Rejuvenated volcanism refers to the reemergence of volcanism after a hiatus of 0.5–2 Ma following the voluminous shield building stage of Hawaiian volcanoes. The composition of the rejuvenated source and its distribution relative to the center of the plume provide important constraints on the origin of rejuvenated volcanism. Near-contemporaneous lavas from the Kaula-Niihau-Kauai ridge and the North Arch volcanic field that are aligned approximately orthogonally to the plume track can constrain the lateral geochemical heterogeneity and distribution of the rejuvenated source across the volcanic chain. Nephelinites, phonolites and pyroxenite xenoliths from Kaula Island have radiogenic Hf, Nd and unradiogenic Sr isotope compositions consistent with a time-integrated depleted mantle source. The pyroxenites and nephelinites extend to the lowest 208Pb/204Pb reported in Hawaiian rocks. These data, along with new Pb isotope data from pyroxenites from the Salt Lake Crater (Oahu) redefine the composition of the depleted end-member of the Hawaiian rejuvenated source at 208Pb/204Pb=37.35±0.05, 206Pb/204Pb = 17.75±0.03, εNd = 9–10, εHf ∼16–17 and 87Sr/88Sr <0.70305. The revised isotope composition also suggests that this depleted component may contribute to LOA and KEA trend shield stage Hawaiian lavas, consistent with the rejuvenated source being part of the Hawaiian plume and not entrained upper mantle. The isotope systematics of rejuvenated magmas along the Kaula-Niihau-Kauai-North Arch transect are consistent with a larger proportion of the rejuvenated depleted component in the periphery of the plume track rather than along its axis.
 The Hawaiian-Emperor Chain is the type example of a deep-seated mantle plume [Morgan, 1971; Sleep, 1990]. Numerous studies have discussed the spatial and temporal distribution of the various compositional components recognized in the Hawaiian lavas, resulting in plume structure models that vary from a concentrically zoned [DePaolo et al., 2001] to vertically and laterally asymmetrically heterogeneous plume [Abouchami et al., 2005; Xu et al., 2007; Farnetani and Hofmann, 2010; Hanano et al., 2010; Pietruszka et al., 2011]. There is, however, considerable debate over the nature and scale of lithological complexity of the Hawaiian plume. For example, several studies suggest that the Hawaiian plume contains recycled eclogite or pyroxenite along with peridotite [Hauri, 1996; Lassiter and Hauri, 1998; Blichert-Toft et al., 1999; Huang and Frey, 2005; Sobolev et al., 2005], whereas others have argued for the presence of ancient depleted peridotite [Norman and Garcia, 1999; Stracke et al., 1999; Kogiso et al., 2003; Bizimis et al., 2005; Frey et al., 2005; Salters et al., 2006; Bizimis et al., 2007]. These models are not mutually exclusive.
 The temporal evolution of many Hawaiian volcanoes includes a main phase of voluminous shield-building tholeiitic basalts followed by renewed volcanism after a hiatus of 0.5–2 Ma [Ozawa et al., 2005]. The lavas erupted during this rejuvenated stage are alkalic, e.g., alkali basalts to nephelinites [Clague and Frey, 1982; Dixon et al., 2008; Garcia et al., 2010], some with mantle xenoliths [White, 1966; Jackson and Wright, 1970]. In addition to the subaerial rejuvenated volcanism on the islands there is also widespread and near synchronous submarine alkalic volcanism on the flanks and distal aprons of some islands and along segments of the Hawaiian Arch that is geochemically similar to the rejuvenated volcanism [Lipman et al., 1989; Clague et al., 1990; Clague and Dixon, 2000; Frey et al., 2000; Yang et al., 2003; Garcia et al., 2008]. An important aspect of the rejuvenated volcanism is that it typically has more depleted isotope characteristics (e.g., lower 87Sr/86Sr, higher 143Nd/144Nd, higher 176Hf/177Hf) relative to shield stage basalts and that of primitive mantle [Stille et al., 1983; Frey et al., 2000; Frey et al., 2005; Fekiacova et al., 2007; Garcia et al., 2010]. The large geographical distribution and the near contemporaneous eruption of rejuvenated type lavas along 300 km of the Hawaiian Chain and on the Hawaiian Arch 200–250 km from the Islands within the last 1 million years [Garcia et al., 2010] suggests that the mechanism(s) and source(s) responsible for the rejuvenated volcanism extend over a wide area. The source of rejuvenated volcanism is thought to be the oceanic lithosphere [e.g., Lassiter et al., 2000] that has been metasomatized by melts from the plume [Yang et al., 2003; Dixon et al., 2008] or a source entirely within the mantle plume [Fekiacova et al., 2007]. Rejuvenated volcanism, therefore, provides key constraints on the compositional complexity and geodynamics of the Hawaiian plume.
 Mantle xenoliths found within the Hawaiian rejuvenated lavas provide additional constraints on the geochemical diversity of the Hawaiian plume and the origin of the rejuvenated volcanism. The garnet pyroxenite xenoliths from Salt Lake Crater (SLC), a vent on Oahu related to the Honolulu Volcanics, are thought to be high pressure (> 20 kb) crystal fractionates from alkali magmas [Frey, 1980; Sen, 1987; Keshav and Sen, 2001; Bizimis et al., 2005; Keshav et al., 2007; Sen et al., 2011]. Thus, they provide access to trapped melts within the lithosphere and potentially allow identification of sources not yet recognized in the surface volcanism. For example, their Hf-Nd isotope systematics suggest the presence of a depleted component within the Hawaiian plume that is distinct from the mantle source of Pacific MORB [Bizimis et al., 2005].
 Here we present the first combined isotope (Hf, Nd, Sr, Pb), investigation of alkali lavas and pyroxenite xenoliths, including garnet pyroxenites, from Kaula Island. We also report Pb isotope data on previously described garnet pyroxenite xenoliths from Salt Lake Crater [Bizimis et al., 2005; Sen et al., 2010; Sen et al., 2011], as well as Hf, Nd, Sr, Pb isotopes on two SLC garnet pyroxenites. Kaula Island is at the W-SW end of the Kauai-Niihau-Kaula ridge, which is near orthogonal to the plume track. Thus, data from the rejuvenated Kaula lavas, together with existing data from Niihau, Kauai and North Arch volcanic fields can provide important constraints on the composition of the rejuvenated source across the plume track. Our new data redefine the composition of the depleted rejuvenated component in the plume and show a greater proportion of the depleted component at the periphery than the plume center.
2. Geological Setting and Sample Description
 Kaula Island is located ∼33 km southwest from the island of Niihau and ∼120 km from the center of Kauai. The Kauai-Niihau-Kaula islands form a volcanic ridge that is approximately orthogonal to the Hawaiian ridge (Figure 1). Kaula Island is a 0.5 km2 remnant of a palagonitic tuff cone with accidental blocks of tholeiite, phonolite, basanite, peridotite, and pyroxenite [Garcia et al., 1986]. Nephelinite is the host lava for the blocks. The petrology, major and trace element chemistry of the Kaula volcanics were reported in Garcia et al. . Mineralogy and chemistry of few pyroxenite xenoliths were described by Garcia and Presti . The phonolites were dated by K-Ar methods at 4.0–4.2 Ma [Garcia et al., 1986], which is older than the rejuvenated volcanism on the neighboring Niihau and Kauai islands [Sherrod et al., 2007; Garcia et al., 2010] but overlaps with the shield-to-postshield transition on Kauai. Thus, the phonolites are considered products of postshield stage volcanism [Garcia et al., 2010]. A basanite accidental block from Kaula gave an age of 1.8 Ma [Garcia et al., 1986] overlapping with the rejuvenated volcanism in the adjacent islands [Garcia et al., 2010]. Nephelinitic juvenile bombs have not been dated but their eruption age is assumed to be younger than the 1.8 Ma basanite accidental block. The similar major and trace element composition of the Kaula nephelinites and basanites with the rejuvenated lavas from Oahu and Kauai, the overlap of eruption ages for Kaula with the Kauai and Niihau rejuvenated volcanism and the presence of xenoliths, suggest that the Kaula lavas are related to the rejuvenated volcanism [Garcia et al., 1986]. Here we report Hf, Nd, Sr, Pb isotope data on three nephelinites and three phonolites including sample KA-36 that was dated at 4.22± 0.25 [Garcia et al., 1986].
 The Kaula pyroxenite xenoliths are clinopyroxenites (samples KA-51, 72, 73, 102, 103, 104) with clinopyroxene and orthopyroxene at > 50 and <40 volume percent modal abundance respectively, and garnet-bearing clinopyroxenites (up to 19 volume percent modal garnet, samples KA-107, 106, 107 109). Olivine and spinel are present from trace amounts to a few volume percent; phlogopite occurs in trace amounts in few samples [Presti, 1982]. Glass patches occur in some pyroxenites and were explained as a reaction with nephelinite host melt [Garcia and Presti, 1987]. The major element chemistry and petrography of the Kaula pyroxenites were originally reported in Presti . We conducted additional mineral major and trace element analyses along with Hf, Nd, Sr, Pb isotope data on clinopyroxenes. The SLC pyroxenites are from the Presnall and Jackson collections housed at the Smithsonian Institution [Bizimis et al., 2005; Keshav et al., 2007].
3. Analytical Methods
 Mineral major element compositions were determined either with a 5 spectrometer JEOL 8500F electron microprobe at the University of Hawaii, or on the 5 spectrometer JEOL Superprobe (JSM 8900R) at the FCAEM, Florida International University, following Keshav and Sen  and Keshav et al. . Some analyses are taken from the study of Presti  using a five spectrometer SX-50 at the University of Hawaii and are reported here for completeness. All of the data are reported in the supporting information1.
 Trace element analyses in clinopyroxene were performed by laser ablation inductively coupled plasma mass spectrometry (LA-ICPMS) either at RSES-ANU using an excimer 193 nm laser coupled to an Agilent 7500 quadrupole ICPMS, or at NHMFL-FSU with a NEW WAVE 213 nm Nd:YAG laser coupled to an Element 1 High Resolution ICPMS using previously reported procedures [Norman et al., 2004; Bizimis et al., 2007]. Trace element concentrations are reported in the supporting information.
 Chemical separations and isotope ratio determinations were performed at the Geochemistry Division, National High Magnetic Field laboratory, Florida State University (NHMFL/FSU), and at the Center for Elemental Mass Spectrometry (CEMS), Department of Earth and Ocean Sciences, University of South Carolina. Clinopyroxene was separated from the pyroxenites by a combination of crushing, magnetic separation and picking under a binocular microscope. The clinopyroxene separates where leached in 6N HCl at ∼100°C for several hours and then rinsed with 18MΩ H2O. 100 mg of hand crushed millimeter-sized chips from some lava samples were picked under the microscope and leached in warm 2.5 N HCl for 1 h to remove possible contamination during sample processing. The chips were dissolved without further powdering. Some lava samples were already powdered and were not leached. Pb isotope compositions of duplicate chip and powder splits gave within error identical values, suggesting no analytical bias between the two processes. All samples were dissolved in Teflon beakers with HF:HNO3 mixture. Pb was extracted using HNO3-HBr acids in anion resin [Manhès et al., 1984]. The eluent from the Pb columns was processed for Hf isotopes after Munkeret al. . Sr and Nd were subsequently separated from the Hf column bulk rock eluent using conventional cation exchange methods, while Nd was further separated from the REE using the Ln-resin (Eichrom, USA).
 Sr isotopes were determined on a Finnigan MAT 262 thermal ionization mass spectrometer (TIMS). The Eimer and Amend (E&A) SrCO3 standard was determined at 87Sr/86Sr = 0.708000 ± 0.000015 (2 standard deviations: 2 stdv, n=15). Nd-isotopes were determined either on the Finnigan MAT 262 TIMS or a Thermo Neptune multi collector ICPMS (MC-ICPMS) at NHMFL/FSU. The LaJolla Nd standard gave 143Nd/144Nd=0.511846 ± 0.000011 (2 stdv, n=10) by TIMS. On the MC-ICPMS sample introduction was by a 50 ul PFA self-aspirating nebulizer coupled to an ESI Apex® introduction system. The LaJolla standard was run every two to three samples. All Nd measurements were fractionation corrected to 146Nd/144Nd = 0.7219 and normalized to 143Nd/144Nd = 0.511850 for the LaJolla standard. Blanks were Nd ∼10pg and Sr < 100pg.
 Pb isotope ratios where determined on the Neptune MC-ICPMS or on the FinniganMAT 262 TIMS. On the MC-ICPMS, Pb isotopes were determined with the Tl addition method [White et al., 2000] using 203Tl/205Tl = 0.418911 to correct for mass fractionation. The NBS 981 Pb standard was determined at 206Pb/204Pb =16.928 ±0.003, 207Pb/204Pb =15.482 ± 0.003, 208Pb/204Pb = 36.669 ± 0.006 (2 stdv, n= 16), and the sample Pb isotope ratios are reported relative to the NBS 981 values of Todt et al. . For the TIMS measurements, all reported Pb isotope ratios were determined on a single analytical session and the data are corrected for fractionation by 0.081 %/amu, based on replicate measurements of the NBS 981 standard and the values reported in Todt et al. . External reproducibility of the standard was 0.04%/amu (n=6). Duplicate clinopyroxene samples run with both methods agree within internal error. Blanks for Pb were ∼50 pg.
 The Hf isotope ratios were determined on a Neptune MC-ICPMS at NHMFL/FSU and at CEMS/USC. At FSU the JMC 475 reference solution was determined at 176Hf/177Hf = 0.282149 ± 9 (2 stdev, n=30; average of 45–90 ng runs). At USC a spilt from the same solution was determined at 176Hf/177Hf = 0.282142 ± 5 (2 stdev, n=10; average of 20 ng runs). All measurements were corrected for fractionation using 179Hf/177Hf = 0.7325 and are reported relative to the accepted JMC-475 value of 176Hf/177Hf = 0.282160. Lu and Yb interferences were corrected online using the instrument software. Duplicates run in both labs agree within error. Blanks for Hf were <40pg. The Sr, Nd, Hf, Pb isotope data are reported in Table 1.
Table 1. Isotope Compositions of Volcanic Rocks From Kaula and Clinopyroxene Separates From the Kaula and SLC Pyroxenites
Hf isotopes at USC, all else at FSU.
Pb isotope by TIMS, all others by MC-ICPMS.
Neph = nephelinite, Phon = Phonolite, (dup) = duplicate. Duplicate dissolution and chemical separation.
Hf, Nd, Sr isotope compositions of the SLC pyroxenites are reported in Bizimis et al. . 2σ = 2 *stdev /√n internal precision of the analyses, listed for last significant digits. εNd and εHf are calculated using CHUR 143Nd/144Nd = 0.512638, and 176Hf/177Hf =0.282785, respectively.
 The clinopyroxenes from Kaula pyroxenites are classified as aluminian augites after Morimoto , with relatively Fe-rich (Mg#= 0.73–0.83, where Mg# = Mg/[Mg+Fe] cations) compositions compared to clinopyroxenes from Hawaiian spinel peridotites with Mg#> 0.88 (Figure 2a). The compositions of clinopyroxenes in the Kaula pyroxenites overlap with those in the SLC pyroxenites (e.g., Na2O versus Mg#, Figure 2a), although the Kaula clinopyroxenes extend to somewhat higher Al contents at a given Mg# (Figure 2b).
 The Kaula and SLC pyroxenite garnet compositions entirely overlap (not shown). Some of the Kaula garnets have spinel cores, as also observed in the SLC pyroxenites [Keshav et al., 2007]. On a Mg#cpx versus Mg#garnet plot, the Kaula samples have slightly higher Mg#garnet at a given Mg#cpx than the SLC pyroxenites and the two trends are subparallel (Figure 3a). Experimental data on pyroxenitic and eclogitic compositions [Hirschmann et al., 2003; Kogiso et al., 2003; Keshav et al., 2004; Pertermann et al., 2004] show coexisting clinopyroxene-garnet pairs with variable Mg#cpx/Mg#gt ratios similar to the SLC and Kaula ranges (Figure 3a). There is no straightforward relationship between KDFe-Mg and P,T or compositional parameters. As the Kaula clinopyroxene-garnet pairs follow the same trend as the SLC, we interpret the Kaula pyroxenite clinopyroxene and garnet compositions to reflect equilibrium conditions.
 Orthopyroxene compositions in the Kaula pyroxenites overlap with those of the SLC pyroxenites on a Mg#cpx versus Mg#opx plot (not shown), but they extend to higher Al2O3 contents at a given Mg# (Figure 3b). In SLC pyroxenites, high-Al orthopyroxenes occur in the contact rims between garnet and spinel [Keshav et al., 2007]. This is also observed in the Kaula pyroxenites where the contact rim between spinel and garnet in samples KA-106 and KA-107 contains Al-rich orthopyroxene (Figure 3b, data in the supporting information). These interstitial orthopyroxenes have variable Al contents even within a few 10s of microns. Garcia and Presti  suggested that the secondary orthopyroxene in sample KA-107 formed during the reaction of an infiltrating nephelinitic melt with a preexisting garnet + clinopyroxene + spinel mineralogy.
4.2. Trace Element Results
 The chondrite-normalized rare earth element (REE) patterns of the Kaula pyroxenite clinopyroxenes show light REE enrichment with convex upward REE patterns, whereas patterns for samples KA-102 and KA-104 are more sinuous (Figure 4a). Compared to the SLC pyroxenites, the Kaula clinopyroxenes generally have higher heavy REE (HREE) concentrations and less fractionated REE patterns. We do not observe any correlation between the degree of HREE depletion in clinopyroxene and the presence of garnet. For example, the two samples with the lowest HREE contents in clinopyroxene (1–2 × chondrite, Figure 4a) do not contain garnet. This was also observed in some SLC pyroxenites [Bizimis et al., 2005]. This implies that although these clinopyroxenes were in equilibrium with garnet, garnet is heterogeneously distributed in the rock and not present in the studied centimeter-sized xenoliths. Primitive mantle normalized trace element patterns of the Kaula clinopyroxenes show positive Sr and negative Pb anomalies but lower Zr-Hf contents than their SLC counterparts, consistent with their lower LREE and MREE concentrations relative to SLC (Figure 4b).
4.3. Isotope Compositions
4.3.1. Pb Isotope Systematics
 The Kaula clinopyroxenes have relatively unradiogenic Pb isotope ratios (Table 1) and extend to lower 208Pb/204Pb ratios that previously reported in Hawaiian shield or rejuvenated lavas (Figure 5a). The three most unradiogenic pyroxenites (KA-51, KA-106, KA-109) include both garnet-bearing and garnet-free samples. Thus, there is no apparent correlation between mineralogy and isotope composition. Pb isotope ratios for Kaula nephelinites overlap those of pyroxenites and extend to less radiogenic compositions than other Hawaiian rejuvenated lavas. The Kaula phonolites have slightly more radiogenic Pb isotope ratios than the nephelinites and overlap other rejuvenated lava compositions from Oahu (Honolulu Volcanics), Niihau (Kiekie Volcanics), and the unradiogenic end of the field defined by rejuvenated stage (Koloa Volcanics) and postshield lavas from Kauai (Figure 5a). The SLC pyroxenite clinopyroxenes partially overlap and extend to less radiogenic Pb isotope ratios than their host Honolulu Volcanics lavas.
 In 206Pb/204Pb versus 208Pb/204Pb space the Kaula and SLC pyroxenites define two slightly offset arrays with slopes that are identical within error (1.1 ± 0.1 and 1.0 ± 0.2, respectively) (Figure 5b). The Koloa Volcanics plot at the radiogenic extension of the Kaula pyroxenite Pb isotope array with a slope of 1.21 ± 0.04 that is within error of the pyroxenites (Figure 5b). Together, the Kaula pyroxenites and lavas and the Koloa Volcanics define an array that is oblique to and crosses the LOA-KEA divide into the LOA field (Figures 5a and 5b). In contrast, the Honolulu Volcanics have a distinctly shallower slope (0.76 ± 0.07) and plot on the KEA side of the divide. The Kalaupapa Volcanics of East Molokai define a slope (0.87±0.20) intermediate to that of the Honolulu Volcanics and the Koloa or Kaula pyroxenites, although with a larger uncertainty due to the limited Pb isotope variability. The North Arch lavas entirely overlap the Kalaupapa lavas Pb isotope values but with only 4 samples and a small range in Pb isotopes it is difficult to obtain meaningful regression lines. Overall, our new Kaula and SLC Pb isotope data, together with other published rejuvenated lava data forms a well-defined array in 206Pb/204Pb versus 208Pb/204Pb isotope space that is distinct from the LOA and KEA shield stage trends, with lower 208Pb/204Pb at a given 206Pb/204Pb than the LOA-shield stage lavas, and distinctly less radiogenic Pb isotopes than KEA-shield stage lavas.
4.3.2. Sr-Nd Isotope Systematics
 The Kaula lavas and pyroxenites have relatively radiogenic Nd and unradiogenic Sr isotopes compared to the LOA and KEA trend shield volcanoes and broadly overlap the compositions of rejuvenated stage lavas and pyroxenites xenoliths in 87Sr/86Sr-εNd isotope space (Figure 5c). The nephelinites and phonolites form a subhorizontal trend with variable Sr isotopes at near constant Nd, as also observed in the Honolulu Volcanics. Phonolite KA-37 has higher 87Sr/86Sr plotting above the Hawaiian shield and rejuvenated lava fields, but overlaps the other two phonolites in all other isotope compositions. In contrast, the Kaula pyroxenites have higher εNd and lower 87Sr/86Sr than the lavas and form a well-defined negative trend toward the most radiogenic Nd isotopes reported for Hawaiian rejuvenated volcanism (e.g., North Arch lavas) and overlap with the SLC pyroxenite trend. The combined Kaula-SLC pyroxenite trend appears subparallel to the negative trend defined by the Koloa Volcanics, but shifted to slightly higher 87Sr/86Sr.
4.3.3. Hf-Nd Isotope Systematics
 The Kaula lavas and pyroxenites have relatively radiogenic Hf and Nd compositions and overlap the SLC pyroxenites, Honolulu Volcanics and Koloa Volcanics data. In εHf versus εNd space the Kaula and SLC pyroxenites form a field that originates at the radiogenic end of the Hawaiian shield stage lava compositions (Mauna Kea lavas) and extends toward relatively radiogenic Hf compositions at a given Nd, away from the Pacific MORB field (Figure 5d). In detail the Kaula pyroxenites have a shallower slope (∼2.2) than the SLC pyroxenites (slope ∼3.6), but considerably steeper than the terrestrial array (∼1.5) and the Pacific MORB (∼1.4) (Figure 5d). The Kaula lavas have less variable Hf isotopes than the Kaula pyroxenites with no obvious distinction between nephelinites and phonolites.
5.1. Pressures and Temperatures of Equilibration of the Kaula Pyroxenite Xenoliths
 Equilibration temperatures for the garnet-bearing samples were calculated using the pyroxene-garnet Fe-Mg exchange thermometers [Ellis and Green, 1979; Ganguly et al., 1996; Nimis and Grutter, 2010], and the Ca-in-orthopyroxene and orthopyroxene-clinopyroxene Ca-Mg exchange thermometers [Brey and Köhler, 1990; Nimis and Grutter, 2010] for all samples. Pressures are calculated with the Al-in orthopyroxene thermobarometer of Brey and Köhler  and Nickel and Green  for the garnet-bearing samples, by solving for pressure and temperature iteratively. For samples without garnet we assume 20 kb pressure. We used the compositions of larger grains and not the interstitial phases. The calculated pressures and temperatures are reported in the supporting information and plotted in Figure 6.
 The Ca-in-orthopyroxene and two-pyroxene thermometers give similar temperatures for all samples, around 1100°C. At these temperatures, the calculated pressures for the garnet-bearing samples are relatively low (<13 kb). The orthopyroxene-garnet Fe-Mg thermometer of Nimis and Grutter  yields temperatures (∼1130 to 1270°C) that are somewhat higher than the pyroxene thermometers and pressures of 11–15 kb. The clinopyroxene-garnet thermometers [Ellis and Green, 1979; Ganguly et al., 1996] give temperatures of ∼1400 to 1500°C, which are 130–400°C higher than the temperatures derived by the pyroxene thermometers, and result in calculated pressures of 23–25 kb. Such pressures and temperatures would place the Kaula garnet pyroxenites above the peridotite dry solidus [Hirschmann, 2000] (Figure 6), so we consider these temperatures unreasonably high. The largest temperature discrepancies between the two pyroxene and clinopyroxene-garnet thermometers are in samples KA-107 and KA-105, which also have opx with the highest Al contents.
 An explanation for the temperature (and resulting pressure) differences between the clinopyroxene-garnet and the pyroxene thermometers could be the presence of Fe+3 in cpx and garnet. Mossbauer Fe3+/Fe2+ data on the SLC pyroxenites [Tibbetts et al., 2010] indicates significant Fe3+ contents in cpx (Fe+3/Fetotal ∼0.3) and garnet (Fe+3/Fetotal ∼0.05), within the range of clinopyroxenes from cratonic peridotites [Woodland, 2009]. Assuming similar Fe+3/Fetotal contents in the Kaula clinopyroxenes and garnets, this would result in clinopyroxene-garnet temperatures that are 150–200°C lower than assuming Fetotal= Fe+2, thus decreasing the discrepancy between the pyroxene and clinopyroxene-garnet thermometers. As the pressure effect on both pyroxene thermometers is relatively small (30°C/10 kb), we take the orthopyroxene-clinopyroxene Ca-Mg exchange thermometer temperatures to broadly reflect those of last equilibration (Figure 6). However, the calculated pressures (8–15 kb) for the garnet pyroxenites are inconsistent with the presence of garnet, which suggests pressures higher that about 18–20 kb in Cr-poor basaltic systems similar to the composition of these pyroxenites [Sen, 1988; Liu and Presnall, 2000; Keshav et al., 2007].
 Additional evidence for the pressure of igneous crystallization may come from phase petrography. Based on experimental data in the CMAS system and the coexistence of olivine +clinopyroxene + garnet it was suggested that the SLC garnet pyroxenites crystallized at >30 kb [Keshav et al., 2007]. Below 30 kb, the olivine +clinopyroxene + garnet assemblage does not exist on the liquidus of the CMAS system [Milholland and Presnall, 1998] and the liquidus phases are clinopyroxene, garnet and spinel. Since the Kaula garnet pyroxenites only have traces of olivine they likely crystallized at <30 kb (<90 km), with spinel as the early crystallizing phase as observed experimentally in similar composition systems [Keshav et al., 2004]. The low calculated pressures (<15 kb) and relatively high temperatures may then reflect transient heating and reequilibration at shallower levels, perhaps during ascent from the mantle. In either case, the Kaula pyroxenites are considerably hotter than the calculated 90–100 Ma oceanic geotherm (Figure 6), the estimated age of the lithosphere under the Hawaiian islands based on the 110Ma age of ODP site 843 SW of Oahu [Waggoner, 1993] and ocean floor age reconstructions [Müller et al., 1997].
5.2. Crystal Fractionation Control on the Pyroxenite Compositions
5.2.1. Major Elements and REE
 The negative correlation between Na2O and Mg# in the clinopyroxenes from the Kaula (and SLC) pyroxenites is generally consistent with magmatic fractionation. Figure 2a shows that progressive clinopyroxene fractionation from a parental magma similar to the most primitive Kaula nephelinites [Garcia et al., 1986] and Honolulu Volcanics [Clague and Frey, 1982] with Mg#∼64 to 67, and Na2O ∼ 2.9 to 3.2 wt%, will create increasingly higher Na and Fe contents in the residual melt and crystallizing clinopyroxene that match well with the Kaula and SLC clinopyroxene Na2O and Mg# systematics. The influence of cocrystalizing garnet is limited as Na is also incompatible in garnet. The negative Mg# versus Al2O3 correlation is also consistent with such a model (Figure 2b), however the larger range of Al2O3 at a given Mg# likely reflects additional variation in the Al contents of the parent melt and possibly the crystallization of garnet where Al is compatible.
 The Nd/Yb ratios in the Kaula and SLC clinopyroxenes, taken as a proxy for LREE/HREE fractionation, increase with decreasing Mg# (Figure 7). Ytterbium is more compatible than Nd in clinopyroxene, and clinopyroxene fractionation will lead to residual melts with increasing Nd/Yb and decreasing Mg# (Figure 7). Garnet crystallization together with clinopyroxene will even further fractionate Nd/Yb in the residual melt. Allowing for variable Nd/Yb in the starting melt composition (Nb/Yb(ch.norm)∼12 to 36, within the range of Kaula lavas and Honolulu Volcanics [Garcia et al., 1986; Yang et al., 2003], fractional crystallization of a pyroxenitic mineralogy with a range of clinopyroxene to garnet ratios can reproduce the observed Kaula and SLC cpx Nb/Yb versus Mg# systematics (Figure 7). Some Kaula samples are best fit with a lower DYbbulk than the SLC pyroxenites which implies less garnet in the crystallizing Kaula mineral assemblage, qualitatively consistent with the overall lower modal abundance of garnet in the Kaula pyroxenites compared to the SLC.
 In summary, the major and trace element systematics suggest that the Kaula pyroxenites formed within the lithosphere (<100 km) as crystallization products of parental melts similar to erupted rejuvenated magmas, in agreement with their generally similar isotope systematics. We emphasize that there are no obvious correlations between major or trace element systematics and the isotope compositions in the Kaula pyroxenites. This suggests that any source heterogeneity reflected in the parental melt major and trace element composition has been obscured by the subsequent fractionation and crystallization of these rocks. Furthermore, the occurrence of garnet pyroxenite xenoliths at Kaula and SLC, some 300 km apart, suggest that high-pressure crystallization of rejuvenated-type magmas may be common beneath the Hawaiian chain, and perhaps in other OIB settings. This likely has implications for the rheology and seismic structure of the oceanic lithosphere, as well as the composition of the recycled oceanic lithosphere, but such discussion is beyond the scope of this paper.
5.3. Constraints on the Source of the Hawaiian Rejuvenated Volcanism
 The radiogenic Hf, Nd, and unradiogenic Sr and Pb isotope compositions of the Kaula nephelinites and pyroxenites are common in Hawaiian rejuvenated volcanism. Recycled oceanic crust has been invoked to explain the radiogenic Sr and unradiogenic Nd isotopes of the shield stage magmas [Hauri, 1996; Huang and Frey, 2005]. The isotopic compositions however suggest that the magmas parental to the Kaula pyroxenites did not contain a significant amount of recycled oceanic crust. It is also unlikely that they represent “secondary pyroxenites” formed as reaction products of a melt from such recycled oceanic crustal lithologies with peridotite [Sobolev et al., 2005].
 We note that SLC garnet pyroxenite NMNH-114954-28A from our study has exsolved orthopyroxene in garnet, inferred as pseudomorphs after majoritic garnet, which would indicate crystallization pressure >5 GPa [Keshav and Sen, 2001]. Sample NMNH-114954-20A has ilmenite exsolution in garnet, and is also thought to have crystallized at >5GPa [Keshav and Sen, 2001]. These two samples indicate crystallization deeper than the seismically imaged lithosphere in Hawaii (88±7 km) [Priestley and Tillman, 1999], and were thought to perhaps represent recycled oceanic crust in the plume [Keshav and Sen, 2001]. However, the radiogenic Nd and unradiogenic Sr isotopes of both samples reported here (Table 1) entirely overlap with other SLC pyroxenites and rejuvenated magmas, and lack the enriched isotope compositions of an eclogite or “secondary pyroxenite” invoked to explain the high silica concentrations of shield stage volcanism [Hauri, 1996; Sobolev et al., 2005]. The isotopic compositions inferred for the source of the shield stage lavas remain elusive in the pyroxenite population.
 There is no correlation between Lu/Hf versus 176Hf/177Hf and Sm/Nd versus 143Nd/144Nd ratios in the Kaula pyroxenite cpx. Similarly, no correlation was observed in the cpx-garnet and bulk rock Lu-Hf and Sm-Nd isotope systematics of the SLC pyroxenites [Bizimis et al., 2005] as well as their bulk Re-Os isotope systematics [Sen et al., 2011]. These data argue against these rocks having formed within the Pacific lithosphere at ∼90 Ma ago at a mid oceanic ridge [e.g., Lassiter et al., 2000]. Therefore we conclude that the measured isotopic compositions in the Kaula and SLC pyroxenites reflect their initial isotopic compositions, and that these in turn reflect those of their mantle sources.
5.3.1. The Depleted Rejuvenated Component
 The Kaula nephelinites and pyroxenites and Hawaiian rejuvenated lava data delineate a “rejuvenated volcanism” array that extends to lower 208Pb/204Pb at a given 206Pb/204Pb than the LOA-trend lavas and has less radiogenic Pb isotope ratios than KEA-trend shield lavas. The Kaula phonolites overlap this array. In 206Pb/204Pb-208Pb/204Pb space, the slopes defined by all pyroxenites and Kaula lavas are parallel, but offset to higher 208Pb/204Pb at a given 206Pb/204Pb, compared to the Pacific MORB field (Figures 5a and 5b), which includes the Garret Transform fault lavas with highly unradiogenic Pb isotopes. This argues against involvement of a Pacific MORB mantle source in the Kaula and the SLC pyroxenites source, in agreement with the conclusions reached for other rejuvenated lava sources [Fekiacova et al., 2007; Garcia et al., 2010]. Extension of the arrays defined by the Kaula, Koloa, SLC pyroxenites and Kalaupapa Volcanics arrays toward unradiogenic 208Pb/204Pb-206Pb/204Pb converge around 208Pb/204Pb = 37.35 ± 0.05, 206Pb/204Pb = 17.75 ± 0.05 (Figure 5b). This convergence suggests that this composition may represent a common end-member component in the source of rejuvenated volcanism. Following Fekiacova et al.  we call this component the Depleted Rejuvenated Component (DRC). By defining the DRC in 206Pb/204Pb-208Pb/204Pb isotope space we can use the correlations between Pb isotopes and the other isotope systems to define its Hf, Nd, Sr isotope compositions.
 In εNd-208Pb/204Pb isotope space (Figure 8a), the coupled unradiogenic Pb and radiogenic Nd isotope compositions of the rejuvenated array, currently best represented by the Kaula pyroxenites, constrain the Nd isotope composition of DRC at εNd ∼9 to 10. In a similar fashion, the Sr-Nd isotope systematics for the Kaula and SLC pyroxenites suggest 87Sr/88Sr <0.70305 for the DRC.
 The different slopes of the SLC and Kaula pyroxenites in εHf-εNd space (Figure 5d) make it difficult to define a single Hf isotope composition for DRC. This is also evident in Hf-Pb isotope space (Figure 8b) where some SLC pyroxenite [Bizimis et al., 2005] and Honolulu Volcanics data [Stille et al., 1983] extend to more radiogenic εHf at a given 206Pb/204Pb than the Kaula lavas and pyroxenites. Nevertheless, the εHf-206Pb/204Pb systematics of the Kaula pyroxenites are highly correlated, and together with the Kaula lavas, Koloa Volcanics and most of the Honolulu Volcanics and SLC pyroxenites constrain the DRC to have εHf =16–17 at 206Pb/204Pb=17.75.
 The radiogenic Nd, Hf and unradiogenic Sr, Pb isotope compositions of DRC suggest a source with relatively low, time-integrated Rb/Sr, U/Pb, Th/Pb and high Lu/Hf, Sm/Nd parent/daughter ratios. These ratios are qualitatively consistent with the partitioning of these elements in the residual mantle during peridotite melting [Salters et al., 2002]. The DRC isotope composition is therefore consistent with a peridotite that is residual after an ancient partial melt extraction event. The relatively radiogenic Hf at a given Nd isotope compositions of the SLC pyroxenites have been explained by mixing between plume (KEA-type) melts and those from an ancient (e.g. >1Ga) depleted mantle [Bizimis et al., 2005]. Peridotites with extremely radiogenic Hf and unradiogenic Os isotopic compositions in Oahu and Kauai were explained as reflecting an ancient depleted recycled mantle lithosphere within the Hawaiian plume [Bizimis et al., 2004; Bizimis et al., 2007; Bizimis et al., 2008]. An ancient depleted lithosphere has also been invoked in the source of the Koolau shield lavas [Salters et al., 2006], and is a long-lived source feature of the Hawaiian-Emperor Chain [Frey et al., 2005]. The above evidence favor an origin of the DRC as an ancient depleted mantle that is part of the plume. Note, however, that the elevated incompatible element concentrations of the rejuvenated Hawaiian alkalic lavas [Clague and Frey, 1982; Yang et al., 2003; Garcia et al., 2010] are hard to reconcile with direct melting of a depleted peridotite. Instead, several studies suggested that metasomatic enrichment of a depleted source shortly before melting is required [Roden et al., 1984; Salters and Zindler, 1995; Reiners and Nelson, 1998; Yang et al., 2003; Dixon et al., 2008].
5.3.2. The Enriched Components in the Rejuvenated Source
 At its radiogenic Pb isotope extension, the array defined by the Kaula lavas and pyroxenites and the Koloa Volcanics crosses the LOA-KEA divide into the LOA region of the 206Pb/204Pb versus 208Pb/204Pb plot (Figures 5a and 5b). In contrast, other rejuvenated lavas (Honolulu Volcanics, Kalaupapa) have shallower slopes and remain on the KEA side of the LOA-KEA divide (Figure 5b). To a first approximation, this suggests the contribution of both LOA and KEA type sources to the rejuvenated volcanism. In detail however, this Kaula-Koloa Volcanics array points to a “Loihi” or “Kea-hi8” [Abouchami et al., 2005] type composition with relatively radiogenic 208Pb/204Pb at a given 206Pb/204Pb, and not to the more isotopically extreme LOA sources of the Makapuu (Koolau) Lanai, or Kahoolawe shield lavas with relatively unradiogenic 208Pb/204Pb (Figures 5a and 5b). This is further evident in εNd-208Pb/204Pb and εHf-206Pb/204Pb space (Figure 8) where there appears to be a compositional continuity from the DRC, through the rejuvenated magmas, toward the KEA and/or Loihi type compositions only, and not toward the other LOA type compositions. It is unclear whether both Loihi and KEA type sources contribute to rejuvenated magmas, or there is simply a large enough variability in the Pb isotope compositions of the KEA source alone to explain the radiogenic Pb isotope variability of the rejuvenated magmas.
 In this context, the previously proposed KEA end-member isotope composition with εHf ∼14, εNd ∼9 at 206Pb/204Pb ∼ 18.8 and 208Pb/204Pb ∼38.35 [Tanaka et al., 2008] appears too radiogenic in Nd and Hf isotopes and falls significantly off the trend defined by the KEA lavas in εNd versus 208Pb/204Pb and εHf versus 206Pb/204Pb space (Figures 8a and 8b). In particular, the slightly negative slope defined by the Mauna Kea HSDP lavas in εNd versus 208Pb/204Pb space requires that the KEA end-member has less radiogenic Nd isotope composition than DRC. Based on the trends in Figures 8a and 8b, we suggest that the KEA end-member should be revised to εNd = 6–7, εHf =11–12, with 87Sr/86Sr = 0.7035–0.7036, 206Pb/204Pb = 18.7–18.8, 208Pb/204Pb = 38.3–38.4 similar to values proposed by Tanaka et al. .
6. Implications for the Geochemical Structure of the Hawaiian Plume
 The DRC proposed here allows us to reinterpret some previously proposed components for the Hawaiian volcanism. The Depleted Makapuu Koolau (DMK) component has been invoked to explain the unradiogenic Pb compositions of the Koolau shield stage and rejuvenated Honolulu Volcanics [Tanaka et al., 2008]. In 206Pb/204Pb-208Pb/204Pb space (Figure 5b), DMK falls between DRC and the Enriched Makapuu Koolau component of Tanaka et al. . In this case, the unradiogenic Pb compositions of the Honolulu Volcanics can be explained by mixing between DRC and the Enriched Makapuu Koolau end-member (Figure 5b) and DMK is not required as a separate entity. If correct, this mixing relationship also implies a contribution from a LOA-type source (Enriched Makapuu Koolau) to the rejuvenated volcanism in Koolau, which is best seen in the Pb isotope systematics of Honolulu Volcanics (Figure 5b).
 The slightly negative slope defined by the Mauna Kea lavas in εNd-208Pb/204Pb space (Figure 8a) toward the DRC and away from the low εNd-low208Pb/204Pb ratios of the extreme LOA compositions, also implies that the DRC may also contribute to the KEA shield / post shield stage volcanism. A mixed DRC-KEA source appears to also contribute to some LOA shield and postshield volcanoes. For example, the so-called “Kalihi” component with 206Pb/204Pb = ∼18.3, 208Pb/204Pb ∼37.9 invoked to be common in the source of the rejuvenated Honolulu Volcanics and the KSDP (Koolau) shield lavas [Fekiacova et al., 2007], is internal to the DRC-KEA compositional range and therefore may be a mixture of these two as opposed to a physically distinct entity in the plume. Moreover, the Hualalai postshield lavas [Hanano et al., 2010], a LOA trend volcano, show a shift from the LOA shield compositions toward the DRC in both εNd versus 208Pb/204Pb and εHf versus 206Pb/204Pb space (Figures 8a and 8b). Finally, a contribution of a depleted component, similar to that sampled during rejuvenated volcanism has been proposed for the Kauai shield lavas [Mukhopadhyay et al., 2003]. The above observations suggest that KEA and DRC may be locally mixed and that DRC may contribute to LOA shield (e.g., KSDP) and postshield (e.g., Hualalai) volcano sources.
 The DRC is recognized in the compositions of xenoliths and lavas erupted in the last ∼4.2 Ma along 300 km of the Hawaiian Island chain (Kauai-Kaula to East Molokai). The compositions of the ∼76 to 81 Ma lavas from Detroit Seamount along the Emperor Seamount chain [Frey et al., 2005] are also consistent with the DRC composition proposed here. These observations are consistent with the DRC being a long-lived feature of the Hawaiian plume, rather than a local component fortuitously incorporated in the source of the Kaula parental magmas. The increased proportion of the DRC in the rejuvenated rather than the shield stage lavas can be explained by melting of a heterogeneous plume, where the more fertile and isotopically enriched components preferentially melt and dominate the chemistry of the voluminous shield lavas. Exhaustion of the enriched components after the shield stage melting event leaves a residual plume mantle with an increased proportion of the more refractory DRC. This depleted plume source subsequently melts downstream from the plume center and at the edges of the Hawaiian swell, either due to a secondary melting zone [Ribe and Christensen, 1999], lithosphere flexing [Bianco et al., 2005], erosion of the lithosphere by the plume [Li et al., 2004; Ballmer et al., 2011], or some combination thereof [Garcia et al., 2010], leading to a higher proportion of DRC in the source of rejuvenated volcanism.
6.1. Rejuvenated Source Heterogeneity Across the Plume Track
 The rejuvenation stage lavas from Kaula, Niihau, Kauai, and North Arch volcanic field fall on a ∼400 km transect that strikes approximately orthogonal to the Hawaiian Islands (Figure 1). The ages of rejuvenated volcanism along this transect approximately overlap. Kaula nephelinites and pyroxenite xenoliths are younger than 1.8 Ma. The Niihau lavas erupted between 2.5 and 0.3 Ma [Clague and Dalrymple, 1987; Moore and Clague, 2004; Sherrod et al., 2007]. The Koloa Volcanics erupted between 2.6 to 0.15Ma [Garcia et al., 2010]. The North Arch lavas erupted between 1.15 and 0.5 Ma based on Mn-coating thickness [Moore and Clague, 2004]. Therefore, the compositions of these lavas provide a snapshot of the geochemical variability of the rejuvenated source across the plume track. There is a systematic increase from southwest to northeast in 208Pb/204Pb ratios for Kaula Niihau, and Kauai rejuvenated lavas, and then a slight decrease for North Arch lavas (Figure 9a). These trends are less apparent in Nd isotopes (Figure 9b). However, the peripheral Kaula pyroxenites and North Arch lavas have, on average, more radiogenic Nd isotopes (143Nd/144Nd = 0.513056 ± 19 and 0.513076 ± 12, 1 standard deviation, respectively) than the Kauai and Niihau rejuvenated lavas (143Nd/144Nd = 0.513037 ± 29 and 0.513039 ± 36, respectively). Similarly Kaula pyroxenites and North Arch lavas have, on average, more radiogenic Hf isotope compositions (176Hf/177Hf = 0.283198 ± 27 and 0.283183± 6 respectively) than the Koloa Volcanics (176Hf/177Hf = 0.283143 ± 29). The Sr isotope compositions are more scattered and do not allow for a meaningful comparison. The limited isotope data from the North Arch volcanic field notwithstanding, the Pb, Nd, Hf isotope patterns imply a larger contribution of the DRC to the rejuvenated magmas at the edges of the plume, than along the plume track.
 The addition of rejuvenated magma data from other locations along the plume track, for example the West Kaena alkali lavas erupted at 1.39–0.38 Ma [Greene et al., 2010], the Oahu Honolulu Volcanics at 0.8–0.1 Ma [Ozawa et al., 2005] and SLC pyroxenites, does not significantly change the observed patterns. It does however add more complexity to the interpretation as it also compares the Kaula-North Arch transect with lavas from along the plume track closer to the present day plume center. Plume material is thought to disperse downstream from the plume center [Ribe and Christensen, 1999; Farnetani and Hofmann, 2010; Ballmer et al., 2011]. However it is unclear how it is distributed with distance from the plume center. To avoid this complexity, here we only discuss the isotopic gradients across the plume track from Kaula to North Arch.
 One interpretation for the larger proportion of the DRC at the periphery of the plume is that it represents entrained mantle [Fekiacova et al., 2007; Ballmer et al., 2011] that is similar to, but compositionally distinct from, the source of Pacific MORB. In this scenario the enriched plume component is not entirely exhausted during the shield stage, and contributes to the rejuvenated volcanism downstream from the plume center (e.g., Kauai), leading to less radiogenic Nd, Hf and more radiogenic Pb isotopes than the peripheral (Kaula, North Arch) rejuvenated volcanism. Alternatively, if there is no significant upper mantle entrainment during plume upwelling [Farnetani et al., 2002] then the DRC might be traced to the source of the plume (perhaps the core mantle boundary) together with the LOA and KEA components [Farnetani and Hofmann, 2010; Farnetani et al., 2012]. The isotope systematics would then require a greater proportion of the enriched components (e.g., LOA, KEA) in the plume center than in the periphery, and proportionally larger contribution from the enriched plume core to the rejuvenated volcanism downstream from the plume center (e.g., Kauai) than the off-axis volcanism (Kaula, North Arch). In either scenario however, there is evidence for a greater proportion of the DRC in the rejuvenated magma source at the periphery of the plume than downstream from the plume center. Additional data from the off-axis rejuvenated volcanism could further test this hypothesis and may help distinguish whether the DRC is entrained mantle versus a component integral to the Hawaiian plume.
 1. The Kaula pyroxenites formed as cumulates within the Pacific lithosphere from melts geochemically similar to the alkalic rejuvenated lavas erupted at Kaula volcano. The broadly similar major, trace element and isotope compositions of the Kaula and SLC pyroxenite xenoliths that erupted some 300 km apart suggest that such high pressure (>20 kb) crystal fractionates might be a common byproduct of Hawaiian volcanism, and perhaps oceanic island volcanism elsewhere.
 2. Nephelinites, phonolites and garnet pyroxenite xenoliths from Kaula Island have radiogenic Hf, Nd, and unradiogenic Sr isotope compositions, consistent with a time-integrated depleted peridotitic source. The Pb isotope compositions of the Kaula pyroxenites, nephelinites and Salt Lake Crater pyroxenites extend to the lowest 208Pb/204Pb reported in Hawaiian volcanism. These new data allow the composition of the isotopically depleted end-member component of the rejuvenated volcanism (DRC) to be redefined as 208Pb/204Pb = 37.35 ± 0.05, 206Pb/204Pb = 17.75 ± 0.03, εNd = 9–10, εHf = 16–17 and 87Sr/88Sr <0.70305. The rejuvenated magma compositions do not trend toward the Pacific MORB compositions, suggesting that DRC is distinct from the source of Pacific MORB. Based on the isotope systematics of the rejuvenated and KEA shield lavas we also propose a refined composition for the KEA end-member at εNd = 6–7, εHf=11–12, with 87Sr/86Sr = 0.7035–0.7036, 206Pb/204Pb = 18.7–18.8, 208Pb/204Pb = 38.3–38.4.
 3. The DRC is recognized in the compositions of xenoliths and lavas erupted in the last ∼4.2 Ma and >300 km apart along the Hawaiian Island chain. The isotopic continuum between KEA shield and rejuvenated magmas suggests a close spatial association between KEA and DRC. The DRC is also recognized in the compositions of the LOA-trend Koolau shield and Hualalai postshield lavas. The compositions of the ∼76 to 81 Ma Detroit Seamount lavas from the Emperor Seamount chain [Frey et al., 2005] are also consistent with the DRC composition proposed here. All these observations are consistent with the DRC being a long-lived component of the Hawaiian plume.
 4. The Kaula pyroxenites and North Arch rejuvenated lavas at the periphery of the plume have on average less radiogenic Pb and more radiogenic Nd and Hf isotope compositions than Kauai and Niihau rejuvenated lavas near the center of the plume track. This suggests a larger proportion of the depleted component at the periphery of the plume than along the Island chain during rejuvenated volcanism.
 These results indicate that the DRC, which is dominantly sampled during rejuvenated volcanism and currently best represented by the Kaula pyroxenites, is also widely available and contributes to the KEA shield, and (at least in some instances) LOA shield volcanism. This depleted component appears to be dispersed within the plume, but with a larger proportion at the periphery than the center.
 We would like to thank James Day, Fred Frey and Al Hofmann for their in-depth comments and suggestions that significantly improved this presentation, and Janne Blichert-Toft for editorial handling. Thanks to Shichun Huang for many discussions on Hawaiian volcanism. Many thanks to Elizabeth Bair and Carl Frisby at USC for help with the analyses. This work was supported by the National Science Foundation grants OCE-0622827, OCE-0820723, OCE-1129280 to MB and EAR-0510482 to MG.