The radiogenic isotopic ratios of Sr, Nd, Hf, and Pb in basaltic lavas associated with major hot spots, such as Hawaii, document the geochemical heterogeneity of their mantle source. What processes created such heterogeneity? For Hawaiian lavas there has been extensive discussion of geochemically enriched source components, but relatively little attention has been given to the origin of depleted source components, that is, components with the lowest 87Sr/86Sr and highest 143Nd/144Nd and 176Hf/177Hf. The surprisingly important role of a depleted component in the source of the incompatible element-enriched, rejuvenated-stage Hawaiian lavas is well known. A depleted component also contributed significantly to the ∼76–81 Ma lavas erupted at Detroit Seamount in the Emperor Seamount Chain. In both cases, major involvement of MORB-related depleted asthenosphere or lithosphere has been proposed. Detroit Seamount and rejuvenated-stage lavas, however, have important isotopic differences from most Pacific MORB. Specifically, they define trends to relatively unradiogenic Pb isotope ratios, and most Emperor Seamount lavas define a steep trend of 176Hf/177Hf versus 143Nd/144Nd. In addition, lavas from Detroit Seamount and recent rejuvenated-stage lavas have relatively high Ba/Th, a characteristic of lavas associated with the Hawaiian hot spot. It is possible that a depleted component, intrinsic to the hot spot, has contributed to these young and old lavas related to the Hawaiian hot spot. The persistence of such a component over 80 Myr is consistent with a long-lived source, i.e., a plume.
 The diverse geochemical characteristics of ocean island basalt (OIB) commonly attributed to mantle plumes, such as Iceland, Galapagos and Hawaii, reflect, in part, processes that create geochemical heterogeneity in the mantle. For example, geochemical data for such OIB have been used to argue that their mantle sources include materials recycled into the mantle at subduction zones, i.e., sediments, upper and lower igneous crust and lithospheric mantle [e.g., Lassiter and Hauri, 1998; Blichert-Toft et al., 1999; Chauvel and Hémond, 2000; Skovgaard et al., 2001; Niu and O'Hara, 2003]. Although the origin of enriched components, commonly referred to as EM1, EM2 and HIMU, in OIB has been discussed extensively [e.g., Zindler and Hart, 1986; Hart, 1988; Hofmann, 1997], the equally important role of depleted components in OIB has received less attention. Compared to primitive mantle, depleted components have relatively low 87Sr/86Sr, high 143Nd/144Nd and high 176Hf/177Hf.
 Like the mantle sources of Icelandic and Galapagos lavas, the source of Hawaiian lavas is geochemically heterogeneous. The scale of this heterogeneity is reflected by geochemical variability within an individual Hawaiian volcano. For example, 87Sr/86Sr decreases and 143Nd/144Nd increases as Haleakala Volcano evolved from shield-stage to postshield-stage volcanism [Chen and Frey, 1985]. Furthermore at a given Hawaiian volcano, rejuvenated-stage lavas have the lowest 87Sr/86Sr and highest 143Nd/144Nd. An important aspect of this temporal trend is the role of MORB-related depleted components, i.e., asthenospheric source of MORB or oceanic lithosphere. Since rejuvenated-stage lavas erupt after a volcano has migrated away from the hot spot, a common interpretation [e.g., Chen and Frey, 1985; Yang et al., 2003] is that MORB-related asthenosphere or lithosphere was a source component.
 Studies of drill core from Detroit Seamount, a 76–81 Ma volcanic complex in the northern Emperor Seamount Chain [Duncan and Keller, 2004] document another example of a depleted source component in lavas attributed to the Hawaiian hot spot [Keller et al., 2000; Regelous et al., 2003; Huang et al., 2005]. On the basis of the proximity of Detroit Seamount to a spreading ridge axis at ∼80 Ma, two alternative hypotheses have been proposed. Keller et al.  proposed that MORB volcanism at the spreading ridge dominated the magma flux originating from the hot spot. In contrast Regelous et al.  emphasized that the trend to unradiogenic Pb isotopic ratios in Detroit Seamount lavas is unlike that of most Pacific MORB; they inferred that melting of a depleted plume component was facilitated by plume ascent beneath young and thin oceanic lithosphere. Using geochemical data for lavas from newly drilled sites on Detroit Seamount, Huang et al.  evaluated these alternative hypotheses. They concurred with Regelous et al.  that Pb isotopic data for lavas from Detroit Seamount define a trend to low 206Pb/204Pb that does not overlap with the field of Pacific MORB. However, Huang et al.  noted that lavas erupted within the Garrett transform fault at 13°28′S on the East Pacific Rise (EPR) have lower Pb isotope ratios than the most extreme lavas from Detroit Seamount. Hence a suitably unradiogenic Pb component is present in the Pacific asthenosphere, but MORB with such unradiogenic Pb isotope ratios are uncommon [Huang et al., 2005].
 We emphasize that there are similarities in Sr, Nd and Pb isotopic ratios between rejuvenated-stage lavas erupted at Hawaiian volcanoes and Detroit Seamount lavas. Note that in our discussion we include the geochemically similar North Arch lavas with rejuvenated-stage lavas [e.g., Yang et al., 2003]. In this paper, we report Hf isotopic data for rejuvenated-stage and Emperor Seamount lavas. In our discussion, we use isotopic data for Sr, Nd, Hf and Pb and incompatible element abundance ratios to evaluate whether the depleted components that contributed to rejuvenated-stage lavas erupted in the Hawaiian Islands (<5 Ma) and lavas forming Detroit Seamount (76–81 Ma) are intrinsic to the Hawaiian plume.
 The samples studied range from some of the oldest lavas associated with the Hawaiian hot spot (Meiji, Detroit, and Suiko Seamounts) in the Emperor Seamount Chain to the relatively young rejuvenated-stage and North Arch lavas erupted in the Hawaiian Islands (Figure 1). All samples analyzed for Hf isotope ratios (Tables 1a and 1b) have been previously analyzed for Sr, Nd and Pb isotope ratios. They include 4 samples from Meiji Seamount obtained during DSDP Leg 19 [Regelous et al., 2003], 28 samples recovered from Detroit Seamount on ODP Legs 145 and 197 [Regelous et al., 2003; Huang et al., 2005], 4 samples from Suiko Seamount obtained during DSDP Leg 55 [Lanphere et al., 1980], 16 samples from rejuvenated-stage lavas, 9 from the Honolulu Volcanics, 7 from the Koloa Volcanics [Reiners and Nelson, 1998; Lassiter et al., 2000], and 4 samples from the North Arch volcanic field which are geochemically similar to rejuvenated-stage lavas [Frey et al., 2000; Yang et al., 2003]. In addition, garnet and clinopyroxene separates from three garnet pyroxenite xenoliths from Salt Lake Crater, Oahu were analyzed for Hf and Nd isotopic ratios.
Table 1a. Hf Isotopic Ratios of North Arch and Hawaiian Rejuvenated-Stage Lavas
Sr, Nd, and Pb isotopic ratios are reported by Lassiter et al. , except for Sample PF-1, which has 87Sr/86Sr = 0.70335, 143Nd/144Nd = 0.513050, 206Pb/204Pb = 18.05, 207Pb/204Pb = 15.44, and 208Pb/204Pb = 37.71 (J. Lassiter, unpublished data).
 Following the procedures described by Blichert-Toft et al. , Hf isotopic compositions were measured by MC-ICP-MS using a VG Plasma 54 at the Ecole Normale Supérieure in Lyon, France. In order to monitor machine performance, the JMC-475 Hf standard was analyzed after every two samples and gave 0.282160 ± 0.000010 (2 sigma) for 176Hf/177Hf throughout this study, corresponding to an external reproducibility of 35 ppm. 176Hf/177Hf was normalized for mass fractionation relative to 179Hf/177Hf = 0.7325. Hafnium total procedural blanks were less then 25 pg. Uncertainties reported on Hf measured isotope ratios are in-run 2σ/√n analytical errors, where n is the number of measured isotope ratios (Tables 1a and 1b).
 For Hf and Nd isotope analysis 100–200 mg of garnet and clinopyroxene were separated from three garnet pyroxenite xenoliths collected at Salt Lake Crater, Oahu (Table 2). Grains with inclusions were excluded. In order to minimize blank levels, the hand-picked minerals were not crushed, but dissolved whole. Prior to dissolution, the mineral grains were acid leached in an ultrasonic bath using both HF and HCl to eliminate grain coatings. Due to the refractory nature of garnet, the garnet separates were dissolved in steel-jacketed Teflon bombs at 160°C for one week. For consistency, the clinopyroxene separates were treated similarly. Upon complete dissolution, a Hf-bearing fraction was first separated from a REE-bearing fraction on a cation-exchange column. Nd was subsequently isolated from the latter on an HDEHP column, while Hf was further purified, first through an anion-exchange column to remove remaining matrix elements, then through a cation-exchange column serving to separate Ti and some Zr from the Hf [Blichert-Toft et al., 1997, 2002; Blichert-Toft, 2001]. Total procedural Nd and Hf blanks were <200 and <25 pg, respectively.
Table 2. Nd-Hf Isotopic Ratios of Clinopyroxene and Garnet From Salt Lake Crater Garnet Pyroxenite Xenoliths
 Isotopic analyses of Nd were carried out by MC-ICP-MS on the VG Plasma 54 instrument in Lyon following the procedures described by Blichert-Toft et al. . In order to monitor machine performance, the La Jolla standard was run systematically before and after each sample and gave, throughout this study, 0.511858 ± 0.000018 for 143Nd/144Nd (two standard deviations), corresponding to an external reproducibility of 35 ppm.
4.1. Comparison of Isotopic Data (Sr, Nd, Hf, and Pb) for Rejuvenated-Stage/North Arch and Detroit Seamount Lavas
Figure 2a shows the well-known observation that despite their relative enrichment in highly incompatible elements, the alkalic rejuvenated-stage and North Arch samples have lower 87Sr/86Sr and trend to higher 143Nd/144Nd than the field for tholeiitic Hawaiian shield lavas. Relative to the EPR MORB field they have higher 87Sr/86Sr and trend to lower 143Nd/144Nd; consequently, the field for EPR MORB does not lie on an extrapolation of the Hawaiian field (Figure 2a; see Table 3).
Table 3. Parent/Daughter Ratios Used to Calculate Initial Ratios for Fields in Figures 2–6
Figure 2b shows that lavas from Detroit Seamount range from overlap with the field for rejuvenated-stage/North Arch lavas to overlap with the EPR MORB field. Although most lavas from Detroit Seamount are also slightly offset from the MORB field to high 87Sr/86Sr, this offset is dependent upon the age corrections. The isotopic overlap of Site 1203 lavas, a 453 m section of Detroit Seamount with tholeiitic basalt overlying alkalic basalt [Huang et al., 2005], with the field for alkalic, rejuvenated-stage/North Arch lavas is surprising. The simplest inference is mixing between MORB-related and plume-related components. At Detroit Seamount an important role for a MORB component is consistent with the proximity of the plume to a spreading ridge center at ∼80 Ma [Keller et al., 2000]. For rejuvenated-stage lavas involvement of MORB-related asthenosphere or lithosphere has been proposed as the volcano migrates away from the hot spot [e.g., Chen and Frey, 1985; Reiners and Nelson, 1998; Yang et al., 2003].
 Although Sr-Nd isotopic variations are consistent with a MORB-related component, plots of Pb isotopic ratios show that the fields for rejuvenated-stage/North Arch lavas and Detroit Seamount lavas range to nonradiogenic ratios that are not typical of EPR MORB (Figures 3a–3e) [Regelous et al., 2003; Huang et al., 2005]. Such low ratios are characteristic of lavas erupted at a spreading center within the Garrett transform fault, but the rarity of such MORB (Figure 3e; see Huang et al.  for a detailed discussion) have led to alternative proposals. For Detroit Seamount, Regelous et al.  proposed that higher extents of melting below a young and thin oceanic lithosphere enabled sampling of a depleted plume component with relatively low Pb isotopic ratios. Similarly, the Pb isotopic trends for rejuvenated-stage and North Arch lavas have led to the conclusion that a MORB-related component is not the depleted component present in these lavas (Fekiacova and Abouchami  and T. Kani et al. (Multiple components involved in North Arch volcanism, Hawaii: Evidence from Pb and Sr isotope compositions of submarine lavas, submitted to Journal of Petrology, 2004; hereinafter referred to as Kani et al., submitted manuscript, 2004), respectively).
 Tholeiitic basalt forming the Hawaiian shields defines a negative 87Sr/86Sr versus 206Pb/204Pb trend (Figure 4). In contrast, West et al. [1987, Figure 3] noted that postshield alkalic lavas from Haleakala volcano define a positive 87Sr/86Sr versus 206Pb/204Pb trend, and they inferred that these alkalic lavas contain a depleted mantle component that is not present in shield lavas. Is this depleted mantle component related to EPR MORB? The field for rejuvenated-stage lavas shows variable 206Pb/204Pb at relatively uniform 87Sr/86Sr; i.e., there is no trend toward the EPR MORB field (Figure 4a). The positive 87Sr/86Sr versus 206Pb/204Pb trend defined by lavas from Detroit Seamount clearly extends to lower 206Pb/204Pb than the EPR MORB field (Figure 4b). As in Figures 3a–3e, Detroit Seamount lavas extend toward the field for lavas erupted within the Garrett transform fault.
 The 176Hf/177Hf versus 143Nd/144Nd trend has provided the strongest evidence that the depleted Icelandic component is unlike MORB [e.g., Kempton et al., 2000; Fitton et al., 2003]; namely depleted Icelandic basalt has higher 176Hf/177Hf at a given 143Nd/144Nd than the field for North Atlantic MORB (filtered to exclude Hf isotopic ratios whose within-run standard errors are >100 ppm [see Fitton et al., 2003, Figure 4]). Although the Pacific MORB field for 176Hf/177Hf versus 143Nd/144Nd is large with numerous outlier points, it is not significantly decreased by using the data quality filter of Fitton et al.  (see Figure 5a). In Figures 5a–5d, rejuvenated-stage Hawaiian lavas are within the broad positive trend defined by Hawaiian shield lavas and Pacific MORB. Bizimis et al. [2003a, p. 56] noted that data (six samples) for the Honolulu Volcanics from Stille et al.  define a significantly steeper 176Hf/177Hf versus 143Nd/144Nd slope than the OIB/MORB array. However, our new and significantly more precise Hf data (9 samples) for the Honolulu Volcanics do not confirm this steep slope (Figure 5d). In general, however, rejuvenated-stage lavas overlap with the Pacific MORB field in 176Hf/177Hf but are offset to lower 143Nd/144Nd (Figures 5a–5d).
 Among Emperor Seamount lavas, Site 884 lavas are distinct in that they plot within the Pacific MORB field (Figure 5b). In contrast, samples from Sites 883, 1203 and 1204 on Detroit Seamount and Meiji and Suiko Seamounts are at the low 143Nd/144Nd boundary of the Pacific MORB field, and they define a steeper 176Hf/177Hf versus 143Nd/144Nd trend than the mantle array defined by most OIB and MORB (Figure 5b). Similarly steep Hf-Nd isotope ratio trends have been reported for Emperor Seamount lavas by Kempton and Barry , Kempton et al. , and Thompson et al. . A steep trend extrapolating to high 176Hf/177Hf at a given 143Nd/144Nd suggests that, like depleted Icelandic lavas [Fitton et al., 2003, and references therein], these Emperor Seamount lavas sampled a depleted component that is geochemically distinct from the depleted MORB source. At the low 176Hf/177Hf end of the Emperor Seamount array, lavas from Meiji Seamount and Site 1203, especially the alkalic lavas, trend to lower 176Hf/177Hf at a given 143Nd/144Nd than the trend of Hawaiian shield lavas. Finally, compared to the field for rejuvenated-stage lavas, Emperor Seamount lavas, except those from Site 884 at Detroit Seamount, are offset to lower 176Hf/177Hf at similar 143Nd/144Nd; this offset is not a result of age corrections (Figures 5c and 5d).
Blichert-Toft et al.  showed that Hawaiian shield lavas define a hyperbolic trend in plots of 176Hf/177Hf versus 206Pb/204Pb, e.g., the trend defined by fields for Koolau, Kahoolawe, Mauna Loa and Mauna Kea in Figure 6a. Because the nearly uniform 176Hf/177Hf at the high 206Pb/204Pb end of this hyperbola is less than that of most MORB (Figure 6a), Blichert-Toft et al.  concluded that the depleted component in Hawaiian shield lavas (e.g., Mauna Kea lavas with relatively low 87Sr/86Sr, high 143Nd/144Nd and high 176Hf/177Hf (Figures 2 and Figures 5a–5d) is not related to MORB. This inference has been questioned by Wang et al. , but Figure 6 of Blichert-Toft et al.  shows that relatively depleted Hawaiian shield lavas, such as Mauna Kea, have lower 176Hf/177Hf than most Pacific MORB. The conclusion that oceanic lithosphere or MORB-related asthenosphere does not contribute to Hawaiian shield lavas has also been reached by Lassiter and Hauri  and Abouchami et al. .
 In a 176Hf/177Hf versus 206Pb/204Pb plot it is evident that rejuvenated-stage lavas are not on the hyperbola defined by shield lavas (Figure 6a); consequently isotopically distinct depleted components have contributed to these two lava groups. Specifically, for shield lavas the high 176Hf/177Hf component has high 206Pb/204Pb whereas rejuvenated lavas with high 176Hf/177Hf have lower 206Pb/204Pb (Figure 6a). In contrast to the positive trend defined by Hawaiian shield lavas, lavas from Detroit Seamount, like MORB, form an elongated negative trend ranging from Site 1203 lavas, which in large part overlap the field for the Mauna Kea shield, to Site 884 lavas which overlap with the age-corrected field for rejuvenated-stage lavas (Figure 6b). This negative trend extends toward the low 206Pb/204Pb end of the Pacific MORB field, which, as in Figures 3b and 3d, is defined by lavas erupted within the Garrett transform fault.
4.2. Meiji Seamount
 Meiji Seamount is the most northerly Emperor Seamount that has been sampled (Figure 1). Drilling on DSDP Leg 19 recovered ∼13m of pillow basalt. The altered whole rock compositions are alkalic [Stewart et al., 1973], but subsequent analysis of a glassy sample by Dalrymple et al.  showed that the samples may be altered tholeiitic basalt. An age of 85 Ma was inferred by Keller et al. , but no reliable radiometric ages are available. The minimum age is 68–70 Ma on the basis of fossils in overlying sediments [Worsley, 1973].
 Meiji samples have 87Sr/86Sr and 143Nd/144Nd similar to basalt from Site 1203 at Detroit Seamount (Figure 2b), but they range to lower 176Hf/177Hf (Figure 5b). Although the Sr-Nd-Hf isotopic data for Meiji samples lie within or close to fields for Detroit Seamount lavas, the age-corrected Pb isotopic data do not (Figures 3a–3e). The anomalously low 206Pb/204Pb of the 4 samples analyzed by Regelous et al.  arises from the high parent/daughter ratios (U/Pb > 1.08; also, U/Th > 1 is anomalously high [see Regelous et al., 2003, Table 4]). The plagioclase phenocrysts in Meiji samples have been variably replaced by potassium feldspar [Dalrymple et al., 1980]. Therefore it is likely that the anomalously high U/Pb of the leached Meiji samples reflect the fluids that converted plagioclase to potassium feldspar. Since this alteration event was relatively young, <38 Ma [Dalrymple et al., 1980; Duncan and Keller, 2004], the use of measured U/Pb to correct to an inferred age of 85 Ma leads to overestimates of the age corrections for 206Pb/204Pb. In contrast, the single Meiji sample studied by Keller et al.  has lower U/Pb (0.5); consequently, it has much higher age corrected 206Pb/204Pb (Figure 3b). Given the difficulty in calculating initial Pb isotope ratios, Meiji samples are not plotted in Figures 4 and 6. Sample 145-2 from Detroit Seamount Site 883 also has anomalously high U/Pb (0.74 [Regelous et al., 2003, Table 4]), and its anomalous location in Figure 3d, i.e., relatively low 206Pb/204Pb and high 208Pb/204Pb, probably reflects an overestimated age correction.
4.3. Garnet Pyroxenite Xenoliths From Salt Lake Crater
 Previous studies have shown that Sr, Nd, Pb and Os isotopic ratios of whole rock garnet pyroxenite xenoliths from Salt Lake Crater, a vent of the rejuvenated-stage Honolulu Volcanics, overlap with the range of ratios in the Honolulu Volcanics [e.g., Okano and Tatsumoto, 1996; Lassiter et al., 2000]. Hence a genetic relationship is inferred. Frey  inferred that these pyroxenites formed as high-pressure cumulate rocks derived from the Honolulu Volcanics, whereas Lassiter et al.  suggested that these pyroxenites formed at a spreading ridge at ∼100 Ma and subsequently were an important source component for the Honolulu Volcanics.
 Our Nd and Hf isotopic data for clinopyroxene and garnet bear on these interpretations (Figure 5a). Firstly, the Hf isotopic data overlap with the range of the Honolulu Volcanics. Secondly, Hf isotopic ratios are similar in the clinopyroxenes and garnets indicating that these phases are equilibrated (Table 2). Nd isotopic data are, however, more complex. Sample 3B15 has equilibrated clinopyroxene and garnet with 143Nd/144Nd also overlapping with the field for the Honolulu Volcanics. However, garnet in the other two samples has lower 143Nd/144Nd than both the coexisting clinopyroxene and the Honolulu Volcanics. Since (Sm/Nd)gt > (Sm/Nd)cpx, the two-point isochron has a negative slope indicating recent mixing with a low 143Nd/144Nd component. In this case the lower 143Nd/144Nd in garnet relative to clinopyroxene is consistent with the higher diffusion coefficient of Nd in garnet relative to clinopyroxene [Van Orman et al., 2002]. Furthermore, in a study of garnet peridotite xenoliths from kimberlites Bedini et al.  find “that Hf is far more resistant to diffusional loss and metasomatism than Nd.”
5. Temporal Evolution of Sr, Nd, and Hf Isotope Ratios Along the Emperor Seamount Chain
Keller et al.  showed that the measured 87Sr/86Sr of tholeiitic basalt decrease from Suiko Seamount (65 Ma) to Detroit Seamount (76–81 Ma). Assuming that age corrections are unimportant, they proposed an increasing proportion of a MORB-related component as the age difference between the seamount and underlying oceanic crust decreased. This interpretation was questioned by Regelous et al.  who noted that given the 25 Myr difference between eruption ages of Suiko lavas and the underlying oceanic crust, “it is therefore unlikely that a ridge could influence the chemistry of hot spot lavas over such a distance.” Regelous et al.  plotted ɛSr and ɛNd versus age along the Hawaiian-Emperor hot spot track. They found decreasing initial ɛSr with increasing age, but there is a reversal in this trend from the ∼81 Ma Site 884 lavas at Detroit Seamount to presumably older lavas from Meiji Seamount (see Figure 7a). There is no corresponding age trend for ɛNd; except for Site 884 lavas, there is considerable overlap in ɛNd for lavas erupted from 43 to 81 Ma [Regelous et al., 2003, Figure 8d]. This result is also apparent in Figure 2b, where, except for Site 884 lavas from Detroit Seamount, lavas from Meiji and Detroit Seamounts define a shallow slope, i.e., significant variation in 87Sr/86Sr but overlap in 143Nd/144Nd. Given the strong correlation between 143Nd/144Nd and 176Hf/177Hf, it is not surprising that there is also no systematic temporal trend for 176Hf/177Hf (Figure 7b).
 The implication is that the processes that led to isotopic variation among lavas from different Emperor Seamounts created more variability in Rb/Sr than in Sm/Nd and Lu/Hf. Given the relatively higher incompatibility, mobility and volatility of Rb relative to Sr, Sm, Nd, Lu and Hf, there are well known examples of decoupling between 87Sr/86Sr and 143Nd/144Nd or 176Hf/177Hf; for example both lower continental crust and HIMU OIB are offset from the field defined by MORB and most OIB to low 87Sr/86Sr for their 143Nd/144Nd (e.g., Halliday et al.  and Hart , respectively). Specifically for Hawaiian lavas, Basu and Faggart  proposed that the offset from the MORB trend to high 87Sr/86Sr (Figure 2a) indicates that an ancient lithospheric component that was subjected to seawater alteration is in the source of Hawaiian magmas. In this case, the proportion of such a component in Hawaiian lavas has increased with decreasing age (Figure 7a).
6. Definition and Origin of Depleted Components in Lavas Associated With the Hawaiian Plume
 Firstly, trends defined by Detroit Seamount lavas in 206Pb/204Pb versus 87Sr/86Sr and 176Hf/177Hf show the following: (1) At one extreme, Site 1203 lavas trend to (Figure 6b) or beyond (Figure 4b) the low 87Sr/86Sr, high 176Hf/177Hf and high 206Pb/204Pb end of the array defined by Hawaiian shield lavas. This end is commonly defined as the Kea end-member [e.g., Mukhopadhyay et al., 2003]. (2) At the other extreme, relative to Hawaiian shield lavas, lavas from Sites 883 and 884 have low 87Sr/86Sr and high 176Hf/177Hf accompanied by relatively low 206Pb/204Pb. Such a depleted component is not expressed in Hawaiian shield lavas. Our discussion focuses on the significance of this component in lavas from Detroit Seamount and the surprisingly similar component expressed in isotopic ratios of rejuvenated-stage/North Arch lavas. The abundance of major and incompatible trace elements in the highly alkalic rejuvenated-stage/North Arch lavas differs substantially from those of the mildly alkalic to tholeiitic lavas from Detroit Seamount [Yang et al., 2003; Huang et al., 2005]. Therefore any similarity in radiogenic isotope ratios (Sr, Nd, Hf and Pb) and their trends in these two suites of lavas is surprising (Figures 2 to 6). In particular, both groups trend to very nonradiogenic Pb isotopic ratios which are lower than those of most recent and ancient Pacific MORB (Figures 3a–3e). This trend could reflect the influence of a component similar to that of the atypical ocean floor lavas erupted in the Garrett transform fault; however, this inference is weakened by the rarity of such MORB (Figure 3e). It is possible, however, that depleted asthenospheric mantle with unradiogenic Pb isotope ratios is accessed only under unusual tectonic conditions; for example, the two-stage melting of a MORB source proposed for lavas erupted within the Garrett transform fault [Wendt et al., 1999], the proximity of the plume and spreading ridge when Detroit Seamount formed, and the second pulse of volcanism forming rejuvenated-stage lavas at Hawaiian volcanoes [Ribe and Christensen, 1999].
 Evidence against the involvement of a depleted component like that expressed in the Garrett transform fault lavas is the steep 176Hf/177Hf versus 143Nd/144Nd trends defined by most Emperor Seamount lavas. This trend extrapolates toward a depleted component unlike that expressed in lavas from the Garrett transform fault (Figure 5b). Site 884 lavas from Detroit Seamount are an exception since they are within the MORB field (Figures 5b, 5c, and 5d). What is the significance of the steep 176Hf/177Hf versus 143Nd/144Nd trend (Figure 5d)? One possibility is that aged mantle lithosphere with relatively high Lu/Hf, reflecting residual garnet, is offset to high 176Hf/177Hf at a given 143Nd/144Nd [see Salters and Zindler, 1995, Figure 3], and in unusual circumstances this material is a magma source. This is the model proposed by Fitton et al.  to explain Hf-Nd isotopic systematics in Icelandic lavas.
 The atypically high Hf isotopic ratios, up to ɛHf = 65, in clinopyroxene from peridotite xenoliths at Salt Lake Crater, a Honolulu Volcanics vent, may be relevant for understanding rejuvenated-stage lavas [Salters and Zindler, 1995; Bizimis et al., 2003a]. Because these clinopyroxenes are relatively enriched in incompatible elements, their high Hf isotopic ratios require processes more complex than aging of residual material. The simplest model proposed by Bizimis et al. [2003a] is mixing of an old, >1 Ga, depleted peridotite that has relatively high 176Hf/177Hf (and 143Nd/144Nd) with small amounts, <5%, of melt with high Nd/Hf ratio, such as lavas forming the Honolulu Volcanics. As shown in Figure 7 of Bizimis et al. [2003a], the mixing lines are highly hyperbolic and result in peridotite with high 176Hf/177Hf at a given 143Nd/144Nd. An alternative and favored model of Bizimis et al. [2003a] is melt-peridotite reaction. Specifically, they propose that a magma, similar to the Honolulu Volcanics lavas reacted with 100 Ma depleted oceanic lithosphere mantle [Bizimis et al., 2003a, Table 3]. Although these models were created to explain anomalously high 176Hf/177Hf in clinopyroxene of peridotite xenoliths at Salt Lake Crater, Bizimis et al. [2003a] suggest that the steep 176Hf/177Hf versus 143Nd/144Nd trend of the Honolulu Volcanics (data of Stille et al.  in Figure 5d) indicates the importance of such reacted melts in forming the Honolulu Volcanics. The validity of this inference is weakened by the absence of Nd-Hf isotope ratio correlation shown by our new data for the Honolulu Volcanics (Figure 5d). Regardless, models involving melts enriched in highly incompatible elements are unlikely explanations for the steep Hf-Nd isotope ratio trends defined by Emperor Seamount lavas with their relatively low abundance of incompatible elements [Huang et al., 2005]. Ancient depleted lithospheric mantle is a more appropriate explanation.
 In summary, we suggest that the trends to unradiogenic Pb isotope ratios and 143Nd/144Nd-176Hf/177Hf systematics in lavas from Detroit Seamount and rejuvenated-stage/North Arch lavas indicate involvement of an aged depleted component that is intrinsic to the Hawaiian plume. Despite the marked compositional differences between these two groups of lavas, their Sr, Nd, Hf and Pb isotopic characteristics are surprisingly similar.
 There are also similarities in abundance ratios of some incompatible elements. Specifically, high Ba/Th, relative to the primitive mantle estimate of 83, is characteristic of all Hawaiian lavas, including rejuvenated-stage lavas [Yang et al., 2003, Figure 10a]. Glasses from Detroit Seamount also have high Ba/Th; in fact among Detroit Seamount samples, Site 884 glasses have the highest Ba/Th, approximately a factor of two greater than the Ba/Th of lavas from the Garrett fracture zone [Huang et al., 2005, Figure 17c].
Fitton et al.  used a Nb/Y versus Zr/Y plot to distinguish a depleted component in the Iceland plume from MORB. The pros and cons of this approach are discussed by Hanan et al.  and Fitton et al. . For Hawaiian shield lavas this approach is less useful because Hawaiian shield lavas plot at the boundary between Icelandic lavas and MORB (Figure 8) [see also Yang et al., 2003, Figure 10b]. However, within the context of a MORB-related or plume-related source for the depleted component it is notable that (1) rejuvenated-stage lavas plot farthest from the MORB field and (2) Site 884 lavas do not overlap with lavas from the Garrett transform fault (Figure 8).
 If a depleted component in the Hawaiian plume was sampled at Detroit Seamount and in the rejuvenated-stage lavas of the Hawaiian Islands, there are important implications. Despite considerable short-term geochemical variability expressed among lavas forming the Hawaiian Islands there is also long-term homogeneity, specifically a depleted component intrinsic to the plume that has been available for 80 Ma. This component has not been sampled randomly, as might be expected of a variably metasomatized source, but is available only under specific conditions that enable melting of a depleted source with relatively low 87Sr/86Sr and 206Pb/204Pb coupled with high 143Nd/144Nd and 176Hf/177Hf.
 What is the mechanism for sampling a depleted component intrinsic to the plume? For Detroit Seamount, Regelous et al. [2003, Figure 11] proposed that an intrinsic depleted refractory component within the Hawaiian plume was accessed as a result of the relatively higher extent of melting achieved when the plume ascended below thin oceanic lithosphere. Additional studies of recently acquired samples from Detroit Seamount, indicating relatively low pressures of melt segregation, are consistent with this hypothesis [Huang et al., 2005].
 For rejuvenated-stage/North Arch lavas, the depleted component has typically been inferred to be depleted oceanic lithosphere [e.g., Yang et al., 2003, Figure 12]. However, numerical models of melting dynamics during plume ascent by Ribe and Christensen  suggest that rejuvenated-stage volcanism results from a secondary melting zone (see their Figures 5a–5d) whereby plume material is initially melted in a primary melting zone above the plume stem followed by a secondary melting zone 300–500 km downstream. They proposed that rejuvenated-stage, but not North Arch, magmas result from this second-stage melting of the plume.
 There are two obvious problems with the hypothesis that rejuvenated-stage lavas arise from a secondary melting zone where melts are derived from the plume-related residue of the initial melting event. Firstly, the very high abundances of incompatible elements in rejuvenated-stage and North Arch lavas suggests a source enriched in these elements relative to primitive mantle [e.g., Clague and Frey, 1982; Frey et al., 2000; Yang et al., 2003]. An incompatible element-rich source with depleted isotopic characteristics could be created by metasomatism during the time gap between the shield and rejuvenated-stage volcanism [e.g., Roden et al., 1984]. Alternatively, derivation of incompatible element-rich magma from a source depleted in incompatible elements is only possible if extents of melting are very low. For example, the most SiO2-undersaturated, nephelinite and nepheline melilitite, lavas of the Honolulu Volcanics are enriched in incompatible elements relative to primitive mantle by factors of 100 to 200 [Yang et al., 2003, Figure 3]. Extents of melting therefore must be <1% and even lower if the residue from the initial melting event had lower abundances of incompatible elements than primitive mantle. The alkalic posterosional (rejuvenated-stage) lavas associated with the Madeira hot spot also have depleted isotopic ratios compared to shield lavas and relative enrichment in highly incompatible elements. Geldmacher and Hoernle  proposed that these posterosional lavas were derived by low degrees of melting of a depleted recycled peridotite intrinsic to the Madeira plume.
 Secondly, Lassiter et al.  found that rejuvenated-stage lavas have higher 187Os/188Os than Hawaiian shield lavas, and most samples of mantle peridotite. A source with relatively high Re/Os is required and a likely explanation is pyroxenite veins in the source of rejuvenated-stage lavas [Lassiter et al., 2000; see also Yang et al., 2003, Figure 13]. Garnet pyroxenite xenoliths from Salt Lake Crater have variable and relatively high 187Os/188Os that overlap with the range for the Honolulu Volcanics [Lassiter et al., 2000]. We also find that Hf isotope ratios of garnet and clinopyroxene from these xenoliths overlap with the range for the Honolulu Volcanics. Moreover, in another study of garnet pyroxenite xenoliths from Salt Lake Crater, Bizimis et al. [2003b] also found that Sr, Nd and Hf isotope ratios in garnet and clinopyroxene overlap with those of the Honolulu Volcanics. Clearly, the Honolulu Volcanics and garnet pyroxenites are genetically related. Frey  proposed that the garnet pyroxenites formed as high-pressure cumulates in rejuvenated-stage magmas. In contrast, Lassiter et al.  proposed that these xenoliths formed from MORB at ∼100 Ma. On the basis of the similar 176Hf/177Hf of garnet and clinopyroxene in these xenoliths (Table 2) [see also Bizimis et al., 2003b], this hypothesis requires that garnet and clinopyroxene were at sufficiently high temperature to maintain Hf isotopic equilibrium.
 Since pyroxenites typically have lower solidi than peridotites [e.g., Hirschmann and Stolper, 1996], it is unlikely that pyroxenites could remain in the residue of the initial melting event. However, pyroxenites within the mantle are likely to have diverse mineral proportions and a wide range of compositions [Hirschmann and Stolper, 1996]. As a result their solidus temperatures are variable. In fact some pyroxenites, such as the silica-deficient Salt Lake Crater garnet pyroxenites, have solidi close to that of mantle peridotite [Kogiso et al., 2003]. Another possible alternative is that some peridotites have high 187Os/188Os [Stracke et al., 1999, 2003], but a recent assessment of Os isotopic ratios in mantle rocks [Chesley et al., 2004] notes that mantle peridotites do not have 187Os/188Os > 0.15 whereas rejuvenated lavas and pyroxenites commonly reach values of 0.16 [Lassiter et al., 2000].
 Isotopic ratios of Sr, Nd, Hf and Pb indicate that 76 to 81 Ma lavas forming Detroit Seamount contain a depleted component which is unlike that expressed in young Hawaiian shields, such as Mauna Kea. Distinguishing characteristics of most Detroit lavas are relatively unradiogenic Pb isotope ratios, a negative 87Sr/86Sr versus 206Pb/204Pb trend, a steep positive 143Nd/144Nd versus 176Hf/177Hf trend, and a negative 176Hf/177Hf versus 206Pb/204Pb trend (Figures 3a–3e–6). A MORB-related component is a suitable end-member for some of these trends but most young and ancient Pacific MORB do not have sufficiently unradiogenic Pb isotope ratios. Lavas erupted within the Garrett transform fault on the EPR have suitably unradiogenic Pb isotope ratios, but they do not lie on the trend of 143Nd/144Nd versus 176Hf/177Hf defined by most lavas from Detroit Seamount and Emperor Seamounts in general. Also lavas from the Garrett transform fault differ from Hawaiian lavas in diagnostic incompatible element abundance ratios, such as Ba/Th and Nb/Y at a given Zr/Y.
 Although the alkalic rejuvenated-stage and North Arch lavas erupted in the Hawaiian Islands are substantially different from Detroit Seamount lavas in major element composition and abundance of incompatible elements, these lava suites have surprisingly similar Sr, Nd, Hf and Pb isotopic ratios, and they define similar trends among these ratios. A possible explanation is that these young and old lavas related to the Hawaiian hot spot sampled a depleted component intrinsic to a long-lived source. This component can be accessed only under special conditions, such as (1) the second-stage melting for rejuvenated-stage lavas proposed for Hawaii by Ribe and Christensen  and for the Madeira hot spot track in the eastern North Atlantic by Geldmacher and Hoernle [2000, Figure 9] or (2) under thin lithosphere resulting from the proximity of the Hawaiian plume to a spreading ridge axis [Regelous et al., 2003]. If this inference is correct, the magma source has contained such a depleted component for ∼80 Myr. Although the geochemical heterogeneity of the Hawaiian plume is well established, a long-term role for a depleted component argues against models invoking sources involving random incorporation of geochemical heterogeneities within the shallow mantle. A long-lived geochemically heterogeneous plume with well-defined and spatially coherent geochemical heterogeneities is inferred [see also Abouchami et al., 2004].
 Supporting evidence for a long-lived plume source also arises from high 3He/4He and Ba/Th ratios. Relatively high 3He/4He are typical of Hawaiian shield lavas and Keller et al.  report R/RA (3He/4He in lava relative to atmospheric ratio) >10 for lavas at Site 1203 on Detroit Seamount. All lavas related to the Hawaiian hot spot, including tholeiitic basalt forming the young shields, the Cretaceous lavas forming Detroit Seamount and alkalic lavas erupted during rejuvenated-stage volcanism and forming the North Arch volcanic field, share a common feature of relatively high Ba/Th [e.g., Yang et al., 2003; Huang et al., 2005]. High Ba/Th distinguishes lavas derived from the Hawaiian hot spot from MORB and most other OIB [Hofmann and Jochum, 1996].
 We thank the editors, R. Duncan and W. White, and reviewers, G. Fitton and K. Hoernle, for their constructive review comments. We also thank J. Lassiter, P. Reiners, and D. Weis for discussion. The garnet pyroxenite xenoliths were provided by D. Velde. This research used samples provided by the Ocean Drilling Program (ODP). The ODP is sponsored by the U.S. National Science Foundation (NSF) and participating countries under the management of Joint Oceanographic Institutions, Inc. Funding for this research was provided by the U.S. Science Support Program and NSF grant EAR-0105557. J.B.T. acknowledges financial support from the French Institut National des Sciences de l'Univers.