Mantle Sources and Geochemical Evolution of the Picture Gorge Basalt, Columbia River Basalt Group

Abstract

The Columbia River Basalt Group (CRBG) is the youngest continental flood basalt province, proposed to be sourced from the deep-seated plume that currently resides underneath Yellowstone National Park. If so, the earliest erupted basalts from this province, such as those in the Picture Gorge Basalt (PGB), aid in understanding and modeling plume impingement and the subsequent evolution of basaltic volcanism. Using geochemical and isotopic data, this study explores potential mantle sources and magma evolution of the PGB. Long known geochemical signatures of the PGB include overall large ion lithophile element (LILE) enrichment and relative depletion of high field strength elements (HFSE) typical of other CRBG main-phase units. Basaltic samples of the PGB have ⁸⁷Sr/⁸⁶Sr ratios on the low end of the range displayed by other CRBG lavas and mantle-like δ¹⁸O values. The relatively strong enrichment of LILE and depletion of HFSE coupled with depleted isotopic signatures suggest a metasomatized upper mantle as the most likely magmatic source for the PGB. Previous geochemical modeling of the PGB utilized the composition of two high-MgO primitive dikes exposed in the northern portion of the Monument Dike swarm as parental melt. However, fractionation of these dike compositions cannot generate the compositional variability illustrated by basaltic lavas and dikes of the PGB. This study identifies a second potential parental PGB composition best represented by basaltic flows in the extended spatial distribution of the PGB. This composition also better reflects the lowest stratigraphic flows identified in the previously mapped extent of the PGB. Age data reveal that PGB lavas erupted first and throughout eruptions of main-phase CRBG units (Steens, Imnaha, Grande Ronde Basalt). Combining geochemical signals with these age data indicates cyclical patterns in the amounts of contributing mantle components. Eruption of PGB material occurred in two pulses, demonstrated by a ~0.4 Ma temporal gap in reported ages, 16.62 to 16.23 Ma. Coupling ages with observed geochemical signals, including relative elemental abundances of LILE, indicates increased influence of a more primitive, potentially plume-like source with time.

1. Introduction

The Columbia River Basalt Group (CRBG) of the Pacific Northwest, USA, is the world’s youngest and least volumetric flood basalt (Figure 1). The CRBG can be subdivided into four main-phase formations, which include the Steens, Imnaha, Grande Ronde, and Picture Gorge Basalts (PGB) [1,2]. Of these formations, the PGB is the least volumetrically extensive and exhibits geochemical signatures that suggest that it represents a separate magmatic system potentially resulting from lithospheric thinning in a back-arc tectonic setting [3,4,5,6]. The PGB is the earliest main-phase unit of the CRBG [7], followed by the Steens Basalt [8]. Thus, a detailed reexamination of PGB geochemistry and magma petrogenesis provides insight into mantle source components and their temporal and/or spatial boundaries. Examining the spatial, temporal, and geochemical similarities between PGB and other CRBG main-phase units like the Steens Basalt, this study assesses the mantle sources and their respective contributions to the petrogenesis of the PGB.

Through the decades of research on the CRBG flood lavas, numerous origins and source materials have been proposed by many workers to contribute to the geochemical diversity exhibited by the basalts [3,4,5,10,11,12,13,14,15]. These source components include a mantle plume, a subcontinental lithospheric mantle that is variably metasomatized by prior subduction zone processes, and various crustal lithologies within the North American plate. Many workers agree that a mantle plume is ultimately responsible for the CRBG [16] and the Snake River Plain–Yellowstone hotspot track [17] to the east, although different non-plume models have been proposed. Examples of some of these models invoke a rapid increase in back-arc extension caused by steepening of the subducting oceanic plate to the west [18,19], a northward opening tear in the descending slab [20], peeling off of the Farallon plate [13], and a similar model of peeling off of the Farallon plate with the added component of lithospheric foundering at the root of the Wallowa Mountains in northeastern Oregon [12].

1.1. Geochemical Relationships between Main-Phase CRBG Formations

Endmember compositions within main-phase formations of the CRBG reflect the contribution of multiple mantle source components, as interpreted by trace element and isotopic data [3,4,5,10,13,21]. In isotopic space, the Grande Ronde Basalts and Imnaha Basalts define their own trend, while the PGB and the Steens Basalt share other distinct geochemical and isotopic traits [3,5,6,21]. Partial melting processes within the various source regions likely also contribute to some of the observed similarities and differences in trace element contents [21]. The Grande Ronde Basalts form a trend emanating from the Imnaha Basalts in isotopic space and could represent crustally contaminated Imnaha Basalts [3,5,10]. The “endmember” identified within the Imnaha Basalts was originally interpreted as re-enriched mantle, possibly the result of added sediment to a depleted mantle component [3]. This mantle source was defined as “C2” by Carlson (1984) [3] and later named the “Imnaha Component” by Wolff et al. (2008) [5], who suggested it to best represent the plume source, although there are no identified OIB with Pb isotopic compositions like C2.

The Steens Basalt is modeled as the result of a depleted mantle component (C1 of Carlson, 1984 [3]) mixing with the Imnaha component [18] or plume source [5]. Within lavas of the Steens Basalt, the proportion of depleted mantle decreases with time, moving stratigraphically upwards through the lower to the upper Steens Basalt [15]. Lower Steens lavas reflect a higher contribution of a depleted mantle component than do upper Steens lavas, suggesting that the proportion of the Imnaha component in the Steens Basalt increases with time. This is evidenced in isotopic space by ⁸⁷Sr/⁸⁶Sr versus ²⁰⁶Pb/²⁰⁴Pb, where the upper Steens Basalt plots closer to the defined Imnaha Component than the lower Steens Basalt [5,6]. This observation also applies to the lower Imnaha Basalts; the depleted mantle component is most prevalent in the earliest erupted Imnaha lavas [2,6,15].

The PGB shares the most geochemical characteristics with the Steens Basalt and the American Bar subtype of the Imnaha Basalts (Figure 2). Similar to the Steens Basalt, PGB magmas are interpreted as being derived from a depleted mantle source, but one that was significantly influenced by prior subduction [4,21]. The observed enrichment of LILE relative to HFSE is one of the more characteristic geochemical signatures in PGB lavas and dikes. This feature has been interpreted as evidence for a fluid-fluxed mantle component (Figure 3) [3,4,5,6,10,21,22,23]. In the context of the entire flood basalt province, it has been suggested that the PGB reflects a subduction overprinted back-arc mantle less influenced by plume magmatism compared to the rest of the CRBG [5].

1.2. Petrogenetic Significance of Picture Gorge Basalt Stratigraphy

The PGB is best exemplified by the type locality at Picture Gorge, where 18 basaltic lava flows are exposed in a continuous stratigraphic section. These lavas erupted from north-northwest trending dikes of the Monument Dike swarm [2,4,23] and can be subdivided based on plagioclase modal abundance; major element concentrations, specifically TiO₂ and MgO; and magnetic orientation (Figure 4) [2,22,23,25,26,27,28]. The previous stratigraphic terminology used to describe these PGB subdivisions is inconsistent and resulted in application of stratigraphic nomenclature that does not abide by the Stratigraphic Code of North America (North American Commission on Stratigraphic Nomenclature, 2005) [29]. Here, we use the term “subunit” to reference these three previously mentioned subdivisions. the Twickenham, Monument Mountain, and Dayville Basalts, and “chemical type” to reference the further subdivisions of these subunits into 17 groupings of lava flows in addition to a distinct composition classified as high-MgO dikes (referred to as “members” by Bailey, 1989) [23].

The high-MgO dikes, identified as potential parental material [23], are enriched in Ba, K, and Rb relative to elements of similar incompatibility, suggesting some crustal contamination or slab-derived fluids fluxing the source. This enrichment in large ion lithophile elements (LILE) is also observed in the Twickenham Basalt subunit and in the Monument Mountain Basalt subunit but to a lesser degree [23]. A decrease in eruptive activity followed extrusion of Monument Mountain lavas, as the subsequent Dayville Basalt is characterized by lava flows that are both spatially limited and compositionally diverse. Overall, the Twickenham and Monument Mountain Basalts exhibit minor compositional variation, which cannot be the result of closed-system fractionation [23], and the younger Dayville Basalt forms its own evolutionary trend characterized by lava flows that are grouped into ten chemical types that are highly variable in their geochemical signatures (Figure 4).

2. Methods

2.1. Major and Trace Element Concentrations

Major and trace element data were acquired at the Peter Hooper GeoAnalytical Lab at Washington State University (WSU) using a Thermo-ARL Advant’XP automated X-ray fluorescence spectrometer (XRF) (Waltham, MA, USA) and an Agilent 7700 inductively coupled plasma mass spectrometer (ICP-MS) (Santa Clara, CA, USA). For sample preparation, basaltic samples were crushed into chips using the Braun Chipmunk at WSU and following sample preparation as outlined in Johnson et al. (1999) [33]. Then, chips exhibiting weathered surfaces were removed to reduce the influence of alteration on the resulting data. Rock chips were further crushed into a powder in a tungsten carbide swing mill. The sample powder was then combined with dilithium tetraborate (Li₂B₄O₇) in a ratio of 2:1, dilithium tetraborate to sample. The sample was fused into a bead in an oven at a temperature of 1000 °C. The beads were reground into powder, and one gram of powder was separated in order to make the ICP-MS bead. Two beads per sample were then made, again at 1000 °C, one for XRF and one for ICP-MS analysis. The XRF bead was analyzed in the XRF instrument, and the ICP-MS bead was dissolved for final analysis. All major and trace element geochemical data for our PGB samples are compiled in Table S2. and Mg# is plotted against a variety of these geochemical parameter for visualization through geochemical evolution in Figure S4.

2.2. Radiogenic Isotope Analyses

Sample preparation and analysis for Sr, Nd, Pb, and Hf were performed at the Carnegie Institution for Science, Earth and Planets Laboratory (EPL) in Washington D.C. Sample preparation and analyses were conducted in 2017 and 2018 during two separate laboratory visits. During both visits, over 10 g of fresh rock chips from each sample were powdered in an agate ceramic bowl to avoid contamination of metals (i.e., Ta from WC grinding bowls). Once the sample was powdered, approximately 50 mg was weighed and dissolved in ~2 mL of concentrated HNO₃ and 4 mL concentrated HF. The mixture in a Savillex Teflon vial was capped and heated on a 90 °C hotplate for ~12–18 h. Samples were heated longer (i.e., 18 h) during the 2017 sample preparation as the result of a laboratory closure. Each sample solution was then evaporated, redissolved twice in 1 mL of concentrated HNO₃, and dissolved again in 4 mL of 4M HCl. Finally, samples were evaporated, dried two more times with 0.5 mL 0.5 N HBr, and redissolved in 3 mL of 0.5 N HBr. Following these acid digestions, elements of interest (Sr, Nd, Hf, and Pb) were separated and purified via column chromatography following the procedures described below.

Lead was separated first by loading the samples (in 3 mL 0.5 N HBr) on 0.04 mL Teflon columns with AG1-X8 100–200 mesh anion exchange resin following the methodology of Carlson et al. (2006) [34]. Following elution of the majority of the elements using 0.5 N HBr, the Pb fraction was collected in 1.5 mL HNO₃, dried on a hotplate, and, after adding 1 mL 0.5 N HBr, loaded onto the columns a second time for purification. The final Pb fraction was collected in 1.5 mL 0.5 N HNO₃. The remaining sample in 0.5 N HBr was then dried, oxidized with 0.5 mL of concentrated HNO₃, evaporated again, redissolved in 5 mL of 0.1 N HF-1 N HCl, and heated for approximately one hour at 80 °C. Next, samples were centrifuged to remove any precipitates and loaded on AG 50 W × 8 resin (200–400 mesh), 20 cm height × 6 mm ID.

The high field strength elements, including hafnium, were separated next by collecting the initial solution and an additional 5 mL of 0.1 N HF-1 N HCl following a procedure modified from Patchett and Tatsumoto (1980) [35]. The eluant was then switched to 2.5 N HCl with 30 mL eluted until the Sr cut was collected in an additional 14 mL of 2.5 N HCl. Switching to 4 N HCl, the light REE was collected in 16 mL after eluting and discarding 12 mL.

Hafnium was purified on 0.6 × 10 cm columns using Eichrom LN 100–150 μm resin. The sample was loaded in 5 mL of 2.5 N HCl and rinsed with 10 mL 2.5 N HCl, 10 mL 6 N HCl, 4 mL of Milli-Q (MQ) water, 60 mL of 0.09 N citric acid-0.45 N HNO₃-1 wt. % H₂O₂, 5 mL of 0.09 N citric acid-0.45 N HNO₃, and 20 mL of 6 N HCL-0.06 N HF. The final Hf sample was collected in 8 mL of 6 N HCl-0.4N HF. Strontium was purified on 0.25 mL Teflon columns with Eichrom Sr resin. The purified Sr sample was collected in 2 mL of 0.05M HNO₃. Neodymium was separated from the other REE following the procedure of Garcon et al. (2018) [36] on columns with 10–50 μm LN spec resin and collected in 9 mL of 0.173N HCl.

Isotopic compositions for Sr and Nd were measured via thermal ionization mass spectrometry (TIMS) on the Thermo-Fisher Triton. Sr was run on single Re filaments loaded in a Ta₂O₅ powder in dilute H₃PO₄-HF. Each ⁸⁷Sr/⁸⁶Sr ratio reported in Table 2 is the average of 279–290 ratios obtained using 8 s integrations with a signal size from 5 to 20 × 10⁻¹¹ A of ⁸⁸Sr. Mass fractionation was corrected to ⁸⁶Sr/⁸⁸Sr = 0.1194 using exponential mass dependency and reported relative to a value of 0.71025 for the NBS987 Sr standard. The Sr standard NBS987 was measured multiple times during the 2017 and 2018 analytical sessions and averaged 0.710240 ± 0.0000027 (n = 3) in 2017 and 0.710249 ± 0.000003 (n = 3) in 2018, where uncertainty is 2σ. For all but one sample, the individual run uncertainty on ⁸⁷Sr/⁸⁶Sr was 0.000003. Uncertainties for all samples are reported in Figure S5.

Each ¹⁴³Nd/¹⁴⁴Nd ratio reported in Table 2 is the average of 199–203 ratios (in 2017) or 233–279 (in 2018) statically obtained using 8 s integrations with a signal size from 1–3 × 10⁻¹¹ A of ¹⁴⁴Nd. Mass fractionation was corrected using exponential mass dependency and ¹⁴⁶Nd/¹⁴⁴Nd = 0.7219. The Nd standard JNdi was measured multiple times during the 2017 analytical sessions and averaged 0.5121117 ± 0.0000023 (n = 3), where uncertainty is 2σ. During the 2018 analytical sessions, the data for JNdi averaged 0.5120940 ± 0.0000017 (n = 3), where uncertainty is 2σ. The difference in values measured for the standard reflects the aging of the Faraday cups, with the lower value obtained on fresher cups. Data for the samples are adjusted to a JNdi value of 0.512115 [37] based on the average value determined for JNdi during the separate analytical sessions.

Isotopic compositions for Pb and Hf were measured on the Nu HR multicollector ICP-MS and Nu Plasma II multicollector ICP-MS. Each Pb ratio reported in Table 2 is the average of 41–50 ratios. Samples that yielded ²⁰⁸Pb signals below 100 mv were not reported. Mass fractionation during the 2017 analyses was corrected to ²⁰⁵Tl/²⁰³Tl = 2.417 and then adjusted for the difference between the NBS981 Pb standards measured every 4 samples and the reported value for the standard [38]. Mass fractionation during the 2018 analyses was corrected to ²⁰⁵Tl/²⁰³Tl = 2.410 and then adjusted for the difference between the NBS981 Pb standards measured every 4 samples and the reported value for the standard [38]. Average values for NBS981 obtained during 2017 analytical sessions for n = 7 measurements were ²⁰⁸Pb/²⁰⁴Pb = 36.714 ± 0.006, ²⁰⁷Pb/²⁰⁴Pb = 15.497 ± 0.003, and ²⁰⁶Pb/²⁰⁴Pb = 16.941 ± 0.003, where uncertainty is 2σ. Average values for NBS981 obtained during 2018 analytical sessions for n = 6 measurements were ²⁰⁸Pb/²⁰⁴Pb = 36.703 ± 0.028, ²⁰⁷Pb/²⁰⁴Pb = 15.495 ± 0.011, and ²⁰⁶Pb/²⁰⁴Pb = 16.938 ± 0.012, where uncertainty is 2σ. Uncertainties for all samples are provided in Figure S5.

Hafnium isotopic compositions were measured by introducing samples into the plasma in 0.08N HNO₃-0.06 N HF via a Teflon nebulizer. Each ratio reported in Table 2 is the average of 52–80 ratios, and the signal size of ¹⁷⁸Hf ranged from 3 to 8 × 10⁻¹¹ A. Mass fractionation was corrected to ¹⁷⁹Hf/¹⁷⁷Hf = 0.7325. The average of JMC475 obtained during the 2017 analytical sessions for n = 3 measurements was ¹⁷⁶Hf/¹⁷⁷Hf = 0.282148 ± 0.000002 (2σ). The average of JMC475 obtained during the 2018 analytical sessions for n = 13 measurements was ¹⁷⁶Hf/¹⁷⁷Hf = 0.282144 ± 0.000002 (2σ). Reported ¹⁷⁶Hf/¹⁷⁷Hf ratios were adjusted based on the average value determined for JNdi during the separate analytical sessions corrected to a JMC475 value of 0.282160. Values for ε_Hf and ε_Nd are calculated using the average chondrite parameters from Bouvier et al. (2008) [39] of ¹⁷⁶Hf/¹⁷⁷Hf = 0.282785 and ¹⁴³Nd/¹⁴⁴Nd = 0.512630. All Sr, Nd, Hf, Pb isotope uncertainty and isobaric interferences for our PGB samples are compiled in Table S3.

2.3. Oxygen Isotope Analyses

In order to better quantify crustal contamination, δ¹⁸O values are revisited with modern techniques applied to both plagioclase mineral separates and basaltic groundmass that was exposed to multiple acid leachings to remove excess alteration. Basaltic groundmass and plagioclase separates were analyzed from 16 basaltic flows. The samples were crushed, sieved to grain sizes between 180 and 150 or 250 µm, magnetically separated via Franz, and leached in acid. Samples were prepared at Oregon State University following procedures for ⁴⁰Ar/³⁹Ar dating, and the remaining groundmass and mineral separates were used for oxygen isotope analysis. Oxygen isotopic compositions were then measured at the University of Oregon via an integrated CO₂ laser fluorination MAT 253 isotope ratio mass spectrometer (IRMS) system [40,41]. The range of material analyzed from each sample was 1.2 to 1.6 mg. During the analytical session, the Gore Mountain Garnet standard (UOG, accepted δ¹⁸O = 6.52‰) was measured multiple times (n = 7) to correct for instrumental drift [41]. Analytical precision of UOG analyses from the session was 0.06% (2σ) and is reported in Table S4.

2.4. MELTS Calculations

MELTS calculations were conducted using version rhyoliteMELTS v1.2x with two distinct primitive compositions identified as PGB [5,7,23]. This version of MELTS is intended for modeling crystallization over 0–20 kb using bulk compositions that correspond to natural magmas; the melt is interpreted to contain dissolved H₂O, but the magma is not quartz saturated [42,43]. Compositions were modeled at 6 different pressures (2, 3, 4, 7, 8, and 9 kb) with 0.1 and 0.5 wt. % water. For each experiment, the pressure stayed constant, and the temperature decreased to 100 degrees C below the wet liquidus temperature, as calculated by MELTS. Utilized partition coefficients are compiled in Table S1, and the results of each individual model are summarized in Table S5.

3. Results

3.1. PGB Composition

3.1.1. Major and Trace Element Compositions

Newly identified basaltic lavas and dikes of PGB composition extend across a wide swath of eastern Oregon. Along with PGB samples from the type locality and the original distribution area described by Bailey (1989) [23], PGB samples from these newly identified areas have relatively primitive to evolved (MgO ~4–9 wt. %) compositions with flat to slightly negatively sloping patterns on mantle-normalized incompatible trace element diagrams, but with troughs (HFSE) and spikes (Ba, K, Pb) typically associated with volcanic arc-related lavas of the Pacific Northwest (Figure 3). Samples of PGB composition exhibit relative depletions in many incompatible elements (i.e., La, Nb, Rb, Hf, Th), overall elevated large ion lithophile element (LILE) to high field strength element (HFSE) ratios, and less LREE enrichment on a mantle-normalized trace element diagram (Figure 2 and Figure 3) relative to other CRBG units.

To gauge fractionation and assess the geochemical variation of the PGB, major and trace elements are plotted against Mg# (molar Mg# = (Mg/(Mg + Fe) × 100)) (Figures S1–S3). Observed trends in both major and trace element space define broad bands on most such geochemical plots. Major element concentrations exhibit larger variability between Mg# 65–45 (i.e., Al₂O₃, MnO, P₂O₅, and FeO*), and both FeO* and MgO exhibit a slight inflection at Mg# of 45, which could be due to crystallization of ilmenite and magnetite.

At the high Mg# end of the compositional arrays, several elements (e.g., Si, Al, Fe, Ca, Ni, Cu, LREE, Pb), and Sr in particular, show a wide range in concentration, suggesting that the primary magmas for the PGB are compositionally variable (Figures S1–S3). Some basalts are characterized by Mg# > 60 and could represent potential parental magmas. The high Mg# PGB samples are best represented by samples of the high-MgO dikes and lava flows exposed along Rattlesnake Road in the southeastern region of the PGB extended distribution (Figure 1,

Table 1. Summary Table of Picture Gorge Basalt Primitive Endmember Samples.

Sample Name	Mg#	87Sr/86Sr	143Nd/144Nd	Age (Ma)
high-MgO dike (HMD) Endmember
MMB D69A	65.4	0.703400	0.512980	no age data
MMB D69B	64.5	0.703066	0.513012
MMB D70A *	66.1	0.703400	0.512990
MMB D70B	64.1	-	-
Rattlesnake Road (RR) Endmember
DM255B	57.4	0.703757	0.512905	15.76
CAH15-007	64.4	0.703691	0.512889	16.22
CAH17-241A *^	59.6	-	-	15.5–16.16
CAH17-241B ^	59.5	-	-
CAH17-242A ^	57.8	0.703565	0.512925
MS-15-18ba ^	58.8	-	-

^{^} Samples collected along Rattlesnake Road (see Figure 1 for location). * Samples used for MELTS modeling.

). Although other samples could have been chosen, samples of Rattlesnake Road were selected as representatives of the parental magma because there are four analyses from this location and their classification as PGB has been further confirmed by the Oregon Department of Geology and Mineral Industries (DOGAMI) (R. A. Houston et al., 2017) [44]. Overall, these high Mg# samples reflect two groups of primitive compositions observed within the PGB, referred to as the high-MgO dike (HMD) and the Rattlesnake Road (RR) groups (Table 1).

3.1.2. Radiogenic and Oxygen Isotope Compositions

Isotopically, PBG lavas have the lowest ⁸⁷Sr/⁸⁶Sr ratios of all main-phase CRBG units, and the most primitive PGB dikes contain the lowest ⁸⁷Sr/⁸⁶Sr ratios observed within PGB [45]. New radiogenic isotope data are reported for 16 sampled basaltic lavas and dikes of the PGB. Samples from within and outside of the previously mapped extent have ⁸⁷Sr/⁸⁶Sr ratios from 0.70341 to 0.70376 and ¹⁴³Nd/¹⁴⁴Nd ratios from 0.51299 to 0.51289 (Figure 5). The only exceptions are samples of PGB lavas from Pole Creek in the Malheur Gorge. These samples have higher ⁸⁷Sr/⁸⁶Sr ratios > 0.7039 and lower ¹⁴³Nd/¹⁴⁴Nd ratios < 0.5129 than the rest of the PGB but resemble the PGB in major and trace element concentrations.

Combining isotopic data with trace element data, samples of the identified primitive groups (HMD and RR) are evaluated to see if they share isotopic similarities. The HMD group is best characterized by three samples in trace element and isotopic space (sample ID: MMB-D69A, MMB-D69B, and MMB-D70A). The RR group is characterized by three samples (CAH16-241A, DM255B, and CAH15-007). Because RR group sample CAH16-241A does not have isotopic data, it is substituted with a similar basaltic lava along Rattlesnake Road (CAH16-242A) that does (Figure 6D–F). Considering trace element and radiogenic isotope data together, samples of the RR group exhibit lower Y/P₂O₅ and higher LREE concentrations at lower ¹⁴³Nd/¹⁴⁴Nd values (Figure 6B). Samples of each group show variable ⁸⁷Sr/⁸⁶Sr ratios relative to other isotope ratios (i.e., ¹⁴³Nd/¹⁴⁴Nd or ²⁰⁶Pb/²⁰⁴Pb, Figure 7C,D). Within the RR group, ⁸⁷Sr/⁸⁶Sr values are >0.7035, and ¹⁴³Nd/¹⁴⁴Nd values are <0.51293. In the HMD group, ⁸⁷Sr/⁸⁶Sr values are <0.7034, and ¹⁴³Nd/¹⁴⁴Nd values are >0.51298 (Figure 7).

Oxygen isotope data for the PGB yield δ¹⁸O values ranging from 5.28 to 6.68 ‰ in groundmass and 5.32 to 5.92‰ in plagioclase (

Table 2. Sr, Nd, Hf, Pb, and δ¹⁸O Isotopic Compositions for the PGB.

Sample Name	Dike or Lava	87Sr/86Sr	143Nd/144Nd	εNd	176Hf/177Hf	εHf	208Pb/204Pb	207Pb/204Pb	206Pb/204Pb	208Pb/206Pb	207Pb/206Pb	204Pb/206Pb	δ18O	δ18O	Mg#	Age	Error
													GM (‰)	PL (‰)		(Ma)	(±2σ)
DM255B	Lava	0.703757	0.512905	5.4	0.283093	10.9	38.62	15.64	18.96	2.04	0.83	0.05	6.68	-	57.4	15.76	0.11
CAH16-179B	Lava	-	-	-	-	-	-	-	-	-	-	-	-	5.92	53.5	16.02	0.08
MS-11-6	Dike	0.703461	0.512974	6.7	-	-	38.70	15.68	18.97	2.04	0.83	0.05	5.55	5.79	53.2	16.06	0.14
CAH16-163	Lava	0.703584	0.512940	6.1	0.283093	10.9	-	-	-	-	-	-	5.99	5.83	45.8	16.18	0.05
CAH15-007	Lava	0.703691	0.512889	5.1	-	-	38.52	15.60	18.89	2.04	0.83	0.05	6.12	-	64.4	16.22	0.06
MC-76-16	Lava	-	0.512957	6.4	0.283128	12.1	-	-	-	-	-	-	-	5.76	52.2	16.23	0.09
CAH16-148	Lava	-	-	-	-	-	-	-	-	-	-	-	-	5.54	49.8	16.62	0.07
CAH16-138	Dike	-	0.512971	6.7	0.283109	11.4	38.43	15.58	18.84	2.04	0.83	0.05	5.75	5.81	50.2	16.70	0.09
CAH16-065	Lava	0.704102	0.512861	4.5	0.283039	9.0	38.59	15.60	18.98	2.03	0.82	0.05	-	-	44.2	16.72	0.03
CAH15-023	Dike	0.703494	0.512964	6.5	0.283100	11.1	38.40	15.57	18.79	2.04	0.83	0.05	5.45	-	56.1	16.88	0.06
CAH16-174A	Lava	0.703528	0.512964	6.5	0.283111	11.5	38.52	15.60	18.85	2.04	0.83	0.05	5.52	-	49.7	16.96	0.07
CAH17-200	Lava	-	0.512956	6.4	0.283126	12.1	-	-	-	-	-	-	5.75	5.72	52.6	17.02	0.03
CAH17-245	Lava	0.703573	0.512941	6.1	0.283097	11.0	38.49	15.58	18.87	2.04	0.83	0.05	5.78	-	47.3	17.14	0.04
CAH17-222A	Lava	0.703412	0.512994	7.1	0.283126	12.1	38.52	15.61	18.86	2.04	0.83	0.05	5.28	5.32	52.1	17.23	0.04
CAH16-073A	Lava	0.703969	0.512791	3.1	-	-	38.93	15.73	19.14	2.03	0.82	0.05	-	-	55.2	-	-
CAH16-171A	Lava	0.703597	0.512930	5.9	0.283071	10.1	38.51	15.60	18.86	2.04	0.83	0.05	-	-	52.7	-	-
CAH16-195	Dike	0.703428	0.512972	6.7	0.283109	11.4	38.44	15.57	18.84	2.04	0.83	0.05	-	-	50.8	-	-
CAH17-242A	Lava	0.703565	0.512925	5.8	0.283084	10.6	-	-	-	-	-	-	-	-	57.8	-	-
CAH16-140	Dike	-	-	-	-	-	-	-	-	-	-	-	-	5.9	55.6	-	-

GM: groundmass. PL: plagioclase.

). At 49–53 wt. % SiO₂, plagioclase δ¹⁸O values straddle the boundary of the δ¹⁸O melt; such thatδ¹⁸O plagioclase = δ¹⁸O melt at ~53 wt. % SiO₂, which is 0.13‰ lighter at 49 wt. % SiO₂ and 0.1‰ heavier at 56 wt. % SiO₂ [40]. When compared with previous work, these data demonstrate a larger range of values from groundmass and a similar overall trend of increasing δ¹⁸O values for younger samples (Figure 13). This trend is observed with the earlier ⁸⁷Sr/⁸⁶Sr isotopic data, where PGB flows lower in the stratigraphy had lower ratios and higher stratigraphic flows had slightly higher ratios. Comparing ⁸⁷Sr/⁸⁶Sr data with δ¹⁸O values collected as part of this study, there is a clear linear correlation (R² = 0.92) (Figure 5D).

3.2. Distinguishing Basaltic Lavas and Dikes of the PGB

Samples of PGB are distinguished from other main-phase CRBG formations (Grande Ronde, Imnaha, and Steens Basalts) by specific geochemical parameters. Relative to the Grande Ronde Basalt, the PGB is easy to differentiate, as it contains higher MgO (mostly > 4 wt. %) and lower concentrations of SiO₂ (mostly < 53 wt. %) and generally lower incompatible trace element contents, for example, La, Y, and Hf (Figure 2). Compared to the Imnaha Basalts, which can be subdivided into two subtypes, Rock Creek (<51% wt. % SiO₂) and American Bar (>51% wt. % SiO₂, cf. Hooper, 1974 [46]), the PGB exhibits a comparable SiO₂ range (48.5 to 53 wt. %) but with lower concentrations of various incompatible elements (i.e., Th, Hf, La) at a similar SiO₂ content (Figure 2).

A more complicated goal is distinguishing the PGB near the northern extent of the Steens Basalt, as both the PGB and the Steens Basalt are coarsely plagioclase-phyric and contain similar SiO₂ ranges. However, they are easily differentiated using Zr/Y ratios—at low SiO_2, the PGB samples have Zr/Y < 4, and at high SiO_2, the PGB samples have Zr/Y < 4.5, while samples of the Steens Basalt have ratios > 4 and > 4.5, respectively (Figure 2). Additionally, PGB samples have lower concentrations of LREE, TiO₂, and Hf at a given SiO₂ value (Figure 2).

3.3. Parental Magmas and Implications for Mantle Sources

Samples of each endmember (i.e., RR and HMD) represent some of the most primitive basalts identified within the PGB (Mg# > 57). The samples defining the HMD endmember plot closer to the C1 depleted mantle component of Carlson (1984) [3], while the samples defining the RR endmember exhibit a more enriched composition (Figure 7C,D). Magmatically evolved PGB samples are also consistent with a petrogenetic model involving two different mantle sources, as compositional variability at a given Mg# cannot be explained by fractionation or contamination alone, as documented in the following discussion.

The two defined primitive groups (HMD and RR) represent PGB endmember compositions and are distinct in both trace element and isotopic spaces with minimal overlapping attributes (Figure 7). Isotopic data suggest that the PGB samples define two populations of ⁸⁷Sr/⁸⁶Sr values, demonstrated by samples representative of both the RR and HMD endmembers. Plots of major and trace element data similarly suggest that samples of both endmember groups show distinct compositional differences. However, sample CAH15-007 plots closer to the HMD endmember array in most major and trace element spaces, and in isotopic geochemical space, this sample clusters with the RR endmember array (Figure 7). These relatively small enrichments in ⁸⁷Sr/⁸⁶Sr could be the result of crustal contamination or reflect larger contributions over time from a less-depleted magmatic source (i.e., with less conspicuous HFSE depletions and LILE enrichments).

Distinct geochemical and isotopic traits distinguish the RR endmember group from the HMD endmember group, suggesting that they were sourced from two different primitive compositions. Compared with other isotopic ratios such as ¹⁴³Nd/¹⁴⁴Nd or ²⁰⁶Pb/²⁰⁴Pb (Figure 7C,D), samples of each exhibit a wider range of ⁸⁷Sr/⁸⁶Sr ratios. For example, ⁸⁷Sr/⁸⁶Sr ratios range between 0.7030 and 0.7034 in the HMD endmember and 0.7035–7038 in the RR endmember. This variability is not observed in ¹⁴³Nd/¹⁴⁴Nd ratios, as the ratio is comparable between endmembers (HMD ~0.5130, RR 0.5129). This could be because Sr is a more fluid mobile element compared to Nd, so minor subduction metasomatism of a mantle source would be likely to result in more variability in Sr isotopic composition (compared to Nd). Higher LREE concentrations and Zr/Y values at higher ⁸⁷Sr/⁸⁶Sr ratios (i.e., >0.7035) in RR endmember samples suggest that some PGB parental magmas already contained slightly elevated ratios (Figure 7).

3.4. Fractionation Models to Account for PGB Geochemical Variability

Given the observed range of compositions even at the high MgO end of the PGB data, we take two endmembers from this dispersion as starting points for modeling the fractionation processes that resulted in the trends towards lower MgO contents. The data array from the RR endmember contains geochemical signatures similar to the earliest PGB subunit, the Twickenham Basalt (Figure 8B), while the data array from the HMD endmember is from high-MgO dikes that are independent of the three previously defined PGB subunits [23,47]. Comparing the three PGB subunits, there is one evolutionary trend for the Twickenham and Monument Mountain Basalts and another for the Dayville Basalts, which are anchored by the high-MgO dikes (Figure 4 and Figure 8). Select trace and major elements distinguish the geochemical evolution of each, such as P₂O₅ (wt. %) for similar Y (ppm) concentrations. This is useful to consider as a ratio (i.e., Y/P₂O₅), as Y has a close correlation with magmatic evolution (Figure 6, Figure 8 and Figure 9). Although Ti and P contents are effective at distinguishing various CRB units, the specific petrogenetic processes responsible for their relative differences remain unclear.

Identified primitive endmembers (HMD and RR) are used as parent compositions for fractionation models in MELTS. These compositions include a sample from the high-MgO dikes (HMD sample ID: MMB D70A, [5]) and a sample from a lava flow located along Rattlesnake Road (RR sample ID: CAH16-241A, this study) (Figure 1, Table 1). Using these two PGB compositions, a total of 24 MELTS models were conducted under variable crustal pressures (2–4 kb and 7–9 kb) and water contents (0.1 and 0.5 wt. % H₂O). During each model, pressure is held constant, and temperature is lowered to 100 degrees below the calculated liquidus in increments of 2–5 degrees C. Model outputs include crystallizing mineral phases and their modal abundance, major element concentrations, and select trace element concentrations calculated using bulk partition coefficients (Di). Select trace elements include Y, La, Ba, Th, Sc, and Cr, where the behavior of each element in the magma is dictated by a bulk partition (or distribution) coefficient (Di). Utilized partition coefficients are summarized in Table S1, and MELTS modeling results are summarized in Table S5.

Model fractionation trends for the HMD endmember composition yielded results between ~75%–82% crystallization at all pressures with 0.1 wt. % H₂O and varied between ~54%–67% when the starting composition contained 0.5 wt. % H₂O. For models conducted under low pressures (2–4 kb), plagioclase and spinel are the first crystallizing phases, followed by olivine and clinopyroxene. At higher pressures (7–9 kb), clinopyroxene and spinel crystallize first, followed by plagioclase. Olivine is only a crystallizing phase under low pressure conditions (i.e., 2–4 kb), and spinel is a minor phase (<1%) under low pressure conditions and ranges from 1%–2.3% with higher pressure. High- and low-Ca clinopyroxene crystallize during each model, except at 2 and 9 kb at 0.5 wt. % H₂O. The high-Ca clinopyroxene (~17 wt. %) crystallizes just a bit earlier than the low-Ca clinopyroxene (~4–7 wt. %). Considering both clinopyroxene populations, the modal abundance of clinopyroxene is similar (~ 48–49) for models run at higher pressures (7–9 kb). Model results also indicate increases in the modal abundance of the earlier formed high-Ca clinopyroxene with increasing pressure, where the later formed and lower-Ca clinopyroxene decreases with increasing pressure. At low pressures (2–4 kb), the modal abundance of both clinopyroxene populations is much more variable and increases with pressure (i.e., 2–3 kb and 3–4 kb).

Modeling results of the 100 °C decrease in temperature using the RR endmember composition indicate maximum fractionation degrees between 74% and 80% at lower pressures (2–4 kb) and 52%–62% at higher pressures (7–9 kb) with 0.1 wt. % H₂O. When the melt contained 0.5 wt. % H₂O, maximum fractionation degrees varied between 58% and 63% at lower pressures (2–4 kb) and 35–45% at higher pressures (7–9 kb) (Table S5). Mineral phases typically crystallized in the following sequence: orthopyroxene, followed by spinel, plagioclase, and clinopyroxene. Similar to the results using the HMD endmember, fractionation of the RR endmember composition predicts two populations of clinopyroxene, a high (>15 wt. %) and a low (<10 wt. %) Ca clinopyroxene. Both the high- and low-Ca populations of clinopyroxene crystallize with every model. For all modeled conditions, the modal abundance of clinopyroxene (both high and low Ca) remains similar, and plagioclase decreases with increasing pressure, and only under low pressure (2 kb) and wet (0.5 wt. % H₂O) conditions will model results predict olivine as a crystallizing phase (Table S5).

Orthopyroxene is an early crystallizing phase predicted in each model using the RR endmember composition (likely due to the elevated SiO₂ content) but never crystallizes during models conducted with the HMD composition (Table S5). Using the predicted modal abundance of crystallizing phases in MELTS with calculated bulk partition coefficients, concentrations for select trace elements were calculated from each model output. Trace element concentrations calculated from each model border the observed concentrations of Th, Ba, and La and could result from mixing of the magmas along the two differentiation trends. Predicted concentrations of Sc and Cr do not reflect the majority of observed PGB compositions (Figure 9 and Figure S2). Finally, the most conspicuous differences between these two primitive PGB endmembers is the enrichment of light rare earth elements (LREE) in the RR endmember composition relative to the HMD endmember composition (Figure S2). These endmembers may reflect differences in source compositions (e.g., amount of subduction metasomatism) or simply the relative depth and degree of melting.

3.5. Evaluating Crustal Contamination

To determine if assimilation of crustal material is playing a role in the petrogenesis of PGB lavas and dikes across its distribution area, K/P and K/Ti are plotted against Mg# (Figure 10). Typically, crustal contamination is evaluated via isotopic compositions, but here, the potential crustal endmembers from the young, accreted terranes are not isotopically distinct enough to have much effect on Sr and Nd isotopic compositions if assimilated. Regardless of age, felsic rocks have much higher K/P and K/Ti than mantle melts (even after crustal fractionation), so felsic crustal additions would show up in these elements no matter how old the crust is that was assimilated. Both ratios K/P and K/Ti will be modified if felsic material is assimilated, as K is concentrated in felsic material, while P and Ti are progressively depleted because of apatite and ilmenite fractionation, respectively. Additionally, if felsic crustal melts were assimilated, this should also be visible in other geochemical parameters such as SiO₂, numerous incompatible elements (i.e., Rb, Nb, La, etc.), and radiogenic and oxygen isotopic data. Incompatible elements and isotopic data exhibit no clear evidence of assimilated granitic material.

The crustal section through which PGB magmas erupted consists of a series of accreted Paleozoic-Mesozoic oceanic volcanic arcs [50]. Although there is a range of compositions present in this crustal basement, we use the composition of the Izee metasediments as a representative of the potential crustal contaminant in the area of the PGB eruptions. Compositions of Izee metasediments relative to all PGB samples are plotted on Ba/Th versus La/Sm (normalized), which is often used to help unravel influences of oceanic crust compared with added sediment (Figure 11) [51]. The addition of such sediment into a magma is important to evaluate, as a volumetrically minor amount of sediment (~2%–3%) can dramatically alter the abundances of the most incompatible elements of an erupting basalt [51]. To investigate whether metasediments sourced from the Izee formation could yield the observed PGB geochemical signatures, we utilize a simple mixing model.

Models were done with the following mixing calculation:

m = xA + B(1 − x)

(1)

Different mixing compositions were investigated at 2, 5, 10, 15, and 20% (x), where (A) is a primitive PGB composition (i.e., HMD or RR), (B) is the average bulk composition for Izee metasediments, and (m) is the resulting contaminated PGB composition. The objective was to determine if assimilation of Izee metasediments (B) could yield observed PGB compositions (Figure 12). Results of this modeling indicate that contamination of PGB by Izee argillitic material would not sufficiently modify concentrations of many HREE, such as Tb, Dy, Yb, and Lu. Elevated HREE of average PGB is more likely to reflect their extensive amount of fractionation. Patterns for all PGB compositions do not reflect as large a depletion (i.e., deep trough) in Nb and Ta concentrations on a mantle normalized trace element diagram compared to the metasediments (Figure 12). One of the most prominent compositional changes of metasediment contribution is in K concentrations, but observed concentrations show no significant evidence that sediment assimilation is a dominant process over magmatic evolution (Figure 10). Concentrations of HREE and HFSE in PGB samples could not be produced from assimilation of Izee metasedimentary material; thus, additional processes need to be invoked to produce the observed compositional array in PGB rocks.

4. Discussion

4.1. Evidence of a Contaminated Mantle?

Lava flows and dikes of the PGB contain the lowest ⁸⁷Sr/⁸⁶Sr ratios of any main-phase formations of the CRBG (Figure 6 and Figure 7), and oxygen isotopes are near the mantle range at ~5.8 ± 0.2‰. When combined with elevated concentrations of Ba and overall high LILE/HFSE ratios, these characteristics have been used to argue sourcing from a back-arc-type mantle [3,6]. Previous petrogenetic modeling suggests that low ⁸⁷Sr/⁸⁶Sr ratios (<0.7030–0.7034) are characteristic of early and less-modified PGB lavas, and any relative increase in ⁸⁷Sr/⁸⁶Sr ratios was attributed to crustal contamination [3,4,5,6]. Relative to the existing PGB stratigraphy, low ⁸⁷Sr/⁸⁶Sr ratios (<0.7032) are observed in only three PGB samples: a Twickenham Basalt flow, a Dayville Basalt flow, and a high-MgO dike [5,23,45]. Higher ⁸⁷Sr/⁸⁶Sr ratios are observed in some of the lavas of the youngest PGB subunit, the Dayville Basalt; however, ⁸⁷Sr/⁸⁶Sr ratios in the Dayville Basalt range from 0.703087 to 0.703713, almost the full range of the entire PGB [5,23]. By contrast, a sample with relatively elevated ⁸⁷Sr/⁸⁶Sr (0.703691) of this study (CAH15-007, RR endmember) is among the most primitive PGB samples in terms of elemental composition (Figure 6). Another relatively primitive PGB sample (DM255B, Mg# 57.4) also has a relatively elevated ⁸⁷Sr/⁸⁶Sr ratio of 0.703757. This sample location at Dooley Mountain represents the northeastern extent of the added PGB distribution and clusters with other samples of the RR endmember in isotopic space (Figure 6).

The previous reported δ¹⁸O values from groundmass for the PGB ranged from 5.6 to 6.1‰, where stratigraphically lower basalts of the PGB have the lowest δ¹⁸O values, identical to the canonical mantle range of 5.7–5.8‰, and stratigraphically higher lavas have slightly elevated δ¹⁸O, up to 6.1 ‰ [4] (Figure 13). This study notes higher δ¹⁸O values in samples with younger ages, in agreement with observed increases with stratigraphy from earlier studies (i.e., Brandon et al., 1993 [4]) (Figure 13). Increases in δ¹⁸O values through time could be generated by prolonged volcanism through the same area gradually increasing the amount of crustal assimilation that was occurring—supported by the observed correlation between δ¹⁸O and ⁸⁷Sr/⁸⁶Sr (Figure 5D). Fractionation of mafic phases (i.e., olivine, clinopyroxene, and plagioclase) from a basaltic magma should minimally (~0.2‰) increase δ¹⁸O, assuming a starting composition of 5.7–5.8‰ in the basalt [40]. During magmatic evolution, δ¹⁸O increases and becomes isotopically heavier in minerals with more covalent bonds, yielding lower δ¹⁸O in more mafic phases and higher δ¹⁸O in the remaining melt [40]. However, δ¹⁸O ratios also increase as a result of crustal contamination, as most continental rocks have δ¹⁸O ratios > 6‰. Observed increases in δ¹⁸O values in more primitive PGB lavas are, then, more likely a result of contamination rather than fractional crystallization (Figure 13).

Assuming that these higher PGB groundmass δ¹⁸O values (6.12 and 6.68‰) are representative of both RR endmember samples (i.e., CAH15-007, DM255B in Table 2), the question becomes if this contamination occurred in the source or later as the magma traversed the crust. The most elevated δ¹⁸O signatures are from two primitive samples of the RR group (CAH15-007 and DM255B, respectively). Both samples are groundmass separates but have undergone acid leaching following ⁴⁰Ar/³⁹Ar procedures to remove excess alteration, which could lead to misleading (i.e., elevated) values. This observation is unique relative to other CRBG units, which often have elevated δ¹⁸O with lower MgO content (i.e., Grande Ronde Basalt). The range in δ¹⁸O in Grande Ronde is too big for magmatic fractionation, so this correlation between δ¹⁸O and MgO is likely reflective of combined fractionation-assimilation [54].

Addition of sediment to magma during crustal contamination affects the Sr isotope ratio of the contaminated magma less than addition of “sediment” to the mantle through subduction zone processes because of the much higher Sr content of a basaltic magma compared to its peridotitic source in the mantle [55]. There are no ⁸⁷Sr/⁸⁶Sr data for Izee material, but Nd isotopic data yield ε_Nd values between −4 and 0, suggesting arc material that is compositionally evolved [56,57] with ⁸⁷Sr/⁸⁶Sr between 0.7045–0.707 based on the mantle correlation of Sr and Nd isotopic composition. δ¹⁸O values for Izee samples are also approximated to be > 9, as is common for most clays and clastic material [40].

To help determine if crustal contamination originated from melting a contaminated mantle source or during fractionation of primary magmas in the crust, δ¹⁸O and ⁸⁷Sr/⁸⁶Sr are plotted against one another, following James (1981) [55] (Figure 14). It is difficult to draw a definitive conclusion, as the crustal composition is not precisely known, but this does suggest that crustal contamination plays a role in petrogenesis of PGB magmas. Observations of PGB isotopic data suggest that source contamination with metasediment primarily increases ⁸⁷Sr/⁸⁶Sr prior to influencing δ¹⁸O values. By contrast, crustal contamination of the magma increases both ⁸⁷Sr/⁸⁶Sr and δ¹⁸O. PGB data are offset from mantle values in ⁸⁷Sr/⁸⁶Sr at mantle δ¹⁸O (source contamination) but show a trend of steeply increasing δ¹⁸O, with minor changes in ⁸⁷Sr/⁸⁶Sr (crustal contamination) likely reflecting assimilation of a high-δ¹⁸O mafic source. Samples of the HMD endmember may represent source contamination of a depleted mantle, whereas the RR endmember may represent a more enriched composition that experienced subsequent contamination.

4.2. PGB Magmas: Source Lithologies and Modification

Source lithologies for PGB lavas and dikes are interpreted to vary between pyroxenite to peridotite based on variable degrees of trace element enrichment and depletion, compared to Ni concentration within olivine phenocrysts [58]. The Ni content in the bulk rock can be skewed as a result of olivine accumulation. However, geochemical data combined with petrographic observations suggest that Ni content in PGB lavas is not driven by olivine accumulation, as our PGB samples that fall along the high-Ni trend are olivine poor (e.g., Soderberg and Wolff, 2023 [58], their Figure 7).

Distinguishing processes that result in incompatible element enrichment can be difficult, especially when the processes of interest have comparable geochemical consequences. Melting a sediment metasomatized mantle or adding sediment to a melt can yield comparable geochemical and isotopic characteristics in the resulting basaltic magma (Figure 9B). Both LILE and LREE can be enriched as a result of slab fluxing in the mantle or addition of sediment in the crust, and both modification scenarios have to be considered. Samples of both primitive PGB endmembers (HMD and RR) contain elevated Ba/Nb ratios (i.e., ~60) compared to MORB and OIB (Figure 9B). With respect to the RR endmember, all four sampled basaltic lavas from along Rattlesnake Road were candidates to represent a parental PGB composition (Mg# 58–59) for modeling in MELTS. However, these four lavas contain highly variable Ba/Nb ratios, ranging from 39 to 100, almost the full range observed in all PGB samples and providing evidence that PGB primary magmas must have been derived from compositionally distinct sources. The variability in Ba/Nb values within the same inferred mantle source (RR endmember) suggests that this source itself varied spatially and that it is more likely a source characteristic than the result of crustal contamination of the magmas while in the crust.

The data require a range in compositions of the primary magmas for the PGB, as shown in a variety of trace element plots (e.g., Ce/P, Ba/Nb). Simply put, magmatic differentiation alone cannot explain the huge range in the data for these ratios, and we demonstrate the minimal consequences of crustal contamination. Taken together, this suggests there must have been a wide range in primary PGB magmas derived from compositionally distinct sources.

The limited degree of crustal contamination could result from the rapid ascent rate of PGB magmas [4]. Earlier modeling tried to calculate the amount of contamination using an isotopic component that represents the crust in the Blue Mountains Province (Paleozoic accreted terranes) and a redefined C1-depleted mantle [3,4,23]. To achieve observed radiogenic Sr and δ¹⁸O values, Brandon et al. (1993) [4] estimated 8%–21% crustal contamination, which was consistent with the estimate of Carlson (1984) [3]. Although the bulk of these rocks provide little isotopic leverage in composition when compared to the CRBG, if the LILE enrichment in PGB lavas is a result of sediment input, metasedimentary Izee rocks are the most representative composition.

Trace element modeling suggests that crustal contamination with Izee metasedimentary material could not yield the observed compositional patterns in some incompatible elements in samples of the PGB, such as Nb, Ta, and some HREEs (Figure 13). Observed Zr/Y ratios in PGB lavas and dikes are highly variable, ranging from ~2 to 4.5 (Figure 9C). Both Zr and Y are similarly incompatible with partition coefficients that make them difficult to decouple during magmatic evolution. Variable Zr/Y values at high MgO suggests more than one contributing mantle source, as these ratios distinguish both primitive PGB endmembers (Zr/Y > 2.5 in RR, Zr/Y < 2.5 in HMD) (Figure 9C). If changes in Zr/Y primarily reflect compositional variability in the mantle source, the RR endmember is more heterogeneous. This is because samples defining the RR group exhibit a larger range in Zr/Y ratios than samples of the HMD group, similar to the variability of Ba/Nb ratios mentioned above (Figure 9B). Alternatively, these Izee sediments may not serve as the most compositionally representative material assimilated by PGB magmas.

High Mg basalts of the PGB contain Sc concentrations between 30 and 40 ppm; meanwhile, more evolved PGB samples contain Sc concentrations between 40 and 50 ppm (Figure 9A). The results of every MELTS model do not predict an increase in Sc but rather a decrease with magmatic evolution, suggesting that the modeling input parameters were not representative of the magmatic conditions at which differentiation occurred. Even under initial conditions that were highly saturated in H₂O (3.0 wt. %), MELTS modeling still did not account for the observed increase in Sc concentrations (Figure 9A). This, combined with higher-than-predicted Cr concentrations, suggests that fractionation of clinopyroxene was not a dominating process in the formation of PGB basalts. Instead, the elevated concentrations of both Sc and Cr trace elements indicate that crystallization of clinopyroxene and potentially orthopyroxene was delayed or suppressed—and MELTS may be overestimating the importance of pyroxene crystallization (Figure 9). The arrival of clinopyroxene on the liquidus can be delayed when fractionation occurs at lower pressures [59]. In order for Sc content to steadily rise, it needs to remain incompatible, and that could possibly be achieved by either significant crystallization of olivine and plagioclase over pyroxene or episodic magmatic recharge at low pressures (< 3 kb) that may buffer basalt magmas at conditions prior to pyroxene crystallization (and may also explain highly variable Cr/Ni ratios among PGB magmas).

4.3. Petrogenetic Relationships to Other CRBG Formations

Of the CRGB formations, PGB shares the most geochemical and isotopic characteristics with the Steens Basalt, with comparable ⁸⁷Sr/⁸⁶Sr ratios (0.7035–0.7038). While observed δ¹⁸O values in PGB samples are not anomalously low, they are slightly lower than in the Steens Basalt. Reported δ¹⁸O values for the Steens Basalt range between 5.91 and 6.75‰ (n = 17), excluding one outlier at 8.5‰ [8]. The majority of reported δ¹⁸O values for PGB range between 5.28 and 5.99‰ (n = 17 of 19, Table 2) with two values > 6‰, which is generally representative of the mantle. Within basaltic lavas and dikes of the CRBG, δ¹⁸O values below ~5.9‰ are proposed to result from hydrothermal alteration by heated groundwater [54]. Since a portion of δ¹⁸O values in PGB are consistently below the average mantle, which differs from other CRBG formations, an alternative consideration could be exposure to hydrothermal processes. However, this interpretation seems unlikely given the correlation of δ¹⁸O values to ⁸⁷Sr/⁸⁶Sr, the minor depletion of δ¹⁸O values, and the large spatial extent and age range of analyzed PGB samples. We also note that Sr could be very sensitive to hydrothermal alteration, which may explain the range of ⁸⁷Sr/⁸⁶Sr at more or less constant ¹⁴³Nd/¹⁴⁴Nd.

4.4. Two Temporal Pulses of Magmatism

Eruptive activity for the PGB spans ~1.4 Ma, as evidenced by numerous high-precision ArAr dates from samples dispersed across a wide region in eastern Oregon. These ages demonstrate that lavas and dikes of the PGB were emplaced during two discrete intervals, a temporal pattern that this research has also identified in main-phase CRBG (Figure 15). Within the PGB, it is inconclusive whether there are clear geochemical trends with time. However, younger ages are often correlated with lower Sc (ppm) concentrations, which may suggest increasing pressure during crystallization and more pyroxene fractionation relative to olivine (Figure 15).

Two of the youngest reported PGB ages represent primitive samples used to characterize the RR group within radiogenic isotope space (samples CAH15-007 and DM255B) (Table 1, Figure 7). These samples erupted during the second and later pulse of PGB activity with ⁴⁰Ar/³⁹Ar ages between 16.22 and 15.76 Ma and contain the lowest TiO₂ values of any of the dated samples (0.91 and 0.99 wt. %, respectively). While these two samples are the only dated samples within either identified endmember (RR or HMD), the other primitive samples that comprise the RR endmember have stratigraphic age control. Multiple basaltic lava flows along Rattlesnake Road both underlie and overlie the rhyolitic Dinner Creek Tuff unit 2, dated at ~15.5 Ma [61]. The exposures here are faulted, but the section overlies the Dinner Creek Tuff unit 1 and underlies unit 3, dated at 16.16 and 15.5, respectively [61]. This stratigraphic relationship constrains eruptive age and suggests that these lavas were emplaced between 15.5 and 16.16 Ma. These basaltic lavas along Rattlesnake Road erupted during the later pulse of PGB volcanism and contain low TiO₂ concentrations (~0.92–1.11 wt. %). In addition to samples of the RR endmember, HMD endmember samples contain the lowest TiO₂ concentrations of any analyzed PGB lava or dike, between 0.72 and 0.77 (wt. %). If these basaltic lavas and dikes do represent two parental components for the PGB, they were not emplaced until the second pulse of PGB eruptive activity (Figure 15).

4.5. PGB Endmember Compositions: Trust but Verify

The modeled HMD endmember composition is sourced from two dikes mapped in the Monument Dike swarm, originally interpreted as compositionally dissimilar to any CRBG geochemical subunit (samples D-16 and D-17 [47]). Although there are no published data, earlier studies also reported basaltic lavas of a similar high-MgO (and low TiO₂) composition in Flat Creek, approximately 20 km southeast of the type locality at Picture Gorge [2,62] (D. Swanson, 201—personal comm.) (Figure 1). Both sample locations (D-16/-17 and Flat Creek) were targeted as part of this study to gain a context of sample exposures, field relationships, and textures. Multiple dikes were sampled at the mapped location of D-16/-17, along with basaltic outcrops exposed at the Flat Creek section. However, none of the basaltic samples collected from either site (D-16/-17 dikes, Flat Creek exposures) contained elevated MgO and depleted TiO₂ (the geochemical composition that characterizes the high-MgO dikes). Thus, our field reconnaissance could not confirm the presence of any basaltic exposure with HMD composition as suggested by earlier workers [2,47,62] (D. Swanson, 201—personal comm.) While these high-MgO dikes are similar in their location and orientation to known PGB dikes of the Monument swarm, they are spatially and geochemically localized—with few samples existing between them and the rest of the PGB in geochemical parameter space (Figure 6, Figure 7 and Figure 8). Thus, although additional samples have been identified with similar compositions, we question whether these high MgO dikes represent a significant and volumetrically dominant parental magma of the PGB.

5. Conclusions

Geochemical compositions of the PGB and Steens Basalt were influenced by mantle contributions from a subduction-related component, likely from prior slab-sourced fluid fluxing [3,4,5,18]. In context to the PGB, this mantle source must be spatially widespread and variable in its relative contribution, as lavas and dikes with PGB composition identified as far east as Lake Owyhee and as far south as Hart Mountain, Oregon.

Samples of PGB composition exposed in both the original and extended distributions exhibit petrochemical diversity that trends toward two primitive endmembers in geochemical and isotopic spaces. MELTS modeling of both endmember compositions (HMD and RR) demonstrates that PGB geochemical variability cannot be attributed to fractionation processes alone. Evolution of the RR composition can account for a larger portion of the observed PGB trace element concentrations compared to the HMD composition, supporting the observation that samples of the HMD endmember are spatially localized and limited—and may represent a different basaltic suite.

Each group displays distinct trends with magmatic evolution, as seen in major and trace element compositional arrays. The HMD group is more sensitive to pressure fluctuations during fractionation, as exhibited by larger compositional changes between MELTS models conducted at different pressures. Predicted crystallizing phases are also dissimilar; olivine crystallizes from the HMD endmember composition but not from the RR endmember composition, while orthopyroxene crystallizes from the RR endmember composition but not from the HMD endmember composition. Results of trace element modeling is inconsistent with increasing concentrations of Sc and Cr (ppm) with magmatic evolution of natural samples. This could suggest suppression of clinopyroxene (or possibly orthopyroxene) fractionation and the continuous recharging of a Sc-enriched melt.

Differences in partial melting, compositionally distinct sources, fractional crystallization, and crustal contamination all play into creating compositional diversity in basalts. The geochemical diversity of PGB magmas cannot be solely attributed to magmatic fractionation, requiring some modification via crustal contamination. Modeling potential crustal contamination to produce the average composition for the PGB or the HMD endmember group, bulk assimilation of compositions like Izee metasedimentary rocks does not result in compositions similar to the PGB. More specifically, Izee rocks exhibit large relative depletions in trace elements such as Nb, Ta, and some HREEs. Mixing models indicate that the assimilating material could not have such significant depletions in these elements to produce average PGB compositional patterns on a mantle-normalized trace element diagram if the contamination occurred in the crust. Crustally derived material, however, does play a role in the generation of PGB, clearly indicated by the observed positive correlation of δ¹⁸O versus ⁸⁷Sr/⁸⁶Sr and enrichment of various LILEs. Considering additional mantle, magmatic, and surficial processes that influence these geochemical traits, the degree of contamination is likely < 10%, given that the majority of the PGB contains δ¹⁸O values below 6‰.

Combining geochemical traits, MELTS modeling results, and reported ages for the PGB, which range from 15.76 to 17.23 Ma, indicates that the most primitive PGB samples erupted during the later episode of PGB volcanic activity. A trend in composition over this time scale more likely reflects changing source compositions or conditions/degree of partial melting—also supported by decreases in TiO₂ (wt. %) with time, the opposite of what would be expected with magmatic evolution. On the other hand, a fluid-fluxed mantle component is present in both temporal periods of PGB volcanic activity and may become more significant over time within each eruptive pulse of volcanism.