Figure Captions (Hofmann, CIDER 2003)

Fig. 1
Sr and Nd isotopes of Mid-Ocean Ridge Basalts (MORB) and marine sediments. The slope of the oceanic basalt array is consistent with the model of Armstrong postulating extensive recycling of continental material into the mantle through sediment subduction. However, this hypothesis is not consistent with the lead isotope data (Fig. 2).

Fig. 2.
Lead isotopes of MORB, OIB, average continental crust and average subducted sediments (GLOSS, . The reference isochron of 1.8 Ga through the oceanic basalt data indicates that the array is dominated by mantle differentiation less than 2 Ga old. If it were dominated by recycled continental material, which has a mean age of at least 2.5 Ga, the array would have to have a much steeper slope.
This diagram is also an illustration of Allègre’s (1969) “lead paradox”: The isotopic compositions of all sampled reservoirs lie overwhelmingly on the right-hand side of the 4.53 Ga geochron. These data therefore require the existence of a hidden reservoir with lead isotopes to the left of the geochron in order to balance the reservoirs represented by the data from the continental and oceanic crust. This reservoir might conceivably lie in the poorly sampled lower continental crust.

Fig. 3.
Examples of primitive-mantle normalized trace element abundance diagrams (“spidergrams”) for representative samples of HIMU (Mangaia, Austral Islands, sample M-11; Woodhead 1996), EM-1 (Pitcairn Seamount sample 49DS1; Eisele et al. 2002), EM-2 (Tahaa, Society Islands, sample 73-190 (White and Duncan, 1996); Average Mauna Loa (Hawaii) tholeiite (Hofmann, unpublished data), average continental crust , average subducting sediment, GLOSS , and Average Normal MORB (Su and Langmuir, 2002). Th, U and Pb values for MORB were calculated from average Nb/U = 47, and Nd/Pb = 26). All abundances are normalized to primitive-mantle values of McDonough and Sun, (1995).
Note the positive peaks for lead, and negative peaks for niobium in average continental crust and sediments. These are complementary to opposite peaks in almost all oceanic basalts. This indicates that, in general, OIBs sources share the general geochemical characteristics of MORB sources, and that they cannot contain large amounts of recycled continental material. However, the smaller magnitude of these peaks in EM-type basalts could be explained by relatively small additions of recycled sediments to EM sources.
Figs. 4, 5, and 6.
Isotope ratios of ocean island basalts (OIBs), taken from Hofmann, 1997, Nature 385, 219-229. The acronym HIMU (high mu) refers to the high U/Pb source ratios required to produce the highly radiogenic lead found in these basalts. EM-1 and EM-2 refer to “enriched mantle” 1 and 2. These isotopic flavors are color coded to show how they translate from Sr-Nd to Pb-Pb isotope diagrams. “Enriched” refers to the more highly incompatible elements and specifically means high Rb/Sr and high Nd/Sm (= low Sm/Nd). The acronyms EM-1, EM-2, and HIMU were introduced by Zindler and Hart (Ann. Rev. Earth Planet. Sci. 14, 493-571, 1986)

Fig. 7 (taken from Fig. 3 of Saal et al., 1998, Science 282, 1481-1484). Lead isotope data for melt inclusions from olivine crystals in two basalt samples from the island of Mangaia (Austral Islands), the most extreme representative of the HIMU-type of oceanic basalts. Mangaia (whole-rock) basalt samples have a restricted range of highly radiogenic Pb isotope ratios. However, the melt inclusions in these samples have a very much wider range of Pb isotopic compositions, showing that the whole-rock basalts are mixtures of much more heterogeneous primary melts. This is evidence that the length scale of isotopic and chemical heterogeneities in the mantle is much smaller that previously suspected on the basis of whole-rock isotope geochemistry. However, we do not know the specific geometry of melt extraction systems, so we cannot infer the precise length scale of heterogeneities, but we may guess that it is greater than 10 cm and smaller than 10 km.

Fig. 8. Nb/U versus 87Sr/86Sr in basalts from the Society Islands, a typical “EM-2” hotspot (White & Duncan, in Basu and Hart, editors, “Earth Processes: Reading the isotopic code”, AGU Monogr. 95, 183-206 (1996). Nb and U have very similar bulk partition coefficients during oceanic mantle melting, so that the Nb/U ratio can be used as a tracer for source compositions. Normal oceanic basalts (both MORB and OIB ) have Nb/U = 47 ± 10, but island arc basalts, continental crust and continent-derived sediments have much lower Nb/U ratios (<10), as well as much higher 87Sr/86Sr ratios. Therefore the correlation of these parameters can be used to identify the presence of sediments (or similar continental components) within hot spot basalts. This seems clearly to be the case here, but it is not automatically clear whether these sediments are integral parts of the mantle plume, or whether they were located in the upper mantle or crust and were essentially “contaminants” in the primary, mantle derived melts.
A very similar but positive correlation is found for Pb/Ce or Pb/Nd ratios versus 87Sr/86Sr. High Pb concentrations are another tracer of continent-derived material, so this correlation is consistent with the sediment-recycling hypothesis. Workman and Hart (ms submitted) propose an alternative interpretation involving two stages, (1) impregnation of oceanic lithosphere by low-degree melts, (2) subducting and aging this lithosphere and (3) recycling this into the Society Isl. mantle plume. This model clearly needs further evaluation.

Fig. 9. Photomicrograph of glassy melt inclusions in olivine crystal (“phenocrysts”) from a Hawaiian (Mauna Loa) lava (from Sobolev et al., 2000, Nature, 404, 986-990). This olivine formed in a deep crustal magma chamber (pressure of 2 to 6 kb), and it incorporated primary, mantle derived melts of highly diverse compositions in these melt inclusions. Trace element abundance patterns are schematically indicated (without labeling the individual elements for reasons of clarity. For more details of these, see fig. 10). Inclusions 76-A, B, and E have patterns similar to ordinary Mauna Loa basalt. 76-C is strongly depleted in highly incompatible elements on the left hand side. 76-B is highly unusual because of its excess Sr (positive spike). It was incorporated earlier (at higher pressure) into the growing crystal than the other inclusions. Among 160 more or less normal inclusions, we found 6 such exotic, Sr-rich inclusions, the chemistry of which is shown and interpreted in Fig. 10.

Fig. 10. Average composition of the six, exotic, Sr-rich melt inclusions in olivine phenocrysts from Mauna Loa volcano. (Sobolev et al., 2000, Nature, 404, 986-990).
These trace element patterns show anomalously high Sr and low Th abundances. The pattern mimics the patterns of typical gabbros from ophiolites (Oman and Gabal Gerf) thought to be similar to gabbros forming the lower half of the oceanic crust. The characteristic trace element features of gabbros are caused by cumulus plagioclase (see the lowermost, “Pl” pattern). However, the included melts were not formed in actual equilibrium with gabbro but with eclogite, the high-pressure form of gabbro. Also, they are isotopically different form present-day oceanic crust in the Pacific Ocean. Therefore, the gabbros must be ancient, and they must have been transformed into eclogite by subduction. In other words, they were recycled through the source of the Hawaiian mantle plume.