitude bigger (Self, 2006; Hildreth & Wilson, 2007). The catastrophic consequences emphasise the importance of a detailed scientific investigation of the mechanisms controlling the generation of these large volumes of rhyolitic magma that can culminate in so-called super-eruptions.

At large, long-lived caldera centres such as Yellowstone, the catastrophic explosive events represent only a tiny fraction of the lifespan of the volcanic centre; the majority of activity is taken up by relatively quiescent effusion of lava and intermittent minor explosive activity. Studying this effusive activity preceding and post-dating large-volume eruptions is critical to understanding the magmatic system as a whole, and can be addressed over shorter timescales via geophysical methods (Chang et al., 2007) and over longer timescales via assessment of the compositional evolution of the reservoir (Vazquez et al., 2009). Such an assessment includes questions of how such huge volumes of magma are generated, stored in the crust and erupted, with a particular focus on the processes and temporal relations in the generation of these magmas.

The formation of rhyolites in the Yellowstone province remains a topic of active debate with two proposed end-member models: (1) a fractionation-dominated evolution with subordinate assimilation in an upper crustal mush zone (Vazquez & Reid, 2002; Vazquez et al., 2009; Girard & Stix, 2010; Stelten et al., 2015); (2) bulk or partial crustal melting initiated by intrusion of basaltic and/or silicic magma as a heat source with limited mass contribution (Bindeman & Valley, 2001; Bindeman et al., 2008; Simakin & Bindeman, 2012). Whereas the first model is mainly based on mineralogical and geochronological studies of young Yellowstone lavas, the second scenario has received significant support from oxygen isotope studies. These have revealed rhyolitic lavas with unusual δ18O characteristics and isotopically diverse phenocrysts, suggesting melting of hydrothermally altered lithologies in the shallow crust. In this study, we evaluate the applicability of these models to the record of lavas erupted prior to Yellowstone’s latest super-eruption at 0·631 Ma (Matthews et al., 2015). The Island Park–Mount Jackson rhyolite series represents a prime example of a series of effusive events that preceded an explosive eruption, but has as yet received little attention. By coupling major and trace elements in whole-rock, glasses and minerals to oxygen and Pb isotopic data and 40Ar/39Ar ages, we obtain a detailed petrological and geochemical characterisation of these lavas to provide a framework for rhyolite generation in Yellowstone and to test whether the mineralogical and geochemical records might be hinting at a forthcoming super-eruption.

GEOLOGICAL BACKGROUND

The Yellowstone volcanic field represents the current focus of volcanism, which has been continuing for at least 16·5 Myr covering vast areas of the NW of the USA. The Columbia River–Snake River Plain–Yellowstone province is compositionally bimodal (basalt–rhyolite) and is most commonly interpreted as the result of a hotspot (Geist & Richards, 1993; Hooper et al., 2007; Wolff et al., 2008). Silicic volcanism in the province was initially widely dispersed (e.g. Coble & Mahood, 2012), prior to becoming focused along the track of the Snake River Plain from around 14 Ma. Here, numerous voluminous ignimbrites and lavas were produced, with more frequent explosive eruptions and higher magmatic temperatures compared with the younger part of the province (Cathey & Nash, 2009; Ellis et al., 2013).

Silicic volcanism at Yellowstone has been continuing for the past 2 Myr and is characterised by large-volume, explosive eruptions separated by periods of relative quiescence during which effusion of rhyolitic lava dominates (Christiansen, 1984, 2001). The explosive eruptions consist of the Huckleberry Ridge Tuff (HRT) at 2·1 Ma (Rivera et al., 2014), the Mesa Falls Tuff (MFT) at 1·3 Ma (Lanphere et al., 2002), and the Lava Creek Tuff (LCT) at 0·6 Ma (Lanphere et al., 2002; Matthews et al., 2015; Wotzlaw et al., 2015). These voluminous deposits provide regionally significant stratigraphic markers and separate Yellowstone volcanism into three volcanic cycles (Christiansen, 2001).

Following the early studies, which considered Yellowstone volcanism as a whole (e.g. Christiansen & Blank, 1972; Doe et al., 1982; Hildreth et al., 1984, 1991; Christiansen, 2001), more recent work has focused separately on the large explosive events and the youngest volcanism after the LCT, namely the Upper Basin Member and Central Plateau Member (e.g. Obradovich, 1992; Gansecki et al., 1996; Bindeman & Valley, 2001; Lanphere et al., 2002; Vazquez & Reid, 2002; Bindeman et al., 2008; Girard & Stix, 2009, 2010; Vazquez et al., 2009; Ellis et al., 2012; Watts et al., 2012). To improve our understanding of the early history of Yellowstone, we investigate the less-documented Island Park domes and Mount Jackson lavas, which were erupted between the MFT and the LCT, thus spanning the second and third volcanic cycles in Yellowstone (Christiansen, 2001).

The Island Park–Mount Jackson (IPMJ) rhyolite series

The Island Park (IP) domes include the undated units Silver Lake dome, Lookout Butte and Elk Butte, as well as Osbourne Butte (1·28 ± 0·01 Ma) and Warm River Butte (1·24 ± 0·02 Ma) [all K–Ar ages by Obradovich (1992)]. These domes are located in a NW–SE-trending zone inferred to be tectonically controlled (Christiansen, 2001), with only the Lookout Butte dome erupted along what is thought to be the caldera related to the MFT (Fig. 1). Historically, the notably small-volume IP domes are thought to represent post-caldera units relating to the preceding MFT-related caldera collapse, whereas the volumetrically larger Mount Jackson (MJ) rhyolite lava flows are interpreted as pre-caldera units of the following LCT eruption (Christiansen, 2001). Units from the MJ series are widely dispersed across the Yellowstone volcanic field and include the Moose Creek flow (1·22 ± 0·01 Ma), Wapiti Lake flow (1·16 ± 0·01 Ma), Flat Mountain Rhyolite (0·929 ± 0·034 Ma), Lewis Canyon Rhyolite (0·853 ± 0·007 Ma), Harlequin Lake flow (0·839 ± 0·008 Ma), Big Bear Lake flow (undated) and Mount Haynes Rhyolite (0·609 ± 0·006 Ma) [all K–Ar ages by Obradovich (1992)]. These lava flows were proposed to trace an arcuate fault structure, which was later exploited during the LCT eruption (Christiansen et al., 2007). Although the origin of such an arcuate structure prior to caldera collapse remains unclear, it could have been the result of piecemeal, caldera-related faulting during the eruption of the earlier HRT. Notably, the MJ rhyolites are morphologically similar to the voluminous rhyolite lavas of the Plateau Member, whereas rhyolites from the IP series are much smaller and have dome-like structures (Fig. 1). Within the MJ lavas lies the Lewis Canyon Rhyolite, which is inferred to represent an unknown number of lavas that are petrologically dissimilar to the other MJ lavas in containing significantly more plagioclase (Christiansen & Blank, 1972; Christiansen, 2001). We therefore mention this unit separately in the results and discussion sections.

Fig. 1

Distribution of IPMJ rhyolites in the Yellowstone area after the geological map of Christiansen (2001). IP domes are shown in purple, MJ lavas in green and the Lewis Canyon (LC) Rhyolite in dark green. Dashed red lines indicate approximate caldera outlines related to eruption of HRT (I), MFT (II) and LCT (III). Black dashed line shows the borders of Yellowstone National Park. Upper left inset shows unit abbreviations and affiliation of units with either the Island Park or Mount Jackson series. Upper right inset shows the location of the Yellowstone volcanic province within North America; dark grey circles mark eruptive centres along the Yellowstone hotspot track.

Fig. 1

Distribution of IPMJ rhyolites in the Yellowstone area after the geological map of Christiansen (2001). IP domes are shown in purple, MJ lavas in green and the Lewis Canyon (LC) Rhyolite in dark green. Dashed red lines indicate approximate caldera outlines related to eruption of HRT (I), MFT (II) and LCT (III). Black dashed line shows the borders of Yellowstone National Park. Upper left inset shows unit abbreviations and affiliation of units with either the Island Park or Mount Jackson series. Upper right inset shows the location of the Yellowstone volcanic province within North America; dark grey circles mark eruptive centres along the Yellowstone hotspot track.

METHODS

Major and trace element analyses

Samples were analysed for bulk geochemistry via X-ray fluorescence (XRF) and inductively coupled plasma mass spectrometry (ICP-MS) at the GeoAnalytical Laboratory at Washington State University, following the procedures described by Johnson et al. (1999). For mineral separates, rocks were fragmented using a SELFRAG at ETH Zurich. Following cleaning, samples were separated further if necessary via heavy liquid separation and minerals were picked by hand under a binocular microscope.

Electron microprobe (EMP) analyses of mafic minerals were performed at the University of Kiel (Germany) with a Jeol JXA 8900 R electron microprobe; 20–30 grains per unit were measured for rim and core composition. Fayalite was analysed at 20 keV and 15 nA, pyroxene and hornblende at 15 keV and 15 nA. Standard measurements for all EMP and laser ablation (LA)-ICP-MS analyses can be found in the

(available for downloading at http://www.petrology.oxfordjournals.org).

Cathodoluminescence (CL) images of 25–30 epoxy-mounted quartz grains per sample were obtained at the Scientific Centre for Optical and Electron Microscopy (ScopeM) at ETH Zurich on a FEI Quanta 200 scanning electron microscope. Single grains were evaluated for zonation grade (strongly zoned, gradual changes or unzoned) and direction (unidirectional bright core–dark rim, dark core–bright rim or oscillatory zoning). Similarly, sanidine grains from five representative units (IP units Osbourne Butte and Silver Lake dome; MJ units Moose Creek flow, Lewis Canyon Rhyolite and Mt Haynes Rhyolite) were imaged for CL prior to trace element analysis.

Analyses of feldspars were performed at ETH Zurich, using a Jeol JXA 8200 electron microprobe at 15 keV and 15 nA. On all grains a minimum of two points were analysed to obtain rim and core compositions. Beam diameter was set to 10 µm and counting times were shortened for Na and K to avoid element mobility. Glass shards were analysed with a beam of 20 µm and 12 nA at 15 kV.

A 193 nm Resonetics Resolution 155 excimer laser ablation system coupled to a Thermo Element XR sector field mass spectrometer was used to determine trace elements in quartz, feldspar and glass at ETH Zurich. Spot sizes were 43 or 67 µm and samples were ablated using 5 Hz for 40 s after 30 s of gas blank acquisition. NIST 612 was used as a primary standard and GSD-1G as a secondary standard for quality control. Data were reduced using SILLS software (Guillong et al., 2008) with SiO2 contents from previous EMP analyses used as an internal standard (or taken as 100% for quartz). Trace element determinations are considered to be precise and accurate within 5% of the reported values, based on long-term reproducibility of a variety of glass standards.

40Ar/39Ar dating

For 40Ar/39Ar geochronology, sanidine separates were handpicked under a binocular microscope from the 0·5–1 mm size fraction, with care taken to avoid inclusions of glass or minerals and visible alteration. Samples were tested with the method of Hynek et al. (2011) to ensure a pure sanidine separate, before being cleaned and packed for irradiation at the CLICIT facility of the OSU reactor. Sixteen to 31 crystals per unit were placed into high-purity Al irradiation disks with samples of Alder Creek (ACs, age 1·2056 Ma) and Fish Canyon sanidine (FCs-EK, age 28·294 Ma; Morgan et al., 2014). Most samples were irradiated for 1 h in June 2013, but were not run due to laboratory issues. They were re-irradiated (along with the same monitors) for 1 h in November 2014 and analysed in January 2015. The Wapiti Lake sample was irradiated for 2 h in October 2014 and analysed in January 2015 (note this sample was not a re-irradiation). Following irradiation and cooling, sanidine samples were fused using a CO2 laser. Extracted gases were subjected to 300 s of purification by exposure to two SAES GP50 getters at room temperature and at 450°C [for full system technical details see Mark et al. (2012)]. Isotope measurements were made using a MAP 215-50 noble gas spectrometer.

All Ar isotope data were corrected for backgrounds (average ± standard deviation from entire run sequence), mass discrimination (calculated from air calibration shots of 7·32 × 10–14 moles 40Ar), and reactor-produced nuclides, before being processed using standard data reduction protocols (Mark et al., 2005). Data are reported according to the criteria of Renne et al. (2009) and relative to the optimization model of (Renne et al. 2010, 2011). We employed the atmospheric argon isotope ratios of Lee et al. (2006), which have been independently verified by Mark et al. (2011), and are consistent with inverse isochron plots for single units. The eruption age of each unit is calculated from the weighted means of eruption-age total fusion ages of single sanidines and is reported as X ± Y/Z (2σ confidence level), where Y is the analytical uncertainty and Z is the full external precision, including uncertainty from the decay constant.

Isotopic analyses

Lead isotope compositions in sanidine crystals were measured using laser ablation multicollector inductively coupled plasma mass spectrometry (LA-MC-ICP-MS), with a Photon Machines G2 excimer laser ablation system and NuPlasma MC-ICP-MS system at Oregon State University. Analytical techniques follow those given by Kent (2008). Analyses were made using a single 85 µm laser spot translated at 5 µm s–1 and a pulse frequency of 7 Hz on each grain. Measured ratios were corrected for mass bias based on measurement of NIST-612 standard glass at similar ablation conditions throughout the analysis session. Repeated analyses on standard reference BCR-2G glasses were typically within 0·01–0·22% of the accepted values.

Quartz and sanidine separates for O isotopic determinations were cleaned using dilute HNO3 to remove any trace of adhering groundmass. Aliquots of 1–2 mg (typically 2–4 grains) were then analysed by CO2-laser fluorination at the University of Oregon using BrF5 as reagent, a Hg diffusion pump fluorine getter and a MAT253 mass spectrometer (Bindeman et al., 2008). The long-term reproducibility of the standards is ± 0·07‰ (1SD) during runs where standards are run concurrently with the unknowns and ±0·2‰ (1SD) during airlock runs where samples are run one by one.

RESULTS

Sample description

Samples were collected from the rhyolitic domes of the Island Park (IP) series (Osbourne Butte, Lookout Butte, Warm River Butte, Elk Butte and Silver Lake dome), as well as from the rhyolitic lava flows of the Mount Jackson (MJ) Member (Moose Creek flow, Wapiti Lake flow, Harlequin Lake flow and Mt Haynes flow), and the Lewis Canyon Rhyolite. The groundmass in the IPMJ (Island Park–Mount Jackson) rhyolites is microcrystalline, often displaying spherulitic textures and flow banding. The IP units Silver Lake dome, Osbourne Butte, Elk Butte and Lookout Butte are only partially devitrified and still contain some glass. The IPMJ rhyolites are porphyritic with the IP rhyolites being more crystal-rich (28–39%) than the MJ rhyolites (13–19%). The mineral assemblage consists of ∼6–26% sanidine, ∼4–17% quartz, ∼2–7% plagioclase, and ∼1–3% Fe–Ti oxides and mafic minerals, such as clinopyroxene, orthopyroxene, amphibole and fayalite (Table 1

Table 1

Modal mineral abundances in per cent determined via point counting on thin sections

 Unit
 
 WR MC SL OB EB LB WL HL LC MH 
Devitrified 
Matrix 70 ·7 69 ·4 61 ·2 60 ·9 71 ·8 67 ·4 86 ·5 81 ·6 82 ·2 81 ·1 
Quartz 11 ·8 16 ·6 8 ·2 8 ·0 7 ·1 11 ·3 4 ·1 9 ·1 3 ·9 5 ·8 
Sanidine 11 ·4 8 ·8 25 ·5 19 ·8 16 ·3 14 ·8 7 ·0 5 ·5 5 ·9 10 ·1 
Plagioclase 3 ·9 3 ·2 2 ·5 7 ·2 4 ·3 4 ·5 1 ·8 3 ·1 7 ·1 2 ·5 
Fe–Ti oxides 1 ·2 1 ·0 1 ·1 1 ·6 0 ·3 1 ·3 0 ·5 0 ·6 0 ·4 0 ·3 
Mafic minerals* 1 ·0 0 ·9 1 ·4 2 ·5 0 ·3 0 ·8 0 ·1 0 ·1 0 ·6 0 ·3 
Accessories <0 ·1 <0 ·1 <0 ·1 <0 ·1 <0 ·1 <0 ·1 <0 ·1 <0 ·1 <0 ·1 <0 ·1 
Mafic minerals          
Clinopyroxene 20 ·3 61 ·8 73 ·6 22 ·5 34 67 ·1 93 ·3 
Orthopyroxene 28 ·6 28 ·4 15 ·4 24 ·2 10 18 ·4 100 100 25 6 ·7 
Fayalite 66 ·7 16 ·2 19 ·5 62 ·5 27 ·2 7 ·9 
Fe-hornblende 4 ·8 35 ·1 3 ·3 2 ·2 20 ·4 
 Unit
 
 WR MC SL OB EB LB WL HL LC MH 
Devitrified 
Matrix 70 ·7 69 ·4 61 ·2 60 ·9 71 ·8 67 ·4 86 ·5 81 ·6 82 ·2 81 ·1 
Quartz 11 ·8 16 ·6 8 ·2 8 ·0 7 ·1 11 ·3 4 ·1 9 ·1 3 ·9 5 ·8 
Sanidine 11 ·4 8 ·8 25 ·5 19 ·8 16 ·3 14 ·8 7 ·0 5 ·5 5 ·9 10 ·1 
Plagioclase 3 ·9 3 ·2 2 ·5 7 ·2 4 ·3 4 ·5 1 ·8 3 ·1 7 ·1 2 ·5 
Fe–Ti oxides 1 ·2 1 ·0 1 ·1 1 ·6 0 ·3 1 ·3 0 ·5 0 ·6 0 ·4 0 ·3 
Mafic minerals* 1 ·0 0 ·9 1 ·4 2 ·5 0 ·3 0 ·8 0 ·1 0 ·1 0 ·6 0 ·3 
Accessories <0 ·1 <0 ·1 <0 ·1 <0 ·1 <0 ·1 <0 ·1 <0 ·1 <0 ·1 <0 ·1 <0 ·1 
Mafic minerals          
Clinopyroxene 20 ·3 61 ·8 73 ·6 22 ·5 34 67 ·1 93 ·3 
Orthopyroxene 28 ·6 28 ·4 15 ·4 24 ·2 10 18 ·4 100 100 25 6 ·7 
Fayalite 66 ·7 16 ·2 19 ·5 62 ·5 27 ·2 7 ·9 
Fe-hornblende 4 ·8 35 ·1 3 ·3 2 ·2 20 ·4 

*Owing to the scarcity of mafic minerals in thin section, mineral abundances of different mafic mineral groups were estimated by number of grains picked after density separation.

WR, Warm River Butte; MC, Moose Creek flow; SL, Silver Lake dome; OB, Osbourne Butte; EB, Elk Butte; LB, Lookout Butte; WL, Wapiti Lake flow; HL, Harlequin Lake flow; LC, Lewis Canyon Rhyolite; MH, Mt Haynes Rhyolite. D, fully devitrified; P, partially devitrified, still contains glass shards.

). Biotite, zircon and allanite occur as accessory phases. Magnetite and ilmenite are usually exsolved. Some units (Warm River Butte and Lookout Butte) contain blebs with myrmekitic intergrowth of K-feldspar and quartz. The Lewis Canyon Rhyolite is petrographically distinct and shows abundant plagioclase (∼7%), with a concurrent decrease in the amount of sanidine and quartz.

Bulk-rock and glass compositions

All IPMJ samples are rhyolitic in composition with the majority being high-silica (>75 wt % SiO2 on an anhydrous basis) rhyolites (Table 2

Table 2

Bulk compositions of IPMJ lavas via XRF and ICP-MS

Unit: Warm River Silver Lake Osbourne Elk Lookout Moose Creek Wapiti Lake Lewis Harlequin Mt Haynes 
 Butte dome Butte Butte Butte flow flow Canyon Rhy. Lake flow 
Series: IP IP IP IP IP MJ MJ MJ MJ MJ 
Sample: BE SRP BE SRP BE SRP BE SRP BE SRP BE SRP JT 13 019 BE SRP BE SRP BE SRP 
 12 027 12 018 12 022 12 024 12 025 12 001  12 043 12 032 12 036 
Major elements via XRF, normalised to 100% anhydrous (wt %) 
SiO2 76 ·11 75 ·54 73 ·97 78 ·43 73 ·54 77 ·73 77 ·77 75 ·74 76 ·28 77 ·65 
TiO2 0 ·17 0 ·20 0 ·22 0 ·11 0 ·24 0 ·13 0 ·12 0 ·22 0 ·21 0 ·14 
Al2O3 13 ·29 13 ·08 13 ·70 11 ·69 13 ·71 12 ·09 12 ·04 12 ·66 12 ·40 12 ·02 
FeO* 1 ·76 2 ·01 2 ·00 1 ·19 2 ·19 1 ·30 1 ·15 2 ·00 1 ·73 1 ·31 
MnO 0 ·05 0 ·04 0 ·04 0 ·02 0 ·03 0 ·03 0 ·01 0 ·03 0 ·02 0 ·02 
MgO 0 ·05 0 ·08 0 ·12 0 ·04 0 ·07 0 ·03 0 ·01 0 ·04 0 ·12 0 ·05 
CaO 0 ·46 0 ·71 0 ·95 0 ·48 0 ·79 0 ·43 0 ·36 0 ·61 0 ·53 0 ·32 
Na23 ·13 3 ·28 3 ·51 3 ·15 3 ·63 3 ·26 3 ·24 3 ·61 3 ·40 3 ·23 
K24 ·95 5 ·04 5 ·45 4 ·85 5 ·65 4 ·97 5 ·30 5 ·06 5 ·28 5 ·24 
P2O5 0 ·02 0 ·03 0 ·03 0 ·04 0 ·14 0 ·02 0 ·01 0 ·03 0 ·02 0 ·01 
Trace elements via LA-ICP-MS (ppm) 
La 88 85 96 71 83 75 61 56 79 53 
Ce 163 156 172 141 139 138 109 130 152 93 
Pr 19 17 20 14 16 15 12 12 16 10 
Nd 64 60 67 49 55 49 41 43 58 33 
Sm 13 12 13 10 10 10 12 
Eu 0 ·9 1 ·3 1 ·6 0 ·5 1 ·9 0 ·5 0 ·4 1 ·6 1 ·1 0 ·4 
Gd 10 ·6 10 ·1 11 ·6 9 ·7 8 ·7 9 ·0 8 ·0 8 ·0 10 ·6 6 ·2 
Tb 1 ·9 1 ·8 2 ·0 1 ·8 1 ·4 1 ·7 1 ·5 1 ·5 1 ·9 1 ·2 
Dy 11 ·3 10 ·6 11 ·9 11 ·5 8 ·6 10 ·3 9 ·8 9 ·7 11 ·8 7 ·6 
Ho 2 ·2 2 ·1 2 ·4 2 ·3 1 ·7 2 ·1 2 ·1 2 ·0 2 ·4 1 ·6 
Er 5 ·6 5 ·6 6 ·4 6 ·5 4 ·6 5 ·6 5 ·7 5 ·5 6 ·5 4 ·5 
Tm 0 ·9 0 ·8 0 ·9 1 ·0 0 ·7 0 ·8 0 ·9 0 ·8 1 ·0 0 ·7 
Yb 5 ·4 5 ·1 5 ·7 6 ·3 4 ·4 5 ·3 5 ·4 5 ·2 5 ·9 4 ·8 
Lu 0 ·8 0 ·8 0 ·9 0 ·9 0 ·7 0 ·8 0 ·8 0 ·8 0 ·9 0 ·7 
Ba 420 622 789 207 1098 180 141 901 575 160 
Th 34 30 29 31 28 33 35 26 29 31 
Nb 49 44 42 43 42 48 50 46 50 52 
48 53 60 55 44 50 53 50 62 35 
Hf 8 ·5 8 ·7 9 ·1 6 ·4 9 ·9 7 ·0 7 ·3 10 ·0 9 ·3 7 ·2 
Ta 3 ·7 3 ·2 3 ·0 3 ·5 2 ·9 3 ·8 4 ·0 3 ·2 3 ·5 4 ·0 
5 ·7 5 ·0 5 ·3 7 ·1 4 ·6 7 ·5 7 ·6 5 ·1 6 ·1 7 ·3 
Pb 32 28 30 30 26 33 25 28 26 29 
Rb 121 147 162 175 141 182 230 166 187 223 
Cs 2 ·0 2 ·3 2 ·7 3 ·2 1 ·7 3 ·6 2 ·5 1 ·8 2 ·6 2 ·8 
Sr 25 37 48 16 61 13 10 53 29 
Sc 3 ·2 3 ·3 3 ·5 1 ·7 3 ·5 1 ·7 1 ·8 3 ·0 2 ·3 1 ·8 
Zr 254 277 303 164 350 181 192 345 290 188 
Zr sat. T (°C)1 843 843 843 795 856 804 807 858 842 806 
Zr sat. T (°C)2 814  ± 17 810  ± 19 806  ± 21 753  ± 17 820  ± 22 763  ± 18 766  ± 18 824  ± 21 805  ± 20 766  ± 18 
Unit: Warm River Silver Lake Osbourne Elk Lookout Moose Creek Wapiti Lake Lewis Harlequin Mt Haynes 
 Butte dome Butte Butte Butte flow flow Canyon Rhy. Lake flow 
Series: IP IP IP IP IP MJ MJ MJ MJ MJ 
Sample: BE SRP BE SRP BE SRP BE SRP BE SRP BE SRP JT 13 019 BE SRP BE SRP BE SRP 
 12 027 12 018 12 022 12 024 12 025 12 001  12 043 12 032 12 036 
Major elements via XRF, normalised to 100% anhydrous (wt %) 
SiO2 76 ·11 75 ·54 73 ·97 78 ·43 73 ·54 77 ·73 77 ·77 75 ·74 76 ·28 77 ·65 
TiO2 0 ·17 0 ·20 0 ·22 0 ·11 0 ·24 0 ·13 0 ·12 0 ·22 0 ·21 0 ·14 
Al2O3 13 ·29 13 ·08 13 ·70 11 ·69 13 ·71 12 ·09 12 ·04 12 ·66 12 ·40 12 ·02 
FeO* 1 ·76 2 ·01 2 ·00 1 ·19 2 ·19 1 ·30 1 ·15 2 ·00 1 ·73 1 ·31 
MnO 0 ·05 0 ·04 0 ·04 0 ·02 0 ·03 0 ·03 0 ·01 0 ·03 0 ·02 0 ·02 
MgO 0 ·05 0 ·08 0 ·12 0 ·04 0 ·07 0 ·03 0 ·01 0 ·04 0 ·12 0 ·05 
CaO 0 ·46 0 ·71 0 ·95 0 ·48 0 ·79 0 ·43 0 ·36 0 ·61 0 ·53 0 ·32 
Na23 ·13 3 ·28 3 ·51 3 ·15 3 ·63 3 ·26 3 ·24 3 ·61 3 ·40 3 ·23 
K24 ·95 5 ·04 5 ·45 4 ·85 5 ·65 4 ·97 5 ·30 5 ·06 5 ·28 5 ·24 
P2O5 0 ·02 0 ·03 0 ·03 0 ·04 0 ·14 0 ·02 0 ·01 0 ·03 0 ·02 0 ·01 
Trace elements via LA-ICP-MS (ppm) 
La 88 85 96 71 83 75 61 56 79 53 
Ce 163 156 172 141 139 138 109 130 152 93 
Pr 19 17 20 14 16 15 12 12 16 10 
Nd 64 60 67 49 55 49 41 43 58 33 
Sm 13 12 13 10 10 10 12 
Eu 0 ·9 1 ·3 1 ·6 0 ·5 1 ·9 0 ·5 0 ·4 1 ·6 1 ·1 0 ·4 
Gd 10 ·6 10 ·1 11 ·6 9 ·7 8 ·7 9 ·0 8 ·0 8 ·0 10 ·6 6 ·2 
Tb 1 ·9 1 ·8 2 ·0 1 ·8 1 ·4 1 ·7 1 ·5 1 ·5 1 ·9 1 ·2 
Dy 11 ·3 10 ·6 11 ·9 11 ·5 8 ·6 10 ·3 9 ·8 9 ·7 11 ·8 7 ·6 
Ho 2 ·2 2 ·1 2 ·4 2 ·3 1 ·7 2 ·1 2 ·1 2 ·0 2 ·4 1 ·6 
Er 5 ·6 5 ·6 6 ·4 6 ·5 4 ·6 5 ·6 5 ·7 5 ·5 6 ·5 4 ·5 
Tm 0 ·9 0 ·8 0 ·9 1 ·0 0 ·7 0 ·8 0 ·9 0 ·8 1 ·0 0 ·7 
Yb 5 ·4 5 ·1 5 ·7 6 ·3 4 ·4 5 ·3 5 ·4 5 ·2 5 ·9 4 ·8 
Lu 0 ·8 0 ·8 0 ·9 0 ·9 0 ·7 0 ·8 0 ·8 0 ·8 0 ·9 0 ·7 
Ba 420 622 789 207 1098 180 141 901 575 160 
Th 34 30 29 31 28 33 35 26 29 31 
Nb 49 44 42 43 42 48 50 46 50 52 
48 53 60 55 44 50 53 50 62 35 
Hf 8 ·5 8 ·7 9 ·1 6 ·4 9 ·9 7 ·0 7 ·3 10 ·0 9 ·3 7 ·2 
Ta 3 ·7 3 ·2 3 ·0 3 ·5 2 ·9 3 ·8 4 ·0 3 ·2 3 ·5 4 ·0 
5 ·7 5 ·0 5 ·3 7 ·1 4 ·6 7 ·5 7 ·6 5 ·1 6 ·1 7 ·3 
Pb 32 28 30 30 26 33 25 28 26 29 
Rb 121 147 162 175 141 182 230 166 187 223 
Cs 2 ·0 2 ·3 2 ·7 3 ·2 1 ·7 3 ·6 2 ·5 1 ·8 2 ·6 2 ·8 
Sr 25 37 48 16 61 13 10 53 29 
Sc 3 ·2 3 ·3 3 ·5 1 ·7 3 ·5 1 ·7 1 ·8 3 ·0 2 ·3 1 ·8 
Zr 254 277 303 164 350 181 192 345 290 188 
Zr sat. T (°C)1 843 843 843 795 856 804 807 858 842 806 
Zr sat. T (°C)2 814  ± 17 810  ± 19 806  ± 21 753  ± 17 820  ± 22 763  ± 18 766  ± 18 824  ± 21 805  ± 20 766  ± 18 
), except for Osbourne Butte and Lookout Butte (74·0 and 73·5 wt % SiO2 respectively). Like most Yellowstone rhyolites, the IPMJ rhyolites are relatively potassic with K2O and Na2O typically ranging around 5·2 wt % and 3·3 wt % respectively.

Only the IP domes and the Wapiti Lake flow contain glass; in all other flows the groundmass is microcrystalline. The glass contains ∼76–78 wt % SiO2 (recalculated anhydrous) in the IP domes and ∼78 wt % SiO2 in the Wapiti Lake flow, with original totals of ∼96–99 wt % (

). Glass shards contain high amounts of Rb (100–640 ppm) and Zr (120–160 ppm) and notably low contents of Sr (<10 ppm) and Eu. The rare earth element (REE) patterns of the glasses mimic those of the bulk-rock samples, with both MJ and IP rhyolites exhibiting ‘seagull-shaped’ REE patterns with a deep negative Eu anomaly and relatively high concentrations in light REE (LREE) and heavy REE (HREE) (Fig. 2). These signatures are typical of rhyolites from the Snake River Plain–Yellowstone volcanic suite (Ellis et al., 2013) and characteristic for hot and dry volcanic settings (Bachmann & Bergantz, 2008).
Fig. 2

Trace element and REE contents normalised to Bulk Silicate Earth (BSE) composition (McDonough & Sun, 1995) for bulk-rocks and glasses (grey field), in units which contain glass (compare Table 1). Note the strong depletions in Ba, Sr and Eu, suggestive of extensive plagioclase fractionation. Depletions in glass are more pronounced owing to the presence of sanidine and plagioclase in the whole-rock.

Fig. 2

Trace element and REE contents normalised to Bulk Silicate Earth (BSE) composition (McDonough & Sun, 1995) for bulk-rocks and glasses (grey field), in units which contain glass (compare Table 1). Note the strong depletions in Ba, Sr and Eu, suggestive of extensive plagioclase fractionation. Depletions in glass are more pronounced owing to the presence of sanidine and plagioclase in the whole-rock.

40Ar/39Ar geochronology

Single-crystal 40Ar/39Ar ages for the IPMJ series range between 1·2943 ± 0·0020/0·0026 and 1·2190 ± 0·0144/0·0146 Ma for the IP domes and between 1·2856 ± 0·0064/0·0066 and 0·7016 ± 0·0014/0·0016 Ma for the MJ rhyolites. Thus, both series overlap by 0·066 Ma. All ages with their respective uncertainties are presented in Table 3

Table 3

40Ar/39Ar eruption ages

Unit Abbr. Series 40Ar/39Ar Analytical Full external MSWD Non-eruption 
   age (Ma) uncertainty precision  age grains 
    2σ 2σ  nnon-eruptive/ntotal 
Warm River Butte WR IP 1 ·2943 0 ·0020 0 ·0026 1 ·44 7/30 
Moose Creek flow MC MJ 1 ·2856 0 ·0064 0 ·0066 0 ·57 0/16 
Silver Lake dome SL IP 1 ·2839 0 ·0112 0 ·0114 0 ·88 0/28 
Osbourne Butte OB IP 1 ·2784 0 ·0052 0 ·0054 1 ·12 1/27 
Elk Butte EC IP 1 ·2777 0 ·0090 0 ·0092 1 ·01 0/29 
Lookout Butte LB IP 1 ·2190 0 ·0144 0 ·0146 0 ·9 9/19 
Wapiti Lake flow WL MJ 1 ·2187 0 ·0158 0 ·0160 0 ·3 0/31 
Harlequin Lake flow HL MJ 0 ·8300 0 ·0072 0 ·0074 1 ·09 3/29 
Lewis Canyon Rhyolite LC MJ 0 ·8263 0 ·0184 0 ·0186 0 ·58 10/27 
Mount Haynes Rhyolite MH MJ 0 ·7016 0 ·0014 0 ·0016 1 ·04 3/29 
Unit Abbr. Series 40Ar/39Ar Analytical Full external MSWD Non-eruption 
   age (Ma) uncertainty precision  age grains 
    2σ 2σ  nnon-eruptive/ntotal 
Warm River Butte WR IP 1 ·2943 0 ·0020 0 ·0026 1 ·44 7/30 
Moose Creek flow MC MJ 1 ·2856 0 ·0064 0 ·0066 0 ·57 0/16 
Silver Lake dome SL IP 1 ·2839 0 ·0112 0 ·0114 0 ·88 0/28 
Osbourne Butte OB IP 1 ·2784 0 ·0052 0 ·0054 1 ·12 1/27 
Elk Butte EC IP 1 ·2777 0 ·0090 0 ·0092 1 ·01 0/29 
Lookout Butte LB IP 1 ·2190 0 ·0144 0 ·0146 0 ·9 9/19 
Wapiti Lake flow WL MJ 1 ·2187 0 ·0158 0 ·0160 0 ·3 0/31 
Harlequin Lake flow HL MJ 0 ·8300 0 ·0072 0 ·0074 1 ·09 3/29 
Lewis Canyon Rhyolite LC MJ 0 ·8263 0 ·0184 0 ·0186 0 ·58 10/27 
Mount Haynes Rhyolite MH MJ 0 ·7016 0 ·0014 0 ·0016 1 ·04 3/29 
. Contrary to conclusions from Christiansen (2001), based on field observations and preliminary K–Ar dates on some of the units, there is no time gap between the IP domes and the MJ rhyolite series, and both the Moose Creek flow (1·2856 ± 0·0064/0·0066 Ma) and the Wapiti Lake flow (1·2187 ± 0·0158/0·0160 Ma) overlap with ages from the IP domes. However, there is a significant time gap between the eruption of the Wapiti Lake flow at 1·2187 ± 0·0156/0·0160 Ma and the eruption of the subsequent Flat Mountain Rhyolite at 0·93 Ma [K–Ar age from Obradovich (1992); not sampled in this study]. Among the MJ rhyolites, the Moose Creek flow is the oldest and erupted close to the MFT-related Henry Forks caldera (Fig. 3). The MJ rhyolite Wapiti Lake flow erupted contemporaneously with the last of the IP domes, but at the northeastern end of the volcanic field. Consistent with observations from other Yellowstone rhyolites (Gansecki et al., 1996, 1998; Lanphere et al., 2002; Dallegge, 2008; Ellis et al., 2012; Rivera et al., 2014; Stelten et al., 2015), many of the IP and MJ lavas show subtly older sanidine grains, which are too old to be part of the eruption-age population (Fig. 3). It cannot be entirely excluded that apparent ‘too-old’ 40Ar/39Ar ages were caused by excess Ar hosted in minute melt or fluid inclusions; however, we note that all dated grains were screened for inclusions under a binocular microscope.
Fig. 3

Spatial distribution and age of IPMJ rhyolites based on 40Ar/39Ar eruption ages determined on sanidine. Coloured regions connect units from the same syn-eruptive cluster as marked with a frame in the bottom panel. The bottom panel shows eruption age as indicated in Table 3 with error bars marking the full external precision (2σ; see text); noted in grey are the two rhyolites not covered in this study: FM, Flat Mountain Rhyolite at c. 0·93 Ma K–Ar age from Obradovich (1992); BB, Big Bear Lake flow [undated, probable age based on stratigraphic relationships as recommended by Christiansen (2001)]. Inset diagrams show frequency distributions for some typical units, with the black continuous line marking the distribution of eruption age grains and the black dotted line the preservation of non-eruptive ages.

Fig. 3

Spatial distribution and age of IPMJ rhyolites based on 40Ar/39Ar eruption ages determined on sanidine. Coloured regions connect units from the same syn-eruptive cluster as marked with a frame in the bottom panel. The bottom panel shows eruption age as indicated in Table 3 with error bars marking the full external precision (2σ; see text); noted in grey are the two rhyolites not covered in this study: FM, Flat Mountain Rhyolite at c. 0·93 Ma K–Ar age from Obradovich (1992); BB, Big Bear Lake flow [undated, probable age based on stratigraphic relationships as recommended by Christiansen (2001)]. Inset diagrams show frequency distributions for some typical units, with the black continuous line marking the distribution of eruption age grains and the black dotted line the preservation of non-eruptive ages.

Quartz

Quartz is a ubiquitous mineral in all IPMJ rhyolites. Crystals are commonly subhedral, rounded in shape and ∼1·5–2 mm in diameter, but are up to 3 mm in the Warm River Butte. Quartz crystals commonly host melt inclusions with sizes from ∼10 to 200 μm. CL imaging reveals a lack of systematic zonation (i.e. the same zonation pattern in the majority of crystals) with most of the grains displaying gradual transitions between several generations of dark and bright zones (Fig. 4). Within a unit, several modes of zonation are observed, from non-zoned grains to a few that are strongly zoned. Quartz grains in most of the units feature darker rims whereas a few grains from the Lewis Canyon (∼15%) and the Silver Lake dome (∼14%) show an abrupt late change to bright rims. Minor discordant growth zones indicate periods of resorption followed by new crystallization. No correlations were observed with changes in crystal size.

Fig. 4

Cathodoluminescence images of quartz crystals sorted by eruption age (top left to bottom right young to old). Titanium concentrations (ppm) measured by LA-ICP-MS are shown next to the analysis spots (white circles). The bottom panels show the relative probability distributions of Ti concentrations in quartz grains from each unit and the number of analyses. The higher Ti contents in the Lewis Canyon Rhyolite compared with other MJ lavas should be noted.

Fig. 4

Cathodoluminescence images of quartz crystals sorted by eruption age (top left to bottom right young to old). Titanium concentrations (ppm) measured by LA-ICP-MS are shown next to the analysis spots (white circles). The bottom panels show the relative probability distributions of Ti concentrations in quartz grains from each unit and the number of analyses. The higher Ti contents in the Lewis Canyon Rhyolite compared with other MJ lavas should be noted.

Ti contents correlate with contents of Al (73–130 ppm) and Li (15–28 ppm) and are in good agreement with relative CL brightness distributions (Fig. 4), consistent with them being the main control of CL brightness [e.g. Leeman et al. (2012) and references therein]. Quartz grains exhibit a wide range of Ti contents (35–206 ppm), similar to those observed in post-LCT rocks (e.g. Vazquez et al., 2009; Girard & Stix, 2010; Leeman et al., 2012). Different crystals in the same unit can be zoned from lower to higher Ti values from core to rim, or from higher to lower Ti values, or show no zonation at all. Thus, they do not record a systematic compositional evolution with time. In most units, the Ti contents in rims of different grains are relatively similar, even when different crystals display a variable intra-grain Ti evolution. All trace element data for quartz and details regarding CL brightness and zonation of grains are provided in the

.

Feldspars

The IPMJ rhyolites contain both sanidine and plagioclase. However, the two phases occur in markedly different contexts. Sanidine is the dominant feldspar within all IPMJ rhyolites, except for the Lewis Canyon Rhyolite, occurring as euhedral, single crystals showing little optical zonation and rarely containing melt inclusions. Sizes range from ∼0·6 to 2·5 mm in diameter with the largest sizes occurring in the Warm River Butte; however, even the largest sanidines observed are smaller than the centimeter-sized sanidines in the MFT (Christiansen, 2001). By contrast, plagioclase is relatively rare in the IPMJ rhyolites with the exception of the Lewis Canyon Rhyolite. Crystal sizes range from 0·2 to 2 mm, commonly displaying small melt inclusions. A few crystals have sieve textures, a feature that has been previously described for the post-LCT lavas (Hildreth et al., 1984; Girard & Stix, 2009; Watts et al., 2012) and can be overgrown by sanidine. Optical zonation is rare and seemingly limited to very large crystals (>0·8 mm). Plagioclase is most commonly found within polymineralic glomerocrysts (see below), which are dominated by pyroxene, plagioclase and Fe–Ti oxides and may contain accessory zircon and apatite with some interstitial groundmass.

Compositionally, all sanidines from the IPMJ series are similar and range from Or43 to Or63; the end-member compositions are from the Lewis Canyon (Or43–55) and the Moose Creek (Or59–63) samples. Intra-grain major element compositional variability is limited, typically less than Or1–2. Plagioclase exhibits compositions between An11 and An36, with the end-members being the Mt Haynes Rhyolite (An11–22) and the Warm River Butte (An21–36). Generally, plagioclase from the IP rhyolites has slightly higher An contents compared with the MJ lavas (Fig. 5). Both IPMJ sanidine and plagioclase resemble previously reported compositions from Yellowstone rhyolites (Hildreth et al., 1984; Gansecki et al., 1996; Girard & Stix, 2009, 2010; Stelten et al., 2015).

Fig. 5

CL images of typical sanidine grains showing the locations of trace element analysis spots and An–Ab–Or ternary plot of sanidines and plagioclase from the IPMJ rhyolites; grey fields mark the compositions of other Yellowstone rhyolites (Hildreth et al., 1984; Gansecki et al., 1996).

Fig. 5

CL images of typical sanidine grains showing the locations of trace element analysis spots and An–Ab–Or ternary plot of sanidines and plagioclase from the IPMJ rhyolites; grey fields mark the compositions of other Yellowstone rhyolites (Hildreth et al., 1984; Gansecki et al., 1996).

Trace element data for sanidine do not distinguish the different units, with the exception of sanidine from the Lewis Canyon Rhyolite, which contains more Sr and Ba. Rather, sanidines within each of the IPMJ rhyolites are characterised by large spreads in trace element compositions between grains, often equally as large as the range of trace element contents within sanidines from the entire IPMJ series. CL imaging of sanidines in a subset of rhyolites reveals diverse modes of zonation between grains, similar to observations in quartz. The IP sanidines show both oscillatory and unidirectional zoning. Strong reverse, mostly unidirectional zoning with bright CL rims occurs in the MJ rhyolites, with strong variations of trace elements such as Ba, Sr, Eu, and Ti between core and rim. Sanidines in the Lewis Canyon rhyolite record Ba contents of up to 10 000 ppm in rims (compared with ∼7500 ppm in cores). To a lesser degree rim enrichment can also be observed in Sr content. The IP sanidines exhibit larger inter-grain variability and show no preferred direction of trace element zonation. Inter-grain variability and possible causes for Ba-enriched sanidine rims are discussed below in more detail.

Mafic minerals

Mafic minerals are generally rare in the IPMJ rhyolites. Previously, these lavas were described as containing clinopyroxene and orthopyroxene in addition to accessory fayalite and Fe-hornblende (Hildreth et al., 1984; Christiansen, 2001). However, in some of the studied lavas, particularly those of the IP series, fayalite and hornblende (Fig. 6) are abundant and form up to 60% of the mafic mineral assemblage (Table 1). Where both pyroxenes are present, clinopyroxene is more abundant. However, there are cases where clinopyroxene (Warm River Butte, Wapiti Lake flow, Harlequin Lake flow) or orthopyroxene (Mt Haynes Rhyolite) is sufficiently rare to be absent in mineral separates.

Despite limited intra-grain variability, compositional ranges in clinopyroxenes are large between grains (En11–38 Wo30–48, Fig. 7). The Mg# averages for each unit range from 32 ± 7 to 44 ± 2, whereas single crystals can reach an Mg# of up to 70. The degree of scatter varies markedly between units; for example, clinopyroxenes from the IP domes Lookout Butte, Osbourne Butte and Silver Lake dome cluster around Mg# = 33 ± 7, with single grains having values as high as 56, whereas grains from Lewis Canyon Rhyolite have values of around 45 ± 2. Orthopyroxene has compositions of En18–65 Wo02–15. Additionally, the Mt Haynes Rhyolite contains pure ferrosilite with >2 wt % MnO and <0·1 wt % MgO. Both ortho- and clinopyroxenes can often be found in crystal aggregates with plagioclase and iron oxides (see discussion below); these aggregate-hosted pyroxenes are compositionally identical to single pyroxenes in the groundmass.

Fig. 6

Examples of mafic minerals in back-scattered electron images (from upper left to bottom right): (1) exsolved magnetite; (2) clinopyroxene with zircon and apatite inclusions; (3) resorbed fayalite; (4) subhedral Fe-hornblende; (5) subhedral orthopyroxene with Fe–Ti inclusions; (6) ‘fresh’ anhedral inclusion-rich fayalite; (7) clinopyroxene overgrowth of orthopyroxene; (8) euhedral fresh clinopyroxene; (9) resorbed fayalite with cracks.

Fig. 6

Examples of mafic minerals in back-scattered electron images (from upper left to bottom right): (1) exsolved magnetite; (2) clinopyroxene with zircon and apatite inclusions; (3) resorbed fayalite; (4) subhedral Fe-hornblende; (5) subhedral orthopyroxene with Fe–Ti inclusions; (6) ‘fresh’ anhedral inclusion-rich fayalite; (7) clinopyroxene overgrowth of orthopyroxene; (8) euhedral fresh clinopyroxene; (9) resorbed fayalite with cracks.

Fig. 7

Pyroxene compositions from the MJ rhyolites (left) and the IP domes (right) illustrating the wide compositional range between pyroxenes within the same unit.

Fig. 7

Pyroxene compositions from the MJ rhyolites (left) and the IP domes (right) illustrating the wide compositional range between pyroxenes within the same unit.

Clinopyroxene REE patterns are similar to those in whole-rocks and show elevated contents of LREE and HREE and a marked Eu anomaly (Fig. 8). Orthopyroxenes show similar patterns, with the majority of grains having LREE > HREE and elevated contents of Zr, Sr and Sc. A minority of both clinopyroxenes and orthopyroxenes have LREE < HREE, lower overall abundances of REE, lack a pronounced Eu anomaly and have Mg# > 60.

Fig. 8

REE contents in mafic minerals normalised to BSE (McDonough & Sun, 1995). Dark grey fields show the REE patterns of the majority of the crystals, which are likely to have grown in a highly evolved rhyolitic melt. Black lines indicate the REE patterns of crystals deviating from the main trend. Crystals depleted in LREE and lacking a characteristic Eu anomaly crystallised in less evolved magma prior to significant plagioclase fractionation.

Fig. 8

REE contents in mafic minerals normalised to BSE (McDonough & Sun, 1995). Dark grey fields show the REE patterns of the majority of the crystals, which are likely to have grown in a highly evolved rhyolitic melt. Black lines indicate the REE patterns of crystals deviating from the main trend. Crystals depleted in LREE and lacking a characteristic Eu anomaly crystallised in less evolved magma prior to significant plagioclase fractionation.

Fayalite is restricted to the IP domes and the Moose Creek flow. Forsterite contents are in the range of 6–15, slightly higher than the 2–7 reported by Vazquez et al. (2009) for post-LCT lavas. Fayalites commonly show altered rims and oxidation features involving the transformation to iron oxides and hydroxides along cracks (Fig. 6). In terms of trace elements, the majority of grains show an evolved character with LREE > HREE and marked Eu anomalies, whereas a minority of crystals have low LREE contents and a less pronounced Eu anomaly (Fig. 8).

Amphiboles are Fe-hornblendes with subhedral shapes, without zoning or alteration features (Fig. 6). They occur commonly in the IP domes, Moose Creek flow and the Lewis Canyon Rhyolite. All hornblendes display pronounced negative Eu anomalies and high contents of LREE and HREE (Fig. 8).

Oxygen isotopes

Oxygen isotope compositions were determined on quartz and sanidine with ranges of 4·7–6·9‰ VSMOW for quartz and 4·5–6·0‰ VSMOW for sanidine (see

). Intra-unit variability is large, reaching 1·3‰ in both quartz and sanidine (Fig. 9). Units with lower oxygen isotope compositions such as the IP domes display more variability in δ18O. Mean square weighted deviations (MSWDs) of quartz populations are mostly higher than the critical MSWD for the analysed number of samples (Mahon, 1996), suggesting significantly more variation than expected for a single population.
Fig. 9

Oxygen isotope δ18O in per mil VSMOW for explosive end-members MFT and LCT and effusive domes and lavas between these end-members (abbreviations and colours as in Fig. 1). Wide boxes indicate probable δ18Omelt values calculated from average quartz compositions for each unit. Grey background bar illustrates the range of ‘normal’-δ18O rhyolites [expected values by fractionation of mid-ocean ridge (MOR) and arc basalts; Bindeman (2008) and references therein]. Upper panel shows the MSWD of the quartz analyses. MSWD > MSWDcritical indicates that the results do not represent a single population. Lower panel represents trends in δ18O throughout the history of the Yellowstone volcanic field (after Hildreth et al., 1984; Bindeman & Valley, 2001).

Fig. 9

Oxygen isotope δ18O in per mil VSMOW for explosive end-members MFT and LCT and effusive domes and lavas between these end-members (abbreviations and colours as in Fig. 1). Wide boxes indicate probable δ18Omelt values calculated from average quartz compositions for each unit. Grey background bar illustrates the range of ‘normal’-δ18O rhyolites [expected values by fractionation of mid-ocean ridge (MOR) and arc basalts; Bindeman (2008) and references therein]. Upper panel shows the MSWD of the quartz analyses. MSWD > MSWDcritical indicates that the results do not represent a single population. Lower panel represents trends in δ18O throughout the history of the Yellowstone volcanic field (after Hildreth et al., 1984; Bindeman & Valley, 2001).

Sanidine δ18O values are 0·5–1·0‰ lower than δ18O for quartz in the same unit. Although such divergence can be expected given the equilibrium fractionation of Δ18O(Qtz–Fsp) = 0·75‰ in silicic magmas at 850 ± 60°C (Bindeman & Valley, 2001), overlaps in δ18O values between quartz and sanidine crystals and the large variation of δ18O within a given rhyolite preclude a closed-system evolution for the IPMJ rhyolites. Taking average quartz values and assuming a fractionation factor of Δ18O(Qtz–melt) = 0·35‰ (Bindeman & Valley, 2001), we estimate magmatic δ18O values between 4·9 and 6·1‰ for the various units. Only the youngest MJ lavas are normal-δ18O rhyolites (defined as 5·8–6·3‰, Bindeman, 2008), whereas the majority of samples show a slight depletion. Values for all units are significantly lower than the δ18O value of ambient crust (6·8–7·2‰, Fig. 9) or of units that show a significant crustal component in oxygen, Pb, Sr and Nd isotope space, such as HRT member C and extra-caldera post-LCT rhyolite (Hildreth et al., 1991).

The mildly depleted δ18O signature of both the IP and MJ lavas is in good agreement with values previously reported for some of the MJ units (Hildreth et al., 1984; Bindeman & Valley, 2001), with the MJ lavas having slightly higher values than the IP domes (Fig. 9). Both the MFT and LCT explosive eruptions have higher δ18O than the intervening lavas. The pattern of falling and recovering δ18O magmatic values following large caldera-forming eruptions is well documented for Yellowstone (Hildreth et al., 1984; Bindeman & Valley, 2001), with the IPMJ series exhibiting similar behaviour on a smaller isotopic scale. Although inter-crystal O isotopic variability in the IPMJ rhyolites is not as large as in the low-δ18O post-LCT lavas (Bindeman & Valley, 2001; Watts et al., 2012), it still records a lack of closed-system fractionation throughout the history of Yellowstone.

Pb isotopes

Pb isotopes in sanidine have been proven to be a useful tool for the distinction of units that are very similar in trace element composition (e.g. Watts et al., 2012; Stelten et al., 2015). Data for the IPMJ rhyolites are in the range of previously published data for Yellowstone and differences between units are slight compared with the overall range of Pb isotope variation in Yellowstone (Fig. 10). Values seem to change from MFT-like to LCT-like compositions, with Pb isotope compositions in the IP domes being close to those of the preceding MFT, whereas values for MJ units are closer to those of the LCT (Fig. 10). We note that only a single translated spot was analysed per grain and that these analyses were obtained prior to CL imaging, and therefore yield no information on whether crystals are zoned in terms of Pb isotopes.

Fig. 10

Pb isotopes in sanidine in IPMJ rhyolites. Inset diagram shows the entire range of Pb isotopic compositions for the entire Yellowstone volcanic field with red-framed box showing the extent of the main diagram. Data for IPMJ rhyolites from this study, MFT from B. S. Ellis et al., (2014), and other Yellowstone rhyolites and basalts from Doe et al. (1982) and Watts et al. (2012).

Fig. 10

Pb isotopes in sanidine in IPMJ rhyolites. Inset diagram shows the entire range of Pb isotopic compositions for the entire Yellowstone volcanic field with red-framed box showing the extent of the main diagram. Data for IPMJ rhyolites from this study, MFT from B. S. Ellis et al., (2014), and other Yellowstone rhyolites and basalts from Doe et al. (1982) and Watts et al. (2012).

Intensive parameters

Zr-saturation thermometry (Hanchar & Watson, 2003; Boehnke et al., 2013) returns temperatures of 795–858°C for bulk-rock compositions, with the Lewis Canyon Rhyolite having the highest inferred temperature (Table 2). Where glass is present, results for glasses are typically 10–20°C higher than temperatures derived from bulk-rock compositions, but within uncertainty. Bulk-rock Zr-saturation temperatures from Boehnke et al. (2013) are 10–40°C lower than those from Hanchar & Watson (2003) and within error of the results of alkali feldspar–liquid thermometry (Putirka, 2008), except for Warm River Butte, which returns slightly lower temperatures (771 ± 23°C instead of 814 ± 17°C). All details and input parameters for thermometry calculations can be found in the

.

Ti-in-quartz thermometry (TitaniQ, Wark & Watson, 2006) was not performed because we were unable to constrain aTiO2 sufficiently. Fe–Ti oxides in these lavas are exsolved, therefore prohibiting calculation of aTiO2 from coexisting mineral pairs. Given the compositional range of quartz grains in these lavas, magmatic conditions with similar aTiO2 in space and time seem unlikely and previously employed activities range from 0·33 to 0·56 between post-LCT lava flows (Vazquez et al., 2009). Ti activities may change as a function of Ti content in the melt and its major element composition, co-crystallizing Fe–Ti oxide composition, oxygen fugacity, and temperature and pressure conditions (e.g. Thomas & Watson, 2012, and references therein). Thus, aTiO2 may have changed significantly throughout the growth history of even a single quartz grain, suggesting that an averaged aTiO2 calculated from coexisting Fe–Ti oxides or TiO2 solubility models provides only an estimate at best.

The previously unrecognised occurrence of amphibole and accessory biotite in the IPMJ rhyolites suggests a sufficient presence of water to create hydrous mineral phases [e.g. > 4·5 wt % H2O for formation of amphibole and >2·5 wt % H2O for biotite in experiments on Yellowstone hotspot track rhyolites by Almeev et al. (2012)]. Estimates of water contents from plagioclase–melt hygrometers (e.g. Putirka, 2008) indicate that about 3·6 ± 1·1 to 5·7 ± 1·1 wt % H2O was present during the formation of the IPMJ rhyolites (see

). Although strongly depending on inferred temperatures, these values are consistent with the generally increasing water contents from the central Snake River Plain to Yellowstone (Almeev et al., 2012; Bolte et al., 2015).

DISCUSSION

Time–space evolution of the Yellowstone volcanic field at 1·3–0·6 Ma

The new 40Ar/39Ar geochronology provides insights into the development of the volcanic field during the period between the MFT and LCT eruptions. Broadly speaking, volcanism migrated from west to east. Post-caldera activity first focused on the area of the MFT-related caldera collapse with the eruption of the IP domes and the Moose Creek flow, before migrating to areas surrounding the younger LCT-related caldera, where activity is characterised by eruptions from widely separated regions.

The overlap between the two magmatic series questions the traditional inference of two different cycles, with the IP domes marking the post-caldera activity of the MFT-forming eruption (second cycle), whereas the MJ succession records pre-caldera activity of the third cycle encompassing the eruption of the LCT (Christiansen, 2001). In terms of both age and geochemical behaviour, the Moose Creek flow is similar to the IP domes; for example, in mafic mineral abundances (Table 1), feldspar and pyroxene major and trace element compositions (Figs 4 and 5), and oxygen and Pb isotopic compositions (Figs 9 and 10). Geochemistry may thus largely be controlled by geographical location and source region rather than affiliation with a magmatic series defined by morphology.

The temporal overlap between the MJ and IP series suggests that the different morphologies of IP domes compared with the lava flows of the MJ series are not controlled by a time-progressive evolution. IPMJ volcanism occurred in unevenly distributed clusters of 2–4 lava flows (Fig. 3), consistent with discrete periods of activity observed in the post-LCT record (Christiansen, 2001; Christiansen et al., 2007). Notably, in some cases lava flows erupted contemporaneously, within uncertainty, with other flows at the opposite edge of the volcanic field (e.g. Lookout Butte and Wapiti Lake flow), suggesting that batches of broadly similar, yet isotopically subtly distinct liquids were present simultaneously within the crust.

The largest time gap in the IPMJ record is the 290 kyr of quiescence starting ∼120 kyr after the MFT eruption between the eruptions of Wapiti Lake flow and the Flat Mountain Rhyolite. The post-LCT record equally contains periods of prolonged quiescence, such as the ∼180 kyr quiescence between the Dunraven Road flow and the following South Biscuit Basin flow (Christiansen et al., 2007; Bindeman et al., 2008). This similarity suggests that prolonged quiescence (>100 kyr) could be a common feature in the effusive record between two large-volume eruptions.

However, temporal gaps may be a function of an incomplete record. Prior to the Mesa Falls eruption, only six small-volume lavas are known (five post-HRT lavas and the earlier Snake River Butte lava; Christiansen, 2001). Between the MFT and LCT eruptions, around 12 lavas are known, with the uncertainty coming from the unknown number of lavas forming the Lewis Canyon Rhyolite. Following the LCT, the youngest period of volcanism contains at least 40 separate lavas and small-volume pyroclastic deposits. This suggests either that the frequency of lavas erupted from Yellowstone has increased by a factor of four after the eruption of the LCT, or, as we prefer, that the younger lavas are much better preserved owing to a lack of later caldera-forming eruptions. Such a preservation bias remains a challenge during investigations of cycles at caldera volcanoes.

Petrogenesis of the IPMJ series

Our detailed mineral-scale characterisation of the IPMJ record provides important constraints for rhyolite generation in Yellowstone. Here we discuss our observations from geochemistry, petrology, and isotopic studies, which together provide a framework for rhyolite petrogenesis, within which the IPMJ series may be assessed.

Glomerocrysts as recorders of melt extraction

As noted above, plagioclase and pyroxenes in the IPMJ rhyolites mostly occur together in glomerocrysts. Notably, quartz and sanidine are absent in these aggregates, despite being the most abundant phases within the overall mineral assemblage (see Table 1). Being dominated by plagioclase and mafic phases, with only limited groundmass (typically <20%), these crystal aggregates are necessarily less evolved in bulk composition than the lavas within which they are hosted (Fig. 11). The stark contrast between the bulk compositions of the crystal aggregates and the host magma has been interpreted as reflecting the extraction of rhyolitic liquid from these crystal aggregates, thus rendering them cumulates (e.g. Ellis et al., 2014). The lack of sanidine and quartz in these aggregates implies that these minerals predominantly crystallised after melt was extracted, leaving behind a refractory cumulate of plagioclase, pyroxene and Fe–Ti oxides (Fig. 11; compare Ellis et al., 2014). The observation that sanidine and quartz occur as large, mostly euhedral to subhedral crystals, whereas smaller plagioclase and mafic minerals are often anhedral, supports this interpretation. We note that the majority of mafic minerals show trace element patterns consistent with growth from an evolved liquid (Fig. 8), with high contents in LREE and HREE and a pronounced Eu anomaly, similar to whole-rock and glass REE patterns in the IPMJ rhyolites (Fig. 2). The crystal aggregates therefore record a fractionation step between two evolved magmas.

Fig. 11

Examples of glomerocrysts from different Island Park domes. Comparison of mineral assemblage within the aggregate with free single crystals indicates no significant compositional differences. Bulk glomerocryst compositions are less evolved compared with typical whole-rock modal compositions owing to the lack of late-forming sanidine and quartz.

Fig. 11

Examples of glomerocrysts from different Island Park domes. Comparison of mineral assemblage within the aggregate with free single crystals indicates no significant compositional differences. Bulk glomerocryst compositions are less evolved compared with typical whole-rock modal compositions owing to the lack of late-forming sanidine and quartz.

In bulk, glass, and pyroxene compositions, all IPMJ rhyolites exhibit notable negative Eu anomalies (Figs 2 and 7). The rarity of plagioclase and the low Sr contents in the erupted liquids (Tables 1 and 2) suggest extensive removal of plagioclase at depth to decrease the bulk Sr content from 190–680 ppm in the basalts (Doe et al., 1982) to 10–60 ppm observed in the IPMJ rhyolites. Similarly, pyroxene fractionation could account for the extremely low contents of MgO in the IPMJ rhyolites in both whole-rocks (0·01–0·12 wt % MgO, Table 2) and glasses (0·03 ± 0·02 wt % MgO).

In the Yellowstone province, identical crystal aggregates have been reported from the young post-LCT rhyolites (Girard & Stix, 2009; Watts et al., 2012), the Heise eruptive centre (Watts et al., 2011), the central Snake River Plain (Cathey & Nash, 2009; Ellis & Wolff, 2012; Ellis et al., 2014) and in the Jarbidge rhyolite (Brueseke et al., 2014). We note that in the Yellowstone system, both low- and normal-δ18O rhyolites contain these crystal aggregates and both groups have identical major and trace elemental compositions [as noted previously by Hildreth et al. (1984)], including low Sr contents. Thus, both rhyolite groups must have experienced large-scale plagioclase and pyroxene fractionation during their petrogenesis and the observed glomerocrysts may reflect a common process of crystal–liquid separation in a mush zone (Vazquez & Reid, 2002; Stelten et al., 2015).

If quartz grew primarily after extraction, the time between liquid extraction from the pyroxene–plagioclase-dominated mush zone and eruption of rhyolite magma is equal to the minimum time for quartz to grow without intermittent resorption stages. Based on quartz sizes and growth rates, Ellis et al. (2014) suggested that timescales of extraction were of the order of a few thousand years for the rhyolites of the Snake River Plain. Timescales inferred in the same fashion for IPMJ rhyolites with maximum quartz sizes of 1·5–2·9 mm and growth rates of ∼10–13 to ∼10–14 m s–1 (Gualda et al., 2012b) result in timescales of ∼1000–9000 years. Such estimates are in good agreement with longevities of crystal-poor magma of ∼10 000 years based on zircon chronology (Rivera et al., 2016; Wotzlaw et al., 2014) and up to 9000 years from 238U–230Th dating of zircon rims (Stelten et al., 2015).

Compositional variability and timescales of mineral entrainment

The large inter-grain variability seen in mineral and oxygen isotope compositions is a pervasive feature in all IPMJ rhyolites. Previously, we noted the large variability in oxygen isotopic compositions in both quartz and sanidine, with MSWD for quartz often exceeding critical MSWD values expected for a single population (Fig. 9). Large compositional differences for trace elements in sanidine and quartz (Fig. 4) and their complex, unsystematic CL zoning patterns (Figs 4 and 5) suggest that different crystals in the same unit did not share common magmatic histories. Additionally, a comparison of trace element contents, such as Ba in glass, with those in the rims of CL-imaged sanidine grains shows that a large number of grains did not grow in equilibrium with their host melt, even if a wide range of published KD(San–melt) values is explored (Fig. 12).

Fig. 12

(a) Ba contents in sanidine rims and cores do not fall into the calculated equilibrium range expected from Ba contents in glass from the same unit. Equilibrium range calculated with KD(San–melt) ≈ 22 in rhyolite from Leeman & Phelps (1981) and KD(San–melt) ≈ 6·7 in high-silica rhyolite (Mahood & Hildreth, 1983). The figure includes only data for those sanidine crystals that were imaged via CL. (b) Average Rb/Sr in San for each unit with error bars indicating one standard deviation. Numbers in italic correspond to number of analyses per unit. The large variability for IP domes and Moose Creek compared with MJ flows should be noted; this indicates a higher entrainment component for units close to MFT-related caldera. (c) Ba ratios between rim and core vs Sr ratios respectively suggest that reverse zoning in San is due to cumulate remelting or crystallization from a less evolved rhyolite melt. The figure includes only data for those sanidine crystals that were imaged via CL.

Fig. 12

(a) Ba contents in sanidine rims and cores do not fall into the calculated equilibrium range expected from Ba contents in glass from the same unit. Equilibrium range calculated with KD(San–melt) ≈ 22 in rhyolite from Leeman & Phelps (1981) and KD(San–melt) ≈ 6·7 in high-silica rhyolite (Mahood & Hildreth, 1983). The figure includes only data for those sanidine crystals that were imaged via CL. (b) Average Rb/Sr in San for each unit with error bars indicating one standard deviation. Numbers in italic correspond to number of analyses per unit. The large variability for IP domes and Moose Creek compared with MJ flows should be noted; this indicates a higher entrainment component for units close to MFT-related caldera. (c) Ba ratios between rim and core vs Sr ratios respectively suggest that reverse zoning in San is due to cumulate remelting or crystallization from a less evolved rhyolite melt. The figure includes only data for those sanidine crystals that were imaged via CL.

Mg-numbers in clinopyroxene are similar for the majority of grains in one unit. Deviating mineral compositions in a small number of grains identify entrainment products not in equilibrium with their host melt, such as rare more mafic pyroxenes (Fig. 7). These pyroxenes could represent remnant crystals from less evolved regions of the magmatic system, as indicated by their lower content of LREE (Fig. 8). Their lack of a pronounced negative Eu anomaly suggests crystallization prior to significant plagioclase fractionation (compare Szymanowski et al., 2015). A basaltic origin is unlikely for these high-Mg# pyroxenes as Yellowstone basalts do not contain pyroxene phenocrysts (Christiansen, 2001).

Orthopyroxene and fayalite are commonly thought to replace one another (e.g. Davidson & Lindsley, 1989) and the abundant resorption features in fayalite suggest that conditions mostly favoured the formation of Fe-rich orthopyroxene, which is abundant in both the Heise and Yellowstone volcanic fields (e.g. Christiansen, 2001; Watts et al., 2011). A mafic mineral assemblage of clinopyroxene, orthopyroxene, fayalite and hornblende, as observed in some of these lavas (Table 1), reflects a disequilibrium mineral assemblage that requires late entrainment of crystals from at least one of these mineral groups.

Unsystematic zoning patterns, variable major and trace element contents and isotopic compositions, and a complex mafic mineral assemblage indicate that crystals did not share a common magmatic history and that a large number of crystals probably did not grow in the melt in which they erupted. This late-stage entrainment seems to be particularly strong in units close to the MFT-related caldera, where chemically homogeneous crystals of different compositions, derived from either the mush or the overlying caldera roof, are assembled in the same magma batch, whereas younger MJ rhyolites contain less variable crystal populations (Figs 9 and 12; see also comments on compositional variability between grains of the same unit in the results sections on the various mineral groups).

The preservation of oxygen isotopic disequilibrium between quartz and sanidine crystals allows the timescales of material entrainment to be estimated. Diffusion rates for oxygen in quartz are about four magnitudes faster than those for feldspars, and quartz crystals of 4 mm diameter would be expected to have fully equilibrated at magmatic temperature after thousands of years (Bindeman, 2008, and references therein). The internal complexities observed in the CL images of the quartz grains suggest a potential for dissolution and reprecipitation processes, which can accelerate isotopic exchange. Thus, the O isotopic disequilibria indicate that quartz was added to the magma less than a few thousand years (at maximum) prior to eruption.

The recognition of subtly older sanidine crystals via 40Ar/39Ar single-crystal geochronology also allows an estimation of the timescale of entrainment. Diffusion calculations suggest that ages of older crystals are indistinguishable from those of phenocrysts after residence times of 1–2 years in the magma (Gansecki et al., 1996), with similar short preservation timescales reported from other studies (Renne et al., 2012). The preservation of such antecrystic sanidine requires that addition occurs immediately prior to eruption. The same process causing tails of older ages in sanidine, could well be responsible for the compositional ranges in all mineral groups, as sanidine is unlikely to be the only antecrystic mineral sampled prior to and during eruption.

Assimilation and crustal component

Recurring patterns of sudden falls in δ18O following caldera collapse represent the major argument for bulk remelting of hydrothermally altered material (Bindeman & Valley, 2001). Extra-caldera, normal-δ18O rhyolites indicate that the processes responsible for the depletion are limited to the spatial extent of the caldera and directly relate to the caldera collapse. Even if magma is stored in a mush-dominated environment, the caldera collapse can be expected to bring hydrothermally altered lithologies to sufficient depths for partial or bulk remelting, thus providing a mechanism for additional entrainment of roof material that could account for the higher crystal contents (Table 1) and larger mineral-scale compositional ranges observed in the units spatially associated with the caldera.

In the same manner, assimilation and remelting of hydrothermally altered roof material following caldera collapse could be responsible for adding sufficient amounts of H2O to the shallow parts of the magma reservoir to allow for crystallization of amphibole and accessory biotite. Although we do not attempt to speculate about the mechanisms of this process, it is noteworthy that hydrous minerals are seemingly limited to units close to the MFT-associated caldera.

IPMJ-like depletions in δ18O are permissive of remelting and incorporation of (1) high degrees of very slightly altered materials or (2) lower degrees of more strongly altered (lower δ18O) protoliths. For a simple estimate, we take the lowest value advocated by Watts et al. (2011) for a hydrothermally altered assimilant of −1‰ and a normal-δ18O rhyolite of 6·3‰ (i.e. value expected from fractionation of mantle-derived basalt), inferring that the depletions in the IPMJ series (+5·0 to +6·1‰) can be explained by ≤25% assimilation. However, the mechanistic processes of assimilating a hydrothermally altered and mineralogically complex protolith remain poorly understood. Given that hydrothermal processes produce a wide variety of alteration assemblages with bulk δ18O spanning more than 10‰, and potentially being lower than the assumed value of −1‰ (Hildreth et al., 1984), the degree of interaction with hydrothermally altered material might deviate strongly.

In terms of radiogenic isotopes, Pb isotope data from this study (Fig. 10) show little variability and are similar to those of the majority of Yellowstone rhyolites (Doe et al., 1982). Even given the strong leverage provided by the Wyoming Craton and the fact that the low Sr contents of these rhyolites make them particularly susceptible to Sr contamination, bulk 87Sr/86Sr values for the IPMJ rhyolites vary between 0·7090 and 0·7139, only slightly higher than that for the most radiogenic Yellowstone basalts of 0·7089 (Doe et al., 1982; Hildreth et al., 1991). Thus, the rhyolites of the IPMJ series do not require significant contamination from crustal lithologies to pass from basalt to rhyolite and small variations probably reflect minor regional differences in the magma source region.

A model for the Yellowstone magmatic system between 1·3 and 0·6 Ma

In the context of the parameters discussed above, we focus here on a coherent model for the evolution of the Yellowstone magmatic system during the IPMJ period of activity. We propose that during IPMJ time, the silicic magmas of the Yellowstone magmatic system were derived from small-scale magma batches residing in a long-lived crystal-rich, mushy reservoir, with the crystal aggregates representing direct samples of this mushy refractory material. These aggregates (Fig. 11) contain mostly plagioclase, pyroxene and Fe–Ti oxides, suggesting that these phases dominate the mush. Aggregates from different lavas have subtly different mineral compositions that are identical to the compositions of the single crystals in that lava, suggesting that some single crystals may represent disaggregated cumulate material. If silicic lavas are a true reflection of the magma compositions beneath their eruption sites, the widely dispersed, synchronous eruption of compositionally almost indistinguishable rhyolite lavas (Table 2, Fig. 2) suggests the presence of a large-volume mushy source region. Small-scale compositional differences, as observed in isotopes and trace elements, indicate that the composition of the mush varied both laterally and with time, suggesting that the reservoir was incrementally built and subject to changes in the focus of main activity over time.

To investigate the conditions at which the erupted magmas were stored prior to eruption, we performed simulations using rhyolite-MELTS (Gualda et al., 2012a). Even though the complex mineral assemblage requires some degree of disequilibrium, we can use equilibrium magmatic modelling to approximate the last stages of crystallization before eruption because (1) recycling mainly focused on compositionally similar material, (2) entrainment of antecrystic mineral phases (disequilibrium on a trace element or isotopic scale) does not preclude them from being stable in the melt, and (3) true disequilibrium assemblages such as the mafic minerals make up a small fraction of the magma (Table 1). We choose Elk Butte as a starting material for rhyolite-MELTS calculations as Elk Butte represents a typical MJIP member in terms of mineral compositions and does not contain any non-eruption age sanidine, and therefore was not significantly affected by crystal contamination. As input parameters, we use a variety of combinations of pressure (0·5–5·0 kbar) and water contents (0–5 wt % H2O) at an oxygen fugacity of QFM [Quartz–fayalite–magnetite; supported by the observation of fayalite, quartz and magnetite in these units and in agreement with experimental data from Bolte et al. (2015)]. Temperature input ranges start at the calculated liquidus of the assemblage and stop at 670°C, with steps of 1°C. We define the solidus as the highest temperature at which rhyolite-MELTS returns <1% melt.

As described above, the mineral record in the MJIP rhyolites may be complex, so we focused on a few important parameters to identify likely storage conditions: (1) the crystallinity of the rhyolite-MELTS run should be within the range of the IPMJ lavas (60–87% melt; Table 1); (2) the composition of the melt should roughly correspond to the glass compositions; (3) the composition of sanidine predicted by rhyolite-MELTS should be similar to the sanidine compositions (Or57–61). These parameters are approximated at conditions of ∼870°C, ∼0·8–1 kbar and ∼2·8–3·3 wt % H2O (see figure and additional details in

).

At these conditions, the mineralogy of the rhyolite-MELTS output is dominated by quartz and sanidine with minor plagioclase, consistent with these phases being dominant in the mineral assemblage (Table 1). The only other crystallizing phase is magnetite, although the identified conditions border the stability field of orthopyroxene. The fact that these conditions cannot reproduce the mafic mineral assemblage is consistent with our interpretation of them originating mostly from cumulate glomerocrysts, which record an earlier petrogenetic stage (see section ‘Glomerocrysts as recorders of melt extraction’).

A magmatic temperature of 870°C agrees well with the temperature inferred from sanidine–melt geothermometry (∼870°C for Elk Butte;

), suggesting that thermometry on late-crystallizing sanidine and melt pairs may result in the most reliable temperature estimates. In terms of water content, the results are slightly higher than previous estimates for rhyolites of the province based on rehomogenised melt inclusions (1·5–2·0 wt %, Befus & Gardner, 2016) and experimental petrology (1·5–2·5 wt %, Bolte et al., 2015) and lower than our estimates based on plagioclase–melt hygrometry (3·6 ± 1·1 to 5·7 ± 1·1 wt % H2O; see ). The latter indicate that plagioclase–melt pairs should be treated with caution because of the longer crystallization range of plagioclase compared with sanidine. Neither significant biotite nor amphibole crystallization is observed at the identified storage conditions or any other explored combination of pressure, temperature and water content, suggesting that crystallization of hydrous mafic minerals is driven by less evolved melt compositions at an earlier petrogenetic stage or higher water contents, or that rhyolite-MELTS is not well calibrated for these phases.

The pressure estimate of 1 kbar translates to a depth of ∼4 km, given the relatively low density of the crust, which is composed of silicic volcanic rocks and hydrothermally altered materials. Similar pressures, ranging between 0·5 and 1·5 kbar, have been reported from other studies (e.g. Befus & Gardner, 2016), albeit for the younger Yellowstone volcanism. These pressure estimates are slightly lower than those for the upper reaches of the geophysically imaged magma body beneath Yellowstone (8–18 km, Husen et al., 2004; Lowenstern et al., 2006; Chang et al., 2007; DeNosaquo et al., 2009; Chu et al., 2010; Farrell et al., 2014), in which the eruptible magmas would collect prior to eruption (e.g. Bachmann & Bergantz, 2008).

Geophysical studies have long been able to detect a large low-Vp body beneath the Yellowstone volcanic field. The proportion of crystal-poor, eruptible magma has been estimated at ∼15% volume via gravity measurements (Krukoski, 2002), ≤30% volume via P-wave arrival times (Chu et al., 2010), and ≤35% volume via strain measurements from seiche waves on Yellowstone lake (Luttrell et al., 2013). Regardless of the method of study or the proportion of crystal-poor melt estimated, geophysical studies overwhelmingly support the occurrence of a crystal mush currently underlying Yellowstone. Although this does not necessarily show that such a situation occurred during IPMJ times, we have no reason to believe that the volcanic system then behaved in a fundamentally different way from today.

Based on the age clustering of the IPMJ rhyolites, magmatic activity occurred in discrete pulses as is clear in the post-LCT record. Reactivation of the magmatic system probably occurred as a result of heat transfer following magma recharge at greater depths [termed ‘defrosting’ by Mahood (1990); compare Wolff et al. (2015)] and is recorded in late rims in some sanidine and quartz crystals. The reverse zonation in Ba in some sanidine crystals, coupled with other trace elements (Sr, Rb, Eu), suggests that these rims result from melting of sanidine and plagioclase from the cumulate pile or crystallised from less evolved rhyolite [e.g. 960 ppm Ba in Biscuit Basin flow (Christiansen, 2001) with KD(San–melt) ≈ 22 (Leeman & Phelps, 1981) would allow crystallization of sanidine with up to 21 000 ppm Ba]. Significant mass transfer from the replenishing magma is not favoured owing to the homogeneity in Pb isotopes (Fig. 10). Crystal aggregates generally do not contain sanidine, thus the crystal aggregates could record the refractory part of the mushy reservoir after remelting of sanidine and plagioclase (Fig. 12).

As noted above, the IPMJ rhyolites show a complex mineralogical record including entrainment of various mineral phases. Within the constraints of the preserved lavas, there appears to be little obvious compositional evolution within the magmatic system during IPMJ times, with the variability occurring in a mineral group in a single unit as large as that observed through time. Although the higher variability in our dataset compared with previous studies could be due to the larger number of analyses, this variation should not be large enough to obscure possible trends if these had the same magnitude as the clear trend towards cooler temperatures and more juvenile isotopic compositions observed in the post-LCT lavas (compare Vazquez et al., 2009). The fact that there is neither a continuous evolution through time nor sudden compositional changes that could hint at a forthcoming super-eruption suggests that (1) reactivation of the magmatic system occurs on a faster timescale than can be recorded in most mineral compositions or (2) a significant preservation bias hinders conclusions on possible eruption triggers. Changes might be preserved in lavas erupted between the Mt Haynes Rhyolite and the LCT, which are now hidden under younger volcanic deposits or were removed by the LCT eruption.

The limited depletion in δ18O observed in the IPMJ rhyolites does not require high degrees of remelting of hydrothermally altered precursors, as is observed for post-LCT low-δ18O rhyolites with diverse phenocrysts. However, a number of similarities between the IPMJ series and the later low-δ18O rhyolites hint that the processes involved in their petrogenesis may not be so distinct. For example, within the IPMJ series the lowest δ18O magmas and those with the highest proportion of hydrous mineral phases are those spatially associated with the MFT-related caldera. This association supports, on a smaller scale, the cannibalisation processes proposed for the youngest Yellowstone lavas. Understanding these processes fully, however, still requires a physical model for the melting of a mineralogically complex protolith with heterogeneous isotopic compositions.

Our combined data support petrogenesis in an incrementally built, mushy magma reservoir with compositional variations in space and time, similar to the recent models of Vazquez et al. (2009) and Stelten et al. (2015) for similarly less depleted post-LCT rhyolites. Although we do not find the same thermochemical trends indicating a cooling and evolving reservoir through time as observed for the short-lived post-LCT rhyolites, we note that the compositional evolution described in these studies was revealed from seven lavas erupted within the space of 130 kyr. Only among the IP domes do we find a similar density of deposits, and only the Mt Haynes Rhyolite erupted within 130 kyr before the LCT. Thus, the potential for direct comparison of the IPMJ period with the post-LCT record remains limited. Potential precursory geochemical signals remain elusive in the IPMJ record.

CONCLUSIONS

Our detailed mineralogical and geochemical study of the IPMJ rhyolites provides new insights into rhyolite generation at Yellowstone, complementing previous studies that focused on the youngest Yellowstone rhyolites. Multiple lines of evidence support rhyolite genesis in an incrementally built, upper crustal mush zone with remobilization of hydrothermally altered material occurring in areas affected by caldera collapse (Fig. 13). The striking geochemical similarity between all the Yellowstone rhyolites suggests that fundamentally similar processes control their generation in the shallow parts of a crystal mush.

Fig. 13

Schematic summary of the petrogenetic processes and characteristics of the IPMJ rhyolite series.

Fig. 13

Schematic summary of the petrogenetic processes and characteristics of the IPMJ rhyolite series.

New ages via 40Ar/39Ar dating of sanidine reveal that the magmatic series IP domes and MJ rhyolites overlap by at least 66 kyr. The decreased eruption frequency in the IPMJ record suggests that large parts of the depositional record are covered by younger lava flows, prohibiting a time resolution high enough to decipher potential precursory activity before eruption of the LCT, if timescales allow minerals to record such processes.

Detailed investigation of the mineralogy of the IPMJ rhyolites indicates a complex assemblage of minerals within the lavas, particularly in the mafic mineral assemblage. Disequilibrium in O isotopic compositions and the preservation of subtly older single-crystal 40Ar/39Ar ages in sanidines in most Yellowstone rhyolites are consistent with late-stage entrainment of compositionally similar material. We infer that the processes giving rise to the variability in sanidine also generate the disequilibrium seen in other minerals. Because there is no correlation between the amount of entrained material and the eruption style, this late-stage contamination probably occurs in shallow, cooler portions of the crystal mush and is enhanced in the area of the preceding caldera collapse. Here, mushy roof material could be brought to sufficient depths to be remobilised and recycled. Additional research needs to be undertaken to assess the preservation potential of antecrystic sanidine in mush-like conditions, with or without partial re-equilibration.

An important conclusion of this work is that both trace elements and isotopes suggest that significant proportions of the crystals found within the IPMJ suite of lavas did not crystallise from the magma in which they erupted, even if the same mineral phase can be stabilised in the melt according to equilibrium magma modelling via rhyolite-MELTS. This appreciation is particularly relevant for the derivation of magmatic parameters and in situ studies (e.g. melt inclusion or diffusion timescale work), which necessarily investigate a limited number of grains owing to the time-consuming nature of the work. In such cases, conclusions derived from a small number of grains must be exported to whole magmatic systems with care.

Plagioclase–pyroxene-dominated glomerocrysts represent direct petrological samples from an underlying crystal-rich mushy zone at depth (Fig. 11). Such crystal aggregates occur in both low-δ18O rhyolites such as the Upper Basin Member rhyolites (Girard & Stix, 2009; Watts et al., 2012) and the mildly depleted rhyolites of the IPMJ series, and thus represent a common part of rhyolite petrogenesis. Depletions in Sr and Eu in whole-rock and glass data support significant plagioclase fractionation at depth, consistent with the scarcity of glomerocryst-forming minerals in hand specimen. Whereas plagioclase and mafic minerals remain trapped at depth and are rarely observed in the erupted magmas, subhedral to euhedral sanidine and quartz grow after extraction of the melt. The timescales for liquid extraction from an upper crustal mush zone, as inferred from quartz growth, are of the order of 10 kyr, consistent with timescales from zircon chronology for both effusive and explosive eruptions at Yellowstone (Rivera et al., 2016; Wotzlaw et al., 2014; Stelten et al., 2015).

Mafic minerals containing lower amounts of LREE and lacking a prominent Eu anomaly could provide glimpses into deeper, less evolved parts of the magmatic system, which are usually not sampled owing to the sheer size of the magma reservoir [estimated today to be as large as c. 4000 km3 with 5–15% melt (Farrell et al., 2014)]. The fact that similar lavas are recurrently produced throughout the entire magmatic history of the volcanic field supports magma genesis in a coherent crystal mush via several melt-rich magma batches that may erupt independently of each other.

FUNDING

This work was supported by an ETH research grant (ETH-05 13-2 to J.T.), funds from Swiss National Science Foundation research grants (SNSF 200021-146268 and SNSF 200021-155923/1 covering O.B. and B.S.E.), an ETH student research travel fund (to J.T.) and an American Philosophical Society Franklin Grant (to B.S.E.). NERC is acknowledged for continued funding of AIF at SUERC, East Kilbride.

SUPPLEMENTARY DATA

for this paper are available at Journal of Petrology online.

ACKNOWLEDGEMENTS

We thank Peter Appel and Barbara Mader (University of Kiel) for help with EMP analyses, Karsten Kunze and the Scientific Center for Optical and Electron Microscopy (ScopeM) at ETH for support with CL images, and Christie Hendrix and Stacey Gunther from the Yellowstone National Park Service for their assistance with research permits (Yellowstone permit YELL-05940). We are grateful to Mark Stelten, Christy Till and James Brophy for their constructive and careful reviews. Wendy Bohrson is thanked for editorial handling and providing additional helpful comments on the paper.

REFERENCES

Almeev
R. R.
Bolte
T.
Nash
B. P.
Holtz
F.
Erdmann
M.
Cathey
H. E.
(
2012
).
High-temperature, low-H2O silicic magmas of the Yellowstone hotspot: an experimental study of rhyolite from the Bruneau–Jarbidge Eruptive Center, Central Snake River Plain, USA
.
Journal of Petrology
 
53
,
1837
1866
.
Bachmann
O.
Bergantz
G. W.
(
2008
).
Rhyolites and their Source Mushes across Tectonic Settings
.
Journal of Petrology
 
49
,
2277
2285
.
Befus
K. S.
Gardner
J. E.
(
2016
).
Magma storage and evolution of the most recent effusive and explosive eruptions from Yellowstone Caldera
.
Contributions to Mineralogy and Petrology
 
171
,
1
19
.
Bindeman
I.
(
2008
).
Oxygen isotopes in mantle and crustal magmas as revealed by single crystal analysis
. In:
Putirka
K. D.
Tepley
F. J.
III
(eds)
Minerals, Inclusions and Volcanic Processes. Mineralogical Society of America and Geochemical Society, Reviews in Mineralogy and Geochemistry
 
69
,
445
478
.
Bindeman
I. N.
Valley
J. W.
(
2001
).
Low-δ18O rhyolites from Yellowstone: magmatic evolution based on analyses of zircons and individual phenocrysts
.
Journal of Petrology
 
42
,
1491
1517
.
Bindeman
I. N.
Fu
B.
Kita
N. T.
Valley
J. W.
(
2008
).
Origin and evolution of silicic magmatism at Yellowstone based on ion microprobe analysis of isotopically zoned zircons
.
Journal of Petrology
 
49
,
163
193
.
Boehnke
P.
Watson
E. B.
Trail
D.
Harrison
T. M.
Schmitt
A. K.
(
2013
).
Zircon saturation re-revisited
.
Chemical Geology
 
351
,
324
334
.
Bolte
T.
Holtz
F.
Almeev
R.
Nash
B.
(
2015
).
The Blacktail Creek Tuff: an analytical and experimental study of rhyolites from the Heise volcanic field, Yellowstone hotspot system
.
Contributions to Mineralogy and Petrology
 
169
,
1
24
.
Brueseke
M. E.
Callicoat
J. S.
Hames
W.
Larson
P. B.
(
2014
).
Mid-Miocene rhyolite volcanism in northeastern Nevada: The Jarbidge Rhyolite and its relationship to the Cenozoic evolution of the northern Great Basin (USA)
.
Geological Society of America Bulletin
 
B30736.30731
.
Cathey
H. E.
Nash
B. P.
(
2009
).
Pyroxene thermometry of rhyolite lavas of the Bruneau–Jarbidge eruptive center, Central Snake River Plain
.
Journal of Volcanology and Geothermal Research
 
188
,
173
185
.
Chang
W.-L.
Smith
R. B.
Wicks
C.
Farrell
J. M.
Puskas
C. M.
(
2007
).
Accelerated uplift and magmatic intrusion of the Yellowstone caldera, 2004 to 2006
.
Science
 
318
,
952
956
.
Christiansen
R. L.
(
1984
).
Yellowstone magmatic evolution: Its bearing on understanding large-volume explosive volcanism
. In:
Explosive Volcanism: Inception, Evolution, and Hazards
 ,
National Research Council. Washington, DC
:
The National Academies Press
, pp.
84
95
.
Christiansen
R. L.
(
2001
).
The Quaternary and Pliocene Yellowstone Plateau Volcanic Field of Wyoming, Idaho, and Montana. US Geological Survey, Professional Papers
 
729-G
.
Christiansen
R. L.
Blank
H. R.
(
1972
).
Volcanic stratigraphy of the Quaternary rhyolite plateau in Yellowstone National Park. US Geological Survey, Professional Papers
 
729-B
.
Christiansen
R. L. L.
Jacob
B.
Smith
R. B.
Heasler
H.
Morgan
L. A.
Nathenson
M.
Mastin
L. G.
Muffler
L. J. P.
Robinson
J. E.
(
2007
).
Preliminary Assessment of Volcanic and Hydrothermal Hazards in Yellowstone National Park and Vicinity. US Geological Survey Open-File Report
 
2007
1071
.
Chu
R.
Helmberger
D. V.
Sun
D.
Jackson
J. M.
Zhu
L.
(
2010
).
Mushy magma beneath Yellowstone
.
Geophysical Research Letters
 
37
. doi:10.1029/2009GL041656.
Coble
M. A.
Mahood
G. A.
(
2012
).
Initial impingement of the Yellowstone plume located by widespread silicic volcanism contemporaneous with Columbia River flood basalts
.
Geology
 
40
,
655
658
.
Dallegge
T. A.
(
2008
).
40Ar/39Ar geochronology of lavas from the Central Plateau Member of Plateau Rhyolite with implications for magma residence times and eruptive reoccurrence intervals, Yellowstone National Park
.
The Mountain Geologist
 
45
,
77
98
.
Davidson
P. M.
Lindsley
D. H.
(
1989
).
Thermodynamic analysis of pyroxene–olivine–quartz equilibria in the system CaO–MgO–FeO–SiO2
.
American Mineralogist
 
74
,
18
30
.
DeNosaquo
K. R.
Smith
R. B.
Lowry
A. R.
(
2009
).
Density and lithospheric strength models of the Yellowstone–Snake River Plain volcanic system from gravity and heat flow data
.
Journal of Volcanology and Geothermal Research
 
188
,
108
127
.
Doe
B. R.
Leeman
W. P.
Christiansen
R. L.
Hedge
C. E.
(
1982
).
Lead and strontium isotopes and related trace elements as genetic tracers in the Upper Cenozoic rhyolite–basalt association of the Yellowstone Plateau Volcanic Field
.
Journal of Geophysical Research: Solid Earth
 
87
,
4785
4806
.
Ellis
B.
Wolff
J.
Boroughs
S.
Mark
D.
Starkel
W.
Bonnichsen
B.
(
2013
).
Rhyolitic volcanism of the central Snake River Plain: a review
.
Bulletin of Volcanology
 
75
,
1
19
.
Ellis
B. S.
Wolff
J. A.
(
2012
).
Complex storage of rhyolite in the central Snake River Plain
.
Journal of Volcanology and Geothermal Research
 
211
,
1
11
.
Ellis
B. S.
Mark
D. F.
Pritchard
C. J.
Wolff
J. A.
(
2012
).
Temporal dissection of the Huckleberry Ridge Tuff using the 40Ar/39Ar dating technique
.
Quaternary Geochronology
 
9
,
34
41
.
Ellis
B. S.
Bachmann
O.
Wolff
J. A.
(
2014
).
Cumulate fragments in silicic ignimbrites: The case of the Snake River Plain
.
Geology
 
42
,
431
434
.
Farrell
J.
Smith
R. B.
Husen
S.
Diehl
T.
(
2014
).
Tomography from 26 years of seismicity revealing that the spatial extent of the Yellowstone crustal magma reservoir extends well beyond the Yellowstone caldera
.
Geophysical Research Letters
 
41
,
3068
3073
.
Gansecki
C.
Mahood
G.
McWilliams
M.
(
1998
).
New ages for the climactic eruptions at Yellowstone: single-crystal 40Ar/39Ar dating identifies contamination
.
Geology
 
26
,
343
346
.
Gansecki
C. A.
Mahood
G. A.
McWilliams
M. O.
(
1996
).
40Ar/39Ar geochronology of rhyolites erupted following collapse of the Yellowstone caldera, Yellowstone Plateau volcanic field: implications for crustal contamination
.
Earth and Planetary Science Letters
 
142
,
91
107
.
Geist
D.
Richards
M.
(
1993
).
Origin of the Columbia Plateau and Snake River plain: Deflection of the Yellowstone plume
.
Geology
 
21
,
789
792
.
Girard
G.
Stix
J.
(
2009
).
Magma recharge and crystal mush rejuvenation associated with early post-collapse Upper Basin Member rhyolites, Yellowstone Caldera, Wyoming
.
Journal of Petrology
 
50
,
2095
2125
.
Girard
G.
Stix
J.
(
2010
).
Rapid extraction of discrete magma batches from a large differentiating magma chamber: the Central Plateau Member rhyolites, Yellowstone Caldera, Wyoming
.
Contributions to Mineralogy and Petrology
 
160
,
441
465
.
Gualda
G. A.
Ghiorso
M. S.
Lemons
R. V.
Carley
T. L.
(
2012a
).
Rhyolite-MELTS: a modified calibration of MELTS optimized for silica-rich, fluid-bearing magmatic systems
.
Journal of Petrology
 
53
,
875
890
.
Gualda
G. A.
Pamukcu
A. S.
Ghiorso
M. S.
Anderson
A. T.
Jr
Sutton
S. R.
Rivers
M. L.
(
2012b
).
Timescales of quartz crystallization and the longevity of the Bishop giant magma body
.
PloS one
 
7
e37492
.
Guillong
M.
Meier
D.
Allan
M.
Heinrich
C.
Yardley
B.
(
2008
).
SILLS: a MATLAB-based program for the reduction of laser ablation ICP-MS data of homogeneous materials and inclusions
. In:
Sylvester
P.
(ed.)
Laser Ablation ICP–MS in the Earth Sciences: Current Practices and Outstanding Issues. Mineralogical Association of Canada, Short Course Series.
 Vancouver, pp.
328
333
.
Hanchar
J. M.
Watson
E. B.
(
2003
).
Zircon saturation thermometry
. In:
Hanchar
J. M.
Hoskin
P. W. O.
(eds)
Zircon. Mineralogical Society of America and Geochemical Society, Reviews in Mineralogy and Geochemistry
 
53
,
89
112
.
Hildreth
W.
Wilson
C. J.
(
1984
).
Compositional zoning of the Bishop Tuff
.
Journal of Petrology
 
48
,
951
999
.
Hildreth
W.
Christiansen
R. L.
O’Neil
J. R.
(
1984
).
Catastrophic isotopic modification of rhyolitic magma at times of caldera subsidence, Yellowstone Plateau Volcanic Field
.
Journal of Geophysical Research
 
89
,
8339
.
Hildreth
W.
Halliday
A. N.
Christiansen
R. L.
(
1991
).
Isotopic and chemical evidence concerning the genesis and contamination of basaltic and rhyolitic magma beneath the Yellowstone Plateau Volcanic Field
.
Journal of Petrology
 
32
,
63
138
.
Hooper
P. R.
Camp
V. E.
Reidel
S. P.
Ross
M. E.
(
2007
).
The origin of the Columbia River flood basalt province: Plume versus nonplume models
. In:
Foulger
G. R.
Jurdy
D. M.
(eds)
Plates, plumes, and planetary processes. Geological Society of America, Special Papers
 , pp.
635
668
.
Husen
S.
Smith
R. B.
Waite
G. P.
(
2004
).
Evidence for gas and magmatic sources beneath the Yellowstone volcanic field from seismic tomographic imaging
.
Journal of Volcanology and Geothermal Research
 
131
,
397
410
.
Hynek
S. A.
Brown
F. H.
Fernandez
D. P.
(
2011
).
A rapid method for hand picking potassium-rich feldspar from silicic tephra
.
Quaternary Geochronology
 
6
,
285
288
.
Johnson
D.
Hooper
P.
Conrey
R.
(
1999
).
XRF analysis of rocks and minerals for major and trace elements on a single low dilution Li-tetraborate fused bead
.
Advances in X-ray Analysis
 
41
,
843
867
.
Kent
A. J. R.
(
2008
).
In-situ analysis of Pb isotope ratios using laser ablation MC-ICP-MS: Controls on precision and accuracy and comparison between Faraday and ion counting systems
.
Journal of Analytical Atomic Spectrometry
 
23
,
968
975
.
Krukoski
J. C.
(
2002
).
A geologic data base (GeoGIS) and three-dimensional inversion for the density structure of the Yellowstone volcanic system
.
MS thesis, University of Utah, Salt Lake City, 96 pp.
Lanphere
M. A.
Champion
D. E.
Christiansen
R. L.
Izett
G. A.
Obradovich
J. D.
(
2002
).
Revised ages for tuffs of the Yellowstone Plateau volcanic field: Assignment of the Huckleberry Ridge Tuff to a new geomagnetic polarity event
.
Geological Society of America Bulletin
 
114
,
559
568
.
Lee
J.-Y.
Marti
K.
Severinghaus
J. P.
Kawamura
K.
Yoo
H.-S.
Lee
J. B.
Kim
J. S.
(
2006
).
A redetermination of the isotopic abundances of atmospheric Ar
.
Geochimica et Cosmochimica Acta
 
70
,
4507
4512
.
Leeman
W. P.
Phelps
D. W.
(
1981
).
Partitioning of rare earths and other trace elements between sanidine and coexisting volcanic glass
.
Journal of Geophysical Research
 
86
,
10193
10199
.
Leeman
W. P.
MacRae
C. M.
Wilson
N. C.
Torpy
A.
Lee
C. T.
Student
J. J.
Thomas
J. B.
Vicenzi
E. P.
(
2012
).
A study of cathodoluminescence and trace element compositional zoning in natural quartz from volcanic rocks: mapping titanium content in quartz
.
Microscopy and Microanalysis
 
18
,
1322
1341
.
Lowenstern
J. B.
Smith
R. B.
Hill
D. P.
(
2006
).
Monitoring super-volcanoes: geophysical and geochemical signals at Yellowstone and other large caldera systems
.
Philosophical Transactions of the Royal Society of London, Series A
 
364
,
2055
2072
.
Luttrell
K.
Mencin
D.
Francis
O.
Hurwitz
S.
(
2013
).
Constraints on the upper crustal magma reservoir beneath Yellowstone Caldera inferred from lake‐seiche induced strain observations
.
Geophysical Research Letters
 
40
,
501
506
.
Mahood
G.
(
1990
).
Second reply to comment of R. S. J. Sparks, H. E. Huppert and C. J. N. Wilson on ‘Evidence for long residence times of rhyolitic magma in the Long Valley magmatic system: the isotopic record in the precaldera lavas of Glass Mountain’
.
Earth and Planetary Science Letters
 
99
,
395
399
.
Mahood
G.
Hildreth
W.
(
1983
).
Large partition coefficients for trace elements in high-silica rhyolites
.
Geochimica et Cosmochimica Acta
 
47
,
11
30
.
Mahon
K. I.
(
1996
).
The New “York” regression: Application of an improved statistical method to geochemistry
.
International Geology Review
 
38
,
293
303
.
Mark
D.
Stuart
F.
De Podesta
M.
(
2011
).
New high-precision measurements of the isotopic composition of atmospheric argon
.
Geochimica et Cosmochimica Acta
 
75
,
7494
7501
.
Mark
D.
Renne
P.
Morgan
L.
Deino
A.
Smith
V.
Ellis
B.
Pearce
N.
(
2012
).
40Ar/39Ar dating of Pleistocene tuffs: an accurate age for the Matuyama–Brunhes geomagnetic reversal (MBGR)
.
American Geophysical Union Fall Meeting Abstracts
 
1
,
07
.
Mark
D. F.
Parnell
J.
Kelley
S. P.
Lee
M.
Sherlock
S. C.
Carr
A.
(
2005
).
Dating of multistage fluid flow in sandstones
.
Science
 
309
,
2048
2051
.
Matthews
N. E.
Vazquez
J. A.
Calvert
A. T.
(
2015
).
Age of the Lava Creek supereruption and magma chamber assembly at Yellowstone based on 40Ar/39Ar and U–Pb dating of sanidine and zircon crystals
.
Geochemistry, Geophysics, Geosystems
 
16
,
2508
2528
.
McDonough
W. F.
Sun
S.-S.
(
1995
).
The composition of the Earth
.
Chemical Geology
 
120
,
223
253
.
Minnis
P.
Harrison
E.
Stowe
L.
Gibson
G.
Denn
F.
Doelling
D.
Smith
W.
(
1993
).
Radiative climate forcing by the Mount Pinatubo eruption
.
Science
 
259
,
1411
1415
.
Morgan
L.
Mark
D.
Imlach
J.
Barfod
D.
Dymock
R.
(
2014
).
FCs-EK: a new sampling of the Fish Canyon Tuff 40Ar/39Ar neutron flux monitor
. In:
Jourdan
F.
Mark
D. F.
Verati
C.
(eds)
Advances in 40Ar/39Ar Dating: from Archaeology to Planetary Sciences. Geological Society, London, Special Publications
 
378
,
63
67
.
Obradovich
J. D.
(
1992
).
Geochronology of the Late Cenozoic volcanism of Yellowstone National Park and adjoining areas, Wyoming and Idaho. US Geological Survey, Open-File Report 92-408
 .
Putirka
K. D.
(
2008
).
Thermometers and barometers for volcanic systems
. In:
Putirka
K. D.
Tepley
F. J.
III
(eds)
Minerals, Inclusions and Volcanic Processes. Mineralogical Society of America and Geochemical Society, Reviews in Mineralogy and Geochemistry
 
69
,
61
120
.
Renne
P. R.
Deino
A. L.
Hames
W. E.
Heizler
M. T.
Hemming
S. R.
Hodges
K. V.
Koppers
A. A.
Mark
D. F.
Morgan
L. E.
Phillips
D.
(
2009
).
Data reporting norms for 40Ar/39Ar geochronology
.
Quaternary Geochronology
 
4
,
346
352
.
Renne
P. R.
Mundil
R.
Balco
G.
Min
K.
Ludwig
K. R.
(
2010
).
Joint determination of 40K decay constants and 40Ar*/40K for the Fish Canyon sanidine standard, and improved accuracy for 40Ar/39Ar geochronology
.
Geochimica et Cosmochimica Acta
 
74
,
5349
.
Renne
P. R.
Balco
G.
Ludwig
K. R.
Mundil
R.
Min
K.
(
2011
).
Response to the comment by W. H. Schwarz et al. on ‘Joint determination of 40K decay constants and 40Ar*/40K for the Fish Canyon sanidine standard, and improved accuracy for 40Ar/39Ar geochronology’ by P. R. Renne et al. (2010)
.
Geochimica et Cosmochimica Acta
 
75
,
5097
.
Renne
P. R.
Mulcahy
S. R.
Cassata
W. S.
Morgan
L. E.
Kelley
S. P.
Hlusko
L. J.
Njau
J. K.
(
2012
).
Retention of inherited Ar by alkali feldspar xenocrysts in a magma: Kinetic constraints from Ba zoning profiles
.
Geochimica et Cosmochimica Acta
 
93
,
129
142
.
Rivera
T. A.
Schmitz
M. D.
Crowley
J. L.
Storey
M.
(
2014
).
Rapid magma evolution constrained by zircon petrochronology and 40Ar/39Ar sanidine ages for the Huckleberry Ridge Tuff, Yellowstone, USA
.
Geology
 
42
,
643
646
.
Rivera
T. A.
Schmitz
M. D.
Jicha
B. R.
Crowley
J. L.
(
2016
).
Zircon Petrochronology and 40Ar/39Ar Sanidine Dates for the Mesa Falls Tuff: Crystal-scale Records of Magmatic Evolution and the Short Lifespan of a Large Yellowstone Magma Chamber
.
Geological Society of America Annual Meeting, Vancouver
.
Journal of Petrology
 
57
,
1677
1704
.
Self
S.
(
2006
).
The effects and consequences of very large explosive volcanic eruptions
.
Philosophical Transactions of the Royal Society of London, Series A
 
364
,
2073
2097
.
Simakin
A.
Bindeman
I.
(
2012
).
Remelting in caldera and rift environments and the genesis of hot, ‘recycled’ rhyolites
.
Earth and Planetary Science Letters
 
337
,
224
235
.
Stelten
M. E.
Cooper
K. M.
Vazquez
J. A.
Calvert
A. T.
Glessner
J. J.
(
2015
).
Mechanisms and timescales of generating eruptible rhyolitic magmas at Yellowstone Caldera from zircon and sanidine geochronology and geochemistry
.
Journal of Petrology
 
56
,
1607
1642
.
Szymanowski
D.
Ellis
B. S.
Bachmann
O.
Guillong
M.
Phillips
W. M.
(
2015
).
Bridging basalts and rhyolites in the Yellowstone–Snake River Plain volcanic province: The elusive intermediate step
.
Earth and Planetary Science Letters
 
415
,
80
89
.
Thomas
J. B.
Watson
E. B.
(
2012
).
Application of the Ti-in-quartz thermobarometer to rutile-free systems. Reply to: a comment on: ‘TitaniQ under pressure: the effect of pressure and temperature on the solubility of Ti in quartz’ by Thomas et al
.
Contributions to Mineralogy and Petrology
 
164
,
369
374
.
Vazquez
J. A.
Reid
M. R.
(
2002
).
Time scales of magma storage and differentiation of voluminous high-silica rhyolites at Yellowstone caldera, Wyoming
.
Contributions to Mineralogy and Petrology
 
144
,
274
285
.
Vazquez
J. A.
Kyriazis
S. F.
Reid
M. R.
Sehler
R. C.
Ramos
F. C.
(
2009
).
Thermochemical evolution of young rhyolites at Yellowstone: Evidence for a cooling but periodically replenished postcaldera magma reservoir
.
Journal of Volcanology and Geothermal Research
 
188
,
186
196
.
Wark
D. A.
Watson
E. B.
(
2006
).
TitaniQ: a titanium-in-quartz geothermometer
.
Contributions to Mineralogy and Petrology
 
152
,
743
754
.
Watts
K. E.
Bindeman
I. N.
Schmitt
A. K.
(
2011
).
Large-volume rhyolite genesis in caldera complexes of the Snake River Plain: insights from the Kilgore Tuff of the Heise Volcanic Field, Idaho, with comparison to Yellowstone and Bruneau–Jarbidge rhyolites
.
Journal of Petrology
 
52
,
857
890
.
Watts
K. E.
Bindeman
I. N.
Schmitt
A. K.
(
2012
).
Crystal scale anatomy of a dying supervolcano: an isotope and geochronology study of individual phenocrysts from voluminous rhyolites of the Yellowstone caldera
.
Contributions to Mineralogy and Petrology
 
164
,
45
67
.
Wolff
J.
Ramos
F.
Hart
G.
Patterson
J.
Brandon
A.
(
2008
).
Columbia River flood basalts from a centralized crustal magmatic system
.
Nature Geoscience
 
1
,
177
180
.
Wolff
J.
Ellis
B.
Ramos
F.
Starkel
W.
Boroughs
S.
Olin
P.
Bachmann
O.
(
2015
).
Remelting of cumulates as a process for producing chemical zoning in silicic tuffs: A comparison of cool, wet and hot, dry rhyolitic magma systems
.
Lithos
 
236
,
275
286
.
Wotzlaw
J.-F.
Bindeman
I. N.
Watts
K. E.
Schmitt
A. K.
Caricchi
L.
Schaltegger
U.
(
2014
).
Linking rapid magma reservoir assembly and eruption trigger mechanisms at evolved Yellowstone-type supervolcanoes
.
Geology
 
G35979.35971
.
Wotzlaw
J.-F.
Bindeman
I. N.
Stern
R. A.
D’Abzac
F.-X.
Schaltegger
U.
(
2015
).
Rapid heterogeneous assembly of multiple magma reservoirs prior to Yellowstone supereruptions
.
Scientific Reports
 
5
, doi: 10.1038/srep14026.

Supplementary data