The 40Ar/39Ar investigations of a large suite of fine-grained basaltic rocks of the Boring volcanic field (BVF), Oregon and Washington (USA), yielded two primary results. (1) Using age control from paleomagnetic polarity, stratigraphy, and available plateau ages, 40Ar/39Ar recoil model ages are defined that provide reliable age results in the absence of an age plateau, even in cases of significant Ar redistribution. (2) Grouping of eruptive ages either by period of activity or by composition defines a broadly northward progression of BVF volcanism during latest Pliocene and Pleistocene time that reflects rates consistent with regional plate movements. Based on the frequency distribution of measured ages, periods of greatest volcanic activity within the BVF occurred 2.7–2.2 Ma, 1.7–0.5 Ma, and 350–50 ka. Grouped by eruptive episode, geographic distributions of samples define a series of northeast-southwest–trending strips whose centers migrate from south-southeast to north-northwest at an average rate of 9.3 ± 1.6 mm/yr. Volcanic activity in the western part of the BVF migrated more rapidly than that to the east, causing trends of eruptive episodes to progress in an irregular, clockwise sense. The K2O and CaO values of dated samples exhibit well-defined temporal trends, decreasing and increasing, respectively, with age of eruption. Divided into two groups by K2O, the centers of these two distributions define a northward migration rate similar to that determined from eruptive age groups. This age and compositional migration rate of Boring volcanism is similar to the clockwise rotation rate of the Oregon Coast Range with respect to North America, and might reflect localized extension on the trailing edge of that rotating crustal block.
The Boring volcanic field (BVF) in the greater Portland and Vancouver metropolitan areas of northwestern Oregon and southwestern Washington (Fig. 1) consists of lava flows that erupted from dozens of monogenetic centers during latest Pliocene and Pleistocene time (Evarts et al., 2009a). Geologically youthful volcanoes are recognizable features in the Portland area, and were referred to collectively as Boring Lava by Treasher (1942a, 1942b) after a group of volcanic-capped hills near the community of Boring, Oregon, ∼20 km southeast of downtown Portland. The volcanic field apparently represents a westward extension of late Cenozoic volcanism of the Cascade volcanic arc; some vents are located as much as 90 km west of the arc axis near Mount Hood (Fig. 1; Peck et al., 1964; Hildreth, 2007).
Despite the obvious hazards and neotectonic implications of young volcanism in an urban setting, surprisingly little was known about the age and composition of the Boring volcanoes until recently. Trimble (1963) mapped the distribution of Boring volcanic rocks and Allen (1975) inferred the locations of dozens of presumed Boring vents largely on geomorphic grounds. Subsequent work in parts of the BVF (Hammond, 1980; Hammond and Korosec, 1983; Madin, 1994) differentiated some Boring flows based on geochemistry and limited isotopic dating, and K-Ar dates for several Boring volcanic centers were provided in Conrey et al. (1996). However, no systematic attempt to determine the eruptive history of the BVF had been undertaken prior to our study.
In order to characterize the BVF fully and assess its neotectonic and hazards significance, an integrated program was initiated involving geologic mapping, petrographic and geochemical analyses, 40Ar/39Ar dating, and paleomagnetic determinations. Our study describes the detailed results of 40Ar/39Ar geochronology and includes details of paleomagnetic polarity and physical stratigraphy. General petrologic study is routinely part of geochronology, but more detailed and systematic studies of BVF paleomagnetism, field mapping, and petrology and geochemistry will be reported elsewhere. A major result of this study is the development of an approach for calculating geologically reasonable ages for samples whose 40Ar/39Ar analyses exhibit nuclear recoil. Such recoil model ages represent 37% of the 40Ar/39Ar ages reported here and are an integral part of the conclusions.
GEOLOGIC SETTING AND DISTRIBUTION
The BVF extends from the Cascade Range westward across the southern part of the Portland Basin, a late Neogene to Quaternary topographic and structural depression within the Puget-Willamette forearc trough of the Cascadia subduction system (Evarts et al., 2009a). Following Allen (1975), the eastern boundary of the volcanic field is placed at long 122°W (Fig. 2), which approximates the position of the Quaternary Cascade arc front elsewhere (Guffanti and Weaver, 1988; Hildreth, 2007). The BVF extends well west of the Portland Basin, occurring west of the Willamette River and the Portland Hills (Fig. 2). The area of the field is ∼4000 km2; ∼500 km2 of this is underlain by locally erupted volcanic rocks. The total eruptive volume for the field is ∼10 km3, issued from ∼80 individual volcanoes, most of which erupted small volumes. The BVF is a monogenetic volcanic field (Connor and Conway, 2000) with vents that erupted compositionally similar lavas over short intervals of time. BVF volcanic centers are chiefly mafic and, like those in the main Cascade arc, are apparently related to subduction of the Juan de Fuca oceanic plate beneath western North America. However, the tectonic position of the field is anomalous, being located in the forearc of the Cascadia convergent margin, well trenchward of the volcanic arc defined by large, long-lived stratovolcanoes such as Mount Hood and Mount Adams (Fig. 1).
The Portland Basin began to form ca. 17 Ma and gradually filled with sediment transported from the east by the ancestral Columbia River (Evarts et al., 2009b). Significant volcanism across the basin, however, did not commence until latest Pliocene time. In the southern and eastern parts of the BVF, locally erupted volcanic rocks overlie a series of mostly low-K tholeiite flows (Evarts et al., 2009a) that were assigned informally by Peck et al. (1964) to the volcanic rocks of the High Cascade Range. These largely Pliocene units evidently issued from vents in the Cascade arc well to the east (Conrey et al., 2004). Earlier workers, including Peck et al. (1964), generally grouped these slightly older, far-traveled flows with the Boring lavas, but we exclude them because they did not erupt from local vents, and may represent a different geologic setting.
The 40Ar/39Ar radiometric dating technique is a variation of the potassium-argon method that uses neutron activation of 39K to 39Ar, and analysis of the resulting argon isotopic ratios to obtain high-precision age estimates (Merrihue and Turner, 1966; McDougall and Harrison, 1999). Details of most 40Ar/39Ar techniques are provided in Appendix 1. The use of the 40Ar/39Ar method in the study of fine-grained volcanic rocks such as those of the BVF, however, has lagged behind its development with mineral separates, despite a number of very important advances in the understanding of nuclear recoil. This study contributes to, utilizes, and documents a model that provides reliable age results in the presence of significant Ar recoil effects, regardless of the presence of an age plateau. This approach to 40Ar/39Ar dating of fine-grained materials is discussed here as a major aspect of this study.
Ar Recoil Effects
The 40Ar/39Ar technique involves neutron irradiation of rock samples for the concurrent measurement of 39Ar, as the surrogate for potassium, and radiogenic 40Ar in order to determine their ages (Merrihue and Turner, 1966; McDougall and Harrison, 1999). When nuclei of 39K capture a neutron, they become unstable and emit a proton, transmuting to 39Ar by the nuclear reaction 39K(n,p)39Ar. Incorporation of a neutron by any atomic nucleus transfers most of the energy of the neutron to the nucleus, elevating it to an excited state. The excited nucleus often decays immediately by emitting a particle, but may emit some or all of the energy as a photon (gamma radiation). The emission of a particle, such as the proton emitted to produce 39Ar, transfers some of the energy to the product nucleus in the form of kinetic energy, displacing it in a direction opposite that of the emitted particle. This displacement is known as nuclear recoil, or the recoil effect. In their study of lunar samples, Turner and Cadogan (1974) were the first to identify loss of 39Ar by nuclear recoil as the cause of negatively inclined 40Ar/39Ar age spectra (age spectra whose ages decrease progressively with increasing percent of 39Ar released). When the site of the 39Ar atom involved is near the margin of the irradiated mineral grain, ejection of the proton results in the recoil of the 39Ar atom with sufficient energy that it may move out of the grain in which it was produced if it recoils toward the grain margin (Turner and Cadogan, 1974). Similarly, neutron activation of 40Ca produces the reaction 40Ca(n,α)37Ar, in which the emission of an alpha particle causes recoil of the 37Ar. Argon atoms recoiling out of a grain may be incorporated by adjacent grains in contact with the source grain or be lost as Ar gas to the space between grains. Argon recoiled from separated mineral grains is generally lost, whereas that from multigrain aggregates, such as rock groundmass or matrix, is more likely to be incorporated into an adjacent grain. Depending on the size and shape of the grains irradiated, loss of 39Ar and 37Ar may have a significant effect on the apparent age determined. When 39Ar is lost from the material analyzed, the potassium atoms it represents are not counted in the analysis, and the potassium concentration of the sample is underestimated proportional to the amount of loss. Because 39Ar is used to determine the amount of 40K, the radioactive parent in the K-Ar technique, its underestimate results in an overestimation of the calculated 40Ar/39Ar age.
Whereas recoil of 39Ar affects 40Ar/39Ar ages directly, recoil of atoms of other Ar isotopes may affect ages indirectly by affecting the corrections made to abundances of 40Ar, 39Ar, and 36Ar for atmospheric and irradiation-produced interferences (Turner and Cadogan, 1974; Onstott et al., 1995; Jourdan et al., 2007). Recoil of 37Ar is the most significant of these because the ratio of calcium-derived 37Ar/36Ar is used to correct for simultaneously generated 36ArCa. The underrepresentation of 37Ar, because of its loss by recoil, results in the inclusion of some 36ArCa as atmospheric 36Ar. Atmospheric 36Ar is used in turn to correct for atmospheric 40Ar, and its overestimation results in an overcorrection of 40Ar and a reduced apparent age. In rocks with limited amounts of radiogenic 40Ar this overcorrection may result in removing more 40Ar than is measured, resulting in a so-called negative age. The effects of 37Ar recoil are most serious in high-calcium and low-radiogenic materials such as young basalts like the Boring lavas.
Huneke and Smith (1976) demonstrated that recoiled 39Ar is not only lost from the near-surface areas of grains, but the recoiled atoms may be embedded in adjacent grains and released according to the Ar-release characteristics of the receiving grains. Their evidence of redistribution of recoiled 39Ar nuclei confirms the interpretation made by Turner and Cadogan (1974) for commonly observed age spectra that decline progressively from older ages at low temperatures to minimum ages at high temperatures, but yield meaningful integrated ages. These negatively inclined age spectra were called the second type of anomalous age patterns by both Huneke and Smith (1976) and Turner and Cadogan (1974), the first type showing only high-temperature decreases in age. In reporting the loss of both 39Ar and 37Ar by recoil from sanidine and plagioclase mineral separates, Jourdan et al. (2007) noted the results of Turner and Cadogan (1974) when observing that reimplantation of recoiled nuclei is greatly diminished in powdered or finely ground samples. In nonvesicular crystalline aggregates (e.g., lunar basalts, Boring lavas), many or even most mineral grains are bounded by other mineral grains or glass. In this case, rather than being lost to the voids between separated grains and escaping from the sample, recoiling 39Ar and 37Ar are commonly redistributed to adjacent grains or glass, which may have Ar-retention characteristics different from the grains from which the atom recoiled. In this way, the effect on the resulting 40Ar/39Ar age spectrum may be amplified, whereas the integrated age may be unaffected, as noted by Turner and Cadogan (1974) and Huneke and Smith (1976).
Recoil effects of 39Ar, producing age spectra with progressively declining ages, were discussed by Turner and Cadogan (1974) and by Huneke and Smith (1976), but the recoil effects of 37Ar in fine-grained rocks have received less attention, as noted by Jourdan et al. (2007). Jourdan et al. (2007) discussed the effects of 37Ar recoil in fine-grained (plagioclase-rich) mineral separates as a function of grain size, showing the decrease in total-gas 40Ar/39Ar and apparent age with increasing net loss of 37Ar and resulting undercorrection for Ca-derived 36Ar. Foland et al. (1993) reported an age spectrum for a fine-grained (45–75 μm), plagioclase-rich (>97%) crystalline aggregate separated from a Jurassic tholeiitic basalt (81-7-2A), which provides an excellent example of 37Ar recoil (Fig. 3). Whereas the lowest temperature fractions of 81-7-2A appear to show minor recoil of 39Ar, probably related to the ≤3% cryptocrystalline matrix reported in the sample (Foland et al., 1993), the ages of the remaining increments increase progressively with temperature from values 23% below the 175.3 Ma total-gas or integrated age to values 15% above it. Apparently unaffected by recoil, a plateau age of 176.8 ± 0.8 Ma is defined for the central 64% of the 39Ar released, followed by elevated ages in high-temperature increments affected by redistributed (migrant) 37Ar (Fig. 3). The positively inclined age spectrum is typical of 37Ar recoil, as 37Ar has been redistributed from sites that release Ar at low temperatures, leading to overcorrection of atmospheric 40Ar and young ages in those increments. Redistribution of this 37Ar into sites releasing Ar at high temperatures results in undercorrection for atmospheric 40Ar in those steps and apparent ages greater than the true age of the material. Contrary to the results for crushed mineral separates, the similar plateau and integrated (total gas) ages for basalt 81-7-2A demonstrate that substantial recoil redistribution of 37Ar, resulting in positively inclined age spectra, may occur without significant net loss of the recoiling species. It is important to note here that this redistribution, affecting gas fractions containing more than one-third of the 39Ar released, does not affect the very tightly constrained plateau age of the remaining 64%. These age spectra are distinctly different from positively inclined age spectra typical of 40Ar loss, as discussed in detail by Turner (1968) and later in the section on 40Ar/39Ar results.
Results for a second Antarctic tholeiite reported by Foland et al. (1993) are also shown in Figure 3 to document a contrasting age spectrum dominated by recoil redistribution of 39Ar. The second age spectrum is from a 75–150 μm, high potassium (∼2% K2O) basalt glass separate, sample 90-75-2 (Foland et al., 1993). The two samples of identical age yield age spectra that are nearly mirror images of each other, although a central plateau is defined in each. The age spectrum of plagioclase-rich sample 81-7-2A is dominated by recoil of 37Ar, with low ages at low temperatures and ages above the plateau at high temperatures, defining a positive sloping pattern. The high-K glassy sample 90-75-2 is dominated by recoil of 39Ar, with ages above the plateau at low temperature and below it at high temperature, defining a negative sloping pattern. In addition to statistically identical central plateaus, the integrated age of each age spectrum is indistinguishable from the plateau, demonstrating substantial recoil redistribution of argon without significant net loss (Fig. 3). From the grain sizes reported by Foland et al. (1993) and recoil dimensions (Turner and Cadogan, 1974; Huneke and Smith, 1976; Onstott et al., 1995; Jourdan et al., 2007), this result should not be altogether surprising. Calculations similar to those of Jourdan et al. (2007) show that recoil loss of 39Ar from equant-shaped grains should be <∼0.5% for the grain sizes reported for the samples in Figure 3 and barely detectable within error under most circumstances. We conclude that significant recoil redistribution of both 39Ar and 37Ar in fine-grained crystalline and glassy aggregates could still result in 40Ar/39Ar age spectra whose central portions yield highly precise ages and whose integrated ages show either little or no net loss of either recoiling species.
Model Ages: Recoil
As shown in Figure 4 and Table 1, where the central fraction modeled is 0.5 or 50% of the 39Ar released, all of the 775 °C (21.62%) and 850 °C (16.20%) steps plus 8.59% from the 700 °C increment and 3.59% from the 925 °C step are within the modeled portion (totaling 50.00%). These proportions of the measured ages are combined as shown above for tR to define the recoil model age for 50%. If the modeled portion is increased to 0.7 or 70%, then 18.59% from the 700 °C step is included in the modeled age along with all of the 775 °C, 850 °C, and 925 °C steps plus 1.07% from the 1000 °C step (Table 1). The uncertainty in the age of any fraction of an increment is taken to be the same as that calculated for the whole. The uncertainty in the model age is propagated by standard techniques, but the mean square of weighted deviates (MSWD) is calculated for the ages included in the central fraction. Where the MSWD is >1.0, errors for each age are multiplied by the square root of the MSWD, increasing the uncertainty in the model age as the dispersion in the ages increases (Ludwig, 2003). In this study recoil model ages were calculated for the central 50% and 70% of the 39Ar released by assuming that all gas released in a step was of the same age, just as plotted in the age spectrum diagram. Model ages calculated in this way are not sensitive to the direction or amount of slope on the age spectrum, but MSWD increases dramatically as slope increases or decreases from zero. Abrupt changes in slope within the central portion have significant effects and make the choice of that portion more subjective. Selection of the central fraction is based on an examination of the age spectrum and on the statistics of the ages calculated, with the intent to avoid portions involving net loss of recoil products and those most seriously affected by recoil effects. We emphasize our use of the term, model, with reference to this age, recognizing that the model may not always be satisfied. Alternative approaches could involve modeling a smooth spectrum and selection of the central portion by minimizing the error in the observed age, but these approaches will be evaluated elsewhere.
Model Ages: Excess Argon
As used here, excess argon refers to 40Ar in the rock exceeding that generated by decay of 40K within components of the rock. This excludes 40Ar inherited from xenocrysts or xenoliths, where it accumulated by natural decay prior to incorporation in the magma and may appear in the age spectrum as radiogenic 40Ar formed in situ. Kelley (2002) provided an excellent discussion of excess Ar and its treatment as a trace element controlled by mineral-melt and mineral-fluid partition coefficients. Kelley (2002; supported by numerous studies cited therein), concluded that the dominant source of excess 40Ar is from fluid and melt inclusions incorporated in the rock during crystallization and/or quenching. Age spectra for samples containing unsupported radiogenic 40Ar or excess Ar have been recognized for a long time (Dalrymple et al., 1975; Lanphere and Dalrymple, 1976). Saddle- or U-shaped age spectra were first recognized as typical of excess Ar (e.g., Lanphere and Dalrymple, 1976), but studies of excess Ar in a wide range of geologic environments have demonstrated a complex variety of other release patterns under special conditions such as high pressure and partial resetting during metamorphism (e.g., Foster et al., 1990; Phillips, 1991; Scaillet et al., 1992; Scaillet, 1996). Saddle-shaped age spectra remain the most commonly observed and more reliably interpreted patterns, however, especially in volcanic rocks. Model ages based on excess Ar are calculated for a number of samples in this study, where the age defining the minimum of a well-developed saddle or U shape is considered the maximum age of the sample. Where this age is represented by more than one increment, the model age is taken as the weighted mean.
Ages determined by 40Ar/39Ar techniques for the Boring lavas are calculated and reported in Table 21 as integrated (total gas), weighted-mean plateau, isochron, and modeled ages. The integrated age is calculated from the sums of all radiogenic 40Ar and potassium-derived 39Ar in all increments of gas analyzed. A plateau age is defined as the inverse-variance weighted mean of that part of an age spectrum composed of contiguous gas fractions that together represent >50% of the total 39Ar released from the sample and for which no difference in age is detected between any two fractions at the 95% level of confidence (Fleck et al., 1977). An isochron age is calculated using the algorithm of York (1968) for 40Ar/36Ar versus 39Ar/36Ar correlation with conventions adopted by Ludwig (2003). Modeled ages, discussed herein, represent the age calculated for a specified fraction of an age spectrum, based on a specific model for the origin of that pattern. Columns in Table 2 labeled “Indicated age” report the age determined to be the best age estimate for the sample from evaluation of the different analytical approaches and from consideration of relevant geologic, paleomagnetic, petrologic, and chemical results. Tables of incremental heating results of individual samples summarized in Table 2 are available in the Supplemental Table2.
Our results identify two groups of dated olivine-phyric flows that were previously considered part of the BVF, but represent older volcanic activity. The earliest group consists of lavas that yield ages of ca. 13.6 Ma and belong to inadequately studied Miocene sedimentary and volcanic units (Table 3). The 40Ar/39Ar results for these Miocene rocks are reported here because of past confusion concerning their relationship with the BVF, but the geology of these older units is not part of this paper.
The second group of (largely) pre-Boring volcanic rocks consists of more voluminous, far-traveled, predominantly low-potassium tholeiite flows that were sourced in the High Cascades to the east and followed drainages, including the ancestral Columbia River, into the Portland Basin (Fig. 2). These rocks, which yield ages between ca. 3.6 Ma and 2.9 Ma (and one age of 1.96 Ma) (Table 3), are excluded from the BVF because their vents are outside the defined area. These are not to be confused with locally erupted low-potassium Boring lavas that are part of the field and for which results are shown in Table 2. Results for early low-potassium tholeiite lavas are included in Tables 3 and 43 despite their eruption from distant vents because they occur within the area of the BVF, bear on the local stratigraphy, and their ages and paleomagnetic directions were measured as part of this study.
The earliest eruptions of the Boring lavas are typified by sample QV02-83 (Fig. 5), basalt from southeast of Oregon City and west of Estacada, that yields a plateau-type age spectrum and an age of 2.609 ± 0.013 Ma (Table 2). Small amounts of gas, released in the earliest and latest increments, probably represent recoil redistribution of small amounts of 37Ar during neutron irradiation from sites releasing argon at lower temperature (e.g., Na-rich groundmass plagioclase) to those releasing at high temperature (e.g., Ca-rich pyroxene). The youngest samples analyzed in this study are from the Beacon Rock volcanic neck along the Columbia River east of Multnomah Falls (QV98-17 , QV01-35B) (Fig. 5). The Beacon Rock samples were collected at two localities and analyzed three times. The first was analyzed twice as a whole-rock sample and yielded widely disparate ages of 55.8 ± 6.1 ka and 84.2 ± 5.1 ka. A second sample was collected and analyzed as a groundmass separate and yielded an age of 58.4 ± 6.4 ka that is in agreement with the younger of the two whole-rock analyses. This age is accepted as the most reliable estimate of the age of the unit.
The remainder of this section relates to representative age spectra of samples characteristic of specific patterns, ranging from plateaus, to positively or negatively inclined spectra, to more complex patterns that may not be successfully interpreted. The age patterns for samples 06BV-G750, QV01-34, and QV03-129 (Fig. 5) exhibit age plateaus in their central portions with ages above the plateau in low-temperature steps and below it in high-temperature fractions in a classic 39Ar recoil pattern (see Fig. 3). The well-defined plateaus show no evidence of age discordance, however, and recoil effects are minor to moderate. As should be expected, recoil model ages are virtually identical to the plateaus. Age spectra of samples such as 03BV-G186, 04BV-G292B, QV03-167, and RC02-134 (Fig. 5), however, show negatively inclined patterns typical of severe 39Ar recoil redistribution with no age plateaus defined, but the ages are also well constrained. The ages of the first two are well controlled as part of a group of four samples located ∼5 km northwest of Larch Mountain that have the same reversed magnetic polarity and nearly the same stratigraphic position. Recoil model ages and integrated ages of all four are within analytical error at 1.5 Ma (Table 2). The last two samples represent lavas erupted very close to the Brunhes-Matuyama reversal at 0.781 Ma (Ogg, 2012). Normal polarity sample RC02-134 was erupted at 0.770 ± 0.050 Ma, shortly after the reversal, whereas sample QV03-167, with an age of 0.784 ± 0.022 Ma, is from a unit with a transitional magnetic orientation (discussed herein as perhaps the best constrained example of agreement between recoil model ages and magnetic polarity).
Recoil redistribution of 37Ar is also apparent in samples of the BVF. Age spectra for samples 02CM-T145, 03CM-T250, and QV01-42 define positively inclined patterns typical of 37Ar recoil, with younger ages in the low-temperature steps and older ages at high temperature (Fig. 5). As noted here, these cannot be confused with results of 40Ar loss, which cannot produce ages less than zero, and whose positively inclined age spectra either flatten (small to moderate loss) or steepen (large to extreme loss) monotonically with increasing temperature (Turner, 1968). In two of these three examples the recoil of 37Ar out of sites where it would be released at low temperature results in total 40Ar/36Ar below that of atmosphere and negative apparent ages. In each of these three samples significant amounts of 37Ar are redistributed to sites where it is released at high temperatures, resulting in undercorrection for atmospheric 40Ar and ages well above the true age of the sample. Age spectra such as those of samples QV01-40 and QV01-44 (Fig. 5) yield central plateaus, but exhibit negative ages in both lowest and highest temperature intervals. These indicate substantial bulk loss of 37Ar, rather than redistribution, even in high-temperature sites where total 40Ar/36Ar is also below that of atmosphere. These cannot be accounted for by redistributed 39Ar because no amount of 39Ar can result in statistically negative ages. Negative 40Ar/39Ar ages at high temperatures indicate that improper corrections of atmospheric 40Ar are being made as a result of bulk loss of 37Ar.
Complex age spectra such as those of samples QV02-56, QV02-94, and QV03-144B (Fig. 5) define patterns with modest saddle or U shapes characteristic of excess Ar. Where amounts of possible excess are modest, recoil of both 39Ar and 37Ar, or bulk loss of 39Ar, possibly with some high-temperature release of redistributed 37Ar, might produce these patterns. True excess Ar patterns, however, are also identified in Boring lava samples such as tunnel samples Tri-Met 807+73, Tri-Met 809+44, and Tri-Met 813+73 (Robertson light rail tunnel west of Portland, i.e., “Tri-Met tunnel”; Fig. 5; Walsh et al., 2011). Where the magnitude of the saddle is great, however, the probability of excess Ar is much more likely. The presence in whole-rock samples of large amounts of coarse-grained olivine, a common source of excess Ar, undoubtedly contributed to this effect in those samples.
Sample-related complexities also contribute to a few age spectra that defy interpretation, probably due to recoil with substantial net loss of the recoiling nuclides. Sample 05BV-G333 (Fig. 5; Table 2) has no plateau, a disparate integrated age, and no paleomagnetic sampling for control. Low ages in both early and late increments suggest net recoil loss. Sample QV03-164 shows large amounts of both 39Ar and 37Ar recoil producing an enormous range in ages, between 0.13 and 1.4 Ma (Fig. 5; Table 2). The recoil model age of 1.13 ± 0.10 Ma is consistent with its reversed polarity, but little confidence can be attached. Sample RC02-129 also shows low ages in both early and late increments and large amounts of recoil redistribution (Fig. 5; Table 2). Its isochron age is invalid due to recoil and its recoil model age and integrated age are discordant. The integrated age for this sample is more consistent with its normal polarity, but together with the other two samples discussed here provides evidence that recoil may result in age spectra that are not interpretable.
PALEOMAGNETISM OF THE BVF
Techniques used in paleomagnetic sampling and measurement of the Boring volcanic rocks are provided in Appendix 2.
The natural remanent magnetization (NRM) intensities for samples of the Boring lavas and intrusions, measured prior to any demagnetization treatments, are ∼1–0.1 A/m. Both of the alternating field (AF) demagnetization procedures were successful in defining stable characteristic directions of magnetization for the rock samples (Fig. 6). Pilot and later stepwise demagnetizations indicate that the specimens contain primarily univectorial magnetizations that form linear components on orthogonal vector plots, which decay toward the origin with increasing demagnetization. Significant secondary magnetic overprints were rare (Fig. 6G), but those found might be due to local reheating from subsequent dike emplacement. Success of AF demagnetization treatments in reducing NRM intensities to <10% of their initial values by 80–100 mT indicates that the magnetic mineralogy in many of these rocks is most likely some form of titanomagnetite. Where AF demagnetization is incomplete and a significant amount of remanent intensity remains (Figs. 6C, 6E), titanohematite is also likely present.
The characteristic magnetizations are inferred to be primary thermal remanent magnetizations (TRM) acquired during initial cooling of the rock unit, and this inference is corroborated by the excellent correspondence of observed paleomagnetic polarities with the geomagnetic polarity time scale (Table 4; Ogg, 2012). Due to the absence of measurable posteruptive deformation in the region, no structural corrections were applied to the in situ site-mean directions.
Site-mean directions for the BVF (Table 4) are shown in Figure 7. For the most part, dispersal of these directions about the long-term dipole direction is associated with paleosecular variation of the geomagnetic field resolvable by the paleomagnetic method over centuries to millennia. Such changes in direction, along with geologic mapping and geochemical studies, have been used to correlate elements within the Boring units, but this analysis will be the subject of a separate report. Here, a unit’s geomagnetic polarity (normal, reversed, transitional, or excursional) and associated 40Ar/39Ar ages, particularly those near polarity boundaries, are compared with an established polarity time scale (Ogg, 2012) to identify any inconsistencies between the paleomagnetic and geochronologic methods. One likely excursional direction (Table 4; Fig. 7) was found in a hand sample from a flow dated as 1.500 ± 0.013 Ma, within the Matuyama polarity chron.
Although paleomagnetic directions can be affected significantly by ancient (and unconstrained) magnetic anomalies and terrain effects that deformed the local geomagnetic field at the time of a rock unit’s cooling and magnetic acquisition (Baag et al., 1995), the paleomagnetic polarity is far more robust and within a geochronologic framework is considered the most reliable indicator of the rock unit’s relative age.
Constraints on the Reliability of Measured Ages
Despite complexities introduced by sample-related issues such as fine grain size and phenocryst- and/or fluid-inclusion–borne excess Ar, 40Ar/39Ar geochronology of the Boring lavas is remarkably successful. The 40Ar/39Ar age spectra, isochron, recoil or excess Ar model ages, and integrated (total gas) ages were calculated for nearly 150 samples (Table 2). Plateaus were defined by data for 92 of these, representing 62% of those studied. Recoil model ages in all 92 samples are concordant with these plateau ages (i.e., agree within the quoted 2σ uncertainties). Although plateaus are not always precisely central, they commonly contain most of the same steps included in the recoil model age, so this is not unexpected. The 40Ar/39Ar isochron ages were calculated for all but two samples, and all except one of these are also concordant with their plateaus for similar reasons (Table 2). It is important that, where no 40Ar/39Ar plateau was defined, recoil model ages and isochron ages that exclude strongly divergent steps were also concordant for all samples except those few with extreme recoil effects or excess Ar. This indicates that recoil ages are a good proxy for plateaus, and that both the recoil model age and the isochron age may represent the age of the sample in the absence of an age plateau where they agree within a 2σ level of uncertainty. The value of recoil model ages in rocks like those of the BVF, however, is demonstrated by the Lookout Point basaltic andesite, which has transitional magnetic polarity (Table 4). The isochron age of sample QV03-167 from this unit is 0.849 ± 0.030 Ma and not within 2σ of a known magnetic reversal. The corresponding recoil model age, however, is 0.784 ± 0.022 Ma (Fig. 5) and compares to an age of 0.781 Ma for the Brunhes-Matuyama boundary (Ogg, 2012). A second sample of the unit was less affected by recoil, yielding plateau, isochron, and recoil model ages that are consistent with the boundary age (Table 4).
Although geologic mapping, petrology, and geochemistry provide constraints on ages through correlation of units, stratigraphic position, and structural level, as shown above, paleomagnetic results provide the strongest independent constraints on the measured ages through comparison to the time scale of geomagnetic reversals (Ogg, 2012; shown graphically in Fig. 8 and in the final column of Table 4). Dated samples near the Brunhes, Matuyama, and Gauss boundaries agree within uncertainties (Table 4; Fig. 6); 6 samples are within 20 k.y. of the Brunhes-Matuyama boundary, 2 of which have transitional magnetic directions. Of the 8 samples within 20 k.y. of the Matuyama-Gauss boundary, 6 also have consistent polarity. The agreement is also reflected in samples erupted during brief periods of the appropriate polarity in the Jaramillo, Olduvai, and Kaena geomagnetic polarity events. Two reversed samples plot within 15 k.y. after the Jaramillo normal polarity event and 11 samples of reversed polarity closely bracket the brief 1.185–1.173 Ma Cobb Mountain normal event, which was not represented in our sampling. The Olduvai event is represented by the basaltic andesite of Oneonta Creek (Tables 2 and 4), and the Kaena event is well represented by low-potassium tholeiites assigned by Peck et al. (1964) to the volcanic rocks of the High Cascade Range (Tables 3 and 4). Of the dated samples for which paleomagnetic results are available (Table 4), only three yield ages that are inconsistent with their observed paleomagnetic polarity beyond their 2σ uncertainties calculated following the approach of Ludwig (2003). The ages of two of these three samples, from the basalt of Canemah, are statistically within the uppermost part of the Gauss, but have reversed polarity (Table 4). The measured ages of four other reversed polarity samples of the earliest lavas of the BVF are also older than the 2.581 Ma age accepted as the Matuyama-Gauss boundary, although these others are not statistically different at an uncertainty of 2σ. If J values consistent with an age of ca. 28.2 Ma for Fish Canyon sanidine (Kuiper et al., 2008; Rivera et al., 2011) were used, these ages would be further increased, suggesting that the Ogg (2012) value of 2.581 Ma may underestimate the best age for that boundary. The third age inconsistent with its polarity is the 0.968 ± 0.008 Ma age measured for the normally magnetized basaltic andesite of Kelly Butte. This age is ∼20 k.y. younger than the age given by Ogg (2012) for the end of the Jaramillo event, slightly greater than the 2σ uncertainty. Because inconsistencies in the data set are so limited, the most probable explanation is that propagated analytical errors inadequately represent the total uncertainties in those samples. We conclude that 40Ar/39Ar ages reported here, including recoil model ages calculated as discussed herein, agree well with paleomagnetic constraints and ages on the same units with no apparent age bias.
Age and Geographic Trends in the BVF
Figure 8 presents a relative probability plot of the 40Ar/39Ar indicated-age results from Table 2. Measured ages included are shown in the upper diagram with their uncertainties. Volcanic rocks with eruptive vents outside the BVF and Miocene units (Table 3) are not included in the figure or in the following discussion. Dated samples, covering ∼80% of the identified volcanic centers in the BVF, document persistent but intermittent small-volume volcanic activity in the Portland Basin during the past 2.7 m.y. The figure includes more than one sample from a center in several cases, but the broad coverage of eruptive centers lends credence to the representative nature of the data set. Based on our sampling, peak activity within the BVF occurred between ca. 1.3 and 0.5 Ma, most of which was erupted from monogenetic centers (Treasher, 1942a, 1942b; Allen, 1975). Breaks of 100 k.y. or more in the record of dated volcanic activity occurred between 2.4 and 2.3 Ma, 2.2 and 2.0 Ma, 1.9 and 1.7 Ma, and 500 and 350 ka (Fig. 8).
Evarts et al. (2009a) summarized the distribution and age of the volcanic centers within the BVF, based on many of the results cited herein. The beginning of BVF volcanism in the southern part of the field with a northward and westward migration of activity within the field was reported (Fleck et al., 2002; Evarts et al., 2009a). To evaluate spatial, temporal, and compositional trends more thoroughly, we elected to subdivide the data set into age groups of sufficient size to provide some measure of statistical relevance, but adequately spaced in time to evaluate spatial change. Using observed breaks in the age data as indicative of true breaks in eruptive activity seemed the most logical approach, but using all the breaks noted here resulted in groups of 5 and 6 ages as well as groups of 31 and 91 ages. To provide a more representative distribution of ages a more arbitrary subdivision was necessary, utilizing shorter breaks in dated activity and combining early periods of infrequent activity. The following age groups were selected, and are shown in Figure 8: (1) 2.7–2.1 Ma, (2) 2.1–1.3 Ma, (3) 1.3 Ma to 900 ka, (4) 900–450 ka, (5) 450–200 ka, and (6) 200–0 ka. Although admittedly subjective, even more arbitrary divisions of the data set gave similar results, including ones based on dividing the data into 6 groups of equal numbers of ages and into 2 equal groups based on composition (discussed herein).
When locations of dated BVF samples are plotted by age groups defined by pulses of activity, these distributions confirm a northward migration of activity and suggest a somewhat more systematic evolution of the BVF (Fig. 9). Evarts et al. (2009a) noted that the eruptive vents for most lava flows of the BVF can be identified, but are not known for all of the flows sampled. This introduces a small uncertainty that could be reduced if the locations of all vents were known or if the ages of each recognizable cone had been determined. This may be practical in the future with more detailed work, but in this study the dated locations are considered to adequately represent the sources of small-volume BVF volcanism. The distributions outlined by the samples define a suite of elongate fields that documents a northward migration of volcanism in the BVF. Fields for the oldest (2.7–2.1 Ma) and the youngest (200–0 ka) age groups are shown by shaded patterns in Figure 9, emphasizing the almost complete absence of overlap in their areas. The 2.1–1.3 Ma group also has little overlap with the youngest volcanism of the BVF, but the distribution of volcanism during this period is in a more central area of the BVF, not extending south of the town of Boring (Fig. 9). It is noteworthy that the fields of all four intermediate aged groups (spanning the 2.1 Ma to 200 ka period) define a northeast-southwest trend between those of the youngest and oldest age groups with significant overlap between them. Locations of points within the fields are dispersed, but geographic trends within these distributions are apparent. To define these trends, linear trend lines, numbered from oldest to youngest, were determined by least-squares regression of the Universal Transverse Mercator (UTM) coordinates for the samples of each age group (Table 5; Fig. 10). The trends range from N46°E to N80°E, and the time-averaged trend of activity in the BVF is represented by a mean trend of N70°E for the six fields. The age-group trends show an irregular clockwise rotation with time that appears to be the result of a greater northward migration of volcanic activity on the western side of the BVF than to the east. Correlation coefficients (R) for the regressions ranging from 0.37 to 0.89 (Table 5) are sufficiently well defined to establish the generally northeast-southwest trends of the age groups and the change in trend after the earliest activity of age group 1.
The geographic centers of the age groups (Fig. 10) migrate north-northwestward, roughly perpendicular to the N70°E average trend of the age groups. To evaluate the effect of the group selection on this migration, 6 different age groups were selected by dividing the data set into equal groups of 24 samples, except for groups 4 and 5, which were given 25 samples. As expected, group centers are different, but the latitude of each group center is north of the next older center, confirming the northward migration defined by the original grouping. This pattern is also confirmed by compositional grouping of the data, as seen in the following discussion.
To address the magnitude of the north-northwestward migration of the age groups, the distances of each center from some reference must be determined. Calculating the perpendicular distances for each of the age-group centroids from an average trend line (N70°E) through the group 6 center provides snapshots of the average location of volcanism in the BVF during each pulse of activity (Fig. 11). These distances and the average age of each group (Table 6) are regressed to calculate an average rate of north-northwestward migration of volcanism in the BVF between 2.7 Ma and the present (Fig. 12). An average migration rate of 9.3 ± 1.6 m/k.y. (mm/yr) (1σ) is calculated in the N20°W direction, normal to the average N70°E trend of the age groups. It is important that the Gales Creek–Mount Angel, Canby-Molalla, Portland Hills–Clackamas River, and Sandy River faults also have northwestward trends, as shown by the dextral displacement of aeromagnetic anomalies (Blakely et al., 2000).
The northeastward trend of age groups of the BVF is broadly parallel to the northeast trend of mapped normal faults within the Portland region (e.g., Wells et al., 1995; Blakely et al., 2000), and to the most probable trend and slip mechanism that produced the Mw 5.2 November 1962 Portland-Vancouver earthquake (Yelin and Patton, 1991). Local extension normal to these trends is consistent with interpretations of active tectonism in this region during late Pliocene and Quaternary time (Yelin and Patton, 1991; Wells et al., 1998; Blakely et al., 2000), and with the anomalous location of Boring volcanism in the Cascadia forearc. Beeson et al. (1985, 1989) concluded that the Portland Basin is a pull-apart basin, formed east of the Portland Hills fault. Whereas the western margin of the basin fits this model with the northwest-trending, dextral Portland Hills–Clackamas River fault (Yelin and Patton, 1991; Blakely et al., 1995, 2000), a basin-bounding fault on the eastern margin is poorly defined by geologic mapping and the Sandy River fault zone is not at the basin margin. Dextral shearing may be more distributed along the eastern side of the basin and extend still farther east, as a substantial portion of Boring vents extend east of the basin. North-northwestward migration of age groups is less apparent and extension may be slower in the eastern part of the BVF. Because the Portland Basin is not an area of high heat flow typical of volcanic fields supplied by large mantle heat sources (Blackwell et al., 1990a, 1990b), eruption of small-volume lavas in the extensional regimes between rotating crustal blocks may provide an explanation for the anomalous forearc location of the BVF. The pattern of migration of volcanism within the BVF suggests that the locus of greatest extension or pulling apart in the Portland Basin may also have shifted to the north-northwest during the past 2.7 m.y. The similarity of the 9.3 ± 1.6 mm/yr migration rate determined for the Boring lavas to rates of northwestward motion of crustal blocks in the Cascadia forearc (England and Wells, 1991; Wells et al., 1998; McCaffrey et al., 2007), and to extension in the northern Basin and Range province (Magill et al., 1982; Wells et al., 1998), suggests that within-block deformation may be comparable in magnitude to, or even accommodate, the majority of differential motion between blocks. In the case of the BVF, the frequency of volcanism could thus reflect periods of locally greater and lesser extension, or block rotation, with the 1.7–0.5 Ma period being the most rapid and the 2.2–1.7 Ma period representing an interval of reduced extension.
Evolution of Potassium and Calcium in the BVF
Measurements of K and Ca are byproducts of 40Ar/39Ar analyses, but because most analyses used groundmass separates, bulk-rock measurements of K2O and CaO are reported in Table 2. Variations in bulk-rock K2O and CaO for the suite of Boring lavas show an impressive correlation with age with an evolution toward more alkaline magmas (Fig. 13). Because BVF activity migrates in a northward direction, a geographic control of the chemical evolution should be considered. Evarts et al. (2009a) noted that no low-K tholeiites were erupted in the BVF after ca. 1.6 Ma, but the trends defined in Figure 13 show a more coherent pattern of chemical evolution. K2O and K/Ca values increase progressively from the oldest (2.6–2.7 Ma) Boring volcanic rocks to the youngest without abrupt increases or decreases (Fig. 13). K2O increases from an average of ∼0.7% in 2.6 Ma samples to ∼1.1% in the youngest groups. CaO decreases progressively from an average of ∼9% in the 2.6 Ma rocks to ∼8% in the most recent group. Because of the inverse variation in K2O and CaO, the average K/Ca in BVF lavas increases by ∼70%, from ∼0.10 to ∼0.17 during this period. Variations among rocks of the same age are substantial as is apparent from the dispersion in K/Ca, but regressions of the data define clear trends in the plots (Fig. 13).
Additional chemical and isotopic studies are needed to define and distinguish between causes for the correlation of age and composition of the lavas, such as time-dependent changes in depth of melting and melt fraction, crustal involvement, or water content. To aid in evaluating the possible effect of geographic control on the chemical trends, the sample suite was divided equally into two compositional groups, one above and one below the median K2O value (0.9355). As expected from age and chemical trends, samples with K2O values higher than the median (mean 1.213% K2O) are younger (mean of 0.848 Ma versus 1.342 Ma) than those with values below the median (mean 0.752% K2O). The locations of samples in each group are plotted in Figure 14 and show similar areal distributions, although the northernmost samples have above-median K2O values and the most southern have below-median values. The geographic center for the above-median K2O group is 5.3 km due north of the lower group, indicating a 10.7 mm/yr migration rate based on the mean ages of the K2O groups. Considering the grouping and averaging of lavas from multiple periods of volcanic activity, this rate compares favorably with the 9.3 ± 1.6 mm/yr average calculated above from activity-based groupings (Fig. 11). It is significant that this grouping of samples by concentration is completely independent of the criteria used to select the age groupings shown in Figure 8, yet defines a similar migration of volcanism within the BVF.
Tectonic Implications of Migration of BVF Volcanism
McCaffrey et al. (2007) noted that the Juan de Fuca plate currently subducts obliquely northeastward beneath North America at a rate of 30–45 mm/yr with most of the relative plate motion taken up by subduction, but that the overriding continental plate is broken into a series of crustal blocks rotating clockwise about nearby poles in the backarc. The resulting northward motion of the Coast Range, at 7.1 ± 0.5 mm/yr (McCaffrey et al., 2007), is similar to the volcanic migration rate discussed here for the BVF. Paleomagnetic and global positioning system (GPS) measurements also indicate a westward increase in the northward motion of the Coast Ranges that results in their clockwise rotation relative to North America (McCaffrey et al., 2007). The greater north-northwestward migration of volcanic activity determined for the western side of the BVF may reflect this westward increase in dextral shear during the past 2.7 m.y.
Both volcanic activity and magmatic compositional changes of the Boring lavas migrate northward within the BVF and both distributions are broad, spanning much of the volcanic field. Because the distribution of volcanic activity within each of the age-grouped intervals (Fig. 9) covers as much as half of the areal extent of the BVF, crustal extension (Yelin and Patton, 1991; Wells et al., 1998; Blakely et al., 2000) is presumed to have had at least a similar breadth. The increasing alkalinity of the Boring lavas and simultaneous northward migration of activity permit both temporal and geographic interpretations of this evolution. Temporal changes that involve increasing depth of melting and/or decreasing melt fraction, possibly due to cooling, could result in increased potassium and decreased calcium in the lavas. Migration of melting into different source materials could also result in compositional evolution of the lavas. For example, Church et al. (1986) suggested that geographic changes in the age and structure of the crust in this region are responsible for differences in Pb isotopic compositions of sulfide minerals in Tertiary ore deposits emplaced in the Cascade mineral belt. They showed significantly lower 206Pb/204Pb between 44°N and 45°30′N, the zone referred to as the Oregon Embayment. The BVF spans the north edge of this zone and the southern edge of the higher K/Ca, 200–0 ka age group and the northern edge of the lower K/Ca, 2.7–2.1 Ma age group meet at precisely 45°30′N (Fig. 9). This coincidence between Pb isotope and eruptive (age) group boundaries is also coincident with the course of the Columbia River through the Cascades. Additional chemical and isotopic studies of the Boring lavas are clearly needed to evaluate their total compositional variation and petrogenesis, but K/Ca results reported here document a trend within the BVF that is consistent with current plate motion studies.
Volcanism in the BVF occurred during late Pliocene and Pleistocene time. With few exceptions, 40Ar/39Ar age spectra in this study exhibit evidence of nuclear recoil effects of 39Ar and/or 37Ar that are related to the fine-grained character of the lavas. The age effects due to recoil of 39Ar exceed those of 37Ar even in these high-Ca rock types, but most samples show less than ∼5% net recoil loss of either of these species. Age control from paleomagnetic polarity, stratigraphy, and available plateau ages from the large data set of Boring volcanic rocks documents the validity of recoil model ages calculated as the integrated ages of gas fractions in the central portions of age spectra. Use of recoil model ages has broad implications for 40Ar/39Ar studies of volcanic fields of fine-grained basaltic rocks, as these rock types offer fewer options for high-quality mineral separates that are available in more silicic rock types. As demonstrated with the BVF, sample coverage within the volcanic field was expanded by almost 40% through the use of this approach.
Grouped by eruptive episode, the geographic distributions of Boring lava samples define a series of elongate, northeast-southwest trends that appear to progress irregularly in a clockwise sense as their centers migrate northward with time. Age groups based on equal numbers of samples show a similar pattern of northward migration, demonstrating that activity-based groupings have no effect on the geographic trend. Periods of greatest volcanic activity occurred from 2.7 to 2.2 Ma, from 1.7 to 0.5 Ma, and from 350 ka to 50 ka, whereas no record of volcanism was found between 2.2 and 2.0 Ma and between 500 and 350 ka. Beginning ca. 2.7 Ma, volcanic activity in the BVF migrated northward at an average rate of ∼9.3 ± 1.6 mm/yr, with the greatest migration on the western side of the field resulting in an apparent clockwise rotation of the age-group trends.
K2O and CaO values in volcanic rocks of the BVF are well correlated with age. Grouped only by concentration, geographic centers of the low (older) and high (younger) K2O groups support a northward migration of volcanic activity, showing an overall rate similar to that calculated for age groupings of the lavas. Together with the elongate northeast-southwest distributions and clockwise rotation of age-group trends, the calculated migration rates are consistent with recent GPS measurements of relative motion of crustal blocks within the U.S. Pacific Northwest.
We thank Ray Wells for support and patience as this study progressed and the extent of the Boring volcanic field (BVF) and its variations were recognized. The technical assistance of James Saburomaru, Charles Holdsworth, Honore Rowe, Dean Miller, Donald Shamp, and Blair Bridges was critical to this project and is greatly appreciated. Ian Madin participated in mapping and sampling of the BVF. Ray Wells, Mike Clynne, Laura Webb, Jan Lindsay, and an anonymous reviewer provided insightful reviews that led to significant improvements to the manuscript. We thank science editors Carol Frost and Shanaka de Silva and associate editor Rebecca Flowers for advice and editorial assistance. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.
APPENDIX 1. 40Ar/39Ar TECHNIQUES
Insofar as possible, we have obtained samples for 40Ar/39Ar dating from the same outcrops drilled for paleomagnetic analyses. However, many samples from drilled outcrops were less suitable for 40Ar/39Ar dating due to weathering, glassy groundmass, or fine grain size. In these cases we selected the geochronology samples from nearby outcrops of the same flow that were correlated by physical continuity of outcrop (where possible), petrography, and geochemistry. All samples used for geochronology were selected by petrographic examination of the rocks in thin section. Vesicular rocks and those exhibiting alteration of either the groundmass or any mineral phase with sufficient potassium to affect the age were rejected. Use of samples with groundmass plagioclase having minimum dimensions <15–20 μm was restricted, and those of <10 μm were generally avoided. Analyses for this study were performed over a period of 10 yr, during which experimental procedures evolved from using crushed whole-rock samples to groundmass separates and some extraction and data acquisition procedures were modified. The 40Ar/39Ar incremental-heating experiments (e.g., Merrihue and Turner, 1966; Dalrymple and Lanphere, 1974; Fleck et al., 1977) were performed on whole-rock samples and groundmass separates of the Boring lavas. In both cases samples of 250–1000 g were first crushed in a jaw crusher and then in a roller mill to the appropriate grain size. Initially, a size range of 250–355 μm was used, but in more recent preparations the roller mill output was reduced to a size range of 180–250 μm for ease in eliminating phenocryst phases. Use of a disk grinder was avoided due to the greater frictional heating that might occur. The crushed materials were washed with tap water and rinsed with deionized water. More recent groundmass separates have also been treated in an ultrasonic bath to remove loose and fine particulates, as well as breaking up and removing some fine glassy selvages within the groundmass. Although uncommon, interstitial carbonate was removed when present with cold 1N HCl followed by a minimum of 3 rinses with deionized water in a sonic bath and resizing. Groundmass separates were prepared by removing the most highly magnetic grains with a hand magnet or by Frantz isodynamic separator at the lowest setting. Depending on the sample mineralogy, olivine and/or coarse plagioclase was removed in a less magnetic fraction with the Frantz. Where magnetic separations were ineffective, generally due to the abundance of finely dispersed magnetite, heavy liquid separation with appropriate densities of acetone-diluted methylene iodide was employed. Groundmass concentrates were then hand-picked for purity.
All samples for 40Ar/39Ar incremental heating were packaged in Cu-foil cylinders and packed compactly in quartz vials for neutron irradiation. Positions of the samples in the vials were measured precisely from flatbed scanner images of the vials after evacuation and sealing. Heights of the top and bottom surfaces of each sample cylinder in the vial were measured at the vial centerline with digital graphics software using 2 images of the vial taken 180° opposite each other. In this manner the average height of any two diametrically opposite points at the edge yields an accurate height for the center of the upper and lower surfaces of the cylindrical sample packet, regardless of any slope on these surfaces relative to the long axis of the vial.
Uncertainties are reported at one standard error of the mean except where stated. Uncertainties in J are typically <±0.2%.
The argon extraction system for releasing argon and removing reactive gases utilizes a temperature-controlled, molybdenum-shielded resistance furnace with a molybdenum crucible and SAES ST-172 getters. The gas is exposed to a 4 A tungsten filament and 125 K cold finger during purification. Argon is extracted from the samples in temperature steps selected to distribute the total released over a minimum of ∼8 steps and extend to temperatures where the potassium-derived 39Ar drops to small fractions of the total released, but some extractions may exceed 15 steps. Early in this study stepwise heating included all argon remaining in each sample following a 200 °C, 12–24 h bakeout of the samples. This included large amounts of atmospheric Ar in the first steps, as release of radiogenic 40Ar and potassium-derived 39Ar is not observed from basalts and andesites under routine laboratory heating times of 5–15 min until temperatures of ∼500 °C are reached. Because these low-temperature steps may introduce large amounts of atmospheric argon, hydrocarbons, N2, O2, H2O, and CO2 to the mass spectrometer without adding additional information to the analysis, degassing of the samples to the pumping system at ∼500 °C was adopted. Argon isotopic analyses were carried out using an MAP216 single-collector mass spectrometer with a Bauer-Signer source, a Johnson MM1 electron multiplier, and peak stepping. Analyses are automated, using machine- and BASIC-language software developed by Brent Dalrymple for the earlier samples and LabView software developed by Andrew Calvert since 2003. Data reduction was done using routines developed by these workers, and with Isoplot (developed by Ludwig, 2003, with updates through 2009).
APPENDIX 2. PALEOMAGNETIC TECHNIQUES
Oriented samples of Boring lavas and shallow intrusions were collected for paleomagnetic analysis primarily from artificial (e.g., road cut, quarry) and stream-bank exposures in order to obtain the freshest possible rock. As previously mentioned, 40Ar/39Ar and paleomagnetic samples were collected within the same units, and every attempt was made to collect both sample sets from the same outcrop. This was not always possible, however, due to the differing rock properties advantageous to each technique (see Appendix 1). Between 8 and 12 paleomagnetic core samples 2.5 cm in diameter were drilled at each site using a portable gasoline-powered drill, and were oriented and marked, prior to removal, using an orienting tool consisting of magnetic and solar compasses and a clinometer. Oriented hand samples were also collected at remote and inaccessible sites, to determine a unit’s polarity, from which core samples were drilled in the laboratory.
Typically, three 2.2-cm-long specimens were cut from each core sample in the laboratory, and the lowermost specimen (i.e., least weathered) was subjected to progressive alternating field (AF) demagnetization to determine the stability and structure of the specimen’s natural remanent magnetization. The laboratory work spanned the years from 1983 to 2010, during which the AF demagnetization procedures changed. Initially, 1 or 2 specimens from a given site were subjected to pilot demagnetization of 5 or more steps at peak inductions of up to 100 mT, and the remaining specimens were treated with a single or blanket demagnetization, usually at peak inductions between 30 and 50 mT, to remove any spurious or secondary components of magnetization identified in the pilot demagnetizations.
By 2010, all specimens were being subjected to 3–5 demagnetization steps of up to 80 mT, and specimen-level characteristic directions were determined using a least-squares method based on principal component analysis (Kirschvink, 1980). Throughout, site-mean directions were calculated using Fisher (1953) statistics, and sample directions beyond two angular standard deviations of the mean were omitted from the site-mean calculation. In only one case (site 4T208; Table 4), intersecting demagnetization planes and Bingham statistics (Onstott, 1980) were used to determine the site’s mean direction. Where possible, multiple core samples were drilled and analyzed from the hand samples, but these results are considered less reliable because of the small number of specimens.