Western Australian Argon Isotope Facility, Department of Applied Geology and John de Laeter Centre for Isotope Research, Curtin University of Technology, GPO Box U1987, Perth, Western Australia 6845, Australia
Berkeley Geochronology Center, 2455 Ridge Road, Berkeley, California 94709, USA
 The Hawaii Scientific Drilling Project recovered core from a 3.5 km deep hole from the flank of Mauna Kea volcano, providing a long, essentially continuous record of the volcano's physical and petrologic development that has been used to infer the chemical and physical characteristics of the Hawaiian mantle plume. Determining a precise accumulation rate via 40Ar/39Ar dating of the shield-stage tholeiites, which constitute 95–98% of the volcano's volume is challenging. We applied40Ar/39Ar dating using laser- and furnace-heating in two laboratories (Berkeley and Curtin) to samples of two lava flows from deep in the core (∼3.3 km). All determinations yield concordant isochron ages, ranging from 612 ± 159 to 871 ± 302 ka (2σ; with P ≥ 0.90). The combined data yield an age of 681 ± 120 ka (P = 0.77) for pillow lavas near the bottom of the core. This new age, when regressed with 40Ar/39Ar isochron ages previously obtained for tholeiites higher in the core, defines a constant accumulation rate of 8.4 ± 2.6 m/ka that can be used to interpolate the ages of the tholeiites in the HSDP core with a mean uncertainty of about ±83 ka. For example at ∼3300 mbsl, the age of 664 ± 83 ka estimated from the regression diverges at the 95% confidence level from the age of 550 ka obtained from the numerical model of DePaolo and Stolper (1996). The new data have implications for the timescale of the growth of Hawaiian volcanoes, the paleomagnetic record in the core, and the dynamics of the Hawaiian mantle plume.
 The Hawaii-Emperor seamount and island chain, generally regarded as the archetype of a hot spot track, forms as the Pacific plate moves over the Hawaiian hot spot magma source. As such, the Hawaii islands (Figure 1) have received much attention from the scientific community seeking to understand the origin, composition, evolution and dynamics of deep-seated mantle plumes [e.g.,Frey et al., 1991; Dixon et al., 1997].
 In 1999, the Hawaii Scientific Drilling Project recovered a core (HSDP-2) that reached a depth of 3097.7 m below sea level (mbsl). In 2006 and 2008, the same drill hole was extended to a depth of 3505.7 mbsl. Overall, the HSDP-2 drill core intersected eruptive products from Mauna Kea (Figure 1), thereby providing a comprehensive chemical and isotopic time series of unprecedented duration for the evolution of a Hawaiian volcano [Sharp and Renne, 2005]. However, it has been suggested, mostly based on elemental and isotopic data, that the eruptive sequence recovered by the HSDP-2 core consists not only of Mauna Kea's eruptive products but also includes interfingered lavas coming from a different volcanic center, possibly the Kohola volcano [Blichert-Toft and Albarède, 2009; Lipman and Calvert, 2011].
 The accumulation rate of the subaerial post-shield alkalic lavas of Mauna Kea is well defined with a rate of 0.9 ± 0.4 m/ka based on relatively precise40Ar/39Ar age constraints [Sharp et al., 1996]. However, this sequence erupted when the volcano had moved away from the rapid melting zone and includes only 2–5% of the total volume of the Mauna Kea volcano (Figure 2). Most of Mauna Kea, like other Hawaiian volcanoes, consists of tholeiitic lavas erupted during the shield stage (95–98% of the volume), when the volcano was above the zone of rapid mantle melting [DePaolo and Stolper, 1996].
DePaolo and Stolper constructed a simplified geologic model of Hawaiian volcano formation (hereafter, D-S model) and proposed that an accumulation rate of 15–21 m/ka would be observed at the core site during Mauna Kea's shield formation. In contrast,Sharp and Renne  observed a substantially lower accumulation rate of 9 ± 3 m/ka based on 40Ar/39Ar ages of the HSDP-2 core. Due to large uncertainties on the published ages, however, all but one of their ages are consistent with the proposed higher rates. Only the deepest (2.8 kmbsl) published age is distinct at the 95% confidence level from the high eruption rate model, though the persistent offset of the40Ar/39Ar ages to values older than the D-S model is more consistent with slower accumulation rates.
 In order to further refine the age-depth relations in the deep part of the HSDP cores, we have intensively applied the40Ar/39Ar dating technique to two samples obtained during the final leg of drilling, from the deepest part of the HSDP-2 core. The samples, from depths of 3278 mbsl and 3321 mbsl, provide new dates near the bottom of the core where the ages predicted by the D-S model (Figure 2) and the ages predicted by extrapolating the previous 40Ar/39Ar dates diverge most significantly, thereby facilitating comparison of the two proposed age-depth models. To evaluate possible sources of bias in the new dates, we analyzed several replicates of each sample using laser and furnace gas extraction methods carried out in two different laboratories. The resulting data are statistically combined and integrated with the ages of the other Mauna Kea core samples.
2. Previous Geochronology
 A significant number of absolute ages have been obtained on the main Hawaii Island volcanic rocks using K/Ar; 40Ar/39Ar and 14C dating [e.g., McDougall, 1969; McDougall and Swanson, 1972; Wolf and Morris, 1996; Cousens et al., 2003; Calvert and Lanphere, 2006; Aciego et al., 2010; Lipman and Calvert, 2011] and more recently by U-Th/He dating [Aciego et al., 2010]. However, most of these ages were obtained on post-shield late Quaternary lavas sampled at the surface of the Island. The HSDP project has brought to the surface ∼3.5 km of continuous core, which has been previously investigated by40Ar/39Ar dating over a significant part of its length [Sharp et al., 1996; Sharp and Renne, 2005]. The core allows probing the evolution of the Island further back in time, but isotopic dating of Hawaiian tholeiites is extremely challenging due to their late Quaternary age, low K2O content (K2O ≈ 0.4 ± 0.1 wt%) [Rhodes and Vollinger, 2004], high concentrations of trapped argon [Sharp and Renne, 2005] and sensitivity to alteration. Samples from deep in the core were erupted and cooled below a deep water column at elevated pressures (40–50 bars) [Stolper et al., 2004]. Such samples retain high levels of volatiles, including up to 0.8% of H2O, 50–100 ppm of CO2 [Seaman et al., 2004] and potentially high levels of mantle-derived argon. From a technical point of view, young volatile-rich samples require adequate purification of the gas to remove any active gas before Ar isotopic measurements in a mass spectrometer. Previous40Ar/39Ar results and approaches have been extensively described by Sharp et al.  and Sharp and Renne  who preferred to use the isochron approach, which determines the trapped (or initial) 40Ar/36Ar from the data for each sample rather than correcting for trapped 40Ar by inappropriately assuming an “atmospheric” 40Ar/36Ar ratio.
 This study is focused on the tholeiitic lavas from the HSDP-2b core. The tholeiitic lavas from the HSDP-1 and HSDP-2 cores (Figures 1 and 2) previously analyzed come from 5 distinct levels and yielded 7 isochron ages ranging from 329 ± 86 ka (357 mbsl) to 674 ± 88 ka (2789 mbsl). All these previous ages have significant uncertainties, ranging from ±67 to ±322 ka (2σ) and leading to large uncertainties in the calculated accumulation rate of 8.6 ± 3.1 m/ka for the shield stage tholeiites of the core. Recent 40Ar/39Ar data showed that the ∼1150 ka Hilo Ridge belongs to the Kohala volcano and suggest that the volume of the Mauna Kea volcano may be smaller than previously thought [Lipman and Calvert, 2011]. Geochemical and isotopic variations down the HSDP-2 core have been attributed to interfingering of a concurrently growing “lost” volcano [Blichert-Toft and Albarède, 2009], possibly the early phase of the Kohala volcano [Lipman and Calvert, 2011]. Therefore, the eruption rate calculated along the HSDP-2 core reflects the growth rate of the eruptive sequence at Hilo, but represents only a maximum eruption rate for the Mauna Kea volcano itself.
3. Sample Description and Methodology
 The two dated lavas are from HSDP-2 unit 14/72 at a depth of 3278 mbsl and unit 19/67 at a depth of 3321 mbsl (Figures 1 and 2). Unit 14/72 is a massive flow of moderately olivine-phryric basalt and unit 19/67 is a pillowed flow of aphyric to sparsely olivine phyric basalt. After crushing, the samples were rinsed with distilled water in an ultrasonic cleaner and inspected using a binocular stereomicroscope. The grains consist of groundmass (including glass, olivine, pyroxene and plagioclase) and olivine and plagioclase phenocrysts. Alteration in lavas of the HSDP core mostly results in K2O and/or 40Ar* loss and crystallization of younger alteration phases [Sharp et al., 1996; Sharp and Renne, 2005]. Such effects can produce apparent ages that may be either too old or too young and must be monitored carefully. The two samples generally do not show visible alteration (e.g., contains mostly fresh olivine crystals) and the few grains with palagonite rims were removed by careful handpicking. The samples were chemically leached for one minute in 2N HF to remove any superficial and crack-filling alteration phases and then rinsed with distilled water in an ultrasonic cleaner. Phenocrysts of olivine and plagioclase were removed as they constitute a potential source of excess40Ar* [Sharp et al., 1996].
3.1. Analytical Procedure at the Berkeley Geochronology Center (BGC)
 One irradiation of 30 min duration (#I352FJ) was performed in 2007 in the Cd-shielded CLICIT facility of the TRIGA reactor at Oregon State University, U.S.A. Samples and standards were placed in wells in machined aluminum discs during irradiation thereby maintaining well-constrained spatial relations and facilitating accurate measurement of the neutron fluence received by each sample. We calculated J-values relative to an age of Alder Creek sanidine (ACs) of 1.193 ± 0.001 Ma [Nomade et al., 2005] and using the decay constants of Steiger and Jäger . Computed J-values are, respectively, 0.00006910 ± 0.0000005 (0.67%) and 0.0026644 ± 0.00006900 (0.30%) for samples HSDP2 14/72 and HSDP2 19/67. We corrected for interfering isotopes (minimized by Cd shielding) using the following factors: (39Ar/37Ar)Ca = (7.60 ± 0.09) × 10−4; (36Ar/37Ar)Ca = (2.70 ± 0.02) × 10−4; and (40Ar/39Ar)K = (7.30 ± 0.90) × 10−4, which are derived from the weighted mean values of 10 years of measurements at the Berkeley Geochronology Center (BGC) of K-Fe and CaSi2 glasses and CaF2 fluorite irradiated at the Oregon State TRIGA reactor. 40Ar/39Ar analyses were performed at the BGC in 2007, using a CO2laser on ∼150 mg of material per sample. To promote uniform heating, the laser beam was enlarged and recombined to yield a 6 × 6 mm beam of nearly uniform intensity, samples were distributed 1–2 grains deep in 5 mm-wide troughs up to 50 mm long, and heated for 90 s during each step by scanning the laser beam. The gas was purified in a stainless steel extraction line using two C-50 SAES getters and a cryogenic condensation trap. Ar isotopes were measured in static mode using a MAP 215–50 mass spectrometer with a Balzers electron multiplier using ∼10 cycles of peak hopping. A more complete description of the mass spectrometer and extraction line is given byRenne et al. . Blank measurements were generally obtained before and after every three sample runs. Mass discrimination (D) was calculated assuming a power law relationship between D and atomic mass [e.g., Renne et al., 2009] and was monitored several times daily. During this study, we used a 40Ar/36Ar value of 295.5 [Nier, 1950] in our calculation (mass spectrometer discrimination and data reduction). We note that (presumably) more accurate values of atmospheric 40Ar/36Ar have been recently determined by Lee et al.  and Valkiers et al. , but using either value will have negligible influence on the resulting ages provided that both the standard and unknown data have been normalized to the same value of the 40Ar/36Ar ratio of atmospheric argon [Renne et al., 2009]. Mean values of D per atomic mass unit of 1.0063 ± 0.0022 and 1.0068 ± 0.0024, respectively, were applied to the analyses of HSDP-2 14/72 and HSDP-2 19/67.
 Ar isotope ratios extrapolated to their values upon inlet to the mass spectrometer and corrected for blank, mass discrimination, and radioactive decay since irradiation are given in Table S1 in the auxiliary material, with errors given at the 1σ level and summarized in Table 1.
Table 1. Summary of Plateau and Inverse Isochron Ages for Lava-Flow Analysesa
Plateau Age (ka, ±2σ)
Total 39Ar Released (%)
Inv. Isochron Age (ka, ±2σ)
40Ar/36Ar Intercept (±2σ)
Preferred Age (ka, ±2σ)
MSWD and probability of fit (P) for plateau and isochron, percentage of 39Ar degassed used in the plateau calculation, number of analyses included in the isochron, and 40Ar/36Ar intercept are indicated. Analytical uncertainties on the ages are quoted at 2 sigma (2σ) confidence levels. Preferred age for each sample is indicated in bold.
1280 ± 180
680 ± 360
300 ± 3
680 ± 360
1120 ± 180
760 ± 360
299 ± 4
760 ± 360
298 ± 484
994 ± 537
286 ± 9
871 ± 302
224 ± 214
764 ± 351
289 ± 10
413 ± 130
608 ± 289
290 ± 5
612 ± 159
339 ± 312
594 ± 198
291 ± 4
Weighted mean of the 4 preferred ages
681 ± 120
3.2. Analytical Procedure at the Western Australia Argon Isotope Facility (WAAIF)
 Samples and ACs standards were irradiated in Al discs as at BGC, were Cd-shielded (to minimize undesirable nuclear interference reactions), and irradiated in 2009 for 15 min in the Hamilton McMaster University nuclear reactor (Canada) in position 5C. Mean J-values varied from 0.00009204 ± 0.00000060 (0.65%) to 0.00009302 ± 0.00000060 (0.65%). The mean value of D was 1.0014 ± 0.0031 per amu. We used correction factors for interfering isotopes of (39Ar/37Ar)Ca = 7.30 × 10−4 (±11%), (36Ar/37Ar)Ca = 2.82 × 10−4 (±1%), and (40Ar/39Ar)K = 6.76 × 10−4 (±32%).
 The 40Ar/39Ar analyses were carried out in 2009 at the Western Australian Argon Isotope Facility at Curtin University, operated by a consortium consisting of Curtin University and the University of Western Australia. The samples, consisting of 300 mg of groundmass per replicate for each sample were step-heated using a Pond Engineering® double vacuum resistance furnace. The gas was purified for 10 min in a stainless steel extraction line using a GP50 and two AP10 SAES getters and a liquid nitrogen condensation trap. Ar isotopes were measured in static mode using a MAP 215–50 mass spectrometer (resolution of ∼600; sensitivity of 2 × 10−14mol/V) with a Balzers SEV 217 electron multiplier mostly using 9 to 10 cycles of peak-hopping. The data acquisition was performed with the Argus program written by M.O. McWilliams and ran under a LabView environment. The raw data were processed using the ArArCALC software [Koppers, 2002]. Full system blanks were monitored every 2 samples. Blanks, constants and Ar isotopic data are given in Table S2 in the auxiliary material, with errors given at the 1σ level.
3.3. Isochron and Age Criteria
 Our criteria for the determination of inverse isochron ages are as follows: isochrons must include at least 70% of the released 39Ar and should include a minimum of 3 consecutive steps. Inverse isochrons include the maximum number of consecutive steps with a probability of fit ≥0.05. We calculated isochron ages relative to an age of ACs of 1.193 ± 0.001 Ma [Nomade et al., 2005] for ACs and using the decay constants of Steiger and Jäger . For the sake of comparison between the new and old data, we did not recalculate the ages of the samples and monitor relative to the new 40K constants proposed by Renne et al. , but we note that that this would result only in an increase of ∼1% (i.e., ∼6–7 ka), negligible compared to the other uncertainties associated with our ages (i.e., ≥±160 ka). The uncertainties in the 40Ar*/39Ar ratios of the monitors are included in the calculation of the integrated and plateau age uncertainties, but not the errors on the age of the monitor and on the decay constant (internal errors only, see discussion by Min et al. ). This is appropriate to compare all the 40Ar/39Ar ages along the HSDP2 core. Detailed 40Ar/39Ar results are shown in Table S2 in the auxiliary material and summarized in Table 1.
 Groundmass separates from two basaltic lavas were each analyzed in triplicate via the 40Ar/39Ar incremental heating technique using broad-beam laser heating at BGC (two analyses) and furnace heating (four analyses) at the WAAIF.
4.1. Results From the BGC
 Samples HSDP2 14/72 and 19/67 yield two plateau ages of 1280 ± 180 ka (MSWD = 0.88; P = 0.59) and 1120 ± 180 ka (MSWD = 0.29; P = 0.99) (Figure 3). High temperature steps of sample 19/67 depart from the plateau and increase toward an apparent age of ∼6 Ma, and Ca/K ratios gradually increase in the same high temperature heating steps, suggesting the presence of excess 40Ar in retentive, Ca-rich minerals such as clinopyroxene. Samples HSDP2 14/72 and 19/67 give two inverse isochron ages of 680 ± 360 ka (2σ; MSWD = 0.17; P = 1.0) over 17 steps and 760 ± 360 ka (MSWD = 0.035; P = 1.0) over 13 steps, respectively (Figure 4). The samples have 40Ar/36Ar trapped ratios of 300 ± 3 and 299 ± 4, respectively, which are biased toward supra-atmospheric values and indicate that the use of the inverse isochron age calculation is appropriate. The large uncertainty on the two ages is due to the extremely low concentration of radiogenic40Ar (40Ar*) in each samples with a spreading factor (S-factor) [Jourdan et al., 2009] of only ∼3% along the isochron (Figure 4) and a radiogenic composition (40Ar*) varying from, respectively, 1.9% to 4.5% and 1.5 to 6.1% for samples 14/72 and 19/67.
4.2. Results From the WAAIF
 Sample HSDP2 14/72 was analyzed twice and yields two plateaus with apparent ages of 224 ± 214 ka (MSWD = 0.62; P = 0.80) and 298 ± 484 ka (MSWD = 0.51; P = 0.94). Sample HSDP2 19/67 was also analyzed twice and gives two plateau ages of 413 ± 130 ka (MSWD = 0.34; P = 0.98) and 339 ± 312 ka (MSWD = 0.21; P = 0.99). The Ca/K ratios of the four analyses generally increase toward high temperature although three of the four analyses describe a more complex Ca/K pattern (Figure 5).
 Two inverse isochron ages of 994 ± 537 ka (MSWD = 0.48; P = 0.95) and 764 ± 351 ka (MSWD = 0.93; P = 0.54) with respective 40Ar/36Ar intercepts of 286 ± 9 and 289 ± 10 were obtained on Sample HSDP2 14/72 (Figure 6). The replicate analyses have been combined on a single inverse isochron resulting in an age of 871 ± 302 ka (MSWD = 0.70; P = 0.90) over 36 heating steps and a 40Ar/36Ar trapped ratio of 287 ± 6 (Figure 7). The S-factors is 4% and the radiogenic content varies from 0% to 5.8%.
 Sample HSDP2 19/67 yielded two inverse isochron ages of 608 ± 289 ka (MSWD = 0.74; P = 0.75) and 594 ± 198 ka (MSWD = 0.93; P = 0.54) along with respective 40Ar/36Ar intercepts of 290 ± 5 and 291 ± 4 (Figure 6). We combine the two replicates on a global inverse isochron that gives an age of 612 ± 159 ka (MSWD = 0.60; P = 0.96) over 32 steps and an intercept 40Ar/36Ar ratio of 291 ± 3 (Figure 7). The significantly higher precision of this age is due to a larger S-factor of 8%, reflecting the greater variation in the radiogenic40Ar content of this sample, which ranges from 0% to 6.7%. Sample 19/67, when analyzed at the WAAIF, yielded a larger S-value than at BGC, likely due to more homogeneous heating in the WAAIF furnace compared to the laser at the BGC. During heating, the laser tends to extract and mix the argon gas from different retentive domains, in other words, each grains sees a slightly different temperature depending on the position of the grain relatively to the laser spot power distribution. As such, furnace heating allows better probing of the radiogenic end-member of a sample than the laser as all the grains are heated at the same temperature at any given step, albeit at the cost of somewhat larger blanks. As the intercept ratios of samples HSDP2 14/72 and HSDP2 19/67 are sub-atmospheric, we consider the two global (and statistically indistinguishable) isochron ages of 871 ± 302 ka and 612 ± 159 ka, respectively, to be the preferred age for these samples.
4.3. Influence of Heat Gradient on Isotopic Ages?
 One of the important questions arisen during the course of this study was the possible influence of the Hawaii Island heat gradient on isotopic ages. The Hawaii Island is volcanically active, and thus, it raises the question of whether the ambient temperature at a depth of few thousand mbsl would be hot enough to progressively open the Ar chronometer of the groundmass (i.e., glass and plagioclase microlites) samples, hence leading to an overestimated growth rate. However, previous studies [Büttner and Huenges, 2003] have shown that the gradient measured within the HSDP-2 borehole is negative during the first 600 mbsl, steady up to 1000 mbsl and then increase of about 18°C per km. Hence the temperature reached at 3300 mbsl (this study) and at the bottom of the lava pile (∼5500 mbsl) are only ∼45°C and ∼80°C, respectively, significantly too cold to reach the Ar closure temperature of K-bearing components of basaltic rock.
5.1. Age of Mauna Kea's Lavas at 3.1 km Depth in the HSDP-2 Core
 The four isochrons, some of which combine several replicate analyses, yielded concordant ages ranging from 612 ± 159 to 871 ± 302 ka (2σ). This yields two weighted mean ages of 792 ± 230 and 636 ± 150 ka for samples 14/72 (3278 mbsl) and 19/67 (3321 mbsl). The two samples come from two different levels of the lava pile and thus cannot be combined in a single inverse isochron since they may contain trapped Ar of distinct compositions. However, the vertical distance between the two lava flows samples is relatively small (∼50 m) and the time span between the two eruptions can be estimated at roughly ∼6 ka using an eruption rate of 9 m/ka [Sharp and Renne, 2005]. As the age difference between the two eruptions is negligible compared to the uncertainties of our ages, we combined the four inverse isochron ages (i.e., using one age per sample and per laboratory; Figure 8) in a single weighted mean age of 681 ± 120 ka (MSWD = 0.84; P = 0.47) for a single depth of 3300 mbsl. This approach has the advantage of reducing the age uncertainties by increasing the counting statistics (i.e., number of ages) at ∼3.3 km-depth. Determining weighted mean ages for samples 14/72 (729 ± 230 ka) and 19/67 (636 ± 150 ka) first, and computing the weighted mean of the two lava flows yields an indistinguishable age of 683 ± 130 ka. We therefore propose that the ∼3.3 km-deep lava flows erupted sub-synchronously at 681 ± 120 ka, with the age difference between the two lava flows being unresolvable.
5.2. Distinct 40Ar/36Ar Trapped Ratios?
 The samples run at the BGC and WAAIF come from the same preparation splits and hence should yield the same trapped 40Ar/36Ar ratios within error. Yet despite yielding indistinguishable isochron ages, the results from the two laboratories yielded distinct intercept 40Ar/36Ar ratios of 300 ± 3 (2σ) and 299 ± 4 (BGC) and, 287 ± 6 and 291 ± 3 (WAAIF), respectively a few units above and below the atmospheric ratio of 295.5 used in this study. Whereas samples containing excess 40Ar are common [e.g., Kelley, 2002], samples with sub-atmospheric trapped Ar ratios (i.e.,40Ar/36Ar < 295.5) are rarer. Sub-atmospheric40Ar/36Ar ratios can be found in recent lavas and are explained by mass-dependent kinetic isotope fractionation during the emplacement of these lavas [e.g.,Matsumoto and Kobayashi, 1995; Ozawa et al., 2006; Morgan et al., 2009; Renne et al., 2009]. Another process that can produce sub-atmospheric Ar isotopic ratios is the percolation of36Ar-rich hot fluids that will simultaneously alter and introduce36Ar in the groundmass [e.g., Baksi, 2007] and/or leach 40Ar* out of the system. Subatmospheric apparent 40Ar/36Ar may also result from the presence of residual Cl in the sample gas, as H35Cl is an isobar for 36Ar whose resolution is far beyond the capability of either of the two instruments. This isobar, if present, would presumably be correlated more with the trapped than the radiogenic component, hence would affect age intercepts less than 40Ar/36Ar intercepts.
 The sample preparations conducted in the two laboratories were similar (e.g., HF leaching, grain picking and phenocrysts removal), but were carried out two years apart and involve 300 mg of material per experiment at the WAAIF compared to 150 mg at the BGC. A possible explanation is that olivine phenocrysts present in the samples, possibly containing minor excess 40Ar*, have been more efficiently eliminated during the second sample preparation, as illustrated by the age increase at high-temperature steps, observed for sample HSDP2 19/67 (Figure 3). However, all the separated material was consumed during the analyses, rendering this hypothesis impossible to verify. Another possibility is that the aliquots run at the WAAIF were more altered than those run at the BGC, yielding a sub-atmospheric40Ar/36Ar ratio for the former. If the latter process is indeed responsible for the discrepancy, this would imply that the final weighted mean age obtained on these samples is a minimum age and that the true emplacement age should be slightly older. However, as said above, the sample preparation is likely to be very similar as in both cases was carefully carried out by the first author of this paper on the same splits and the isochron does not show any departure from a 2 reservoir mixing line suggesting that another process is responsible for the distinct trapped ratios. A third possible explanation is a different fractionation induced by the extraction line/mass spectrometer system of the two different laboratories, but repeated air pipette measurements over several months before and after experiments show no departure from the discrimination values of ∼1.0065 and ∼1.0014 for the BGC and WAAIF respectively. However, these values are obtained at ambient temperature and do not account for any possible internal fractionation due to the furnace or the laser extraction system. In both laboratories, several unrelated samples ran close to the samples analyzed during the course of this study yielded trapped 40Ar/36Ar ratios of atmospheric composition seemingly ruling out a systematic bias of one of the laboratory. Hence, the reason for the discrepancy between the different 40Ar/36Ar trapped ratios of the trapped Ar is not completely understood. However, it is important to note that whether the isotopic fractionation is geologically or laboratory-induced, the resulting40Ar/36Ar intercept ratio is included in the age calculation of both samples and that the ages obtained are identical within error, suggesting that this process does not affect the outcome of these analyses.
5.3. Eruption Rate of the “Mauna Kea” Shield Phase
 Our new age of 681 ± 120 ka has been used in conjunction with the seven previous 40Ar/39Ar isochron ages ranging from 329 ± 86 ka (357 mbsl) to 674 ± 88 ka (2789 mbsl) that are distributed along the tholeiitic stage along the core. The combination of these ages is shown in Figure 6 and yield an integrated linear growth rate of 8.4 ± 2.6 m/ka (MSWD = 1.17; P = 0.32), indistinguishable from the rate proposed by Sharp and Renne , but significantly lower than the growth rate of 15–20 m/ka suggested by the D-S model [DePaolo and Stolper, 1996].
 We note that Farnetani and Hofmann  have constructed an alternative model for Hawaiian volcano growth that predicts an accumulation rate of 10–15 m/ka, the lower range of which is consistent with our measured rate. Our measured rate is also similar to the ∼10 m/ka estimated for Kilauea tholeiitic lavas [Lipman et al., 2006].
 Nevertheless, the eruption rates for Mauna Kea mentioned above rely on the assumption that all eruptive products sampled along the HSDP2 come exclusively from the Mauna Kea volcano. As mentioned previously, there is evidence that below ∼1950 mbsl, the eruptive products come from at least two distinct volcanic centers. Therefore, as four of the seven ages used to calculate a growth rate of ∼8.4 m/ka are below ∼1950 mbsl, this rate is in fact a maximum value for the eruption rate of Mauna Kea and rather describes the total accumulation rate in the core hole. The statistical test (P = 0.32) associated with our eruption rate calculation suggest that all the age data define a linear accumulation rate within error, although the large uncertainties associated with the isotopic ages could mask some level of complexity, e.g., a faster accumulation rate below 1950 mbsl due to combined contributions from Mauna Kea and another volcanic center [e.g., Lipman and Calvert, 2011]. Nevertheless, all dates are consistent with a constant rate. Along with the isotopic data of Blichert-Toft and Albarède , these observations suggest that the proportion of volcanic products from sources other than Mauna Kea is relatively small.
5.4. Ages Along the HSDP-2 Core
 Adopting a linear eruption rate throughout most of the shield phase, a modeled regression age at a given depth can be back calculated using the following first-degree equation:
Where D is the depth in m and t is the eruption age in ka.
 The 2σ uncertainty on the age at any given depth, ignoring correlation of errors between the ages, is given by:
The 2σ error on the regression age varies between ±81 Ma (youngest samples) and ±86 Ma (onset of the shield stage), and is smaller than the errors on the individual ages. This illustrates the advantage of integrating a data set containing several age measurements. In addition, pooling ages on a regression curve has the benefit of including all data from the core and deriving a mean rate independent of minor local rate fluctuations.
 For example, the eruption age at 3.3 km derived from the regression is 664 ± 83 ka (2σ). Of course this approach is valid only if the eruption rate is approximately linear, but a linear rate is statistically compatible with the eight age data obtained along the core (P = 0.32) (Figures 2 and 9). If we extrapolate the measured eruption rate to the seafloor (∼6000 mbsl), the age of the basal tholeiitic lavas beneath the core hole can be estimated to be 985 ± 88 ka. From this calculation, the maximum duration of the tholeiitic stage of Mauna Kea is estimated to be ∼750 ka. Following Lipman and Calvert  and using an (early) alkalic stage duration of ∼150 ka, similar to what is observed for Kilauea, we can estimate that the onset of the volcanism on the Cretaceous seafloor occurred between ca. 1050 and 1250 ka.
 At 3300 mbsl, our core-integrated model age of 664 ± 83 ka is statistically indistinguishable from our new isotopic age, but diverges at the 95% confidence level from the D-S model which predicts a younger age of ∼550 ka (no error margin assigned;Figure 2). The D-S model was based on several parameters, such as the volume of the magma capture, that are difficult to confidently estimate with precision. Another source of the discrepancy could be a slightly higher density of the submarine section compared to the subaerial part, not accounted for in the model [Bryce et al., 2005]. In any case, the 40Ar/39Ar ages obtained from the HSDP-2 core suggest that the physical and chemical evolution of Mauna Kea at the bottom of the volcanic pile and observable in the HSDP-2 core took place on a slightly longer timescale than the one initially postulated by the D-S model and is in better agreement with the more recent model ofFarnetani and Hofmann . In addition, we note that the accumulation rate determined in this study is in agreement with the approximated rate of ∼7.6 m/ka calculated using the ages of two magnetic excursions identified in the first 1000 mbsl of the core [Steveling et al., 2003].
5.5. Implications for Studies of the HSDP-2 Core
 The age-depth relations determined in this study have implications for the calculation of the scale length of the isotopic variations in the Hawaiian plume. Based on helium, neodymium, strontium, osmium, hafnium, and lead isotopic ratios,Bryce et al.  showed that isotopic values drift through time (e.g., ϵNd values vary from +7.5 to +6.5) and that it is possible to calculate the size of the isotopic heterogeneities of the Hawaiian mantle plume based on a time-dependent calculation involving the isotopic compositions along the HSDP-2 core.Bryce et al. plotted the possible size of isotopic anomalies as a function of the age and composition of the samples. The age they used for each sample was based on the theoretical D-S age model which, as we propose in this study and was previously proposed bySharp and Renne , underestimates the age of the deep section of the tholeiitic stage. As an example, we have replotted in Figure 10 the isotopic data measured by Bryce et al. [2005, and references therein] using the age of the magmatic pile derived from the present study (Figures 2 and 9). Although the shift of the data is not prominent, when it is compared to the size-model curves of isotopic heterogeneity, it implies that the size of the isotopic anomalies is likely to be few km smaller than previous hypothesized (e.g.,3He/4He and 87Sr/86Sr plots). The later hypothesis is in agreement, although to a lesser extent, with the study of the Pu‘u ‘Ō eruption (Kilauea volcano), suggesting that the plume heterogeneities are smaller than estimated using the HSDP core data alone [Marske et al., 2008].
 Finally, these new ages also have implications for other time-dependent studies carried out along the HSDP-2 core such as, for example, the measurements of the paleomagnetic record in the core [Steveling et al., 2003] and the paleomagnetic field intensity [e.g., Weiss et al., 2007] and might be inverted to adjust the parameters used in the D-S model [DePaolo and Stolper, 1996]. In any case, we recommend that any future isotopic investigation of the HSDP-2 core and/or Mauna Loa lava pile use our new results to estimate the ages of samples.
 We applied the laser and furnace 40Ar/39Ar dating techniques using two different laboratories to two groundmass samples from two sub-contemporaneous lava flows located at a depth of ∼3300 mbsl within the HSDP-2 core. The new40Ar/39Ar data allow us to reach the following conclusions:
 1) We obtained four isochron ages ranging from 612 ± 159 ka to 871 ± 302 ka, all with P ≥ 0.90. As the age difference between the two lava flows are likely to be less than ∼6 ka, we calculated a single weighted mean age from the four isochrons of 681 ± 120 ka (MSWD = 0.84; P = 0.47), which we interpret as the mean eruption age of the two ∼3.3 km deep Mauna Kea lava flows.
 2) Considering a total of eight 40Ar/39Ar isochron ages obtained on tholeiitic lava-flows located between ∼400 and ∼3300 mbsl in the HSDP-2 core indicates a linear accumulation rate at the HSDP core site of 8.4 ± 2.6 m/ka (MSWD = 1.17; P = 0.32). This rate suggests that the eruption of the shield stage took place over ∼750 ka, a longer timescale than predicted by some models of Hawaiian volcano growth.
 3) The accumulation rate of Mauna Kea volcanic rocks in the core hole may be lower than the measured 8.4 m/ka below 1950 mbsl if one or more additional volcanoes contributed to the section recovered by the HSDP-2 core.
 4) Combining all eight available ages for tholeiitic rocks of the HSDP cores allow calculating a combined age-depth model where the ages of the tholeiitic rocks of the core can be estimated with a typical uncertainty of ∼80 ka (2σ).
 5) The revised timescale of the HSDP-2 core suggests that the physical and chemical evolution of Mauna Kea observable in the HSDP-2 core took place on a longer timescale than some previous estimates, with resulting implications for the timescale of the growth of Hawaiian volcanoes, the paleomagnetic record in the core, and the dynamics of the Hawaiian mantle plume.
 Tim Becker (BGC) and Adam Frew (WAAIF) are thanked for analytical assistance. We thank D. DePaolo for providing the samples and P. Vasconcelos for discussion. We acknowledge the support from the HSDP NSF grant. The work at BGC was supported by the Ann and Gordon Getty Foundation. We thank A. Koppers, B. Cousens and two anonymous reviewers for constructive reviews of earlier versions of this paper.