Evolution of the Continents and the Atmosphere Inferred from Th-U-Nb Systematics of the Depleted Mantle

See allHide authors and affiliations

Science  05 Mar 1999:
Vol. 283, Issue 5407, pp. 1519-1522
DOI: 10.1126/science.283.5407.1519


Temporal evolution of depleted mantle thorium-uranium-niobium systematics constrain the amount of continental crust present through Earth's history (through the niobium/thorium ratio) and date formation of a globally oxidizing atmosphere and hydrosphere at approximately 2.0 billion years ago (through the niobium/uranium ratio). Increase in the niobium/thorium ratio shows involvement of hydrated lithosphere in differentiation of Earth since approximately 3.8 billion years ago. After approximately 2.0 billion years ago, the decreasing mantle thorium/uranium ratio portrays mainly preferential recycling of uranium in an oxidizing atmosphere and hydrosphere. Net growth rate of continental crust has varied over time, and continents are still growing today.

The nature and time scales of differentiation of Earth's silicate portion into depleted mantle and continental crust, as well as the effects of a changing atmosphere on this process, remain uncertain (1). Continental crust is highly enriched in many trace elements. In general, the degree of enrichment follows the order of compatibility (2); the most incompatible elements are most strongly enriched in the continents. There are a few prominent exceptions to this rule (Pb, Ti, Nb, and Ta) and we show that the temporal changes in Th, U, and Nb distribution between the continental crust and the depleted mantle (that portion of the mantle from which the continents have been extracted) can be used to reconstruct important aspects of Earth's evolution.

Thorium, U, and Nb behave as strongly incompatible elements during melting of the mantle at midocean ridges (2) and have similar bulk distribution coefficients (3). Large differences in ratios of Th, U, and Nb between the depleted mantle and the continental crust are therefore not expected. However, continental crust of all ages has a deficit in Nb relative to U and Th, expressed as Nb/U and Nb/Th ratios of <10 and <5, respectively (4). These ratios are much lower than those of the undifferentiated, primitive mantle (30 ± 3 and 8 ± 1) (5). Depleted mantle has a complementary overabundance of Nb leading to a Nb/U ratio of 44.5 ± 2.5 and a Nb/Th ratio of 18.5 ± 1.2 (5). If the temporal evolution of Nb/U and Nb/Th ratios in depleted mantle can be established (1, 5), the different behavior of U and Th versus Nb during formation of continental crust can be exploited to infer the mass of continental crust present through geological time.

There is, however, a serious complication in this elegant concept: the present-day Th/U ratio of the depleted mantle is ∼2.6 (6), which is significantly lower than the time-integrated Th/U value of ∼3.75 (6) estimated from the Pb isotope composition of midocean-ridge basalts (MORBs). This observation is termed the second terrestrial Pb paradox (6). It could either be the result of a temporal change in the Th/U ratio of depleted mantle or a result of recycling of continental Pb (derived from crust with a mean Th/U ratio of ∼4.2) back into the mantle. Implication of the first possibility is that the evolution of Nb/U and Nb/Th in depleted mantle would have differed and would therefore not provide an unequivocal estimate of the mass of continental crust present through time.

We estimated the temporal evolution of these ratios, using suites of rocks inferred to be derived from the depleted mantle and ranging in age from 3.8 billion years (Ga) to the present day (Table 1) (7). When Th/U ratios of depleted mantle–derived samples are plotted against age (Fig. 1A), the data define a smoothly decaying curve that can be described by a second-order polynomial fitEmbedded Imagewith r 2 = 0.989 and aget in millions of years (Ma). The calculated present-day (t = 0) and chondritic values (t = 4550 Ma) are 2.55 and 4.18, respectively, in excellent agreement with independently constrained estimates of 2.5 to 2.7 and 3.91 to 4.28 (8, 9). The mean time-integrated depleted mantle Th/U ratio obtained from the second-order polynomial fit to the compiled data (Fig. 1A) is 3.4. This value is lower than the Th/U ratio based on the Pb isotope composition of MORB, ∼3.75 (6). Solution to the second Pb paradox thus involves a mantle with a time-integrated Th/U of 3.4 (as determined above), plus recycling of some continental Pb into the mantle (9). In theory, Th and U could have been significantly fractionated during formation of continental crust, in which case the Th/U ratio would portray crustal volume through time. If this were the case, however, the mean age of the resulting continental crust would be younger than 2.0 to 2.2 Ga (10). Therefore, evolution of Th/U in the depleted mantle must reflect an additional process. Recycling of continental crust has been suggested as the most likely reason for decoupling of U from Th (11). Staudigel et al. (12) proposed that U would be preferentially recycled into the mantle, but only when the atmosphere and hydrosphere were sufficiently oxidizing to render it mobile. Under present-day oxidizing conditions, Th/U ratios of river waters range between 0.05 and 0.1. This reflects the fact that U is more readily leached from weathered continental crust than is Th, and is mainly transported in dissolved form (13). By contrast, Th enters the estuaries as suspended particle load and as a result, the Th/U ratio of seawater is even lower than that of river water (13). Under an oxidizing atmosphere U is preferentially transported into the open ocean where it is incorporated into marine sediments and altered oceanic crust, which are eventually subducted.

Figure 1

Evolution of Th, U, and Nb ratios of ancient depleted-mantle–derived rocks from compilation in Table 1. (A) Curve for Th/U ratios versus age. Filled circles are data points from Table 1, which were used to obtain a second-order polynomial fit (given in text). Open circles represent target values, which were not included for fit calculation (present-day E-MORB = 2.44 ± 0.16, chondrite average = 4.1 ± 0.1). (B) Curve for Nb/Th ratios versus age. All data points are connected by a smoothed interpolation (smoothing factor of 50%). (C) Curve for Nb/U ratios versus age. All data points are connected by a smoothed interpolation (smoothing factor 60%). Note the dissimilarity in the behavior of Th and U relative to Nb.

Table 1

Compilation of Th, U, and Nb data for depleted mantle–derived rocks (error = 1 SD). ɛNdexpresses the fractional difference between the initial143Nd/144Nd ratio of the rocks and the corresponding value of this ratio in the chondritic uniform reservoir at the time of crystallization of the rocks, expressed in units of 104. Fm., formation.

View this table:

In the mantle samples, the temporal evolution of Nb/Th and Nb/U ratios differ (Fig. 1, B and C). The Nb/Th ratio generally increased with time, slowly between the present day and ∼2 Ga, strongly between 2 and 3.5 Ga, and very slowly during earliest Earth history. This contrasts with the Nb/U ratio, which increased slowly during the early and middle Archaean and then strongly during late Archaean and early Proterozoic to a maximum value of ∼58 at 2.0 Ga. The Nb/U ratio then decreased to the present-day value of 44.5.

Thus, the Nb/U and Nb/Th ratios in depleted mantle both evolved to significantly higher values than those of primitive mantle. There are two plausible mechanisms capable of efficiently fractionating Nb from U and Th. (i) The presence of rutile, a stable phase in the eclogite assemblage formed by subduction of oceanic crust, significantly changes the bulk distribution coefficient for Nb (14) because Nb substitutes for Ti. Thus, melt in equilibrium with rutile has a Nb deficit relative to U and Th. (ii) Dehydration of the subducting slab leads to depletion of the oceanic crust in some elements (for example U, Th, and Rb), which are transported to the melting environment of the overlying mantle wedge. The degree of depletion depends on the distribution coefficients between the eclogite assemblage and fluid (not melt). Uranium and Th are significantly more mobile than Nb, and thus the fluid will be characterized by low Nb/U and Nb/Th ratios (14). This indicates that growth of continental crust must have, at least in part, involved subduction zone–induced melting.

The Nb/U and Nb/Th curves can be converted to curves of crust volume versus age (CvA curves) by assuming that there was 0% crust present for chondritic values and 100% crust present for the present-day MORB values. However, these two proxies yield very different topologies (Fig. 2A). The CvA curve based on Nb/Th ratio agrees well with independent CvA estimates (Fig. 2B): the post-Archaean CvA curve of Reymer and Schubert (15), which is based on geophysical considerations, and the CvA curve of Kramers and Tolstikhin (9), which satisfies both terrestrial Pb paradoxes as well as the Nd isotope database (16). We therefore combined these three proxies to yield a new composite CvA curve (solid line in Fig. 2B).

Figure 2

Curves for crust volume versus age. (A) Comparison of crust volumes estimated through the Nb/Th ratio (filled circles, connected by solid line) and through the Nb/U ratio (open squares, connected by dotted line). Estimation is based on the assumption that the chondritic values represent 0% crust and that present-day MORB average represents 100% of present continental crust volume. Note that the Nb/U-based estimate predicts the former existence of a substantially more voluminous continental crust. (B) Comparison of crust volumes estimated from geophysical data by Reymer and Schubert (15) (solid diamonds), transport-forward modeling by Kramers and Tolstikhin (9) (open squares), and the Nb/Th evolution of the depleted mantle in (A) (solid circles). Solid line represents composite parameterization for the crust volume versus age (eight data points from Reymer and Schubert, 12 data points from Kramers and Tolstikhin, and all 10 Nb/Th data points). Relative to the present-day volume or mass of continental crust, the percent volume or mass of continental crust (y) present at any given age (t, in megannum) can be expressed as a seven-order polynomial: y = 100 – 1.5742 × 10–2 × t – 4.2617 × 10–5 × t 2 + 9.9543 × 10–8 × t 3 – 7.663 × 10–11 × t 4 + 2.5738 × 10–14 × t 5 – 3.9626 × 10–18 × t 6 + 2.3005 × 10–22 × t 7 withr 2 = 0.984.

In contrast, if the Nb/U ratio is converted into a CvA curve (Fig. 2A), using the approach described above, it predicts that late Archaean to early Proterozoic continental crust was volumetrically similar (1), if not more voluminous than continental crust today (17). This hypothesis of constant, or even decreasing, crustal volume since the late Archaean is not supported by other constraints: (i) it is not recorded in the estimated Nd isotope evolution of depleted mantle (16); (ii) it fails to account for the observed crustal age distribution (10); and (iii) the required high recycling rates (comparable to, or greater than, estimated production rates of ∼1 km3a–1) would erase the future Pb paradox (9, 11).

The Nb/U ratio of the depleted mantle therefore not only reflects the volume of continental crust, it is also influenced by the preferential recycling of U (over Th and Nb). The kink in the Nb/U evolution (Fig. 1C) dates the time of formation of an atmosphere and hydrosphere that were sufficiently oxidizing to allow U transport in the form of UO2 2+ (11, 12). This interpretation is supported by the observation that the Th/U ratio in cratonic shales increased from ∼3.5 in the Archaean to ∼4.2 in the Proterozoic and ∼4.6 in the Phanerozoic in spite of a relatively constant upper-crustal Th/U ratio through time (4).

The evolution of Th/U in depleted mantle can be quantitatively assessed using the constraints derived from Nb/Th and Nb/U systematics. The evolution of the Th/U ratio between ∼3.8 and 2.0 Ga is interpreted to be solely a result of extraction of continental crust. The strong decrease in Th/U ratio after 2.0 Ga is interpreted to result from preferential U recycling, superimposed on extraction of more continental crust. From our parameterization of the CvA curve, a hypothetical Th/U evolution of the depleted mantle can be calculated by equating the evolution from a chondritic value of 4.18 with 0% continental crust and the Th/U ratio at 2.0 Ga of 3.64 with ∼67% continental crust present (Fig. 3). This model represents the Th/U evolution of depleted mantle in a permanently anoxic Earth. The observed Th/U ratio and the predicted ratio, based on crust volume alone, show agreement between 4.55 and 2.0 Ga; however, the curves deviate between 2 Ga and the present day. If the hypothetical Th/U curve is calculated with an anchor point outside 1.8 to 2.2 Ga, the fit worsens rapidly in the section before 2.0 Ga. We therefore conclude that this age range (1.8 to 2.2 Ga) reflects timing of development of a pandemic oxidizing atmosphere. This is supported by the recent discovery of laterites 2.0 to 2.2 Ga in age and by cessation of deposition of banded-iron formation (18).

Figure 3

Comparison of the evolution of the observed Th/U ratio (open squares, approximated by a second-order polynomial fit; Fig. 1) with a hypothetical Th/U ratio in an anoxic Earth (solid circles, where no preferential U recycling was possible). The hypothetical evolution is calculated assuming that the decrease of the Th/U ratio in the depleted mantle is purely due to extraction of continental crust (discussed in text).

This model can be independently tested with Pb isotope systematics of upper continental crust. To satisfy mass-balance considerations (using a chondrite ratio of 4.1 and a present-day depleted-mantle ratio of 2.55), the mean Th/U ratio of present-day continental crust must be ∼4.5. The hypothetical Th/U evolution of an anoxic Earth predicts that the mean continental crustal Th/U ratio was lower (<4.25) between 4.55 and 2.0 Ga. The inescapable implication (which is independent of exact Th/U ratios) is that the Th/U ratio of the continental crust has increased between 2.0 Ga and the present day. Our model is strengthened by the observation that post-Archaean samples of upper crust show the predicted increase in initial208Pb/206Pb isotope ratios (9).

The observed present-day depleted mantle is enriched in U by ∼30% compared to the hypothetical mantle in an anoxic Earth (with no preferential U recycling; Fig. 3). The U required for this surplus corresponds to only ∼6% of that hosted in continental crust. The amount of U eventually recycled into the depleted mantle therefore only represents a small fraction of the total U budget that entered the oceans (and eventually marine sediments and hydrated oceanic lithosphere). Calculation of a reliable crustal recycling rate from the difference in U concentration between the observed and the modeled depleted mantle is not possible, because U is highly mobile and incompatible during subduction zone dehydration (14). Nevertheless, an important feature of the observed Th/U evolution of the depleted mantle (Fig. 3) is that continental recycling increased a few hundred million years after the onset of formation of a globally oxidizing atmosphere. This pattern indicates that an increase in the level of free oxygen in the atmosphere accelerates the rate of continental recycling (erosion rates). Although the cause of this phenomenon is unknown, it is possible that a change in subaerial biological activity could have played a major role in enhancing continental weathering (16, 18).

Continental growth rate has varied substantially over geological time. Strong net growth is recorded between 3.0 and 2.0 Ga, but generally slowed down after 2.0 Ga due to increased erosion. Superimposed on this pattern is the effect of Proterozoic supercontinent assembly (Nena, Rodinia, Gondwana) resulting in small total continental circumference, and hence, decreased total subduction melt flux at that time. Renewed increase of net crustal growth from ∼250 Ma to the present day indicates faster continental growth during times of continental dispersal.

  • * To whom correspondence should be addressed. E-mail: k.collerson{at}


View Abstract

Navigate This Article