Length scales of mantle heterogeneities and their relationship to ocean island basalt geochemistryINTRODUCTIONCONSTRAINTS FROM OSMIUM ISOTOPES OF OIBPERIDOTITE-PYROXENITE INTERACTIONS IN THE MANTLESubsolidus InteractionsInteractions in Basalt Source RegionsSummaryQUANTITATIVE MODELS OF DIFFUSION AND MELT SEGREGATIONDiffusive Equilibration Within Two-Layer Composite MediaMelt Segregation by CompactionCompetition Between Diffusive Equilibration and Melt SegregationThreshold Porosity for Melt InterconnectionChief Parameters Affecting Diffusive EquilibrationChief Parameters Affecting CompactionRESULTS AND DISCUSSIONSurvival of Pyroxenite Signature During Partial MeltingSurvival of Heterogeneous Bodies Before MeltingFate of Melts Expelled From Pyroxenite DomainsSUMMARY AND CONCLUDING REMARKSREFERENCESdoi:10.1016/S0016-7037(03)00419-8Length scales of mantle heterogeneities and their relationship to ocean island basaltgeochemistryTETSU KOGISO,1,2,*MARC M. HIRSCHMANN,1and PETER W. REINERS31Department of Geology and Geophysics, University of Minnesota, Minneapolis, MN 55455, USA2Institute for Frontier Research on Earth Evolution, Japan Marine Science and Technology Center, Yokosuka, Kanagawa 237-0061, Japan3Department of Geology and Geophysics, Yale University, New Haven, CT 06520, USA(Received December 17, 2002; accepted in revised form June 9, 2003)Abstract—The upper mantle is widely considered to be heterogeneous, possibly comprising a “marble-cake”mixture of heterogeneous domains in a relatively well-mixed matrix. The extent to which such domains arecapable of producing and expelling melts with characteristic geochemical signatures upon partial melting,rather than equilibrating diffusively with surrounding peridotite, is a critical question for the origin of oceanisland basalts (OIB) and mantle heterogeneity, but is poorly constrained. Central to this problem is thecharacteristic length scale of heterogeneous domains. If radiogenic osmium signatures in OIB are derived fromdiscrete domains, then sub-linear correlations between Os isotopes and other geochemical indices, suggestingmelt-melt mixing, may be used to constrain the length scales of these domains. These constraints arise becausepartial melts of geochemically distinct domains must segregate from their sources without significantequilibration with surrounding peridotite. Segregation of partial melts from such domains in upwelling mantleis promoted by compaction of the domain mineral matrix, and must occur faster than diffusive equilibrationbetween the domain and its surroundings. Our calculations show that the diffusive equilibration time dependson the ratios of partition and diffusion coefficients of the partial melt and surrounding peridotite. Comparisonof time scales between diffusion and melt segregation shows that segregation is more rapid than diffusiveequilibration for Os, Sr, Pb, and Nd isotopes if the body widths are greater than tens of centimeter to severalmeters, depending on the aspect ratio of the bodies, on the melt fraction at which melt becomes interconnectedin the bodies, and on the diffusivity in the solid. However, because Fe-Mg exchange occurs significantly morerapidly than equilibration of these isotopes under solid-state and partially molten conditions, it is possible thatsome domains can produce melts with Fe/Mg ratios reflecting that of the surrounding mantle but retainingisotopic signatures of heterogeneous domains. Although more refined estimates on the rates of, and controlson, Os mobility are needed, our preliminary analysis shows that heterogeneous domains large enough toremain compositionally distinct in the mantle (as solids) for ⬃109yr in a marble-cake mantle, can produce andexpel partial melts faster than they equilibrate with surrounding peridotite. Copyright © 2004 Elsevier Ltd1. INTRODUCTIONMantle heterogeneity is well established from geochemicaland geophysical studies, but the distribution, scale, magnitude,and origin of mantle heterogeneities remain controversial(Brandon et al., 1998; Kamber and Collerson, 1999; Normanand Garcia, 1999; Rudnick et al., 2000). An essential constrainton the origin and consequences of mantle heterogeneity is thescale of heterogeneous domains. Geophysical observations in-dicate that heterogeneity may be present on the scale of 104mor greater, (e.g., Ishii and Tromp, 1999; Kaneshima and Helf-frich, 1999; Zhao, 2001) but do not currently allow resolutionof finer scale features. Regional geochemical variations of OIB,however, require mantle heterogeneity on a 103-m scale (Gastet al., 1964; Zindler and Hart, 1986) and some geochemically-based models of basalt source regions call on heterogeneitiesranging from decimeter-scale veins found in orogenic perido-tite massifs (Alle`gre and Turcotte, 1986; Reisberg et al., 1991)(but see also Blichert-Toft et al., 1999) to kilometer-scalebodies of subducted oceanic crust (Hauri, 1996; Hanyu andKaneoka, 1997). Presumably, mantle heterogeneities have arange of scales, but quantitative constraints on their size dis-tributions in basalt source regions is important because theirdimensions may strongly affect the dynamics of melting andthe geochemistry of integrated melts. Knowledge of their sizesmay also have implications for the possible origins and historyof such domains.Hofmann and Hart (1978) showed that centimeter-sized het-erogeneities can remain compositionally distinct in the solidmantle for billions of years, but that partial melting mighthomogenize even much larger bodies (⬃20 m) over time scalesas short as 105to 106yr. Consequently, they argued that smallcompositional heterogeneities can be maintained over the life-time of the Earth, but are erased by diffusion in partiallymelting basalt source regions. In recent years, evidence forsmall-scale heterogeneities in basalt source regions hasmounted, most notably from trace element and isotopic studiesof melt inclusions in phenocrysts from basalts (Saal et al.,1998; Sobolev et al., 2000), and apparent isotopic discrepanciesbetween oceanic basalts and their inferred abyssal peridotitesources (Salters and Dick, 2002). Partly in response to theselines of evidence, models invoking decimeter-scale mafic veinsto explain features of basalt geochemistry have proliferated. Atthe same time, modern models for melt extraction and U-seriesisotopes of oceanic basalts both indicate that melt extractiontime scales are 103yr or less, rather than the 105to 106yr* Author to whom