A1. Analytical Methods
Glass major element compositions were analyzed by electron microprobe at the U.S. Geological Survey in Menlo Park (Table 2 and Table A1) using natural and synthetic standards as described by Clague and Frey . Major element contents for whole-rock samples, except 10 samples, were analyzed by X-ray fluorescence (XRF) with FeO, H2O+, H2O−, and CO2 analyzed by classical wet chemical techniques at the U.S. Geological Survey (USGS) laboratories in Denver, Colorado, and Menlo Park, California, respectively (Table 3). Ten additional samples were analyzed by XRF at University of Massachusetts, Amherst (Table 3). Data for one sample, 71WMOL-4, analyzed by both laboratories (Univ. Mass and USGS) agree within 3% except for MnO (Table 3).
Table A1. Glass Compositions for Dikes and Artifactsa
| 89KAA-1 (dike)||−157.1915||21.2072||49.5||3.32||15.0||11.72||0.16||6.41||9.83||3.42||0.98||0.56||0.01||100.9|
| WMO-11 (dike)||−157.1484||21.1394||53.6||3.27||13.6||12.40||0.18||4.11||8.08||2.63||1.02||0.67||0.01||99.6|
| 89FDD-1 (dike)||−157.1518||21.1367||51.4||2.55||14.5||11.20||0.16||6.70||10.26||2.24||0.51||0.30||0.01||99.8|
| WFD (dike)||−157.1505||21.1375||52.1||2.50||14.3||11.16||0.16||6.69||10.58||2.50||0.51||0.38||0.00||100.9|
| WFD-1 (dike)||−157.1505||21.1375||52.0||2.66||14.6||11.41||0.17||6.33||10.31||2.58||0.55||0.33||0.02||101.0|
Trace elements for the seven dive samples (T307Rx) were analyzed at the Geoanalytical Laboratory at Washington State University. Information on methods, precision, and accuracy for samples analyzed at this facility are available at http://www.wsu.edu/∼geolab/. Trace element abundances for all the other samples were determined at MIT by ICP-MS (Table 4) using a Fisons VG Plasmaquad 2 + S with both internal and external drift monitors. Trace element results are reported as the mean of duplicate analyses obtained on different days. The chemical procedures and estimates of accuracy and precision were discussed by Huang and Frey . The 2 sigma uncertainty for BHVO-2, which is analyzed as an unknown sample, is better than ±3%. Ten of the fourteen samples with REE-Y enrichment were not analyzed by ICP-MS but their trace element abundances were determined by XRF and Instrumental Neutron Activation Analysis (INAA) at USGS, Denver.
Samples for Sr, Nd, Hf and Pb isotopic analyses were chosen so as to encompass the full range of compositions (Table 5). Two dive samples (T307R1 and R4) were analyzed at Carleton University on a Finnigan MAT261 multicollector mass spectrometer running in static mode. For these two samples, analyses were done on two powder splits. The Sr splits were washed for eight days in hot 6 M HCl, whereas the splits for Nd and Pb were washed in hot 1.5 M HCl overnight. Subsequently they were rinsed twice with ultra-pure H2O before dissolution in HF-HNO3. Information on methods, precision and accuracy for these two samples (T307R1 and R4), and values for standards analyzed at this facility, is given by Cousens et al. .
Hafnium isotopic analyses (and Nd isotopic ratios for acid-leached (in hot 6 M HCl) submarine samples T307R1, R2, R4, R6, R7 and R8) were determined at the Ecole Normale Supérieure in Lyon (ENSL) following the procedure described by Blichert-Toft et al. . Hafnium isotopic compositions were measured by MC-ICP-MS at ENSL using a Nu Plasma 500 HR coupled with a desolvating nebulizer Nu DSN-100. The analytical procedure was similar to that of Blichert-Toft et al.  with the exception that potential W isobaric interferences on mass 180 were monitored on mass 183 instead of mass 182. The Faraday cups were positioned to collect masses 173 (Yb monitor; L3), 175 (Lu monitor; L2), 176 (Hf, Lu, Yb; L1), 177 (Hf; Ax), 178 (Hf; H1), 179 (Hf; H2), 180 (Hf, Ta, W; H3), 181 (Ta monitor; H4) and 183 (W Monitor; H5). On-line mass fractionation-corrected corrections for Yb, Lu, Ta and W isobaric interferences were either zero (Lu and Yb) or zero to negligible (Ta and W). Sixty ratios, in 3 blocks of 20 ratios each, were measured for each sample with an integration time of 10 seconds s per scan. In order to monitor machine performance, the JMC-475 Hf standard was analyzed systematically in alternation with samples and gave 0.282162 ± 0.000012 (2 sigma; n = 28) for 176Hf/177Hf during the three run session of the present samples, corresponding to an external reproducibility of 0.35ɛ (Tables 5 and 6). 176Hf/177Hf was normalized for mass fractionation relative to 176Hf/177Hf = 0.7325 using an exponential law. Hafnium total procedural blanks were less than 20 pg for all sample batches. Uncertainties reported on Hf measured isotope ratios are in-run 2σ/√n analytical errors in last decimal place, where n is the number of measured isotope ratios.
The Nd for the six samples analyzed at ENSL for their Nd isotope compositions was recovered from the CaMg-fluoride precipitates left over from the Hf separation chemistry and purified by a two-step procedure using first a cation-exchange column to separate the REE fraction and then an HDEHP column to isolate Nd. Neodymium isotopic compositions likewise were measured by MC-ICP-MS at ENSL using the Nu Plasma HR coupled with a desolvating nebulizer Nu DSN-100 and the same approach as for the Hf isotopic measurements. The Faraday cups were positioned to collect masses 140 (Ce monitor; L3), 142 (Nd, Ce; L2), 143 (Nd; L1), 144 (Nd, Sm; Ax), 145 (Nd; H1), 146 (Nd; H2), 147 (Sm monitor; H3), 148 (Nd, Sm; H4) and 150 (Nd, Sm; H5). Samarium isobaric interference corrections on Nd were zero. Our in-house JMC Nd standard (batch #801149A) was run in alternation with the samples and gave 0.512126 ± 0.000012 (2 sigma; n = 3) for 143Nd/144Nd during the single short run session of the present samples (Table 5), and corresponds within error to the accepted value of the La Jolla Nd standard of 0.511846 as determined by cross calibration measurements. 143Nd/144Nd was normalized for mass fractionation relative to 146Nd/144Nd = 0.7219 using an exponential law. The Nd total procedural blank was less than 20 pg for the single sample batch analyzed in this study. Uncertainties reported on Nd measured isotope ratios are in-run 2σ analytical errors in last decimal place.
Twelve samples were analyzed for Pb isotopic ratios by the triple-spike technique at the Max-Planck-Institut fur Chemie (MPI) following the procedure of Abouchami et al. . The average ratios measured for NBS-981 by the triple-spike method are 206Pb/204Pb = 16.9447 ± 0.0015, 207Pb/204Pb = 15.5024 ± 0.0015 and 208Pb/204Pb = 36.7350 ± 0.0034 (2 sigma) on the basis of 12 runs during the course of this study. The data reported in Table 5 were normalized to the values of Galer and Abouchami . Data obtained by the triple-spike method and conventional TIMS at MIT (see following text) have comparable machine in-run uncertainties. They agree within 0.1% for 206Pb/204Pb and the data sets are highly correlated (Table 5 and Figure A1a). For 207Pb/204Pb and 208Pb/204Pb both data sets agree within 0.2%, although they are more scattered which is likely due to the larger uncertainties in mass fractionation correction for conventional TIMS data (Figure A1).
Figure A1. Comparison of Pb isotopic ratios analyzed by conventional TIMS and the triple-spike method. The 2 sigma errors shown are the maximum in-run uncertainties which are similar for conventional TIMS and triple-spike data. Both data sets are highly correlated for 206Pb/204Pb, but they are more scattered for 207Pb/204Pb and 208Pb/204Pb. Two samples which were analyzed in duplicate by conventional TIMS (filled symbols) are shown in the figures. The dashed lines connect the duplicates in Figures A1a, A1b, A1c, and A1d. The lines in Figures A1e and A1f link the different data for the same sample. Filled pink squares are TIMS duplicates.
Download figure to PowerPoint
All other isotopic data were obtained at MIT using the following procedures. In order to avoid the effects of postmagmatic alteration on isotopic ratios, a 6.0 M HCl multistep acid leaching procedure was used prior to dissolving in HF-HNO3 [Xu et al., 2005]. After verification of complete dissolution, samples were evaporated and the residual cake was dissolved overnight in 1 mL 6.0 M HCl, and aliquots were taken for Sr (0.25 mL), Pb (0.25 mL) and Nd (0.5 mL) isotope analysis. The aliquots were dried on a hot plate. The Sr aliquot was dissolved using 0.5 mL concentrated 14 M HNO3 and then fluxed on a hot plate for at least one hour before drying. Then 0.5 mL 3.5 M HNO3 was added and digested for 24 hours. The Nd and Pb aliquots were dissolved using 3 mL 1 M HCl and 0.5 mL 1.1 M HBr, respectively, and digested for 24 hours.
Sr was separated from other elements using 50 μL columns filled with Eichrome Sr-Spec resin. The Sr aliquot was loaded onto the column followed by rinsing with 1.2 mL 3.5 M HNO3. Sr was then eluted with 800 μL MQ H2O. A drop of 0.1 M H3PO4 was added to the eluate before drying.
Pb was separated by anion exchange using HBr. The Pb aliquot was loaded onto 120 μL columns containing Eichrom AG1-X8 anion exchange resin. The columns were then washed with 0.5 mL 1.1 M HBr and 0.5 mL 2 M HCl, and Pb was eluted with 1 mL 6 M HCl. A drop of 0.1 M H3PO4 was added to the eluate before drying. For Ca- and Mg-rich samples, this procedure was repeated to improve separation of Pb from Ca.
Separation of Nd and Sm for isotopic analysis requires two ion exchange procedures. The first column separates out the rare earth elements from the bulk rock solution, whereas the second column separates Nd from the other rare earth elements. Rare earth element separation from bulk rock solution utilizes an 8 cm3 column containing Bio-Rad AG® 50W-X8 resin. The Nd aliquot was loaded onto the column and washed by 12 mL 1 M HCl followed by 60 mL 3 M HCl, which removes the bulk of the Fe and Al in the solution, and then washed by 5 mL 3 M HNO3, which removes the bulk of the Ba. The change from HCl to HNO3 also allows some separation of Nd and Sm from Ce and La on the first column. The rare earth element concentrate was eluted with 30 mL 3M HNO3 and dried down. The concentrate was dissolved in 100 μL 0.3 M HCl for the second column, on which Nd was separated from the other rare earth elements using 0.3 and 0.5 M HCl on a 5 cm3 column filled with Eichrome LN-Spec resin.
Sr, Nd and Pb were run on a thermal ionization multicollector mass spectrometer (GV Isoprobe-T) at MIT. Sr was loaded in phosphoric acid on Re filaments with TaCl5 activator, and run in dynamic mode with an average 88Sr signal of 3 V. Pb was loaded on Re filaments with phosphoric acid and silica gel, and run in static mode with an average 208Pb signal of 1–1.5 V. Nd was loaded with phosphoric acid on the Re side filaments of a triple filament assembly, and run in dynamic mode as Nd metal with an average 142Nd signal of 1–1.5 V. The blanks for Sr, Nd and Pb were 300 pg, 100 pg and 10 pg, respectively. See Table 5 footnotes for normalization procedures, precision estimates and data for Sr and Nd standards.
The Pb analyses are corrected for fractionation using the NBS-981 standard. The average ratios measured for NBS-981 are 206Pb/204Pb = 16.896 ± 0.016, 207Pb/204Pb = 15.437 ± 0.022 and 208Pb/204Pb = 36.527 ± 0.070 (2 sigma) on the basis of 61 runs during the course of this study. The MIT isotope laboratory routinely uses a fractionation correction of 0.12 ± 0.03%/amu, based on the values of Todt et al. . Considering the uncertainties of in-run and mass fractionation correction, the 2 sigma reproducibility is better than 0.1% for 206Pb/204Pb. One of the two full procedure duplicates agrees within 0.1% and the other agrees within 0.2% (Table 5); triple-spike data suggest that there is measurable sampling heterogeneity [Abouchami et al., 2000]. The reported data in Table 5 are normalized to the values of Galer and Abouchami  for NBS-981, which require a fractionation correction of 0.135%/amu, and are within the uncertainty of the routinely used fractionation corrections.
A2. Postmagmatic Alteration
Except for historic eruptions Hawaiian subaerial lavas have been affected to variable extents by postmagmatic alteration; the effects of such alteration must be evaluated before using geochemical data to constrain magmatic processes and magma sources. The most common observation is that low-temperature, subaerial alteration in the Hawaiian environment results in loss of K, Rb, Ba and U [e.g., Feigenson et al., 1983; Frey et al., 1994]. We estimated K2O loss by identifying samples with anomalously low K2O/P2O5 (<1.29) (Figure A2); such samples range widely to anomalously high K/Rb (260 to 2669), Ba/Rb (15 to 436 relative to the 11.3 value of most fresh oceanic basalts [Hofmann and White, 1983]), Nb/U (49 to 102) and low Ba/Th (57 to 140) (Figure A3). In our isotopic analyses some samples with anomalous abundance ratios were included, but we used a multistep acid-leaching procedure in an effort to recover the isotopic characteristics of the magmas. The well-defined 87Sr/86Sr versus 143Nd/144Nd correlation (Figure 8a) suggests that the leaching was successful in removing the relatively high 87Sr/86Sr material caused by alteration [e.g., Huang et al., 2005c].
Figure A2. K2O versus P2O5 for West Molokai whole rocks and glasses. All the glasses (green squares) and whole rocks with K2O/P2O5 > 1.29 (filled pink squares and circles) define a strong positive trend, whereas whole rocks with K2O/P2O5 < 1.21 (open squares and circles) are offset to lower K2O at a given P2O5. We inferred that samples with K2O/P2O5 < 1.21 have lost K; these samples were corrected for K loss by adding K2O so that they plot on the regression line derived from the “normal” samples. Data sources: this study, Macdonald and Katsura , and Shinozaki et al. .
Download figure to PowerPoint
Figure A3. K2O/P2O5 versus K/Rb, Ba/Rb, Ba/Th, and Nb/U for West Molokai lavas. Tholeiitic lavas with K2O/P2O5 < 1.3 have experienced loss of K, Rb, U, and Ba (for some lavas) during postmagmatic alteration.
Download figure to PowerPoint
There are, however, other postmagmatic processes; in particular, some Hawaiian lavas have been enriched in REE and Y but typically not in other incompatible elements [e.g., Clague, 1987; Fodor et al., 1989, 1992; Frey et al., 1994]. Although the process is not well-understood, the REE and Y enrichment is apparently caused by formation of a groundmass rhabdophane-type phosphate [Fodor et al., 1989; Cotten et al., 1995]. This process can be recognized by anomalously high La/Nb and low Zr/Y, commonly accompanied by a relative depletion in Ce, which is inferred to reflect an oxidizing environment. Fourteen of 40 West Molokai tholeiitic whole rocks have La/Nb > 1.25 and Zr/Y < 3.5 (Figure A4). The dramatic effect of such alteration is evident on a mantle-normalized plot for incompatible element abundances (Figure A5). Although there is no strong evidence that this REE-Y enrichment affects isotopic ratios [Clague, 1987; Cotten et al., 1995], i.e., the REE-Y mobilization and deposition is highly localized, we did not analyze these West Molokai lavas for isotopic ratios.
Figure A4. Zr/Y versus (a) (La/Nb)PM and (b) Ce/Ce* for West Molokai lavas. Subscript PM designates normalized to primitive mantle value of Sun and McDonough . Ce* is Ce abundance interpolated from primitive mantle normalized abundances of La and Nd. Some West Molokai lavas range to very low Zr/Y (<3.5) and (La/Nb)PM (>1.2) ratios, and these lavas are inferred to reflect a REE-Y enrichment process. Eight out of ten lavas with REE-Y enrichment also have a negative Ce anomaly (see Figure A5). Error bars shown are 2 sigma uncertainties. Data for Hawaiian shield-stage lavas from GEOROC (http://georoc.mpch-mainz.gwdg.de/georoc/).
Download figure to PowerPoint
Figure A5. Incompatible trace element abundances in West Molokai tholeiitic shield lavas normalized to the primitive mantle [Sun and McDonough, 1989]. The gray field shows the range for 17 tholeiitic basalts; only lavas with >4.5% MgO analyzed by ICP-MS are shown. Data points are shown for samples with anomalous ratios involving rare earth elements (REE) and Y that are interpreted as an alteration feature [Clague, 1987]. Sample 74WMOL-6 is the most extreme example, but three additional samples have anomalously high La/Nb, Nd/Sr, Sm/Zr, Eu/Ti, and Y/Zr.
Download figure to PowerPoint