142Nd Evidence for Early(94.53 Ga) Global Differentiationof the Silicate EarthM. Boyet*and R. W. CarlsonNew high-precision samarium-neodymium isotopic data for chondriticmeteorites show that their142Nd/144Nd ratio is 20 parts per million lowerthan that of most terrestrial rocks. This difference indicates that most (70 to95%) of Earth’s mantle is compositionally similar to the incompatibleelement–depleted source of mid-ocean ridge basalts, possibly as a result ofa global differentiation 4.53 billion years ago (Ga), within 30 million years ofEarth’s formation. The complementary enriched reservoir has never beensampled and is probably located at the base of the mantle. These data in-fluence models of Earth’s compositional structure and require revision of thetiming of global differentiation on Earth’s Moon and Mars.Radiogenic isotope tracers are often used tounderstand the chemical evolution of planetarybodies. In one system, Sm decays to Nd viatwo radioactive decay schemes:146Sm-142NdEhalf-life T1/20 103 million years (My)^ and147Sm-143Nd ET1/20 106 billion years (Gy)^).Both Sm and Nd are refractory lithophile (pre-fer silicates over metal) elements, whose rel-ative abundances should not be affected byeither volatile loss or core formation. The long-lived147Sm-143Nd system has been widelyused to trace planetary-scale processes such asthe evolution of the bulk silicate Earth (BSE,defined as all the Earth except for its metalliccore) and its chemical differentiation into crustand mantle over Earth_s history. The early epicof Earth_s differentiation is better investigatedwith the short-lived chronometer146Sm-142Nd,because of the lack of available samples fromEarth_s first 500 My of existence and the sen-sitivity of most long-lived radioactive systemsto resetting by events occurring later in Earth_shistory. Recent studies have reported small142Nd/144Nd excesses in a small number ofsamples from 3.8 Ga from the Isua SupracrustalBelt, Greenland (1, 2), which is the firstevidence for differentiation of the silicateportion of Earth that must have occurred closein time to the well-documented early differen-tiation of the Moon (3)andMars(4, 5). Thestandard assumption in using this radiometricsystem to model the geochemical evolution ofEarth is that the BSE has Sm/Nd,143Nd/144Nd,and142Nd/144Nd ratios approximately equal tothose of chondritic meteorites, the buildingblocks of the planet. Chondritic meteoritesshow a relatively limited range in Sm/Nd(147Sm/144Nd from 0.1932 to 0.2000), withcorresponding143Nd/144Nd from 0.512525 to0.512722, when fractionation is corrected to146Nd/144Nd 0 0.7219 (6–8), and the averagechondritic values of 0.1966 and 0.512638 (6)have been used to represent the BSE model val-ues for 25 years. Here we present Sm-Nd datafrom chondrites that overlap previous Sm/Ndand143Nd/144Nd ratio determinations but indi-cate a measurable difference of142Nd/144Ndratios between chondrites and all terrestrialsamples. If the BSE has an Sm/Nd ratio withinthe range measured for chondrites, the higher-than-chondritic142Nd/144Nd ratio of terrestrialmaterials requires that the silicate Earth experi-enced a global chemical differentiation duringthe lifetime of146Sm, resulting in high and com-plementary low Sm/Nd ratio reservoirs that haveremained separate over all of Earth_shistory.Terrestrial Sm-Nd evolution. The mostactive volcanic system on Earth, the globalocean ridge system, erupts magmas with143Nd/144Nd ratios considerably higher thanany value measured for bulk chondritic mete-orites. This value reflects the high Sm/Nd ratioand the general depletion of the oceanic mantlein those elements that selectively partition intomelts (the so-called incompatible elements), be-cause Nd is more incompatible than Sm. Theincompatible element depletion of the mid-oceanridge basalt (MORB) source generally is mod-eled as resulting from the extraction of the in-compatible element–rich continental crust overEarth’s history. To explain the high Sm/Nd in theMORB source purely by continental crust ex-traction from a mantle that initially had chon-dritic Sm/Nd, only roughly one-third to one-halfof the mantle can be as incompatible element–depleted as the MORB source (9–11). The restof the mantle has typically been assumed to haveremained primitive, with a chondritic Sm/Ndratio. However, some observations are difficultto reconcile with this model (12, 13). For exam-ple, recent seismic imaging of the mantle sug-gests that convective motion may stir at leastthe upper three-quarters of the mantle (14), ifnot most of it. Also, the oldest crustal rocks onEarth (dating to 4 Ga) have superchondritic143Nd/144Nd ratios, requiring an episode of evenearlier crustal extraction, although only volu-metrically insignificant amounts of this pre–4-Ga crust survive at Earth’s surface today(15, 16). Other evidence that the incompatibleelement depletion that now characterizes theMORB source mantle is an old feature, not re-lated to the formation of the current continentalcrust, is the presence of excess129Xe (derivedfrom the decay of 17-My–half-life129I) andPu-fission Xe in the MORB, which indicatethat the source must have been outgassed within50 My of Earth’s formation (17, 18).RESEARCH ARTICLESDepartment of Terrestrial Magnetism, Carnegie Insti-tution of Washington, 5241 Broad Branch Road,N.W., Washington, DC 20015, USA.*To whom correspondence should be addressed.E-mail: [email protected]. 1.142Nd/144Nd ratios measured for chon-drites and eucrites compared to the La Jollaterrestrial Nd standard (e142Nd). All chondritesand basaltic eucrites have negative e142Ndvalues outside the external analytical error ofT 0.07 e units (2s) (shaded area). Cumulateeucrites have positive e142Nd values in agree-ment with their high Sm/Nd, resulting fromigneous processes on their parent body. Theerror bars correspond to the internal precision(2smean). Terrestrial samples (MORBs, kimber-lites, and carbonatites of different ages andcollected in diverse locations) measured usingthe same procedure (27) have been added todemonstrate the significant excess of 0.2 eunits in all the terrestrial material (samples andstandard) relative to the mean chondriticvalue. All terrestrial samples were measuredseveral times using the same procedures aswere used for the chondrites. The uncertaintiesreported on the mean are 2s.22 JULY 2005 VOL 309 SCIENCE www.sciencemag.org576Chondritic142Nd/144Nd. Because of thesmall variations