Geochemistry, Geophysics, Geosystems

Geochemical structure of the Hawaiian plume: Sr, Nd, and Os isotopes in the 2.8 km HSDP-2 section of Mauna Kea volcano

Authors


Abstract

[1] Sr, Nd, and Os isotopic measurements were made on 110 Mauna Kea lava and hyaloclastite samples from the drillcore retrieved from the second phase of the Hawaii Scientific Drilling Project (HSDP-2). The samples come from depths of 255 to 3098 meters below sea level, span an age range from 200 to about 550-600 kyr, and represent an ordered record of the lava output from Mauna Kea volcano as it drifted a distance of about 40 km over the magma-producing region of the Hawaiian hot spot. The deepest (oldest) samples represent the time when Mauna Kea was closest to the center of the melting region of the Hawaiian plume. The Sr and Os isotopic ratios in HSDP-2 lavas show only subtle isotopic shifts over the ∼400 kyr history represented by the core. Neodymium isotopes (ɛNd values) increase systematically with decreasing age from an average value of nearly +6.5 to an average value of +7.5. This small change corresponds to subtle shifts in 87Sr/86Sr and 187Os/188Os isotope ratios, with small shifts of ɛHf, a large shift in 208Pb/204Pb and 208Pb/207Pb values, and with a very large shift in He isotope ratios from R/RA values of about 7–8 to values as high as 25. When Mauna Kea was closest to the plume core, the magma source did not have primitive characteristics for Nd, Sr, Pb, Hf, and Os isotopes but did have variable amounts of “primitive” helium. The systematic shifts in Nd, Hf, Pb, and He isotopes are consistent with radial isotopic zoning within the melting region of the plume. The melting region constitutes only the innermost, highest-temperature part of the thermally anomalous plume mantle. The different ranges of values observed for each isotopic system, and comparison of Mauna Kea lavas with those of Mauna Loa, suggest that the axial region of the plume, which has a radius of ∼20 km, is a mixture of recycled subducted components and primitive lower mantle materials, recently combined during the formational stages of the plume at the base of the mantle. The proportions of recycled and primitive components are not constant, and this requires there be longitudinal (vertical) heterogeneity within the core of the plume. The remainder of the plume, outside this plume “core zone,” is less heterogeneous but distinct from upper mantle as represented by mid-ocean ridge basalt (MORB). The plume structure may provide a detailed view of mantle isotopic composition near the core-mantle boundary.

1. Introduction

[2] Mantle plumes play a significant role in the processes of heat and mass transfer between Earth's deep interior and its surface. Studies of mantle plumes have impacted hypotheses about plate recycling [e.g., Hofmann and White, 1982] and adiabatic decompression melting in Earth's mantle [e.g., Watson and McKenzie, 1991], but the physical and chemical structures of mantle plumes are difficult to explore directly. Significant advances have been made on the basis of geochemical and geochronologic study of flood basalts [Richards et al., 1989], geophysical studies of plume regions [e.g., Ni et al., 2002; Wolfe et al., 2002; Montelli et al., 2004], and numerical models of Hawaii [e.g., Ribe and Christensen, 1999] and generic thermochemical plumes [e.g., Farnetani et al., 2002; Samuel and Farnetani, 2003]. High-resolution studies of plume volcanic products provide a means to refine hypotheses of the chemical structure of mantle plumes and will provide constraints for future generations of geodynamic models of mantle plumes and mantle convection [e.g., Wüllner and Davies, 1999; Tackley, 2000a, 2000b; Farnetani et al., 2002; Samuel and Farnetani, 2003; Lowman et al., 2004].

[3] The Hawaiian plume offers an advantageous setting for unraveling chemical and physical mantle structures because its tectonic setting is less complicated than several other deep plumes (e.g., Iceland and Afar). The Hawaiian plume is long-lived [Dalrymple et al., 1980; Keller et al., 1995], hot (∼300°C hotter than ambient mantle [Li et al., 2000]) and deep-seated, as indicated by large seismic anisotropy at depths near the core-mantle boundary (CMB) layer [e.g., Fouch et al., 2001]. To gain further insight on the structure and dynamics of mantle plumes, the Hawaii Scientific Drilling Project (HSDP) was designed to recover a significant portion of the eruptive history of Mauna Kea volcano to provide a coherent geophysical and geochemical data set with a semi-continuous time-stratigraphic context. By capturing as much as possible of a volcano's shield stage, the HSDP samples provide a high-resolution geochemical “time series” of a Hawaiian volcano recording the changes in the nature of lava output as the volcano drifts across the melting region of the plume. A unique outcome of the HSDP is the capability to compare Mauna Kea products erupted during a significant portion of its shield-building phase, at approximately the same volcanic stage as those of active Hawaiian-island volcanoes (Mauna Loa and Kilauea).

[4] In this paper we report high-resolution Sr, Nd and Os isotopic data from selected subaerial and submarine Mauna Kea lavas from the second stage of coring in the HSDP (HSDP-2). We then use these data, in combination with other isotopic, trace and major elemental data from the HSDP-2 suite of studies, to consider the differences in isotopic patterns encountered at different stages of volcano growth. These records, when coupled with geochemical data from other Hawaiian shield volcanoes, provide information about the spatial arrangement of the distinct geochemical components (i.e., the scales of heterogeneity) in the Hawaiian plume and thus have implications for the structure of the plume source at the base of the mantle.

2. Samples and Methods

2.1. Overall Stratigraphy of the HSDP-2 Drillhole

[5] The first phase of the HSDP commenced with the drilling and coring of a 1039 meter hole near Hilo, Hawaii (HSDP-1); the second phase (HSDP-2) was cored to a depth of 3098 meters at a site approximately 2 km south of the first hole (Figure 1). Figure 2 summarizes the key stratigraphic relationships in the two drillcores. A detailed discussion of the stratigraphy of the two cores is given elsewhere [e.g., Stolper et al., 1996, 2004; DePaolo et al., 2001; Hawaii Scientific Drilling Project-2, 2000, core logs; Huang and Frey, 2003; Rhodes and Vollinger, 2004; Seamen et al., 2004]. The lower 2843 m of the HSDP-2 core consists of Mauna Kea volcanic products. The subaerial to submarine boundary in Mauna Kea lavas occurs at 1082 meters below sea level (mbsl). The submarine section consists of four different rock-types: hyaloclastite, pillow lava, “massive” basalt, and “intrusive.” Hyaloclastite and pillow lava constitute >90% of the section and are the expected rock types based on models of volcano growth and internal structure [Walker, 1990]. Massive basalt units are defined on the basis of the occurrence of a thick monolithologic section lacking glassy margins or a matrix. The massive units may represent large clasts from hyaloclastites [Stolper et al., 2004]. The “intrusives” encountered in the HSDP-2 core are short sections of core (<1 m) bounded by thin zones of alteration. The intrusives may be dikes supplying pillow lavas [cf. Hawaii Scientific Drilling Project-2, 2000, core logs]. All of the samples in the geochemical reference suite taken from above 2230 mbsl in the core, including all but one of the massive samples and most of the hyaloclastite, are degassed [Seaman et al., 2004]. These rocks were extruded subaerially and were then transported and finally deposited in a submarine environment. A majority of the samples taken from core depths greater than 2230 mbsl are undegassed [Seaman et al., 2004], which indicates that they were erupted under water at substantial depth. Major and trace element chemical data for samples of the geochemical reference set are provided by Rhodes and Vollinger [2004], Feigenson et al. [2003], and Huang and Frey [2003]; other isotopic data are presented by Althaus et al. [2003], Blichert-Toft et al. [2003], Chan and Frey [2003], Eisele et al. [2003], Wang et al. [2003], and Kurz et al. [2004].

Figure 1.

Location sketch-map showing the big island of Hawaii, the volcanoes of the island of Hawaii (Hu, Hualalai; Ki, Kilauea; Ko, Kohala; ML, Mauna Loa; MK, Mauna Kea), and the locations of drillholes HSDP-1 (1993; total depth 1.1 km) and HSDP-2 (1999; total depth 3.1 km). The submerged Mauna Kea shoreline is marked in red at roughly 500 m depth; inferred continuation of the shoreline at depth is marked by the dashed red line. The contour interval is 500 m.

Figure 2.

(a) Schematic stratigraphic column outlining major units and boundaries in the HSDP-2 core stratigraphy. Simplified after Stolper et al. [2004]. (b) Age data for Mauna Kea lavas of the HSDP-2 core are given by Sharp and Renne [2005]. The model age curve (dashed line) is parameterized employing a lava accumulation rate model of DePaolo and Stolper [1996] assuming a 10 cm/yr plate velocity and a 130 ka age for the cessation of Mauna Kea growth. Preliminary data were fit according to a simple quadratic equation so that the model accumulation rate in the submarine section varied between 15 and 21 mm/yr.

2.2. Analytical Procedures

[6] Powders from the Mauna Kea portion of the HSDP-2 geochemical reference suite were used for these measurements. Analytical details for the preparation of samples for isotopic study are provided in Appendix A. Strontium isotopic fractionation was corrected with an internal normalization of 86Sr/88Sr = 0.1194. Nd isotopic compositions were normalized to 146Nd/142Nd = 0.636151 to correct for within-run fractionation and are reported in this study relative to the chondritic uniform reservoir (CHUR) value of 0.511836. Standards run over the course of the analyses provided mean values (with 2σ) of the Berkeley Nd Ames standard (n = 35) of 143Nd/144Nd = 0.510981 (0.000017, 2σ) and for BCR-1 (n = 5) 143Nd/144Nd = 0.511876 (0.000004, 2σ). The ɛNd values reported in Table 1 are normalized to ɛNd = 0 for BCR-1 [cf. Wasserburg et al., 1981]. The average measured value (n = 107) for NIST SRM 987 during the course of this study was 0.710287 (0.000023, 2σ). The 87Sr/86Sr values reported in Table 1 are normalized to a NIST SRM 987 value of 0.710250.

Table 1. Sr, Nd, and Os Isotopic Composition of Mauna Kea Lavas From the Geochemical Reference Suite of the HSDP-2 Core
SampleDepth, mbslUnit Description87Sr/86SrɛNd143Nd/144Nd187Os/188Os
SR0121-4.40246.2subaerial0.7036267.20.512204 
SR0124-3.90252.9subaerial0.7035627.30.512212 
SR0125-6.25256.5subaerial0.7036087.00.512195 
SR0127-4.75261.7subaerial0.7035317.30.5122080.1332
SR0129-5.20267.5subaerial0.7035527.20.512202 
SR0131-6.92274.4subaerial0.703623   
SR0133-8.20281.3subaerial0.703518   
SR0137-5.98293.0subaerial0.7035407.40.512214 
SR0141-7.90305.8subaerial0.7034827.40.512215 
SR0148-8.50326.7subaerial0.7035397.40.512216 
SR0157-6.25353.0subaerial0.7035627.10.5122010.1293
SR0167-5.90378.4subaerial0.7036247.20.512204 
SR0175-5.25398.1subaerial0.7036066.60.512172 
SR0184-2.80421.2subaerial0.7035917.20.5122030.1297
SR0193-0.00443.6subaerial0.7035477.10.512198 
SR0205-1.30467.8subaerial0.7036076.60.512172 
SR0212-8.20490.9subaerial0.7036126.80.512184 
SR0222-2.00516.2subaerial0.7035836.50.512168 
SR0232-8.50542.1subaerial0.7035976.80.512186 
SR0240-3.30563.5subaerial0.7035836.70.5121760.1296
SR0256-0.95589.0subaerial0.7035977.20.512203 
SR0267-6.85615.8subaerial0.7035656.80.5121830.1284
SR0276-7.85636.0subaerial0.7035956.60.512174 
SR0284-1.75658.3subaerial0.7035996.80.512182 
SR0294-7.65678.6subaerial0.7036436.70.512176 
SR0300-6.50695.9subaerial0.7035836.30.512158 
SR0311-4.40724.1subaerial0.7036026.90.512187 
SR0328-3.10759.8subaerial0.7035596.60.5121730.1290
SR0340-1.00793.6subaerial0.7036016.90.512189 
SR0346-5.60812.7subaerial0.7035956.30.5121570.1302
SR0354-7.75833.9subaerial0.7036076.50.512168 
SR0372-2.80871.2subaerial0.7035826.70.5121810.1292
SR0379-3.00888.4subaerial0.7035796.50.512170 
SR0392-4.30921.8subaerial0.7035846.80.5121850.1293
SR0401-2.85948.9subaerial0.7035986.90.512188 
SR0413-3.10984.2subaerial0.7036137.00.512193 
SR0423-3.651012.4subaerial0.7035526.70.512181 
SR0431-8.501037.7subaerial0.7036086.10.512149 
SR0441-9.101061.2subaerial0.7036316.60.512175 
SR0450-3.551083.7massive0.7035716.80.5121850.1296
SR0455-7.401098.2massive0.7036226.50.512168 
SR0472-1.001123.2massive0.7035756.50.512171 
SR0490-1.501229.6hyaloclastite0.7036186.80.512186 
SR0502-4.851265.2massive0.7036416.60.5121750.1295
SR0518-0.801311.9massive0.7036036.60.512172 
SR0531-4.401352.6hyaloclastite0.703645   
SR0545-8.351395.0hyaloclastite0.7035036.90.5121910.1290
SR0548-8.001404.1massive0.7036006.50.512166 
SR0560-7.501435.4hyaloclastite0.7035916.80.512185 
SR0574-1.901474.7hyaloclastite0.7035656.80.5121850.1295
SR0582-10.001497.7massive0.7035406.70.512180 
SR0594-8.701521.4massive0.7035676.70.512180 
SR0603-8.901548.2hyaloclastite0.7035107.00.5121920.1299
SR0604-2.501549.3massive0.7035337.00.512194 
SR0622-7.101581.2hyaloclastite0.7035746.70.512180 
SR0630-6.201605.0massive0.7035986.20.512154 
SR0641-1.001636.3massive0.7035926.60.512176 
SR0655-4.001678.7hyaloclastite0.7035586.50.512170 
SR0664-5.101705.5hyaloclastite0.7035916.60.512171 
SR0675-6.901739.3hyaloclastite0.7036396.70.512177 
SR0683-5.751763.2massive0.7036196.50.512168 
SR0694-9.001794.8massive0.7034506.90.512188 
SR0705-0.151823.2hyaloclastite0.7036796.20.512152 
SR0709-13.351852.0hyaloclastite0.703540  0.1297
SR0714-11.551883.6intrusive0.7035246.70.512179 
SR0720-18.251921.6intrusive0.7036086.40.512166 
SR0723-13.701933.8hyaloclastite0.7035906.50.512168 
SR0732-1.101973.8hyaloclastite0.7036286.20.512154 
SR0741-7.902009.8pillow0.7036506.10.512150 
SR0750-12.452062.7pillow0.7036535.90.512138 
SR0756-13.252098.6pillow0.7036296.00.512145 
SR0762-4.602123.7pillow0.7036526.20.5121520.1305
SR0768-11.202157.4hyaloclastite0.7035796.90.512188 
SR0776-17.702209.5hyaloclastite0.7035776.60.512173 
SR0778-3.202218.2hyaloclastite0.7035986.80.5121860.1300
SR0791-9.502280.2pillow0.7036016.40.5121620.1297
SR0796-6.702300.2pillow0.7036066.30.512156 
SR0800-13.202321.6pillow0.7036206.20.512151 
SR0814-14.402357.0pillow0.7036416.70.5121780.1302
SR0826-20.602414.1pillow0.7035886.20.512153 
SR0836-5.802467.3pillow0.7036376.00.5121440.1283
SR0842-2.352503.5massive0.7036176.10.512149 
SR0846-2.802525.4hyaloclastite0.7035476.80.512186 
SR0850-5.952550.9hyaloclastite0.7035786.90.512188 
SR0855-0.102581.8hyaloclastite0.7035096.40.512165 
SR0860-8.102615.0pillow0.7035546.50.5121690.1307
SR0871-13.002654.1pillow0.7036615.90.512138 
SR0891-15.102730.2pillow0.7035716.30.5121600.1299
SR0896-2.402759.3hyaloclastite0.7036136.40.512163 
SR0899-2.452770.9pillow0.7036506.30.512158 
SR0907-1.652789.9pillow0.7035576.50.512169 
SR0913-2.402825.8intrusive0.7035416.60.512175 
SR0916-1.152837.6pillow0.7035406.80.5121850.1288
SR0930-15.852919.5pillow0.7035806.50.512170 
SR0939-18.102961.0pillow0.7035166.80.512186 
SR0940-18.352967.8intrusive0.7036116.00.5121420.1305
SR0954-8.003009.2pillow0.7035436.80.512185 
SR0956-18.353019.0intrusive0.7035356.50.512170 
SR0964-4.303058.0pillow0.7035516.80.5121830.1300
SR0967-2.753068.9pillow0.7035906.80.512181 

[7] Rhenium and osmium abundances and Os isotopic compositions were measured on selected samples spanning the entire depth of the Mauna Kea portion of the HSDP-2 drill core. Rhenium and osmium concentration data are presented by Lassiter [2003]. Os isotopic values were evaluated from purified samples loaded onto Pt filaments with a mixed Ba(OH)2-NaOH electron emitter and analyzed as OsO3 using a modified Finnigan MAT262 run in negative-ion mode. Further details may be found in Appendix A.

3. Results and Discussion

3.1. Age Model for Mauna Kea Lavas

[8] An essential context for the data presented is the age of the lavas cored by the HSDP. The available data indicate that the youngest Mauna Kea lava in the cores has an age of slightly less than 200 ka [Sharp et al., 1996; Sharp and Renne, 2005]. The deepest sample for which an age has been determined is at a depth of about 2800 meters and has an age of 600 ka (Figure 2b). The uncertainties in the ages of the samples from the submarine part of the section are relatively large due to the small amounts of radiogenic 40Ar and complications in the Ar release patterns [Sharp and Renne, 2005]. The determined 40Ar-39Ar ages for the HSDP-2 lavas between a depth of 382 meters and 2798 meters can be fit with a straight line that corresponds to an accumulation rate of 9 meters per thousand years. Models for the growth of Hawaiian volcanoes, although still crude [DePaolo and Stolper, 1996], suggest that the accumulation rate of lavas and hyaloclastites should decrease systematically upsection. For the HSDP-2 core samples, a simple age model was developed, based on the DePaolo and Stolper [1996] lava accumulation rates. This model was developed prior to the completion of the Ar-Ar dating, but nevertheless fits the data within the 1-sigma error limits. This age model, depicted in Figure 2b, is used in the other papers that report HSDP-2 results, and is retained here for consistency. In general, the lavas in the lowermost 1 km of the core appear from Ar-Ar dating to be somewhat older than the model values. The lower accumulation rate may be partly due to the effect of an overall slightly higher density of the submarine section relative to the subaerial section [Moore, 2001], which is not accounted for in the DePaolo and Stolper [1996] models, or to inaccuracies in the parameterization of melt supply used in the age model.

3.2. Isotopic Stratigraphy of Nd and Sr

[9] The MgO and SiO2 contents of the lava samples are plotted in Figure 3. To facilitate discussion it is useful first to categorize the samples in terms of major element chemical compositional trends as shown in Figure 3 and discussed by Rhodes and Vollinger [2004] and Huang and Frey [2003]. Three compositional end-members can be defined from the whole rock XRF data [Rhodes and Vollinger, 2004], as well as from electron microprobe data from the HSDP-2 glass suite [Stolper et al., 2004]. The tholeiitic shield lavas are separable into “high-silica” and “low-silica” groups. A third group consists of alkaline and transitional lavas found near the top of the cored section (see Figure 2). The two tholeiite groups define linear (olivine control) trends. The high-silica group has a typical SiO2 content of 50% at MgO = 8%, whereas the low-silica group has SiO2 of slightly above 48% at MgO = 8%. The low-silica tholeiites are defined here as those samples where SiO2 < 51.5 − 0.28*MgO; the line separating the high- and low-silica tholeiites is shown in Figure 3. Many of the postshield alkaline and transitional lavas do not plot along olivine control trends and extend to much lower values of SiO2.

Figure 3.

Classification of the samples in the geochemical reference suite by petrology and MgO-SiO2 chemistry (data from Rhodes and Vollinger [2004]). In this figure and those of HSDP-2 data that follow, the following symbols are used: red circle, subaerial lavas; black squares, “massive lavas”; green diamonds, hyaloclastites; inverted triangles, intrusive lavas; upright triangles, pillows. The dashed line represents SiO2 = 51.5 − 0.28*MgO, a way to separate low- and high-silica Mauna Kea volcanic products.

[10] The Nd and Sr isotopic data for HSDP-2 geochemical reference suite samples are provided in Table 1 and plotted against depth and age in Figures 4 and 5, and against each other in Figure 6. The ɛNd values of the HSDP-2 samples span a narrow range from +5.9 to +7.4 (Figure 4a). However, as shown in Figure 5a, there is a systematic progression with age from values averaging ∼+7.0 at 200 ka to values averaging ∼+6.5 at 550 ka. This trend was evident in samples from the shallow HSDP-1 drill hole [Lassiter et al., 1996], but is extended here over an additional 150 kyr in age and to lower ɛNd values. The scatter of ɛNd values over stratigraphic intervals corresponding to about 10 kyr in age is roughly 1 unit of ɛNd. Hence the short timescale variability is similar in magnitude to that of the longer timescale drift. The drift of ɛNd values from +7.5 to +6.5 over 350 kyr is attributed to the changing position of Mauna Kea relative to the plume axis, and may represent radial zoning of the plume in terms of Nd isotopes. The isotopic zoning apparent in the Nd isotopic data corresponds with systematic shifts in He and Pb isotopic compositions as discussed below. In samples from below 2230 m depth below sea level, the high-silica lavas systematically have somewhat higher ɛNd values in comparison to low-silica lavas. The correlation between silica content and isotopic values also applies for helium (low-silica lavas have higher 3He/4He [Kurz et al., 2004]), hafnium [Blichert-Toft et al., 2003], 206Pb/204Pb, and 207Pb/204Pb [cf. Blichert-Toft et al., 2003; Eisele et al., 2003] in the Mauna Kea lavas of the HSDP-2 core.

Figure 4.

(a) Nd isotopic and (b) Sr isotopic composition of HSDP-2 lavas as a function of drilling depth in meters below sea level. The solid symbols are high-SiO2 products; the open symbols are the low-SiO2 products as defined by Figure 3. Volcanic products are keyed by symbol as described in Figure 3.

Figure 5.

(a) Nd isotopic and (b) Sr isotopic composition of HSDP-2 lavas as a function of model age (see Figure 2). Symbols as described in Figure 4.

Figure 6.

The relationship between Sr and Nd isotopic compositions in (a) HSDP-2 Mauna Kea lavas (b) other Hawaiian volcanoes. Symbols as described in Figure 4. The correlation coefficient describing a linear correlation between Sr and Nd isotopic compositions in the submarine lavas is 0.65. Fields in Figure 6b are drawn using data from Chen and Frey [1985], Chen et al. [1991, 1996], Hofmann et al. [1984, 1987], Kurz and Kammer [1991], Kurz et al. [1987, 1995], Leeman et al. [1994], Roden et al. [1994], Staudigel et al. [1984], Stille et al. [1986], West and Leeman [1987], West et al. [1988], and White and Hofmann [1982]. HSDP-1 Mauna Kea data are from Lassiter et al. [1996]. Note that the published values of Lassiter et al. [1996] have been corrected by 0.24 ɛ units for these comparisons. See Appendix A for discussion and justification.

[11] The 87Sr/86Sr ratios of the HSDP-2 samples span a relatively narrow range from 0.70345 to 0.70368 (Figures 4b and 5b). The shift in 87Sr/86Sr over the entire length of the core is considerably subtler than the shift in ɛNd, and far smaller than the corresponding shifts in He and Pb isotopes. Within the main tholeiitic sequence (rocks older than about 300 ka), the upper limit value of 87Sr/86Sr shifts upward by about 0.00005 between 300 ka and 530 ka (Figure 5b). The short timescale (∼10 kyr) isotopic variability is quite small in the subaerial section (0.00005), larger in the submarine section (0.00012), but in both cases equal to or greater than any longer timescale shift of the ratios. Thus for Sr isotopes, the plume apparently has only a small variation within the melting region, and the smaller-scale heterogeneities are of comparable magnitude with the overall radial zoning. The contrast in behavior between the Sr isotopic data from HSDP-2 and other isotope ratios is an important observation that is considered in more detail in a later section.

[12] The relationships of Sr and Nd isotopes with each other as well as other isotopic and geochemical measurements in the core are provided in Figures 678910. In each of the detailed x-y plots, the fields of the youngest (200–250 ka) and oldest (500–550 ka) lavas in the core are indicated. The Sr and Nd isotopic compositions are negatively correlated (Figure 6a). The low-SiO2 pillow lavas define the high-87Sr/86Sr, low-ɛNd extreme, whereas the high-SiO2 lavas cover a wider range of 87Sr/86Sr and ɛNd. The post-shield subaerial lavas define the high-ɛNd extreme. The comparison to other Hawaiian volcanoes (Figure 6b) shows that Mauna Kea lavas are similar to those of Kilauea in Sr and Nd isotopic compositions, slightly different from those of Kohala, and mostly distinct from those of Mauna Loa, Haleakala, and Kahoolawe, so-called Loa-trend volcanoes. Loihi has a relatively wide range of Nd and Sr isotopic values, some of which overlap the values found in the HSDP-2 core samples. This comparison between Mauna Kea as sampled by HSDP and the other volcanoes, however, is not definitive insofar as a much smaller stratigraphic range has been sampled for the other volcanoes.

Figure 7.

The relationship between Nd and Hf isotopic compositions in (a) HSDP-2 Mauna Kea lavas (symbols as described in Figure 4) and (b) other Hawaiian volcanoes. Hf data for HSDP-2 lavas are from Blichert-Toft et al. [2003]. Fields are drawn using data from Stille et al. [1986], Blichert-Toft and Albarède [1999], Blichert-Toft et al. [1999, 2003], and Gaffney et al. [2004].

Figure 8.

Nd-Pb isotopic relationships in HSDP-2 Mauna Kea and Hawaiian lavas. (a) ɛNd versus 206Pb/204Pb in HSDP-2 Mauna Kea lavas, (b) ɛNd versus 208Pb/204Pb, (c) in HSDP-2 Mauna Kea lavas, and (d) 208Pb/204Pb versus ɛNd in Hawaiian lavas. HSDP-2 Pb isotopic data (Figures 8a–8c) are from Blichert-Toft et al. [2003]. For Figure 8d other Hawaiian fields are drawn on the basis of the compilation of analyses in GEOROC (http://georoc.mpch-mainz.gwdg.de/georoc/), including many of the studies identified in Figure 6. Symbols as described in Figure 4.

Figure 9.

Relationships between Nd and O isotopic data for HSDP-2 MK lavas. Oxygen data are from Wang et al. [2003]. Symbols as described in Figure 4.

Figure 10.

3He/4He (R/Ra) versus ɛNd for HSDP-2 MK lavas. Helium data are from Kurz et al. [2004]. Symbols as described in Figure 4.

[13] The relationship between ɛNd and ɛHf in the HSDP-2 Mauna Kea lavas (Hf data from Blichert-Toft et al. [2003]) has some unexpected features despite the small variations (Figure 7a). Most of the tholeiitic shield stage samples show a correlation between ɛNd and ɛHf, with the ɛHf variations being about twice as large as those of ɛNd (Figure 7b), as is typical of Pacific mid-ocean ridge basalt (MORB) and many other oceanic volcanic suites [cf. Chauvel and Blichert-Toft, 2001]. The young, low-SiO2 subaerial lavas are displaced by about +1 unit of ɛNd from the trend defined by the other samples. This shift suggests that the source material melting to produce the alkaline lavas is higher in Sm/Nd than would be expected on the basis of Lu/Hf. This difference was not observed for the HSDP-1 samples [Lassiter et al., 1996; Blichert-Toft and Albarède, 1999]. As shown in Figure 7b, the HSDP-2 lavas differ from those of Mauna Loa mainly in the ɛNd values; the range of ɛHf values for the two volcanoes is similar. Overall, among the lavas of the island of Hawaii, there is a larger range in ɛNd than in ɛHf, and so-called “Kea trend” volcanoes (Mauna Kea, Kilauea, Kohala) are distinguished from the “Loa-trend” volcanoes (Loihi, Mauna Loa, Hualalai) mainly by ɛNd values.

[14] The ɛNd values correlate poorly with 206Pb/204Pb (Figure 8a) but correlate well with both 208Pb/204Pb (Figure 8b) and 208Pb/207Pb (Figure 8c). The low-silica pillow lavas are distinctive in both diagrams, but whereas they follow the same trend as the other samples with respect to 208Pb/204Pb and 208Pb/207Pb, their 206Pb/204Pb is distinctly lower. The excellent correlation of ɛNd values and 208Pb/207Pb, as depicted in Figure 8c, implies that Sm/Nd and Th/U fractionation is linked in the magma sources, as is expected for melting in non–island arc settings [Sims et al., 1995, 1999]. Figure 8d shows the relationship between 208Pb/207Pb and ɛNd for HSDP-2 MK lavas in the context of other Hawaiian lavas. The MK lavas follow the main trend, where in general high 208Pb/207Pb is correlated with lower ɛNd. The Koolau shield, Kahoolawe, Lanai, and recent Mauna Loa lavas follow a distinct trend to much lower ɛNd values.

[15] Neodymium and oxygen isotopic compositions show some correlation in HSDP-2 MK lavas (Figure 9). The main bulk of submarine lavas and hyaloclastites have “normal” δ18O values of 5.0 ± 0.16. Most of the subaerial lavas have lower δ18O values. The preferred interpretation of Wang et al. [2003] is that the low δ18O values are due to lithosphere interaction. Such a process would also be expected to displace ɛNd values to slightly higher values. However, we do not ascribe the Nd isotopic variations to this process because the amount of interaction required to produce significant shifts in ɛNd would also impact Os isotopic ratios, which we do not observe in any of the shield stage lavas as noted in the discussion below.

[16] Mauna Kea lavas sampled by HSDP-2 demonstrate a strong correlation between ɛNd and helium isotopes (Figure 10). The overall range of 1.2 epsilon units corresponds to a shift in helium R/Ra values from 7 to 23. The Mauna Kea samples with the highest R/Ra values and lowest ɛNd are those that fall into the low-SiO2 group near the bottom of the cored section.

3.3. Os Isotopic Stratigraphy

[17] Rhenium and osmium concentration data are presented and discussed in a separate paper [Lassiter, 2003]. Os abundances range from 116 ppt to almost 2 ppb, and correlate broadly with MgO and Ni content. The positive correlation between Os concentrations and MgO content is consistent with the removal of Os from the magma during olivine fractionation. Although recent studies suggest that Os is not compatible in olivine itself [Burton et al., 1999, 2002], Os is compatible in both Cr-spinel [e.g., Walker et al., 2002] and in sulfides, and these phases typically co-precipitate with olivine in many tholeiitic basalts. Rhenium concentrations vary from 111 to 914 ppt, but show no correlation with MgO content or any other chemical or isotopic variables. However, Re abundances increase systematically with increasing depth in the core. This correlation reflects Re loss during degassing of shallow submarine and subaerial lavas and retention of Re in lavas erupted at significant water depth [Bennett et al., 2000; Lassiter, 2003].

[18] Osmium 187Os/188Os ratios span only the narrow range of 0.1283 to 0.1305 for shield stage lavas (Table 1 and in Figures 11 and 12). Shield lavas from other Hawaiian volcanoes span a much larger range in 187Os/188Os (∼0.128–0.148 [Lassiter and Hauri, 1998]). One HSDP-2 sample, a subaerially erupted low-silica subalkaline (postshield) lava, has high 187Os/188Os. This lava has especially low Os concentration and low MgO, and the high 187Os/188Os is interpreted as resulting from interaction with the oceanic crust. The rest of the core samples show characteristics that are analogous to those shown by Nd and Sr isotopes. The oldest samples include those with the highest 187Os/188Os ratios, and the upper limit of this ratio tends to decrease with decreasing eruption age. The oldest lavas, however, are also quite heterogeneous, and the short-time-scale variability is slightly larger than the longer-term trend from the bottom to the top of the HSDP-2 section. Although the range of Os isotopic variability is limited (0.128–0.131), there is some correlation with Nd and He isotopes (Figure 12). The higher 187Os/188Os ratios tend to be associated both with lower ɛNd and higher 3He/4He.

Figure 11.

Os isotopic composition of HSDP-2 lavas as a function of (a) drilling depth in meters below sea level and (b) model age. Symbols as described in Figure 4.

Figure 12.

The relationship between Os and (a) Nd isotopic compositions and (b) He isotopic signatures in HSDP-2 Mauna Kea lavas. He data from Kurz et al. [2004]. Symbols as described in Figure 4.

[19] Most HSDP-2 Mauna Kea lavas have slightly lower 187Os/188Os than those of HSDP-1 samples, which have reported 187Os/188Os values ranging from 0.1294 to 0.1322 [Lassiter and Hauri, 1998]. Because the HSDP-1 Mauna Kea lavas and the subaerial portion of the HSDP-2 core sample approximately the same period of Mauna Kea volcanism, such systematic differences in isotopic composition of lavas from the two drill cores are not expected. The differences are attributable to artifacts of the analytical procedures employed in the earlier study (see Appendix A for further discussion). Similar subtle differences between HSDP-1 and HSDP-2 lavas have been observed for other isotopic systems, and in the case of Pb isotopes appear to be related to the fact that identical (strong) leaching procedures do not yield exactly reproducible results [Eisele et al., 2003]. For the present purposes, the discrepancy between the HSDP-1 and HSDP-2 results for Os is of little consequence. The results presented here cover a larger stratigraphic range and can be used to assess the Os isotopic variations independent of the HSDP-1 reported values.

[20] The Os isotopic data pertain to the origin of the high- and low-silica lavas in the Mauna Kea tholeiitic shield section. Feigenson et al. [2003] have suggested that the low-silica samples are preferentially melted from enriched components in the plume, possibly pyroxenites or eclogites with high modal garnet. Such components should have radiogenic Os isotopes, and so we would expect systematically higher 187Os/188Os in the low-silica samples. This is not observed.

3.4. Mauna Kea Versus Mauna Loa

[21] The variations of the Nd, Sr, 208Pb/207Pb and Os isotopic signatures in the lavas from HSDP are compared to those of Mauna Loa and Kilauea in Figure 13. The new results from HSDP-2 provide a means to examine a longer geochemical time series within an individual volcano. When compared to studies of historical samples of Hawaiian volcanoes [e.g., Pietruszka and Garcia, 1999], shifts in Nd, Sr and 208Pb/207Pb isotopic composition seen over depth intervals in the HSDP core corresponding to about 10 kyr are comparable in magnitude to those observed for the ∼300 year history of Kilauea summit lavas (see Figure 13). Mauna Loa lavas, on the other hand, have larger isotopic shifts over timescales of 2 kyr or less, especially for the lavas erupted since about 10 ka.

Figure 13.

(a) Nd, (b) Sr, (c) 208Pb/207Pb, and (d) Os isotopic composition in Mauna Kea and Mauna Loa lavas as a function of age. Data are taken from this study and Kurz et al. [1987, 1995, 1996], Kurz and Kammer [1991], Bennett et al. [1996], Hauri et al. [1996], Lassiter et al. [1996], Hauri and Kurz [1997], Brandon et al. [1999], Pietruszka and Garcia [1999], and Blichert-Toft et al. [2003]. Factors used to construct the Mauna Loa age model are described by DePaolo et al. [2001]. Strontium data are normalized to NIST SRM-987 = 0.710250. Note that the Mauna Kea HSDP-1 lavas are keyed separately from HSDP-2 lavas in Figure 13d. These Os isotopic compositions were determined using different analytical procedures as described in Appendix A. Although there is some nonsystematic bias between the two sets of procedures for preparation of samples for isotopic study, Mauna Loa Os isotopic signatures are distinctly more radiogenic than Mauna Kea isotopic signatures.

[22] The Nd, Sr and Os signatures of both Mauna Kea and Mauna Loa lavas show some of the characteristics that are to be expected from a simple radially zoned plume model. This model is discussed in more detail below, but broadly speaking, we expect the ɛNd values to increase with time (i.e., decreasing age), and 87Sr/86Sr and 187Os/188Os to decrease with time. Mauna Loa lavas, however, demonstrate a second type of variation that cannot be effectively modeled with radial variations in the plume. The Mauna Kea trend of increasing ɛNd from 550 ka to 200 ka, and the Mauna Loa trend of increasing ɛNd from 200 ka to about 40 ka are the expected trends. The Mauna Loa ɛNd values are lower than those of Mauna Kea. This can be attributed to the fact that Mauna Loa crossed the plume closer to the plume axis, and because the effects of the plume core persist in the later stages of Mauna Loa's growth due to viscous coupling of the plume with the moving Pacific plate [DePaolo et al., 2001]. The differences could also indicate some NE-SW asymmetry in the plume as suggested by Abouchami et al. [2005]. In the age range 0–40 ka, the Mauna Loa lavas show a reversal to low ɛNd values that is not predicted by either the radial or asymmetric model. This reversal is evident also in the Sr isotopic data, and for Sr appears to begin at 150 ka or even at 200 ka. To account for this reversal it is necessary to have longitudinal (vertical) variations of ɛNd, 87Sr/86Sr (and, in considering Figure 13c, also 208Pb/207Pb) along the axis of the plume. The effects of the anomalous material are evident also in the lavas of Hualalai [DePaolo et al., 2001]. This geochemically anomalous “blob” is also apparent in the He isotopic data, insofar as the Mauna Loa lavas trend sharply toward low 3He/4He over the past 40 kyr [Kurz and Kammer, 1991; Kurz et al., 2004]. The effects of the anomalous material are found only in the volcanoes that traversed the plume along a track that crossed near the plume center; they are not found in the Mauna Kea record, or in what is known of the Kohala record. Os isotopic ratios show analogous differences between Mauna Loa and Mauna Kea (Figure 13d).

[23] Our conclusion is that the Hawaiian plume has two types of geochemical structure. Radially, it has relatively systematic variation of isotopic ratios outward from the “outer core” of the plume to the fringes of the melting region. Within the core of the plume, the isotopic ratios are much more heterogeneous and vary longitudinally (vertically), approximately along the plume axis, by a larger amount than they do in the radial direction. In the following sections we evaluate the radial and nonradial isotopic variations in the plume in more detail.

3.5. Geochemical Zoning in the Hawaiian Plume

[24] In this section we evaluate the isotopic variations measured in the HSDP core in the context of a mantle plume that has predominantly radial isotopic variations. There most certainly are additional isotopic heterogeneities in three dimensions, but the general dependence of isotopic ratio on age for most of the isotopes measured suggests that there is radial structure [e.g., Lassiter et al., 1996; Hauri et al., 1994, 1996; DePaolo et al., 2001]. We discuss nonradial structure and other models for the Hawaiian plume in section 3.7.

[25] Figure 14 shows the isotopic ratios versus age for samples from the HSDP-2 core for the isotopes of Nd, Os, He, 208Pb/207Pb, Hf and Sr. Data other than those reported here are from Blichert-Toft et al. [2003], Eisele et al. [2003], and Kurz et al. [2004]. In each figure the lower limit of the vertical axis is set to approximate values expected for ambient asthenospheric upper mantle based on data from East Pacific Rise MORB, and the upper limit is the most distinctive value observed for lavas from the island of Hawaii. Consequently, the scale of the variations as represented in the figures is comparable between isotopic systems. Our analysis is similar to that described by DePaolo et al. [2001] for He and Nd isotopes, but here we consider other isotopes and evaluate primarily a strictly radial symmetry for the plume geochemistry. The objective is to compare the data to what might be expected on the basis of a simple radial plume model. This comparison will enable us to estimate what the isotopic composition range of the plume axis material is, and the consistency among the observed variations for different isotopes.

Figure 14.

(a) Helium, (b) neodymium, (c) strontium, (d) osmium, (e) hafnium, and (f) 208Pb/207Pb isotopic ratios versus model age for HSDP-2 Mauna Kea lavas. HSDP-2 data (solid circles) are from this study, Blichert-Toft et al. [2003], Eisele et al. [2003], and Kurz et al. [2004]. Sr and Nd HSDP-1 data (open circles in Figures 14b and 14c) are from Lassiter et al. [1996] and are renormalized to 87Sr/86Sr = 0.710250 for NIST SRM 987 and ɛNd = 0 for BCR-1. Hf HSDP-1 (open circles in Figure 14e) data are from Blichert-Toft and Albarède [1999]. Settings for ambient asthenospheric upper mantle (lower limit of vertical axis) determined by typical East Pacific Rise MORB values (see Table 2). Superposed curves represent the scale length for isotopic variations in the Hawaiian plume. See text for discussion.

3.5.1. Thermal and Velocity Structure of the Hawaiian Plume

[26] The model we use as a dynamical basis for our evaluation is that of Ribe and Christensen [1994, 1999] (hereafter referred to as RC1999), with slight modifications based on scaling relationships from Hauri et al. [1994] and Zhong and Watts [2002]. Certain characteristics of the plume structure are important for the interpretation. In particular, the “width” of the plume is much different if one considers temperature, as opposed to upwelling velocity or magma generation rate. And the “width” of the isotopically anomalous material in the plume may or may not correspond to, for example, the width of the thermal anomaly. Differences can provide important and unique information about the plume source.

[27] The RC1999 model includes calculations of mantle flow and melt generation caused by a circular temperature anomaly placed in the upper mantle at a depth of 400 km. The (horizontal) distribution of the potential temperature excess (Δθ) in the resulting plume is described by

equation image

where r = 0 corresponds to the location of the axis of the plume, and aθ is a measure of the thermal radius of the plume. The total width of the thermal anomaly is approximately 4aθ, which is full width at about 2% of the maximum temperature. The calibration of the RC1999 model is based on the height and width of the Hawaiian swell, as well as the total rate of magma generation. For the preferred parameterization, aθ ≈ 65 km, and Δθo = 300K at a depth of about 170 km under Hawaii; a depth slightly greater than that at which melting begins. The amplitude of the potential temperature anomaly is set so that the appropriate amount of melt is generated to make the volcanoes of the island of Hawaii, using the McKenzie and Bickle [1988] parameterization of peridotite melting, the plume has a viscosity that is appropriate to the height and width of the Hawaiian swell, and an appropriate thermal buoyancy flux is generated.

[28] According to Ribe and Christensen [1999], the Hawaiian swell topography constrains the buoyancy flux and the minimum viscosity of the plume. The buoyancy flux (B) is estimated to be about 4000 kg/s, of which about 25% is “depletion buoyancy” caused by removal of high-density components from the mantle by partial melting and melt extraction. The minimum plume viscosity must be ≥5 × 1017 Pa-s [Ribe and Christensen, 1999], otherwise the Hawaiian swell is not high enough relative to its width. These two constraints allow derivation of self-consistent values for the upwelling velocity and temperature structure of the plume. Zhong and Watts [2002] developed a similar model but did not consider melting. For comparison, Zhong and Watts [2002] conclude that aθ is between 50 and 70 km, and that Δθo ≈ 400°C. Although not explicitly provided in the RC1999 model, the values of maximum plume upwelling velocity is close to Wo ≈ 0.36 m/yr, based on their maximum calculated melting rates.

[29] On the basis of the calculations of Hauri et al. [1994], for a plume with a Newtonian, temperature-dependent viscosity, the upwelling velocity varies with radial position approximately as

equation image

where Wo is the axial upwelling velocity and W′ is about 4 cm/yr. The constant upwelling velocity at large r in this approximation is of no consequence because very little of the buoyancy flux is contributed by the fringes of the thermal anomaly, and no melting occurs there. The models of Hauri et al. [1994] suggest that for axial plume viscosity of order 1018 Pa-s, the value of aW is about 0.6aθ. This result implies that the length scale for variation of upwelling velocity is aW ≈ 40 km.

3.5.2. Radius of the Melting Anomaly in the Hawaiian Plume

[30] The ∼100 km-thick lithosphere under Hawaii restricts melting to upwelling mantle with high potential temperature [Watson and McKenzie, 1991; Ribe and Christensen, 1999]. Consequently, the radius of the melting region is much smaller than the radius of the temperature anomaly. For the parameterizations used in the RC1999 model, to a sufficient approximation the magma supply varies radially as

equation image

where Go is about 0.05 m3/m2/yr, and aG ≈ 35 km. The maximum radius of the melting region occurs at a depth of about 120 km, where the plume is spreading beneath the lithosphere. Consequently, the nominal width of the melting region in the RC1999 model, as described by aG(120 km) ≈ 35 km, corresponds to a smaller width at a depth of 400 km of aG(400 km) ≈ 20 to 25 km. This width is only about 30% of the radius of the thermal anomaly, and about half of the radius of the upwelling velocity anomaly. The mantle that undergoes melting has Δθ > 0.85Δθo. This means that if the ambient mantle potential temperature is 1300°C, and the maximum plume potential temperature is 1600°C, then only those parts of the plume with potential temperature higher than 1555°C undergo substantial melting.

[31] DePaolo and Stolper [1996] and DePaolo et al. [2001] favor a narrower melting region than that employed in the RC1999 model, suggesting that the value of aG(120 km) appropriate to the Hawaiian plume is about 25 km rather than 35 km. This difference is not large, and could be mainly a function of the melting behavior of the mantle or the viscosity structure of the plume, both of which have uncertainties that could generate this difference. The various radii are illustrated in Figure 15. The conclusions reached below would require significant reevaluation only if the thermal and upwelling velocity radii are substantially smaller than those we assume.

Figure 15.

Illustration of the radii of the melting region and thermal anomaly beneath Hawaii as a function of depth.

3.5.3. Volcano Sampling and Radial Variation of Melt Fraction

[32] The total melt production from the Hawaiian plume is estimated to be 0.15 to 0.2 km3/yr [Watson and McKenzie, 1991; DePaolo and Stolper, 1996; Ribe and Christensen, 1999]. The typical eruption rate of a Hawaiian volcano during its shield-building phase is between 0.03 and 0.12 km3/yr (Kilauea is estimated to be erupting 0.1 km3/yr at present). The magma erupting from a single volcano during its main stage of shield building therefore comes from a substantial fraction of the melting region of the plume (20 to 70%), and hence the geochemical signals are averages representing mixtures of melts from the axis of the plume as well as melts from the fringes of the melting region. To relate radial plume geochemical variations to geochemical variations in the lava output, one must have a model to specify how the volcano samples the plume magma output, and how melting processes affect the concentrations of elements in magma coming from different parts of the plume. Such models have been described previously [DePaolo and Stolper, 1996; DePaolo et al., 2001].

[33] As the volcano drifts over the plume, carried by the moving Pacific plate, the model posits that the volcano samples plume magma from a circular region of radius 25 km centered on the volcano summit (Figure 16). The 25-km scale length for magma generation (aG) is smaller than that in the RC1999 model in an effort to provide a more accurate model prediction of the lifetimes of Hawaiian volcanoes. The parameter Go is set so that the volcano volumes are reproduced and hence it is proportional to the inverse of the plate velocity relative to the plume. For a plate velocity of 10 cm/yr relative to the melting region, Gmax is equal to 5 cm/yr, which corresponds to a total magma production of 0.17 km3/yr for the plume. The resulting magma generation map (Figure 16) has a shape that is broadly consistent with numerical models [Watson and McKenzie, 1991; Ribe and Christensen, 1999] although the width of the melting region is smaller and it has radial symmetry instead of being slightly asymmetric along the NW-SE axis that corresponds to the relative velocity vector of the Pacific plate and the plume.

Figure 16.

Magma generation map beneath Hawaii after DePaolo et al. [2001], based on the model of DePaolo and Stolper [1996]. Superimposed on the melt supply model are the volcano “tracks” as a function of age and an illustration of the magma capture area used in the parameterization. The HSDP portions of the volcano tracks are indicated in red. The capture area is illustrated around Loihi for clarity. Loihi, erupting both tholeiitic and alkalic lavas, is likely on the edge of the magma capture zone [cf. DePaolo et al., 2001].

[34] In the model, the magma supplied to each volcano is derived from a circular region that spans a range in isotopic composition and a range in magma generation rate. This stipulation is arbitrary in that it is not based on physics; but it is defensible insofar as it generates appropriate magma supply to each volcano and matches individual volcano volumes reasonably well [Baker et al., 2003]. The average isotopic value of the magma supply to a particular volcano as a function of time (or position of the volcano) must be calculated by weighting the isotopic values in the plume by an assumed magma generation rate, and also by the concentration of the element in the magma. Both the magma generation rate and the concentrations of trace elements in magmas vary with distance from the plume axis. Our model assumes initially that the concentration of each trace element is constant in the original (unmelted) plume mantle (i.e., does not vary with radial position), and that the melt fraction, which is the integrated amount of melting from the bottom to the top of the melting region, varies with radial position according to

equation image

For Fmax we use the value of 0.2 [after Ribe and Christensen, 1999; Watson and McKenzie, 1991]. The melt generation rate as a function of radial position, G(r), is approximately equal to the product of the upwelling velocity W(r) and the vertically integrated melt fraction F(r) [cf. Spiegelman and Elliott, 1993]. Both W(r) and F(r) must vary with distance from the plume axis. However, because aW (≈60 km at the mean depth of melting) is substantially larger than aG, (≈25 km) as discussed above, upwelling velocity varies much more slowly with r than does magma generation rate. Consequently, the melt generation rate, G(r), is approximately proportional to melt fraction, F(r) which implies that aFaG = 25 km. Using this approximation for aF, and FMAX = 0.2, the concentration of each chemical element in the supplied magma is calculated from equation (4) and the equilibrium bulk melting equation:

equation image

where Ci(r) is the concentration of element i in the produced magma as a function of radial distance from the axis of the plume, Di is the solid/liquid distribution coefficient for element i (assumed to be a constant), and Cpi is the concentration of element i in the plume mantle before melting begins. Equation (5) applies for steady state porous flow magma migration in an upwelling column of mantle rock undergoing adiabatic decompression melting [Spiegelman and Elliott, 1993]. The net result of the concentration weighting is that, for incompatible trace elements, the fringes of the melting region, where the melt fraction is relatively small, get extra weight as a result of high concentrations in the liquids, and hence have more influence on the average isotopic composition of the magma supply than would be calculated on the basis of the magma generation rate alone (cf. Asimow and Langmuir [2003] for a recent discussion of this effect at mid-ocean ridges). More compatible elements (Os for instance) are weighted essentially by the magma generation rate because, according to equation (5), COs(r) varies little with r. The distribution coefficients used to calculate concentrations for our model are DNd = 0.024, DSr = 0.017, DPb = 0.020, DHe = 0.001, DHf = 0.028, and DOs = 5. These values, which are typical based on the data summarized on the GERM website (http://www.earthref.org) need only be approximate, as the model sensitivity to this parameter is low. Because of the low sensitivity to the distribution coefficient values, there is effectively no difference between the D values for Nd, Sr, Pb and Hf, and in the calculations described below, D = 0.02 is used for all four elements.

3.5.4. Radial Variations of Isotopic Ratios

[35] The considerations discussed above are important for evaluating the geochemical structure of the plume. The lavas erupted from Hawaiian volcanoes are derived from a relatively small-diameter inner core of the plume and hence sample only a small fraction of the thermally anomalous mantle [Ribe and Christensen, 1999]. As shown in Figures 14a–14f, however, the lavas cored from Mauna Kea show isotopic variability over the 400 kyr history represented. This requires that the melting region be zoned in isotopic composition, even though it constitutes only the innermost part of the plume. In the case of helium isotopes, this innermost core of the plume contains most of the range of variation in 3He/4He observed for the Hawaiian plume.

[36] To assess the patterns in Figure 14, we evaluate a model where the isotopic values in the plume, Rp, are a function of radius according to

equation image

where RAMB is the mantle value remote from the plume axis, RMAX is the maximum (or minimum) isotopic value at the axis of the plume, r is radial distance from the plume axis, and “aR” is a scale length for isotopic variations to be evaluated from the data. The value of aR is estimated separately for the isotopes ratios of each element (Nd, Sr, Pb, Hf, Os). The isotopic values need not be arranged in this way, but the approach allows assignment of a radial length scale to the isotopic variations that corresponds to the other length scales discussed above for the thermal, velocity and melting structure of the plume.

[37] The predictions of the model described above are juxtaposed with the data in Figure 14. For each graph, reference curves are shown for different values of aR. The baseline model is aR = 100 km, which is the approximate 1/e radius of the thermal anomaly at the depth (120 km) where the melting region is widest. The other curves are for aR = 50 km and for aR = 25 km. The chosen values are bracketed by the Pacific MORB values (for RAMB) and the most anomalous lavas from the island of Hawaii (for RMAX). The values of both RAMB and RMAX are subject to uncertainty. The curves shown indicate the expected magnitude of isotopic variations with time, and how they correspond with isotopic variation with position in the plume. Changing the values of RAMB or RMAX corresponds to a telescoping of the family of curves in the y axis direction.

[38] The general features of the model can be appreciated by inspection of Figure 14a. If the isotopic anomaly corresponds in size to the thermal anomaly (i.e., aR ≈ 100 km), then within the HSDP core, the isotopic ratio should not vary greatly and should be close to the maximum, axial value in the plume (RMAX). At the other extreme, the isotopic ratios in the youngest lavas approach the ambient upper mantle value (RAMB) only if the radius of the isotopic anomaly is less than the radius of the melting region, which in turn is about 1/4 the radius of the thermal anomaly (i.e., aR ≤ 25 km).

[39] The observed variations of isotopic ratio with time in the HSDP-2 core are broadly consistent with the radial plume model. The inferred value for the isotopic anomaly radius (aR) is different for different isotopic ratios. For He isotopes (Figure 14a) the amount of variation shown by the isotopic measurements in the HSDP core is close to what is expected for aHe = 15–25 km, roughly 1/6 of the radius of the temperature anomaly. This conclusion, that the He-isotope anomalous material in the plume is much narrower than the thermal anomaly, is essentially that reached by DePaolo et al. [2001] for He isotopes considering all of the data available for the island of Hawaii. No value of aHe can capture the combination of very high 3He/4He ratios in the age range 450–550 ka and the low values for age <300 ka. The simple model we are discussing here does not account for radial differences in the helium concentration in the plume. One way to produce a pattern that resembles the data more closely is to assign a higher helium concentration to the innermost core of the plume. However, the heterogeneity displayed by the lavas in the age range 450–550 ka is significant [Kurz et al., 2004]. The model curves for aHe = 15–25 km capture a majority of the data. The units with especially high 3He/4He are all low-silica types and are interspersed in the section with high-silica units with lower 3He/4He. The overall conclusions are that the high-3He/4He material is confined to the axial region of the plume, but also that this axial material is highly heterogeneous with respect to He isotopes and contains both high- and low-3He/4He mantle.

[40] For Nd (Figure 14b) the relatively subdued radial zoning of ɛNd requires that the length scale for the Nd isotopic anomaly is in the range aNd = 30–50 km, larger than that for helium but still significantly smaller than the thermal anomaly length scale. For Nd the value of RMAX is estimated from the Kilauea data and the values of older Mauna Loa lavas (Figure 13). As noted by DePaolo et al. [2001], however, the small variation of Nd isotope ratios may partly result from coupling of the plume to the overriding Pacific lithosphere. A plausible distribution of ɛNd values in the plume melting region is shown in Figure 17a, constructed in much the same way as the map given by DePaolo et al. [2001] based on the geochemical measurements summarized in Table 2, but with the added constraint that the initial estimate is described by the Gaussian function represented by equation (6), and including a simple linear trail approximately, but not exactly, along the trend of the Hawaiian chain. The available data are fit reasonably well with a value of aNd = 35 km and RMAX = 5, and with the trend of the isotopic trail rotated about 10–20° relative to the trend of the Hawaiian ridge between Maui and Hawaii. This reconstruction compares well with that given by DePaolo et al. [2001].

Figure 17.

(a) Neodymium and (b) strontium isotopic maps of the Hawaiian plume. Procedure used to create maps is after DePaolo et al. [2001], employing the values in Table 2.

Table 2. Input Parameters for Sr and Nd Isotopic Maps of the Hawaiian Plumea
VolcanoAge, kaMean 87Sr/86SrRange 87Sr/86SraMean ɛNdRange ɛNda
Loihi00.70350.7034–0.70366.14.8 –8.3
Mauna Loa2000.703670.70365–0.703705.55.3–5.7
Mauna Loa1000.70360.7036–0.703665.5–6.5
Mauna Loa00.70380.7036–0.70404.53.2–7.4
Hualalai00.70370.7036–0.70385.55.3–5.7
Kilauea00.70360.7035–0.70376.55.9–7.1
Mauna Kea5000.70360.70346–0.703686.65.9 –7
Mauna Kea3000.70360.70355–0.703637.37.1–7.5
Mauna Kea2000.703550.7035–0.70367.4 
Mauna Kea00.70340.7033–0.70357.56.6–8.5
Kohala2400.70370.7036–0.70386.65.7–7.4
Kohala1000.70350.7033–0.70367.16.8–7.4
North Arch 0.70310.70304–0.70311  
Pacific MORB 0.70260.70215–0.70271  

[41] For Sr isotopes, there is very little variation with age in the HSDP section. All of the lavas have a moderately strong “plume” isotopic signal that is clearly distinguishable from the ambient mantle values, which are characterized here as having the average Pacific MORB value of 0.7026 (see Table 2) but may range between those of the North Arch volcanic field (∼0.70304 [Frey et al., 2000]) to the lowest Pacific MORB values of ∼0.70215 [Hamelin et al., 1984]. Radial variation in 87Sr/86Sr in the plume must be mainly confined to the region peripheral to the melting region and hence is essentially unrecorded in the lavas. The HSDP data can be fit well by setting RMAX to 0.7037, and aSr = 70–100 km. The value of RMAX = 0.7037 is consistent with modern values of Kilauea lavas, which should be sampling the modern central portion of the plume. The inferred width of the Sr isotope anomaly implies that Sr isotopes in the plume scale approximately with the excess temperature anomaly, which is in stark contrast to the He isotopic variation. As with Nd, the Sr isotopic variations in the HSDP-2 core may be subdued as a result of plume-lithosphere coupling. In Figure 17b, we show a plausible two-dimensional Sr isotopic “map” for the Hawaiian plume melting region, constructed in a manner analogous to the Nd isotopic map using aSr = 70 km. The map shows some of the same features as the Nd map, including the slight departure of the trend of the trail from the Hawaiian ridge axis and the necessity for an anomalous region located beneath the present position of the Mauna Loa summit that accounts for the recent eruption from Mauna Loa of lavas with high 87Sr/86Sr (as well as low ɛNd, low 3He/4He, and high 187Os/188Os).

[42] The relationships between the isotopic maps shown in Figure 17 and the Mauna Kea and Mauna Loa stratigraphic time series data (Figure 13) are shown in Figure 18. The dashed lines in the figure show the predicted time dependence of the Nd and Sr isotopic parameters over the past 600 kyr corresponding to the isotope ratio maps. The Mauna Kea trend is monotonic and in the “expected” sense. The slope of the Mauna Loa trend is opposite in sign, reflecting the anomalous blob inferred to be located near the present position of the Mauna Loa summit. The model captures only roughly the rapid shift of the values in the Mauna Loa lavas since 40 ka.

Figure 18.

Time series of Hawaiian island volcanic products compared with predictions of isotopic trends (dashed lines) predicted from the isotopic maps of Figures 17a and 17b).

[43] The high Kd value for Osmium implies that the isotopic values of the plume core should predominate because Os is contributed in proportion to the magma supply. Hence the relatively low 187Os/188Os values and the minimal variation through the HSDP core can only be fit with a low value of RMAX ≈ 0.131. This value for RMAX is consistent with values for modern Kilauea lavas, but lower than almost all of the values from Mauna Loa. The difference with Mauna Loa can be attributed to the character of the anomalous “blob” of material that is evident in the Nd and Sr isotope maps of Figure 17.

[44] Hafnium isotopic variations fit expectations for the lavas in the age range 320–550 ka (Figure 14e), and suggest a value of aHf (≈25–50 km) that is similar to that inferred for Nd isotopes. The HSDP-2 Hf data for lavas younger than 320 ka diverge from model predictions. Hf data from lavas from this time period from the HSDP Pilot hole (open circles, Figure 14e), however are more consistent with model predictions.

[45] The 208Pb/207Pb data show strong time variation in the HSDP lavas (Figure 14f). They cover most of the range seen in Hawaiian lavas from all of the islands. The older lavas fit the model curves reasonably well for aPb = 50 km. The lavas younger than 300 ka diverge from the aPb = 50 km curve. There is very little variation of 207Pb/204Pb in the core; hence both the 208Pb/204Pb and the 208Pb/207Pb ratios reflect mainly variations of 208Pb.

[46] The differences in the inferred values of the length scale that describes the isotopic variations (aR), are roughly similar for Nd and Hf, for Sr and Os, and for He and 208Pb. We have not evaluated in detail the effects of radial zoning with respect to the concentration of these elements. Some of the differences may be made smaller if there are significant and systematic differences in concentration between the core zone of the plume and material surrounding it. However, with the possible exceptions of He and Os, we do not expect the differences in concentration within the plume to be large enough to change the model results as shown in Figure 14. For Os, variations in concentrations will not significantly alter the inferences made here. For He, there appears to be a need for substantial concentration variation to explain the data.

3.5.5. Isotopic Discontinuities in the HSDP Section

[47] Several of the isotopic parameters exhibit discontinuous changes of character at model age values of 300 ka and 450 ka (Figure 14). The 300 ka discontinuity is apparent in 208Pb/207Pb and 206Pb/204Pb, where the values shift downward substantially, and in ɛHf, where the values also shift downward and depart from the expected trend. The 300 ka time horizon corresponds to the age at which 3He/4He ratios are no longer elevated relative to MORB values. At this time ɛNd values become less variable, and 87Sr/86Sr values slightly more variable. The 300 ka age also corresponds fairly closely with a dramatic drop in the lava accumulation rate (Figure 2b), but is prior to the appearance of alkalic lavas in the core. Lavas older than 450 ka have high and highly variable 3He/4He, somewhat more variable 87Sr/86Sr, high and relatively variable 208Pb/207Pb. The 450 ka model age corresponds roughly to the boundary between subaerial and submarine lavas.

3.6. Isotopic Structure of the Lowermost Mantle

[48] The combination of the HSDP-2 data from Mauna Kea, the comparison with Mauna Loa data (see Figure 13), and data from other Hawaiian volcanoes (Figure 17) suggest that the most radical departures from typical upper mantle isotopic compositions are contained in the core of the plume. Using the dimensions that apply to the plume before it starts spreading due to interaction with the lithosphere, this “plume core” is a zone ∼30 km wide (full width) that is embedded within the 260 km-wide thermal anomaly that constitutes the plume. The typical characteristics of this plume core zone are high but variable 3He/4He, moderately high 87Sr/86Sr (0.7037), low ɛNd (+5), low ɛHf, (+11), and high 187Os/188Os (0.131) relative of ambient upper mantle as sampled by Pacific MORB. However, the core zone also is longitudinally (vertically) heterogeneous and has a large range of isotopic values (3He/4He = 10 to 32 Ra, 87Sr/86Sr = 0.7037 to 0.7044, ɛNd = 0 to +6, ɛHf = 0 to +11, and 187Os/188Os = 0.130 to 0.140). The region outside this core zone is less heterogeneous vertically within the plume, distinct from typical upper mantle compositions sampled by MORB, and radially zoned outward from the plume core. This “outer core” of the plume typically has values of 3He/4He = 8 to 15 Ra, 87Sr/86Sr = 0.7035 to 0.7037, ɛNd = +6 to +7, ɛHf = +12 to +14, and 187Os/188Os = 0.128 to 0.131.

[49] This description of the plume structure is illustrated for Nd, Sr, and He isotopes in Figures 192021, which show the inferred distribution of He isotope ratios in the plume (Figure 19; adapted from DePaolo et al. [2001]), the range of Nd and He isotopic values in the plume plotted against estimated potential temperature (Figure 20), and the range of He and Sr isotopic compositions plotted against depth in the mantle (Figure 21). The plume core has the highest potential temperature, shows the most heterogeneity (Figure 20), and has the highest 3He/4He ratios (Figure 19). The fringe of the melting region (“outer core” zone) has more systematic structure and is transitional to the ambient mantle characteristics. However, a large proportion of the plume is invisible in that it does not melt and consequently cannot be characterized through studies of the volcanic products (Figure 20).

Figure 19.

Inferred distribution of He isotope ratios in the Hawaiian plume [cf. DePaolo et al., 2001] in comparison to the thermal anomaly beneath Hawaii.

Figure 20.

(a) Nd and (b) He isotopic values in the plume as a function of estimated potential temperature. The greatest degree of heterogeneity is associated with the hottest mantle.

Figure 21.

Schematic model of the Hawaiian plume. The central core zone of the plume is inferred to come from either the core-mantle boundary or is a tendril of a basal dense layer that is entrained by the plume along its axis. The material in the plume core is highly heterogeneous, containing materials that appear to be recycled oceanic crust as well as material that could be regarded as approximating “primitive.” An essential aspect of the plume core mantle is that the character of what is being melted under the volcanoes changes with time. The plume material that is just outside the plume core, which is referred to in the text as the “outer core” of the plume, is less heterogeneous than the plume core but still distinct from the ambient upper mantle. High 3He/4He is found only in the plume “core zone;” high 87Sr/86Sr is found in both the plume core and “outer core” zones.

[50] If the plume comes from a thermal boundary layer (TBL) at the base of the mantle, then the isotopic structure observed in the Hawaiian melting region can be mapped to the geochemical structure of the TBL. The core of the plume corresponds to the material closest to the core-mantle boundary, and a traverse from the core of the plume to the fringe of the plume corresponds to a vertical traverse upward through the TBL (Figure 21) and lower mantle. The models of Farnetani et al. [2002] and Hauri et al. [1994] provide indications of the amount and scale of entrainment of lower mantle materials and suggest that all of the material that passes through the relatively narrow melting region comes from the thermal boundary layer. The highest temperature material coming up the axis of the plume flows into the plume conduit from a large area at the base of the mantle. This material should be heterogeneous since broad regions of the base of the mantle are being telescoped into the narrow axial region of the plume. Because the most “anomalous” isotopic compositions, particularly the highest 3He/4He ratios, are confined to a small inner core of the plume, it follows that these isotopic characteristics apply to only a thin layer at the base of the mantle.

[51] An alternative, but in many ways similar, model for the plume source is that it is both a thermal and compositional boundary layer at the base of the mantle, not at the core-mantle boundary but at the top of a dense layer that mantles the core and corresponds to the seismically defined D'' layer [cf. Lay and Helmberger, 1983; Lay et al., 1998; Bréger and Romanowicz, 1998]. According to Jellinek and Manga [2002], the fixity of plumes can be explained by the likely situation that this dense material is drawn into “peaks” under buoyant upwellings, and the peaks persist for long time periods, focusing the upward flow and maintaining the position of the plume. The axial part of the plume can be composed of entrained material from the dense layer. In this model the material in the axial region of the plume would come from the dense layer, and the material in the “outer core” of the plume should be from the base of the overlying main lower mantle reservoir.

[52] In the TBL model, since the plume “core” is material derived from very close to the core-mantle boundary (probably within a few kilometers), it is tempting to attribute the source of helium with especially high 3He to the core, if in fact it is reasonable to expect the core to have sufficient He of the appropriate isotopic composition [cf. Porcelli and Halliday, 2001; van Keken et al., 2002]. The presence of this signal in the plume would require transport of helium out of the core into at least the basal few hundred meters or more of the mantle. In the model of a dense layer at the base of the mantle, the high and variable 3He signal would be attributable to the dense layer. Samuel and Farnetani [2003] suggest that it may be possible to sequester primordial helium in a dense layer over the age of the Earth, given a sufficiently large density contrast (>2.4%).

[53] A particularly important feature of the HSDP-2 data is the observation that the 3He/4He ratios of the lavas decrease to the ambient upper mantle value of R/Ra ≈ 8 defined by MORB (Figure 14a) well below the top of the section. This implies that the lower mantle that is immediately above the basal part of the TBL or immediately above the dense layer is similar to MORB mantle and does not have high 3He/4He. Hence the HSDP data give no support to a mantle model where a large fraction of the lower mantle has high 3He/4He [e.g., Kellogg et al., 1999]. High 3He/4He may be a feature that is restricted to only the basal mantle layer, at least under Hawaii, and the bulk of the lower mantle may be similar to MORB mantle with respect to 3He/4He. However, whereas the He signal is restricted to the lowermost lower mantle, the Sr isotope signal is not. The Sr isotope ratio (≈0.7035–0.7037) of the melting region of the Hawaiian plume may be characteristic of a much larger volume of the lower mantle. This ratio is clearly displaced from the average MORB value (≈0.7026) and from “bulk Earth” Sr (≈0.7045). The difference in the physical distribution of the He and Sr isotopic signals suggests that it is unlikely that they have the same origin.

[54] The conclusions reached here are consistent with the idea that the strength and longevity of plumes may be enhanced by the presence of relatively cool lower mantle material [Jellinek and Manga, 2002]. If the lower mantle is cool as a result of the presence of a high proportion of deeply subducted oceanic lithosphere, the temperature/buoyancy contrast with the thermal/compositional boundary layer will be maximized, and plumes are more likely to be formed and more likely to persist for geologically long times. Hence the fact that Hawaii is a strong, persistent plume is consistent with the He isotopic data, which suggest that the ambient lower mantle under Hawaii is similar in isotopic composition to the upper mantle.

3.7. Plume Models in Hawaii

3.7.1. Nonradial and Longer-Term Isotopic Variations in the Hawaiian Plume

[55] The radial plume model describes with a certain degree of success many geochemical features of lavas from the island of Hawaii. There are other isotopic variations that are not mainly radial. One type of variation is shorter-term (∼10 kyr) fluctuations in isotopic ratios about the general trends that we have ascribed to radial structure [e.g., DePaolo, 1996; Blichert-Toft et al., 2003; Eisele et al., 2003]. Another type of variation has been ascribed to a northeast-southwest asymmetry that is discussed in terms of two distinct volcanic chains: the so-called “Loa trend” (which on Hawaii comprises Hualalai, Mauna Loa, and Loihi volcanoes) and the “Kea trend,” which comprises Kohala, Mauna Kea, and Kilauea [Frey and Rhodes, 1993; Abouchami et al., 2000; Eisele et al., 2003].

[56] Short-term fluctuations of isotopic ratios (see Figures 4, 5, 11, 13, and 14) indicate that there are reasonably large-amplitude, short-length-scale isotopic variations in the plume that are superimposed on any larger-scale (such as radial) structure. DePaolo [1996] discussed the significance of these variations. Magma generation and transport processes in the plume tend to attenuate the shortest length scale isotopic variations (<100 m). The major process responsible is physical dispersion in the melt-producing zone. Magma mixing is not a major attenuation process during the shield stage because magma chambers are small in comparison to the magma flux such that chamber residence times are short (<1000 years). The sampling frequency for the HSDP-2 core is about 2000–5000 years, so the variations observed are sufficiently long that they are unlikely to be affected much by magma chamber storage. DePaolo [1996] estimated that the longitudinal (vertical) dispersivity in the melting column in Hawaii is in the range 100–1000 meters. This implies that there is sufficient mixing in the melt zone (in the vertical direction) to homogenize effectively isotopic ratios over vertical distances of order 1 km. For upwelling velocities in the range 10–50 cm/yr, this distance corresponds to timescales of 2000 to 10,000 years. Because of the large capture area for magma for each volcano (a diameter of about 50 km), there is a large amount of averaging in the horizontal direction. The isotopic composition of any erupted lava therefore can represent an average over a disc-shaped region (actually a bowl-shaped region because of the radial variation in upwelling velocity) in the plume with vertical dimension ≤1 km, and a radius of ∼25 km. The data (Figure 13) show variations of most isotopic parameters on timescales of 2000–10000 years (e.g., ±0.5 units of ɛNd and ɛHf, several units of R/Ra for helium). These variations should map to variations in the plume of similar amplitude at length scales of ∼1 km, or to variations of somewhat larger amplitude at shorter length scales. The large variations of He isotope ratios [Kurz et al., 2004] over times of only a few thousand years may result from inhomogeneous radial sampling of the magma capture area.

[57] We have concentrated on modeling the recent geochemical structure of the Hawaiian plume as represented by the volcanoes of the island of Hawaii. It is, however, clear from inspection of geochemical trends such as that depicted in Figure 6b that the composition of the plume has varied over longer time periods as well. The recent numerical experiments of Samuel and Farnetani [2003] are relevant to the possible origins and scales of heterogeneity in plumes. In their numerical experiments, they investigated the efficiency of mixing between a basal dense layer with primordial helium and overlying mantle enriched in subducted components, as these materials are tapped at the plume source and brought up in a plume without complete mixing. Several plumes formed in the numerical simulations have large variations in the peak He isotopic signal (e.g., 10–15 R/RA versus 30 R/RA) along the central axis of the plume as a function of depth (and thus time). These variations occur on length scales as small as ∼500 km to 600 km, which, given plume axial upwelling velocities of about 0.5 m/yr correspond to a timescale of about 1 Myr. Such temporal variation in the core of the plume in a radial-zoning model could easily explain the range of values encountered in shield lavas of older volcanoes such as Koolau [Roden et al., 1994] and Kauai [Mukhopadhyay et al., 2003]. However, the Mauna Loa data, as well as recent studies of the Koolau volcano indicate that even finer-scale vertical variations exist in the axial region of the plume. The rapid transitions in isotopic character seen in Mauna Loa (Figures 12 and 17) suggest length scales in the 10–50 km range, and large amplitude variations of all isotopic ratios at these shorter length scales. These results suggest that there is not only mixing between major reservoirs near the base of the mantle (e.g., a dense primordial layer and overlying mantle with subducted components), but that at least the dense primordial layer may have large amplitude heterogeneity at small length scales.

[58] In the radial model presented here, differences between the so-called Loa and Kea trend volcanoes are produced as a result of the Loa trend volcanoes passing over the plume at a position closer to the plume axis, and to the “trail” of the plume trending in a direction that is rotated about 10–20° relative to the trend of the volcano tracks over the plume (Figures 14 and 16). Other differences are produced by longitudinal (vertical) heterogeneity along the plume axis, in particular the “blob” of anomalous plume material near the present location of Mauna Loa (Figure 16). The model presented here shows that it is possible to account for the Loa-Kea geochemical differences with radial structure in the plume. We also suggest that the existence and extent of the differences is time dependent and related to longitudinal heterogeneity in the plume rather than a strict northeast-southwest asymmetry.

[59] The orientation of the “trail” of the Hawaiian plume, as depicted in Figure 17 does not follow precisely the trend of Hawaiian Ridge, the latter being the trend inferred for the velocity vector of the Pacific plate relative to the plume. This configuration is discussed by DePaolo et al. [2001]. Although the difference between the two trends appears arbitrary, it should be noted that the trend of the Hawaiian ridge over the past 2 million years is N30°W, which is much more northerly than the long-term trend of the Hawaiian archipelago (N65°W). Hence there is reason to believe that there may be a difference between the orientation of the Hawaiian ridge and the sub-lithospheric position of the plume and its geochemical trail under the lithosphere.

3.7.2. Comparison With Other Hawaiian Plume Models

[60] The plume model presented here is a refinement of the DePaolo et al. [2001] study. The essential characteristic of these models is that geochemical variations in Hawaiian lavas are placed in a quantitative geodynamic context using models available for thermo-chemical plumes and their interaction with the lithosphere. This approach contrasts with interpretations based on “cartoons” of the Hawaiian plume [e.g., Hauri, 1996; Hauri et al., 1996; Lassiter et al., 1996; Kurz et al., 1996] which do not address the complex relationships between plume thermal and velocity structure, plume geochemical structure, melting in the plume, and volcano growth. The main distinction of our models is that they effectively capture many of the geochemical variations within the ∼400 kyr history of the Mauna Kea portion of the Hawaiian drillcore, and at the same time account for the growth of the six major volcanoes of the island of Hawaii and the geochemical differences between them.

[61] One limitation of our model is that it represents only the mean isotopic values of lavas erupted from each volcano as a function of time. Because of the large sampling area associated with each volcano, the model can only capture changes that occur over time periods of 100 to 300 thousand years or longer. As the data show (Figure 14), there are short-timescale (≤10,000 year) variations that are similar in magnitude to the longer timescale trends. Even smaller-scale heterogeneities have been inferred on the basis of Pb isotopic heterogeneity in melt inclusions [Kobayashi et al., 2004]. Small-scale heterogeneities are not explicitly represented in the model, but can be accommodated by superimposing smaller length scale heterogeneity on the larger-scale structure shown in Figure 17.

[62] Longer-term (1–5 Myr) geochemical trends of Hawaiian volcanoes (depicted in Figures 6b, 7b, and 8d) demonstrate that the hypothetical structure of the Hawaiian plume, as characterized by the <1 Ma magmas erupted by the volcanoes of the island of Hawaii, does not show the more extreme isotopic values found in some of the older Hawaiian volcanoes. Our model can accommodate these more extreme values as simply amplified examples of the features exhibited by the youngest Mauna Loa lavas. We attribute such signatures to (1) the high degree of heterogeneity of the “core zone” of the plume and (2) the fact that the characteristics of the core zone that is passing through the melting region change on timescales of <105 to >106 years. This aspect of our model is consistent with the interpretations of Blichert-Toft et al. [2003], Eisele et al. [2003], and Abouchami et al. [2005] which also call for longitudinal (vertical) heterogeneity in the plume. There is also agreement with Mukhopadhyay et al. [2003], who call for temporal changes in the proportion of recycled materials in the Hawaiian plume. Numerical simulations of Samuel and Farnetani [2003] demonstrate physical mechanisms by which heterogeneous mixtures of recycled and undegassed material are brought to shallow levels of the mantle. Two other results of Samuel-Farnetani numerical simulations are significant to this debate: relative proportions of the recycled and undegassed materials in the simulations change over time as the plume “ages” and a cross-section across their simulated plumes reveals quasi-radial (as opposed to bilateral) symmetry.

[63] Measurements of Pb isotopic time series of Hawaiian volcanic products are interpreted by Eisele et al. [2003] and Abouchami et al. [2005] in terms of bilateral symmetry in the Hawaiian plume. The Kea trend volcanoes of the NE side of the island show many fairly consistent differences with those of the Loa trend on the SW side of the island. Abouchami et al. [2005] cite the relative homogeneity and consistency of the isotope ratios in the Kea trend volcanoes as evidence that heterogeneities in the plume are strung out into long vertical filaments (“spaghetti”) that allow melts derived from these parts of the plume to remain isotopically unchanging for 1 million years or more. This characteristic of the Hawaiian plume we ascribe to the “outer core zone” which is much less heterogeneous than the inner core zone of the plume. Abouchami et al. [2005] argue that the Loa-trend volcanoes require that the SW side of the plume be different from the NE side. We show in Figure 17 how the parts of the Loa-trend volcanoes that have so far been sampled, can be different for Sr and Nd isotopes from the Kea trend volcanoes even with radial symmetry in the plume. We cannot at this time prove that bilateral asymmetry does not exist, but Abouchami et al. [2005] have similarly also not proven that it does exist. It is also unclear how this bilateral asymmetry is manifested, in terms of both spatial distribution and amplitude of isotope anomalies, so that the observations they cite can be captured in the context of the magma-capture area model (Figure 16) or a similar approach that accounts for the magma supply to the volcanoes. Ultimately this debate will have to be addressed with more extensive data from volcanoes to the southwest of the axis of the Hawaiian plume. Mahukona volcano could potentially shed some light on this argument, but there are no existing high-resolution geochemical time series from this volcano. The few existing Mahukona He isotopic data (3He/4He in olivine phenocrysts ∼21 R/Ra [Garcia et al., 1990]) could support either model.

4. Summary and Conclusions

[64] We have presented new Nd, Sr and Os isotopic data from the Mauna Kea portion of the 3 km HSDP-2 core. The drill core represents a 350–400 kyr record of the lava output of the volcano as it drifted 35–40 km across the ∼100 km-wide main melting region of the Hawaiian mantle plume. The isotopic shifts (ɛNd, 87Sr/86Sr, and 187Os/188Os) with depth/age are small but systematic. The small variations of these three isotope systems contrast with relatively large variations of 3He/4He [Kurz et al., 2004] and 208Pb/204Pb [Eisele et al., 2003]. The variations of ɛHf [Blichert-Toft et al., 2003] are similar in magnitude to those of ɛNd; the two parameters correlate in the shield stage tholeiitic lavas but not in the late- to post-shield lavas (younger than 300 ka).

[65] A radially zoned plume model is used to evaluate the data. Because of the thick lithosphere under Hawaii, only the highest temperature, axial region of the plume melts, so the HSDP lava section represents only this inner part of the plume. If the isotopic anomalies associated with the Hawaiian plume are distributed like excess temperature, which has an approximately Gaussian radial profile with a 1/e length scale of 70–100 km in the melting region, then the HSDP core should show very little variation from top to bottom. This expectation is realized for Sr isotopes, 206Pb, and 207Pb, and to a lesser degree for Nd, Hf, and Os isotopes. For He and 208Pb, however, the large and systematic variation within the HSDP core indicates that the isotopic anomalies are much narrower than the thermal anomaly, i.e., restricted to the axial region of the plume. Since the radial structure of the plume can be mapped to the vertical structure of the boundary layer from which the plume originates at the base of the mantle, the data indicate that the 3He and 208Pb anomalies are restricted to a thin layer at the base of the mantle, whereas the other isotope signatures in the plume characterize a thicker zone at the base of the mantle. Although the isotopic data conform broadly to a radially zoned plume, there are also short-length scale isotopic heterogeneities superimposed on the larger-scale radial zoning. The data suggest that most of the mantle beneath Hawaii, between the core-mantle boundary and the uppermost mantle need not have elevated 3He/4He, and that the origin of the plume 3He signature cannot be the same as that of the Sr, Nd, Os and Hf isotopic signatures.

[66] The HSDP-2 data are used in conjunction with data from the literature on other volcanoes from the island of Hawaii to produce a revised Nd isotopic map of the Hawaiian plume [after DePaolo et al., 2001] and a new Sr isotope map. An important component of these maps is the comparison between Mauna Kea and Mauna Loa lavas. Conclusions from these comparisons are that much of the isotopic structure of the plume can be captured by a radial model, so that there is no requirement for a northeast-southwest asymmetry in the plume, and that the axial region of the plume has pronounced longitudinal (vertical) heterogeneity on length scales of 10 to 500 km. The implication is that the base of the mantle, which should supply the axial region of the plume either from a thermal boundary layer or from a high-density mantle layer, is both anomalous and highly heterogeneous.

[67] The modeling of the plume isotopic structure presented here is still crude because the physical interaction of the plume with the overriding Pacific lithosphere creates a complicated relationship between plume structure and the melting region. Improvement of our understanding of the structure of the plume will require a new generation geodynamic model that incorporates melting and tracers, and may also require a formulation for reactive transport that captures the effects of percolating magma on the composition of previously melted solid residue that is coupled to the lithosphere and carried over the melting region.

Appendix A

A1. Sr and Nd Isotopic Analyses

[68] Powders from the Mauna Kea portion of the HSDP-2 geochemical reference suite were used for these measurements. As discussed by Rhodes and Vollinger [2004], these powders were prepared from chips that were repeatedly rinsed in deionized water and then pulverized, following procedures developed by Rhodes [1996]. The samples were all recovered from depths of between 230 and 3070 meters below the present sea surface. The pore fluid in the section encountered during the first phase of the Hawaiian Scientific Drilling project has a composition essentially the same as seawater, and hence has a relatively high 87Sr/86Sr as well as a high Sr concentration (8 ppm [Thomas et al., 1996]). While lenses of freshwater were encountered during drilling of HSDP-2, we assumed conservatively that the majority of the core was subjected to interaction with seawater at depth for timescales of hundreds of thousands of years. We employed leaching to remove any nonmagmatic Sr signatures that were acquired as a result of diagenetic processes. Leaching tests show that a very strong leaching procedure (hot HCl-HNO3) is required to remove the secondary Sr. The corresponding loss of Sr from these leaching procedures varied between our test samples from a loss of <25% to ∼75% of the original Sr in the sample. Analyses of residues from repeated leaching procedures on the hot HCl-HNO3 leached residue yielded no decrease in Sr isotopic ratios; thus we concluded from our preliminary leaching tests that boiling the powders in a mixture of strong HCl and HNO3 resulted in the most effective leaching. The leaching procedure is less critical for Nd isotopic measurements, because of the very low Nd concentrations in seawater, but the Nd isotopic measurements are carried out on leached samples as well, since both Sr and Nd isotopic measurements were made on the same dissolved powder aliquots.

[69] Approximately 200–250 mg aliquots of geochemical reference suite powders were boiled for one hour in mixed 8 N HNO3 – 6 N HCl and then ultrasonically homogenized. After several ultrasonic rinses in nanopure H2O and decantations, the residue was then digested in a HF-HNO3-HClO4 mixture. Samples for Sr isotopic study were prepared using standard ion chromatographic separation techniques using Eichrom Sr spec resin. Purified Sr salts were loaded in Ta2O5 on rhenium filaments for analysis. The strontium isotopic measurements were made on a VG sector-54 multicollector mass spectrometer operating in dynamic mode. Isotopic fractionation was corrected with an internal normalization of 86Sr/88Sr = 0.1194. The long-term average for NIST SRM 987 at the CIG is 0.710283 (0.000018, 2σ), and the average for NIST SRM 987 during the HSDP-2 analyses was 0.710287 (0.000023, 2σ). The reproducibility of the strontium isotopic values reported in Table 1 was evaluated in two ways. One method employed separate loads of an internal HSDP-2 standard that was analyzed in each magazine containing HSDP-2 samples, a total of seventeen times. The average, with 2σpopulation, for the HSDP-2 internal standard (SR-694-9.00) was 0.703484 (0.000028). The second way we established the reproducibility was through separate preparation of replicate samples through the leaching-dissolution-separation process. These duplicates are reported in Table A1. We have renormalized all Sr isotopic data we discuss in this paper to a NIST SRM value of 0.710250 and have plotted them with 2σ of 0.000025, a conservative measure of the reproducibility. For comparison, the NIST SRM 987 value relevant to the Sr isotopic analyses of HSDP-1 lavas as reported by Hauri et al. [1996] and Lassiter et al. [1996] was 0.710240 (0.000038, 2σ).

Table A1. Replicate Analyses of Sr and Nd Isotopic Compositions of Mauna Kea HSDP2 Samples
 87Sr/86Sr87Sr/86SrMean Value 
Sr duplicate analyses
    SR630-6.200.7036050.7036130.703609 
    SR705-0.150.7036770.7036820.703680 
    SR723-13.700.7035720.7036220.703597 
 
SampleɛNd Run 1ɛNd Run 2ɛNd Run 3Mean Value
Nd duplicate analyses
SR125-6.257.056.98 7.02
SR127-4.757.307.22 7.26
SR240-3.306.616.69 6.65
SR284-1.756.676.80 6.74
SR714-11.556.646.696.786.70
SR723-13.706.436.51 6.47
SR756-13.255.846.23 6.04
SR778-3.206.756.92 6.84
SR850-5.956.896.86 6.88
SR896-2.406.406.37 6.39

[70] Neodymium elemental separations were carried out using a three-column procedure: first, sample solutions were loaded onto a small Ag 50 × 4 ion chromatographic exchange column where the major elements were eluted in dilute HCl; secondly, the sample solution was loaded onto a column containing Eichrom TRU-Spec resin wherein REE elements were concentrated following procedures adapted from Luo et al. [1997]. Separation of the rare earth elements was subsequently carried out using standard ion chromatographic procedures described by DePaolo [1978]. Purified Nd salts were then loaded in 5 N HNO3 onto rhenium filaments; Nd was run as NdO+ in an environment ∼10−6 mbar oxygen pressure. Nd isotopic measurements were also made on the multicollector mass spectrometer operating in dynamic mode; Nd isotopic compositions were normalized to 146Nd/142Nd = 0.636151 to correct for within-run fractionation and are reported in this study relative to the chondritic uniform reservoir (CHUR) value of 0.511836. The results presented in Table 1 reflect analytical results corrected for oxygen and other rare earth element oxide interferences (e.g., Pr and Sm). The maximum acceptable correction was ∼1 ɛ unit, and replicate analyses were carried out for those samples requiring corrections between 0.5 and 1 ɛ unit (see Table A1). Measurements of standards over the course of the HSDP-2 analyses provided mean values (with 2σ) of the Berkeley Nd Ames standard (n = 35) of 143Nd/144Nd = 0.510981 (0.000017) and 143Nd/144Nd = 0.511876 (0.000004) for BCR-1 (n = 5). On the basis of evaluation of both the BCR-1 analyses and the project-duration values for the Berkeley Ames standard, the Nd isotopic values reported in Table 1 are normalized to ɛNd = 0 for BCR-1 (for the HSDP-2 data, this amounts to a correction of ɛNd of 0.7).

[71] The reproducibility of the Nd isotopic values for HSDP-2 lavas was determined via several sets of duplicates (Table A1). Mean values of the duplicate analyses are provided in Table 1. As is evident in Table A1, reproducibility of Nd isotopic values for HSDP-2 lavas was for the most part excellent, better than 0.2 ɛ units.

[72] During the course of the HSDP-2 project, it became apparent that there was a history since 1994–1995 of drifting Nd isotopic values as measured by the Berkeley VG sector 54 multicollector mass spectrometer. This shift was not entirely convincing when it first started in 1994–1995, as it was within 2σ limits for individual runs (∼0.000012 shift in 143Nd/144Nd). Through time and continued analyses of standards, of both BCR-1 and the more frequently analyzed UCB Nd solution (which has a 143Nd/144Nd isotopic value of the BCR-1 minus 0.000897 ± 7), it has become apparent that the drift in Nd isotopic values requires correction back to BCR-1 values of ɛNd = 0. Accordingly all data in Table 1 were corrected from the measured values by an ɛ value of 0.7 units. Corrections are also required for some published HSDP data with which we compare our values. The Mauna Loa Nd isotopic data of DePaolo et al. [2001] are corrected by 0.7 ɛ units. The HSDP-1 Nd isotopic data for most of the samples reported by Lassiter et al. [1996] and Hauri et al. [1996] are corrected by 0.24 ɛ units. The exceptions to this are samples R166-5.25 and R-189.850 which required no correction as they were run on the single collector mass spectrometer or prior to late 1994 when the drift on the multicollector began. Samples R197-0.80 and R215-7.20 required more substantive corrections of 0.48 and 0.7 ɛ units, respectively, because they were run at a different time from the rest of the HSDP-1 samples.

A2. Os Isotopic Analyses

[73] The analytical techniques utilized for Os extraction and purification in this study are significantly different than those used in the initial Os-isotope study of the HSDP-1 lavas [Hauri et al., 1996; Lassiter and Hauri, 1998]. In the initial HSDP-1 studies, Os was preconcentrated using a NiS fusion technique and extracted via double distillation into HBr [cf. Hauri and Hart, 1993]. Os was then further purified using a single Chelex resin bead. The procedures used in this study represent a significant improvement over those utilized in the initial study of the HSDP-1 lavas. For example, total procedural Os blanks in the current study averaged ∼300 fg/g, whereas in the earlier Lassiter and Hauri [1998] study, procedural blanks averaged ∼1 pg/g. In the present study, Os blank 187Os/188Os isotopic signatures are ∼0.24. Reported 187Os/188Os (Table 1) are blank corrected, using the Os concentrations reported by Lassiter [2003]. Blank corrections are in all cases significantly smaller than analytical uncertainty. 187Os/188Os measured on a MPI in-house standard during the course of this study was 0.1070 (±4‰, 2σ). 187Os/188Os measured on a DTM in-house standard at MPI during the same period was 0.1742 (±4‰, 2σ).

[74] We have reanalyzed several Mauna Kea and Mauna Loa samples from HSDP-1 to evaluate whether the Os-isotope differences between the HSDP-1 and -2 samples are real or are an analytical artifact. We have also analyzed aliquots of the DTM in-house Os standard to evaluate possible interlaboratory biases. A comparison of the DTM and MPI Os-isotope data is presented in Table A2. The average 187Os/188Os value for four analyses of the DTM J-M shelf standard at MPI was 0.17415 (±4‰, 2σ), indistinguishable from the DTM average of 0.1742 (±3‰ 2σ). However, the reanalyzed Mauna Loa and Mauna Kea samples have consistently less radiogenic 187Os/188Os values than the originally reported analyses. Furthermore, the two reanalyzed HSDP-1 Mauna Kea lavas have 187Os/188Os values of 0.1279 and 0.1289, consistent with the HSDP-2 Mauna Kea 187Os/188Os values reported here but lower than the range of 187Os/188Os values previously reported for HSDP-1.

Table A2. Comparison of Os Isotopic Analyses at DTM and MPI
HSDP-1 Sample187Os/188Os (DTM)187Os/188Os (MPI)
R103 (ML)0.13524a0.13323
R153 (ML)0.13286a0.13222
R365 (MK)0.12967b0.12790
R413 (MK)0.13278b0.12891

[75] We have carefully reexamined the earlier HSDP-1 Os-isotope analyses to determine the source of discrepancy between the original and the new data. Although the procedural blanks were much higher in the original study than in the current study, all samples were blank-corrected, and an underestimate of the procedural blank by 100% would still not lower the 187Os/188Os values of the original HSDP-1 analyses sufficiently to bring them in line with the new analyses. However, there do appear to be weak positive correlations between measured 233/236 (equivalent to 185Re16O3), 234/236 (186Os16O3), and 235/236 (187Re16O3, 187Os16O3) ratios in the earlier HSDP-1 analyses. The slope of these correlations is inconsistent with an origin from either Re interference (which would affect the 233/236 and 235/236 ratios, but not 234/236) or from residual mass-fractionation correction errors. Instead, we suggest that minor organic interferences may have compromised some of the original HSDP-1 analyses. The Chelex bead cleanup step in particular appears to be less effective at removing organic interferences than the microdistillation technique currently in practice. Thus it appears that the systematic offset of the HSDP-1 samples to more radiogenic 187Os/188Os values relative to the HSDP-2 lavas is an artifact. Given the narrow range of the HSDP-2 Os-isotope data and the consistency of the reanalyzed HSDP-1 samples with other Mauna Kea samples from HSDP-2, we believe that the new MPI data are more reliable than the earlier HSDP-1 data. We note however that although the apparent discrepancy between the earlier HSDP-1 Mauna Kea Os-isotope data and the new HSDP-2 data are significantly larger than the reported analytical uncertainty, the offset is small compared to the total range of 187Os/188Os values reported for Hawaiian shield lavas as a whole. Therefore the basic conclusions drawn from inter-volcano comparisons in the original studies of Os-isotope variations in HSDP-1 Mauna Kea and Mauna Loa lavas [Hauri et al., 1996; Lassiter and Hauri, 1998] remain valid.

Acknowledgments

[76] We gratefully acknowledge the efforts of and discussions with the HSDP-2 on- and off- site scientific teams. We thank Tom Owens for assistance with the analytical work and Lisa Hammersley for assistance with Figure 1. We are indebted to an anonymous referee, Fred Frey, Mike Garcia, Al Hofmann, Associate Editor Ed Stolper, and Editor Bill White for their extremely thorough reviews that substantially improved this manuscript. Support for this project was provided from the NSF Continental Dynamics Program through grant EAR-9528544 to D.J.D. Support for scientific drilling was also provided by the International Continental Drilling Program. Support for the Berkeley Center for Isotope Geochemistry is also provided by the Director, Office of Energy Research, Basic Energy Sciences, Chemical Sciences Division of the U.S. Department of Energy under contract De-AC03-76SF00098.

Ancillary