We report major and trace element and Sr and Nd isotope data for a 283 m thick sequence of postshield lavas drilled during Ocean Drilling Project Leg 197 from Site 1205, Nintoku Seamount, an approximately 56 m.y. old volcano from the Emperor Seamount Chain. At least 25 subaerially erupted lava flow units were sampled, of which all but one are alkali basalts. Clasts of hawaiite and mugearite are present in a conglomerate overlying the basement sequence. Similar rock types characterize the postshield stage of many young Hawaiian volcanoes. Major and trace element, and age-corrected Sr and Nd isotope compositions of the Nintoku lavas are similar to those of young postshield lavas from the Hawaiian Islands. Concentrations of highly incompatible elements tend to increase, La/Yb and Nb/Zr ratios increase, and 87Sr/86Sr decreases with decreasing depth in the drill core. The more evolved rock types can be related to the underlying alkali basalts by fractional crystallization of olivine, clinopyroxene, plagioclase, and Fe-Ti oxides, all of which occur as phenocryst or significant groundmass phases in these lavas. However, variations in the trace element and isotopic compositions of the Nintoku lavas indicate derivation from a heterogeneous mantle source. The isotopic and trace element compositions of the lavas can be modeled using a modified version of the Chen and Frey (1985) mixing model for Hawaiian lavas, in which small-degree melts of the ∼60 m.y. old oceanic lithosphere beneath Nintoku, previously fertilized by small-degree melts soon after its formation at a ridge axis, are mixed with melts from the Hawaiian plume. The contribution from the oceanic lithosphere became more important with time as Nintoku Seamount moved away from the Hawaiian plume axis. Together with other data from Leg 197, our results show that the 87Sr/86Sr ratios of not only shield but also postshield lavas from the Emperor Seamounts increased with decreasing age between ∼80 and 50 m.y. ago. For alkalic postshield lavas, this could result from either increased melting of the oceanic lithosphere beneath volcanoes situated on younger oceanic crust or lower degrees of incompatible trace element enrichment in the lowermost portions of younger oceanic lithosphere.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The intraplate Hawaiian Ridge-Emperor Seamount Chain is an approximately 6,000 km long, age-progressive chain of volcanic islands and seamounts that has been created as the Pacific plate moved over the Hawaiian hot spot (Figure 1). Active volcanism occurs on the Hawaiian Islands at the eastern end of the Hawaiian Ridge and the volcanoes become progressively older to the west and north. The oldest preserved volcanoes, at the northern end of the Emperor Seamount Chain (ESC) at Meiji and Detroit Seamount, have an age of approximately 76–80 Ma [Keller et al., 1995; Duncan and Keller, 2004]. Lavas from the ESC thus preserve a record of Hawaiian hot spot geochemistry since the Late Cretaceous. The Emperor Seamount Chain is built on Pacific lithosphere ranging in age from about 80 to less than 10 million years older than the central volcanoes [Mammerickx and Sharman, 1988], providing a opportunity to examine the role of the thickness and fertility of the lithosphere on the chemistry of intraplate magmatism. Hawaiian postshield lavas are believed to be derived in part from the oceanic lithosphere [e.g., Chen and Frey, 1983; Reiners and Nelson, 1998; Yang et al., 2003].
 Recent studies have shown that the geochemistry of lavas from the ESC vary dramatically from one end to the other. Tholeiitic and alkalic lavas from Meiji and Detroit Seamounts are depleted in highly incompatible trace elements compared to their younger counterparts from the Hawaiian Islands [Keller et al., 2000; Regelous et al., 2003]. Initial 87Sr/86Sr ratios of tholeiitic and alkalic lavas from Detroit and Meiji Seamounts (5580 and 5800 km from Kilauea, respectively) extend to lower values than lavas from younger Hawaiian-Emperor Seamounts [Keller et al., 2000; Regelous et al., 2003; Huang et al., 2005]. The depleted compositions of the oldest Emperor lavas could reflect the fact that these seamounts were formed on young, thin lithosphere close to a former spreading center [Mammerickx and Sharman, 1988]. For example, the composition of Detroit Seamount lavas may be a consequence of magma mixing between MORB and OIB melts [Kinman and Neal, 2002], increased entrainment of depleted upper mantle into the Hawaiian plume, or partial melting of depleted upper mantle close to the ridge [Keller et al., 2000], or be the result of higher average degrees of melting beneath thin lithosphere resulting in an increased contribution of relatively refractory, incompatible trace element depleted mantle components [Regelous et al., 2003; Frey et al., 2005; Ito and Mahoney, 2005a, 2005b].
 The subaerial Hawaiian Islands afford a detailed look at the evolution of volcanoes produced by the Hawaiian hot spot. Most young volcanoes pass through a series of evolutionary stages. Preshield volcanism is small in volume and predominantly alkalic in composition and is followed by voluminous amounts of tholeiitic basalt that erupt to form the bulk of the shield. After a short period of quiescence, postshield alkali basalts and their differentiates are erupted and form a thin discontinuous cap (<1% of total volcano volume) over the tholeiitic base [Powers, 1955; Macdonald and Katsura, 1964; Engel et al., 1965; Clague and Beeson, 1980]. In some cases, postshield alkali basalts have been found intercalated with tholeiitic basalts. Occasionally, renewed volcanism at isolated vents may occur up to 2.5 m.y. after eruption of the postshield basalts. These so-called posterosional lavas are strongly alkalic and comprise <0.1% of the total volcano volume. Not all stages are developed on all volcanoes, for example, Koolau and Lanai volcanoes lack a postshield alkalic stage. Compared to the associated shield tholeiites, the younger alkalic lavas have higher concentrations of incompatible trace elements, higher ratios of more to less incompatible elements (e.g., Nb/Zr, La/Yb), and extend to silica undersaturated compositions.
 In this study, we report major and trace element and Sr and Nd isotope data for 33 samples of volcanic rock from a 283 m thick sequence of alkalic postshield lavas drilled from Nintoku Seamount during Ocean Drilling Program (ODP) Leg 197, Site 1205. The relatively thick, stratigraphically controlled sequence of postshield lavas at Nintoku provides an opportunity to examine the evolution of the Hawaiian plume through geochemical variability of ∼56 Ma postshield lavas, and compare them to present-day postshield basalt compositions.
2. Nintoku Seamount
 Nintoku Seamount is located at 41°N, 175°E in the central part of the ESC, and formed at approximately 56 Ma [Dalrymple et al., 1980; Duncan and Keller, 2004]. The upper 32 m of the volcanic basement of this seamount was previously drilled during Deep Sea Drilling Project (DSDP) Leg 55, Site 432, where three flow units of alkali basalt were recovered [Kirkpatrick et al., 1980]. Site 1205 of ODP Leg 197 is located at the north-west edge of the summit region of Nintoku Seamount in 1310 m water depth, about 100 m south-west of Site 432. Hole 1205A encountered basement at 42.7 meters below seafloor (mbsf) and penetrated 283.3 m into a sequence of basaltic lavas and interbedded sediments and soil horizons (Figure 2). The minimum age of the basement rocks at this site is constrained to be 53.6–54.7 Ma on the basis of the nanofossil assemblage (zone NP10, Early Eocene) recovered in Core 197-1205A-5R immediately overlying volcanic basement. Dalrymple et al.  obtained an 40Ar-39Ar age of 56.2 ± 0.6 Ma for a sample of olivine and plagioclase phyric alkali basalt from DSDP Site 432. 40Ar-39Ar plateau ages from 6 whole rock samples and one feldspar separate from Site 1205 gave a mean age of 55.51 ± 0.22 Ma [Duncan and Keller, 2004], with no significant difference in age between samples from the top and bottom of the section.
 At least 25 lava flow units were recovered at Site 1205. These included both a'a and pahoehoe flow types, which were commonly vesicular and have weathered flow top and inter-flow soil horizons indicating that the sequence was erupted subaerially and that eruption rates were relatively low [Tarduno et al., 2002].
 The lavas drilled at Site 1205 vary from aphyric to plagioclase-phyric to plagioclase-olivine-phyric basalts with texture varying from subophitic to subtrachytic [Tarduno et al., 2002]. The groundmass is typically composed of plagioclase, olivine, clinopyroxene, titanomagnetite, and glass. Preliminary shipboard major element analyses of the lava flow units showed that most are composed of alkali basalt. A conglomerate containing clasts of alkali basalt, hawaiite, and mugearite overlies the basement at Sites 1205 and 432. Alkali basalts and hawaiites with similar chemistry and mineralogy characterize the postshield alkalic stage of many young Hawaiian volcanoes, including Mauna Kea [Frey et al., 1990], Kohala [Lanphere and Frey, 1987], and Haleakala [Chen et al., 1990].
3. Samples and Analytical Methods
 A total of 33 samples were analyzed (Table 1). Of these, six samples are clasts of volcanic rock that were extracted from the conglomerate unit directly overlying the volcanic basement. Drillcore samples were crushed to 0.5–1.0 cm sized chips and the freshest material was handpicked. Vesicle fillings, veins, and highly altered material were avoided. The chips were rinsed in distilled water, dried, and powdered in an agate swing mill at the Max Planck Institut fur Chemie, Mainz.
Table 1. Summary of Unit Number, Recovery Depth, and Rock Type of Samples Recovered From Hole 1205A
moderately to highly plagioclase-phyric basalt
moderately to highly plagioclase-phyric basalt
sparsely to moderately olivine-plagioclase-phyric basalt
moderately plagioclase-olivine-phyric basalt
moderately plagioclase-olivine-phyric basalt
moderately plagioclase-olivine-phyric basalt
plagioclase-olivine basalt autoclastic breccia
sparsely plagioclase-olivine-phyric to aphyric basalt
sparsely plagioclase-olivine-phyric to aphyric basalt
sparsely to moderately plagioclase-olivine-phyric basalt
sparsely to moderately plagioclase-olivine-phyric basalt
sparsely to moderately plagioclase-olivine-phyric basalt
sparsely to moderately plagioclase-olivine-phyric basalt
aphyric to moderately olivine-phyric basalt
moderately to highly olivine-phyric basalt
sparsely to moderately olivine-phyric basalt
moderately to highly plagioclase-olivine-phyric basalt
moderately olivine-plagioclase-phyric basalt
aphyric basalt breccia
aphyric to sparsely plagioclase-phyric basalt
aphyric to sparsely plagioclase-phyric basalt
olivine-plagioclase basalt breccia
olivine basalt breccia
aphyric to moderately olivine-plagioclase-phyric basalt
 Loss on ignition (LOI) was determined by heating between 0.1 and 0.5 g of sample powder in platinum crucibles at 1025°C for 4 hours. The samples were weighed before and after removal from the oven and total weight percent LOI was determined.
 Major element concentrations were determined by inductively coupled plasma-optical emission spectroscopy (ICP-OES) at the University of Notre Dame. Approximately 0.1 g of sample powder was fused with 0.5 g of lithium metaborate in carbon crucibles at 1025°C for 30 minutes in a Fisher Scientific Isotemp programmable muffle furnace. Immediately upon removal from the oven, the molten pellets were quenched in 5% HNO3, which was then transferred to polypropylene bottles and brought to 100 g with 5% HNO3. This solution was then placed in an ultrasonic bath until the fused glass was completely dissolved. This solution was then used for ICP-OES analyses. Machine drift was monitored by analyzing a standard solution every four samples (i.e., every four test tubes). Calibration was performed by taking the blank- and drift-corrected counts of a selected standard reference material, or SRM (two or three different SRMs were replicated at least twice during each run including BHVO-1, BHVO-2, BIR-1, and BPL-1), and dividing by the accepted elemental concentrations [Shafer et al., 2004]. The background signal (the raw counts per second or cps) for the analytical blanks were generally low (<0.8% of standard reference material cps, except for P2O5, which ranged from 11–48% of the reference material cps).
 Trace element abundances of the agate-milled powders were quantified by ICP-mass spectrometry (ICP-MS) using a PlasmaQuad II quadrupole machine at the University of Notre Dame [Neal, 2001]. Blank levels were around 100 cps for the REE and the standard deviation for replicated SRMs was generally less than 10% (REE < 5%, except for Yb and Er). In all ICP analyses, dissolution was performed on unignited powders.
 Strontium and Nd isotope analyses were carried out at Bristol University. Before dissolution, sample powders designated for Sr isotope analysis were subjected to a sequential leaching procedure in order to remove alteration products. About 4 g of rock powder was weighed into a Teflon beaker to which 5–10 ml of distilled 6M HCl was added. The samples were placed in an ultrasonic bath for 2 hours and the acid was changed every 30 minutes. The samples were then leached in hot (80°C) 6M HCl for a further 5 hours with the acid being changed every 60 minutes. The residue was soaked in deionized water at 80°C for 30 minutes, rinsed twice, and dried. This leaching procedure has been shown to yield reliable initial Sr isotope data for old, altered Hawaiian lavas [see Regelous et al., 2003; Huang et al., 2005].
 For Sr isotope analysis, about 0.1 g of leached rock powder was spiked with 87Rb and 84Sr and digested in HF-HNO3. After treating with 15M HNO3 and 6M HCl, the sample was dissolved completely in dilute HCl. Rubidium and Sr were separated from the rock matrix on cation exchange columns in 2.5M HCl. The Rb fraction was evaporated to dryness, redissolved in 3M HNO3, and cleaned of Sr by passing it through a second column containing 50 μl of Sr Spec resin.
 Neodymium isotope analyses were conducted on unleached sample powders. Approximately 0.1 g of rock powder was digested using HF-HNO3-HCl, dissolved in 3M HNO3, and passed through a column containing 0.1 ml of exchange resin. After washing with 3M HNO3, the rare earth elements (REE) were recovered in 2.5M HCl. Neodymium was separated from the other REE using HDEHP columns and dilute HCl. All reagents used were quartz distilled and total procedural blanks for Sr and Nd were below 400 pg and 50 pg, respectively.
 Strontium and Nd isotope measurements were conducted using a Finnigan “Triton” thermal ionization mass spectrometer in static mode. These measurements were corrected for instrumental mass fractionation using exponential fractionation corrections and assuming 86Sr/88Sr = 0.1194 and 146Nd/144Nd = 0.7219. The NBS-987 Sr and J&M Nd standards gave 0.710247 ± 8 (n = 12) and 0.511114 ± 5 (n = 8), respectively, over the period of the analysis.
 Rubidium isotope compositions were determined by multicollector ICP-MS using methods adapted from Waight et al. . The Rb fraction was dissolved in 0.6M HCl + 0.02M HF in order to obtain a Rb concentration of ∼100 ppb. About 2 ml of this solution was transferred to a small vial and approximately 250 ng of natural Zr was added. Sample solutions were introduced into a Finnigan “Neptune” double focusing MC-ICP-MS via a Cetac Aridus microconcentric nebulizer and desolvating system equipped with a PFA spray chamber, at an uptake rate of 50 μl/minute. Typical sensitivity for Rb was ∼200–400 Volts/ppm. Masses 85, 87, 88, 90, and 91 were measured simultaneously on Faraday cups. The contribution of Sr to mass 87 was monitored via mass 88 and in all cases the correction applied (using the 88Sr/87Sr ratio obtained for individual samples by TIMS) was negligible. An individual analysis typically consisted of 50 integrations of 8 seconds. 0.6M HNO3 followed by a 0.6M HCl + 0.02M HF was used to flush the system between analyses. The measured 85Rb/87Rb ratio was corrected for instrumental mass fractionation (0.3–0.4%/amu) using an exponential correction and assuming 90Zr/91Zr = 4.5882. On the basis of repeat analysis of rock standards, the reproducibility of Rb concentration measurements is better than 0.6%.
 We divide the Site 1205 Nintoku Seamount lavas into an upper and lower alkalic series, which are separated by Unit 19B (tholeiitic basalt).
4.1. Major Elements
 Whole rock major element concentrations and LOI values of 33 samples of volcanic rock from Nintoku Seamount are presented in Table 2. Major element totals (total + LOI) vary from 97.05–105.38, although much of the variation is caused by 5 samples (Table 2). The major element totals of the other 28 samples are 100 ± 2 wt.%. Sample 35R-3, 108–114 from Unit 19b, which contains accumulated olivine, is the only sample that plots in the tholeiitic basalt field (Figure 3). Although alteration may increase Na2O and K2O contents (see Discussion), the high abundances of the more immobile elements such as TiO2 and P2O5 and incompatible trace element concentrations (e.g., Nb, Zr) of the Nintoku lavas compared to Hawaiian tholeiites confirm that the former are indeed alkalic. The most evolved rock types are the conglomerate clasts from core 5R, which were not present as flow units deeper in the recovered basement section at Site 1205. The significant variation in composition of the lava clasts from the conglomerate indicate that they were derived from several discrete flow units.
Table 2. Major Element Abundances for Leg 197, Site 1205 Basaltsa
 MgO contents vary from 1.74 to 14.9 wt.%. As MgO abundances decrease, CaO, MnO, and FeOT (total Fe as FeO) concentrations decrease whereas P2O5, K2O, Na2O, SiO2, Al2O3, and TiO2 increase (Figures 4a–4h). The hawaiite/mugearite clasts from the conglomerate have higher TiO2 and P2O5 contents than their younger counterparts from Mauna Kea, Haleakala, and Kohala volcanoes.
4.2. Trace Elements
 Trace element contents of the Nintoku Seamount lavas from Site 1205 are presented in Table 3. Nickel, Cr, Sc, and V generally increase down the section. One sample from Unit 19b, which contains accumulated olivine, has the highest Ni content (1135 ppm) while the hawaiites and mugearites contain the lowest concentrations of compatible trace elements. The Ni contents of the alkali basalts (<234 ppm) are too low for these lavas to represent primary melts of mantle peridotite, indicating that they have undergone fractional crystallization within the crust. The concentration of the highly incompatible elements that are immobile during alteration (e.g., Zr, Nb) vary by a factor of ∼13, with the highest concentrations in the most evolved lavas (up to 91 times primitive mantle abundances). Chondrite-normalized REE patterns of the Nintoku lavas are similar to those of similar rock types from the Hawaiian Islands (Figure 5). Hawaiites and mugearites have higher concentrations of incompatible trace elements and steeper REE patterns (La/Yb varies from 12.6–17.0) compared to the alkali basalts (La/Yb < 10.0). Ba/Th ratios (113–162) of the Nintoku lavas are high compared to most other intraplate oceanic lavas, but are similar to other Hawaiian lavas [Yang et al., 2003].
Table 3. Trace Element Abundances of Leg 197, Site 1205 Basaltsa
 Age-corrected 87Sr/86Sr ratios for the leached sample powders range from 0.703130–0.703245 and initial 143Nd/144Nd ratios range from 0.512927–0.512990 (Table 4). Epsilon-Nd and ɛSr values of the Site 1205 lavas are within the range of postshield lavas from Mauna Kea, Haleakala, and Kohala volcanoes, although ɛSr is at the lower end of this range (Figure 6). One sample of alkali basalt from Site 432 analyzed by Regelous et al.  has lower ɛSr than any postshield lavas from the Hawaiian Islands. There is no systematic variation of Sr or Nd isotope composition with degree of differentiation (Figure 7). The older lavas from deeper in the borehole tend to have the highest ɛSr values (Figure 8f), as discussed in the next section. ɛSr correlates negatively with ratios of more to less incompatible trace elements, such as La/Yb (Figure 9).
Table 4. Sr and Nd Isotope Data for Leg 197, Site 1205 Basalt
5.1. Chemical and Isotopic Variations With Depth (Stratigraphic Age) in the Site 1205 Drill Core
 There are significant variations in lava composition that correlate with stratigraphic position in the Site 1205 drill core. With decreasing depth (decreasing age), incompatible element concentrations and ratios of more- to less-incompatible elements (e.g., La/Yb, Nb/Zr) generally increase, and initial 87Sr/86Sr decreases (Figure 8). Epsilon-Nd shows no consistent correlation with depth. Similar compositional variations with age have been reported for the Kula postshield lavas of Haleakala volcano [Chen et al., 1990], although the variation in Sr and Nd isotope ratios of the Nintoku lavas is less than half that observed in the Kula volcanics. Samples with the lowest 87Sr/86Sr values occur in the upper part of the Site 1205 drill core tend to have higher incompatible trace element contents, and have higher La/Yb and Nb/Zr ratios (Figure 9).
 As on Mauna Kea, and in contrast to Haleakala, alkali basalts and their differentiates (hawaiites and mugearites) are not intercalated in the postshield stage of Nintoku Seamount, but instead the more evolved rocks appear to have formed a discontinuous cap on the alkali basalts. Although the more evolved rock types are present only as clasts within the conglomerate unit immediately above basement at Site 1205, the predominance of hawaiite and mugearite compositions suggest that the stratigraphically highest lava units in this area were dominated by these rock types. The origin and petrogenetic significance of the temporal variations in lava composition are discussed further in the following sections.
5.2. Influence of Alteration on Chemistry
 Many of the Site 1205 lava units have undergone subaerial weathering immediately after eruption and are highly altered. Flow tops are variably weathered and oxidized, and soil horizons separate many of the flows [Tarduno et al., 2002] (Figure 2). The lava flow units subsequently underwent seawater alteration after subsidence of the summit of Nintoku Seamount below sea level. Secondary alteration minerals are abundant. For example, olivine is often completely replaced by iddingsite and calcite; glass in the groundmass and in flow margins is often replaced by clay minerals. Vesicles are partially to completely filled with clay minerals, Fe-oxyhydroxides or zeolites, and calcite veins occur throughout the section. Loss on ignition (LOI) values range from 1.3 to 7.3%, reflecting the presence of secondary carbonate and water-bearing clay minerals in most of the samples, despite great care taken to sample only the freshest material. Given the abundance of secondary minerals, alteration is likely to have affected the concentrations of K, Na, and to a lesser extent, Ca, Mg, and P in the most altered samples. K2O/P2O5 ratios of the Site 1205 lavas are more variable, and generally lower, than fresh postshield lavas from Haleakala (Figure 10), probably reflecting preferential loss of K during alteration [Frey et al., 1994].
 Few of the samples analyzed have Ce/Pb, Rb/Ba, or Nb/U within range of fresh oceanic basalts (Figure 11), indicating that Pb, U, Rb, and possibly Ba were mobile during alteration. The relatively high Ce/Pb and Nb/U ratios suggest that most samples have lost U and Pb (the highest measured Nb/U ratios indicate 65% of original U content removed). Therefore, noting the influence of secondary processes on the primary chemistry of the Site 1205 lavas, we use those elements that are immobile during alteration (REE and HFSE) to investigate the petrogenesis of the Nintoku basalts [Ludden and Thompson, 1979; Bienvenu et al., 1990].
5.3. Role of Fractional Crystallization, Crystal Accumulation, and Crustal Contamination
 Hawaiites and mugearites are considered to be the products of fractional crystallization of clinopyroxene-rich cumulates from alkalic magmas at depths of ∼15 km [West et al., 1988; Frey et al., 1990; Chen et al., 1991]. The major element characteristics of the hawaiite and mugearite samples from Nintoku Seamount can be modeled by using methods adapted from Albarede . The hawaiite/mugearite samples define a fairly linear extension in major element versus MgO variation diagrams from the alkalic basalts from Core 6R-2 (the youngest igneous basement recovered), suggesting that the evolved samples formed by fractional crystallization of a parental melt with major element characteristics similar to Core 6R-2 alkalic basalts (Figure 4). After applying a least squares fit to the major element versus MgO content of the hawaiite/mugearite and Core 6R-2 alkalic basalts, an estimate of the general parental melt was made. Using the concentrations of highly incompatible elements Nb, Th, and La from sample 5R-2, 90–94, an alkali basalt clast contained in the conglomerate with the hawaiite/mugearite samples that has major element abundances similar to the parental melt estimate, an estimate of amount of fractional crystallization was made assuming that the bulk partition coefficient, D, for these elements is approximately 0. If this is the case, then the standard Rayleigh fractional crystallization equation reduces to CL = CO/F, where CO is the trace element concentration of this alkali basalt clast and CL is the average trace element concentration in the hawaiite/mugearite samples. Once an estimate of F is made, the major element composition of the cumulate that fractionated from the hawaiite/mugearite parental melt can also be estimated using simple mass balance approaches. Using a constrained linear least squares method presented by Albarede , the cumulate composition was recast into proportions of mineral phases present as phenocrysts in the hawaiite/mugearite lavas. The Site 1205 lavas commonly contain olivine, plagioclase, clinopyroxene, and titanomagnetite as phenocryst phases. The theoretical cumulate composition can be represented as the vector y contained within the matrix A, which has 5 columns representing the cumulate and mineral (olivine, plagioclase, clinopyroxene, and Fe-Ti oxide) compositions and 11 rows for the 11 major elements modeled. The compositions of the mineral phases were taken from electron microprobe data collected on the phenocryst phases from the Nintoku hawaiite/mugearite samples during DSDP Leg 55 [Clague et al., 1980]. Manipulation of matrix A leads to the unconstrained solution, the vector x. The constrained solution (the proportions of mineral phases needed to reproduce the cumulate composition) can be calculated after recalculating the unconstrained solution to force the mineral phase proportions to sum to unity. The fractional crystallization (37–58%) of a crystal assemblage composed of plagioclase, clinopyroxene, olivine, and Fe-Ti oxides in the proportions 46:30:14:10 can reproduce the trends seen in the hawaiite/mugearite samples, except for MnO (Figure 12). Model parameters are presented in Table 5.
The “Initial” and “Final” compositions refer to the end-point compositions of the least squares fit of the Nintoku data presented in the MgO-variation diagrams. The “Cumulate” is the composition of the solid that needs to be removed to account for Nintoku major element variations assuming the fraction of liquid remaining at the Final compositions is 0.45 (see text for justification). This composition was recast into proportions of the four mineral phases listed above, which were then used to recalculate the cumulate composition, given as Cum* (see text for full explanation).
 This model predicts significant plagioclase, clinopyroxene, and Fe-Ti oxide fractionation, the effects of which are only partially borne out in the trace element characteristics of the hawaiites/mugearites. Scandium is compatible in clinopyroxene [Hart and Dunn, 1993; Jenner et al., 1993; Green, 1994] and the hawaiite/mugearite lavas have corresponding low Sc concentrations (<15.5 ppm) that is consistent with significant clinopyroxene fractionation. However, Sr and Eu concentrations in the hawaiite/mugearite lavas are not depleted, which is inconsistent with extensive fractionation of plagioclase. Certain characteristics of the Nintoku samples suggest mitigating factors. The high Sr abundances are coupled with high P2O5 abundances, which suggests the presence of minor mineral phases such as apatite may be controlling the Sr budget in these samples. Petrography indicates that the dominant Fe-Ti oxide species is titanomagnetite [Tarduno et al., 2002] which requires relatively high oxygen fugacity conditions to crystallize (e.g., QFM). This would keep Eu in the 3+ oxidation state, thereby limiting the uptake of Eu into the fractionating plagioclase.
 This model can be quantitatively tested by modeling the trace element variations between the hawaiite/mugearite and the alkali basalt clasts using the mineral proportions and percent fractional crystallization derived from the major element model. Figure 13 shows that the majority of incompatible trace elements can be adequately predicted using calculated mineral proportions, except for Sr, possible reasons for which were discussed above. Modeled concentrations for all elements except Sr, Y, Yb are within 25% of measured concentrations for samples 5R-2, 21–25 and 5R-2, 96–100. For sample 5R-2, 31–34 only Sr, Y, Er, and Yb are outside this range. Samples 5R-2, 74–77 and 5R-2, 114–116 show worse agreement; the predicted HREE concentrations are systematically higher in these samples than the actual concentrations, however, increasing clinopyroxene fractionation could lower the model HREE predictions.
5.4. Role of Melting Processes and Source Heterogeneity in Creating Variation Within Nintoku Lavas
 Major and trace element and isotopic compositions indicate that the Nintoku lavas cannot all be related simply by closed-system fractional crystallization from a single parental magma composition. At MgO contents of ∼7 wt.%, there is a considerable range in SiO2, TiO2, FeOT, MnO, CaO, P2O5, and K2O in the alkali basalt samples (Figures 4a–4h). To some extent, these major element variations could reflect variations in the degree and pressure of melting, such that the Nintoku lavas are derived from crystallization of a range of melt compositions. However, the significant variation in Sr isotope composition, and the correlation of 87Sr/86Sr with incompatible trace element ratios (Figure 9) indicates that source heterogeneity was an important factor in creating variation with the Nintoku lavas.
 Mixing of two sources, or melts derived from them, followed by variable amounts of fractional crystallization could account for the trace element and isotopic variation, although the scatter in plots such as Nb/Zr versus 87Sr/86Sr requires both end-members to be heterogeneous (Figure 9). Most importantly, simple two-component mixing does not readily explain the systematic variations in trace element and isotopic composition with time that are observed in the lavas from Site 1205.
5.5. Model for the Formation of the Nintoku Postshield Basalts
 Two characteristics of the Nintoku alkali basalts must be accounted for in any model of their genesis. First, there are systematic variations in incompatible trace element and Sr isotope ratios with age: 87Sr/86Sr ratios decrease and La/Yb, Nb/Zr ratios increase with decreasing age (Figure 8). Second, lavas with the most enriched trace element compositions (highest La/Yb, Nb/Zr) were derived from a source with long-term depletion in highly incompatible elements (low 87Sr/86Sr). Similar trace element-isotopic systematics characterize the transition from shield tholeiitic lavas through postshield alkali lavas to rejuvenated stage alkalic lavas of many Hawaiian volcanoes [e.g., Chen and Frey, 1985], and have also been recognized within the postshield alkalic lavas of Haleakala [Chen et al., 1990].
 Several models have been proposed to account for the combination of trace element enriched lavas (high Rb/Sr, low Sm/Nd) with depleted isotope signatures (low 87Sr/86Sr, high 143Nd/144Nd) from Hawaii [Chen and Frey, 1983, 1985; Frey and Roden, 1987; Reiners and Nelson, 1998; Yang et al., 2003]. Very small (<0.2%) degrees of melting of depleted mantle sources, or reactive melt transport processes are likely to result in decoupling between trace element and isotopic ratios and non-linear variations between concentrations of incompatible elements (such as Nb and Zr), unlike those observed in Hawaiian posterosional stage lavas [e.g., Reiners and Nelson, 1998].
 Instead, many models attempting to explain the origin of Hawaiian alkali lavas invoke small-degree melts of a source that was relatively recently (<400 m.y. ago) enriched in incompatible trace elements. Chen and Frey [1983, 1985] proposed that partial melts of the Hawaiian plume are mixed with 1–4% of very small degree melts (0.1–1.0%) of a depleted source (oceanic lithosphere or the upper mantle source of MORB). This multicomponent mixing model could account for much of the trace element and isotopic variations in lavas from Haleakala. Mixing of approximately constant proportions of the enriched source with variable-degree melts of the depleted source will result in negative correlations in La/Ce or La/Yb versus 87Sr/86Sr diagrams. A similar model, in which small-degree melts of the MORB source mantle are added to the Hawaiian plume source, and this mixed source then melted to various degrees, has been used to explain the compositions of Koloa posterosional lavas from Kauai [Reiners and Nelson, 1998] and the posterosional Honolulu Volcanics of Oahu [Roden et al., 1994].
 An alternative model involves mixing of a depleted source with 0.5–2.0% of small-degree melts (∼2%) from a more enriched source [Clague and Dalyrmple, 1988]. This is similar to metasomatic models in which small-degree partial melts of the enriched source are added to the depleted oceanic lithosphere, which subsequently undergoes partial melting to form the parental magmas of Hawaiian lavas [Wright, 1984]. In this model, the melt with the greatest contribution from the low 87Sr/86Sr component must have been enriched by the smallest-degree melts from the plume in order to obtain a negative correlation between 87Sr/86Sr and La/Ce or Nb/Zr. Yang et al.  favored such a model to explain the compositions of the Honolulu Volcanics because recently determined partition coefficients imply that extremely small degree melts (<0.1%) of the depleted source would be required to explain the high La/Ce ratios of the Hawaiian posterosional lavas when using the Chen and Frey [1983, 1985] model. An alternative possibility, explored more fully below, is that the low 87Sr/86Sr end-member (oceanic lithosphere) is more enriched in trace elements than assumed by Chen and Frey , perhaps by incorporation of small-degree melts of MORB source soon after its formation at the MOR axis. Below, we evaluate which of these models can best explain the trace element and isotope compositions of the Nintoku lavas.
5.5.1. Model 1: The Chen and Frey-Type Model
 Considering the compositional similarities between Nintoku and Haleakala postshield lavas, we first applied a multicomponent mixing model similar to that proposed by Chen and Frey . Distribution coefficients and model parameters are given in Table 6. Figure 14a illustrates that the Nintoku lavas can be generated by mixing of 88.4–89.2% undepleted mantle (UD) and 11.6–10.8% low-degree (0.79–1.93%) partial melts of MORB source. The amount of incipient melt mixed with UD is greater at Nintoku than for the Kula Series postshield basalts from Haleakala (97–98% UD, 2–3% MORB source), and the amount of partial melting of the MORB source required to fit the Nintoku data (0.79–1.93%) is higher than predicted for Haleakala (0.2–0.5%) [Chen and Frey, 1985].
Table 6. Reservoir Compositions and Partition Coefficientsa
All reservoir composition values are in ppm. UD, undepleted mantle, from Chen and Frey ; MS, MORB source, from Chen and Frey ; NM, normal mid-ocean ridge basalt, from Sun and McDonough , except that 87Sr/86Sr was left the same as MORB source; RDS, refertilized depleted source, created by mixing 0.5% incipient melts of MS with NM in 0.5:99.5 proportions (see text for full description); DML, depleted mantle lithosphere, from Yang et al. ; HP, Hawaiian plume, from Yang et al. . Mineral modes for calculation of the bulk partition coefficients for the model compositions above were adapted from those given by Chen and Frey , Halliday et al. , and Yang et al. . The olivine:orthopyroxene:clinopyroxene:garnet:phlogopite:amphibole proportions in the source and melt for Models 1 and 2 are as follows: Model 1, 0.6:0.25:0.1:0.05:0:0 and 0.1:0.1:0.4:0.4; Model 2, 0.55:0.25:0.15:0.05:0:0 and 0.1:0.1:0.4:0.4. Mineral proportions for MS, FDS, and melt for Model 3 are 0.706:0.2:0.05:0.02:0.002:0.02, 0.6:0.25:0.1:0.05, and 0.1:0.1:0.4:0.4, respectively. Ol, olivine; cpx, clinopyroxene; opx, orthopyroxene; gt, garnet; phl, phlogopite; amp, amphibole. Partition coefficients:
 Following Chen and Frey , four samples that have olivine as the only phenocryst phase were selected to quantitatively test the model. Since the model predicts only the hypothetical parental magma composition, it is necessary to correct the trace element concentrations of the selected samples for olivine fractionation in order to estimate the parental melt of each sample. Olivine addition dilutes incompatible trace element concentrations but does not radically alter trace element ratios. Equilibrium olivine was added in 1% increments until the major element composition of each sample was in equilibrium with Fo90 [Chen and Frey, 1985]. The Fe/Mg exchange coefficient between olivine and melt was 0.30 [Ford et al., 1983]. This addition of olivine was then used to back-calculate the trace element abundances of the parental magma assuming that olivine was the only fractionating phase. Using the corrected La abundance, we calculated how much partial melting of the hybrid UD-MORB source was required to achieve the corrected La abundance of each sample. After partially melting the hybrid source by the amount calculated above, the model results were then compared to the olivine-fractionation-corrected trace element abundances. The results are presented in Table 7.
 The agreement of the actual data with model predictions is generally good. In the following discussion of the model results, samples which show discrepancies of greater than or less than 25% of the actual data are considered to be in poor agreement. Predicted Rb concentrations are systematically low (21–39%) while Nb predictions are low for two of the four samples quantitatively studied (by 27% and 29% for samples 10R-2, 0–5 and 33R-2, 70–75, respectively). Thorium predictions are within 7–20% agreement and Sr predictions are within 9%, except for sample 41R-2, 18–23 (20%). Neodymium, Sm, and Hf all agree within 15%, many within 5%, of the actual data. Overall, the model adequately predicts trace element concentrations.
5.5.2. Model 2: The Yang et al.-Type Model
 We next applied the model of Yang et al.  to the Nintoku data. In this model, the procedure is the basically the same as that used by Chen and Frey , except incipient melts of Hawaiian plume source are mixed with depleted mantle lithosphere. The negative correlation between 87Sr/86Sr and La/Ce for the Nintoku data requires those samples created by the largest degree of melting to contain the smallest contribution from the low 87Sr/86Sr source (Figure 14b). Mixing proportions vary from 96.1–98.5% MORB source and 3.9–1.5% incipient melts from the Hawaiian plume (at 1.84–7.15% partial melting). Quantitative testing was performed in the same manner as above. The full results of the model are presented in Table 7. The predicted Sr, Nd, Sm, and Hf concentrations were substantially lower than those measured in the lavas (> 47% lower). Predicted Rb concentrations agreed quite well with actual data, although Rb concentrations are likely to have been affected by alteration. Thorium and Nb also show close approximation to the actual data. However, the poor reproducibility of the REE and Hf forces us to conclude that this model is not applicable to the petrogenesis of the Nintoku lavas.
5.5.3. Model 3: Modified Chen and Frey-Type Model
 The third model is a modified version of that proposed by Chen and Frey . In this model, a more enriched composition for the depleted end-member is used. Undepleted mantle is used as the enriched end-member. There is some evidence, discussed more fully in the next section, that oceanic lithosphere may be enriched by small-degree melts of MORB source soon after its formation at the ridge axis. In this model, oceanic lithosphere (N-MORB) is metasomatized by a 0.5% degree partial melt of MORB source (in 99.5:0.5% proportions, respectively) to create the “fertilized” depleted end-member. Melting of this source varies from 0.98–2.49% (as opposed to 0.79–1.93% in Model 1) and mixing proportions vary from 1.79 to 2.03% fertilized depleted end-member and 98.21 to 97.97% Hawaiian plume source (Figure 14c). The model was quantitatively tested in the same manner as the other two models and the results are presented in Table 7. Predicted trace element concentrations are significantly better than in the previous two models. Predicted Rb concentrations were all within 25% of the actual concentration (all but one sample is within 8%), which is substantially better than Model 1 and similar to the predicted concentrations from Model 2. Niobium was consistently low (> 25% for two of the four samples), which is similar to Model 1. Neodymium, Sm, Hf, and Th all agree within 16% of the actual concentration, many within 5%. The degree of incipient melting of the depleted source is higher than in Model 1 (1.78% partial melting on average versus 1.41% for Model 1) and the proportion of incipient melt mixing with Hawaiian plume source is much less than Model 1 (∼2% incipient melt versus ∼11% incipient melt in Model 1). As a result, smaller volumes of the oceanic lithosphere are required to melt. The amount of partial melting of the hybrid source required is also less than in Model 1 (8.55% versus 19.88% for Models 3 and 1, respectively). The results of the three models are summarized in Figure 15.
 The chemical and isotopic variations with time recorded in the Site 1205 lavas indicate that an increasing proportion of the incompatible trace element enriched, low 87Sr/86Sr end-member contributed to melting over the period of time represented by the Site 1205 drillcore, coupled with a lower degree of melting of this component. These temporal trends are perhaps the result of the drift of Nintoku Seamount away from the hottest part of the Hawaiian plume.
5.6. A Physical Model for Nintoku Postshield Magmatism
Chen and Frey  envisaged that Hawaiian postshield lavas were formed by melts from the Hawaiian plume mixing with incipient melts of oceanic lithosphere, which formed the wallrock of the plume conduit. The relative contribution from the radiogenic 87Sr/86Sr plume source decreased with time as the oceanic plate drifted away from the axis of the plume [Chen and Frey, 1985].
 A role for the ∼90 m.y. old Pacific lithosphere in the generation of Hawaiian alkalic lavas is supported by evidence for residual phlogopite or amphibole in the source of the Honolulu Volcanics and North Arch lavas [Clague and Frey, 1982; Class and Goldstein, 1997; Frey et al., 2000; Yang et al., 2003]. These phases are unlikely to be stable in the convecting upper mantle, but may exist in the lower P-T conditions of the oceanic lithospheric mantle [Class and Goldstein, 1997]. The relatively radiogenic Os isotopic composition of many Hawaiian posterosional lavas is inferred to have been inherited from pyroxenite veins within the ∼90 m.y. old Pacific oceanic lithospheric mantle beneath Hawaii [Lassiter et al., 2000]. Basu and Faggart  noted that compared to MORB, Hawaiian postshield lavas are offset to higher ɛSr for a given ɛNd, and suggested that they contain a seawater-altered oceanic lithosphere component.
 We recognize that the modeling presented above is dependent on the partition coefficients, assumed end-member compositions, and modal mineral proportions of sources and crystallizing melts. These basic assumptions make the model predictions non-unique, however, the modeling does serve to show that fertilization of the oceanic lithosphere is a viable process for generating the postshield basalts of Nintoku seamount. In particular, fertilized oceanic lithosphere removes the need to invoke very small degrees of melting to account for the incompatible trace element enrichment of Hawaiian postshield and posterosional lavas. The fertilization model satisfies constraints imposed by Os isotopic compositions, which imply a significant age for the pyroxenite veins that contribute to postshield magmatism [Lassiter et al., 2000] and complies with evidence for residual phlogopite/amphibole in the source of Hawaiian postshield lavas.
 Recently, Frey et al.  have suggested that the depleted (low 87Sr/86Sr) component in both the oldest Emperor Seamount lavas, and younger Hawaiian post erosional lavas is the same, and that it is intrinsic to the Hawaiian plume and distinct from the source of MORB. This model has some difficulty in explaining the incompatible trace element enriched compositions of Hawaiian post shield alkalic lavas, although it is possible that heat consumed by melting of the high 87Sr/86Sr “veins” inhibits melting of the more refractory matrix until the “veins” are exhausted, at which point they are more depleted in incompatible trace elements than the matrix.
5.7. Constraints From Sr Isotope Variations With Time on Petrogenesis of Hawaiian-Emperor Postshield Lavas
 Previous studies have shown that ɛSr values of tholeiitic lavas decrease northward along the Emperor Seamount Chain (Figure 16a). The origin of the depleted isotopic compositions of the older Emperor tholeiites is debated, but likely reflects the fact that the oldest Emperor Seamounts were formed close to a former spreading center (Figure 16b). The major and trace element compositions of Detroit Seamount lavas, which indicate that they were derived by larger degrees of melting at lower average pressure than young Hawaiian lavas [Huang et al., 2005]. Keller et al.  suggest that the depleted compositions of older Emperor lavas are the result of plume-ridge interaction, such that the Hawaiian plume entrained depleted upper mantle and young, hot lithosphere. Alternatively, the depleted compositions of these lavas may result from melting of a depleted, refractory mantle component that contributes to melting only when the Hawaiian plume is situated beneath young, thin lithosphere where decompression melting extended to shallower depths [Regelous et al., 2003; Frey et al., 2005; Huang et al., 2005]
 Together with other Leg 197 data [Huang et al., 2005], our new results show that ɛSr values of postshield lavas also decrease northward along the ESC (Figure 16a). Alkalic lavas from ODP Sites 883 and 1204 on Detroit Seamount, which are considered to represent postshield lavas [Huang et al., 2005] have lower ɛSr than any postshield or posterosional lavas reported from the Hawaiian Islands. Two samples of alkali basalt from Suiko Seamount have ɛSr values that are intermediate between those of postshield alkalic lavas from the Hawaiian Islands and from Detroit Seamount [Lanphere et al., 1980]. ɛSr values of most Nintoku lavas are within the range of Hawaiian postshield lavas, but lie at the depleted end of this range (Figures 6 and 16).
 The lower ɛSr values of older Emperor postshield lavas could reflect the fact that ɛSr values of the plume-derived end-member was lower when the Hawaiian plume was situated close to a ridge axis, so that plume-lithosphere melt mixtures from which the alkali basalts were derived also had less radiogenic Sr. However, in this case, tholeiitic lavas would be expected to have a greater range in ɛSr, and to show a more pronounced decrease in ɛSr with distance northward along the ESC. The available data suggest that this is not the case (Figure 16). We can rule out the possibility that postshield lavas from the older Emperor Seamounts contain a larger contribution from the low 87Sr/86Sr, incompatible trace element enriched, lithospheric end-member, because this does not explain the low incompatible trace element concentrations and low La/Ce, Nb/Zr in alkalic lavas from Detroit, compared to young Hawaiian postshield lavas. Similarly, ingrowth of 87Sr in old, metasomatized lithosphere with high Rb/Sr does not explain the low incompatible element concentrations of the Detroit lavas.
 One possible explanation for the observed temporal variation in ɛSr of postshield lavas is that younger oceanic lithosphere has lower incompatible trace element concentrations and lower 87Sr/86Sr. If the oceanic lithosphere is fertilized over periods of several tens of m.y. [Niu et al., 2002; Niu and O'Hara, 2003], by small-degree melts with higher 87Sr/86Sr than normal depleted upper mantle, then the 87Sr/86Sr and incompatible trace element content of the bulk oceanic lithosphere will increase with time, and postshield lavas from seamounts formed on older crust will inherit higher ɛSr and La/Ce.
 Alternatively, a higher degree of melting of the low 87Sr/86Sr, lithospheric end-member might also explain the lower ɛSr, La/Ce and incompatible trace element concentrations of the older Emperor postshield lavas, if more depleted components, with lower 87Sr/86Sr, have higher melting temperatures. Increased decompression melting might be expected in the lower portions of young, thin oceanic lithosphere which would be uplifted to a greater extent above the Hawaiian plume. These two hypotheses could perhaps be tested by comparing volumes of postshield stage lavas along the ESC, although we note that postshield lava volumes on the Hawaiian Islands are highly variable.
 1. Alkali basalts and their differentiates from Nintoku Seamount have similar major and trace element compositions to lavas from the postshield stage of young volcanoes from the Hawaiian Islands. Least squares modeling indicates that fractional crystallization of olivine, plagioclase, clinopyroxene, and Fe-Ti oxides can account for much of the chemical variation within the lava sequence.
 2. A modified version of the Chen and Frey  model for the origin of Hawaiian postshield lavas, in which melts of the Pacific oceanic lithosphere, previously metasomatized by small-degree upper mantle melts, are mixed with melts of Hawaiian plume mantle is proposed to explain the incompatible trace element and isotopic compositions of the Nintoku lavas, without the need to invoke very small degrees of melting.
 3. Sr and Nd isotope compositions of postshield lavas from Nintoku Seamount are within the range of postshield lavas from the Hawaiian Islands, but lie at the depleted end of this range. ɛSr values of postshield, as well as shield lavas decrease northward along the Emperor Seamount Chain. This could reflect either (a) increased melting of the low 87Sr/86Sr end-member beneath volcanoes situated on younger oceanic lithosphere, or (b) less incompatible trace element enrichment (metasomatism) in the lower portions of younger oceanic lithosphere.
 We are deeply indebted to the officers and crew of the R.V. JOIDES Resolution for the help and professionalism in obtaining drill cores during ODP Leg 197. We would like to thank Dennis Birdsell and the Center for Environmental Science and Technology (CEST) at the University of Notre Dame for help with ICP-OES analysis. We thank Adam Kent and associate editor Bob Duncan for reviews. M.R. gratefully acknowledges a grant from the Deutsche Forchungsgemeinschaft while at Mainz. This research used samples provided by the Ocean Drilling Program (ODP), which is sponsored by the U.S. National Science Foundation and participating countries under management of the Joint Oceanographic Institutions Inc. Funding for the research was provided by grants form the United States Science Support Program to C. R. Neal.