Global isotopic signatures of oceanic island basalts

Abstract

Sr, Nd and Pb isotopic analyses of 477 samples representing 30 islands or island groups, 3 seamounts or seamount chains, 2 oceanic ridges and 1 oceanic plateau [for a total of 36 geographic features] are compiled to form a comprehensive oceanic island basalt [OIB] data set. These samples are supplemented by 90 selected mid-ocean ridge basalt [MORB] samples to give adequate representation to MORB as an oceanic basaJt end-member. This comprehensive data set is used to infer information about the Earth's mantle. Principal component analysis of the OIB+MORB data set shows that the first three principal components account for 97.5% of the variance of the data. Thus. only four mantle end-member components [EMI, EMII, HIMU and DMM] are required to completely encompass the range of known isotopic values. Each sample is expressed in terms of percentages of the four mantle components, assuming linear mixing. There is significant correlation between location and isotopic signature within geographic features, but not between them, so discrimination analysis of the viability of separating the oceanic islands into those lying inside and outside Hart's (1984, 1988) DUPAL belt is performed on the feature level and yields positive results. A "continuous layer model" is applied to the mantle component percentage data to solve for the spherical harmonic coefficients using approximation methods. Only the degrees 0-5 coefficients can be solved for since there are only 36 features. The EMI and HIMU percentage data sets must be filtered to avoid aliasing. Due to the nature of the data, the coefficients must be solved for using singular value decomposition [SVD], versus the least squares method. The F-test provides an objective way to estimate the number of singular values to retain when solving with SVD. With respect to the behavior of geophysics control data sets, only the degree 2 spherical harmonic coefficients for the mantle components can be estimated with a reasonable level of confidence with this method. Applying a "delta-function model" removes the problem of aliasing and simplifies the spherical harmonic coefficient solutions from integration on the globe to summation over the geographic features due to the properties of deltafunctions. With respect to the behavior of geophysics control data sets, at least the degree 2 spherical harmonic coefficients for the mantle components can be estimated with confidence, if not the degrees 3 and 4 as well. Delta-function model solutions are, to some extent, controlled by the nonuniform feature distribution, while the continuous layer model solutions are not. The mantle component amplitude spectra, for both models, show power at all degrees, with no one degree dominating. The DUPAL components [EMI, EMil and HIMU], for both models, correlate well with the degree 2 geoid, indicating a deep origin for the components since the degrees 2-3 geoid is in ferred to result from topography at the core-mantle boundary. The DUPAL and DMM components, for both models, correlate well [and negati vely! at degree 3 with the velocity anomalies of the Clayton-Comer seismic tomography model in the 2500-2900 km depth range [immediately above the core-mantle boundary]. The EMil component correlates well [and positively] at degree 5 with the velocity anomalies of the Clayton-Comer model in the 700-1200 km depth range, indicating a subduction related origin. Similar positive correlations for the geoid in the upper lower mantle indicate that subducted slabs extend beyond the 670 km seismic discontinuity and support a whole-mantle convection model.