Abstract

Volcanic rock and mantle xenolith compositions in the Sierra Nevada (western United States) contradict a commonly held view that continental crust directly overlies asthenosphere beneath the Sierran range front, and that ancient continental mantle lithosphere (CML) was entirely removed in the Pliocene. Instead, space-time trends show that the Walker Lane is the principle region of mantle upwelling and lithosphere removal in eastern California, that lithosphere loss follows the migration of the Mendocino Triple Junction (MTJ), and that the processes of lithosphere removal are not yet complete beneath the Sierra Nevada and its range front. Key evidence is provided by volcanic rock compositions. The 87Sr/86Sr ratios for mafic volcanics of the Sierra (MgO > 8%) are mostly > 0.705, and 143Nd/144Nd < 0.5127, much unlike eastern sub-Pacific asthenosphere (where 87Sr/86Sr < 0.7027 and 143Nd/144Nd > 0.5129), but very much like CML. Similarly, Sierran volcanics carry CML-like trace element ratios, with La/Nb > 3 and Th/Nb > 1, values that are significantly higher than asthenosphere-derived melts (La/Nb < 1.5, Th/Nb < 0.08). Spinel-bearing mantle xenoliths contained in Pleistocene–Holocene volcanics from the Sierran range front also have a CML composition, with 87Sr/86Sr and ɛNd ratios that range to 0.7065 and –3.6, respectively. New estimates of melt extraction depths using Si activity and mineralogy-sensitive trace element ratios (Sm/Yb, Lu/Hf) show that CML extends from the base of the crust (40 km) to depths of 75 km beneath the range front, and to 110 km for Pliocene volcanics of the southern Sierra. This means that garnet-bearing lithologies could not have been dislodged from beneath the southern Sierra until after the Pliocene. Only in the Walker Lane do young (0.18 Ma) volcanic rocks, from the Coso volcanic field, approach asthenosphere-like compositions, which occurs only 20 Ma after MTJ arrival. Temporal trends show that MTJ arrival at any given latitude south of 37°N signals lithosphere heating, probably due to asthenosphere that upwells to replace the sinking Farallon plate. Partial melts of the asthenosphere, and perhaps the asthenosphere itself, intrude into and cause heating and partial melting of overlying CML; this culminates after 10 Ma. After 20 Ma, CML becomes highly degraded and asthenosphere-derived melts are dominant. North of 37°N, volcanic rocks approach asthenosphere-derived compositions to the west, not the east, and 87Sr/86Sr ratios increase from 18 to 0 Ma, indicating that this region has entered a phase of lithosphere heating, but not yet a phase of lithosphere removal.

We propose a new model of lithosphere degradation, where asthenosphere or its partial melts pervasively invade CML beneath the Walker Lane. This process is now nearly complete beneath Coso, and is migrating west, so that it is only partly complete at the southern Sierra range front, or within the Sierra Nevada, at any latitude. This model of intermixed asthenosphere and lithosphere better explains the compositions of volcanic rocks and their included xenoliths, and the remarkably consistent S-wave receiver function data, which show a 70-km-thick lithosphere beneath the Sierra Nevada. If the upper mantle is warm CML, permeated by partial melts, this model may also explain low P- and S-wave velocities.

INTRODUCTION

There is a fundamental conflict in our understanding of basaltic volcanism in and adjacent to the Sierra Nevada and in some seismology-based interpretations of underlying mantle. Geochemical studies clearly show that Miocene to Holocene volcanism in the Sierra Nevada (western United States), at least from the Lake Tahoe region southward, requires partial melting of an enriched continental mantle lithosphere (CML) source (e.g., Leeman, 1974; Beard and Glazner, 1995; Cousens, 1996; Lee, 2005; Putirka and Busby, 2007; Cousens et al., 2008; Blondes et al., 2008). This is true even at the southeastern margin of the range, where it was argued (Cousens et al., 2008) that enriched mantle must be beneath the very young Big Pine, Coso, and Long Valley volcanic fields. Geochemical studies of CML thinning place maximum thinning to the east of the Sierra Nevada range front (e.g., Ormerod et al., 1988; DePaolo and Daley, 2000). However, some recent seismic data yield low P- and S-wave velocities in the uppermost mantle beneath the Sierra Nevada, data sometimes interpreted to indicate that continental crust directly overlies asthenosphere in the southern Sierra (e.g., Fliedner and Ruppert, 1996; Wernicke et al., 1996; Savage et al., 2003; Jones et al., 2004; Zandt et al., 2004), and perhaps as far north as Lake Tahoe (Frassetto et al., 2011). These seismic studies are in conflict with S-wave receiver function data indicating that the lithosphere is 70 km thick throughout the range (Li et al., 2007). These contrasting views provide an unclear view of lithosphere degradation and volcanism in the wake of Miocene–Holocene plate reorganization (e.g., Atwater and Stock, 1998). We examine the volcanic record of the Sierran mantle, and propose a new model of lithosphere degradation to rectify geochemical and seismic observations.

DATA

To understand the interplay of volcanism and tectonic forces, volcanic space-time-composition patterns are key. Here we present new (Supplemental Table 11, yellow cells) and published data (Table 1; Supplemental Table 22) that allow us to construct a longitudinal profile, and two latitudinal profiles across the Sierra Nevada mountain range. The new data include major oxide (n = 78), trace element (n = 26), and isotopic (87Sr/86Sr and 143Nd/144Nd; n = 109) compositions from the central Sierra Nevada and from Dixie Mountain (Figs. 1B and 2). Of special interest is whether and how Sierran space-time trends are affected as one moves north into the southern Cascades and east into the Walker Lane. Our analysis thus includes published data from volcanic fields in California (or Nevada, when such fields cross a state boundary) and so extend north to the Shasta–Medicine Lake area, and east into the Walker Lane belt (Table 1; Fig. 1B). The spatial profiles obtained from the new and published data point to profound geographical and temporal controls on Sierran volcanic compositions, and illustrate the modern extent of lithosphere degradation.

Data Filters

As noted by others (e.g., Leeman, 1970; Lum et al., 1989; Fitton et al., 1991; Farmer et al., 2002; Blondes et al., 2008), mafic Cordilleran volcanic compositions reflect mantle sources and processes, rather than crustal contamination. Except for the southern Sierra, the selection of rocks with MgO > 6% (or SiO2 < 55%) provides an effective filter against crustal assimilation for most isotopes and trace elements. In the southern Sierra, we only use rocks with MgO > 10% (or SiO2 < 50%), to guard against crustal contamination.

METHODS

Isotope ratios and trace element concentrations were determined at Carleton University and the Ontario Geological Survey inductively coupled plasma–mass spectrometry laboratory, Sudbury, respectively (see Cousens, 1996). New and published 143Nd/144Nd ratios are corrected for instrumental fractionation using 146Nd/144Nd = 0.72190. Major oxides were determined by X-ray fluorescence at the California State University at Fresno (see Busby et al., 2008b). Age dates, sample locations, and field relationships were described in Busby et al. (2008a, 2008b), Busby and Putirka (2009), Hagan et al. (2009), Koerner et al. (2009), and Roullet (2006).

To build on prior work on partial melting conditions (Van Kooten, 1980; Baker et al., 1994; Wang et al., 2002; Leeman et al., 2005; C.T. Lee, 2011, personal commun.), we use the Si activity barometers of Putirka (2008) and Lee et al. (2009). As a test of the barometers, we compare pressure-temperature (P-T) estimates for more than 440 experimental data saturated with olivine + orthopyroxene ± other phases (hydrous and anhydrous); here we calculated T rather than use experimental T as input, so as to mimic how the models are applied to natural samples (Figs. 3B, 3C). We observe little systematic error for T estimates, although Equation 4 in Putirka et al. (2007) is a more precise thermometer. However, the Lee et al. (2009) models overestimate P (P > 3 GPa); the Putirka (2008) barometer slightly underestimates P (>3.5 GPa), but is much more accurate, in the range of 3–5 GPa. Because volcanic rocks of the southern Sierra Nevada might not be saturated with orthopyroxene (opx) (Elkins-Tanton and Grove, 2003), we also test whether Si activity can be buffered by olivine (ol) + clinopyroxene (cpx) (Carmichael et al., 1970). We thus calculate P for 154 experimental liquids saturated with ol + cpx ± other phases, but without opx, using Equation 42 of Putirka (2008) and experimental T as input. For these experiments, errors are comparable to those for ol + opx–saturated experiments, indicating that the Putirka (2008) barometer can be applied to ol + cpx systems without recalibration (Fig. 3C).

We can only apply the Si activity barometers to rocks that have undergone no crustal assimilation, so we only consider rocks with MgO > 8% generally, and MgO > 10% for Pliocene southern Sierra volcanics (see Results and Discussion). Mafic whole-rock compositions are corrected by olivine addition (Fo90) to achieve equilibrium with mantle olivine of Fo90.54 (the highest Fo reported by Feldstein and Lange, 1999), at KDFe-Mg = 0.32. These calculations assume mean FeO/Fe2O3 = 0.74 on a weight percent basis (from Feldstein and Lange, 1999), and our barometers use 2% H2O in the primitive magma (Feldstein and Lange, 1999; Putirka and Busby, 2007; methods in Putirka et al., 2011). For comparisons of Fe, we use Fe2O3t (total Fe as Fe2O3), because most Sierran rocks are analyzed for Fe2O3t. At Lassen, FeOt is reported, and we recalculate Fe cations to obtain Fe2O3t, (for all practical purposes Fe2O3t = FeOt/0.91); this calculation has no implications for Fe3+/Fe2+, which undoubtedly varies widely in the Sierra (Feldstein and Lange, 1999).

We also estimate melting conditions based on trace element ratios. Garnet-bearing peridotite is only stable at >70–75 km, while spinel peridotite is stable at shallower depths (Longhi, 1995; Kinzler, 1997). Sm/Yb and Lu/Hf are effective for distinguishing between these sources because Yb and Lu are strongly compatible in garnet, but incompatible in spinel and other peridotite phases (e.g., Salters, and Longhi, 1999). For partial melting and fractional crystallization calculations, partition coefficients for major oxides are from Walter (1998) and Gaetani and Grove (1998). For trace elements, partition coefficients are from Salters and Longhi (1999) and Gaetani et al. (2003). Liquid compositions are calculated using fractional crystallization, partial melting (Rayleigh, 1896; Gast, 1968), and assimilation–fractional crystallization (AFC) models (DePaolo, 1981), and energy-constrained AFC (Bohrson and Spera, 2001). Amphibole fractionation models use mineral compositions from Blatter and Carmichael (2001) and trace element partition coefficients from Davidson and Wilson (2011) and Luhr and Carmichael (1980). Phlogopite-liquid partitioning is based on Righter and Carmichael (1996). Source compositions are taken from mantle xenoliths (Beard and Glazner, 1995; Lee et al., 2001; Lee, 2005). Mantle mineralogies are from Workman and Hart (2005) and Walter (1998) (spinel peridotite = 68% olivine, 10% clinopyroxene, 20% orthopyroxene, and 2% spinel; garnet peridotite = 61% olivine, 13% clinopyroxene, 21% orthopyroxene, and 5% garnet; for details see Putirka et al., 2011). Workman and Hart's (2005) model is used for an asthenosphere composition.

BACKGROUND AND PRIOR WORK

Volcanic centers of the Sierra Nevada have been the object of study for many decades, at least since Ransome's (1898) studies of the central Sierra Nevada. We provide a brief review of key aspects relevant to our study.

Tectonic Setting

The Mendocino and Rivera Triple Junctions formed when the Pacific-Farallon spreading ridge impinged North America ca. 30 Ma (Atwater, 1970; Glazner and Supplee, 1982). As the Mendocino Triple Junction (MTJ) migrates north, the Pacific–North American plate margin converts from a convergent to a transform boundary (the north-propagating San Andreas fault; Atwater, 1970), and arc-related volcanic activity ceases. This change yields the arc/post-arc transition documented in the ancestral Cascades (Cousens et al., 2008). To test for MTJ-related volcanic patterns, we calibrate the time at which the MTJ arrives at any given latitude: 
graphic
Equation 1 is derived from the initial position of the MTJ at 30 Ma (29.65°N; Atwater, 1970) and MTJ positions of Atwater and Stock (1998; fig. 11 therein; note that the MTJ stalls in the Sierra at 4 Ma). To obtain paleo-MTJ latitude as a function of age, we have: 
graphic
The regression error is ±1 Ma for Equation 1 and ± 0.2° for Equation 2. We use data from Snow and Wernicke (2000) to reconstruct latitudes of volcanic rocks in the southern Walker Lane, although in practice, most such rocks are so young that the corrections are trivial. Volcanic rocks erupted north of the MTJ are subduction related and noted as pre-MTJ in Table 1; those that erupt south of the MTJ (post-MTJ) do so after subduction has ended. Neogene volcanics of the Tahoe region and the central Sierra Nevada contain rocks that erupted both prior to and after MTJ arrival, and so are noted as transitional in Table 1.

Another tectonic aspect of interest is that the physiographic Sierra Nevada crosses a major tectonic boundary. In the central Sierra Nevada (Fig. 1) the locus of Mesozoic plutonism shifts to the east (Kistler, 1974) (Fig. 1), but the topographic crest of the range continues north as the range transitions into the Cascades (Wakabayashi and Sawyer, 2001). This structural boundary is evident in the initial 87Sr/86Sr [(87Sr/86Sr)i] and 143Nd/144Nd ratios of Mesozoic plutons (Kistler and Peterman, 1973; DePaolo and Farmer, 1984). For example, where (87Sr/86Sr)i > 0.706, south and east of the central Sierra, older, thicker Precambrian continental crust (or sediments derived from such; Saleeby et al., 1987) and CML are presumed to underlie the range; where (87Sr/86Sr)i < 0.706, younger accreted terranes underlie the Sierra. The importance of this boundary is evident in map views of bedrock geology (Fig. 1A). In the presence of a little-faulted Mesozoic batholith, Cenozoic volcanic rock outcrops are rare, but to the north volcanic rocks dominate the landscape. This outcrop pattern is not a product of erosion in the high parts of the range (Putirka and Busby, 2007; Busby and Putirka, 2009), but instead reflects the fact that local density contrasts have a profound impact on magma transport (Lister and Kerr, 1991). Even in Hawaii, where volcanic shields are dominantly basaltic, high-density picrites stagnate at the base of the crust (Garcia et al., 1995). In the southern Sierra Nevada, thick granitic crust clearly inhibits upward transport of all but the lowest density silicate liquids (Putirka and Busby, 2007).

Continental Versus Asthenosphere Mantle Sources, and Lithosphere Degradation

It is now well established that asthenosphere and CML can be distinguished based on isotope and trace element ratios. Leeman (1970, 1974) and Hedge and Noble (1971) were perhaps the first to recognize that elevated 87Sr/86Sr ratios (0.704–0.708, as in the Sierran Province of Leeman, 1974) in Cordilleran basalts require partial melting of CML that has been isolated from the convective mantle for 1–3 Ga. The idea of an enriched mantle source beneath the Sierra Nevada has since been verified by numerous subsequent studies of both volcanic rocks and xenoliths (e.g., Ormerod et al., 1991; Bradshaw et al., 1993; Beard and Glazner, 1995; Rogers et al., 1995; Cousens, 1996; Beard and Johnson, 1997; Blondes et al., 2008). Leeman (1970, 1974) further recognized that within the continental U.S., some basaltic rocks approach oceanic-like 87Sr/86Sr ratios (0.702–0.703); such rocks erupt after episodes of extensional strain, and so indicate replacement of older continental lithosphere with convective asthenosphere (e.g., Walker and Coleman, 1991; Fitton et al., 1991; Daley and DePaolo, 1992; DePaolo and Daley, 2000). Mantle sources can also be characterized by trace element ratios. CML is enriched in large ion lithophile elements (LILE) that are fluid mobile (e.g., Rb, Ba, Sr, Th, U, Pb), and depleted in high field strength elements (HFSE, e.g., Nb and Zr). These trace element signatures may have been imparted to the lithosphere by dehydration of the Laramide-age subducted slab (e.g., Rogers et al., 1995; Humphreys et al., 2003; Lee, 2005), or much earlier (Beard and Glazner, 1995), but in any case, CML has high LILE/HFSE ratios, such as high La/Nb, Th/Nb, and Ba/Zr, when compared to asthenosphere (e.g., Rogers et al., 1995; DePaolo and Daley, 2000; Plank, 2005).

Removal of CML is thought to be related to the development of the MTJ and the San Andreas fault system (Atwater, 1970; Atwater and Stock, 1998). In this model, upwelling asthenosphere may partially melt and erode CML. Ducea and Saleeby (1996, 1998a) first established key evidence in this regard, noting temporal and spatial changes in Sierran mantle xenoliths. In the southern Sierra, xenoliths hosted in 8–10 Ma volcanic rocks (to the west) contain garnet peridotite, while at the range front (to the east) volcanic rocks of 0.8–0.005 Ma host spinel peridotite. Lee et al. (2001) further showed that shallow-seated Miocene xenoliths are heated while deep-seated xenoliths are cooled. These observations indicate some form of mantle degradation in the Late Miocene, and removal of garnet-bearing lithosphere from beneath the range front by the Pleistocene.

Several observations, however, lead us to be skeptical of recent suggestions that lithosphere removal is complete beneath the Sierra Nevada (Frassetto et al., 2011), or that high K2O Pliocene volcanics signal lithosphere removal (Farmer et al., 2002). Mantle xenoliths carried by Pleistocene volcanics at the range front have 87Sr/86Sr as high as 0.7065, and ɛNd as low as −3.7, reflecting a mantle source that is ca. 820 Ma or older (Beard and Glazner, 1995). Van Kooten (1980) (see also Putirka and Busby, 2007) further suggested that southern Sierra Pliocene volcanics are derived by partial melting of garnet peridotite, which requires melting depths >70 km (Kinzler, 1997). Evidently, CML must have been retained beneath the range in the Pliocene. In the central Sierra, high K2O volcanism is synchronous with range front faulting at 11–10 Ma (Putirka and Busby, 2007). High K2O volcanism most likely signals episodes of high tensile stresses in the Walker Lane, which releases low F (where F is melt fraction) (high K2O) magmas from the mantle (Putirka and Busby, 2007; Busby and Putirka, 2009; Takada, 1994).

RESULTS AND DISCUSSION

Given the current uncertainties of lithosphere degradation (as opposed to delamination, degradation encompasses both mechanical and chemical processes), we use volcanic space-time-composition patterns to reevaluate Sierran tectonics.

North-South Profile of Isotopic Compositions; Effects of Crustal Assimilation

New and existing isotope data for Sierran volcanics exhibit first-order spatial features that require explanation. Ratios of 87Sr/86Sr increase (Fig. 4A; unfiltered data) while 143Nd/144Nd ratios decrease (Fig. 4B) from Mount Shasta south to 37°N (e.g., Death Valley), and follow trends that match those for Mesozoic granitoids of the Sierra Nevada batholith. South of 37°N, this trend appears to reverse, but Death Valley and Coso volcanics, which have the lowest 87Sr/86Sr and highest 143Nd/144Nd, are younger than 6 Ma and erupted in the Walker Lane, a region of high extensional strain (e.g., DePaolo and Daley, 2000). Unusual volcanic rocks at Deep Springs (Beard and Glazner, 1998) also define isotopic extremes. However, within the Sierra Nevada there is no reversal, but a flattening of isotopic trends from 37°N to 35.75°N. All volcanic isotopic ratios plot on a single near-linear trend (Fig. 5), indicative of two-component mixing between an asthenosphere-like source exemplified by mafic Cascades lavas, and an enriched source. Sierra granitoids, however, cannot represent the enriched component (Figs. 5 and 6): many volcanic rocks with high 87Sr/86Sr and low 143Nd/144Nd have high MgO (>8 wt%) and low SiO2 (<50%; see following discussion), and Pliocene southern Sierra volcanics with high Sr and moderate 87Sr/86Sr contents cannot be generated by AFC because AFC predicts higher than observed SiO2 and lower than observed MgO (Fig. 6). AFC of felsic crust explains evolved compositions, with SiO2 > 57% for Pliocene lavas, and SiO2 > 62% and MgO < 6% for most other flows (see following discussion). Mafic volcanic rocks, however, must inherit their isotopic character from the mantle (Leeman, 1970; Ormerod et al., 1988; Lum et al., 1989; DePaolo and Daley, 2000; Blondes et al., 2008), as do their Mesozoic plutonic counterparts (Kistler, 1990).

Partial Melting, Assimilation, and Magma Evolution from Major and Trace Elements

Sierran volcanics exhibit key contrasts in major oxides, which can be used to decipher the conditions of mantle melting. Most volcanics have very similar MgO-SiO2, MgO-Fe2O3t, MgO-Al2O3, and MgO-CaO trends (Fig. 7). However, southern Sierra volcanics have distinctly low Al2O3 and CaO at high MgO compared to other Sierran lavas. Southern Sierra volcanics are also the most enriched in TiO2, Na2O, K2O, and P2O5 at high MgO. Also distinct are Walker Lane volcanics at Coso, Death Valley, Long Valley, and Big Pine, which are enriched in TiO2, K2O, and/or P2O5. As with 87Sr/86Sr, AFC processes are incapable of generating elevated TiO2, Na2O, K2O, or P2O5 at high MgO (Fig. 7); either source-region enrichments or low degrees of partial melting (low F) are required. As to the latter, partial melts of Sierran peridotite xenoliths (Mukhopadhyay and Manton, 1994; Beard and Glazner, 1995) explain elevated TiO2, Na2O, K2O, and P2O5 when F = 0.08–0.01 (Fig. 7). In addition, low Al2O3 and CaO contents for southern Sierra magmas are simultaneously explained, if 2%–5% garnet is retained in the mantle residue. These results are in agreement with prior work (Van Kooten, 1980; Dodge and Moore, 1981; Lange and Carmichael, 1996) and consistent with the idea that tensile stresses (in the Walker Lane) favor eruption of low-F magmas (Takada, 1994; Putirka and Busby, 2007).

Rocks with >10% MgO are primitive enough to plot on an olivine control line (FC1 in Fig. 7) (see Powers, 1955; Wright and Fiske, 1971). This is evidenced by near-constant Fe2O3t at high MgO, which indicates that the bulk distribution coefficient for Fe, DFe ≈ 1 (so olivine is the only phase to precipitate). More evolved compositions are best explained as mixtures of liquids produced along curve FC1, followed by subsequent AFC involving precipitation of olivine + clinopyroxene + plagioclase + hornblende + magnetite + apatite (AFC curve in Fig. 7). Hornblende is likely a ubiquitous fractionating phase at SiO2 > 54%, at least in the central Sierra Nevada. Davidson et al. (2007) showed that Dy/Yb versus SiO2 (Fig. 7I) is an effective means to identify hornblende (hbl) fractionation, since forumla, leading to decreases in Dy/Yb with increased SiO2. Southern Sierra and Long Valley lavas do not exhibit internal SiO2-Dy/Yb variations, but most Sierran lavas south of Lassen plot on a hornblende fractionation trend (Fig. 7I) emanating from those compositions. Volcanic rocks from the Cascades (Lassen, Shasta, Medicine Lake) have much lower Dy/Yb at low SiO2. Given that CML is likely absent beneath Lassen, the lithosphere is probably less thick there, and so lower Dy/Yb reflects spinel peridotite partial melting (curve PM3), whereas thicker, colder CML to the south allows for a greater contribution from garnet peridotite–derived partial melts (curve PM1, Fig. 7I) yielding higher Dy/Yb ratios.

Mantle Source Mineralogy and Composition

The latitudinal patterns of Figure 4 clearly indicate at least two mantle sources: one with time-integrated enrichments of Sm/Nd and Rb/Sr to the south and another lacking such enrichments to the north. Are such contrasts accompanied by contrasts in mantle mineralogy (pyroxenite or phlogopite-bearing peridotite) or degree of fluid-based enrichments, such as K-metasomatism (see models by Van Kooten, 1980; Feldstein and Lange, 1999; Farmer et al., 2002)?

As to metasomatism, no special enrichments are evident. Indices of fluid enrichments in the Cordillera (e.g., Rogers et al., 1995; DePaolo and Daley, 2000), like La/Nb and Th/Nb ratios, decrease from north to south (Figs. 8A, 8B), while Ba/Zr ratios are approximately constant (Fig. 8C), even though Ba is highly enriched in many southern Sierran volcanics (Fig. 9A). These relationships do not support the idea that high K2O volcanics in the central and southern Sierra derive from an especially fluid-metasomatized source. Either enrichments along the Sierra are equal (Ba), or some fluid-mobile elements (Th, La) were actually lost from the southern Sierran mantle prior to late Cenozoic volcanism.

As to a phlogopite-bearing mantle, Dodge and Moore (1981) rejected such a source because K/Rb ratios are too high in southern Sierra volcanics. As a further test, we compare K and Ba, which are both highly compatible in phlogopite (Righter and Carmichael, 1996), to Sr and Rb, respectively. A phlogopite-bearing residue should yield basalts with low Ba/Rb and high Sr/K2O, but partial melts of a source with just 0.5% residual phlogopite yield liquids with Ba/Rb ratios that are too low to explain Sierran volcanic rocks with MgO > 6% (Fig. 9A), even when using a source that has the highest Ba contents observed among all Cordilleran mantle xenoliths (Lee, 2005). This might appear to contradict results from Elkins-Tanton and Grove (2003), but those experiments do not yield a multiple saturation point that contains garnet, which we know must be a residual phase for Sierran Pliocene volcanics (Dodge and Moore, 1981; Van Kooten, 1980; Putirka and Busby, 2007; this study). Therefore, the Elkins-Tanton and Grove (2003) multiple saturation point cannot represent a residual mantle source for these rocks; their work, however, sheds light on the conditions at which phlogopite is stable in the mantle. At depths of 100 km, phlogopite only occurs if H2O ≥ 6 wt%, much greater than the 2% H2O needed for eruption (Feldstein and Lange, 1999; Putirka and Busby, 2007) or the 3–5 wt% H2O estimated by Feldstein and Lange (1999), based on phlogopite barometry and saturation (Righter and Carmichael, 1996). A key misunderstanding appears to be that because phlogopite occurs in a volcanic rock, it must therefore be present in the mantle source. Leucite provides a perfect counterexample: it is present in high K2O volcanics both as a groundmass (e.g., Van Kooten, 1980; Feldstein and Lange, 1999) and phenocryst phase (Beard and Glazner, 1998), and yet no proposal for a leucite-saturated mantle has been made. Rather, both leucite and phlogopite form as precipitates from high K2O liquids after their removal from their mantle sources, a result supported by the absence of phlogopite in mantle peridotites (e.g., Beard and Glazner, 1995) that contain as much as 0.09% K2O, i.e., more than enough K2O to explain the highest K2O in Sierran volcanics (F ≈ 2%).

A pyroxenite source can also be rejected for most Sierran volcanics. If pyroxenite sources represent ancient partial melts trapped in the mantle (e.g., Leeman and Harry, 1993), they should have high Rb/Sr and 87Sr/86Sr and yield partial melts with high Sm/Yb. However, in the central Sierra Nevada 87Sr/86Sr does not vary with Sm/Yb. In addition, using forumla = 0.1 (e.g., Gaetani et al., 2003), even the most Sr-rich pyroxenite xenolith from Lee et al. (2006) yields Sr and Sr/K2O ratios that are too low compared to natural lavas (Fig. 9B). Experiments on high K2O liquids (Schmidt et al., 1999) yield forumla > 0.3, making the problem worse. However, while such a source is unnecessary, we cannot exclude a garnet pyroxenite source for high Sm/Yb and high K2O Pliocene volcanics.

In any case, partial melts of Sierran peridotite xenoliths (Mukhopadhyay and Manton, 1994; Beard and Glazner, 1995; Lee, 2005) explain all major and trace elements in Sierran volcanics. Big Pine and Oak Creek peridotite xenoliths, for example, explain high K2O, TiO2, Na2O, and P2O5 if F ≈ 1%–3% (Figs. 7E–7H). Whereas the mean of Sierran peridotite xenoliths (Lee, 2005) has insufficient Ba (curve C, Fig. 9A), xenolith Ki5–8B (Cima) has more than enough Ba (curve A) to explain volcanic compositions. Similarly, while Ki5–8B has insufficient Sr, the mean of peridotite xenolith compositions explains Sr remarkably well (Fig. 9B). These results show clearly that (1) the Sierran mantle is enriched (Lee, 2005), (2) variations in F, rather than special local source enrichments (Moore and Dodge, 1980; Van Kooten, 1980), control volcanic compositions, and (3) heterogeneities in the Sierran basalt source region are encompassed by peridotite xenoliths from the region, so do not require special residual mantle mineralogies.

Depths of Melt Extraction from the Mantle

Because enriched CML exists in most Sierran volcanic source regions, at least south of 39.6°N (Tahoe area) (Fig. 4), thermobarometry can be used to delimit the depths at which CML is located. Van Kooten (1980) used an early version of the Si activity barometer (Nicholls et al., 1971), and derived melt extraction estimates of 3.3 and 4.1 GPa (100–125 km) for two southern Sierra volcanic samples. Those estimates imply that removal of garnet-bearing lithologies beneath the Sierra could not have been complete until after the Pliocene. Farmer et al. (2002) suggested that melt extraction occurs at ∼40 km, based on phlogopite phenocryst barometry of Feldstein and Lange (1999), who computed pressures of 1.2–1.6 GPa, equal to depths of 44–57 km [for the Sierra Nevada we use: depth (km) = 9.69 + 3.03 P + 0.00054(P – 32.736)2 + 0.0001714(P – 32.736)3 – 0.0000037(P – 32.736)4, where P is in kbar]. However, as shown, the mantle source did not contain residual phlogopite (Fig. 9). The 40 km depth estimate from phlogopite phenocrysts instead indicates ponding of magma at the base of the Sierran crust (40 km; Fliedner and Ruppert, 1996). To test this idea, we calculate P-T conditions for clinopyroxene phenocrysts from Feldstein and Lange (1999; using methods in Putirka et al., 2003): one clinopyroxene formed at nearly atmospheric pressure (0.1 kbar), while three others yield a mean of 14.8 ± 0.03 kbar, and another three yield an average of 10.6 ± 0.16 kbar. These latter pressures correspond to depths of 53 and 39 km, respectively, and are equivalent to crust thickness within model error (±10 km), although the 53 km estimate may indicate partial crystallization in the uppermost mantle, which is not uncommon (Putirka, 2008; Putirka and Condit, 2003; Mordick and Glazner, 2006). To better examine mantle melt extraction depths, we use Sm/Yb and Lu/Hf ratios and Si activity models.

Pliocene rocks from the southern Sierra have the highest Sm/Yb and lowest Lu/Hf ratios observed among all mafic volcanics in the Sierra Nevada (Figs. 10A, 10B). Both ratios require partial melting of a garnet-bearing source, which also explains low Al2O3 at high MgO (Fig. 7C), and elevated Dy/Yb at low SiO2 (Fig. 7I). Most other Sierran volcanics have Sm/Yb and Lu/Hf ratios that indicate partial melting in the spinel peridotite field, or mixtures of spinel peridotite and garnet peridotite partial melts. We emphasize that these latter models are not unique: at Big Pine, a contribution from a garnet peridotite source appears required when using a mean Sierran xenolith mantle composition (e.g., Sm = 0.6 ppm, Yb = 0.12 ppm). However, if we reduce Yb in the source to 0.2 ppm Yb, which is well within the observed range for Sierran xenoliths (Lee, 2005), Sm/Yb ratios from spinel peridotite partial melts are >5 and have >2.3% K2O, and so explain both Sm/Yb ratios and Yb and K2O concentrations. Southern Sierra volcanics, however, are a different matter. Even the most infertile xenolith compositions (from Lee, 2005), with the lowest Yb and Lu contents, fail to yield the lowest Lu/Hf and highest Sm/Yb ratios in the southern Sierra. Moreover, such infertile sources would yield magmas with much lower than observed CaO (Fig. 10C), and vastly lower Yb (<0.24 ppm, compared to the observed 0.4–2.2 ppm) and Lu (<0.03 ppm, compared to observed 0.1−0.3 ppm). In any case, the Sierran mantle is probably not so depleted. South of Lassen, Al2O3 contents vary widely but CaO contents are similar (Fig. 10D), indicating that highly infertile sources are absent. We thus conclude that southern Sierra volcanics require a garnet peridotite source, while other Sierran lavas are derived from a spinel peridotite source (<75 km), ± mixing of garnet peridotite–derived melts (>75 km).

Si activity barometers (Putirka, 2008; Lee et al., 2009) provide an independent test of trace element–based results. These barometers yield remarkably similar P-T estimates for most Sierran volcanics (Fig. 11A). Pressure estimates are closest [P of Lee et al. (2009)P of Putirka (2008) = 0.11 GPa on average) when the Putirka (2008) barometer is matched with the thermometer of Equation 4 from Putirka et al. (2007). However, the differences are increased only slightly when using the thermometer of Equation 2 from Putirka et al. (2007) [P of Lee et al. (2009)P of Putirka (2008) = 0.15 GPa on average], and are closer still at <3 GPa. These tests indicate that our P-T estimates are robust at P ≤ 3 GPa, and our estimates of 44–74 km at Big Pine are remarkably close to the 40–75 km estimates of C.T. Lee (2011, personal commun.).

Melt extraction depths fall into two groups. Samples from Lassen yield a mean of 33 ± 9.7 km [for Lassen, depth (km) = 0.0007P3 – 0.0408P2 + 3.8266P + 0.3476, where P is in kbar], while remaining Sierran suites yield 40–75 km depths (53 ± 15 km on average). P estimates positively correlate with Sm/Yb (Fig. 11A), which indicates that Sm/Yb is a valid proxy for mean depth of melt extraction. Some samples from the southern Sierra, however, exhibit a low P compared to their high Sm/Yb (Fig. 11A), but since even high MgO southern Sierra volcanics are affected by crustal assimilation (Fig. 5; Farmer et al., 2002; Ducea and Saleeby, 1998b), calculated P may be too low. For example, K2O/P2O5 is a recognized index of crustal assimilation (e.g., Farmer et al., 2002) and southern Sierra samples plot between a southern Sierran basalt and granitoid (Fig. 11B) along a mixing curve for K2O/P2O5 versus P. The effect of AFC is to increase SiO2, which lowers calculated P by as much as 2–3 GPa in this case (note that MgO is nearly 8%, even at F = 60%). In contrast, assimilation has almost no effect on Sm/Yb (Fig. 11A).

These results, in combination with isotope ratios, require partial melting of CML (Figs. 8 and 9) 40–75 km beneath the modern range front and central Sierra, and 40–110 km beneath the southern Sierra during the Pliocene (Fig. 12). These estimates allow for loss of garnet-bearing mantle from beneath the range front, but indicate that garnet-bearing CML was present during the eruption of Pliocene Sierran magmas.

Implications for Lithosphere Removal

Taken together, these results indicate that the transition from a mantle wedge beneath Lassen to thick, cold CML beneath the southern Sierra provides a primary control on volcanic rock compositions. First, depths of melt extraction from the mantle (Fig. 13B) are greater to the south than to the north. Second, just as isotope ratios shift to CML-like values to the south (Fig. 4), in rocks with >8% MgO, concentrations of incompatible elements and oxides, like K2O (Fig. 13A), increase to the south (Putirka and Busby, 2007), indicating lower degrees of partial melting in that direction. Third, indices of fluid inputs (e.g., La/Nb) do not vary with depth of melt extraction (Fig. 13C). Fourth, isotope ratios fall into two groups: a mostly northern low-P suite, which approaches mid-oceanic ridge basalt mantle–like isotope ratios, and a southern high-P suite, with a CML signature (Figs. 13D, 13E). Mesozoic granitoids show a similar isotope-space pattern (Figs. 3 and 13B; Kistler and Peterman, 1973; DePaolo and Farmer, 1984). In concert, these observations indicate greater partial melting depths and lower F beneath thicker, and presumably cooler, lithosphere to the south, with little depth or spatial control on mantle enrichments. Since melting depths are everywhere as shallow as 40 km, CML heating reaches the base of the crust, which may result from a pervasive intrusion of asthenosphere or asthenosphere-derived partial melts into overlying CML, which would in turn explain xenolith heating and/or cooling systematics observed by Lee et al. (2001).

Space-Time Patterns

Some key questions are, where and when is CML degraded, and what triggers the process? Latitudinal sections across the Sierra Nevada (Figs. 14 and 15) and Equation 1 provide clues. In the westernmost part of the southern Sierra, the (87Sr/86Sr)i 0.706 line plots in the same region as does the 0.706 line of Mesozoic granitoids, as recognized by Kistler and Peterman (1973). This coincidence supports the idea that cratonic North American mantle lithosphere underlies regions where (87Sr/86Sr)i > 0.706 (Kistler, 1990). However, while Mesozoic rocks retain high (87Sr/86Sr)i to the east (Kistler, 1990), 87Sr/86Sr ratios for very young Walker Lane volcanic rocks (e.g., Coso) approach asthenosphere-like values, a result that has been attributed to extension and thinning of the mantle lithosphere, related to evolution of the MTJ (Ormerod et al., 1988). In the southern Sierra, the spatial transition is illustrated by 87Sr/86Sr ratios, which have CML-like values for Pliocene volcanics erupted within the range, but decline to moderate values for Pleistocene–Holocene Big Pine volcanics at the Sierran range front (Fig. 14A), and decline further to approach asthenosphere-like values for young volcanics at Coso (at ∼117.8°W). Mantle xenoliths at Big Pine further show that CML must still exist at the range front, with 87Sr/86Sr ratios that exceed values found for Pliocene volcanics (Fig. 14A), and 143Nd/144Nd in host volcanics as low as 0.512194, lower than southern Sierran Pliocene lavas (which range to 0.51222).

Why do only very young Coso volcanics approach an asthenosphere composition? We use Equation 1 and data in Figure 14A to better test a model first proposed by Ormerod et al. (1988), that the evolving North American–Pacific plate boundary may control volcanic isotope compositions. In Figures 14C and 14D, 0 Ma marks the time of MTJ arrival at a given latitude, while positive values to the right of 0 Ma indicate the lag time between MTJ arrival and eruption; these times are compared to isotopic compositions. Volcanic rocks reveal movement away from asthenosphere and toward a CML source, then back toward asthenosphere. We interpret these trends to reflect partial melting of asthenosphere as it upwells to replace a sinking Farallon plate. Partial melts from the asthenosphere, or perhaps the asthenosphere itself, induce lithosphere heating, and so CML degrades over a 20 Ma time interval. Only Coso volcanics are sufficiently young (Figs. 14C, 14D) and far enough east (Fig. 14A) to be affected by maximum CML degradation, so as to yield a strong asthenosphere component.

North of 37°N, the temporal patterns for Sierran volcanics are different; there, 87Sr/86Sr ratios trend to higher values to the east regardless of age (Fig. 15A). South and east of Lassen, 87Sr/86Sr ratios only increase with decreasing age (Fig. 15B); this most likely indicates a still-increasing phase of CML heating. Many of these rocks erupted prior to MTJ arrival, but are probably still affected by Pacific–North American plate reorganization, as illustrated by McQuarrie and Wernicke (2005).

SUMMARY AND SYNTHESIS

Space-time-composition trends confirm some old views of, and provide some new insights into, Sierran tectonics. First, we show that Sierran Cenozoic volcanic rocks demarcate a (87Sr/86Sr)i = 0.706 line similar to that defined by Mesozoic granitoids, a line that marks the western edge of North American CML (Kistler, 1990; Fig. 14A). Unlike Mesozoic granitoids, however, young Cenozoic volcanics in the southern Sierra transition to asthenosphere-like values to the east, in the Walker Lane (Fig. 14A). The Walker Lane is thus a region of asthenosphere upwelling and CML degradation, and both processes are triggered by MTJ arrival (Figs. 14C, 14D). Heating of the CML occurs over a span of ∼12 Ma, at which point CML degrades and is nearly gone by 20 Ma, post-MTJ arrival at any given latitude (Figs. 14C, 14D). In the southern Sierra, Miocene volcanism signals slab-window opening, while Pliocene volcanism marks the culmination of CML heating; Coso and Big Pine volcanics mark steps toward nearly complete CML removal, as lithosphere degradation encroaches upon the Sierra Nevada (Fig. 14A). North of 37°N, in contrast, asthenosphere upwelling in the Walker Lane is too recent to allow complete CML removal. There, only a CML heating phase is evident (Fig. 15B), which is affected not just by a north-migrating MTJ, but also by a west-propagating ancestral Cascades arc (Colgan et al., 2011; Busby and Putirka, 2009).

Silicate melt barometry reveals the depth extent of Sierran CML, which extends from the base of the crust (40 km) to ∼75 km beneath most of the range, and to ∼110 km beneath the southern Sierra, at least in the Pliocene. CML thus must still be partially intact to depths of 75 km beneath most of the Sierra Nevada, from 35°N to 39°N. These results conflict with conclusions drawn by some seismic studies, which state or imply that Sierran crust directly overlies asthenosphere (Fliedner and Ruppert, 1996; Savage et al., 2003; Jones et al., 2004; Frassetto et al., 2011), but are consistent with those of Li et al. (2007), who indicated a 70-km-thick mantle lithosphere beneath the range front, and Ormerod et al. (1988) and DePaolo and Daley (2000), who illustrated comparable spatial patterns of lithosphere removal.

To reconcile geochemical and seismic observations, we present a new model of lithosphere degradation (Fig. 16A) adapted from Zandt et al. (2004), Li et al. (2007), and Frassetto et al. (2011). We posit that asthenosphere upwells with arrival of the MTJ and opening of a slab window (Fig. 16B) (Atwater, 1970; Atwater and Stock, 1998). This initiates asthenosphere partial melting, and these partial melts, and perhaps the asthenosphere, invade CML (Figs. 14C, 14D), which then degrades (Fig. 16A) (J. Saleeby, 2011, personal commun.). In the central Sierra, the situation is more complex: there, MTJ arrival is preceded by a southwest-migrating arc of volcanism, which culminates in the formation of the ancestral Cascades (Busby and Putirka, 2009; Colgan et al. 2011). The merging of these tectonic spatial trends is observed in the central Sierra, where ancestral Cascade volcanism begins at 16 Ma and is interrupted by a burst of high K2O volcanism at 10 Ma. The high K2O volcanics (latites) emanate from transtensional faults near the modern range front, and signal the birth of the Sierra Nevada microplate there (Busby and Putirka, 2009), but still carry a CML isotopic signature. Such transtensional stresses characterize the Walker Lane (Bursik, 2009; Muffler et al., 2008), where most pulses of high K2O volcanism likely mark episodes of transtensional strain (Putirka and Busby, 2007; Busby and Putirka, 2009), not lithosphere degradation.

This new model is not necessarily in conflict with prior work on Sierran mantle xenoliths or seismic studies. Our model allows for the loss of garnet-bearing lowermost lithosphere (>70 km depths; Fig. 12) beneath the eastern Sierra (Ducea and Saleeby, 1996), and garnet-bearing crust can be lost by lateral mass transfer (Bird, 1991; Zandt et al., 2004; Le Pourhiet et al., 2006). DePaolo and Daley (2000), for example, estimated initial lithosphere thicknesses of 100 km, comparable to our estimates beneath Pliocene Sierran rocks, but 25–30 km less than our estimates for lithosphere thicknesses at the range front. Since garnet is stable only below 70–75 km, our projected loss of CML is sufficient to eliminate garnet-bearing peridotite (Fig. 12). As to observed P- and S-wave velocities, low velocities may reflect the retention of small amounts of melt in a degrading CML, the presence of which can have a profound impact on seismic signals (Hammond and Humphreys, 2000). Our model also appears to be consistent with the Schmandt and Humphreys (2010, fig. 12 therein) imaging of transitional-velocity mantle beneath the range front, and is in excellent agreement with the Li et al. (2007) 70 km estimate for lithosphere thickness and the Ito and Simons (2011) estimate that asthenosphere lies between 80 and 220 km beneath California.

If asthenosphere and mantle lithosphere sources intermix, volcanic rocks should plot on a mixing curve between partial melts generated from both sources. To illustrate, we compare 87Sr/86Sr to Sr and K2O concentrations (Fig. 17). At least three mantle components appear to be required; an asthenosphere component (DMM, depleted mid-oceanic ridge basalt mantle; Workman and Hart, 2005; 87Sr/86Sr = 0.703, Sr = 20 ppm, K2O = 0.006 wt %), and two CML sources: CML1, with 87Sr/86Sr = 0.707 (Sr = 80 ppm, K2O = 0.07 wt%; see Beard and Glazner, 1995; Lee, 2005), and CML2, with 87Sr/86Sr = 0.7055. This range suggests that the widths of the CML fields in Figures 14C and 14D (from DePaolo and Daley, 2000) are real, rather than a product of mixing between a singular CML source and asthenosphere. For completeness, our models assume higher, equal, and lower degrees of partial melting in the CML sources compared to DMM, since CML may be equally fertile as, or more fertile than, DMM; but to explain the heating observed by Lee et al. (2001), we expect that TDMM > TCML. In addition, Coso volcanics are clearly not derived from pure or unmodified asthenosphere (Figs. 14C, 14D, and 17); Coso is only explained if its source contains 0.02 wt% K2O (compared to 0.006% K2O in DMM; Workman and Hart, 2005). Regardless of the precise components or means of mixing, a new view of mantle degradation is needed, and that provided by Figure 16 might be useful for simultaneously explaining seismic and geochemical observations.

We thank Cathy Busby and her students, Jeanette Hagan, Alice Koerner, and Ben Melosh, for their field work, which has been key to our interpretation of the tectonic significance of Cenozoic volcanism in the Sierra Nevada. We also thank the students of EES 101 Igneous and Metamorphic Petrology classes of 2003–2004, for their participation in this project. Putirka greatly appreciated discussions with George Zandt, whose comments helped to inspire the model presented in Figure 16. We also appreciate reviews of early drafts of this manuscript by Jason Saleeby, Cin-Ty Lee, and Lang Farmer, and Jason's allowing us to view an advance copy of a paper in review. The final manuscript was greatly improved by thoughtful reviews by Yildrim Dilek, Lang Farmer, and Cathy Busby. This research was supported by National Science Foundation grants EAR-0711276 and EAR-0711150 (to Putirka and Busby), EAR-0421272, and EAR-0313688 (to Putirka).

1Supplemental Table 1. Excel File of geochemical data for volcanic rocks from the Central Sierra Nevada. If you are viewing the PDF of this paper or reading it offline, please visit http://dx.doi.org/10.1130/GES00728.S1 or the full-text article on www.gsapubs.org to view Supplemental Table 1.
2Supplemental Table 2. Excel File of geochemical data for volcanic rocks from the Sierra Nevada and California portion of the Walker Lane. If you are viewing the PDF of this paper or reading it offline, please visit http://dx.doi.org/10.1130/GES00728.S2 or the full-text article on www.gsapubs.org to view Supplemental Table 2.