Pacific Center for Isotopic and Geochemical Research, Department of Earth and Ocean Sciences, University of British Columbia, Vancouver, British Columbia, Canada
I. G. Nobre Silva, Pacific Center for Isotopic and Geochemical Research, Department of Earth and Ocean Sciences, University of British Columbia, 6339 Stores Road, Vancouver, BC V6T 1Z4, Canada. (email@example.com)
 The 3500 m deep Hawai'i Scientific Drilling Project core provides a ~680 kyr record of the magmatic history and source components of Mauna Kea volcano. We report high-precision Pb-Sr-Nd isotopic compositions of 40 basalts from the last 408 m of the final drilling phase (HSDP2-B and HSDP2-C) and show that these lowermost basalts represent the early shield stage of Mauna Kea's growth history. Two sample groups are distinguished based on their isotopic variability compared to the rest of the core. Over a depth interval of 210 m (3098.2–3308.2 mbsl), the basalts show very restricted isotopic variation and represent sampling of a relatively homogeneous source. Samples from the bottom 192 m record the largest range of 206Pb/204Pb and 208Pb/204Pb in the core, reflecting the greater isotopic variability of the earlier stages of volcanism compared to subsequent stages. The heterogeneity of Mauna Kea lavas is explained by mixing variable proportions of four distinct components intrinsic to the Hawaiian mantle plume. One of these components, Kea, is a prevalent and long-lived composition within the Hawaiian plume, whereas the other three components are involved at different stages of the volcano's history and contribute to the short-term isotopic variability of Mauna Kea. The compositional similarity of the Kea component to “C” and to the super-chondritic bulk-silicate Earth suggests that Kea may be part of the primitive mantle of a non-chondritic Earth. Other Pacific oceanic island basalts share Kea-like compositions, indicating that the Kea component is a common, widespread composition within the Pacific deep mantle.
 The age-progressive Hawaiian-Emperor volcanic chain in the Pacific Ocean is the classic example of intraplate hotspot volcanism attributed to a deep-seated mantle plume [e.g., Morgan, 1971; Courtillot et al., 2003; DePaolo and Weis, 2007]. This simple tectonic setting, together with the distant location from plate margins and large rates of magma flux, makes Hawaiian volcanoes one of the best places to study the chemical evolution and structure of mantle plumes and of the deep mantle [e.g., Staudigel et al., 1984; Sleep, 1990; Campbell and Griffiths, 1991; Hauri et al., 1994; DePaolo et al., 2001; Bryce et al., 2005; Farnetani and Hofmann, 2009; Weis et al., 2011]. As the Pacific plate moves across the Hawaiian mantle plume, at a velocity of 9–10 cm/yr, individual volcanoes grow and evolve through several stages (pre-shield, shield, post-shield, and a much later one, rejuvenated) as they sample different areas of the plume's melting region [e.g., Clague and Dalrymple, 1987]. The geochemistry of continuous eruptive sequences of individual volcanoes hence provides a record of the geochemical time series of the plume's melting region [e.g., Hauri et al., 1996; Lassiter et al., 1996; DePaolo et al., 2001; Bryce et al., 2005].
 Compositional variations occur over various time scales during the eruptive life of a Hawaiian volcano. Both short-term (decadal to centennial) and long-term (millennial) isotopic variations are recognized in shield lavas of Hawaiian volcanoes [e.g., Pietruszka and Garcia, 1999; Marske et al., 2007; 2008; Mauna Loa, Kurz and Kammer, 1991; Kurz et al., 1995; Rhodes and Hart, 1995; DePaolo et al., 2001; Mauna Kea, Blichert-Toft et al., 2003; Eisele et al., 2003; Kurz et al., 2004, Kauai, Salters et al., 2006; Fekiacova et al., 2007], all within the larger time scale (≥1 Ma) of volcano growth and compositional evolution [e.g., Frey et al., 1990; Garcia et al., 2006]. The ~3500 m deep core recovered by the Hawai'i Scientific Drilling Project (HSDP) constitutes the longest stratigraphically controlled record of the magmatic output of a single volcano sampled thus far. The analysis of the temporal isotopic covariations within the first ~3100 m of core allowed the identification of compositional heterogeneities within the Hawaiian plume as well as their maximum and minimum sizes [e.g., Blichert-Toft et al., 2003; Eisele et al., 2003; Abouchami et al., 2005; Bryce et al., 2005]. Additionally, direct comparison with lavas from other Hawaiian volcanoes erupted at equivalent stages of volcano growth led to the formulation of new models for the chemical structure of the Hawaiian plume and its deep mantle source [e.g., DePaolo et al., 2001; Blichert-Toft et al., 2003; Kurz et al., 2004; Abouchami et al., 2005; Bryce et al., 2005; Farnetani and Hofmann, 2010, Weis et al., 2011].
 Hawaiian volcanoes younger than 5 Ma form two sub-parallel chains, termed Kea and Loa [Dana, 1849; Jackson et al., 1972], that are systematically distinct chemically and isotopically [e.g., Tatsumoto, 1978; Frey and Rhodes, 1993; Abouchami et al., 2005]. The geochemical differences between volcanoes along the two chains reflect mixing of variable proportions of at least three distinct components [e.g., Staudigel et al., 1984; Stille et al., 1986; Eiler et al., 1996; Tanaka et al., 2008]. The distribution of these components is a matter of considerable debate with two main models of plume structure being invoked to explain such long-term geochemical differences: a concentrically zoned plume [e.g., Hauri et al., 1994, 1996; Lassiter et al., 1996; DePaolo et al., 2001] and a bilaterally asymmetrical plume [Abouchami et al., 2005; Weis et al., 2011; Huang et al., 2011]. Both purely concentrically and bilaterally zoned plume models are, however, unable to fully explain all the geochemical variability amongst Hawaiian basalts. As a result, numerous variations of these plume structure models incorporating vertical heterogeneities within the upwelling plume have been proposed [e.g., Frey and Rhodes, 1993; Blichert-Toft et al., 2003; Kurz et al., 2004; Abouchami et al., 2005; Bryce et al., 2005; Huang et al., 2005, 2009; Marske et al., 2007; Xu et al., 2007; Blichert-Toft and Albarède, 2009; Hanano et al., 2010; Farnetani and Hofmann, 2010].
 Any model for the Hawaiian mantle plume structure must account for the longevity of the Hawaiian Kea component. Kea-like compositions have been recognized in lavas from the Hawaiian-Emperor seamount chain back to >85 Ma [Regelous et al., 2003; Abouchami et al., 2005; Tanaka et al., 2008] and in mid-Cretaceous basalts preserved in Kamchatka [Portnyagin et al., 2008]. Recently, Weis et al.  showed that Loa-like isotopic compositions extend back at least 5 myr along the Hawaiian chain. Based on their comparative study of Loa and Kea shield lavas, Weis et al.  argued that the geochemical differences between Kea and Loa trend volcanoes can be traced to the core mantle boundary. The more variable and enriched compositions of Loa volcanoes are explained by sampling of the heterogeneous Pacific ultra low velocity zone (ULVZ) by the Loa side of the Hawaiian mantle plume, whereas the “average,” less heterogeneous compositions of Kea volcanoes are explained by sampling of deep Pacific mantle [Weis et al., 2011].
 The last 408 m of the HSDP sampled a series of submarine tholeiitic basalts that erupted in the early shield-building stage of Mauna Kea volcano. In this study, we use the Pb, Sr, and Nd isotopic compositions of 40 whole-rock samples to identify short-term isotopic fluctuations that reflect compositional heterogeneities sampled in the early phase of growth of Mauna Kea. We examine the >680 kyr isotopic record of Mauna Kea's magmatic history to evaluate both the short- and long-term isotopic variations within the drill core and provide constraints on the chemical structure of the Hawaiian mantle plume and deep Pacific mantle.
2 Hawai'i Scientific Drilling Project: Geological Setting and Core Stratigraphy
 The Hawai'i Scientific Drilling Project was a multidisciplinary international scientific effort to test the models of growth and chemical evolution of Hawaiian volcanoes by systematically sampling a continuous stratigraphic sequence of lavas from a large Hawaiian volcano [Stolper et al., 1996; DePaolo et al., 2001]. Mauna Kea was chosen since it is the youngest Hawaiian volcano that has completed its life cycle of major growth stages (pre-shield, shield, post-shield) [e.g., Clague and Dalrymple, 1987; Frey et al., 1990; 1991; Stolper et al., 1996; Garcia et al., 2007; Stolper et al., 2009]. The HSDP drill sites were located on the northeast side of the island of Hawai'i, on the east flank of Mauna Kea volcano, in an abandoned rock quarry adjacent to Hilo International Airport (Figure 1). Almost 250 m of Mauna Loa lavas overlies Mauna Kea lavas at this location, which allowed for comparative studies between the two consecutive volcanoes.
 The success of the first phase of the project (“pilot hole” HSDP1) [e.g., Stolper et al., 1996] led to the deep drilling of the HSDP2, a ~3500 m deep core that was recovered in two phases over a total period of almost 5 years. The main drilling phase of the HSDP2 was carried throughout 1999, during which time the hole reached 3110 meters below sea level (mbsl) [Garcia et al., 2007]. The second drilling phase was accomplished in two stages. Starting in 2003 with the enlargement of the hole's diameter, coring began in late 2004 and proceeded until early 2005 (phase B). Coring restarted in late 2006 and reached completion in early 2007 at 3508 mbsl (phase C) [Stolper et al., 2009].
 On the basis of observed contacts and variations in mineralogical, lithological, and structural features, a total of 389 lithological units were identified within the HSDP2 core [Garcia et al., 2007; Stolper et al., 2009] and five depth zones were recognized (Figure 1c): (1) sub-aerial Mauna Loa lavas (surface to 246 mbsl), (2) sub-aerial Mauna Kea lavas (246 to 1079 mbsl), (3) shallow submarine Mauna Kea lavas (1079 to 1984 mbsl), (4) deep submarine Mauna Kea lavas (1984 to 3098 mbsl), and (5) deep submarine Mauna Kea lavas from the final drilling phase (3098 to 3508 mbsl) [Rhodes and Vollinger, 2004; Garcia et al., 2007, Stolper et al., 2009]. In this final phase of drilling (HSDP2-B and HSDP2-C), 44 flow and intrusive units were identified, consisting primarily of submarine pillow lavas (~60%), with lesser amounts of hyaloclastites (~17%), massive volcanic units (~12%), pillow breccias (~10%), and intrusive units (~1%) [Stolper et al., 2009; Rhodes et al., 2012].
 The youngest dated Mauna Kea sample from the main phase of the HSDP2, an alkalic basalt at ~277 mbsl, has a 40Ar/39Ar age of 236 ± 16 ka, and the deepest dated sample, at a depth of 2789 mbsl, has an age of 683 ± 82 ka [Sharp and Renne, 2005]. Assuming a linear fit to the 40Ar/39Ar ages, Sharp and Renne  determined mean accumulation rates of ~9 m/kyr and ~0.9 m/kyr for the shield and post-shield sequences, respectively, over the 400 kyr of volcanism recorded by the ~2.7 km thick section of Mauna Kea basalts. Recently, Jourdan et al.  refined the age-depth relations in the deeper part of the HSDP2 core. After obtaining indistinguishable mean 40Ar/39Ar ages of 683 ± 130 ka for two tholeiitic basalts recovered from depths of 3278 and 3321 mbsl during HSDP2-B, these authors proposed that these ~3.3 km deep lava flows erupted at 681 ± 120 ka. No age has been determined on samples from the deepest 200 m of core (phase C). Assuming the 8.4 m/kyr shield lava accumulation rate of Jourdan et al. , modeled regression ages for samples from these deeper 200 m are inferred to range between ~668 and ~688 kyr.
3 Samples and Analytical Methods
 Forty Mauna Kea basalts from the whole-rock reference suite of the HSDP2-B and HSDP2-C were analyzed for Pb-Sr-Nd isotope compositions. The major and trace element compositions of these samples are reported in Rhodes et al. , and Hf-Pb-Nd isotope compositions were previously reported by Blichert-Toft and Albarède . All basalts are tholeiitic and cover a wide range of MgO contents (6.6–25.4 wt%). Most samples have high SiO2 contents (>50 wt%) and based on their trace element concentrations belong to type-1 (high SiO2, high Zr/Nb) and type-4 (high SiO2, low Zr/Nb) magma groups [Rhodes and Vollinger, 2004; Rhodes et al., 2012]. Exceptions are samples R6-2.15-3.0 and R184-1.15-2.1, two intrusive units at 3098.2 and 3400.9 mbsl, respectively, which are most similar to type-3 (low SiO2, low Zr/Nb) lavas; and samples R210-3.3-3.95 and R219-5.55-6.2, 7.25-7.7, at 3472.2 and 3500.9 mbsl, respectively, which have compositions most similar to those of Mauna Loa volcano [Rhodes and Vollinger, 2004; Rhodes et al., 2012].
 All isotopic analyses were performed on whole-rock powders (except for sample 26-102-R163-2.1-2.6 that was in glass chip form) that were prepared, rinsed, and pulverized following the procedures of Rhodes  and Rhodes and Vollinger . Given the time gap between the two sample recovery phases of the last 408 m of the HSDP2 core, the isotopic measurements presented in this study were obtained in two analytical sessions (mid-2006 for the 22 samples from phase B and early 2008 for the 18 samples from phase C). All chemical purification and mass spectrometric analyses were carried out in Class 100 and Class 10,000 clean laboratories, respectively, at the Pacific Center for Isotopic and Geochemical Research (PCIGR) at the University of British Columbia. For all samples, the Pb, Sr, and Nd isotopic compositions were determined on the same sample dissolution. All sample solutions were passed twice through Pb anion exchange columns for Pb purification. The fractions washed out from the Pb columns, containing all other sample matrix elements, were then processed through other chromatographic ion exchange columns for Sr and Nd purification, following the sequential chromatographic purification methods described in Weis et al. . Given the importance of efficiently removing alteration phases and any extraneous Pb contaminant to get accurate and reproducible Pb isotopic compositions of oceanic basalts [Hanano et al., 2009, Nobre Silva et al., 2009], all samples were thoroughly acid-leached prior to digestion and isotopic analysis following the sequential acid leaching procedure of Weis et al.  and Nobre Silva et al. [2009; 2010]. All mass spectrometric analyses by MC-ICP-MS and TIMS followed the procedures detailed in Weis et al. , and standard values measured during the analytical sessions are reported in the footnote of Table 1. Total procedural blanks were ~50, 180, and 45 pg for Pb, Sr, and Nd, respectively, which are negligible compared to the elemental concentrations in the samples. Two rock reference materials (USGS BHVO-2 and Hawaiian rock Kil-93) were also processed together with the sample set. Their values are reported in Table 1, together with the sample results.
Table 1. Pb, Sr, and Nd Isotopic Compositions by MC-ICP-MS and TIMS of Mauna Kea Samples From the Last Drilling Phases (B and C) of the HSDP2 Core
Pb isotopic ratios by MC-ICP-MS, normalized relative to the SRM 981 TS-TIMS reference values of Abouchami et al. ; the 2SE is the absolute error value of the individual sample analysis (internal error). During the two analytical sessions over which isotopic measurements were obtained (mid-2006 for the 22 samples from phase B and early 2008 for the 18 samples from phase C), analyses of the SRM 981 Pb standard (n = 140, and 36) yielded mean values of 206Pb/204Pb = 16.9406 ± 0.0013 and 16.9420 ± 0.0016, 207Pb/204Pb = 15.4964 ± 0.0025 and 15.4987 ± 0.0020, and 208Pb/204Pb = 36.7138 ± 0.0084 and 36.7156 ± 0.0062, respectively.
Sr isotopic ratios measured by TIMS, normalized relative to the SRM 987 standard solution value of 87Sr/86Sr = 0.710248 [Weis et al., 2006]; the 2σ error is the absolute error value of the individual sample analysis (internal error) reported as ×106. During the course of these analyses, the average 87Sr/86Sr value of the SRM 987 Sr standard was 0.710256 ± 0.000015 (n = 10) in 2006 and 0.710246 ± 0.000004 (n = 9) in 2008.
Nd isotopic ratios measured by TIMS (HSDP2-B samples) and MC-ICP-MS (HSDP2-C samples), normalized relative to the La Jolla standard solution value of 143Nd/144Nd = 0.511858 [Weis et al., 2006] and to the Rennes Nd standard value of 143Nd/144Nd = 0.511973 [Chauvel and Blichert-Toft, 2001], respectively. During the course of analyses, the average 143Nd/144Nd values of the La Jolla Nd standard was 0.511852 ± 0.000010 (n = 10) and the average value for the Rennes Nd standard was 0.512000 ± 0.000009 (n = 40).
εNd values are calculated using the 143Nd/144Nd CHUR value = 0.512638 [Jacobsen and Wasserburg, 1984].
rep = replicate analysis of the same sample solution by MC-ICP-MS.
dup = full procedural duplicate analysis of the same sample.
Sample from Kilauea's summit Pu'u 'O'o eruption collected in 1993 by M. O. Garcia, used as in-house reference material.
Isotopic values for USGS BHVO-2 are in good agreement with the published values by Weis et al. , and the values for Kil-93 are in good agreement with the in-house values for this Hawaiian rock (Nobre Silva et al., 2013).
4.1 Stratigraphic Variations in Pb-Sr-Nd Isotope Compositions
 Basalts from the last 408 m (phases B and C) of the HSDP2 core yield 206Pb/204Pb = 18.3033–18.6936, 207Pb/204Pb = 15.4707–15.4993, 208Pb/204Pb = 37.924–38.271, 87Sr/86Sr = 0.703513–0.703631, and 143Nd/144Nd = 0.512968–0.513011 (Table 1 and Figure 2). The Pb and Nd isotopic values are mostly similar to those of the overlying, younger lavas (above ~3098 mbsl; Figure 2) [e.g., Eisele et al., 2003; Blichert-Toft et al., 2003; Bryce et al., 2005] and are in good agreement with those of Blichert-Toft and Albarède  (Appendix A in the Supporting Information).1
 Between 3097.7 and 3308.2 mbsl, the isotopic variability of these older basalts is very restricted (206Pb/204Pb = 18.4953–18.5960, 208Pb/204Pb = 38.187–38.204, 87Sr/86Sr = 0.703595–0.703631, and 143Nd/144Nd = 0.512968–0.512990). Type-1 basalts are the most abundant in this interval, and the observed range of isotopic compositions correlates with their limited variation in major element oxide and trace element contents [Rhodes et al., 2012]. These 210.5 m represent the longest interval (~10 kyr) of restricted geochemical variation observed in the HSDP2 core; the relative variations in 206Pb/204Pb, 86Sr/87Sr, and 143Nd/144Nd are ~11×, ~4×, and ~3×, respectively, smaller than those observed for younger basalts (Figure 2). The exception is sample R6-2.15-3.0, a massive intrusive rock at 3098.2 mbsl with low SiO2 and Zr/Nb [Rhodes et al., 2012] that has isotopic compositions identical to other type-3 samples recovered higher in the core stratigraphy during the main phase of the HSDP2 [Blichert-Toft et al., 2003; Eisele et al., 2003; Bryce et al., 2005].
 Basalts from the last 192 m of drill core (3313.5 to 3505.7 mbsl) are most similar to type-4 basalts and show significantly greater Pb isotopic variations (Figure 2 and Table 1). As also noted by Blichert-Toft and Albarède , this group of 19 samples yields the largest range of Pb isotopic compositions (206Pb/204Pb = 18.3033–18.6936, 208Pb/204Pb = 37.924–38.291) compared to basalts from the entire ~3200 m of core above (Figure 2). These deeper basalts extend the Pb isotopic range (206Pb/204Pb and 208Pb/204Pb) of Mauna Kea to compositions that are both significantly more and less radiogenic than previously observed for this volcano [e.g., Abouchami et al., 2000; Eisele et al., 2003; Blichert-Toft et al., 2003]. This distinction is less obvious in Sr and Nd isotopic compositions (Figure 2). Collectively, the lowermost 192 m of basalts in the core display ~2× smaller Sr and Nd isotopic variations in comparison to the younger samples [Bryce et al., 2005] and show a systematic progression of increasing 87Sr/86Sr and increasing εNd with depth (Figure 2). In this deeper section of the core, few samples have major and trace element compositions (e.g., SiO2(13) > 49%; Zr/Nb >13) that trend toward compositions observed at Mauna Loa volcano [Rhodes et al., 2012]. Their Pb, Sr, and Nd isotope ratios are however distinct from those of Mauna Loa, especially their Nd isotopes, which are most similar to late-shield and post-shield lavas found in the upper Mauna Kea section of the drill core [e.g., Bryce et al., 2005; Hanano et al., 2010].
4.2 Isotope Correlations
 The Sr, Nd, and Pb isotopic compositions of the tholeiites from the HSDP2-B and HSDP2-C show some degree of correlation, consistent with the trends formed by other Hawaiian volcanoes (Figures 3-6). With respect to Sr-Nd isotopes, these older tholeiites lie within the compositional range of Mauna Kea, at the depleted end (low 87Sr/86Sr, high 143Nd/144Nd) of the array formed by the Hawaiian Islands (Figure 3). Compared to other Mauna Kea basalts, the 87Sr/86Sr and 143Nd/144Nd values of HSDP2-B and HSDP2-C basalts are displaced toward the lower and higher limits, respectively, of the isotopic ranges. In Sr-Pb and Nd-Pb isotope diagrams (Figures 4 and 5), these basalts extend from the compositional trend defined by other Mauna Kea and Kea-trend shield basalts toward slightly lower 87Sr/86Sr and higher 143Nd/144Nd values, similar to the compositions of the late stage, post-shield lavas from Mauna Kea and Kohala [Holcomb et al., 2000; Eisele et al., 2003; Bryce et al., 2005; Hanano et al., 2010].
 The HSDP2-B and HSDP2-C basalts form two distinct Pb isotope arrays that intersect at the radiogenic end (Figure 6), extending the range of 206Pb/204Pb and 208Pb/204Pb of Mauna Kea to significantly more radiogenic values, similar to those of “ancestral” Kilauea lavas [Kimura et al., 2006]. One Pb isotope array overlaps with Mauna Kea's main Pb isotopic compositional field, best represented by the “Kea-mid8” array defined by the majority (>75%) of the younger HSDP2 samples [Eisele et al., 2003]. The second array is constituted by samples R6-2.15-3.0 and R184-1.15-2.1 (two low SiO2 and low Zr/Nb intrusive units at 3098.2 and 3400.9 mbsl, respectively) that plot together with the low SiO2 basalts recovered higher in the core stratigraphy that define the “Kea-hi8” array [Eisele et al., 2003]. Samples R210-3.3-3.95 and R219-5.55-6.2; 7.25-7.7, at the bottom of the core, extend the Pb isotopic compositions of Mauna Kea to significantly less radiogenic values, plotting in between the “Kea-mid8” and “-lo8” Pb arrays. Although these samples have elemental compositions and 208Pb/204Pb that are within the range of values for Mauna Loa, their 206Pb/204Pb and 207Pb/204Pb (not shown) are higher, comparable to those of lavas from Kohala, Kilauea, and post-shield stage of Mauna Kea (Figure 6).
5.1 HSDP2: A Record of the Evolution of a Single Volcano or the Output of Different Volcanoes?
 Hawaiian volcanoes evolve through several growth stages that are marked by changes in composition and eruption rates as they are carried by the Pacific plate across the magma-production region of the Hawaiian mantle plume [e.g., Chen and Frey, 1983; Clague and Dalrymple, 1987; Moore and Clague, 1992]. Oceanic lithosphere does not significantly influence the isotopic compositions of Hawaiian lavas [e.g., Lassiter et al., 1996; Blichert-Toft et al., 2003; Fekiacova et al., 2007; Marske et al., 2007; Hanano et al., 2010]. The geochemical variations within the stratigraphic sequence of lavas derived from a single volcano therefore reflect the temporal variations within its magma source [e.g., Kurz and Kammer, 1991; Rhodes and Hart, 1995; Lassiter et al., 1996; Rhodes, 1996; Pietruszka and Garcia, 1999; DePaolo et al., 2001; Bryce et al., 2005; Abouchami et al., 2005; Marske et al., 2007, 2008]. As different volcanoes grow to form the Hawaiian Islands, their lavas overlap and interlay, with older volcanoes becoming partially covered by lavas from younger volcanoes [e.g., Moore and Clague, 1992; DePaolo and Stolper, 1996]. Mauna Kea is built on top of ~6 km thick Cretaceous oceanic crust plus pelagic and clastic sediments and, on the flank of the adjacent, older Kohala volcano [e.g., Moore and Clague, 1992; DePaolo and Stolper, 1996].
 It has been suggested that the low SiO2-high 208Pb*/206Pb* lavas encountered in the deeper section of the HSDP2 core may not represent the variable output of Mauna Kea but instead that of Kohala [Holcomb et al., 2000] or of another unknown volcano [e.g., Stolper et al., 2004; Blichert-Toft and Albarède, 2009]. The isotopic similarities between basalts dredged along the Hilo Ridge below ~1100 mbsl and sub-aerial lavas from Kohala volcano (Pololu and Hawi volcanic rocks), plus the older ages of the deeper section of the Hilo Ridge, suggest that it may be part of Kohala's southeast rift zone rather than belonging to Mauna Kea [Holcomb et al., 2000; Lipman and Calvert, 2011]. Based on the elemental and isotopic distinction between HSDP2 lavas shallower than 3098 mbsl and known Kohala samples, Rhodes and Vollinger  and Stolper et al.  argued against the presence of Kohala lavas within this section of the HSDP2 core, although they did not dismiss the possibility of encountering lavas from this volcano at deeper levels.
 The last phase of drilling of the HSDP2 extended the core to ~3500 mbsl. Within the deepest 408 m of the core, high SiO2 lavas of variable Zr/Nb are the most abundant [Rhodes et al., 2012]. The two low-SiO2 intrusive units found in this section have isotopic compositions that are indistinguishable from other Mauna Kea samples, except for their higher 208Pb/204Pb and 208Pb*/206Pb* values, that are Loihi-like (Figures 2-6). These two units may be feeder dikes for the other low SiO2-high 208Pb*/206Pb* lavas encountered higher in the core stratigraphy [Rhodes et al., 2012]. Some of the basalts at the bottom of the drill core do show isotopic similarities to late-shield (Pololu) and post-shield (Hawi) lavas from Kohala, as suggested by Holcomb et al. , but also to post-shield lavas from Mauna Kea [e.g., Eisele et al., 2003; Bryce et al., 2005; Hanano et al., 2010] (Figures 3-6). To assume that these samples in question are derived from Kohala implies that ~680 kyr ago, this volcano was reaching the end of its evolution. This is not supported by growth models for Kohala, which at this time should have been in its vigorous tholeiitic shield-building stage [Moore and Clague, 1992; Lipman and Calvert, 2011], nor it is supported by the ages of sub-aerially exposed Kohala post-shield lavas (~175–450 ka) [Clague and Dalrymple, 1987; Aciego et al., 2010]. Although little is known about the history and interactions between Hawaiian volcanoes [Stolper et al., 2004], it is difficult to geometrically reconcile the existence of an unknown volcano older than Mauna Kea [e.g., Baker et al., 2003]. The overall isotopic consistency of the basalts deeper in the HSDP2 core stratigraphy with those from the overlying younger Mauna Kea lavas (Figures 3-6) attests to the continuity of the Mauna Kea section and to the relative homogeneity of the Mauna Kea source throughout its evolution.
 The ~3263.7 m long Mauna Kea section of the HSDP2 core represents the continuous record of the last ~680 kyr of volcanic activity of the Mauna Kea volcano as it crossed ~60–80 km over the melting region of the Hawaiian mantle plume [e.g., Stolper et al., 2009]. Assuming a model ~1.5 Ma lifetime for typical Hawaiian volcanoes [e.g., Garcia et al., 2006], it is unlikely that the early growth history (pre-shield stage) of Mauna Kea was sampled by the deep HSDP2 core. This is supported by the fact that the deeper section of the core is constituted solely of tholeiitic basalts. Nevertheless, the duration of individual volcanoes is likely to vary depending on the proximity of the volcano track to the center of the hotspot's melting region, which influences the amount of magma supplied to each volcano [e.g., DePaolo and Stolper, 1996; DePaolo et al., 2001; Baker et al., 2003]. According to the model of DePaolo and Stolper , Mauna Kea likely started to grow at ~1.050 Ma. This would place the bottom of the HSDP2 core (>680 kyr) close to the transition between the pre-shield and shield growth stages.
 The older basalts recovered from deeper in the HSDP2 core stratigraphy show greater isotopic variability compared to the overlying younger tholeiitic basalts and are isotopically similar, especially in 206Pb/204Pb, 87Sr/86Sr, and 143Nd/144Nd, to Mauna Kea post-shield lavas (Figures 2-6). Together with the overall lower Zr/Nb of these basalts, characteristic of Kilauea and Loihi [Rhodes and Vollinger, 2004, Rhodes et al., 2012], this suggests that the deeper basalts of the HSDP2 core may represent the very early shield stage of Mauna Kea's growth. At this time, the degrees of melting would have increased enough to produce tholeiitic compositions; however, the volcano's capture zone must have still been close enough to the edge of the plume's melting region to sample a similar compositional domain to that sampled later during the post-shield phase.
5.2 HSDP2 Isotope Variability and the Hawaiian Source Components
 The isotopic heterogeneity amongst Hawaiian shield basalts is explained by mixing of at least three isotopically distinct source components [e.g., Staudigel et al., 1984; Stille et al., 1986; Eiler et al., 1996; Hauri et al., 1996]. These include: a relatively “depleted” component (with low 87Sr/86Sr, 3He/4He, 207Pb/204Pb, and δ18O, and high 206Pb/204Pb, 208Pb/204Pb, 143Nd/144Nd, and 176Hf/177Hf), best observed in basalts from Kilauea (especially from the Hilina bench) [e.g., Chen et al., 1996; Abouchami et al., 2005; Kimura et al., 2006] and Mauna Kea, referred to as the “Kea” component; (2) a “modestly depleted” component (with low 87Sr/86Sr; high 3He/4He, 143Nd/144Nd, and 176Hf/177Hf; and higher 208Pb/204Pb), best expressed in basalts from Loihi, referred to as the “Loihi” component; and (3) an “enriched”component (with high 87Sr/86Sr, 207Pb/204Pb, and δ18O, and low 206Pb/204Pb, 208Pb/204Pb, 143Nd/144Nd, and 176Hf/177Hf), best recognized in basalts from Koolau and Lanai, referred to as the “Koolau” component [e.g., Eiler et al., 1996]. Two end-member compositions have been proposed to contribute to the Koolau component [e.g., Tanaka et al., 2002; 2008; Fekiacova et al., 2007]. The enriched Koolau end-member, observed in the Makapuu stage lavas, is widely referred to as the “enriched Makapuu component” (EMK). However, two different isotopic compositions have been attributed to the depleted Koolau end-member. Fekiacova et al.  considered that the Kahili stage lavas represent best the depleted Koolau end-member and referred to it as the “Kahili” component, whereas Tanaka et al.  considered the depleted Koolau end-member to also include rejuvenated-stage lavas and termed it the “depleted Makapuu component” (DMK).
 The isotopic variability of a single Hawaiian volcano can similarly be explained by mixing different proportions of these components. Koolau (or enriched Makapuu) and Loihi components are extreme compositions of the Loa trend volcano variability, and Kea is the major component in Kea trend volcanoes [e.g., Eiler et al., 1996; Kimura et al., 2006; Tanaka et al., 2008; Weis et al., 2011]. Based on the Pb isotope variations within the first 3100 m of the HSDP2 core, Eisele et al.  recognized the need of four distinct components in the Mauna Kea source to explain the geometry of the three “Kea” Pb arrays. The Pb isotope variability of the basalts recovered in the deeper 408 m of the HSDP2 core is consistent with this isotopic end-member scenario, involving a radiogenic Pb end-member, most similar to the Kea component, and three other distinct end-members of unradiogenic Pb isotopic compositions with distinct 208Pb/204Pb, most similar to “Loihi,” EMK, and DMK components (Figure 6). Whereas Kea is a common component throughout the long-term evolution of Mauna Kea [Tanaka et al., 2008], “Loihi-,” EMK-, and DMK-like components were involved at different stages of the volcano's eruptive history and each contributed to the short-term isotopic variability of Mauna Kea as represented by the different “Kea-hi8,” “-mid8,” and “-lo8” arrays of Eisele et al. , respectively (Figure 6). The two low-SiO2 intrusive units within the deeper stratigraphic section of the core plot along the “Kea-hi8” array of Eisele et al. , together with the other low SiO2-high 208Pb*/206Pb* lavas higher in the stratigraphic section of Mauna Kea. This restricted group of basalts of Loihi-like characteristics may reflect the presence of a local small-scale “Loa” heterogeneity within the melting regime of Mauna Kea [Abouchami et al., 2005; Rhodes et al., 2012]. It may also indicate that the component responsible for producing low SiO2-high 208Pb*/206Pb* lavas was sampled early in Mauna Kea's history and not only for a short time period (~45 kyr, corresponding to a depth interval of ~900 m) of its evolution. Basalts at depths of 3098.2 to 3308.2 mbsl form the longest interval (~10 kyr) of limited compositional variation within the sampled output record of Mauna Kea and likely reflect the sampling of a more homogeneous domain within the magma source region of Mauna Kea.
5.3 Chemical Structure of the Hawaiian Plume During the Growth of Mauna Kea
 The geochemical differences observed between volcanoes younger than 5 Ma along the two sub-parallel Kea and Loa volcanic chains likely result from sampling different components within different domains of the Hawaiian mantle plume [e.g., Tatsumoto, 1978; Frey and Rhodes, 1993; Hauri et al., 1996; Blichert-Toft et al., 2003; Abouchami et al., 2005; Weis et al., 2011]. The similarity of the Sr and Nd isotopic compositions of basalts from the deeper section of the HSDP2 core to those of lavas erupted during the post-shield stage indicates that during both the early and late stages of Mauna Kea's growth, the volcano sampled a compositionally similar domain within the plume's melting region. This would be consistent with a concentrically zoned chemical structure of the Hawaiian plume [e.g., Hauri et al., 1994, 1996; Lassiter et al., 1996; DePaolo et al., 2001; Bryce et al., 2005]. However, the Pb isotopic compositions of the deeper basalts of Mauna Kea's HSDP2 core differ from the respective post-shield lavas, having higher 208Pb/206Pb (Figure 6). This indicates that early shield stage lavas sampled a domain of the plume characterized by higher Th/U than the post-shield stage lavas. In a comparative study of the geochemistry of post-shield and shield lavas from consecutive Kea and Loa volcano pairs, Hanano et al.  concluded that the post-shield lavas on the Big Island retain their Kea- and Loa-like Pb isotope signatures, which does not support a concentrically zoned plume structure.
 The isotopic similarity of the older Mauna Kea HSDP2 basalts to pre-historic and young Kilauea basalts [Kimura et al., 2006; Marske et al., 2007, 2008; Blichert-Toft and Albarède, 2009] supports the proposals that heterogeneities within the Hawaiian plume are sampled by the melting regions of consecutive Kea volcanoes [Abouchami et al., 2005; Farnetani and Hofmann, 2010]. Vertical heterogeneity appears to be an intrinsic feature of the Hawaiian plume that is superimposed on radial heterogeneity derived from the thermal structure of the plume [e.g., Hauri et al., 1994, 1996]. The Sr and Nd isotopic compositions of Hawaiian basalts follow the concentric thermal structure of the Hawaiian plume, increasing and decreasing, respectively, during the lifetime of Hawaiian volcanoes as the potential temperature of the plume varies from the periphery to the core [e.g., DePaolo et al., 2001; Bryce et al., 2005]. In contrast, the Pb isotopic compositions sampled during the lifetime of Hawaiian volcanoes support an aspect of bilateral asymmetry in the distribution of compositional heterogeneities in the Hawaiian plume. These heterogeneities may be restricted to the “core zone” of the plume [Bryce et al., 2005] or distributed through the entire plume radius [e.g., Abouchami et al., 2005; Farnetani and Hofmann, 2010].
5.4 The Nature of the Hawaiian Kea Component and the Deep Pacific Mantle
 The enriched nature of the Koolau component is consistent with incorporation of ancient subducted oceanic crust and sediments into the source of the Hawaiian plume [e.g., Lassiter and Hauri, 1998; Blichert-Toft et al., 1999; Tanaka et al., 2008], whereas the depleted nature of the Kea component continues to be a matter of debate. On the basis of Sr and Nd isotopic compositions, Kea was first interpreted to result from entrainment of depleted asthenospheric mantle [Lassiter et al., 1996], but O, Os, and Pb isotopic studies precluded this hypothesis and instead support assimilation of Pacific lithosphere [Eiler et al., 1996], or the presence of recycled oceanic lithosphere [Lassiter and Hauri, 1998] or “young HIMU” material (recycled oceanic crust younger than 1.5 Ga) in the Hawaiian plume source [Thirlwall, 1997; Eisele et al., 2003]. Some of the basalts recovered in the bottom 408 m of the HSDP2 core, together with the pre-historical Kilauea basalts from the Hilina bench [Kimura et al., 2006], have the most “depleted” Sr, Nd, and Hf isotopic characteristics and the most radiogenic Pb compositions of all Hawaiian shield basalts. This suggests that during the early shield-building phase of a Kea trend volcano, the Kea component is sampled in higher proportions compared to later periods of volcano growth. The compositions of these basalts erupted early in the growth history of a Kea trend volcano thus provide insight into the nature of the Hawaiian Kea component.
 The correlation between high 143Nd/144Nd, 208-7-6Pb/204Pb, and low 87Sr/86Sr of Kea compositions reflects derivation from a source that developed high Sm/Nd and (U,Th)/Pb with lower Th/U and Rb/Sr over time. Such geochemical characteristics are normally attributed to a HIMU-like source, generated by recycling of ancient subduction-modified oceanic lithosphere [e.g., Hofmann and White, 1982; Zindler and Hart, 1986; Chauvel et al., 1992; Stracke et al., 2005; Willbold and Stracke, 2006]. Compared to Kea, HIMU-like compositions have much more radiogenic Pb isotopic signatures, which indicates that the Kea component must be derived from a source with significantly lower μ (238U/204Pb) values to produce 206Pb/204Pb ratios below 19.
 The Hawaiian plume is one of the most productive mantle plumes on Earth, having erupted ~7 × 106 km3 of volcanic material over the formation of the Hawaiian-Emperor chain [Vidal and Bonneville, 2004], and Kea-like compositions have been erupted since the mid-Cretaceous [e.g., Regelous et al., 2003; Abouchami et al., 2005; Portnyagin et al., 2008; Tanaka et al., 2008]. When comparing radiogenic (Pb, Sr, Nd, Hf) isotopic compositions of Hawaiian basalts to other Pacific Ocean island groups, the Kea component occupies an intermediate position in binary diagrams, toward which the general trends formed by other OIB (from EM-I, EM-II, and HIMU) converge (Figures 7 and 8). Basalts from the great Ontong Java plateau [Tejada et al., 2004] and from the Wrangellia oceanic plateau [Greene et al., 2008] also trend toward Kea-like compositions. The isotopic characteristics of Kea are not very different from those of the common mantle component “C” [Hanan and Graham, 1996] or “PREMA” [Zindler and Hart, 1986]. This is in agreement with the proposition of Weis et al.  that Kea trend volcanoes, compared to Loa trend volcanoes, sample “average” Pacific mantle compositions. In the super-chondritic Earth model reference frame [e.g., Boyet and Carlson, 2005, 2006; Caro et al., 2008; Caro and Bourdon, 2010; Jackson et al., 2010], the isotopic compositions of Kea are very close to those of bulk-silicate Earth. The resemblance of the Hawaiian Kea component to non-chondritic primitive mantle compositions [Jackson et al., 2010; Jackson and Carlson, 2011] suggests that Kea is itself part of the primitive mantle of a non-chondritic Earth. Moreover, the presence of Kea-like compositions amongst other Pacific ocean island basalts suggests that this non-chondritic mantle composition is a common and widespread composition within the Pacific deep mantle.
 High-precision Sr, Nd, and Pb isotopic compositions of basalts from the bottom 408 m of the ~3500 m deep HSDP2 drill core on Mauna Kea reveal compositional continuity with the overlying basalts and the presence of two distinct sample groups. Basalts from the top 210 m form the longest interval (~10 kyr) of limited variation within the sampled output record of Mauna Kea and correspond to the sampling of a more homogeneous domain within the magma source region of Mauna Kea. Basalts from the bottom 192 m show the largest range of variation in their Pb isotopic compositions, extending the isotopic compositions of Mauna Kea to significantly more radiogenic values, similar to those of “ancestral” Kilauea lavas, and also to significantly less radiogenic values, similar to those of late-shield and post-shield lavas (<400 kyr) from Mauna Kea and Kohala. Based on their ages (>680 kyr) and compositional characteristics, these basalts likely erupted in the very early shield phase of Mauna Kea, rather than representing part of the output of Kohala or another unknown volcano. The isotopic compositions of the HSDP2 basalts are consistent with the presence of four source components during the growth of Mauna Kea and include the “Kea,” “Loihi,” EMK, and DMK components. Kea is the prevailing component throughout the evolution of Mauna Kea, whereas the remaining three components are involved in different stages of the volcano development and contribute to the short-term isotopic variability of Mauna Kea basalts. In Pb-Sr-Nd isotope binary diagrams, Kea occupies an intermediate position toward which the general trends formed by other Pacific Ocean island groups (from EM-I, EM-II, and HIMU) converge. The Kea component is not only the common composition within the Hawaiian mantle plume but also a common composition within the deep Pacific mantle.
 We thank Bruno Kieffer and Claude Maerschalk for help in the clean lab and with TIMS analyses, and Jane Barling for assistance in operating the MC-ICP-MS. We are grateful to Donald DePaolo for providing the samples and to Michael Garcia, Albrecht Hofmann, and Elspeth Barnes for scientific discussions. We thank reviewers William White and J. Michael Rhodes, and Editor Joel Baker for their insightful comments. I. Nobre Silva was supported by the POCTI program of the Fundação para a Ciência e Tecnologia (Portugal). This research was funded by NSERC Discovery Grants (Canada) to D. Weis and J. S. Scoates.
All Supporting Information may be found in the online version of this article.