5.1. Superchondritic 142Nd/144Nd in the Mantle Sampled by OIBs and MORBs
 Together with previously published results [Boyet and Carlson, 2006; Caro et al., 2006; Andreasen et al., 2008; Murphy et al., 2010], our results show that all MORB and OIB lavas–from 5 different mid-ocean ridge locations and 9 different hot spots–host142Nd/144Nd that is identical to the terrestrial standard (Figure 1 and Table 2). For several different reservoirs to evolve simultaneously in this brief time interval (<30 Ma) to have the exact same Sm/Nd (6% higher than chondrite) and 142Nd/144Nd (18 ± 5 ppm higher than chondrite) would be fortuitous. Therefore, we suggest that all measured modern terrestrial mantle reservoirs that contribute to OIB and MORB volcanism descend from a single early formed reservoir with a superchondritic Sm/Nd ratio (Figure 2).
Figure 2. 143Nd/144Nd evolution of a chondritic and non-chondritic Earth. Both models are consistent with the observation that all accessible modern, mantle-derived terrestrial samples have142Nd/144Nd 18 ± 5 ppm higher than chondrites, and were derived from an early formed (<30 Ma after accretion) precursor reservoir with Sm/Nd ∼6% higher than chondrites. This precursor reservoir has a present-day143Nd/144Nd of ∼0.5130 (ε143Nd = 7.2 [Jackson and Carlson, 2011]) and is (bottom) an EDR (early depleted reservoir) if BSE is chondritic (which implies the existence of an EER), or (top) a non-chondritic BSE. HIMU, EM1, EM2, DMM and most of the continental crust (excluding only a few, rare Archaean/Hadean terranes) were extracted from either the EDR (if the BSE is chondritic) or the BSE (if the BSE is not chondritic). In the chondritic Earth case (bottom), terrestrial samples withε143Nd < 0 (143Nd/144Nd < 0.51263) are considered enriched relative to BSE, and are considered depleted if ε143Nd > 0 (143Nd/144Nd > 0.51263); in the non-chondritic Earth case (top), terrestrial samples withε143Nd < 7.2 (143Nd/144Nd < 0.5130) are considered enriched, and are considered depleted if ε143Nd > 7.2 (143Nd/144Nd > 0.5130). The choice of reference frame is important, as the bulk of OIB 143Nd/144Nd measurements plot within a range of 143Nd/144Nd ratios of 0.51263 and 0.5130. The ε143Nd of the highest 3He/4He Baffin Island lavas is also indicated [Stuart et al., 2003; Starkey et al., 2009], and coincides with the value of the EDR or the non-chondritic BSE. A ∼2 Ga formation age of the EM and HIMU reservoirs is chosen to be consistent with the secondary isochron age of the mantle array in207Pb/204Pb vs 206Pb/204Pb isotopic space [e.g., Zindler and Hart, 1986]. A 3 Ga formation age of DMM is chosen to be consistent with Workman and Hart . The figure is adapted from Jackson et al. [2007b]. ε143Nd = (143Nd/144Ndsample/143Nd/144NdCHUR − 1) × 104, where 143Nd/144NdCHUR = 0.51263 [Bouvier et al., 2008].
Download figure to PowerPoint
 The origin of this progenitor reservoir is still unknown. The progenitor reservoir is either a non-chondritic BSE or the high Sm/Nd “depleted” reservoir (EDR) produced in an early differentiation event where an early enriched reservoir (EER) was extracted from the mantle [Boyet and Carlson, 2005; Caro et al., 2008; O'Neill and Palme, 2008; Caro and Bourdon, 2010; Jackson et al., 2010]. The non-chondritic BSE and the EDR are geochemically identical entities, characterized by the same superchondritic Sm/Nd (6% higher than chondrites),143Nd/144Nd (0.5130) and 142Nd/144Nd (18 ± 5 ppm higher than chondritic). An EER is only needed if the BSE has chondritic relative abundances of refractory lithophile elements (Figure 2). This study and previous studies have identified no geochemical evidence that any surface rock has been derived from the EER [e.g., Boyet and Carlson, 2006; Caro et al., 2006; Andreasen et al., 2008; Murphy et al., 2010]. Either the EER has been convectively isolated from participating in surface magmatism throughout Earth history, or there is no EER and the BSE does not have chondritic relative abundances of the refractory lithophile elements.
 Two explanations that have been proposed for a BSE deficient in the more incompatible of the refractory lithophile elements are impact erosion of planetesimal crusts prior to accretion to Earth [O'Neill and Palme, 2008; Caro et al., 2008] or volcanic ejection of partial melts from small planetesimals [Warren, 2008]. In modeling the trace element composition of the EDR/non-chondritic Earth [e.g.,Boyet and Carlson, 2005; Jackson et al., 2010; Jackson and Carlson, 2011] an important feature is the relative smoothness of the chondrite-normalized patterns when plotting the refractory incompatible elements according to their incompatibility during the relatively shallow mantle melting involved in the production of MORB or OIB [e.g.,Hofmann, 1988]. The shape of the incompatible element pattern of the EDR/nonchondritic BSE thus is most consistent with the loss of a partial melt generated at relatively low pressure [Boyet and Carlson, 2005] such as might occur on small planetesimals prior to their accretion to Earth. Alternatively, if this differentiation occurred on Earth, then a possible explanation of the EDR trace element pattern would be removal of an early formed terrestrial crust either through impact erosion or its subduction and permanent storage in the deep mantle. The timing of these early events—collision erosion or subduction of early crust—is not well known, but is likely to have happened well before 30 Ma, as suggested by recent modeling efforts [Korenaga, 2009] and the recent revision of the 146Sm half-life from 103 Ma to 68 Ma [Kinoshita et al., 2012].
 An important consequence of a non-chondritic BSE model is as follows: Because all modern terrestrial mantle reservoirs were derived from this non-chondritic BSE, a non-chondritic BSE becomes the standard for determining whether a reservoir is considered “enriched” (<0.5130) or “depleted” (>0.5130). For comparison, in a chondritic world, all reservoirs with143Nd/144Nd > 0.51263 [Bouvier et al., 2008] are considered depleted, and reservoirs with 143Nd/144Nd < 0.51263 are enriched. Therefore, it is the range of 143Nd/144Nd ratios from 0.51263 to 0.5130 that is affected by a change of reference frame: considered “depleted” in a non-chondritic world, lavas with143Nd/144Nd ratios from 0.51263 to 0.5130 would be considered “enriched” in the non-chondritic BSE model. The importance of this subtle difference is illustrated with HIMU lavas. In the non-chondritic reference-frame, extreme HIMU lavas from Mangaia with143Nd/144Nd of 0.51286 are considered enriched, not depleted. The non-chondritic BSE model suggests a history of incompatible element enrichment for the HIMU reservoir, instead of a history of incompatible element depletion. If HIMU forms from recycled oceanic crust [e.g.,Hofmann and White, 1982], then the history of isotopic enrichment in HIMU, as implied by a non-chondritic BSE, is consistent with the suggested origin of HIMU.
 Indeed, with the exception of lavas from Koolau, lavas erupted from the highest-flux mantle plume, Hawaii, generally have geochemically depleted143Nd/144Nd ratios (>0.51263) relative to a chondritic Earth. This implies a history of depletion in the mantle source of Hawaiian lavas. However, if the Earth is not chondritic, the bulk of Hawaiian shield lavas are actually enriched (143Nd/144Nd < 0.5130 [see Jackson and Carlson, 2011]), which implies a history of geochemical enrichment for this plume relative to the bulk composition of the silicate Earth. Enrichment of the Hawaiian plume source is also consistent with its Pb isotopic composition that plots to the right of the Pb geochron, whereas a depleted source would be expected to plot to the left of the geochron. Therefore, the non-chondritic reference frame provides a view that is more consistent with the standard model for the origin of mantle plumes, which maintains that plumes are buoyantly upwelling regions ofenriched crustal material that was subducted into the mantle in the geologic past [Hofmann and White, 1982; White and Hofmann, 1982]. Indeed, most of the global OIB database has 143Nd/144Nd that lies in the range of 0.51263 to 0.5130 (the median lies near 0.5130 [Zindler and Hart, 1986; Jackson and Carlson, 2011]), suggesting that, in a non-chondritic reference frame, most plume-derived lavas sampled geochemically enriched material relative to the BSE. By contrast, in a chondritic world, approximately 90% of OIB lavas are considered depleted (143Nd/144Nd > 0.51263) relative to BSE, which is not consistent with the hypothesis that plumes originate from recycling of ancient subducted enriched crustal materials.
 Whether any of the EDR (or non-chondritic BSE) has survived differentiation or mixing with recycled crust over Earth history to survive in the modern mantle is not clear. The143Nd/144Nd predicted for the progenitor reservoir (0.5130) is similar to that measured in lavas with the highest terrestrial mantle 3He/4He ratios [Boyet and Carlson, 2006; Caro et al., 2008; Caro and Bourdon, 2010; Jackson et al., 2010]. The high 3He/4He mantle reservoir, including lavas with the highest 3He/4He (50 Ra) from Baffin Island [de Leeuw et al., 2010], have 142Nd/144Nd identical to all other modern terrestrial lavas and 18 ± 5 ppm higher than chondrites. The highest 3He/4He Baffin Island lavas also have Pb-isotopic compositions that lie near the geochron, which is consistent with an ancient origin for this mantle reservoir [Jackson et al., 2010].
5.2. Is the Hidden Early Enriched Reservoir Simply “Hidden” in the Measurement Precision?
 We cannot exclude the possibility that a hidden EER exists in the deep mantle, but if it exists, it is not efficiently entrained in mantle plumes and is thus not sampled at Earth's surface. Entrainment of basal mantle layers by plumes is the subject of significant discussion, and Bourdon and Caro  review the limits on the amount of a basal layer than can be entrained. Here, we place limits on how much EER can be entrained without measurably modifying the Nd isotopic composition of the EDR. We model the minimum amount of EER mantle (that contributes to a melt) necessary to generate a clearly resolved 142Nd/144Nd anomaly relative to the terrestrial standard (i.e., −10 ppm relative to the terrestrial standard). We assume an EER composition from Carlson and Boyet  (142Nd/144Nd ranges from −38 to −54 ppm relative to the terrestrial standard, with Nd concentrations ranging from 2.2 to 8.9 ppm, respectively), and we assume that the EER mantle was diluted with mantle material with 1.0 ppm Nd (similar to the EDR composition of Carlson and Boyet ) and 142Nd/144Nd identical to the terrestrial standard. This scenario requires that approximately 2–14% of the Nd in the mantle plume is from the EER to generate a 142Nd/144Nd anomaly of −10 ppm. A similar calculation, with similar results, notes that with one exception [Upadhyay et al., 2009], there has been no variation in 142Nd/144Nd in mantle-derived rocks since 3.5 Ga [e.g.,Bennett et al., 2007]. Using the same EER and EDR concentrations mentioned above, the lack of a secular trend in 142Nd/144Nd in mantle-derived rocks since 3.5 Ga allows no more than the same 2 to 14 weight % entrainment of the EER into the EDR. The calculation shows that, given current measurement precision, 2 to 14 weight % of an EER may be entrained in plumes or into the EDR without generating a measurable difference142Nd/144Nd. This neither supports nor denies the presence on an EER, but does point out the relative insensitivity of the Nd system in resolving the presence of geochemical reservoirs left over from early Earth differentiation because of the limited fractionation of Sm and Nd and the low initial abundance of 146Sm.