Accessory mineral U–Th–Pb ages and ⁴⁰Ar/³⁹Ar eruption chronology, and their bearing on rhyolitic magma evolution in the Pleistocene Coso volcanic field, California

Abstract

We determined Ar/Ar eruption ages of eight extrusions from the Pleistocene Coso volcanic field, a long-lived series of small volume rhyolitic domes in eastern California. Combined with ion-microprobe dating of crystal ages of zircon and allanite from these lavas and from granophyre geothermal well cuttings, we were able to track the range of magma-production rates over the past 650 ka at Coso. In ≤230 ka rhyolites we find no evidence of protracted magma residence or recycled zircon (or allanite) from Pleistocene predecessors. A significant subset of zircon in the ~85 ka rhyolites yielded ages between ~100 and 200 Ma, requiring that generation of at least some rhyolites involves material from Mesozoic basement. Similar zircon xenocrysts are found in an ~200 ka granophyre. The new age constraints imply that magma evolution at Coso can occur rapidly as demonstrated by significant changes in rhyolite composition over short time intervals (≤10’s to 100’s ka). In conjunction with radioisotopic age constraints from other young silicic volcanic fields, dating of Coso rhyolites highlights the fact that at least some (and often the more voluminous) rhyolites are produced relatively rapidly, but that many small-volume rhyolites likely represent separation from long-lived mushy magma bodies.

Introduction

An increasing number of studies find that long “crystal residence times” are typical of silicic domes (see review by Reid 2003). In contrast, it appears that large volume eruptions contain datable crystal populations that are more homogeneous and on average closer in age to eruption (Simon et al. 2008). The reason for this is unclear, but may be due to the greater energy input required to produce large eruptions. Further studies bearing on the inverse correlation between crystal (i.e., “magma”) residence time with erupted volume, and moreover on criteria for the possible cause, appear warranted. The spectrum of bimodal continental volcanism ranges from systems that are composed of small monogenetic basaltic centers and rare, small-volume rhyolites (e.g., Coso) to voluminous, dominantly silicic caldera systems (e.g., Long Valley). The existence of numerous small-volume rhyolite domes of the Coso volcanic field located in eastern California in the absence of a voluminous caldera-related eruption presents the opportunity to study the evolution of an end-member among long-lived silicic magma centers and to improve our understanding of the time scales of silicic magma processes. Further dating of Coso rhyolites provides an opportunity to explore and refine ideas on how large shallow silicic magma systems develop and how quickly their magmatic and eruptive behavior evolves (e.g., Bacon et al. 1981).

In this paper we report ⁴⁰Ar/³⁹Ar sanidine ages and U–Th–Pb accessory mineral ages for late Pleistocene Coso rhyolite lavas. The ⁴⁰Ar/³⁹Ar data yield more precise eruption ages than the K–Ar ages previously reported for Coso rhyolites and other volcanic rocks by Duffield et al. (1980). Several extrusions at Coso are nearly aphyric and contain little to no feldspar. In such cases we apply similar ⁴⁰Ar/³⁹Ar methods to potassium-rich obsidian samples, which along with feldspar ages, provide an improved eruption chronology. Crystallization and melt differentiation occur at magmatic temperatures at times that may significantly predate eruption (cf. Simon et al. 2008; and references therein). In favorable cases a comparison between isotopic dating of accessory minerals and the age of individual eruptions can be used to constrain the time scales over which physiochemical variations in magmas are produced (e.g., Vazquez and Reid 2002). At Coso the age distribution, chemistry, and textures of zircons from the ~0.6 Ma Devils Kitchen rhyolite led Miller and Wooden (2004) to suggest that the erupted magma was remobilized intrusions and/or near-solidus crystal mush that were emplaced in the crust over an ~200 ka period.

There is little evidence for cognate crystal recycling or protracted magma residence time when ⁴⁰Ar/³⁹Ar eruption ages are compared to zircon and allanite crystallization ages in the younger (≤230 ka) Coso rhyolites. Moreover, during this time period we find clear evidence for crustal assimilation and crystal inheritance as indicated by Mesozoic zircon ages. The absence of protracted residence times and of evidence for recycling in the younger (≤230 ka) Coso rhyolites contrast sharply with the discrete, i.e., 10’s to 100’s ka age spans reported to represent pre-eruption crystal residence (“magma residence times” senso lato) in the older ~0.6 Ma Devils Kitchen rhyolite, as well as at other large silicic magma centers, e.g., the Whakamaru Tuff at Taupo Volcano, New Zealand (Brown and Fletcher 1999), some post-caldera and the caldera-related rhyolites at Yellowstone (Bindeman, et al. 2001; Vazquez and Reid 2002), and a number of Long Valley rhyolites (Reid and Coath 2000; Simon and Reid 2005; Simon, et al. 2007). These differences amplify our need to distinguish between evidence for protracted fractional crystallization differentiation within long-lived shallow magma reservoirs and evidence for remobilization of crystals formed in relatively short-lived melt bodies (cf. Bachmann and Bergantz 2003; Bacon and Lowenstern 2005; Charlier et al. 2005; Schmitt and Simon 2004) and ultimately why there exists such a range of behavior within and among long-lived silicic magma centers.

Geology of Coso volcanic field

The Coso volcanic field of California is located at the margin of the Basin and Range province, east of the Sierra Nevada and Owens Valley (Fig. 1). The Coso field is underlain by pre-Cenozoic granitic plutons, dioritic to gabbroic plutons, and metamorphic basement rocks (Duffield et al. 1980). Collectively these units compose the Coso Range that is structurally separated from the main body of the composite Sierra Nevada batholith along the Sierra Nevada fault zone. Although a single K–Ar biotite age of ~90 Ma has been reported, most of the granitic plutons in the Coso Range are inferred to be Late Mesozoic, whereas most of the mafic plutons and metamorphic rocks appear to be somewhat older (Duffield et al. 1980). Chen and Moore (1979) report concordant ~150 Ma U–Pb zircon ages for dikes related to the Independence swarm, trending northwest–southeast in the Alabama Hills (northwest of the Coso Range) and in the Argus Range (east of Coso). There are younger east-west-trending dikes intruding ~102 Ma granitic plutons locally in the northern part of the Range that contain zircon as young as ~83 Ma that were probably emplaced during the Late Cretaceous (Kylander-Clark et al. 2005).

Fig. 1

Map showing Pleistocene volcanism at Coso in eastern California (after Bacon 1982). Sampling locations for dated rhyolites and granophyre drill cuttings from geothermal well shown for reference

Miocene and Pliocene basalt flows that represent the initiation of Late Cenozoic volcanism are now tilted, whereas more recent flows are flat lying indicating that uplift of the Coso Range and the onset of volcanism appear to have been nearly contemporaneous (Duffield et al. 1980). Monastero et al. (2005) report that the Quaternary rhyolite field and active geothermal system are located in a releasing bend of the dextral strike-slip Sierra Nevada fault zone above a nascent metamorphic core complex. Monastero et al. (2005) concluded that magma generation was related to crustal thinning, shallowing of the mid-crust, and proximity to the asthenospheric mantle.

Interest in the Coso geothermal system has prompted multidisciplinary research on the volcanic field since the 1970s (e.g., see JGR, volume 85, 1980 special issue), including geologic mapping (Duffield and Bacon 1981) and K–Ar geochronology (Duffield et al. 1980; Lanphere et al. 1975). Bacon et al. (1981) assigned the 38 Quaternary rhyolite extrusions to seven groups, each of which consisted of chemically similar units that had similar apparent eruption ages. Although the precision and accuracy of ⁴⁰Ar/³⁹Ar dating techniques are typically superior to those of dates obtained by the K–Ar method, which may be biased by excess ⁴⁰Ar contamination (e.g., Renne et al. 1997) or incomplete extraction of argon from sanidine (e.g., McDowell 1983), the K–Ar results are adequate to define the general history of the Coso volcanic field. Volcanism began ~6 Ma. Since ~4 Ma, ~35 km³ of lava has erupted in the Coso Range and about 30 km³ of this was emplaced in the Pliocene between ~4 and 2.5 Ma (Duffield et al. 1980). In this study we have focused on the Pleistocene rhyolites (≤1.1 Ma), ~2 km³ of extruded rhyolite likely related to the active magma and geothermal systems. Heat-flow measurements (Combs 1980) imply that the volume of intruded silicic magma may be as much as 100 times greater (Bacon et al. 1981). Similar to that characteristic of the Pliocene buildup, the volume of silicic eruptions during the Pleistocene increased (since ~0.25 Ma) toward the present (Duffield et al. 1980).

Pleistocene volcanism at Coso is bimodal in composition and is composed of basalt and rhyolite. Basaltic vents tend to lie outside of the focus of silicic magmatism at any given time, an observation often interpreted as evidence that basaltic extrusion is hindered by the existence of a low-density magma body (cf. Bacon 1985). Moreover, frequent seismicity (Walter and Weaver 1980), as well as teleseismic P-delay tomography (Reasenberg et al. 1980) provides evidence for a magma body with a current estimate for the top of the body at ~5 km below the surface (Wilson et al. 2003). Pyroxene thermobarometry of Pliocene (~3.5 to 2 Ma) and Pleistocene (~1.1 to 0.1 Ma) basalts at Coso suggests that crystallization occurred at ~15 km (~550 MPa) and then ~25 to 30 km (~900 MPa) depths, respectively (Mordick and Glazner 2006). Manley and Bacon (2000) used the Al-in-hornblende barometer to show that mineral growth within the ~0.6 to 0.3 Ma rhyolites imply that the depth to the top of the silicic magma body was ~10 km (~270 MPa) and that by ~0.04 Ma it had risen to ~5.4 km (140 MPa). Following Bacon et al. (1981), Mordick and Glazner (2006) concluded that Pleistocene basalts likely provided heat needed to generate Pleistocene rhyolite and may have been trapped beneath the Pleistocene rhyolite magma body, a possible explanation for why the Pleistocene basalts records greater depths than the Pliocene basalts.

Whether or not a single long-lived shallow magma chamber was present throughout Pleistocene time, Duffield et al. (1980; and references therein) reported that some level of geothermal actively likely existed during this interval. Duffield et al. (1980) described evidence for fumaroles thought to be related to the ~0.6 Ma Devils Kitchen dome and the adjacent Nicol area, and also noted past hydrothermal activity near Sugarloaf Mountain and Coso Hot Springs based on a K–Ar age of 234 ± 22 ka for an overlying basalt flow that is unaffected by thermal spring activity. More recent evidence of geothermal activity is demonstrated by pollen encapsulated in a travertine deposit at Wheeler Prospect (~2 km south of Coso Hot Springs) that yields an age of ~12 ka (Kovac et al. 2006; and reference therein). Today the abundant fumaroles and rigorously boiling mud holes, high seismicity (Walter and Weaver 1980), high heat flow (~120 to ~1,045 mW/m² in the center of the field, which is above present day Basin and Range levels ≤110 mW/m²; Combs 1980), and the generation of >250 MW of electrical power (Monastero 2002) are manifestations of the active geothermal system.

Description of rhyolites and dated materials

At least 38 rhyolite domes and lava flows were extruded at Coso during the Pleistocene. The dome names (numbers), their field relationships, and their major and trace element compositions used here to develop models for the petrogenetic and eruption history of the magma system come from Duffield et al. (1980) and Bacon et al. (1981). In general, Coso rhyolites are crystal-poor with the two oldest domes (38 and 28) and a third dome (5) located at the north end of the field being the exceptions. Typical crystal contents range from ~1 to <0.001 wt.% (Bacon et al. 1981). Despite their low crystal contents, Coso rhyolites commonly contain xenocrysts that can be distinguished by composition and morphology. Detailed descriptions of their mineral assemblages (including the presence of multiple populations of ferromagnesian silicate phases) can be found in Manley and Bacon (2000). Representative rhyolites have been selected for this study and for completeness a brief description and their mineral assemblages are included in Table 1. Samples were collected by J. Vazquez (VCD series) and by W. Amidon, K. Farley, and P. Renne (PRCO series) separately under the guidance of F. Monastero. The granophyre cuttings from injection well 46A-19RD obtained during the Coso Enhanced Geothermal Systems Experiment were provided by J. Moore. Allanites from dome 24 come from a sample collected by C. Bacon. The “Devils Kitchen” rhyolite (dome 28) was sampled twice (PRCO-5 and PRCO-6). Miller and Wooden (2004) report that almost all Devils Kitchen zircons are euhedral, but that many have anhedral, embayed cores with euhedral overgrowths. Mafic enclaves and xenocrysts are also present (Bacon and Metz 1984). Three samples from dome 5 were collected (VCD03, PRCO-2, PRCO-3). PRCO-2 and PRCO-3 are from the surface of the dome obtained ~100 m apart. Obsidian was obtained from domes 17 (VCD01) and 20 (VCD06), but contain inadequate amounts of feldspar. Likewise, from the sample material collected of dome 25 (VCD09) we obtained insufficient accessory minerals for dating purposes. Rare aphyric mafic enclaves and forsteritic olivine xenocrysts have been reported for both dome 24 and 25 lavas (Bacon and Metz 1984; Manley and Bacon 2000). The allanites from domes 6, 24, and 26 are up to ~100 μm in size and are anhedral to subhedral while zircons are typically subhedral and up to ~200 μm in size with inclusions of apatite and glass. Both allanite and zircon are microphenocrysts in the holohyaline groundmass. Injection well 46A-19RD is located ~1 km north of dome 24 about a third of the distance to Sugarloaf Mountain (dome 26). The subsurface granophyre was sampled as drill cuttings from a depth interval of 11,420–11,440 ft (~3,485 m) in well 46A-19RD (Kovac et al. 2006).

Table 1 Mineralogy and sample description of studied Coso rhyolites and granophyre

Methods

⁴⁰Ar/³⁹Ar dating of obsidian and alkali feldspar

Obsidian samples were crushed to produce fragments of brown glass, free of phenocrysts, and cleaned in distilled water with an ultrasonic disintegrator. Feldspars were separated from crystal-poor obsidian, perlitic, and pumiceous glasses. Alkali feldspars were concentrated by heavy liquid and magnetic separation techniques, and cleaned by short ultrasonic baths of dilute hydrofluoric acid followed by distilled water. Alkali feldspars were distinguished from residual plagioclase using immersion oils and petrographic methods, and then selected by hand for analysis. Two irradiations were used. The VCD-series materials, along with 37 Alder Creek feldspars (1,193 ± 1 ka; Nomade et al. 2005), used to monitor the neutron fluence, were irradiated at the Oregon State University TRIGA reactor in the Cd-lined CLICIT facility for 30 min within a series of stacked aluminum disks, permitting precise monitoring of both vertical (J = ~1.341 × 10⁻⁴ to ~1.350 × 10⁻⁴) and lateral (negligible) flux gradients. The PRCO-series materials, along with 54 Alder Creek feldspar crystals, were irradiated in a similar fashion in a single aluminum disk [J = ~1.41 (±0.01) × 10⁻⁴]. Total fusion and step-heating feldspar and glass analyses were performed. Glasses and three to eight grain feldspar aliquots were analyzed by laser step-heating following procedures described previously (Nomade et al. 2005; and references therein). Age spectra derived from laser step-heating of samples from previously dated domes show that excess ⁴⁰Ar contamination likely biased some K–Ar results. However, ⁴⁰Ar/³⁹Ar analyses of the studied rhyolites with disturbed model ages (i.e., assuming atmospheric initial Ar), but well-defined Ar isochrons are considered to provide accurate eruption ages (Fig. 2). When comparing U–Th–Pb crystal ages and ⁴⁰Ar/³⁹Ar eruption ages, it is important to consider the intercalibration bias between the two systems. This bias, mainly a product of error in the electron capture decay constant for ⁴⁰K (Min et al. 2000), is small (ca. 1%) compared with the differences considered in this paper. For comparison with existing data, a shift upward of ~10 ka to the age of the Devils Kitchen rhyolite is needed to corrected for the bias. This is made using the ⁴⁰K decay constants of Mundil et al. (2006), and the ⁴⁰Ar/³⁹Ar standard basis of Renne et al. (1998).

Fig. 2

⁴⁰Ar/³⁹Ar ages for feldspar and glass of selected high-silica rhyolite domes and lava flows of the Coso volcanic field. Dated domes include eight samples from four of seven chemically distinct Pleistocene rhyolite groups. Analyses are alkali-feldspar except for Sugarloaf Mountain which is obsidian and include: a Multi-grain step-heating and total fusion isochron and plateau ⁴⁰Ar/³⁹Ar ages of the Group 2 Devils Kitchen rhyolite (CO-5, CO-6), b multi-grain step-heating and total fusion isochron and plateau ⁴⁰Ar/³⁹Ar ages of the Group 4 dome 5 rhyolite (CO-2, CO-3, and VCD-03), c multi-grain step-heating ⁴⁰Ar/³⁹Ar isochron ages for Group 6 rhyolite domes 24 (VCD08), 25 (VCD09), and 36 (VCD07), and d multi-grain step-heating isochron and plateau ⁴⁰Ar/³⁹Ar ages for Group 7 Sugarloaf Mountain (VCD05), Cactus Peak (VCD01), and dome 31 rhyolites (VCD02)

Ion microprobe U–Th and U–Pb ages of zircon and allanite

Accessory minerals were separated by heavy liquid techniques. U–Th disequilibrium dating was performed on individual zircon and allanite crystals using the UCLA Cameca ims 1270 secondary ion mass spectrometer (ion microprobe), based on techniques from Reid et al. (1997), Schmitt (2006), and Vazquez and Reid (2004). An <20–50 nA mass-filtered ¹⁶O⁻-beam was focused into a ~35 μm × 30 μm oval spot. Secondary ions were accelerated at 10 keV with an energy bandpass of 50 eV and analyzed at a mass resolution of <4,800 for zircon and ~9,000 for allanite using an axial electron multiplier collector in peak jumping mode. For zircon, relative sensitivities for ²³⁸UO and ²³²ThO were calibrated by measuring the radiogenic ²⁰⁶Pb/²⁰⁸Pb ratio of concordant reference zircons AS-3 and 91,500 (Paces and Miller 1993; Wiedenbeck et al. 1995). Ion microprobe analysis of allanites from the Bishop Tuff (777 ka) and Mesa Falls Tuff (1.2 Ma) yield ²³⁰Th/²³⁸U within error of secular equilibrium, as expected for allanites older than ~500 ka (Reid et al. 1997). Coso zircons found to be in secular equilibrium were analyzed for their ²³⁸U–²⁰⁶Pb crystallization ages, employing established methods (e.g., Dalrymple et al. 1999; Quidelleur et al. 1997). When possible and appropriate (i.e., young), individual U–Pb ages were ²⁰⁷Pb-corrected for common Pb using a typical whole rock initial common ²⁰⁷Pb/²⁰⁶Pb ratio of 0.818 (this study). The common lead composition is likely to be largely intrinsic, i.e., incorporated at the time of crystallization and trapped within inclusions, but could also be partly due to contamination by Southern California anthropogenic Pb, which has a similar composition (Sañudo-Wilhelmy and Flegal 1994). The ²⁰⁷Pb-corrected ages are also adjusted for initial U–Th disequilibrium due to the deficit in radiogenic ²⁰⁶Pb originating from the initial deficit in ²³⁰Th relative to ²³⁸U following Schärer (1984). The magnitude of this deficit can be estimated by comparing the Th/U ratios in zircon to those of the host magmas (i.e., glass, Table 3). The absolute effect of this correction (typically adding 80–90 ka, e.g., Simon et al. 2008) is accurate to ~10 ka, allowing for the range of whole rock Th/U ratios observed. Most applications of this correction assume that the melts were in secular equilibrium with respect to ²³⁸U–²³⁰Th when the zircons crystallized. However, ²³⁰Th excesses during crystallization will lead to calculated ages that overestimate the true ages by 4 ka per 5% ²³⁰Th excess. One of our sampled rhyolites (Dome 24) contains an ~10% ²³⁰Th excess, whereas the Sugarloaf rhyolite contains an ~5% ²³⁸U excess. Whole rock samples were analyzed for their U and Th isotope compositions by thermal ionization mass spectrometry (TIMS) at UCLA following the procedure of Bohrson and Reid (1998). Blanks were negligible (<0.01% for U and Th). Analysis of UCSC-Th-A standard during analyses yielded a weighted mean ²³²Th/²³⁰Th of 170,254 ± 0.3% (2 σ, n = 3). The whole rock compositions of the Coso extrusions were assumed to be representative of the magma from which the granophyre zircons grew, and were used to calculate model ²³⁸U–²³⁰Th crystallization ages.

Ion microprobe zircon Ti abundances

Selected zircons in this study were analyzed for Ti abundances subsequent to U–Pb and/or U–Th dating. The UCLA Cameca ims 1270 ion microprobe was tuned for high mass resolution analysis at a mass resolving power M/ΔM = 8,000 (at 10% peak height), sufficient to largely resolve interferences such as ⁴⁸Ca⁺ and ⁹⁶Zr²⁺. A ~10 nA ¹⁶O⁻ primary beam was focused to a ~10 μm spot, and following a 120-s pre-sputter period with a 10 μm × 10 μm raster to minimize surface contamination, ⁴⁸Ti⁺ and ²⁸SiO⁺ intensities were measured simultaneously for 60 s using two electron multipliers. Ti concentrations were calculated using relative sensitivity factors determined on SL13 zircon (6.32 ± 0.33 ppm Ti; Harrison and Schmitt 2007). External reproducibility of Ti/Si relative sensitivity is estimated from replicate analyses on NIST 610 glass over a 2-day analysis period, which yielded a standard deviation of ~1% (relative; n = 10). All sputter pits were inspected for beam overlap on inclusions or adherent glass using a petrographic microscope, and only results for spots that were entirely contained within zircon are reported here (Tables 4, 5).

Results

Feldspar and obsidian ⁴⁰Ar/³⁹Ar eruption ages

⁴⁰Ar/³⁹Ar ages for eight Pleistocene domes, sampling four of the seven previously defined eruption groups, are summarized in Table 2 (Fig. 2a–d and Appendix). The ⁴⁰Ar/³⁹Ar dating confirms that most silicic extrusions at Coso occurred in the range ~600 to ~60 ka, in general similar to that reported by previous K–Ar studies, but provides substantially better temporal resolution among individual eruptions. Sanidine from two samples from dome 28 (Group 2) the “Devils Kitchen” rhyolite was dated. Multi-grain step-heated sanidine analyses yield isochron ages of 611 ± 6 ka (all reported errors are given at 1σ or 68% confidence unless stated otherwise) (⁴⁰Ar/³⁶Ar₀ = 297 ± 3, MSWD = 0.6) and 612 ± 7 ka (⁴⁰Ar/³⁶Ar₀ = 296 ± 11, MSWD = 1.1) for samples PRCO-5 and PRCO-6, respectively. Multi-grain total fusion sanidine analyses yield isochron ages of 606 ± 7 ka (⁴⁰Ar/³⁶Ar₀ = 305 ± 5, MSWD = 1.2) and 618 ± 7 ka (⁴⁰Ar/³⁶Ar₀ = 299 ± 12, MSWD = 1.0), for samples PRCO-5 and PRCO-6, respectively. All 23 measurements (step-heating and total fusion) from both Devils Kitchen rhyolite samples considered together yield an isochron age of 613 ± 6 ka (⁴⁰Ar/³⁶Ar₀ = 298 ± 2, MSWD = 0.9) that is within error of, but more precise than, the reported 587 ± 18 ka K–Ar date (Duffield et al. 1980).

Table 2 ⁴⁰Ar/³⁹Ar eruption ages for select Pleistocene Coso rhyolites

Three samples were dated from dome 5 (Group 4), the northern most Pleistocene rhyolite at Coso. Seven separate multi-grain sanidine isochron ages for sample VCD03 range from ~204 to ~240 ka. An isochron fit to the combined data (Fig. 2b) reveals scatter toward greater ⁴⁰Ar/³⁶Ar₀ isotope ratios that are consistent with presence of excess ⁴⁰Ar. Variable proportions of excess and atmospheric ⁴⁰Ar preclude the applicability of isochrons, which require a binary mixture of components. To mitigate this problem, individual analyses with comparatively high ⁴⁰Ar/³⁶Ar₀ were incrementally excluded from below the inverse isochron until a sufficiently high probability of fit (0.1) was obtained. The resultant isochron age is 216 ± 6 ka (⁴⁰Ar/³⁶Ar₀ = 296 ± 13, MSWD = 1.5), which is ~45 ka older than the nominal age for Group 4 rhyolites and significantly older than its existing 170 ± 11 ka K–Ar date (Bacon et al. 1981). This model age is supported by sanidine measurements from the two other samples of dome 5 (PRCO-2 and PRCO-3) that yield multi-grain step-heating isochron ages of 219 ± 6 ka (⁴⁰Ar/³⁶Ar₀ = 283 ± 6, MSWD = 1.0) and 214 ± 4 ka (⁴⁰Ar/³⁶Ar₀ = 292 ± 2, MSWD = 1.5), respectively. Multi-grain total fusion sanidine analyses yield isochron ages of 217 ± 9 ka (⁴⁰Ar/³⁶Ar₀ = 292 ± 5, MSWD = 1.6) and 215 ± 4 ka (⁴⁰Ar/³⁶Ar₀ = 298 ± 3, MSWD = 0.8), for samples PRCO-5 and PRCO-6, respectively. If all 28 measurements (step-heating and total fusion) from all three dome 5 samples are considered together an isochron age of 215 ± 3 ka (⁴⁰Ar/³⁶Ar₀ = 293 ± 2, MSWD = 1.2) is determined. The significance of the systematically subatmospheric trapped ⁴⁰Ar/³⁶Ar₀ ratios for these two samples is unclear, but may derive from cosmogenic ³⁶Ar (Renne et al. 2001) accumulated in these exposed surface samples.

The three dated southern domes are older than their reported mean Group 6 age of ~90 ka. It is probable that the anomalously young K–Ar dates reflect incomplete extraction of Ar from feldspar and other phases contained in the analyzed obsidian. Domes 24 (VCD08) and 25 (VCD09) yield similar multi-grain feldspar isochron ages of 115 ± 4 ka (⁴⁰Ar/³⁶Ar₀ = 309 ± 7, MSWD = 0.2) and 117 ± 6 ka (⁴⁰Ar/³⁶Ar₀ = 297 ± 5, MSWD = 0.4) that are nominally 30–40 ka older than their reported 84 ± 32 and 86 ± 24 ka K–Ar ages (Lanphere et al. 1975), respectively. Both composite isochrons are composed of two separate multi-grain sanidine step-heating analyses. In total, the four separate isochrons range in age from ~113 to ~119 ka. For the third previously undated Group 6 dome (dome 36, VCD07), an isochron populated with five multi-grain anorthoclase step-heating analyses yields an age of 148 ± 6 ka (⁴⁰Ar/³⁶Ar₀ = 296 ± 2, MSWD = 0.2). The separate multi-grain anorthoclase isochron ages range from ~141 to ~160 ka.

Analyses of three Group 7 samples suggest that eruptions in this group appear to have occurred in (at least) two events. The small previously undated dome 31 (VCD02) from the northern part of the Coso field yields a multi-grain step-heated sanidine isochron age of 56 ± 9 ka (⁴⁰Ar/³⁶Ar₀ = 330 ± 40, MSWD = 1.0). Individual multi-grain analyses (also used in the composite isochron age) yield a plateau age of 66 ± 5 ka MSWD = 1.5 (see Fig. 2d). Although the data are sparse, we believe that the eruption age of dome 31 is similar to the reported ~60 ka age of Group 7 rhyolites. Coeval eruption of Sugarloaf and Cactus Peak domes, however, appear to have occurred at ~86 ka: ~40 ka later and ~15 ka earlier, respectively, than their K–Ar dates would indicate (Lanphere et al. 1975). Step-heating analyses of glass from Sugarloaf dome 26 (VCD05) gives an isochron age of 86 ± 2 ka (⁴⁰Ar/³⁶Ar₀ = 299 ± 2, MSWD = 0.9) and multi-grain step-heating analyses of sanidine from Cactus Peak dome 6 (VCD01) yield a combined isochron age of 84 ± 7 ka (⁴⁰Ar/³⁶Ar₀ = 306 ± 5, MSWD = 1.5). Considered separately, the two multi-grain step-heating analyses of Cactus Peak sanidine yield isochron ages of ~86 and ~96 ka, respectively. Obsidian from four Coso domes, including domes 17 and 20, in which there was insufficient feldspar for dating, and dome 6 (Cactus Peak) and dome 31, which were dated by feldspar, contain largely atmospheric Ar (>50 to 99%) and provide no reliable age information.

U–Th isotope results and ²³⁸U–²³⁰Th allanite and zircon model crystallization ages

The ²³⁸U–²³⁰Th–²³²Th compositions of allanite, zircon, their associated lavas, and zircon from the granophyre cuttings sample, as well as ²³⁸U–²³⁰Th model ages of zircon and allanite (see Reid et al. 1997), are listed in Tables 3 and 4. Approximately 50% of all of the zircons dated (n = 43) in this study yielded ²³⁸U–²³⁰Th results within error of secular equilibrium and these were reanalyzed for their ²³⁸U–²⁰⁶Pb ages (see Table 5). Representative secondary electron and cathodoluminescence images of analyzed zircon and allanite are shown in Fig. 3. Using whole rock compositions to account for initial ²³⁰Th/²³²Th during crystallization (Reid et al. 1997), allanites from dome 24 (VCD08) yield model ²³⁸U–²³⁰Th ages that cluster at ~115 ka, whereas allanites from Cactus Peak (dome 6, VCD01) and Sugarloaf (dome 26, VCD05) give ages that cluster around ~85 ka. Two allanites yield ages that are >300 ka, one each from the Cactus Peak and the Sugarloaf rhyolites. Excluding the >300 ka crystal, Cactus Peak allanites yield a weighted mean ²³⁸U–²³⁰Th model age of 85 ± 10 ka (95% CI, n = 6, MSWD = 1.0), and a Sugarloaf allanite gives a model age of 86 ± 8 ka (Fig. 4a). The four Cactus Peak zircons give a weighted mean ²³⁸U–²³⁰Th model age of 81 ± 7 ka (95% CI, n = 4, MSWD = 0.5) that is indistinguishable in age from the coexisting allanites as well as the eruption age measured by ⁴⁰Ar/³⁹Ar dating. A zircon from Sugarloaf yields an older model ²³⁸U–²³⁰Th age of 104 ± 9 ka (Fig. 4a). Based on their similar trace element concentrations (Bacon et al. 1981) and ⁴⁰Ar/³⁹Ar ages of the rhyolites (Table 2), the allanite and zircon model ages may be grouped. A combined weighted mean model age for zircons and allanites from Cactus Peak and Sugarloaf rhyolites (domes 6 and 26) is 83 ± 8 ka (95% CI, n = 12, MSWD = 1.0). A combined internal isochron ²³⁰Th/²³²Th–²³²Th/²³⁸U age for the allanite and zircon from domes 6 and 26 is 84 ± 4 ka (2 σ, n = 12, MSWD = 1.0, Fig. 5). The U/Th ratio of the allanites shows a twofold variation of 0.0036 to 0.0074 that does not correlate with ²³⁸U–²³⁰Th model age (see Fig. 4a).

Table 3 U–Th isotope results and ²³⁸U–²³⁰Th ages for Coso allanites

Table 4 U–Th isotope results and ²³⁸U–²³⁰Th ages for Coso zircons

Table 5 U–Pb isotope compositions and zircon ages

Fig. 3

Backscattered-electron (BSE, left) and cathodoluminescence or secondary electron (CL/SE, right) images of dated accessory minerals. I: Representative image of subhedral to anhedral Pleistocene zircon. II: Representative image of euhedral Mesozoic zircon. III: Representative image of subhedral to anhedral allanite. A allanite, G glass, Z zircon, cpx clinopyroxene

Fig. 4

²³⁸U–²³⁰Th ages and (²³⁸U)/(²³²Th) compositions for allanite and zircon from selected Coso rhyolites. a Cactus Peak and Sugarloaf rhyolites, b Dome 24, and c Granophyre drill core 46A-19RD. Symbols are defined in figure. Uncertainty of ion microprobe spot analyses are 1 σ. ⁴⁰Ar/³⁹Ar eruption ages (this study) included for reference. Vertical shaded bands show uncertainty for weighted mean ²³⁸U–²⁰⁶Pb (dashed) and ²³⁸U–²³⁰Th isochron (dot-dashed) ages for Pleistocene zircon from the granophyre

Fig. 5

Internal U–Th isochrons for zircon from granophyre drill cuttings and for zircons and allanites from Cactus Peak and Sugarloaf Mt. Symbols defined in figure. Shaded error envelopes are 2σ

Allanites from dome 24 (VCD08, Group 6) yield a weighted mean age of 117 ± 44 ka (95% CI, n = 4, MSWD = 0.2) and exhibit variation in the U/Th ratio of 0.0054 to 0.0079 (Fig. 4b). Allanite from dome 24 is indistinguishable in age from the time of eruption as determined by ⁴⁰Ar/³⁹Ar dating.

The granophyre sample is altered and consists of drill cuttings, and therefore it is impossible to reliably determine its whole rock ²³⁸U–²³⁰Th isotope composition. Using the whole rock ²³⁸U–²³⁰Th composition of Cactus Peak rhyolite, zircon model ²³⁸U–²³⁰Th ages give a weighted mean age of 192 ± 11 ka (95% CI, n = 9, MSWD = 0.2; Fig. 4c). An internal isochron age of $218_{{ - 46{\text{ka}}}}^{{ + 60{\text{ka}}}}$ (95% CI, n = 8, MSWD = 0.57) is determined for the granophyre zircons. This age is less precise because of the limited spread in zircon Th/U ratios as compared to lava (melt)-zircon pairs, but overlaps with the weighted mean model age, which demonstrates that our assumptions related to host melt composition are reasonable (Fig. 5). The isochron age for the granophyre zircons is distinct from the zircon and allanite ages from the Group 7 rhyolites, but similar to the ~215 ka ⁴⁰Ar/³⁹Ar eruption age of dome 5. Three zircons from the granophyre sample 46A-19D and five additional crystals from the lavas yield ²³⁸U–²³⁰Th in equilibrium. These crystals were dated by the U–Pb method (see Table 5).

U–Pb isotope results and ²³⁸U–²⁰⁶Pb zircon crystallization ages

Many dated zircons exhibit Mesozoic ²³⁸U–²⁰⁶Pb ages (Table 5; Fig. 6). After correction for disequilibrium, Pleistocene zircons from the granophyre sample yield a weighted mean ²³⁸U–²⁰⁶Pb age of 195 ± 17 ka (95% CI, n = 5, MSWD = 1.1) that is in agreement with their ²³⁸U–²³⁰Th age (Figs. 4, 5). ²³⁸U–²⁰⁶Pb analysis of other zircon from Cactus Peak, Sugarloaf, and the granophyre sample 46A-19D (that were not analyzed for ²³⁸U–²³⁰Th) reveals grains with little radiogenic Pb (²⁰⁷Pb/²⁰⁶Pb ≥0.6) as well as grains that yield Mesozoic ages (Table 5). The Mesozoic zircons define a ²³⁸U–²⁰⁶Pb age distribution with peaks at ~95 Ma, ~145 to 155 Ma, and ~180 to 200 Ma (Fig. 7). A single zircon analysis yields an age of ~60 Ma. Zircons from granophyre cuttings have ages that fall within the older Mesozoic age groups, but their origin(s) remain ambiguous because of potential contamination from country rock during drilling.

Fig. 6

Tera Wasserburg diagram showing ²³⁸U–²⁰⁶Pb zircon ages from granophyre drill core. Error bars and shaded envelope are 1σ. Inset shows Mesozoic xenocrysts. Concordia shown corrected for initial U–Th disequilibrium

Fig. 7

Histogram and relative probability curve showing bimodal (Pleistocene and Mesozoic) accessory mineral age distribution for Coso domes 6, 24, 26 and granophyre

Compositional zoning in dated allanite and zircon

The U/Th ratios of zircon interiors vary from ~1.0 to ~2.3 among those that exhibit clear ²³⁸U–²³⁰Th disequilibrium. This range is indistinguishable among the grains from the different extrusions and the granophyre. The U/Th ratios of the interior of the eight zircons that are in ²³⁸U–²³⁰Th equilibrium exhibit a greater range from ~0.5 to ~2.6. In general, unpolished surface analyses (“up” in Table 4) yield higher U/Th ratios that vary from ~2.4 to ~4.3 (in one case up to 12.6), indicating increasing U/Th during crystallization. Rimward enrichment of incompatible elements is consistent with magmatic zircon growth (e.g., Claiborne et al. 2006). The Pleistocene and Mesozoic zircons are zoned in backscatter-electron imaging as well as cathodoluminescence images (Fig. 3). The zircons display internal zoning that is truncated by thin rims. The Mesozoic zircons are commonly euhedral to subhedral, whereas the Pleistocene zircons are subhedral to anhedral. There is no observed grain size distinction between the two age populations. Backscattered-electron imaging reveals that individual Coso allanites are not significantly zoned (Fig. 3).

Ti-in-zircon thermometry

Zircon crystallization temperatures using Ti-in-zircon thermometry (Ferry and Watson 2007; Watson and Harrison 2005) were calculated from Ti abundances measured by ion microprobe (Tables 4, 5). Based on the presence of quartz in essentially all Coso rhyolites (Manley and Bacon 2000), granophyre 46A-19RD, as well as plutonic and metamorphic country rocks, model Ti-in-zircon temperatures assume silica activity (a _SiO2) = 1. Titania activities (a _TiO2) were estimated using: (1) the compositions of coexisting Fe–Ti oxides (cf. Wark et al. 2007) reported by Manley and Bacon (2000), and (2) the rutile saturation model of Hayden et al. (2006) using whole-rock composition (Bacon et al. 1981) and the Watson and Harrison (1983) zircon saturation temperature. For Mesozoic xenocrysts a _TiO2 was assumed to be unity. Pleistocene zircons from granophyre 46A-19RD and Cactus Peak and Sugarloaf Mt. rhyolites yield crystallization temperatures between 610 and 740°C, with averages of 690°C (for a _TiO2 = 0.41) and 660°C (for a _TiO2 = 0.62). In contrast, xenocrystic zircon Ti abundances are higher by at least a factor of 2, and their model crystallization temperatures range from 740 to 800°C (for a _TiO2 = 1). These temperatures will be higher if a _TiO2 is <1, as is typical for granitic magmas (Watson and Harrison 2005).

Discussion

The accuracy and concordance of eruption and pre-eruption ages of Coso rhyolites

Alkali feldspar phenocrysts are ideal for dating eruption of Quaternary rhyolites by the ⁴⁰Ar/³⁹Ar method. The young ages of the Coso rhyolites make their ⁴⁰Ar/³⁹Ar analysis insensitive to the interpreted age bias due to ⁴⁰K decay constant errors (i.e., Min et al. 2000). The exception is the 623 ± 6 ka Devils Kitchen rhyolite that after correction is ~10 ka older than measured (i.e., 613 ka). Faced with the possibility of insufficient ⁴⁰Ar* in single crystals we chose to perform replicate multi-grain total fusion and step-heating analyses. In some cases excess ⁴⁰Ar* due to trapped melt inclusions or partially degassed xenocrysts (cf. Ton-That et al. 2001) may complicate interpretation of the results, as appears to be the case for feldspar from sample VCD03, one of three samples dated from dome 5. As explained above, to mitigate this problem step-heating data were deleted stepwise from below the inverse isochron until a reasonably high probability of fit (0.1) was obtained. The resultant isochron age of 215 ± 6 ka is indistinguishable from the ~214 to 220 ka isochron ages determined for the other two dome 5 samples.

For those Coso rhyolites where obsidian is present, non-hydrated glass may be dated by ⁴⁰Ar/³⁹Ar, as was previously done using K–Ar methods (Duffield et al. 1980; Lanphere et al. 1975). Excess Ar due to open system evolution (i.e., consistent with the existence of old zircon and allanite ages) may explain the difference between the glass date of ~255 ka for dome 5 (VCD-03; Table 2) and the ~215 ka feldspar ages. Notably, the inverse isochron fit to the glass data reveals excess scatter as indicated by an MSWD of 5.1. The problematic glass data likely reflect excess ⁴⁰Ar* and/or atmospheric contamination possibly coupled with kinetic fractionation of Ar isotopes during disequilibrium exchange between atmospheric and magmatic reservoirs. Alternatively, it may have experienced cryptic secondary mobility of K and Ar due to hydration, although this is considered to be less likely (Morgan et al. 2007). In contrast, step-heating analyses of glass from Sugarloaf Mountain dome 26 (VCD05) define a well-behaved inverse isochron that yields an age of 86 ± 2 ka with a MSWD of 0.9. This age is indistinguishable from Sugarloaf Mountain dome allanite model age of ~86 ka and the Cactus Peak allanite model age of ~85 ka. The glass ⁴⁰Ar/³⁹Ar age is also consistent with the ~84 ka internal U/Th isochron age calculated from zircon and allanite (Fig. 5).

There is excellent agreement between the ⁴⁰Ar/³⁹Ar and the U–Th–Pb accessory mineral dating for individual rhyolites performed in this study (Fig. 4a, b). In contrast, the 623 ± 6 ka ⁴⁰Ar/³⁹Ar eruption age from the two Devils Kitchen samples appears older than the youngest zircon (“rim”) dates reported for the Devils Kitchen rhyolite by Miller and Wooden (2004) that appear to exhibit a bimodal age distribution with the younger maximum at ~550 ka. A similar discrepancy exists between the 587 ± 18 ka K–Ar date reported by Duffield et al. (1980) and the youngest zircon rim dates that because of this difference was considered unreliable by Miller and Wooden (2004). Regardless, Miller and Wooden (2004) provide compelling evidence from the zircon (“core”) measurements that crystallization initiated at least ~150 ka prior to the eruption. A majority of the reported zircon rim ages (n = 19/21) are within error (1 σ) of and define a “broad hump” that straddles the 623 ± 6 ka ⁴⁰Ar/³⁹Ar date, although it is unclear whether the reported age range for the rim measurements originates from analytical or geological variability. The youngest zircon core ages (Miller and Wooden 2004) are noteworthy because they are in excellent agreement with the new ⁴⁰Ar/³⁹Ar eruption age determination. The apparent bimodal age distribution of the zircon “rim” measurements reported by Miller and Wooden (2004) may be spurious due to the fact that uncertainties related to Th and U abundances in the concentration standard used and the assumed melt composition for the initial U–Th disequilibrium correction were not propagated into the reported errors. The latter point deserves further explanation: it is conceivable that a less evolved component contaminated the melt shortly prior to eruption without permitting significant zircon growth. This scenario is supported by the abundant mafic enclaves in the Devils Kitchen rhyolite (Bacon and Metz 1984). Based on such a scenario and the range of potential mafic magma and hybridized inclusion compositions reported at Coso (e.g., Bacon and Metz 1984), the U/Th ratio of the rhyolite compositions in which the zircon grew could have been different than measured. If addition of a relatively high-U/Th component occurred then this could make the reported zircon ages for the Devils Kitchen rhyolite too young. Notably, the whole rock uranium concentrations measured in this study are approximately half those expected for the younger rhyolites and therefore the actual abundance in Devils Kitchen may also be lower. Because of this possibility; i.e., that the assumed U/Th ratio for the melt from which the zircon crystallized may be too high, the zircon ages could be an extra ~5 ka too young. In addition, this age discrepancy may be due, in part, to a ²³⁰Th excess related to the melt generation process, i.e., derived from a zircon bearing mush or wall-rock based on the uranium-series disequilibrium measured in this study for the younger rhyolites.

The Pleistocene zircon ages (~200 ka) from the subsurface granophyre sample demonstrate that magma was emplaced into the shallow crust at the time, close to extrusion of dome 5 at ~215 ka. Based on the measured zircon age uncertainties it is conceivable that magma that ultimately became the granophyre was related to Group 3 or 4 extrusions, possibly either as a proximal intrusive body or a residual crystal mush. Nevertheless, the narrow zircon age span (≤20 ka) observed in the granophyre indicates that it likely crystallized rapidly and therefore probably represents a short-lived magma body.

Significance of bimodal zircon age populations in Coso rhyolites

The Devils Kitchen rhyolite contains by far the greatest crystal content and the most enclaves of any Coso rhyolite (Bacon et al. 1981). The porphyritic mafic enclaves themselves are contaminated with gabbroic and alkali-granitic crystal mush and show evidence of commingling at mafic and rhyolite magma contacts (Bacon and Metz 1984). These observations, along with the fact that the Devils Kitchen contains Quaternary zircons that significantly pre-date eruption (Miller and Wooden 2004) suggest that this low-temperature rhyolite reflects eruption of a near solidus magma and/or crystals liberated from Quaternary intrusions rejuvenated, perhaps episodically, by mafic influx (Bacon and Metz 1984; Miller and Wooden 2004). The incorporation of mafic enclaves into the Devils Kitchen magma likely occurred shortly before its eruption.

Disequilibrium features are common to all Coso rhyolites. However the salient features of the Devils Kitchen rhyolite contrast strongly with the crystal-poor (<2%) nature of the younger rhyolites. The abundance of Mesozoic zircon observed in the younger rhyolites (this study) and their absence in the Devils Kitchen rhyolite reinforces this distinction. If the younger rhyolites were directly differentiated from any residual Devils Kitchen magma, then sufficient time or thermal energy would be needed to dissolve the inherited zircon (e.g., Watson and Harrison 1983). Recycling might represent adjustment of the magma reservoir margins near the time of eruption. Any widening of the conduits or reservoir may have led to the introduction of zircon xenocrysts from Mesozoic wall-rocks. Alternatively, the younger rhyolites may represent transient and discrete magma bodies that interacted with country rock. In this case, the younger rhyolites reflect discrete batches of fresh highly-evolved rhyolitic magma that not only require extensive fractional crystallization, but also involve melting and/or assimilation of local country rocks and then rapid eruption.

The narrow age range of Quaternary zircon along with the abundant Mesozoic zircons present in the granophyre (46A-19D) resemble the bimodal zircon age populations observed in the Cactus Peak and Sugarloaf Mt. rhyolites, and imply that: (1) the intrusion formed independently from the Devils Kitchen rhyolite, or (2) that by ~215 ka all of the Devils Kitchen-related residual crystal mush was processed (i.e., excluded from the younger magmas because it was erupted, dissolved, or crystallized). It is possible that the Mesozoic zircons in the granophyre represent contamination during the drilling through overlying wall-rock. Hence, the Mesozoic zircons cannot be unambiguously interpreted as magmatic xenocrysts.

The Coso rhyolites are unusual among silicic extrusions in that a majority of their zircons are clearly xenocrysts derived from the much older wall-rocks of the magmatic system. In other rhyolites (e.g., Bachmann et al. 2007), widely distributed ages for zircons make distinctions between autocrysts, xenocrysts, and antecrysts ambiguous (e.g., Charlier et al. 2005). Classic inheritance (sensu Gulson and Krogh 1973) in several silicic extrusions has been identified by ion microprobe spot analyses of crystal cores, e.g., in crystal-rich units like the Tertiary Amalia Tuff of New Mexico (Johnson et al. 1989), the Fish Canyon Tuff of Colorado (Lanphere and Baadsgaard 2001), and the ~1.2 Ma Ongatiti ignimbrite from the Taupo volcanic zone, New Zealand (Brown and Smith 2004). Similar evidence for the incorporation of basement material also exists for a couple crystal-poor units at Taupo (the ~20 ka unit Ω and 6–7 ka unit G; Charlier et al. 2005). Unit Ω largely contains Pleistocene zircon and exhibits less than ~15% basement aged zircon—much less abundant than the ~50% or more found in the post-230 ka Coso rocks—and as inherited zircon cores rather than as whole crystals. Unrecognized zircon inheritance could lead to spurious evidence for significant Quaternary pre-eruption crystallization. At Coso, inheritance does not lead to spurious intermediate ages, i.e., the Devils Kitchen exhibits a range of Quaternary zircon that significantly pre-date eruption, but no Mesozoic ages have been obtained, whereas the younger Coso rhyolites exhibit little evidence of zircon growth prior to eruption, yet contain abundant Mesozoic xenocrystic zircon.

The differences reported among zircon age populations contained in individual Coso extrusions are distinct (e.g., the range in time by which they pre-date their respective eruptions) and we believe that they have geologic significance. However, the implications inferred here (and elsewhere), given these apparently distinct age “signatures”, make it important to stress that most interpretations are limited by sampling size (~10’s of zircon ages). Where sampling is limited, spot selection and analytical changes may lead to non-representative age spectra and therefore the significance placed on discrete subpopulations, e.g., apparent peaks in zircon age probability density function plots should be interpreted with caution. Nevertheless, the abundance of zircon core ages that significantly pre-date eruption of the Devils Kitchen rhyolite is greater than can be explained by analytical uncertainties alone. Likewise, if the younger Coso rhyolites had had a similar duration of crystallization (≥150 ka), despite their smaller sampling size, statistical arguments would suggest that at least three Quaternary ages ≥100 ka older than eruption would be expected but this is not observed. The oldest magmatic zircon age in the younger rhyolites is not resolvable from eruption (at the ~15 ka level).

Evidence for rhyolite source components and crustal assimilation

Evidence of inherited zircon, as well as radiogenic isotope compositions, suggests that a crustal source component was important in the generation of at least some of the Coso rhyolites (e.g., Bacon, et al. 1984; Glazner et al. 2007; Miller et al. 1996; Monastero et al. 2005). The positive and fairly homogenous ε_Nd values (+1.0 to +2.5 ε-units) of the Pleistocene rhyolites led Miller et al. (1996) to conclude that much of the mass of the rhyolites originates from mantle-derived basalt, either indirectly by partial remelting of underplated mafic intrusions or directly by differentiation of asthenospheric melts. In a written communication J. Miller indicated that the Devils Kitchen is more mantle-like (ε_Nd ≈+4 ε-units) than the more voluminous younger Pleistocene lavas (≈+2 ε-units). Notably, Manley and Bacon (2000) have suggested that the Coso magma system is shallowing and possibly growing. The shift to lower ¹⁴³Nd/¹⁴⁴Nd can be attributed to greater amounts of crustal incorporation during rhyolite evolution and may reflect this growth (Miller 1999), although significant amounts of upper crustal assimilation would require significant preheating because of the greater energy requirements that are needed to heat up relatively cool shallow crust (Dufek and Bergantz 2005).

Why do the highly evolved, i.e., Sr-poor, Coso rhyolites retain relatively primitive isotopic affinities if they have incorporated significant amounts of Mesozoic basement, as implied by the zircon age data reported here? Is it possible that the assimilant was gabbroic and that the isotopic compositions used by, e.g., Glazner et al. (2007) are not entirely representative of the subsurface Mesozoic basement in the zone of rhyolite genesis and/or storage? Mafic intrusions may locally contain abundant zircon (e.g., Paces and Miller 1993) and if partially remelted may generate rhyolites (e.g., Garrison et al. 2007). For example, the Sr and Nd isotopic compositions of lower crustal cumulates and restites associated with Sierrian granitoids erupted in Miocene lavas (Ducea and Saleeby 1998) have isotopic signatures that are similar to those reported by Monastero et al. (2005). Thus, they could be candidates for parental sources that produced the rhyolites by a scenario that involved partial melting, followed by extensive feldspar fractionation in order to generate their extreme depletion in Sr, Ba, and Eu concentrations (Bacon et al. 1981). Alternatively the assimilant may have had a relatively low Sr abundance and therefore have had little leverage on the Sr isotope systematics, e.g., it could have been metasediments where the zircons were recycled previously from Mesozoic basement.

Given the more primitive Nd isotopic composition of the Devils Kitchen rhyolite as compared to the isotopic composition of post-230 ka rhyolites, the latter may reflect generation of discrete melts from (1) partial melting of a hybrid of early Pleistocene mafic rocks that were intruded into the Mesozoic granitoids (i.e., a source that was a physical mixture of Sr-rich mafic rocks with a juvenile isotopic composition and Sr-poor zircon bearing granites) or (2) an evolving magma reservoir that includes rapid piecemeal differentiation of mantle-derived mafic magmas that intruded and partially entrained relatively deep granitic country rock just prior to eruption. Both would have been facilitated by underplating of mafic magma, which is consistent with the mafic xenoliths and enclaves found in some Coso rhyolites (Bacon and Metz 1984). The latter might better explain the unexpected “reversed” crystal morphology of the zircon (Mesozoic grains are euhedral whereas the Pleistocene grains tend to be subhedral, see Fig. 3) and the inverted temperature estimates of the ~85 ka rhyolites (Fig. 8, see below), and does not necessitate differential retention of early inherited zircon. In general, either scenario can explain the evolved compositions of the Pleistocene Coso lavas, their more primitive mantle-derived and relatively homogenous isotopic compositions, and evidence for both mafic xenocrysts (hornblende ± pyroxene ± olivine) and old zircon.

Fig. 8

Upper panel shows episodic changes in the La/Nd incompatible element ratio of Coso rhyolites through time. Inset shows trends of La/Nd in cogenetic suites of silicic magmas related by fractional crystallization of typical mineral assemblages that, at higher degrees of differentiation, include allanite as the dominant light rare-earth element rich accessory phase. The change from increasing to decreasing La/Nd at ~68–72% SiO₂ reflects the greater compatibility of La relative to Nd after allanite saturation, with decreasing La/Nd caused by allanite fractionation. AT low-intermediate SiO₂ Ammonia Tanks Tuff (Mills et al. 1997), CP Cordillera Paine granite (Michael 1984), LP La Pacana caldera (Lindsay et al. 2001), CG Cape Province Granites (Scheepers and Smit 1995), KG Kleivan granite series (Petersen 1980), and Toba lavas (Wark et al. 2001). Zigzag pattern defined by the Coso rhyolites suggests more complicated open system evolution (see text). Lower panel shows Fe–Ti oxide eruption temperatures, calculated zircon saturation temperatures, and Ti-in-zircon temperatures for Coso rhyolites. Symbols defined in figure. Gray vertical band links the La/Nd ratios to the temperature estimates for the ~85 ka domes

Temperature estimates and zircon stability

The stability of zircon in Coso rhyolites is evaluated by comparing calculated zircon saturation temperatures (Watson and Harrison 1983) to pre-eruption melt temperatures. Zircon saturation temperatures of ~735 to 770°C were determined from the whole rock compositions reported in Bacon et al. (1981). When evaluated chronologically, rather than displaying a secular cooling trend, results for the younger (≤150 ka) Coso rhyolites exhibit a +25°C zircon saturation temperature maximum relative to the older rhyolites (e.g., Devils Kitchen, Fig. 8). By ~90 ka, zircon saturation temperatures had returned to/or dropped slightly below pre-230 ka temperatures.

Pre-eruption temperatures of ~735 to 760°C for individual rhyolites determined by Fe–Ti oxide thermometry (Manley and Bacon 2000) mimic the trend defined by zircon saturation temperatures. The average temperatures and standard deviations of temperatures computed from multiple oxide pairs for individual rhyolites are plotted in Fig. 8. In most cases the oxide temperatures are equal to or apparently lower than the zircon saturation temperatures. The exceptions are the oxide temperatures of the ~85 ka Cactus Peak and Sugarloaf Mt. rhyolites, which are ~20 to 30°C higher than their calculated zircon saturation temperatures (see Fig. 8). Although this temperature inversion is at the upper limit of the uncertainty of the oxide thermometer (±20°C) the systematic increase in both the oxide and zircon saturation temperatures prior to this point lend support to the interpretation that the Coso magmas increased in temperature. Alternatively, the varying temperatures could just indicate discrete melt batches with distinct histories.

Given the calculated zircon saturation and Fe–Ti oxides temperatures, the low Ti-in-zircon temperatures (~700°C) of the Pleistocene zircon contained in the granophyre and Cactus Peak and Sugarloaf Mt. rhyolites (Fig. 8) are surprising, but consistent with heating after zircon crystallization. In comparison, the average temperature for the Mesozoic zircon is at least 50°C higher than the Pleistocene Ti-in-zircon temperatures. The origin of Mesozoic zircons in Cactus Peak and Sugarloaf Mt. rhyolites is likely linked to their inverted eruption and zircon saturation temperatures (i.e., melt temperatures were probably above the solidus of the surrounding country rocks). As shown by Watson and Harrison (2005) Ti-in-zircon measurements provide maximum zircon crystallization temperature constraints, in part because the a _TiO2 (=0.42, see earlier Section “Ti-in-zircon thermometry”) used to calculate the temperatures are minimum values (i.e., higher a _TiO2 leads to lower crystallization zircon temperatures). In addition, contamination from cryptic Ti-rich mineral inclusions (e.g., ilmenite) would result in spuriously high Ti-in-zircon temperatures. Accordingly, the Pleistocene zircon crystals likely grew from rhyolite magmas stored as discrete, near-solidus crystal-mushes and/or intrusive bodies that underwent relatively rapid heating just prior to eruption possibly in response to a fresh batch of basaltic magma and/or that the zircon were remobilized just prior to eruption from local, cooler “zero-aged” crystal-mush and/or intrusive bodies.

Coso rhyolite compositional evolution

Like other high-silica rhyolites, those at Coso exhibit little variability among major element abundances (Bacon et al. 1981). Based on their trace element concentrations, e.g., low Sr and Ba and high Rb and Cs (Bacon et al. 1981), differentiation via fractional crystallization was important in their genesis. However, relative changes in trace elements are difficult to explain by simple, single-reservoir, closed system fractionation. For example, Rb/Sr ratios fluctuating (from ~20 to ~260) between successive Coso rhyolite groups imply tapping of discrete magma(s). The ratio La/Nd has been demonstrated to be useful for tracking rhyolite differentiation because of fractionation and differential partitioning related to, light rare earth element-rich accessory minerals (e.g., phases such as allanite, chevkinite, and/ or monazite; Miller and Mittlefehldt (1982) and Vazquez and Reid (2004), see inset in Fig. 8). In allanite-saturated rhyolites, compositional shifts toward higher La/Nd likely equates to input of more mafic magma, although addition of evolved magma where allanite saturation has not occurred could also drive up La/Nd values. Allanite occurs in most Pleistocene rhyolites at Coso (Manley and Bacon 2000), and La/Nd is variable and oscillates through time (Fig. 8). The fluctuating behavior of La/Nd at Coso contrasts with the relatively monotonic increase in La/Nd delineated by rhyolites from the Long Valley magma system that presumably reflects the general shift toward greater input of more primitive magma through time in that system (Simon et al. 2007; see Fig. 9). The fluctuating La/Nd ratios of Coso rhyolites likely suggest that transient and discrete rhyolites characterized the Coso system over time. Magma evolution of a single magma body without rejuvenation would be possible if addition of high La/Nd contaminants to the erupted magma at two alternative times occurred. Contamination by highly evolved basement materials (i.e., low La/Nd ratios) may also have played a role (e.g., there is a significant amount of LREE-depleted peraluminous Mesozoic granite exposed locally, see unpublished basement map of R. Whitmarsh, complied in 1997). Alternatively, a single body of rhyolite could have been periodically recharged and thermally rejuvenated by less evolved melts. If volcanism at Coso was the result of a single evolving magma reservoir, then it is likely that at least two pulses of more primitive rejuvenation occurred (first at ~625 ka and then at ~150 ka). The younger rejuvenation pulse at ~150 ka is the most dramatic because it coincides with a pronounced increase in both zircon saturation and Fe–Ti oxide eruption temperatures. The chemical “recovery” of the differentiating reservoir occurred rapidly, within approximately 40 ka (Fig. 8), but thermal relaxation lagged behind as indicated by the observation that pre-eruption oxide temperatures remained elevated. The presence of abundant euhedral zircon xenocrysts in the ~85 ka lavas, that contain partially resorbed Pleistocene zircon and allanite grains, is conspicuous because zircon dissolution in hydrous rhyolite at ~760°C would occur rapidly, i.e., within only hundreds of years depending on number of factors such as degree of undersaturation and size (Watson and Harrison 1983). This suggests that the rhyolite body continued to grow by assimilation of country-rock after it experienced recharge and thermal and compositional rejuvenation. It follows, given evidence for zircon inheritance in these rhyolites, that the decreased La/Nd values in the ~85 ka rhyolites might also be the result of crustal assimilation, possibly related to the observed thermal pulse rather than due to very rapid crystallization differentiation. For the ~625 ka positive La/Nd excursion characteristic of Devils Kitchen rhyolite, a concomitant temperature spike is absent. Given clear evidence for mafic injection into the Devils Kitchen magma, this could imply that it erupted from a separate magma body that underwent a distinct compositional (i.e., contaminant) evolution history. This is further evidence that volcanism at Coso tapped at least two (and possibly multiple) distinct magma bodies. Nevertheless, eruption of the Devils Kitchen magma must have occurred soon enough after injection of the mafic material such that the intruding magma could not have had time to equilibrate chemically with the resident rhyolite. In summary, Coso rhyolites reflect discrete periods of magma genesis dominated by open system behavior whose chemical signatures reflect variable combination of fractional crystallization differentiation, mafic injection and mixing, and at least in some younger rhyolites crustal contamination.

Fig. 9

Upper panel shows characteristic increase in rhyolite production (steeper cumulative volume curves) through time at Coso and Long Valley fields. Volume estimates based on Bacon (1982) and Bailey (1989); Metz and Bailey (1993), respectively. Lower panel compares La/Nd melt evolution at Coso to Long Valley. Trace element compositions come from Bacon et al. (1981) for Coso rhyolites and Metz and Mahood (1991); Schmitt and Simon (2004) for glasses and Anderson et al. (2000); Peppard et al. (2001); Schmitt and Simon (2004); Wallace et al. (1999) for quartz hosted melt inclusions for Long Valley rhyolites. Solid arrows (Coso) and dashed arrow (Long Valley) used to clarify temporal trends and should not be mistaken for an implied genetic linkage between sequential eruptions (cf. Simon et al. 2007)

Continuity of the rhyolitic magma reservoir in space and through time

On average it appears that the erupted volumes of rhyolite have increased (Bacon et al. 1981; see Fig. 9), the storage depth of rhyolite has decreased, and rhyolite eruption temperatures have risen (Manley and Bacon 2000; see Fig. 8). Geophysical methods that have claimed to have successfully detect extensive bodies of shallow magma elsewhere (e.g., Eaton et al. 1975) have not demonstrated the existence of shallow magma bodies at Coso, but frequent seismicity (Walter and Weaver 1980), as well as teleseismic P-delay tomography studies (Reasenberg et al. 1980; Wilson et al. 2003), show evidence for a low-velocity body in the middle crust (top of the body at ~5 km below the surface). Given that the volume of intrusive silicic magma at Coso may be as much as 100 times greater than the erupted volume (Bacon et al. 1981) the magma reservoir(s) may have been larger at certain times in the past. Similarities between the extruded domes such as eruption ages, zircon and allanite crystallization ages, and initial Th isotope ratios suggest that at times the erupted part of the Coso rhyolitic magma reservoir was well mixed and integrated (e.g., ~85 ka). These isotopic data lend support to the petrogenetic significance of the geochemical groups of Bacon et al. (1981). The distance between Cactus Peak and Sugarloaf Mountain domes (Fig. 1), implies that at ~85 ka petrologically and geochemically similar magma breached the surface at vents spaced ~5 km apart. This may indicate that the subvolcanic reservoir was of significant lateral extent at the time of eruption. However, given the general alignment of domes extruded at that time, the aspect ratio of this intrusion might be extreme, i.e., a dike that propagated from a reservoir focused at the center of the rhyolite field (Bacon et al. 1980). Regardless, demonstrating the possibility that a substantive reservoir existed at certain times in the past does not necessarily indicate that there was a single magma reservoir evolving through time. The Coso magma system is likely to have waxed and waned over time, as suggested for Long Valley by Reid and Coath (2000), Simon and Reid (2005), and Simon et al. (2007).

The significance of accessory mineral crystallization ages for estimates of magma residence time scales and rhyolite production rates

The age constraints of Coso rhyolites suggest that the record of pre-eruption crystallization contained in individual extrusions can be brief (e.g., post-230 ka rhyolite, this study) or protracted (e.g., Devils Kitchen rhyolite, Miller and Wooden 2004). The record determined here for the post-230 ka rhyolites in which there are only near eruption crystal ages is uncommon among silicic extrusions (cf. TIMS ages reported by Crowley et al. 2007), whereas protracted ages like that of the Devils Kitchen are more common (i.e., among ion probe studies summarized in Simon et al. 2008). Although the volcanic history of Coso is distinct from many long-lived silicic volcanic centers, the diversity of pre-eruption zircon and allanite age spans appears analogous to that reported for individual rhyolites at Long Valley (Heumann et al. 2002; Reid and Coath 2000; Reid et al. 1997; Simon and Reid 2005; Simon et al. 2007).

At issue is how to interpret the evidence for whether silicic magma reservoirs remain thermally viable for protracted amounts of time in the upper crust. The approach taken here that directly measures the difference between crystallization age and eruption age, which is usually interpreted as a minimum storage time for magma in its reservoir (e.g., Davies et al. 1994; Reid et al. 1997; van den Bogaard and Schirnick 1995; Vazquez and Reid 2004), assumes that this duration of crystallization places some measurable estimate on the differentiation time and existence of melt in the shallow crust. It remains unclear as how to best translate this evidence into rhyolite production rates (Charlier et al. 2005; Christensen and DePaolo 1993; Davies et al. 1994; Halliday et al. 1989; Jicha et al. 2005; Knesel et al. 1999; Michaut and Jaupart 2006; Reid and Coath 2000; Simon and Reid 2005; Wolff and Ramos 2003). The existence of a long-recognized relationship between the repose interval between explosive eruptions and their volumes (Smith 1979; Trial and Spera 1990) has led to the long-held view that the duration of silicic magma accumulation is proportional to erupted volume (see inset in Fig. 10). It is probable that the general correlation between repose interval and volume likely represents a Gaussian relationship akin to that observed between earthquake magnitude and frequency (Gutenberg and Richter 1944). This relationship may be useful for understanding magma production rates at the global scale when averaged over significant time periods.

Fig. 10

Upper panel shows the pre-eruption time interval for 24 rhyolites based on comparisons between radioisotopic age constraints (e.g., ⁴⁰Ar/³⁹Ar eruption and U–Th–Pb crystallization ages) and magma volume estimated from eruption volume. Data represent a range of volcanic settings from all over the world and come from sources in Simon et al. (2008), as well as from Brown and Smith (2004). Explosive rhyolites (triangles) exhibit relatively restricted pre-eruption time intervals (≤120 ka), whereas smaller effusive rhyolites (diamonds) trend toward greater time intervals (dashed line). Inset shows inferred pre-eruption magma residence times and volumes of volcanic rocks. Data sources are in Carrasco-Núñez and Riggs (2008), de Silva and Gosnold (2007), Reid (2003), and Spera and Crisp (1981). Yellow (light-shaded) fields for mafic magmas defined in Reid (2003). Red (dark-shaded) boxes for evolved magmas show range of uncertainty in volume and repose interval (~inverse eruption frequency). Lower panel shows pre-eruption crystallization intervals for select Pleistocene Coso and pre-caldera Long Valley rhyolites, DK Devils Kitchen (Miller and Wooden 2004), CP Cactus Peak (this study), and OGM older Glass Mountain, YGM younger Glass Mountain, BT Bishop Tuff (Reid and Coath 2000; Simon and Reid 2005; Simon, et al. 2007). The general inverse relationship between size and pre-eruption crystallization interval displayed among different silicic fields (i.e., upper panel) exists among individual eruptions at both systems. Dashed arrows show that Pleistocene Coso and pre-caldera Long Valley magma systems produced larger eruptions with shorter pre-eruption intervals over time

The radioisotopic results from Coso and other rhyolites (see Fig. 10) suggest that at least some voluminous rhyolites can be produced relatively rapidly, and many small-volume rhyolites likely reflect magma storage, differentiation, and fractionation in long-lived mushy magma bodies. In addition, Pleistocene Coso and pre-caldera Long Valley rhyolites show similar temporal trends from: (1) small-volume eruptions that exhibit protracted pre-eruption crystallization intervals to (2) progressively larger-volume rhyolites with dateable crystal evidence preserved only for smaller intervals between early crystallization and eruption. Increased mantle-derived mafic magma production and influx into the silicic magma systems can explain the general characteristics of both systems (e.g., Simon et al. 2007).

Conclusions

Geochemical, geochronological, and isotopic affinities among spatially distinct extrusions at Coso suggest a magmatic evolution characterized by waxing and waning of the system. Distinct zircon and allanite age populations, trace element compositions, and isotopic compositions between sequential eruptions indicate rapid magma differentiation and transient storage of shallow magma bodies over the ~1.1 Ma life time of the system. The unusually high proportion of zircon xenocrysts in the Cactus Peak and Sugarloaf Mountain rhyolites indicate inheritance and assimilation in relatively low temperature magma (<800°C). Based on their considerable xenocryst populations, fluctuating incompatible trace element ratios, and Coso’s prograde thermal history, it can be seen that some Coso rhyolites were periodically subjected to rejuvenation by intrusion of less evolved magma and characterized by crustal assimilation. More specifically, the succession of Coso rhyolite extrusions either represents: (1) a single long-lived reservoir that incorporated and/ or was rejuvenated by a less evolved component by ~625 ka, then differentiated (by fractional crystallization or contamination by a more evolved component) by ~215 ka, was again rejuvenated by a less evolved component by ~150 ka, and finally incorporated Mesozoic crust as the melts got hotter, consistent with the xenocryst ages showing up in the ~85 ka rhyolites, or (2) the existence of a number of distinct transient magma bodies with relatively distinct proportions of sources that formed and evolved more or less independently.