The early Eocene (52–44 Ma) was a time of tectonic reorganization and widespread magmatism in Washington, Oregon, and British Columbia (west coast of the United States and Canada) that culminated with establishment of the Cascade arc. Details of this tectonic transition remain enigmatic, and diverse scenarios involving ridge-trench interaction, slab breakoff, and/or plume magmatism have been proposed. This study focuses on the ca. 48 Ma Basalt of Summit Creek, a ∼1500-m-thick sequence of subaerial lavas that erupted during the interval of time between accretion of the Siletzia terrane to the west and inception of the Cascade arc. The sequence dominantly consists of moderately evolved tholeiitic basalts (9.3–3.1 wt% MgO; Mg# = 0.66–0.29) and scarce rhyolites that erupted in the forearc but are chemically and isotopically similar to oceanic basalts of the Crescent Formation, found ∼100 km to the west as part of the Siletzia terrane. Basalt of Summit Creek lavas lack subduction signatures (e.g., high field strength element depletions) and require a mechanism whereby asthenospheric melts were able to reach the surface having undergone little or no chemical interaction with fluids derived from the subducting slab or metasomatized mantle. We suggest that the accretion of Siletzia led to breakoff of the Farallon slab, and that Basalt of Summit Creek lavas formed by decompression melting of mantle that upwelled through the rupture. Slab breakoff typically produces a short-lived linear magmatic belt; the Basalt of Summit Creek appears to be part of such a belt, previously unrecognized, that parallels the Siletzia boundary and formed between 50 and 48 Ma.

Throughout Washington, Oregon, and British Columbia (west coast of the United States and Canada), the early to middle Eocene was a time of unusually widespread magmatism, extension, and tectonic reorganization. Within a span of less than 10 m.y., between ca. 52 and 44 Ma, major tectonic events in this region included accretion of the oceanic Siletzia terrane in western Washington, Oregon, and British Columbia (Wells et al., 2014), voluminous and compositionally diverse magmatism in the Challis-Kamloops belt (Fig. 1), which extends from central Washington northward into Canada and as far inland as Montana (Ewing, 1980; Dostal et al., 1998; Bordet et al., 2014), broad regional uplift (Dumitru et al., 2013), extension manifested by core complex formation (Kruckenberg et al., 2008; Gordon et al., 2009), rapid basin subsidence (Evans and Ristow, 1994; Eddy et al., 2016), dike swarm emplacement (Tabor et al., 1984; Miller et al., 2016), and establishment of the Cascade arc (Vance et al., 1987; Christiansen and Yeats, 1992; Sherrod and Smith, 2000).

Geologic events occurring during this time have been the subject of intensive research within the region. However, our understanding of the broader tectonic environment remains incomplete. It is well established that the west coast of North America has been a convergent margin since at least the Cretaceous (Dickinson and Snyder, 1978; Ewing, 1980), but within this framework, several models have been proposed to account for the Eocene structural and magmatic evolution of the region. Ewing (1980) attributed the width of the Challis-Kamloops belt to changes in the dip of the subducting Farallon slab. Noting the crustal character of Eocene volcanic rocks in eastern Washington and the contemporaneous extension, Morris et al. (2000) ascribed this activity to collapse of thickened crust, with no explicit role for mantle processes. More recently, based largely on plate-motion reconstructions and/or geochemistry, several authors have proposed that Challis-Kamloops belt magmatism developed during the passage of one or more slab windows beneath the region (Thorkelson and Taylor, 1989; Breitsprecher et al., 2003; Haeussler et al., 2003; Madsen et al., 2006). Alternatively, seismic anomalies in the mantle beneath the northwestern United States have been interpreted as detached fragments of the Farallon slab (Sigloch et al., 2008; Schmandt and Humphreys, 2011), suggesting that slab breakoff followed by rollback could have been the causal mechanisms for magmatism and uplift.

To better understand the tectonic processes that occurred during the Eocene, we studied the Basalt of Summit Creek (BSC), a sequence of basalts and minor rhyolite exposed within the boundaries of the younger Cascade arc in Washington state (Fig. 1). These lavas erupted in the critical time window during or slightly after the accretion of Siletzia but before the inception of the Cascade arc. In addition, the location of the BSC, east of the accreted oceanic rocks but west of the Challis belt, affords a rare opportunity to connect events occurring in the continental margin or forearc region with those occurring farther inboard. In this paper, we combine major-element, trace-element, and Sr-Nd-Pb isotopic data with high-precision U-Pb geochronology to constrain the origin and tectonic setting of the BSC. These results provide insights into the history of the Farallon slab during and immediately after the accretion of Siletzia at the latitude of Washington, as well as the magmatic and tectonic manifestations of this event found farther inland.

Regional Geologic Context

The basement of the Cascadia forearc region consists of Paleocene and Eocene basalts that comprise the Siletzia terrane (Fig. 1). This terrane extends ∼500 km from southern Oregon to southern Vancouver Island and consists of basalt with subordinate sedimentary and silicic volcanic rocks (Wells et al., 1984; Massey, 1986; Babcock et al., 1992; McCrory and Wilson, 2013; Wells et al., 2014). A notable feature of Siletzia is its great thickness. Mapping by Babcock et al. (1992) showed that >16 km of basalt flows are exposed in Washington, while seismic data indicate that the total thickness of the terrane thrust beneath Oregon is up to 20–30 km (Trehu et al., 1994). While an extreme stratigraphic thickness for the terrane has been previously suggested (Babcock et al., 1992), recent U-Pb dating has shown that the terrane is structurally thickened in Washington (Eddy et al., 2017). In Washington, rocks belonging to Siletzia are mapped as the Crescent Formation, which is dominantly composed of submarine and subaerial basalts that display a spectrum of mid-ocean-ridge basalt (MORB) and oceanic-island basalt (OIB) geochemical and isotopic affinities (Snavely et al., 1968; Muller, 1980; Babcock et al., 1992; Pyle et al., 2009). Recent geochronologic investigations suggest that Siletzia was emplaced between 56 and 48 Ma (Wells et al., 2014; Eddy et al., 2017). However, in Washington the age range of the basalts is limited to between 53 and 48 Ma (Eddy et al., 2017).

The large magma volumes calculated from the area-thickness data (∼2 × 106 km3; Trehu et al., 1994; Wells et al., 2014), relatively short duration of volcanism, Pb, Hf, Nd, and Sr isotopic evidence of a plume component, and plate reconstruction models that place the Yellowstone plume near the Oregon coast during the Paleogene have led to the conclusion that Siletzia originated as an oceanic plateau or chain of islands produced where a hotspot interacted with a spreading ridge, much like Iceland (Duncan, 1982; Wells et al., 1984; Massey, 1986; Murphy et al., 2003; Madsen et al., 2006; Pyle et al., 2009; McCrory and Wilson, 2013; Wells et al., 2014; Eddy et al., 2017; Phillips et al., 2017). This terrane accreted to North America during the Eocene, but the event was probably not synchronous along strike. Shortening within forearc sedimentary sequences in Washington indicates that the northern part of the terrane accreted between 51.3 and 49.9 Ma (Eddy et al., 2016); a 48.7 Ma sill in southwest Washington may record extensional stresses associated with establishment of the new subduction zone after accretion was complete (Moothart, 1993; Wells et al., 2014). The short span of time between eruption of Siletzia and its accretion to North America is consistent with the terrane having formed close to the continental margin. Postaccretion (ca. 42–35 Ma) mafic volcanism along the continental margin in Washington and Oregon has been attributed to subduction of a plume-influenced spreading ridge (Chan et al., 2012) or interaction between a plume and the continental margin (Parker et al., 2010; Wells et al., 2014).

The docking of Siletzia triggered a major reorganization of the North American margin at the latitude of Washington, Oregon, and British Columbia: The existing subduction zone was abandoned, and a new one was established farther to the west, outboard of the newly accreted terrane (Sigloch et al., 2008; Schmandt and Humphreys, 2011). Prior to this event, from ca. 65 to 50 Ma, there is only sparse evidence of magmatism in western Washington (Miller et al., 2016), although widespread Challis-Kamloops belt activity was under way to the east of the modern Cascades (Ewing, 1980; Thorkelson and Taylor, 1989; Dostal et al., 1998; Breitsprecher et al., 2003; Haeussler et al., 2003; Madsen et al., 2006; Bordet et al., 2014).

The establishment of the Cascade arc has traditionally been dated to ca. 37–36 Ma, although isolated exposures of calc-alkaline volcanic rocks suggest activity may have begun as early as 44 Ma (Christiansen and Yeats, 1992; Sherrod and Smith, 2000; MacDonald et al., 2013; Dragovich et al., 2016). In the central Washington Cascades, exposures of early Eocene igneous rocks are scarce, but among those with arc geochemical traits, there is an age gap between a volumetrically minor 53.7 Ma silicic tuff (this study) and ca. 42 Ma (Vance et al., 1987), a hiatus that coincides with Siletzia accretion. The BSC lavas were emplaced during this interval, along with six other rock units in western Washington that together define an arcuate belt, ∼160 km long, that parallels the eastern boundary of Siletzia in the subsurface as inferred from seismic data (Fig. 1; Trehu et al., 1994). Rocks in this belt cluster between 50 and 48 Ma in age and are notable for their petrologic diversity (Dragovich et al., 2016; Eddy et al., 2016). All of the volcanic units are bimodal, and many of the mafic rocks, including those in the BSC, have oceanic rather than arc chemical affinities (Tabor et al., 2000; Tanner and Tepper, 2016). Conversely, most of the felsic rocks in this belt, including tuffs and the S-type granites of the Mount Pilchuck stock and Bald Mountain pluton, have isotopic and petrologic traits indicative of crustal melting (Tabor et al., 2002; Dragovich et al., 2016).

Field Relations and Previous Geochronology

The BSC consists of an ∼1500-m-thick section of subaerial basalt flows, interbedded with minor sandstone, shale, and tuff, that unconformably overlie highly deformed turbidites of the Rimrock Lake Inlier (Fig. 2), a dome-like uplift of Mesozoic basement rocks (Walsh et al., 1987; Miller, 1989). At least five scattered exposures of early to middle Eocene volcanic and sedimentary rocks exist along the margins of this uplift, but the section at Summit Creek is the oldest, thickest, and most complete (Vance et al., 1987; Walsh et al., 1987). The BSC is overlain by the late Eocene–Oligocene Ohanapecosh Formation, a voluminous package of subaerial and subaqueous tuffs and andesitic flows that range in age from 36 to 28 Ma and are interpreted as an early expression of the Tertiary Cascade arc (Vance et al., 1987). In the Summit Creek area, Ohanapecosh beds and BSC flows have the same steep (>60°) westward dips, suggesting they are concordant and underwent folding together, likely between 27 and 25 Ma (Vance et al., 1987), and implying that the BSC was approximately flat-lying prior to that time.

Summit Creek lava flows typically strike NNW to NNE and dip >60° to the west. These orientations appear consistent throughout the outcrop area; there is no indication of major internal deformation within the unit. Excellent exposures occur along U.S. Highway 12, which cuts through the section roughly perpendicular to strike (Fig. 2). Here, the flows are generally not more than a few meters thick, aphanitic to sparsely porphyritic, and massive or sparsely vesicular. Off the highway, outcrop is discontinuous, and indications of flow thickness or orientation, such as flow tops or vesicle horizons, are rarely discernible. Surface oxidation, calcite veins, vesicle fillings, and other signs of alteration and weathering are common, as are younger andesitic dikes and sills related to the Tertiary Cascade arc. Discontinuous patches of sedimentary rock and tuff occur at the contact between the BSC and the underlying basement rock. Along Highway 12, these include ∼8 m of black shale overlain by ∼12 m of fluvial sandstone and pebble conglomerate–bearing clasts derived from the underlying Mesozoic units (Vance et al., 1987).

Vance et al. (1987) determined an age range of 55–45 Ma for the BSC, bracketed by a 55 ± 3 Ma U-Pb zircon age on an underlying lapilli tuff (which has arc chemical affinities and is part of an earlier magmatic episode that predates the BSC) and two zircon fission-track ages of 46.1 ± 2.7 Ma and 44.0 ± 3.9 Ma from tuffs inferred to be on strike with the upper part of the basalt section. Two other small exposures of Eocene volcanic units that overlie the Rimrock Lake Inlier south of the BSC (Fig. 2) were also dated by Vance et al. (1987) and yielded fission-track ages of 41.8–42.4 ± 3.8 Ma. These units are roughly age equivalent to the late Eocene Northcraft volcanics that occur to the west (Phillips et al., 1988) and probably represent early stages of Cascade arc magmatism. In the field, these units clearly differ from the BSC in that they contain significant proportions of intermediate and felsic rocks.

Sample sites (available in the BSC sample locations supplement in the GSA Data Repository Item1) encompass the BSC outcrop area and include two transects roughly perpendicular to the strike of the section: one along US Highway 12 east of the junction with WA Highway 123 (Fig. 2), and a second roughly five km farther north in the Gifford-Pinchot National Forest. Samples were chosen for analysis based on stratigraphic position and extent of post-eruptive alteration and weathering. Mineralogical and textural properties of BSC lavas were determined by petrographic examination of 11 representative samples. Mineral compositions (feldspar, pyroxene, Fe-Ti oxides) were measured on a JEOL 733 electron microprobe at the University of Washington.

Zircons were separated from two silicic volcanic rocks and dated using chemical abrasion-isotope dilution-thermal ionization mass spectrometry (CA-ID-TIMS) following methods slightly modified from Mattinson (2005) and described in Appendix A of Eddy et al. (2016). All measurements were made on either the VG Sector 54 or Isotopix X62 thermal ionization mass spectrometer (TIMS) at the Massachusetts Institute of Technology using the EARTHTIME 205Pb-233U-235U isotopic tracer (Condon et al., 2015; McLean et al., 2015) and are presented in Table 1. We corrected for Pb fractionation using α = 0.25 ± 0.04 % (2σ, Sector 54) or 0.18 ± 0.04 % (2σ, X62) based on repeat measurements of the NBS-981 Pb isotopic standard and U fractionation using the known ratio of 233U/235U in the EARTHTIME 205Pb-233U-235U (ET535) isotopic tracer and assuming a zircon 238U/235U of 135.818 ± 0.045 (2σ: Hiess et al., 2012). We assume that all 204Pb arises from laboratory contamination and corrected for this contamination using an isotopic composition of common lead of 206Pb/204Pb = 18.145833 ± 0.475155 (1σ abs.) 207Pb/204Pb = 15.303903 ± 0.295535 (1σ abs.) and 208Pb/204Pb = 37.107788 ± 0.875051 (1σ abs.) based on 149 procedural blanks measured in the MIT isotope geochemistry laboratory between 2009 and 2015. A correction for initial secular disequilibrium in the 238U-206Pb decay chain due to exclusion of Th during zircon crystallization was done using a [Th/U]magma = 2.8 ± 1 (2σ, abs.), which encompasses the range of [Th/U] seen in most felsic tuffs (Machlus et al., 2015). All data reduction was done using the UPb_Redux software package (Bowring et al., 2011) using the methods outlined in McLean et al. (2011) and the decay constants presented in Jaffey et al. (1971). Uncertainties are reported in the format ± X/Y/Z, where X is the analytical uncertainty, Y incorporates uncertainty related to the calibration of the ET535 tracer, and Z includes the uncertainty in the Jaffey et al. (1971) decay constants. We use the expected 2σ variability in the mean square of weighted deviates (MSWD, Wendt and Carl, 1991) to assess whether a group of zircon dates forms a single population within analytical uncertainty, and report weighted mean eruption/deposition ages for the two samples below.

A total of 39 BSC samples were trimmed to remove weathered material and prepared for whole rock isotopic and major and trace element analyses, the results of these analyses are contained in Table 2. Samples for isotopic and trace element analysis were pulverized in an alumina shatterbox; all other samples were pulverized in tungsten carbide. Powdered samples were fused using LiBO4, dissolved in HNO3, and analyzed for their major element compositions using inductively coupled plasma-optical emission spectrometry (ICP-OES), either at the University of Puget Sound or at ALS minerals in Vancouver, British Columbia. Loss on Ignition (LOI) was determined gravimetrically after firing for one hour at 1000 °C. Trace element compositions of 28 samples were also measured by inductively coupled plasma-mass spectrometry (ICP-MS) at ALS Minerals.

Isotopic analyses (Sr-Nd-Pb) of seven samples that reflect the chemical and stratigraphic diversity of the BSC were performed at the University of Washington using a Nu Instruments multicollector (MC) ICP-MS. Strontium and Nd were separated according to the procedures outlined in Nelson (1995). Analyses followed Brach-Papa et al. (2009) for Sr and Gaffney et al. (2007) for Nd. Lead separation and MC-ICP-MS analytical procedures were described in Harkins et al. (2008). To assess the isotopic effects of alteration three samples were analyzed in duplicate, one aliquot of each leached with 6M HCl at 93°C for 30 minutes prior to dissolution. The results of these leaching experiments (Table 3) indicate that for Nd the differences between leached and unleached samples are comparable to analytical uncertainty (0.1-0.3 εNd); while for Pb, although leaching effects are greater (up to 0.15% for 208Pb/204Pb), the effect is not large enough to significantly alter our interpretations. The impact of leaching on measured 87Sr/86Sr ratios is up to 160 ppm and not always toward a less radiogenic composition, although initial ratios of leached samples tend to be lower (0.70315–0.70331) than those of the unleached samples (0.70320–0.70389). This suggests while Sr and Nd isotopic data generally covary as expected, a portion of the 87Sr/86Sr variation could be the result of low temperature alteration.


Most BSC lavas are aphanitic to sparsely porphyritic, and some contain amygdules. Normally zoned plagioclase (An66-An28) is the dominant phenocryst, ranging up to 3 mm long and in some cases forming glomerocrysts. Clinopyroxene (Mg# 0.87–0.68) is common in the groundmass but rare as a phenocryst. Fe-Ti oxides are present mainly in the groundmass and constitute over 10% of the mode in some of the more differentiated rocks. The groundmass is typically microcrystalline, accompanied in some cases by interstitial patches of green secondary minerals that may represent altered glass. Minor alteration is pervasive and includes amygdule fillings, veins of calcite, oxidation of mafic silicates, and partial replacement of primary minerals by epidote, sericite, and other secondary phases.

U-Pb Zircon Geochronology

Sample SC12–20 was collected from an ash-flow tuff near the base of the section and gives a weighted mean age of 53.744 ± 0.032/0.043/0.072 Ma (n = 7, MSWD = 1.8), which we interpret as the age of eruption/deposition (Fig. 3A). This tuff has arc-like chemical affinities such as high field strength element (HFSE) depletions and large ion lithophile element (LILE) enrichment, indicating that it was part of an earlier magmatic episode that predates both the BSC and establishment of the Cascade arc. This tuff provides a maximum age for the onset of BSC magmatism. Vance et al. (1987) obtained a U-Pb age of 55 ± 3 Ma from this unit (sample JV310). Sample NC-MPE-390 was collected from a rhyolite near the top of the section. The dispersion seen in all nine zircon dates from this sample greatly exceeds that expected for a single population (MSWD = 50). However, the youngest four grains overlap within uncertainty and give a weighted mean 206Pb/238U date of 48.086 ± 0.086/0.055/0.075 Ma (n = 4, MSWD = 1.8), which we interpret as the age of eruption/deposition (Fig. 3B). For this unit, Vance et al. (1987) determined a fission-track age of 46.1 ± 2.7 Ma (sample JV232).


The BSC is bimodal in composition, consisting dominantly of tholeiitic basalts and basaltic andesites with minor rhyolite (Table 2; Fig. 4; full BSC data set is available in the BSC major- and trace-element data supplemental material in the Data Repository Item). Considerable variation is present in Mg# (molar Mg/[Mg + FeTotal]), which ranges from 0.66 to 0.29 (and to 0.08 if rhyolitic samples are included). In general, as Mg# decreases, CaO contents decline, TiO2 and P2O5 contents rise to a maximum at Mg# ∼0.4 and then decrease, Na2O and K2O contents increase steadily, and Al2O3 contents remain about the same (Fig. 5). However, major-element data show considerable scatter: CaO, for example, commonly varies by 3–4 wt% among samples with the same Mg# (Fig. 5). Trace-element concentrations show considerable variation and generally coherent trends, excluding a few samples for which there is isotopic evidence suggestive of source heterogeneity. With lower Mg#, Cr (550 to <10 ppm) and Ni (225 to <5 ppm) concentrations decrease markedly, while Zr concentrations increase steadily by a factor of 2–3, and V rises to a maximum at Mg# ∼0.4 and then decreases (Fig. 5). With differentiation, other incompatible trace elements (e.g., rare earth elements [REEs], Ba, Zr, Nb, Th, U, Hf) increase by a factor of 3–4 (and up to 10 or more if rhyolitic samples are included). All BSC samples show moderate light (L) REE enrichment (La/YbN = 1.2–5.9; Fig. 6A) that varies modestly among samples with similar Mg# (∼1–3 at Mg# = 60). Europium anomalies are small (Eu/Eu* = 1.2–0.9), except in the rhyolite samples (Eu/Eu* = 0.3–0.5). Primitive magmas (Mg# >62, Cr >300 ppm) are restricted to the lower part of the section; aside from this, no stratigraphic trends in composition are evident.

Roughly two thirds of the analyzed BSC samples have distinctly low K2O and Rb contents, a characteristic clearly visible on a MORB-normalized spidergram (Fig. 6B). There is no correlation between alkali depletion and either Mg# or stratigraphic position, but in general the alkali-depleted samples also have lower concentrations of REEs, particularly LREEs, and HFSEs than their non-alkali-depleted counterparts (Fig. 6). Low-temperature alteration could have removed K and Rb but is unlikely to have affected REEs or HFSEs. In addition, there is no evidence that the alkali-depleted samples are more altered, as average LOI contents of the alkali-depleted samples (3.3 wt%) and non-alkali-depleted samples (3.4 wt%) are similar. These observations suggest that alkali depletion could be a primary feature of some BSC samples.

Initial εNd values cluster between +6.6 and +5.6, whereas initial 87Sr/86Sr values are more variable, ranging from 0.70316 to 0.70389 (Table 3). These data overlap with the Sr-Nd isotopic compositions of Cascade arc rocks in general (Fig. 7A), but they have higher εNd and lower 87Sr/86Sr values than samples from nearby Mount Rainier (Sisson et al., 2014). Compared to Crescent basalts (Tepper et al., 2008; Pyle et al., 2009; Sisson et al., 2014), BSC lavas display a nearly identical range of initial εNd values but a more restricted range of initial 87Sr/86Sr, with no ratios >0.704 (Figs. 7A and 7B). The BSC data define a slightly inclined linear array (Fig. 7A) that has no apparent correlation with indices of differentiation (e.g., Mg#, Cr content, or K2O content) or stratigraphic position.

Measured Pb isotopic compositions of most BSC samples (Table 3; Figs. 7C and 7D) plot on or slightly below the Northern Hemisphere reference line (NHRL; Hart, 1984). These samples lie within the broad field of data for Crescent basalts (Pyle et al., 2009; Chan et al., 2012; Haileab et al., 2012; Sisson et al., 2014; Phillips et al., 2017). They are more radiogenic than modern Juan de Fuca MORB (White et al., 1987; Hegner and Tatsumoto, 1987) but generally less radiogenic than the late Eocene plume-related Grays River Volcanics (Chan et al., 2012). Pb concentrations of most BSC samples are below detection by the ICP-MS used (∼2 ppm), which precludes calculation of initial Pb isotopic compositions. However, due to their low U and Th contents, even if Pb concentrations of these lavas were as low as typical MORB (0.5 ppm; Doe, 1994), the samples would still straddle the NHRL (Figs. 7C and 7D). Three BSC samples have higher 207Pb/204Pb values (15.589–15.593) than the others and cluster above the NHRL (Fig. 7C). These basalts (Mg# = 0.64–0.53) also have higher U and Th contents, and generally higher La/Yb and Ba/Nb, and one shows HFSE depletions on a spidergram.

Magma Evolution

Low Mg# values and low Cr concentrations indicate that most BSC lavas have undergone significant differentiation. To assess the role of fractional crystallization, we used MELTS (Ghiorso and Sack, 1995) to model the crystallization of potential parent magmas as well as constrain the pressure(s) at which differentiation occurred and the water content of the system. These calculations were performed assuming a parental magma similar in composition to the more primitive BSC samples (50.7 wt% SiO2, 14.3 wt% Al2O3, 0.04 wt% K2O, Mg# = 0.66). Pressure was varied from 0.1 to 10.0 kbar at constant water content of 0.2 wt% H2O, and water contents were varied from 0.1 to 1.0 wt% H2O at constant pressure (P) of 5 kbar.

The main effect of increasing either pressure or water content was to decrease the proportion of plagioclase relative to clinopyroxene in the fractionating assemblage. The impact of this shift on melt composition was most apparent in Al2O3 contents, and the best match between MELTS results and BSC data was obtained with H2O <0.5 wt% and P = 1–5 kbar (Fig. 8), indicating that differentiation occurred mainly at mid- to upper-crustal depths. Using these conditions, the full range of BSC basalts and basaltic andesites (Mg# = 0.65–0.30) can be modeled by a maximum of 64%–77% fractionation of an assemblage dominated by clinopyroxene (52%–68%) and plagioclase (30%–43%) with lesser spinel, olivine, and orthopyroxene. While clinopyroxene phenocrysts are rare, MELTS results do not necessarily conflict with petrographic results. BSC samples are typically phenocryst poor, suggesting that early formed crystals, particularly ferromagnesian minerals, were separated from the magma prior to eruption. Further fractionation of a similar assemblage plus minor apatite and Fe-Ti oxides lead to highly evolved compositions (Mg# <0.20) that are similar to BSC rhyolites (Fig. 8). However, these rhyolites have concave-upward REE patterns (Dy/YbN <1; Fig. 6A) that point to involvement of amphibole (Davidson et al., 2007), a phase that cannot be modeled with the existing version of MELTS. In view of the wide compositional gap that separates the rhyolites from the other BSC lavas (Figs. 5 and 6), we consider a cogenetic relationship with the basalts questionable. Melting of an amphibole-bearing lower-crustal source may be a better explanation than extensive crystal fractionation for derivation of the felsic rocks.

We also investigated the possibility that isotopic variation within the basalts is the result of assimilation-fractional crystallization (AFC) of Mesozoic basement rock (DePaolo, 1981). Our AFC modeling, using the most radiogenic Mesozoic arkose (08RR1001; Sisson et al., 2014) over a range of R (mass assimilated/mass crystallized) and f (melt fraction remaining) values, failed to reproduce the observed Sr-Nd isotopic variations except in cases where extreme distribution coefficients forumla were used. These modeling results and the lack of a correlation between isotopic enrichment and degree of differentiation suggest that assimilation of basement rock was not the primary cause of Sr-Nd isotopic variation among BSC samples. Mantle source heterogeneity (discussed below) and low-temperature alteration that preferentially affected Sr are more likely explanations, the latter of which is consistent with our leaching experiments and with high δ18O values reported for the BSC and other Eocene rock units in the area (Sisson et al., 2014).

Mantle Sources and Heterogeneity

The majority of BSC samples are modestly LREE enriched and have convex, OIB-like patterns on MORB-normalized spidergrams (Fig. 6). Pb isotope compositions of BSC lavas lie along or slightly below the NHRL, are more radiogenic than Juan de Fuca MORB, are less radiogenic than the plume-influenced Grays River Basalts, and are distinct from the field of Cascade arc rocks (Figs. 7C and 7D). Taken together, these chemical and isotopic data indicate that the majority of BSC lavas were derived from an enriched MORB (E-MORB) source that lacked residual garnet.

For most BSC lavas, the traits described above and the absence of Ta-Nb depletions (Fig. 6B) suggest that the subduction-modified mantle wedge was not a source component. MELTS modeling, which suggests parental magmas had <0.5% H2O, further indicates that the source was relatively anhydrous. However, a few lavas have elevated 207Pb/204Pb and HFSE depletions, which point to involvement of a source with a subduction signature, most likely mantle wedge that was modified during earlier subduction. Alternatively, the Pb isotope compositions of these samples, which are shifted toward the compositions of local Mesozoic sedimentary basement rock (Sisson et al., 2014), could be a result of crustal contamination. However, crustal contamination is less likely given that these samples display minimal evidence of fractionation (e.g., high Cr, Eu/Eu* ∼1).

Heterogeneity of the mantle source may also account for some of the incompatible trace-element variations observed among BSC samples. For example, the diversity in Rb and K2O contents among BSC lavas could reflect variable alkali depletion of the source. An alternative possibility is that K2O and Rb were removed during alteration. However, alteration is unlikely to produce the corresponding depletions in LREEs and HFSEs, and there is no observed difference in LOI measurements between samples with and without alkali depletion.

Relationship to the Crescent Formation Basalts

A notable feature of BSC lavas is their chemical and isotopic similarity to basalts of the Crescent Formation, which are of the same age but crop out ∼100 km to the west. On variation diagrams, the two units show extensive overlap in both major- and trace-element composition (Fig. 5). Samples from both units also have similar LREE enrichments and similar diversity in K and Rb concentrations that can be divided into K- and Rb-depleted and K- and Rb-undepleted groups as described above for the BSC (Figs. 6A and 6B). Sr-Nd-Pb isotopic compositions of BSC samples are within the range of data for Crescent samples but less diverse (Fig. 7). Some of the isotopic diversity seen in Crescent samples, in particular, those samples with 87Sr/86Sr >0.704, is probably due to seawater interaction (James et al., 2003). The subaerial BSC samples would not have experienced such alteration and thus have a more limited range of Sr isotopic compositions. Overall, the similarities in age and composition between the BSC and Crescent Formation basalts suggest both units shared a common or at least similar mantle source that was enriched relative to MORB source mantle and moderately heterogeneous in composition.

The age and petrologic similarities of the BSC to Crescent Formation lavas raise the possibility that the BSC could be a displaced block of Crescent Formation that was thrust ∼100 km inland. Seven lines of evidence argue against this hypothesis: (1) There is no indication of large-scale faulting or deformation within or at the base of BSC, and the contact with the underlying Mesozoic rocks is mapped as depositional (Schasse, 1987; Vance et al., 1987). (2) Fluvial sediments interbedded with BSC lavas near the base of the section contain clasts derived from the nearby Mesozoic basement (Vance et al., 1987). These sediments are also lithologically distinct from turbidites and other marine sediments associated with the Crescent Formation (Babcock et al., 1994). (3) BSC flows have the same steep westward dips as the overlying Ohanapecosh Formation, suggesting that the deformation that tilted the BSC did not occur until after Ohanapecosh deposition in the Oligocene. (4) BSC lavas differ from those of the Crescent Formation in that they are entirely subaerial and do not preserve Sr isotopic evidence of interaction with seawater. (5) BSC lavas display no evidence of the prehnite-pumpellyite-facies to lower-amphibolite-facies and lower-blueschist-facies metamorphism that affected the Crescent Formation (Timpa et al., 2005; Hirsch and Babcock, 2009). (6) A few BSC samples show elemental and/or isotopic evidence of derivation from subduction-modified mantle and/or contamination by arc crust, which is known to exist beneath the Summit Creek area (Sisson et al., 2014). Similar arc signatures are unknown among Crescent lavas. (7) Seismic data (Obrebski et al., 2014) indicate that Siletzia does not extend in the subsurface beneath the Mount Rainier area, which also encompasses the Summit Creek area. In summary, field, geochemical, isotopic, and geophysical data all suggest that while BSC lavas originated from similar mantle sources as the Crescent Formation basalts, they erupted in their present setting well inboard of the continental margin.

Tectonic Setting

The tectonic setting in which BSC lavas erupted is constrained by four main observations: (1) The vast majority of these lavas are chemically and isotopically similar to basalts of the Crescent Formation, most likely reflecting a similar petrogenesis. (2) They were emplaced in the forearc region of an active subduction zone (Ewing, 1980; Madsen et al., 2006). (3) They ascended through the mantle wedge but display no chemical signatures that characterize arc magmas. (4) They predate the oldest Cascade arc rocks by >4 m.y. Taken together, these constraints require a mechanism whereby asthenospheric mantle was able to rise through the subducting slab and undergo partial melting during ascent through the wedge without incorporating metasomatized mantle. During the Eocene, two possible tectonic scenarios could meet these requirements: tearing of the Farallon slab (Sigloch et al., 2008; Schmandt and Humphreys, 2011), and passage of a slab window (Thorkelson and Taylor, 1989; Breitsprecher et al., 2003; Madsen et al., 2006). In both scenarios, a discontinuity in the downgoing slab would allow asthenospheric mantle to rise into the mantle wedge and undergo decompression melting.

Alternatively, lavas with MORB- or OIB-like compositions can originate in arcs through melting of heterogeneous mantle sources (Leeman et al., 1990; Jagoutz et al., 2011). Some Quaternary low-K olivine tholeiites in the southern Washington Cascades are broadly similar in composition, although not identical, to the BSC (Leeman et al., 2005). This raises the possibility that BSC magmas could have originated within the mantle wedge, eliminating the need for a break in the slab. However, this scenario requires that mantle indistinguishable from the Crescent basalt source was present in the wedge prior to the accretion of Siletzia, and it underwent melting immediately afterward. While such a process cannot be unequivocally ruled out, during this period rapid, low-angle subduction would have served to both cool the mantle wedge and reduce its thickness. This would have inhibited melting of the mantle wedge and probably explains the scarcity of magmatism in western Washington between 65 Ma and accretion of the Siletzia terrane (Gutscher et al., 2000; Miller et al., 2016).

Instead, we suggest that BSC magmatism was a consequence of slab breakoff. Slab breakoff occurs when buoyant material, such as continental crust, an island arc, an oceanic plateau, or young lithosphere, approaches the trench and resists subduction. The continued downward force of slab pull leads to necking and then tearing (detachment) of the subducted portion of the plate, which can trigger mantle upwelling and subsequent magmatism (Davies and von Blanckenburg, 1995). In the northwestern United States, this would have occurred when Siletzia was accreted to the continental margin at ca. 50 Ma. Evidence of slab detachment in this case is provided by seismic tomography, which reveals a high-velocity “curtain” extending to ∼600 km depth beneath Idaho; this “curtain” is interpreted as a stalled remnant of the Farallon plate that broke off after Siletzia accretion (Sigloch et al., 2008; Schmandt and Humphreys, 2011).

The forearc setting of BSC volcanism and its timing relative to Siletzia accretion are consistent with a breakoff model based on comparisons with: (1) other localities where magmatism is attributed to this process, and (2) the results of numerical modeling in the literature. Miocene forearc volcanism in Baja California, which has been linked to slab breakoff (Pallares et al., 2007), represents a possible analog for the BSC. Normal calc-alkaline volcanism in Baja largely ceased ca. 12.5 Ma as the Pacific-Farallon oceanic spreading ridge approached the subduction zone, and it was followed by forearc magmatism that lacked arc traits from ca. 14.2 to 7.5 Ma. These forearc rocks define a belt ∼600 km long and ∼100 km wide along which magmatism was roughly synchronous. This belt lies 100–200 km inboard of the trench and developed very shortly after the cessation of normal arc magmatism; it is inferred to overlie the gap that opened where the slab detached. The size, location relative to the trench, timing relative to accretion, and chemical traits of Baja forearc magmatism are similar to those of the BSC and the other Eocene belt units emplaced in Washington, suggesting that these units also formed in a slab breakoff setting.

Results of modeling predict that breakoff should occur from <5 to >20 m.y. after the onset of collision and at depths of 35–300 km. However, there is general agreement that breakoff occurs earlier and at shallower levels for younger, slower, warmer, and/or weaker slabs (van de Zedde and Wortel, 2001; Andrews and Billen, 2009; van Hunen and Allen, 2011; Duretz et al., 2014). For situations involving accretion of a spreading ridge, a scenario that may be analogous to the accretion of Siletzia, three-dimensional models predict that strain within the slab becomes great enough to slow subduction when the ridge is within 150 km of the trench, and that the subducted slab will break off within 2 m.y. of initial slowing, when the ridge is 100–115 km from the trench. Tearing is predicted at a depth of 40–55 km and 190–225 km inboard of the trench (Burkett and Billen, 2010). Both the timing and location of BSC eruptions are consistent with these models.

Magmas generated in response to slab breakoff are compositionally diverse and result in short-lived, linear magmatic belts. Davies and von Blanckenburg (1995) suggested that elevated heat flow commonly results in melting of the metasomatized mantle wedge and overlying crust rather than the upwelling asthenosphere itself, and consequently the magmatism tends to be calc-alkaline and felsic. However, where breakoff occurs at shallow depths, as is predicted for a warm young slab, the upwelling asthenosphere may undergo melting and produce magmas with tholeiitic or alkaline compositions (Davies and von Blanckenburg, 1995; van de Zedde and Wortel, 2001). This has been documented in areas of suspected breakoff magmatism, including Newfoundland (Whalen et al., 1996) and Baja California (Ferrari, 2004; Pallares et al., 2007). Thus, the compositions of BSC magmas are consistent with those observed in other areas with suspected breakoff-induced magmatism, and specifically those where breakoff is thought to have occurred at shallow depth.

An alternative tectonic scenario for BSC magmatism is passage of a slab window, an event that has been proposed as the cause of widespread Eocene extension and magmatism in Washington and British Columbia (Haeussler et al., 2003; Madsen et al., 2006). Eruption of basalts with oceanic affinities in the forearc has been attributed to slab window magmatism in a number of arcs, including the Cascades (Chan et al., 2012). Plate-motion reconstructions indicate that the Kula-Farallon and/or Kula-Resurrection Ridges intersected with the North American margin near the latitude of Washington, Oregon, and British Columbia during the Eocene, producing one or more slab windows (Babcock et al., 1992; Breitsbacher et al., 2003; Haeussler et al., 2003; Madsen et al., 2006). However, their precise locations and orientations through time are not well constrained.

Although the location and petrologic characteristics of the BSC are compatible with either a slab window or slab breakoff model, four lines of evidence favor the latter: (1) After Siletzia was accreted, the Cascadia subduction zone jumped westward to its present position. This shift must have been preceded by slab breakoff and was coincident in time with the eruption of the BSC. (2) Between ca. 53 and 45 Ma, a wave of dynamic/thermal uplift (Smith et al., 2014) and magmatism (Tepper, 2016) migrated southwestward (trenchward) across the northwestern United States. This activity is attributed to rollback of the Farallon slab following the accretion of Siletzia, and in other arcs, rollback has been identified as a precursor to slab breakoff (Lucente and Margheriti, 2008; Ozawa et al., 2015). (3) Trenchward migration of uplift and magmatism was roughly perpendicular to the margin, which one would not expect from a migrating slab window, because a slab window would have a significant component of margin-parallel motion. (4) Seismic images of detached Farallon slab fragments (Sigloch et al., 2008; Schmandt and Humphreys, 2011) provide independent evidence of slab breakoff.

A growing body of geochemical, geophysical, and geochronological data, combined with new reconstructions of plate motions, provides support for the hypothesis that the Coast Range basalts of Oregon and Washington originated as an oceanic plateau or chain of islands that was accreted to the margin of North America (e.g., Wells et al., 2014). The accretion of this oceanic terrane, Siletzia, was a significant episode in the Tertiary geologic history of the northwestern United States and British Columbia, but the details of the event, and of its broader impact throughout the region, are only now beginning to emerge (Johnston and Acton, 2003; Wells et al., 2014; Eddy et al., 2016; Tepper, 2016).

Breakoff of the subducting Farallon slab would have been one of the consequences of Siletzia accretion (Fig. 9A), and high-velocity seismic domains in the mantle beneath eastern Washington and Idaho have been interpreted as remnants of this slab (Sigloch et al., 2008; Schmandt and Humphreys, 2011). Another geological manifestation of slab breakoff is a short-lived linear belt of igneous activity localized above the region where asthenospheric mantle upwells through the zone of rupture. We suggest that the BSC is part of such a belt, not previously recognized, that roughly follows the leading subsurface edge of Siletzia for >150 km in western Washington (Figs. 1 and 9B). Evidence supporting a slab breakoff origin for the BSC and other units within this belt includes their restricted range of ages (50–48 Ma; coincident with or slightly after Siletzia accretion), which appear synchronous along its length, their location (in the former forearc region), and their petrologic traits (some units with MORB-OIB affinities, others with crustal melting signatures). Eventually, subduction resumed outboard of the accreted terrane, resulting in renewed arc magmatism. In Washington, this resulted in the earliest Cascade arc magmatism at ca. 44 Ma (Fig. 9C; Christiansen and Yeats, 1992; Sherrod and Smith, 2000; MacDonald et al., 2013; Dragovich et al., 2016).

The BSC represents an important link between magmatic and tectonic events occurring along the continental margin and those occurring further inboard. Age constraints provided by these lavas combined with improved knowledge of the timing of Siletzia accretion (Wells et al., 2014; Eddy et al., 2016) make it possible to better assess the extent to which slab breakoff, likely accompanied by rollback, might have been responsible for other geologic events, in particular, magmatism, extension, and uplift, that occurred in Washington, Oregon, and British Columbia during the Eocene.

Funding for this research was provided by National Science Foundation (NSF) grants EAR-1119252 to J. Tepper and EAR-1118883 to S. Bowring. We thank Ken Clark, Mike Valentine, and Sam Berkelhammer for assistance in the field. Eddy is indebted to S. Bowring for support and encouragement during the early parts of this project. We also acknowledge the Fall 2010 Igneous Petrology class at the University of Puget Sound (Tacoma, Washington) for the data they collected; this study began as their class project. We also thank K. Stuewe, E. Todd, D. Canil, and an anonymous reviewer for their thoughtful comments.

1GSA Data Repository Item 2018359, Basalt of Summit Creek major- and trace-element data, sample locations, and regional data and sources, is available at or on request from
Gold Open Access: This paper is published under the terms of the CC-BY-NC license.