## Abstract

Niobium-92 (^{92}Nb) decays to zirconium-92 (^{92}Zr) with a half-life of 36 million years and can be used to place constraints on the site of *p*-process nucleosynthesis and the timing of early solar system processes. Recent results have suggested that the initial^{92}Nb/^{93}Nb of the solar system was high (>10^{−3}). We report Nb-Zr internal isochrons for the ordinary chondrite Estacado (H6) and a clast of the mesosiderite Vaca Muerta, both of which define an initial^{92}Nb/^{93}Nb ratio of ∼10^{−5}. Therefore, the solar system appears to have started with a ratio of <3 × 10^{−5}, which implies that Earth's initial differentiation need not have been as protracted as recently suggested.

Some extinct radionuclides were sufficiently abundant at the start of the solar system that they produced variations in the abundance of their daughter isotopes in early-formed objects (1). Such nuclides provide information about late-stage presolar nucleosynthetic sites and the time scales over which the early solar system formed and first differentiated (2–4). Of considerable interest in this regard is^{92}Nb, which decays by electron capture with a half-life of 36 ± 3 million years (My) to ^{92}Zr (5, 6).^{92}Nb is a shielded nuclide that forms by the*p*-process only. The *p*-process is a nucleosynthetic process that occurs in supernovae and produces proton-rich nuclides. Therefore, the initial abundance of^{92}Nb provides information on stellar nucleosynthesis before the start of the solar system.

Both Nb and Zr are refractory and lithophile, except under reducing conditions when Nb may become siderophile (7). Within the Earth the dominant Nb/Zr fractionation mechanisms are silicate partial melting (8) and the crystallization of accessory minerals such as zircon, ilmenite, and rutile. The Zr isotopic compositions of early reservoirs can therefore vary in response to early differentiation processes and can help in dating planetary differentiation, if the initial abundance of ^{92}Nb (or^{92}Nb/^{93}Nb) was sufficiently high.

Early results demonstrated that the initial^{92}Nb/^{93}Nb of the solar system was <0.007 (9). Evidence of formerly live ^{92}Nb was first identified in a Nb-rich rutile from the iron meteorite Toluca (10). An initial ^{92}Nb/^{93}Nb of 1.6 (±0.3) × 10^{−5} was inferred. More recently, three studies using a multiple-collector inductively coupled plasma mass spectrometer (MC-ICPMS) proposed that the initial^{92}Nb/^{93}Nb ratio of the solar system was higher by two orders of magnitude (∼10^{−3}) (11–13). Such a high value limits the possible sites for *p*-process nucleosynthesis and places specific constraints on the time scales for the development of differentiated reservoirs on the Earth and Moon. For example, it has been argued that large silicate reservoirs in the Earth and Moon formed >50 My after the start of the solar system (12). However, such a result requires that the Toluca rutile grew very late (>200 My). In contrast, a combined Zr isotopic and U-Pb age study of an early zircon from a eucrite implied that the initial^{92}Nb/^{93}Nb ratio was <10^{−4}(14).

Here, we used the internal isochron approach to determine the initial^{92}Nb/^{93}Nb ratio of the solar system. Two meteorites were studied in which ilmenite with high Nb/Zr is in textural equilibrium with other phases having intermediate Nb/Zr. Because the half-life of ^{92}Nb is long (36 My) and the level of uncertainty concerning the initial solar system abundance is more than two orders of magnitude, the critical concern is not the exact age of the meteorite (provided it is reasonably early) but the acquisition of a reliable isochron from a cogenetic suite of phases that remained undisturbed after their formation. Therefore, we used the equilibrated but only weakly shocked (S1) H6 ordinary chondrite Estacado, which has a single generation of ilmenite, and a eucritic clast from the mesosiderite Vaca Muerta.

Using a Nu Plasma MC-ICPMS, we were able to measure*ɛ*
^{92}Zr with an external precision of ±0.3*ɛ* units (2σ standard deviation) for samples with only 50 ng of Zr (15). The Nb/Zr ratios were determined by inductively coupled plasma dynamic reaction cell mass spectrometry (ICP-DRCMS) without chemical separation (15, 16).

To check for terrestrial Zr isotope anomalies and test our ability to measure the isotopic composition of Zr extracted from different matrices, we analyzed 12 early zircons [aliquots of those described by (17)], two basalts and a rhyolite (from Iceland), a lherzolite (Cameroon Line), and ilmenite (from Russia) (15). The analyzed zircons have ages of up to 4100 Ma (million years ago) (17) and are among the oldest terrestrial minerals. Variations in the Zr isotopic composition of old rocks or minerals are expected only if the initial ^{92}Nb abundance is extremely high. For a solar system initial^{92}Nb/^{93}Nb ratio of ∼10^{−5}, there will be no detectable Zr isotopic effects even in the oldest terrestrial zircons. Earlier studies (12, 14, 18) reached differing conclusions regarding the homogeneity of terrestrial Zr. However, our ^{92}Zr/^{90}Zr,^{91}Zr/^{90}Zr, and^{96}Zr/^{90}Zr ratios for terrestrial samples are identical with the value obtained for a standard solution of National Institute Standards and Technology Standard Reference Material 3169 Zr and do not indicate any variation in the Zr composition of Earth (15). This implies a homogenous distribution of Zr isotopes for the bulk silicate Earth.

The mineral fractions (15) and whole-rock data for Estacado and the Vaca Muerta clast both define isochrons, but with limited variation in ^{92}Zr/^{90}Zr (Fig. 1). The Estacado whole-rock sample and the mineral fractions of olivine/pyroxene and chromite all agree with the terrestrial Zr isotopic composition within analytical uncertainty. However, the ilmenite with a Nb/Zr ratio 20 times that of chondritic (∼0.065) yields an *ɛ*
^{92}Zr of +1.0 ± 0.4. The results for Vaca Muerta are similar (Table 1). The slope of the regression line defines an initial ^{92}Nb/^{93}Nb at the time of closure of 1.2 (±0.6) × 10^{−5} for Estacado (Fig. 1A) and 0.6 (±0.3) × 10^{−5} for Vaca Muerta (Fig. 1B). These values are effectively identical to the Toluca results (10) and are consistent with the zircon data for the Camel Donga eucrite (14).

A high initial ^{92}Nb/^{93}Nb on the order of 10^{−3} was obtained for specific calcium-aluminum-rich inclusions (CAI) from Allende (12, 13). The CAIs are believed to be among the earliest objects that formed in the solar system (19). The production rate of cosmogenic^{92}Zr is too small to generate significant variations in the calculated initial ^{92}Nb/^{93}Nb ratios (20). Both ^{92}Zr and ^{96}Zr isotope anomalies have been reported for CAIs (13). Therefore, the possibility remains that CAIs preserve a component with a presolar Zr isotope signature, analogous to those reported for presolar graphite grains (21). A heterogeneous distribution of^{92}Nb in the early solar system could also cause different estimates of the initial abundance of ^{92}Nb. This would require a variability in the initial^{92}Nb/^{93}Nb between 10^{−3} and 10^{−5} and should produce measurable ^{92}Zr anomalies in whole-rock meteorites. However, the observed^{92}Zr isotopic variations in bulk chondrites (13) have not been yet confirmed (12, 22). This indicates that further work is needed to confirm and understand these reported^{92}Zr anomalies.

One way to reconcile the various results is resetting or slow cooling (>200 My) of the samples yielding low values: Toluca, Vaca Muerta, Camel Donga, and Estacado. This is inconsistent with a considerable amount of independent age data. Estacado (H6) yields a K-Ar age (4450 Ma) only 30 My younger than Kernouvé (another H6 chondrite) (23), which is likely to have experienced the same events as all H6 chondrites. The Pb-Pb age of phosphates extracted from Kernouvé (H6) and Guarena (H6) indicate that these meteorites underwent a metamorphic event or cooling around 45 My and 61 My, respectively, after the formation of the Allende CAIs (19). The H6 chondrites are, therefore, early objects and their equilibrated textures cannot be of late origin. The Sm-Nd chronology of Vaca Muerta clasts (24) and the U-Th-Pb systematics of Vaca Muerta zircons (25) yield similar ages of 4480 ± 190 Ma and 4563 ± 15 Ma, respectively.

To provide more direct time constraints, we have acquired Hf-W isotope data for some of the Estacado and Vaca Muerta mineral separates (15). ^{182}Hf (half-life = 9 My) decays to^{182}W such that W isotopic variations can only be produced within the first 50 My of the solar system. Small excesses of^{182}W have previously been detected in Vaca Muerta zircons (26). We find large W isotopic variations among the aliquots that we have analyzed for Zr (15). Sulfide equilibrates readily, yet the two troilite fractions of Estacado display negative*ɛ*W values of –2.46 ± 0.47 and –2.05 ± 0.56, indicating that the latest time of sulfide/silicate equilibration was 9 ± 3 My and 11 ± 4 My, respectively, after the origin of the solar system, assuming that the initial^{182}Hf/^{180}Hf was 2.4 × 10^{−4}. The W isotopic compositions and Hf/W ratios for the Vaca Muerta pyroxene and feldspar separates (15) define a^{182}Hf/^{180}Hf of 7.68 (±0.8) × 10^{−5}. Therefore, these phases appear to have last equilibrated 14.8 ± 1.3 My after the start of the solar system. Even if this error is increased somewhat to allow for systematic Hf-W uncertainties, there is little doubt that Estacado and Vaca Muerta last underwent internal W isotope equilibration within the first 20 My of solar system history. Therefore, the apparent discrepancy with other recently published Nb-Zr data cannot be explained easily by later equilibration. The Estacado initial ^{92}Nb/^{93}Nb of 1.2 (±0.6) × 10^{−5} should be very close to the solar system initial abundance within a factor of 2 (Fig. 2). The slightly lower initial ^{92}Nb/^{93}Nb of Vaca Muerta is consistent with the Hf-W evidence for a later closure of the phases in the mesosiderite relative to Estacado. However, both initial^{92}Nb/^{93}Nb values are identical within the analytical errors, consistent with independent evidence for early planetesimal differentiation (27, 28). With a low initial^{92}Nb/^{93}Nb ratio of ∼10^{−5}, the timing of terrestrial core formation, the growth rate of Hadean continents, and the longevity of the lunar magma ocean (12, 29,30) can now be considered to be underconstrained from Nb-Zr systematics (Fig. 2).

The low initial abundance of ^{92}Nb also reopens the range of possibilities that can be considered for *p-*process nucleosynthesis. As a *p-*only nuclide, ^{92}Nb is predominantly produced in supernovae by photodisintegration (31). The modeling of other processes such as spallation synthesis (10, 31, 32) shows that these can be only a minor contributor in the synthesis of *p-*nuclei because they fail to explain the solar abundance pattern. The calculated production ratios for ^{92}Nb/^{93}Nb, as predicted by several models for Type Ia and Type II supernovae (33–36), are in the range of 2.1 × 10^{−3} to 9.2 × 10^{−3}. Only one Type II model (11, 37) yields a higher production ratio of 0.35, and is able to explain the high abundance of stable*p*-process ^{92}Mo in the solar system. However, it is inconsistent with the data presented here.

The range 2.1 × 10^{−3} to 9.2 × 10^{−3}, in contrast, is in good agreement with our results. To deduce the abundance of ^{92}Nb at the time of the formation of the solar system from this range, it is important to consider two factors: (i) the free decay time interval between the last nucleosynthesis and the formation of the solar system, and (ii) the amount of ^{92}Nb at the start of this interval, which is a function of the duration of nucleosynthesis. The 36-My half-life of^{92}Nb enables the use of an averaging model (38), which includes continuous nucleosynthesis at a constant rate (39,40). Assuming a 10,000-My period of nucleosynthesis, such a model predicts that 0.5% of the total ^{92}Nb produced is present at the end of nucleosynthesis. This factor applied to the calculated production ratios (33–36) yields an initial^{92}Nb/^{93}Nb for the solar system of 1 × 10^{−5} to 4.6 × 10^{−5}, if the free decay interval is not considered. Our estimate based on Estacado falls into this range. Taking into account the free decay interval [up to 50 My (10)] may further lower the calculated value by a factor of 2 or more. Nonetheless, most production ratio yields of different supernova models (type Ia and type II) are quite similar to our measured value of 1.2 (±0.6) × 10^{−5}. Thus, our data provide only limited constraints on the type of supernova that produced the ^{92}Nb initially present in the solar system.

↵* To whom correspondence should be addressed. E-mail: maria{at}erdw.ethz.ch