Abstract

Small masses of intraplate mafic plutonic and volcanic rocks are exposed in the Navajo volcanic field (26–24 Ma), Dulce dike swarm (25–20 Ma), and western San Juan Mountains (7–0.6 Ma) along the eastern boundary of the Colorado Plateau. Geochemical and Sr-Nd isotopic data for these rocks were employed to assess the time-space variations in melt compositions and to investigate the contributions from different mantle and crustal sources during magma production.

Geochemical data signify that post–26 Ma mantle melts produced in the Four Corners region were heterogeneous in composition, dominantly alkaline, and distinguished by trace-element patterns with elevated LILE and LREE. These magmas were generated by melting of metasomatized subcontinental lithospheric mantle (SCLM) with minor contributions from a mantle source that produced melts with geochemical affinities of oceanic island basalts. Nd and Sr isotopic data reveal that some melts could have experienced minor contamination (<1%) with upper crust, whereas the Dulce magmas formed by contamination of mantle melts with up to 40% lower crust or melting of SCLM with low time-integrated Sm/Nd. Relative to minettes in the Navajo volcanic field, mafic rocks in the western San Juan Mountains have similar Nd and Sr isotopic compositions but are lower in some chemical values (Mg#, K2O, Rb, Zr, La/Lu, La/Ta, and Th/Yb) and higher in others (Al2O3, Na2O, and CaO). We postulate that these different chemical traits are evidence of heterogeneity in the lithospheric mantle. Voluminous magma production involving mantle melts from 35 to 26 Ma could have modified the chemistry of the lithospheric mantle beneath the western San Juan Mountains. Alternatively, this region may be underlain by SCLM with distinct chemical attributes imparted by subduction processes during Proterozoic arc accretion. Our work provides a snapshot of broader regional trends wherein heterogeneity in mantle reservoirs was a major contributor to the variable compositions of mafic alkaline rocks produced over the Colorado Plateau and adjacent Southern Rocky Mountains after 26 Ma.

INTRODUCTION

Intraplate mantle melts were emplaced along the eastern edge of the Colorado Plateau in the Four Corners region after 26 Ma, producing three distinct fields of ultra to medium potassic mafic rocks (Figs. 1 and 2; Table 1). This record of episodic mantle magmatism is concentrated in zones of incipient crustal extension (Gonzales et al., 2010; Gonzales, 2015) defined by northeast-trending faults and dikes that span the northern margins of the San Juan Basin (SJB on Fig. 1). Mantle magmatism in the Four Corners region is broadly linked to regional, latest Paleogene to Neogene extension that created the Rio Grande rift (Fig. 1) (e.g., Smith, 2004).

The three fields of mafic rocks in the Four Corners area were placed in the upper crust in relative close proximity from 26 to 20 Ma and 7 to 0.6 Ma (Figs. 1 and 2). These samplings of mantle magmas thus afford an opportunity to investigate compositions of melt sources over time and evaluate the factors that influenced magmatism during the shift from compressional to extensional regional tectonics. In this contribution, we present a comprehensive comparison of geochemical and isotopic (Sr and Nd) data (Tables 2 and 3) for the different groups of mafic rocks (Fig. 2). Each group is distinguished by major- and trace-element compositions that define discrete but diffuse fields on chemical plots. Amongst the different groups, there is a gradational shift in chemical compositions, but the Sr and Nd isotopic data for the majority of these rocks are similar with a few notable distinctions. We postulate that the compositional variation amongst the different groups was influenced primarily by heterogeneity in metasomatized mantle lithosphere with a Proterozoic ancestry. The chemical signatures inherited from the SCLM were possibly modified by variable amounts of contamination with lower and upper crust. This investigation provides further insight into mantle-source heterogeneity and magmatic processes that influenced compositional variations in Oligocene to Pleistocene mantle rocks on a broader regional scale (Fig. 2) (e.g., Alibert et al., 1986; Leat et al., 1988; Johnson and Thompson, 1991; Tingey et al., 1991; Gibson et al., 1992, 1993).

REGIONAL GEOLOGIC SETTING

Latest Mesozoic to Cenozoic magmatism on the Colorado Plateau and adjacent Southern Rocky Mountains (Fig. 1) is attributed to shallow subduction of the Farallon plate during the Laramide orogeny, followed by slab rollback and regional Neogene extension (e.g., Coney and Reynolds, 1977; Humphreys, 1995; Humphreys et al., 2003; Chapin et al., 2004; Smith, 2004; Farmer et al., 2008; Chapin, 2012). It is hypothesized that starting around 30 Ma asthenospheric melts underplated and heated the metasomatized lithosphere (Schulze et al., 2015) at depths of 120–150 km beneath the eastern Colorado Plateau in response to the progressive rollback and delamination of the Farallon plate (e.g., Ehrenberg, 1979; Esperanca and Holloway, 1987; Roden et al., 1990; Smith et al., 1991; Tingey et al., 1991; Riter and Smith, 1996; Lee et al., 2001; Usui et al., 2002, 2003; Smith et al., 2004; Farmer et al., 2008). In the Four Corners region, melting of the lithospheric mantle produced alkaline mafic magmas that were emplaced in the upper crust in several pulses after 26 Ma (Fig. 2). The spatial distribution of these melts (Fig. 1) was broadly focused on northeast trends aligned with deep crustal fracture zones that formed by accretion and assembly of 1.8–1.75 Ga arc assemblages during the Yavapai orogeny (ca. 1.71 Ga) (e.g., Tweto and Sims, 1963; Tweto, 1980; Warner, 1980; Karlstrom et al., 2005). This is supported by seismic data from Gilbert et al. (2007) that documents distinct northeast variations in the crust-Moho structure on the eastern edge of the Colorado Plateau; they attribute these variations to crustal accretion and variable mantle metasomatism during the Proterozoic. Inheritance of 1.8–1.4 Ga zircons in 75–4 Ma felsic to intermediate plutonic rocks also reveals that Proterozoic basement contributed to the production of Cenozoic crustal magmas along this northeast-trending zone (Gonzales, 2015).

Geochemical and isotopic data for volcanic rocks in the Southern Rocky Mountains volcanic field (SRMVF) (Fig. 1) lend evidence that alkaline mantle melts contributed to the voluminous production of crustal magmas involved in volcanism from 32 to 25 Ma (e.g., Lipman et al., 1978; Slack and Lipman, 1979; Johnson, 1991; Riciputi et al., 1995; Lipman, 2007; Farmer et al., 2008; Lake, 2013; Lake and Farmer, 2015). Most of the mafic magmas remained at depth, but some were emplaced at shallow levels on the margins of the 29–27 Ma (e.g., Bove et al., 2001) San Juan–Silverton calderas (e.g., ca. 27 Ma Mount Sneffels–Stony Mountain gabbro-diorite stock; SN on Fig. 2). Continued emplacement of mantle melts within zones of minor extension (Gonzales, 2009, 2013, 2015; Gonzales et al., 2010) on the eastern edge of the Colorado Plateau (Fig. 1) likely contributed to elevated crustal geotherms that promoted small-volume crustal melting and production of post–26 Ma granitic to dioritic plutons in southwestern Colorado (Lipman et al., 1976; Lipman, 1989, 2007; Gonzales, 2015).

Although mafic magmas were intermittently emplaced at upper crustal levels across the Four Corners region after 26 Ma (Figs. 1 and 2), the greatest production was from 26 to 20 Ma (summarized in Gonzales, 2015). The ages and distribution of mafic rocks in the region define three distinct fields: (1) 26–24 Ma diatreme-dike complexes in the Navajo volcanic field (NVF); (2) 25–20 Ma dikes in the northern extension of the Dulce swarm; and (3) 7–0.6 Ma dikes and minor lava flows in the western San Juan Mountains (WSM) (Fig. 2). These Oligocene to Pleistocene mafic rocks are similar in age and composition to alkaline mafic dikes, lava flows, and cinder cones in northwestern Colorado (Leat et al., 1988; Gibson et al., 1993) and alkaline mafic dike swarms exposed on the Wasatch Plateau and Capitol Reef area in Utah (Gartner and Delaney, 1988; Tingey et al., 1991) (Fig. 1).

The Navajo volcanic field is distinguished by diatreme pipes and related maar craters, plugs, and shallow intrusive masses that span the east-central margin of the Colorado Plateau (Figs. 1 and 2) (Gregory, 1917; Williams, 1936; Appledorn and Wright, 1957; Akers et al., 1971; Naeser, 1971; Roden et al., 1979; Laughlin et al., 1986; Nowell, 1993; Gonzales et al., 2010). The latest Miocene alkaline mafic flows of the Hopi Buttes are exposed southwest of the Navajo volcanic field (e.g., Fitton et al., 1988; Alibert et al., 1986) (Fig. 1). Laughlin et al. (1986) and Nowell (1993) constrained the timing of NVF magmatism at 28–19 Ma, but more recent Ar-Ar analyses indicate that most of the magmatism was from 26 to 24 Ma (Gonzales et al., 2010; Nybo et al., 2011; Carrara, 2012; Nybo, 2014) in agreement with K-Ar ages of Roden et al. (1979) for minette at Buell Park. Nybo (2014) concluded that the ca. 9 Ma range in ages reported by Laughlin et al. (1986) and Nowell (1993) was probably a reflection of argon loss or excess argon in some of the samples analyzed.

Among the potassic rocks (K2O > Na2O) in the NVF (Figs. 1 and 2), minette is volumetrically the most abundant (minette as defined by Rock, 1977, 1991; Mitchell, 1995, 1997; Woolley et al., 1996; Steckeisen, 1979). The minettes are distinguished by phenocrysts of phlogopite + diopside ± olivine in groundmass dominated by phlogopite and sanidine (Table 2). Dikes of katungite (comparable to potassium-rich olivine melilitolite of Le Maitre et al., 1989) are exposed in a small swarm near Hasbidito Creek in Arizona (HC in Fig. 2). The katungites have phenocrysts of olivine and phlogopite in groundmass composed chiefly of phlogopite + melilite + perovskite + apatite ± clinopyroxene. Rocks with compositions that are transitional (Nowell, 1993) to minette and katungite are exposed in an east-west–trending dike swarm near Newcomb (NC on Fig. 2). These rocks have similar mineral assemblages to NVF minettes, although minor nepheline and melilite was identified by Nowell (1993) in some samples.

The 25–20 Ma Dulce dike swarm (Gonzales, 2015) crystallized from mantle melts that invaded a zone of minor extension on the northeast edge of the San Juan basin. The dike zone is roughly 100 km in length by ∼30 km in width (Fig. 1). Farther to the northeast, this swarm melds with 20–26.5 Ma “dacite” dikes that radiate from the Platoro caldera complex in Colorado (Lipman, 2007; Lipman and Zimmerer, 2016). This relationship further hints at a magmatic connection (noted by Gonzales, 2015) of postcaldera crustal melts and mantle magmas on the east margin of the Colorado Plateau. We collected samples (Tables 1 and 2) from dikes exposed in Colorado within the northern extent of the Dulce swarm (Fig. 2). These rocks are texturally similar to NVF minettes but differ slightly in mineralogy (Table 1). They have K2O > Na2O (0.12–1.36, Table 2), and contain abundant phenocrysts of diopsidic augite and phlogopite in fine-grained groundmass dominated by calcic plagioclase. On the basis of these characteristics (Tables 1 and 2), we classify the Dulce rocks as kersantite (after Rock, 1977; Mitchell, 1995, 1997; Woolley et al., 1996; Steckeisen, 1979).

A small number of 7–4 Ma (Gonzales, 2015) mafic dikes were emplaced in the western San Juan Mountains (Bush et al., 1959, 1960; Bromfield, 1967) in close association with late Miocene to Pliocene felsic to intermediate hypabyssal plutons (Lipman et al., 1976; Wareham, 1991; Cunningham et al., 1994; Wareham et al., 1998; Gonzales, 2015). These dikes are exposed in a north to northeast zone that extends for ∼100 km (Fig. 1). They were classified into two types of “lamprophyres” on the basis of mineral composition and chemistry by Bromfield (1967). One group is herein classified as kersantite (Rock, 1977; Steckeisen, 1979; Mitchell, 1995, 1997; Woolley et al., 1996) because the rocks contain phenocrysts of diopsidic augite, biotite, and minor olivine in groundmass dominated by plagioclase and lesser amounts of sanidine or orthoclase (Table 1). A few rocks in this group (identified with PV in Table 2) are distinguished by poikilitic sanidine in the groundmass, which was also observed in some NVF minettes. The other group of WSM “lamprophyres” are characterized by 1–3 mm phenocrysts of clinopyroxene and green to brown amphibole set in fine-grained groundmass containing biotite, opaque minerals, and feldspar (dominated by plagioclase). The dominance of pyroxene + amphibole in these rocks, together with groundmass assemblages, define these rocks as spessartite (after Rock, 1977; Mitchell, 1995, 1997; Woolley et al., 1996; Steckeisen, 1979).

A basaltic lava flow exposed on Specie Mesa near Placerville represents the most recent product of mantle magmatism in the region (Fig. 2; Tables 1 and 2). The flow erupted along a series of north to northwest-trending normal faults, and on the basis of field relationships it was interpreted as Quaternary (Bush et al., 1960). A new 40Ar/39Ar whole-rock analysis constrains the age of the flow at 614 ± 5 ka (personal commun., M. Heizler, 2016, New Mexico Geochronology Research Laboratory).

MAJOR-ELEMENT CHEMISTRY

Seventy-eight samples were analyzed for major- and trace-element geochemistry (Table 2); a summary of methods and analytical parameters are provided in the Appendix. The majority of the samples collected showed no visible signs of alteration in outcrop, and petrographic analyses confirmed an absence of alteration in most samples. The exceptions were several NVF and WSM samples in which olivine phenocrysts were partially altered to chlorite ± serpentine, and a few Dulce samples in which biotite and pyroxene were rimmed by secondary chlorite. Some of the outcrops sampled within the study area (Fig. 1) contain calcite in cavities, veins, and hydrothermal-breccia fillings. Field and petrographic evidence suggests that most of the calcite was related to magmatism and not to postmagmatic alteration. Irrespective of their origin, these calcite-rich zones were avoided during sampling. Petrographic analyses reveal that secondary calcite only occurs in the groundmass of some Dulce samples, which likely contributed to their high loss on ignition (LOI). The relatively high LOI values listed on Table 2 are similar to those (2–9 wt%) documented for “unaltered” alkaline mafic rocks worldwide (e.g., Roden and Smith, 1979; Leat et al., 1988; Wallace and Carmichael, 1989; Tingey et al., 1991; Gibson et al., 1992; Gibson et al., 1993; Nowell, 1993; Gibson et al., 1995; Righter and Rosas-Elguera, 2001; Kheirkhah et al., 2015). For phlogopite-bearing rocks, the high LOI is attributable, in part, to high modal proportions of phlogopite, which can contain up to 4–5 wt% H2O+ (table 18 in Deer et al., 1978; Alietti et al., 1995). The modal percentages of phlogopite (phenocrysts and groundmass) in our samples (Table 2), especially NVF rocks, vary from 10% to 40% and contributed to high loss of H2O+ during major-element analyses. Given the mineral assemblages in our samples, and lack of evidence for secondary alteration, we attribute the high LOI values to primary rock composition. The high LOI and low totals (∼95%) for two of the WSM spessartite samples (L-10-EL-FM-01C and L-11-EL-CM-06; Table 2), however, are problematic. The major-oxide data for these two samples are listed on Table 1 and plotted on the various diagrams, but these data are questionable.

The wide range of crystallization ages (2–7 m.y.) for mafic rocks in any given field (Tables 1 and 2) makes it difficult to assess magmatic relationships of the rocks and complicates the use of chemical models for quantitative assessment of differentiation processes such as crystal fractionation. Models to assess different degrees of source melting are also inhibited by the lack of trace-element compositions for mantle peridotites found in the region. Nevertheless, general geochemical patterns are defined by our data, and these allow a qualitative assessment of source heterogeneity and contamination processes.

Most of the samples we analyzed (Table 2) plot in the alkaline field on the total alkali silica (TAS) diagram (Fig. 3) (Le Maitre et al., 1989). There is a wide range of SiO2 (34–58 wt%) and Na2O + K2O (2–10 wt%) on the plots with considerable overlap of data amongst the three rock groups (Table 2). Samples from the Navajo volcanic field show the greatest spread on the TAS plot, but there is also notable variation for the WSM and Dulce samples.

For all samples analyzed (Table 2), there is a wide range of Mg# values (80–40, defined as 100*[Mg2+/(Mg2+ + Fe2+]) relative to SiO2 concentrations of 34–58 wt% (Fig. 4). The NVF samples are distinguished by a continuum of SiO2 from 34 to 55 wt% with Mg# from 60 to 80. Most of the WSM and Dulce samples have 40–55 wt% SiO2, but the Mg# for a majority of WSM rocks is 55–65, whereas those for the Dulce samples plot around 50. The K2O versus SiO2 plot (Fig. 4, boundaries from Peccerillo and Taylor, 1976) highlights the distinctions between the different rock groups, despite some overlap. Navajo volcanic field katungites with 34–35 wt% SiO2 and 1.5–3.5 wt% K2O plot in a distinct cluster in the low-silica range of the shoshonite field. Most of the NVF minettes are ultrapotassic and plot in the shoshonitic field but have a wider range in K2O concentrations compared to the katungites. Minette samples with the lowest SiO2 overlap the field of NVF transitional rocks in the low K2O part of the shoshonitic field, as do the majority of the WSM samples. In contrast, the Dulce samples plot from the shoshonitic field into the high- to medium-K fields, reflecting their overall lower K2O concentrations (Table 2).

Navajo volcanic field katungites are the least evolved of all rock types with 33–37 wt% SiO2, high Mg#, and highest overall wt% MgO, FeO*, and CaO (Fig. 4). The katungites have relatively uniform K2O/Na2O from 1.03 to 1.95, and are distinguished by high proportions (wt%) of normative olivine (18.4–27.3), nepheline (5.7–10.5), and leucite (8.2–13.7). Navajo volcanic field minettes show the greatest variation in major-element chemistry but overall are silica undersaturated and potassic to ultrapotassic with K2O/Na2O of 0.70–4.84. These chemical trends are reflected by variable but generally high normative orthoclase (12.2–43.3 wt%), and normative nepheline (0.1–12.2 wt%), in most samples. Navajo volcanic field transitional rocks plot between the katungites and minettes on the Harker diagrams (Fig. 4) and are distinguished by normative olivine (3.1–11.5 wt%) + nepheline (2.8–7.4 wt%) ± leucite (2.7–22.8 wt%), and Or > (An + Ab).

The major-element oxides of NVF minette + transitional samples define distinct linear trends. With increasing SiO2 concentrations, there are positive correlations for Al2O3, K2O, and Na2O and negative correlations for MgO, FeO*, CaO, and TiO2. Nowell (1993) argued for 11% to 21% closed-system crystal fractionation of diopside + phlogopite + apatite + magnetite from primary melts to produce the range of chemical compositions in NVF minettes. Alibert et al. (1986) also interpreted the variations in MgO, Cr, Ni, and REE as the result of fractionation of olivine + pyroxene from primary melts, and Roden et al. (1990) proposed that chemical variations in minettes at Buell Park were caused by fractional crystallization of primitive melts derived from sources that were isotopically heterogeneous. These models define samples with high MgO, Ni, and Cr as representatives of primitive or primary mantle melts. A challenge to this assumption comes from several experimental studies (Wyllie and Sekine, 1982; Sekine and Wyllie, 1983; Lloyd et al., 1985). These experiments produced a wide range of alkaline melt compositions by partial melting of heterogeneous subcontinental mantle composed of phlogopite + clinopyroxene + orthopyroxene without involving crystal fractionation from a “primary” melt. The linear variations in major- and trace-element compositions for NVF minettes hint that fractional crystallization was a factor, but it probably was not the primary cause of all of the chemical variation observed. In contrast to the NVF minettes and transitional rocks, the Dulce and WSM data (Table 2) do not exhibit well-defined trends (Fig. 4).

Western San Juan Mountains spessartites are distinguished by generally higher Al2O3 and K2O and a wider range of K2O/Na2O (0.36–6.58) relative to WSM kersantites (0.81–2.82). Most of the WSM dike rocks and the Specie Mesa basalt (Table 2) are distinguished by normative (Ab + An) >> Or, olivine (0.6–10.6 wt%), nepheline (0.1–12.1 wt%) ± leucite (0.9 wt%). Major-element values for the Specie Mesa basalt (Tables 1 and 2) mostly plot in the clusters defined by the Dulce samples (Fig. 4).

The Dulce rocks generally have lower Mg#, CaO, K2O, and K2O/Na2O (0.12–1.36) relative to NVF and WSM rocks (Fig. 4). Most of the samples have normative (Ab + An) >> Or + olivine + nepheline, but four samples (DU3, DU5, DU6, and DU10; Table 2) are quartz normative (0.7–1.6 wt%). The normative quartz could be attributed to crustal contamination in these rocks.

TRACE-ELEMENT CHEMISTRY

Plots on Figure 5 further reveal the distinct groupings of chemical data for different mafic rock suites in the region (Fig. 2; Table 2). On the plot of La/Ta versus K2O (Fig. 5A), the Dulce samples have the lowest K2O with La/Ta ratios of 10–90, and the data cluster around average regional lower crust. Most of the NVF katungites plot in the upper extent of the oceanic island basalt (OIB) field with La/Ta ratios of 0–40 and 1–4 K2O, together with a few NVF minette samples and most of the WSM rocks. The majority of the NVF minette and transitional samples have K2O from 3 to 8 wt% and a greater range in La/Ta (10–150). The variations in K2O could in part be attributable to fractional crystallization, but the fields defined are consistent with heterogeneous mantle sources. The data for Dulce dike rocks are compatible with mixing of mantle melts and lower crust.

A plot of La/Ta against K/Ta*1000 (Fig. 5B) shows a continuum from NVF katungites (OIB) on the lower end of the K/Ta spectrum to NVF minettes on the upper end. The overall trends are similar to the results defined by Leat et al. (1988) for middle to late Cenozoic mafic rocks in northwestern Colorado (Fig. 1); the sources for these rocks were interpreted as asthenospheric (OIB) (Group 1) to subcontinental mantle (Group 2).

The plot of Ta/Yb against Th/Yb (Fig. 5C) after Pearce (1983; Pearce and Peate, 1995) also displays distinct fields with well-defined ranges in Th/Yb for all three rock groups. Navajo volcanic field katungites plot in the upper part of the mantle array in the OIB field, whereas the majority of the NVF minettes and transitional rocks plot between the katungites and potassic “arc” rocks (shoshonitic field). Although the Th/Yb ratios for the WSM and Dulce samples are lower than for NVF samples, there is a similar shift in the Ta/Yb ratios from enriched mantle into the shoshonitic and calc-alkaline fields. The overall trends defined on this plot might be explained by variable mixing of melts from enriched asthenospheric mantle and metasomatized lithospheric mantle or heterogeneity in mantle sources.

Plots of La/Lu relative to concentrations of compatible (Ni and Cr) and more incompatible trace elements (Ba, Nb, Rb, and Zr) (Fig. 6) for samples show a wide variation in concentrations over a range of La/Lu from 50 to 1000. Ba and Zr show a general positive correlation with increasing La/Lu, for all samples, but there are no definitive trends for the other elements. Katungite samples have the least amount of element variation and higher overall Nb, Ni, Cr, and Zr concentrations. There is a general progressive decrease in La/Lu and element concentrations from NVF > WSM > Dulce samples.

The chondrite-normalized, rare-earth element (Sun and McDonough, 1989) (Fig. 7) patterns for all samples display an enrichment in light rare-earth elements (LREEs) relative to heavy rare-earth elements (HREEs) (LaN/YbN = 10–100) and relatively flat HREE patterns (DyN/YbN = 1.2–1.5). A subset of NVF minette-transitional rocks has the highest normalized LREE, whereas the lowest are exhibited by several samples of WSM samples. These patterns are similar to those of potassic magmas formed in subduction margins.

Mid-ocean ridge basalt (MORB)–normalized multi-element plots (after Pearce, 1983) (Fig. 8) show variable but overall high enrichment in incompatible large-ion lithophile elements (LILEs). The patterns for NVF katungite samples are broadly similar to those for oceanic island basalts but with higher LILE and LREE. Navajo volcanic field minette-transitional samples are slightly more enriched in LILE compared to WSM or Dulce samples. Excluding the katungite samples, the MORB-normalized patterns for all samples are distinguished by moderate to pronounced negative Ta-Nb troughs and subtle Zr-Hf negative anomalies that are similar to patterns of rocks generated in continental arcs. Roden et al. (1990) argued that Ta-Nb depletion in “felsic” minettes (sanidine rich) in the NVF could reflect fractionation of Ta-Nb–rich mineral phases from more primitive magmas, but this is not supported by the fact that some NVF minettes (samples at Johnson Canyon, Red Wash, and Boundary Butte) with high proportions of sanidine and phlogopite have higher Nb concentrations relative to olivine-bearing minettes (Table 2). Tingey et al. (1991) proposed that the Ta-Nb troughs could simply be the results of enrichment of LILE and LREE relative to Ta and Nb (and Zr-Hf) in a metasomatized source that interacted with enriched fluids or magmas. For the NVF transitional and minette samples, the normalized Th values define distinct peaks on the MORB-normalized plots relative to Ta and Ba, whereas the WSM and Dulce samples lack this pronounced Th enrichment. The lower normalized Th could be explained by fractionation of phlogopite, hornblende, or pyroxene from the melts or the presence of a Th-rich mineral in the melt source.

Sr-Nd ISOTOPE SIGNATURES

Age-corrected Sr and Nd isotopic ratios for samples from the different rock groups (Table 3) have higher 87Sr/86Sr(i) ratios and lower εNd (t) with respect to depleted asthenospheric mantle (MORB) (Fig. 9). Relative to bulk silicate earth (BSE), the majority of the isotopic data (Table 3) are moderately radiogenic in 87Sr/86Sr(i) (0.7050–0.7060) and slightly enriched or depleted in 143Nd/144Nd, giving εNd(t) values of +2 to –2. These trends are analogous to those documented in some Oligocene to Pliocene alkaline mafic rocks in northwestern Colorado (Fig. 1), defined by increasing Sr isotope compositions (0.703–0.707) and near-constant 143Nd/144Nd (0.5122–0.5124) (Leat et al., 1988). The variable εNd(t) with increasing 87Sr/86Sr(i) in the majority of NVF and WSM samples (Fig. 9) indicates that lithospheric mantle source was heterogeneous or there was contamination of the melts by upper continental crust as argued by Leat et al. (1988) for the mafic rocks in northwestern Colorado (Fig. 1).

Bulk-rock Nd and Sr isotope data for NVF samples (Table 3) are similar to data reported in previous publications (Alibert et al., 1986; Roden et al., 1990; Nowell, 1993; Carlson and Nowell, 2001). Minette and transitional samples mostly plot within the field delineated for the Colorado Plateau subcontinental lithospheric mantle (SCLM on Fig. 9), which straddles BSE. Several of the NVF katungites have lower 87Sr/86Sr(i) ratios at ∼0.7039 and plot between the average field for OIB and PREMA (prevalent mantle of Zindler and Hart [1986]) and BSE (Rollinson, 1993).

Nd and Sr isotopic data for the majority of the WSM dike rocks, and Specie Mesa basalt, cluster around BSE within the field defined by NVF data. Pb isotopic data reported for some WSM rocks (Lake and Farmer, 2015) also overlap the range of Pb isotopic data for NVF minettes (Carlson and Nowell, 2001) and are significantly more radiogenic than the depleted mantle. Several WSM samples have higher 87Sr/86Sr(i) ratios (0.706710 and 0.708064), which could reflect minor contamination with 87Sr-rich upper continental crust (Fig. 9; Table 3). The Dulce samples (Table 3) are distinguished by near-constant 87Sr/86Sr(i) values (∼0.706) and εNd(t) values of –4.2 to –7.5. These εNd(t) values are more negative than those reported for samples of Dulce dikes in northern New Mexico (–2.6 to –3.3 in Gibson et al., 1993).

Roden et al. (1990) argued against crustal contamination as the cause of variation in 87Sr/86Sr(i) in NVF minettes at Buell Park. They cited the positive correlation of Ce/Yb ratios with increasing εSr(t) (noted by Alibert et al., 1986) as supporting evidence, since contamination with upper crustal rocks is expected to produce low Ce/Yb ratios. Roden et al. (1990) proposed instead that the variations in 87Sr/86Sr(i) ratios were likely caused by minor source heterogeneity, a hypothesis that was further supported by isotopic studies of Alibert et al. (1986), Nowell (1993), and Carlson and Nowell (2001).

The variations in Nd and Sr isotopic compositions of post-26 Ma rocks in the Four Corners region (Fig. 2) could reflect heterogeneity in melt sources or different degrees of crustal contamination. We contend that contamination of melts by upper continental crust was not significant. The high Sr concentrations for most of the mafic melts (900–2500, Table 2) would tend to buffer the effects of contamination by crustal rocks with low Sr concentrations. Our model-mixing line (Fig. 9) between the regional SCLM and average upper continental crust indicates that the Sr isotopic ratios of several WSM samples could be produced by contamination of mantle melts with less than ∼0.6% upper crust. The overall variations in the isotopic signatures, however, do not lend support to significant contamination with upper crust. Further evidence against such contamination is provided by 87Sr/86Sr(m) ratios of clinopyroxene (cpx) crystals in selected samples of NVF minette and transitional rocks (Fig. 10; Table 4). Similar core to rim 87Sr/86Sr(m) ratios in the crystals show that there were no notable changes in the isotopic compositions of the melts during crystallization, as would be expected if crustal assimilation happened.

The Sr and Nd isotope data of Dulce samples could indicate variable contamination of mantle melts by mafic lower crust with low time-integrated Sm/Nd (e.g., as documented by Dungam et al., 1986; Johnson et al., 1990; Johnson and Thompson, 1991). The model-mixing line between SCLM and lower “productive” crust shows it is plausible that the Dulce samples were contaminated with 10% to 45% lower crust and up to 0.5% upper crust (Fig. 9). A similar process of lower crustal contamination is argued for basalts on the southern edge of the Colorado Plateau (Alibert et al., 1986), and Pliocene basalts erupted on the Taos Plateau in New Mexico (Dungan et al., 1986). Alternatively, the Dulce magmas could have originated from melting of a mantle source with a composition similar to lower mafic crust.

Plots of Nd and Sr isotope data against SiO2, Th/Yb, and La/Nb (Fig. 11) are used to further assess the influence of different mechanisms (source heterogeneity, crustal contamination, and fractional crystallization) on magma compositions. The varied and negative εNd(t) and similar 87Sr/86Sr(i) with increasing SiO2 for the Dulce samples (Fig. 11; Table 3) are compatible with contamination of mantle magmas by lower crust with low time-integrated Sm/Nd or melting of mantle rocks with a similar composition. The slight increase in 87Sr/86Sr(i) and minor decrease in εNd (t) with increasing SiO2 for NVF samples are consistent with heterogeneity in mantle sources or minor contamination with upper crust. Some WSM samples plot on a trend of increasing 87Sr/86Sr(i) with near-constant SiO2, while other samples plot on a trend of similar 87Sr/86Sr(i) with increasing SiO2. Neither of these trends support contamination of melts with upper continental crust but could reflect compositional variations in the mantle source. Plots of 87Sr/86Sr(i) against Th/Yb and La/Nb (Fig. 11) also do not reveal trends indicative of crustal contamination. Since the Th/Yb and La/Nb ratios in mantle-derived rocks are sensitive to contamination of typical continental crustal with higher LREEs/high field strength elements (HFSEs) (Taylor and McLennan, 1985), a distinct relationship of higher 87Sr/86Sr(i)with increasing element ratios are expected. A few NVF samples with high 87Sr/86Sr(i) show slight increases in La/Nb that might reflect contamination, but the majority of the data do not. Navajo volcanic field katungites and WSM dike rocks with variable 87Sr/86Sr(i) and similar La/Nb and Th/Yb ratios could, however, be explained by compositional diversity in the mantle source.

Except for the Dulce dike samples, overall trends in our Nd and Sr isotopic data do not provide definitive support for contamination by continental crust, although we cannot exclude the possibility of minor assimilation during emplacement. We suggest that the trends are explained mostly by heterogeneity in the melt sources.

DISCUSSION

Oligocene to Pleistocene alkaline mantle melts were emplaced along the eastern margin of the Colorado Plateau in three separate fields (NVF, Dulce, and WSM) (Figs. 1 and 2) during intraplate regional extension, following subduction of the Farallon plate in the western United States (Fig. 12). Major- and trace-element data (Table 2) for rocks from each field reveal (Figs. 46) distinctions in magma compositions. This is contrasted by the similarity of Nd and Sr isotopic signatures, regardless of age. The chemical data also show that melts were enriched in LILE and LREE compared to depleted mid-oceanic ridge basalt (Fig. 5), indicating a long-term contribution of SCLM to mantle magmatism after 26 Ma.

Navajo volcanic field katungites have distinct chemical compositions relative to other mafic rocks in the Four Corners (e.g., lack of Ta-Nb depletion on Fig. 8). The hypothesis that the katungites formed by <2% melting of carbonated lithospheric peridotite with OIB affinities (Alibert et al., 1986; Roden et al., 1990; Nowell, 1993; Carlson and Nowell, 2001) is supported by the low SiO2, high CaO, high Zr/Hf, and low La/Ta (e.g., Rudnick et al., 1993). In addition, the chemical data for these rocks consistently cluster within or near the OIB field on different plots (Fig. 5). It is argued that asthenospheric mantle with OIB signatures was the source for some post–26 Ma mafic rocks on the southern Colorado Plateau (Group 1 rocks of Alibert et al., 1986), northwestern Colorado (Group 1 rocks of Leat et al., 1988; Gibson et al., 1992; Gibson et al., 1993), and Wasatch Plateau of Utah (Tingey et al., 1991) (Fig. 1). There is no conclusive evidence that the NVF katungites are derivatives from melting of enriched asthenospheric mantle (source of OIB). Compared to average OIB, the katungites have higher overall LILE and LREE and more radiogenic Sr isotopic ratios that plot close to bulk earth (Fig. 9). If the katungites were generated by melting of enriched asthenosphere, then the chemical and isotopic data indicate that the melts likely interacted with the lithospheric mantle to some degree (Fig. 9). The extremely low SiO2, high MgO, normative mineralogy, and lower overall 87Sr/86Sr ratios of these rocks argue that crustal contamination was not the cause of their elevated LILE and LREE relative to oceanic island basalts.

The compositions of other mafic rocks in the region (NVF minette-transitional rocks, Dulce dikes, and WSM dikes and flows) preclude an origin by direct melting of asthenospheric mantle as revealed by elevated LILE and LREE/HREE (Fig. 7), range of Nd and Sr isotopes (Figs. 9 and 11), and arc-like trace-element patterns (Fig. 8). The data are consistent with partial melting of a heterogeneous, metasomatized mantle, similar to alkaline magmas worldwide (Bell and Blenkinsop, 1989). The majority of NVF and WSM samples (Tables 13) are similar in composition to Oligocene to Pliocene mafic rocks in northwestern Colorado (Group 2 rocks of Leat et al., 1988; Johnson and Thompson, 1991; Gibson et al., 1993) and mica-bearing alkaline mafic rocks on the Wasatch Plateau (Tingey et al., 1991); these mafic rocks are also interpreted as products of melts from the subcontinental lithospheric mantle.

The Nd and Sr isotopic compositions of the 25–20 Ma Dulce samples are consistent with contamination of melts by lower continental crust. The isotopic signatures of these rocks are similar to Oligocene volcanic rocks in the Southern Rocky Mountain volcanic field (SRMVF on Fig. 1) (e.g., Lipman et al., 1976; Riciputi and Johnson, 1990; Johnson and Thompson, 1991; Riciputi, 1991; Riciputi et al., 1995; Lake, 2013; Lake and Farmer, 2015) that Lake and Farmer (2015) hypothesized were mixtures of mantle magmas with up to 40% mafic lower crust. Alternatively, the Dulce magmas could have formed by partial melting of lithospheric mantle with isotopic affinities similar to those documented for Proterozoic SCLM beneath southern Colorado (Johnson and Thompson, 1991). Leat et al. (1988) also argued that some mafic rocks in northwestern Colorado (Fig. 1) with low 143Nd/144Nd ratios (0.5124–0.5119) and near-constant 87Sr/86Sr ratios (∼0.704) formed by variable contamination of OIB magmas with mafic lower crust (e.g., Kay et al., 1978; Leeman et al., 1985; Dudas et al., 1987) or melting of Proterozoic lithospheric mantle with similar isotopic compositions. The progressive decrease in epsilon Nd values in Dulce rocks from northern New Mexico (–2.6 to –3.3; Gibson et al., 1993) into southern Colorado (–4.2 to –7.5) (Table 3) reveals either variable contamination with lower crustal rocks or regional compositional variations in the SCLM.

The NVF and WSM rocks (minettes and transitional) have similar Nd and Sr isotopic compositions (Fig. 9) but noticeably different concentrations of most major and trace elements (Figs. 46; Table 2), revealing a major shift in melt-chemical compositions over time (from 26 to 24 to 7–0.6 Ma). Relative to NVF rocks, those in the WSM are distinguished by lower MgO (Mg#), K2O, Rb, Zr, La/Lu, La/Ta, La/Nb, and Th/Yb but generally higher Al2O3, Na2O, and CaO. The fact that the Nd and Sr isotopic compositions of NVF and WSM are similar precludes crustal contamination as the cause of the distinct chemical compositions of WSM rocks.

It is argued that 270,000–400,000 km3 of mafic magma were extracted from the mantle during Oligocene volcanism in the SRMVF (Fig. 1) (Riciputi and Johnson, 1990; Riciputi, 1991; Riciputi et al., 1995; Farmer et al., 2008; Lake and Farmer, 2015). Johnson (1991) concluded that there was a significant contribution of SCLM melts (up to 50%) to the crustal magmas, causing an overall shift to more mafic compositions in the continental crust and local recycling of lower continental crust into the mantle (e.g., Farmer et al., 1991; Hildreth et al., 1991). The extensive melting in this region may have produced a mantle-melt column at least 20–25 km thick (Farmer et al., 2008) in which the more fusible constituents of the SCLM were extracted, leaving it “depleted” in some elements. This idea is supported by experimental melting of phlogopite-bearing lherzolite over 1000° to 1300 °C (Condamine and Médard, 2014): this study is applicable since the SCLM on the eastern edge of the Colorado Plateau contains phlogopite-bearing garnet lherzolite (e.g., McGetchin and Silver, 1970; McGetchin, and Besancon, 1973; Hunter and Smith, 1981; Smith, 1987). Condamine and Médard (2014) demonstrate that higher degrees of melting of phlogopite-bearing mantle could produce potassic melts with increasingly higher MgO (2–12 wt%), FeO (1–7 wt%), and CaO (3–10 wt%) but progressively lower concentrations of Na2O (5–1 wt%), K2O (5–3 wt%), and Al2O3 (20–14 wt%). This process could leave the mantle source depleted in all of these elements but to different degrees. The differences in major- and trace-element concentrations in the post-7 Ma WSM dike rocks might thus reflect chemical changes in the SCLM as a result of voluminous magma production in the SRVMF from 35 to 26 Ma. Variations in the “fertility” of Proterozoic SCLM from partial melting events from 1.8 to 1.4 Ga could have influenced the higher degrees of melting beneath the WSM (Lee et al., 2001; Gilbert et al., 2007). It is also feasible that lithospheric mantle under the western San Juan Mountains inherited a distinct chemical composition during assembly and alteration of Proterozoic arc terranes as proposed by Gilbert et al. (2007). They concluded that a northeast-trending zone in the lithosphere that extends beneath western Colorado represents a ca. 1.8 Ga subduction-terrane boundary along which ancient mantle rocks were “serpentinized.” Alteration of the Proterozoic mantle could have modified the chemical composition of the SCLM in certain zones and may have influenced melt compositions in the late Cenozoic.

Another possible explanation for the differences in chemical composition of NVF and WSM rocks is that a relatively “homogeneous” mantle source with similar isotopic compositions (Fig. 9) underwent different degrees of melting. The melting experiments of Condamine and Médard (2014) produced mafic melts with similar variations in major elements as those in NVF and WSM rocks. In this case, however, it is difficult to reconcile the relatively high Mg# and Ni (and comparable Cr) for NVF rocks, given that they formed by 0.5–1.5% melting of metasomatized SCLM under the Colorado Plateau (e.g., Alibert et al., 1986; Nowell, 1993). The lack of trace-element data for mantle xenoliths on the Colorado Plateau at this time, however, inhibits a test of the idea that different degrees of partial melting caused the distinct compositions of NVF and WSM rocks.

We contend that Oligocene to Pleistocene mafic rocks in the Four Corners region crystallized from magmas that were derived from melting of metasomatized and heterogeneous subcontinental lithospheric mantle enriched in LILE and LREE but with broadly similar Nd and Sr isotopic compositions. The chemical and isotopic data argue that the SCLM beneath the Four Corners region has traits similar to the lithospheric mantle in subduction zones. Navajo volcanic field katungites hint at contributions from the enriched asthenospheric mantle, but this prospect is not clearly supported by the data. The regional SCLM was built by assembly of Proterozoic arcs from 1.8 to 1.4 Ga (e.g., Karlstrom et al., 2005; Gonzales and Van Schmus, 2007) and further modified by Farallon hydration and metasomatism in the Laramide from 75 to 30 Ma (e.g., Smith 1979, 1995; Alibert et al., 1986; Broadhurst, 1986; Roden et al., 1990; Wendlandt et al., 1993, 1996; Usui et al., 2002, 2003; Smith et al., 2004; Smith, 2010; Schulze et al., 2015). Mantle xenoliths in NVF rocks (Smith, 2004) have constrained carbonate-silicate mantle sources of different ages beneath the Colorado Plateau, providing further evidence of the composite magmatic-tectonic history for the regional SCLM. Epsilon Nd values for NVF and WSM rocks lie on the Nd-evolution paths for the most primitive mafic rocks in the Paleoproterozoic complex (Gonzales and Van Schmus, 2007) in southwestern Colorado, consistent with the idea that metasomatized Proterozoic SCLM contributed to the generation of Cenozoic mantle magmas. It is thus conceivable, and reasonable, that the chemical heterogeneity in the subcontinental lithospheric was established during the Proterozoic.

CONCLUSIONS

The different groups of Oligocene to Pleistocene mafic rocks in the Four Corners area (Fig. 2) preserve distinct geochemical variations in both space and time (Figs. 2 and 12). Melt production in any given zone of magmatism (Fig. 2) persisted for 2–7 m.y. and created generations of magmas with broadly similar isotopic and geochemical signatures, reflecting the prolonged similarity in source and process. We argue that chemical variations in the mantle source (SCLM) ± crustal contamination were the dominant influence on the variations in isotopic and geochemical compositions of mafic rocks produced after 26 Ma. The close spatial and temporal relationship of mantle melts emplaced in the region with postcaldera intermediate to felsic plutons from 26 to 3 Ma (Gonzales, 2015) also give insight into the enduring influence of mantle melts on regional trends of crustal magmatism in the transition from convergent tectonics to continental rifting.

Navajo volcanic field rocks were generated by small degrees of melting (<2%) (Nowell, 1993) of SCLM with little or no crustal contamination (revealed by Sr isotopic compositions of cpx crystals; Table 4). The distinct chemical and isotopic signatures of melts generated from 25 to 20 Ma beneath the Dulce field were influenced either by melting of lithospheric mantle with low 143Nd/144Nd ratios or mixing of mantle melts with up to 40% lower crustal rocks (Fig. 9). Mafic dikes and flows in the WSM crystallized over ca. 7 Ma from magmas with Nd and Sr isotopic signatures similar to NVF melts but with some distinct differences in major- and trace-element compositions. We argue that the WSM magmas retained a persistent “memory” of chemical modification in the mantle that was influenced by extensive melting (from 35 to 26 Ma) that carried large batches of mafic magmas into the crust. Chemical modification of the lithospheric mantle in this zone could also have been influenced by magmatism and alteration related to subduction processes in the Proterozoic.

Our work contributes to a body of regional research (Fig. 1) that reveals the importance of mantle heterogeneity in the production of alkaline mafic rocks after 26 Ma on the Colorado Plateau (e.g., Alibert, 1994; Alibert et al., 1986; Roden et al., 1990; Tingey et al., 1991; Nowell, 1993; Carlson and Nowell, 2001; Smith et al., 2004; Bailley, 2010) and adjacent Southern Rocky Mountains (e.g., Leat et al., 1988; Johnson and Thompson, 1991; Gibson et al., 1992). It also allows consideration of the role of heterogeneous, ancient subcontinental lithosphere in the production of continental alkaline mafic rocks worldwide.

We thank the Navajo Division of Natural Resources for providing access to the Navajo Nation to collect samples. A special thanks to Arnold Clifford for helping us navigate to locations in the Navajo volcanic field. We also want to thank Michael Roden, Eric Christiansen, and an anonymous reviewer for their detailed and critical reviews, which improved the quality of this manuscript. This project was funded in part by the National Science Foundation grant 0911290 and several Faculty Development Grants at Fort Lewis College.

APPENDIX. ANALYTICAL METHODS AND PARAMETERS

Major- and Trace-Element Analyses

Only unaltered samples or samples with very minor alteration (determined by petrographic analyses) were used for chemical analyses. Whole-rock samples were powdered and analyzed by either Activation Laboratories (methods 4B and 4B2, http://www.actlabs.com/files/Canada_2015.pdf) or SGS Laboratories (http://www.sgs.com). FeO in all samples was determined by the titration method.

SGS Laboratories

Major elements were determined by inductively coupled plasma–atomic emission spectroscopy (ICP-AES) using a lithium-metaborate/tetraborate fusion. Trace elements were determined by inductively coupled plasma–mass spectrometry (ICP-MS) using a sodium peroxide fusion.

In addition to standards analyzed by SGS, internal University of Texas standard (BBB) and a U.S. Geological Survey (USGS) standard (BHVO-2) were analyzed, as well as replicate samples. Compared to mean values of the standards, the SGS values were all within 2-sigma levels of error, and most were within 1-sigma levels.

Activation Laboratories

The following description was taken directly from the Activation Laboratories Web site. Samples for major-element analyses are prepared and analyzed in a batch system. Each batch contains a method reagent blank, certified reference material and 17% replicates.

Prior to sample fusion, the loss on ignition (LOI), which includes H2O+, CO2, S, and other volatiles, was determined from the weight loss after roasting the sample at 1050 °C for 2 h. Samples are mixed with a flux of lithium metaborate and lithium tetraborate and fused in an induction furnace. The molten melt is immediately poured into a solution of 5% nitric acid containing an internal standard and mixed continuously until completely dissolved (∼30 min). The samples are run for major oxides and selected trace elements (Code 4B) on a combination simultaneous/sequential Thermo Jarrell-Ash ENVIRO II ICP or a Varian Vista 735 ICP. Calibration is performed using seven prepared USGS and CANMET certified reference materials. One of the seven standards is used during the analysis for every group of ten samples.

For trace-element concentrations, samples fused under code 4B2 are diluted and analyzed by Perkin Elmer Sciex ELAN 6000, 6100, or 9000 ICP/MS. Three blanks and five controls (three before the sample group and two after) are analyzed per group of samples. Duplicates are fused and analyzed every 15 samples. Instrument is recalibrated every 40 samples. Repeated measurements of standards indicate that analytical errors are at the 2 sigma level (95%) (A. Hoffman, Activation Laboratories Ltd., 2016, personal commun.).

Whole-Rock Sr and Nd Isotopic Analyses

  • (1) Unaltered rock samples and an internal standard from the University of Texas were analyzed for whole-rock Nd and Sr isotopes at the Thermal Ionization Mass Spectrometry Laboratory at the University of Colorado. Samples were dissolved in acid and separated using column chromatography, following the methods of Pin et al. (1994). The 87Sr/86Sr ratios were measured using four-collector static mode on a Finnigan-MAT six-collector solid source mass spectrometer.

  • (2) Total procedural blanks averaged 100 pg for Nd and ∼1 ng for Sr.

  • (3) 87Sr/86Sr ratios were measured using the four-collector static mode. The present-day measured 87Sr/86Sr ratios were recalculated to determine initial ratios using reported radiometric ages (Gonzales, 2015). Sixteen measurements of SRM-987 for samples analyzed by Lake (2013) yielded a mean value for 87Sr/86Sr of 0.71026 ± 0.00002 (2σ), and eight measurements of SRM-987 (Gonzales, this paper) yielded a mean 87Sr/86Sr = 0.71028 ± 0.00002 (2σ).

  • (4) 143Nd/144Nd analyses were done in a three-collector dynamic mode. Twelve measurements of the La Jolla Nd standard yielded a mean 143Nd/144Nd of 0.511832 ± 8 (2σ). Measured 143Nd/144Nd ratios were normalized to 146Nd/144Nd = 0.7219.

Sr Isotopic Analyses of Clinopyroxene Single Crystals

  • (1) All isotopic analyses were performed on a NewWave UP213 LA unit coupled to a NuPlasma HR MC-ICP-MS instrument, housed in the Department of Geological Sciences, University of Cape Town, South Africa. The instrument is a double-focusing ICP-MS fitted with 12 Faraday detectors, three discrete dynode ion counters, and one channeltron ion counter in a fixed-position collector array. Unique variable zoom optics manipulate the ion beam to achieve coincidence and alignment of ion beams of interest. All data were collected on five of the 12 Faraday detectors. Regular Faraday amplifier-gain calibrations ensured relative stability between detectors.

  • (2) Instrumental mass fractionation was corrected using the exponential law and a fractionation factor based on the measured 86Sr/88Sr ratio and the accepted 86Sr/88Sr value of 0.1194.

  • (3) Potential surface contamination was removed by rapidly sweeping the laser along the path of interest (250 mm spot size for analyses, 750 mm line length, 10 Hz repetition rate, 50 μm/s translation rate, 0.25 mJ energy, and 0.5 J/cm2 fluency) prior to the analysis with a narrower beam, higher energy, and slower speed (150–200 mm spot size, 750 mm line length, 20 Hz repetition rate, 5 mm/s translation rate, ±1.35 mJ energy, and ±4.35 J/cm2 fluency).

  • (4) An in-house standard clinopyroxene (JJG1424 cpx) was analyzed during analyses. Two separate runs generated an average value of 0.7048 ± 0.0004 (n = 15) and 0.7049 ± 0.0001 (n = 4). The accepted value determined for this standard by solution multicollector–inductively coupled plasma–mass spectrometry analysis is 0.70495.

  • (5) Detailed descriptions of the analytical methods and parameters are provided in Copeland et al. (2008), Radloff et al. (2010), and Le Roux et al. (2014).