Combined U-Th/He and 40Ar/39Ar geochronology of post-shield lavas from the Mauna Kea and Kohala volcanoes, Hawaii

Late Quaternary, post-shield lavas from the Mauna Kea and Kohala volcanoes on the Big Island of Hawaii have been dated using the {sup 40}Ar/{sup 39}Ar and U-Th/He methods. The objective of the study is to compare the recently demonstrated U-Th/He age method, which uses basaltic olivine phenocrysts, with {sup 40}Ar/{sup 39}Ar ages measured on groundmass from the same samples. As a corollary, the age data also increase the precision of the chronology of volcanism on the Big Island. For the U-Th/He ages, U, Th and He concentrations and isotopes were measured to account for U-series disequilibrium and initial He. Single analyses U-Th/He ages for Hamakua lavas from Mauna Kea are 87 {+-} 40 ka to 119 {+-} 23 ka (2{sigma} uncertainties), which are in general equal to or younger than {sup 40}Ar/{sup 39}Ar ages. Basalt from the Polulu sequence on Kohala gives a U-Th/He age of 354 {+-} 54 ka and a {sup 40}Ar/{sup 39}Ar age of 450 {+-} 40 ka. All of the U-Th/He ages, and all but one spurious {sup 40}Ar/{sup 39}Ar ages conform to the previously proposed stratigraphy and published {sup 14}C and K-Ar ages. The ages also compare favorably to U-Th whole rock-olivine ages calculated from {sup 238}U - {sup 230}Th disequilibria. The U-Th/He and {sup 40}Ar/{sup 39}Ar results agree best where there is a relatively large amount of radiogenic {sup 40}Ar (>10%), and where the {sup 40}Ar/{sup 36}Ar intercept calculated from the Ar isochron diagram is close to the atmospheric value. In two cases, it is not clear why U-Th/He and {sup 40}Ar/{sup 39}Ar ages do not agree within uncertainty. U-Th/He and {sup 40}Ar/{sup 39}Ar results diverge the most on a low-K transitional tholeiitic basalt with abundant olivine. For the most alkalic basalts with negligible olivine phenocrysts, U-Th/He ages were unattainable while {sup 40}Ar/{sup 39}Ar results provide good precision even on ages as low as 19 {+-} 4 ka. Hence, the strengths and weaknesses of the U-Th/He and {sup 40}Ar/{sup 39}Ar methods are complimentary for basalts with ages of order 100-500 ka.


Introduction
Hawaiian lavas are used extensively to probe the chemical composition of the Hawaiian mantle plume (DIXON et al., 1997;FEIGENSON et al., 1983;FREY et al., 1991;FREY et al., 1990;WEST et al., 1988). Multiple chemical components (HART et al., 1992;RODEN et al., 1994), radial (DEPAOLO et al., 2001) and asymmetric (ABOUCHAMI et al., 2005) zonation of the mantle plume source have been invoked to explain the chemical heterogeneity found in the volcanoes. The ability to characterize the temporal evolution of the volcanoes, and to tie the lava geochemistry to the structure of the mantle plume, is critically dependent on accurate dating of the lavas.
Dating has proven to be challenging when using 40 Ar/ 39 Ar technique, because they are young (< 750 ka) and have low concentrations of potassium (COUSENS et al., 2003;SHARP and RENNE, 2005). In this study U-Th/He measurements of phenocrystic olivine (ACIEGO et al., 2007) and 40 Ar/ 39 Ar measurements of groundmass (e.g., SHARP and RENNE, 2005) are applied to Late Quaternary lava flows from Hawaii to further test the U-Th/He method on basalts and to improve the detailed geochronology of the youngest volcanoes. This work is in conjunction with a larger study of the trace element and isotopic compositions of the post-shield stage lavas of the Big Island, including samples from Hualalai (HANANO et al., in review). Detailed geochronology is required in order to accurately compare temporal compositional variations of historical lavas from Hualalai with those from the Mauna Kea and Kohala volcanoes.
Hawaiian lavas are challenging targets for the U-Th/He method because they typically have a large component of trapped helium and low concentrations of uranium and thorium. For this study, we focus on transitional tholeiitic to alkalic lavas, which are largely degassed, and have higher K contents, theoretically allowing high precision 40 Ar/ 39 Ar measurements. Future work will test the U-Th/He method on shield stage tholeiitic basalts, which are traditionally more difficult to date using 40 Ar/ 39 Ar and may have lower U and Th concentrations, but are older and have abundant olivine.
We present data on post shield lavas from the older Kohala and younger Mauna Kea volcanoes, which constitute the northwest section of the Island of Hawaii (Figure 1).
On the Kohala volcano, the northern-most on the island, the volcanic units are classified into two groups: the Polulu Volcanic member, containing the transitional tholeiitic to alkali basalts, and the overlying Hawi Volcanic member, the evolved alkalic cap lavas which range in composition from hawaiitic to trachytic. The Kohala volcano entered the post-shield alkalic stage at about 400 to 500 ka (WOLF et al., 1997). On the Mauna Kea volcano, the lower, transitional basalts are grouped into the Hamakua Volcanic member and the upper, evolved alkalic cap lavas are named the Laupahoehoe Volcanic member (STEARNS and MACDONALD, 1946). Mauna Kea entered the post-shield alkalic stage at about 100 ka (WOLF et al., 1997). For this work, we sampled both sequences of basalts, but found only the Polulu and Hamakua basalts had high enough abundance of phenocrystic olivine for U-Th/He work.

Sample collection and descriptions
Samples were collected from lava flows on the flanks of the Mauna Kea and Kohala volcanoes (Figure 1), exact locations and elevations are summarized in Table 1. The collection points were road and gulch cuts, where the samples could be collected from more than one meter below the original flow surface to minimize cosmogenic 3 He and 4 He production, and more than 1m above the base of the flow which should minimize quenching effects (e.g. glassy groundmass) on the 40 Ar/ 39 Ar ages. At the collection points, sampled lava flows had no direct overlying units and were less than 50m thick.   (MCDOUGALL, 1969). Figure 3 shows the stratigraphic relationship between the collected samples and the nearest age markers.

2.2a U-Th/He.
Rock samples containing olivine were crushed to pea size, a split taken for whole rock powdering, and the remainder sieved, and re-crushed. Olivine grains in the size range 850 to 1000 μm were magnetically separated and handpicked. After picking, the olivine separates, approximately 1 g of material, were air abraded to remove the effects of alpha implantation from the decay of groundmass uranium on the helium concentration or alpha ejection loss from the phenocrysts (ACIEGO et al., 2007;BLACKBURN et al., 2007;MIN et al., 2006). Several attempts were made to separate enough microphenocrysts from samples AMK3, AMK11, AKA2, and AKA7, but the amount of material was not sufficient for helium and U-Th/He analysis. In order to remove enough material by abrasion for microphenocrysts on the order of 100 μm in diameter, more than 70% of the mass must be removed, thereby requiring more than 2 grams of olivine grains to start, an amount unattainable with the 5 kg sample sizes collected.
After abrading, the olivine grains were cleaned, air dried, and loaded into a magnetic mortar and pestle for crushing. The crushing in vacuo releases the trapped (initial) helium component leaving the radiogenic and cosmogenic components. Release of the trapped component was optimized to minimize the effects of overcrushing or undercrushing the samples. Overcrushing can result in the release of radiogenic 4 He (e.g. HILTON et al., 1999) while undercrushing can result in trapped 4 He remaining, between 0.2 and 10% (e.g. WILLIAMS et al., 2005;KURZ et al., 1996). Samples were crushed using 300 beats in 5 minutes, then sieved to remove the remaining pieces larger than 100 μm, which may have magmatic helium remaining.
The <100 μm size fraction was loaded into platinum packets; powder weights ranged from 0.37 to 0.82 g. The total possible contribution of U and Th from the Pt foil was less than 24 pg. The Pt foil packets were loaded into a resistance furnace designed for low abundance U-Th/He work. Gas release was measured at three temperatures: a 300 o C extraction step to remove any adsorbed gases, a 1500 o C step to melt and release the cosmogenic 3 He and the radiogenic 4 He, and a third 1600 o C step to check that gases were fully released. In all cases we found that the gas concentrations released at the 300 o C and 1600 o C steps were at blank level, therefore numbers reported in Table 2 are the blank subtracted 1500 o C step. Extracted gases were purified on a series of getters and the helium concentrated by absorption on a charcoal trap prior and release directly into the mass spectrometer prior to measurement. Helium abundance and isotopic measurements were conducted on a VG5400 at Lawrence Berkeley National Laboratory equipped with a Faraday cup and an electron multiplier operating in pulse counting mode. Abundance measurements were calibrated using an aliquot of air and a reference sample of helium of known isotopic composition: R = 2.4 Ra where Ra is the helium isotopic composition of air ( 3 He/ 4 He = 1.39 x 10 -6 ). The detection limit for 3 He on the multiplier is 5 x 10 -11 nmol; in theory the same detection limit for 4 He although the blanks are significantly higher. Blanks were run prior to each sample for both crushing and heating, and varied between 1.3 and 4.0 x 10 -6 nmol 4 He for the crushers and 1.0 and 3.6 x 10 -6 nmol 4 He for the furnace, 3 He blanks were at the detection limits.
After total gas extraction, the samples packets are retrieved and the fused sample Isolation of U and Th was accomplished using Tru-Spec® column resin following established procedures (LUO et al., 1997). U and Th isotopic and concentration measurements were made at the Woods Hole Oceanographic Institution. U and Th concentration analyses were done by isotope dilution on a ThermoFinnegan Element 2 ICP-MS operating in pulse counting mode. Samples were introduced to the mass spectrometer via a CETAC Aridus desolvator. Background counts were evaluated by peak scanning between masses 227 and 240. Standard NBS960 was measured in between every sample to correct for mass fractionation using the natural 238 U/ 235 U ratio. Samples were measured in triplicate, and the uncertainty in the concentrations, 0.75-1%, reflects the external reproducibility of the repeat measurements. U and Th isotopic compositions were measured on a ThermoFinnegan Neptune MC-ICP-MS.
Thorium and uranium isotopic compositions were measured statically with 232 Th, 238 U, and 235 U in Faraday cups and 230 Th and 234 U in the SEM. Thorium measurements were made with the RPQ filter on, resulting in 85% transmission, abundance sensitivity of 50ppb over 2 amu, and tail corrections of 232 Th on 230 Th of ~0.3%. Sample measurements were bracketed with measurements of UCSC ThA, which was used to correct for mass bias and SEM/Faraday gain of the 232 Th/ 230 Th. Sample measurements for uranium were corrected for mass bias using an internal normalization, the natural 238 U/ 235 U ratio, and bracketed with NBS U10 measurements to determine SEM/Faraday gain. WHOI 's analytical protocols for measuring U and Th isotopes and concentrations are detailed in (BALL et al. 2007 andSIMS et al., 2008a). Accuracy of the spike compositions, and thereby the concentration measurements, and isotopic measurements were monitored by the measurement of rock standard TML, which is well known to have an [ 230 Th/ 238 U] activity ratio of one (see. e.g. SIMS et al., 2008a).
The TML powders dissolved and spiked at the same time as the olivine samples had a [ 230 Th/ 238 U] activity ratio of 1.01, which is within the analytical uncertainties of the measurements. Uranium isotopic compositions for all samples were found to be within error of equilibrium, 234 U/ 238 U activity ratios were 1±0.01. The analytical techniques used for U-Th/He dating are identical to those found in (ACIEGO et al., 2007).

2.2b 40 Ar/ 39 Ar.
Lava rock chunks were crushed into fine chips. Phenocrysts were removed using conventional Frantz magnetic separation. Groundmass grains (300-500 microns) that showed no sign of alteration were further handpicked and leached in diluted (2N) HF for one minute and then thoroughly rinsed with distilled water in an ultrasonic cleaner.
One irradiation of 15 minutes duration was performed in the Cd-shielded (to minimize undesirable nuclear interference reactions) CLICIT facility of the TRIGA reactor at Oregon State University. Samples were irradiated in aluminum discs along with the Alder Creek sanidine standard, for which an age of 1.193 Ma is adopted (NOMADE et al., 2005). 40 Ar/ 39 Ar analyses were performed at the Berkeley Geochronology Center using a CO 2 laser. The gas was purified in a stainless steel extraction line using two C-50 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 mostly using 10 cycles of peak-hopping. A more complete description of the mass spectrometer and extraction line is given in (RENNE et al., 1998). Blank measurements were generally obtained before and after every three sample runs. The correction factors for interfering isotopes correspond to the weighted mean of 10 years of measurements of K-Fe and CaSi 2 glasses and CaF 2 fluorite in the OSTR reactor: ( 39 Ar/ 37 Ar) Ca = (7.60±0.09)x10 -4 ; ( 36 Ar/ 37 Ar) Ca = (2.70±0.02)x10 -4 ; and ( 40 Ar/ 39 Ar) K = (7.30±0.90)x10 -4 . Ages were calculated using the decay constants of (STEIGER and JAGER, 1977). J-and mass discrimination values range from 0.0000680 ± 0.0000003 (0.43%) to 0.0000701 ± 0.0000001 (0.19%) and from 1.00634 ± 0.00216 to 1.00682 ± 0.00242 per dalton (atomic mass unit), respectively. Our criteria for the determination of age plateaus are: (1) to include at least 70% of 39 Ar; and (2) to be distributed over a minimum of 3 consecutive steps agreeing at 95% confidence level and satisfying a probability of fit of at least 0.05. Plateau ages are given at the 2σ level and are calculated using the mean of all the plateau steps, each weighted by the inverse variance of their individual analytical error, and assuming that the initial 40 Ar/ 36 Ar ratio is that of air (295.5 by convention (STEIGER and JAGER, 1977) ). A more recent determination of atmospheric 40 Ar/ 36 Ar (LEE et al., 2006) yields indistinguishable ages because this value is also used to determine mass discrimination and the effects almost entirely cancel out. Integrated ages (2σ) are calculated using the total gas released for each Ar isotope. Data were also cast in inverse isochron diagrams, and in cases where the 40 Ar/ 36 Ar intercept ratio is statistically higher than the atmospheric value, the inverse isochron age is used. Inverse isochrons include the maximum number of consecutive steps with a probability of fit ≥ 0.05. Complete descriptions of the analytical procedure are given in  and (NOMADE et al., 2005). Detailed 40 Ar/ 39 Ar results are shown in Appendix 1 and summarized in Table 2.

U, Th, and He concentrations and isotopic compositions
The U, Th and He concentrations are compiled in Table 2. The 4 He concentrations in olivine are in the range 0.64 to 4.5 x 10 -5 nmol/g; three of the samples have concentrations lower than 1.8 x 10 -5 nmol/g. The low concentrations limit the accuracy with which 4 He concentration can be measured, and if the U and Th concentrations of these samples are representative of Hawaiian olivine, it means that the lower limit age for which the U-Th/He method can be useful for Hawaiian basalt geochronology is about 50 ka using our measurement techniques. Only one sample (AMK7) had cosmogenic 3 He (4.59 x 10 -9 nmol/g) after crushing. Gas released during crushing has a helium R/Ra (helium isotopic composition normalized to air) of 7.5, while the gas released in melting has an R/Ra of > 150 for sample AMK7. All other samples also had crush-release helium compositions between 6.7 and 12 R/Ra, and corresponding concentrations of 3 He released in heating below detection limits.
Hawaiian lavas reported in the literature have a large range of trapped helium concentrations, from 2.2 to 1560 x 10 -5 nmol/g (KURZ et al., 2004). Even the lower limit of this range is comparable to the amount of radiogenic helium measured in our samples. The post-shield lavas, however, are apparently more thoroughly degassed, as indicated by the low amounts of helium released in crushing. The crushing step yielded small amounts of helium that we assume that if there was any trapped helium remaining in the sample at the heating stage, it was minor compared to the amount of helium released in heating. In the worst case, and 10% of the gas remained after crushing (e.g. KURZ et al., 1996), all of the samples would in fact have much younger ages, which would make the U-Th/He ages less consistent with the 40 Ar/ 39 Ar ages.
However, if the crushing and sieving procedures done were as adequate as those in (WILLIAMS et al., 2005), and less than 1% of the trapped component remained than the difference in age would be insignificant.
The olivine U and Th concentrations are compared to those of the whole rocks in Table   1. According to trace element partitioning studies, olivine should contain virtually no U and Th (<0.05 ppb) if it forms in equilibrium with typical basalt liquid (BEATTIE, 1993). The olivine U and Th concentrations are far higher than expected based on published distribution coefficients; instead of a concentration ratio between olivine and whole rock of ca. 10 -5 , the measured ratios are about 0.01 to 0.1. The olivine U and Th concentrations are much more variable than those of the whole rocks, and as shown in the duplicate measurements of AMK12; olivine samples from the same lava flow have variable U and Th concentrations. This variability demonstrates that the U and Th are likely held in inclusions, and therefore the necessity to measure He concentration and U and Th concentrations on the same olivine fraction. The Th/U ratios of the olivine samples are typically lower than those of the whole rocks. The relatively large differences between olivine Th/U and whole rock Th/U indicates that mineral inclusions rather than melt inclusions play the largest role in determining the U-Th concentrations in olivine. Our observations of oxides, phosphates and plagioclase inclusions within the grains is consistent with this hypothesis; PEATE et al. (1996) observed highly variable Th/U ratios in mineral separates of magnetite and plagioclase.
Associated MSWD and P range from 0.25 to 0.94 and from 0.51 to 0.99 respectively (Figure 4 a,b and Table 3). Associated errors are reported as 2 sigma uncertainties within the text.
Sample AMK7 yielded a well-defined plateau age of 123±5ka. For this sample, the percentage of radiogenic 40 Ar* is relatively high (11 -17%), The isochron age (116 ±14 ka) agrees very well with the plateau age and yielded a 40 Ar/ 36 Ar intercept value of 298±4 indistinguishable from the argon atmospheric ratio. For samples AMK3, AMK11, AMK13, AKA5, and AKA7 the fraction of radiogenic 40 Ar* (less than 10%) and their K concentration (i.e. ranging from 0.3 to 1.2%) are significantly lower than for sample AMK7 which limits the age precision, although the estimated initial 40 Ar/ 36 Ar are within 1% of the air value which lends confidence to the plateau age estimates.
Samples AKA2 and AMK12 yielded plateau ages according to our definition of a plateau, but the 40 Ar/ 36 Ar intercept values (313±22 and 305±6; 2σ) are higher than the atmospheric value and their age spectra follow a slight saddle-shaped pattern. These features suggest the presence of excess 40 Ar*. For these samples, we use the isochron age calculation, which should provide a better estimate of the crystallization age . Additionally, AMK12 exhibits a strong tilde-shaped age spectrum. This additional shape suggests that this sample underwent 39 Ar and 37 Ar redistribution during the neutron irradiation (JOURDAN et al., 2007;ONSTOTT et al., 1995). If this is the case, the plateau and isochron calculation cannot be confidently used to define the age of the sample. Furthermore, the fraction of radiogenic 40 Ar* is 6 -10% lower and the estimated initial 40 Ar/ 36 Ar is 6 ±7% higher than the air value. The low percentage of 40 Ar* in comparison to the uncertainty in the initial 40 Ar/ 36 Ar makes the AMK12 age the least reliable of those obtained.
Most of the samples show increasing age and Ca/K over the last 10-20% of the spectrum, at high temperature. These steps also depart from the isochron mixing lines, arguing for a distinct excess 40 Ar* reservoirs included in refractory Ca-rich phases (i.e. interstitial pyroxene). These steps were not included in the plateau and isochron age calculation. Overall, all but one (AMK12) of the 40 Ar/ 39 Ar ages obtained in this study are in agreement with their inferred stratigraphic ages given by previous K/Ar and 14 C dates ( Figure 3). Furthermore, the precision of these new 40 Ar/ 39 Ar ages far surpasses the precision obtained by K/Ar dating on similar lavas, which have uncertainties of 10-30% as shown in Figure 3.

U-Th/He Ages
Ages were calculated using the measured U, Th and He concentrations and isotopic compositions coupled with the U-Th/He age equation given in (FARLEY et al., 2002) and (ACIEGO et al., 2003). For samples with U-series out of radioactive equilibrium, a correction factor must be applied to take into account variations from secular equilibrium. For samples older than 20 ka and less than 1 Ma, this departure from equilibrium will be dominated by the Th-U fractionation. Therefore, the correction is based on the estimate of the initial U-Th disequilibrium (initial 230 Th/ 238 U activity = D 230 ) at the time of helium closure. D 230 can be calculated using either the concentrations of U and Th within the whole rock and olivine separates or by using the 230 Th/ 238 U of the olivine (see FARLEY et al. (2002) and ACIEGO et al. (2007) for discussion). In this case we report the ages calculated directly from the olivine; use of the whole rock-olivine concentrations change the ages by -10% for samples AMK12 and AMK13 and by +10% for sample AMK7. Calculated ages are shown in Table 2 and range from 354 ± 54 to 87 ± 40 ka (2 sigma uncertainties). In general, relatively small amounts of radiogenic 4 He limit the precision of the calculated ages from ±30 to ±50 ka. Unlike the 40 Ar/ 39 Ar ages, there is no additional information (plateau quality, isochron fits, estimate of initial 40 Ar/ 39 Ar) with which to assess the quality of the age determinations. However, for one sample, AMK12, we have duplicate ages of 87±40 and 91±36 ka, which are identical. Unfortunately, that sample has the lowest quality 40 Ar/ 39 Ar age determination of the samples we measured.

3 He C Age
AMK7 is unique because it was collected from a narrow, shallow gully where it was not possible to collect a sample completely shielded from cosmic radiation exposure.
However, although exposed, there was significant cosmic ray shielding due to obstruction of the gully face and the opposite gully wall. We calculated a minimum exposure age of the sample based on the 3 He concentration and a production rate. An average equatorial, sea level production rate for 3 He in olivine, 103 atoms g -1 yr -1 , was scaled for latitude and elevation to 415 atoms g -1 yr -1 (DUNAI, 2001) and again to account for 50% azimuthal shielding and a surface dip angle of 90 o (DUNNE et al., 1999) resulting in a production rate of 101 atoms g -1 yr -1 . The calculated age is ~28 ka. This age is not the age of the bottom of the gully, but is an integrated age based on the increasing exposure of the rock as the gully was cut. Given that the production rate of cosmogenic He is negligible more than 10cm away from the exposed surface, we can infer that the gully was close to 1 m deep at least 28 ka, which is consistent with an eruption age of 120 ka.

Comparison of 40 Ar/ 39 Ar, U-Th/He, and U-series ages
As discussed earlier, the Mauna Kea summit lavas have the best age constraints because of the broad glacial moraine coverage. Sample AMK7 must be older than 15 ka, because the Makanaka glacial moraine overlies it. It is also likely to be older than the 100ka age of the overlying Laupahoehoe Volcanics. The new U-Th/He age of 119±26 ka and 40 Ar/ 39 Ar plateau age of 123±5 ka confirm this hypothesis. The 3 He C age of ~28 ka suggests a slow incision rate that is consistent with the aridity of this region of the Mauna Kea volcano -both the sample collection point and the drainage area for the gully are east of the coastal wet areas on the west coast of the island (EHLMANN et al., 2005). Again, we underscore that for this sample, the 40 Ar/ 39 Ar age and the U-Th/He age agree well.
For sample AMK12 we have the poorest agreement between the U-Th/He results (89±28 ka) and the 40 Ar/ 39 Ar result (239±84 ka). As noted above, the Ar results for this sample are not likely to be as reliable as those of the other samples due to the combined effects of low 40 Ar* and an uncertain initial 40 Ar/ 36 Ar. While there are both large vertical and lateral stratigraphic distances between the collected samples and the closest previously dated samples, all of the available ages in the region where this sample was collected are between 70 and 150 ka (WOLF et al., 1997). Hence we infer that in this instance the duplicated U-Th/He age may be accurate whereas the 40 Ar/ 39 Ar age is spuriously old, possibly beyond estimated uncertainty. The fact that this sample is the most tholeiitic in composition (lowest alkalinity, Table 1) highlights the crux of this work: 40 Ar/ 39 Ar is a powerful dating tool even for young samples, but tholeiites require an alternate dating method, such as U-Th/He.
The minimum age of sample AKA5 from Kohala is constrained by the ~137 ka K-Ar age for a unit (MCDOUGALL, 1969) located 200m stratigraphically higher. The sampled flow is also within the Polulu Volcanic series (Figure 1), which has a documented age range of 250-500 ka based on several previous K-Ar analyses (MCDOUGALL, 1969;MCDOUGALL and SWANSON, 1972). The calculated U-Th/He age of 354±54 and the 40 Ar/ 39 Ar age of 450±40 ka, therefore, are both broadly compatible with the previous data although statistically distinguishable.  (Table 1) from each whole rock -olivine pair. The calculated isochron age would be the eruption age if the olivine and whole rock had identical initial 230 Th/ 232 Th and had remained undisturbed since eruption. There are few olivine U-Th isochron data available in the literature with which to compare these results, so we are not certain how well the requirement of identical initial 230 Th/ 232 Th is likely to be met. One possibility is that the olivine grains did not have an identical initial 230 Th/ 232 Th to the host lava because they are xenocrystic rather than phenocrystic.
Based on the petrographic analysis, this is unlikely, but can not be completely ruled out. In previous work on olivine in basalts (SIMS et al., 2007) it was found that internal U-series isochrons that include olivine separates form linear trends, but more accurate and precise age results are generated by removing the olivine from the age calculation, likely because the olivine grains were xenocrystic or antecrystic.
The whole rock and olivine samples from AKA5 both lie on the equiline and have U/Th ratios that differ only slightly. The U-Th data for AKA5 do not define an age, but are consistent with the age being in the 350 -450 Ka range as determined by the other methods. Sample AMK7 has an OL-WR U-Th age of 163 ± 9 ka, which is somewhat older than our new 40 Ar/ 39 Ar and U-Th/He ages of ca 120 ka. This difference could be an indication that the olivine in this sample is partly xenocrystic, which could skew the U-Th age to older values but not the U-Th/He age. Sample AMK13 also has a well-defined OL-WR U-Th age of 102 ± 11 ka, which is indistinguishable from the U-Th/He age (111±24 ka) and slightly younger than the 40 Ar/ 39 Ar (143 ± 22 ka) age. The two olivine analyses from sample AMK12 give two distinct ages of 20 ± 9 ka and 50 ± 10 ka. The older age is closer to the U-Th/He age but much younger than the 40 Ar/ 39 Ar age for this sample. This may be further evidence that the 40 Ar/ 39 Ar age for AMK12 is too old.
With the exception of sample AMK7, all of the samples have systematically younger U-Th/He ages than 40 Ar/ 39 Ar ages. As discussed above, for AMK12, the most likely cause of this discrepancy is 40 Ar* excess and/or 39 Ar and 37 Ar recoil. However, the reason for the age difference is much less clear for the two other samples. AKA5 and AMK13 both yielded 40 Ar/ 36 Ar intercepts of atmospheric composition on the isochron plot, thereby suggesting that no excess 40 Ar* component is present in these samples.
There are two possible reasons for the U-Th/He to produce erroneously low ages. The first possibility is diffusive loss, where the higher diffusivity of helium than argon in the crystallized lava flow would result in the observed difference in age. However, as (HART, 1984) has shown, olivine in lava flows with thicknesses less than 50m cool too rapidly for helium loss to occur. Similarly, heat from overlying lavas would dissipate too rapidly for the samples to lose helium, at least at these collection locations where the thicknesses of overlying lavas is 0-10m. The second possibility is a systematic error in the estimation of U-series disequilibria (ACIEGO et al., 2007). For the young samples (<300 ka), there is a general agreement (within 10%) between initial 230 Th/ 238 U (D 230 ) disequilibrium calculated using the olivine and the initial value calculated using a Th-U fractionation model for crystals and melts (e.g. FARLEY et al., 2002). Therefore, for these samples we are confident in the errors in the U-Th/He ages due to U-series disequilibria. However, sample AKA5 does have an age that falls in the range of maximum possible error due to uncertainty in D 230 , between 300 ka and 1 Ma (FARLEY et al., 2002;ACIEGO et al., 2007), which could result in uncertainties up to 12%. In this case, using the Th-U fractionation model provides a best estimate for the D 230 , which lowers the error to 2-5%, well below the difference between the U-Th/He age and the 40 Ar/ 39 Ar age.

Implications for future U-Th-He work
A potentially important overall observation is that the U-Th/He ages are consistently either equal to or younger than the 40 Ar/ 39 Ar (Figure 6, uncertainties shown are 1sigma). The sample with the highest percentage of radiogenic 40Ar* is the one that has the best agreement. While the U-Th/He ages are relatively imprecise, this method may yield accurate results as demonstrated by result on sample AMK7 and these results are encouraging. In any case, as noted above, the uncertainty in the 40 Ar/ 39 Ar ages may be as high as that for U-Th/He when the samples have low percentages of radiogenic 40 Ar.
Thus, our results suggest that the U-Th/He chronometer may be a valuable additional tool for dating young mafic volcanic rocks. Substantially more work will be needed, however, before we can be confident about the generality of our conclusions. If the reliability of olivine U-Th/He technique can be demonstrated, we may be able to develop additional insight about which aspects of the Ar data are indicators of unreliable ages by comparing 40 Ar/ 39 Ar and U-Th/He ages. This work reinforces the conclusions (ACIEGO et al., 2007) that U-Th/He dating can be usefully applied to dating basalts in the age range of 50 -500 ka, and provides further evidence about the reliability of the U-Th/He method by comparison to the 40 Ar/ 39 Ar ages on the same samples. The 40 Ar/ 39 Ar ages determined here also strengthen the possibility of using 40 Ar/ 39 Ar to precisely measure ages of increasingly younger alkalic basalts, down to the range of radiocarbon dating, although the comparisons with U-Th/He ages suggest that careful attention must be paid to the percentage of radiogenic 40 Ar measured and the pattern defined by the age spectrum (i.e. sample AMK12 having 6-10% of 40 Ar* and a tilde-shaped age spectrum).
The U-Th/He dating method using olivine of course requires that the samples contain olivine phenocrysts. As noted above and shown by the fewer U-Th/He dates, samples with ~1% microphenocrysts have inadequate olivine for U-Th/He dating. This could be considered a disadvantage in that K-Ar and 40 Ar/ 39 Ar ages can be determined on groundmass and hence are more widely applicable. On the other hand, volcanic groundmass phases may be more susceptible to cryptic alteration that can affect the age determination. Olivine phenocrysts that are useful for U-Th/He dating are also large enough that alteration can be assessed optically. Even in samples where there is some olivine alteration it may be possible to isolate unaltered olivine. However, the possibility of incomplete separation of the helium reservoirs within the olivine remains an issue that has to be carefully considered. Incomplete release of trapped helium would result in older calculated U-Th/He ages. And, while over-crushing could release radiogenic and cosmogenic helium, a significant amount of in situ produced helium (>1%) is not likely to be released unless longer crushing times and greater crushing force is used (MOREIRA and MADUREIRA, 2005). Ultimately, more comparison between U-Th/He and 40 Ar/ 39 Ar ages are desirable to fully assess the validity of the former technique, and a particularly interesting comparison will be for submarine lavas, where it is well known that there are issues with incomplete degassing of Ar (DALRYMPLE and MOORE, 1968).
A stringent test of the U-Th/He method will come in applying it to a wider range of lava compositions. Other ocean island lavas, (e.g. the Azores, Canary, Comores Islands, Samoa) have similar U and Th concentrations to the alkalic and transitional lavas measured in this work, between 0.6 to 7 ppm U BOURDON et al., 2005;CHABAUX and ALLEGRE, 1994;CLAUDE-IVANAJ et al., 1998;CLAUDE-IVANAJ et al., 2001;SIMS and HART, 2006;SIMS et al., 1995;2008b).
For these lavas, assuming U and Th distribution coefficients of order 0.01, the radiogenic helium production will allow U-Th/He ages to be measured in the same age range as in this work. Other ocean island (e.g. Galapagos, Iceland) and mid-ocean ridge basalts (e.g., EPR) basalts have lower U and Th concentrations, between 0.01 and 0.6 ppm (HEMOND et al., 1988;LUNDSTROM et al., 1999;SIMS et al., 2002;KOKFELT et al., 2003;STRACKE et al. 2003;KOKFELT et al., 2005). Therefore, even given optimal analysis conditions of large sample sizes and low blanks, the U-Th/He method will be limited to an older age range, greater than 300 ka. At these low concentrations, the measurement of the 230 Th/ 238 U disequilibria within the olivine will be especially difficult. However, the measurement may be unnecessary if the Th-U fractionation model is valid. Even older samples (> 1 Ma) have the advantage of 230 Th/ 238 U activity ratios close enough to one for multiple half-lives to make the 230 Th/ 238 U disequilibria irrelevant in calculating the U-Th/He age (FARLEY et al., 2002).
One additional complication for future use of the U-Th/He method on OIBs and MORBs is the likely higher initial helium concentration. If samples have both a high initial helium concentration and cosmogenic helium, distinguishing between the radiogenic and initial components of 4 He will be difficult, leading to large errors in the age. However, one advantage of submarine samples is that they lack cosmogenic He.

Implications for Hawaiian plume dynamics
One application of these new ages is to interpret the spatial-temporal evolution of the volcanoes, and in particular, their relationship to the plume source. One way to do this is to map the source of the lava flow, the individual vents, relative to the location of maximum melt supply at the time of eruption (DEPAOLO et al., 2001). The petrology and geochemistry of the lavas can then provide information about the section of the plume it is sampling: the source material via radiogenic isotopes and melting dynamics via U-series isotopes (see e.g. SIMS et al., 1999). However, the combination of geochemical and spatial evidence depends on having reliable ages, which provide the basis for this paleo-mapping. The geochronology of the Big Island has largely been constrained by K-Ar and 14 C ages; the sheer number of ages per stratigraphic unit (20-25; WOLFE AND MORRIS, 1996) provides a "brute force" basis for the age ranges assigned because the standard deviation of the mean for all of the ages is relatively low ( < 10%). But, individual K-Ar ages have poor errors -as much as 50%, therefore reconstruction of the vent locations could be in error by as much as 40 km, the radius of the melting region of the plume. Finer scale analysis of the plume structure and temporal evolution requires more accurate, precise ages, such as those in this work.

Conclusions
The U-Th/He method, applied to olivine phenocrysts in four postshield basalt lavas from the Mauna Kea and Kohala volcanoes in Hawaii yield ages for lavas in the age range 90 to 350 ka. The uncertainty in the ages is estimated to be the larger of ±10% or 20 ka at the 1-sigma level, although duplicate measurements on one lava agree to within a few percent. The age determinations are consistent with previous geologic mapping and geochronological data from the island of Hawaii. Olivine-whole rock U-Th ages measured on the same samples also agree reasonably well with the U-Th/He ages; and the observed discrepancies could have petrological significance. 40 Ar/ 39 Ar ages measured on groundmass from the same four samples yield identical ages in one case, slightly older ages in two cases, a much older (2x) age in one case. The degree of agreement between the 40 Ar/ 39 Ar ages and the U-Th/He ages; the best agreement is for the sample with the largest percent radiogenic 40 Ar* and identical plateau and isochron ages, the worst agreement is for the tholeiitic sample with a clearly identified perturbed age spectrum, low 40 Ar* and higher-than-atmospheric 40 Ar/ 39 Ar trapped component.
For the two intermediate cases, it is not clear yet why we observe some age discrepancy between the two methods and further calibration work is needed. Samples with insufficient olivine for U-Th/He dating yielded robust 40 Ar/ 39 Ar ages, indicating the advantages of 40 Ar/ 39 Ar technique for groundmass samples.
The results presented here are encouraging regarding the applicability of U-Th/He geochronology using olivine phenocrysts in sub aerially-erupted ocean island basalts.
The analytical uncertainty in the U-Th/He ages depends on the He content and age of olivine, and limits the usefulness of the method for samples like those measured here to ages that are greater than about 50 ka. The data from this study and that of (ACIEGO et al., 2007) show the method to be useful for lavas in the age range from 50 to 500 ka, and that the U-Th/He ages can complement 40 Ar/ 39 Ar ages. Further work needs to be done to evaluate other circumstances where the method can complement existing techniques, such as for both subaerial and submarine Quaternary shield stage tholeiitic basalts.    (MCDOUGALL, 1969;WOLF and MORRIS, 1996;WOLF et al., 1997).   Comparison of measured 40 Ar/ 39 Ar ages with U-Th/He ages; 40 Ar/ 39 Ar ages are plateau ages except for AMK12, which is an isochron age. Error bars are 1-sigma, and initial Ar composition is noted. Argon plateau ages are older than U-Th/He ages in all samples.

Figure 7.
Reconstructed vent locations of the sampled lavas relative to the Hawaiian plume based on the 40 Ar/ 39 Ar and U-Th/He ages, and Pacific plate motion of N30W at 9 cm/yr. Table 1.
Description of samples.