Main

Crustal differentiation processes produce a large range of Rb/Sr ratios, because of the different partitioning characteristics of Rb and Sr within the crust (DRb < DSr ≪ 1). This results in a marked increase in Rb/Sr with increasing SiO2 in crustal rocks (Fig. 1). 87Rb decays to 87Sr with a long half-life (∼48.8 Gyr) compared with the age of Earth, and so source rocks with different 87Rb/86Sr will develop a range of 87Sr/86Sr isotope ratios along linear evolution paths with time (Fig. 1 inset). Thus, the ‘time-integrated’ 87Rb/86Sr ratio can be estimated from changes in Sr isotopes with time, and used as a proxy for the bulk composition of the newly formed continental crust. In contrast, other isotope systems classically used in crustal evolution studies, such as Lu–Hf or Sm–Nd, have more restricted parent/daughter ratios in differentiated magmatic rocks. For example, 176Lu/177Hf ratios typically range between ∼0.009 for felsic upper crust and ∼0.022 for mafic lower crust, in contrast to 87Rb/86Sr that ranges from 0.74 to 0.087 in felsic to mafic crust.

Figure 1: Correlation between Rb/Sr and SiO2 in crustal rocks, from a compilation of 96,465 magmatic rocks.
figure 1

Dots (vertical error bars indicate 2 s.e.m.) represent median Rb/Sr values calculated for each 1% SiO2 interval. The best fit to the data is represented by the orange curve. Inset: Two-stage model of evolution for Sr isotopes used to back-calculate the 87Rb/86Sr of the juvenile continental crust (that is, during Stage 1), from Sr–Nd isotopes in the crust sampled today. Purple arrows indicate how the back-calculation was done. Examples for the evolution of a mafic source (green path) and a more felsic source (red path) are presented. Data from GEOROC (http://georoc.mpch-mainz.gwdg.de/georoc).

Full size image

Our approach uses a two-stage model to calculate the 87Rb/86Sr ratio of the juvenile continental crust (Fig. 1 inset). Stage 1 is the period of Sr isotope evolution of this juvenile crust, from its extraction from a depleted mantle reservoir until the formation of a derivative crustal melt. Stage 2 is the period from the crystallization of those melts (typically with higher 87Rb/86Sr ratios) until the present day. The following assumptions were used in the calculation of the Rb/Sr ratios in Stage 1: Nd model ages (TDM) for given crustal samples are a reasonable estimate of the time since new continental crust separated from the mantle; the juvenile continental crust had a Sr isotope composition similar to that of a depleted mantle reservoir (Fig. 1, brown curve) at the time of its formation. The purple arrows with associated yellow numbers (Fig. 1 inset) indicate how the ‘juvenile crust’ (Stage 1) 87Rb/86Sr were back-calculated: the 87Sr/86Sr of the rock is measured (hexagon); the 87Sr/86Sr ratio at the time of crystallization is calculated (star); the Stage 1 87Rb/86Sr is calculated from the slope of the segment between the 87Sr/86Sr ratio of the mantle at the time of crust formation (brown dot) and the Sr isotope ratio of the crustal melt at the time it crystallized (stars). Examples for a mafic source (87Rb/86Sr = 0.087, green path) and a more felsic source (87Rb/86Sr = 0.442, red path) are presented, and a step-by-step method for the calculation of Rb/Sr ratios in Stage 1, along with parameters and references used in those calculations, is presented in the Supplementary Information.

Stage 1 Rb/Sr ratios of crustal source rocks for over 13,000 volcanic and plutonic rocks with crust formation ages ranging from the Hadean to the Phanerozoic are summarized in Fig. 2. As a result of the increased uncertainty in the calculation of Rb/Sr in Stage 1 for rocks with a difference between their Nd model age (TDM) and crystallization age (CA) of less than 300 Ma, only rocks with TDM − CA > 300 Ma were selected (see Supplementary Information). Before ∼3 billion years ago (Ga) the Rb/Sr of the bulk juvenile continental crust remained around 0.03 and since then the Rb/Sr ratios progressively increased to a maximum value of ∼0.08 in the Mesoproterozoic era, whereupon they decreased slightly towards the present day. The composition of juvenile continental crust has become more differentiated with time and, using the relationship in Fig. 1, the calculated Rb/Sr ratios can be converted into estimated average SiO2 contents (Fig. 2, right y axis). These range from ∼48–50% SiO2 before 3 Ga up to a value of ∼57% SiO2 in the Mesoproterozoic.

Figure 2: Variation of Rb/Sr ratios in juvenile crust (Stage 1 in Fig. 1) as a function its formation age.
figure 2

Based on a compilation of 13,125 analyses (see Supplementary Dataset 1). Trends in the evolution of Rb/Sr data were determined by employing maximum-likelihood statistics: a gamma distribution was fitted to the histograms of Rb/Sr within sliding windows of 50 Ma widths, and to account for uncertainties in the time axis we used an overlap of 6 windows. The advantage of fitting a probability density function to the data is that we directly obtain a best model fit with associated confidence levels (95% and 68%) for each window. The model SiO2 on the right y axis is calculated from the Rb/Sr versus SiO2 relationship in Fig. 1.

Full size image

These results highlight that: the composition of the juvenile continental crust was predominantly mafic before 3 Ga (Rb/Sr ∼ 0.03 and SiO2 ∼ 48–50%); there was little fractionation of Rb/Sr between new continental crust and the mantle before 3 Ga; and ∼3 Ga marks the onset of a gradual change in the composition of juvenile continental crust towards more intermediate compositions. The mafic composition of new continental crust in the Archaean eon has been inferred from estimated crustal source Lu/Hf ratios in regional studies2,7. It is consistent with recent models that suggest that continents were mostly mafic in composition8 and below sea level6,9 during this period, and the estimated SiO2 values are similar to those of the low-SiO2 node of the bimodal distribution of SiO2 that characterizes much of the exposed Archaean crust10. This indicates that our approach yields realistic Rb/Sr ratios and SiO2 contents. Finally, the low Rb/Sr ratios of juvenile crust before ∼3 Ga are consistent with recent studies that suggest that the upper mantle was not depleted throughout the Hadean and early Archaean2,11.

A key feature of the results summarized in Fig. 2 is the shift to higher Rb/Sr and, by implication, to higher SiO2 at ∼3 Ga. Primitive arc magmas have low Rb/Sr and SiO2 contents12 and both increase with increasing crustal thickness as illustrated with data from central and south America, which is taken as a representative ‘modern’ site for the generation of new continental crust (Fig. 3). In areas where the crust is 20–60 km thick, there is a positive correlation between the Rb/Sr of the average crust and the crustal thickness (Fig. 3), with Rb/Sr ratios increasing from ∼0.04 at ∼20 km to ∼0.15 at ∼60 km. A sharp increase is observed for crust >60 km, which has Rb/Sr ∼ 0.25, and this is attributed to enhanced crustal melting. A positive correlation is also observed for SiO2 (Fig. 3, inset), although it is best observed for 25–55-km-thick crust as thicker crust (55–70 km) has restricted SiO2 around 60–61%.

Figure 3: Correlation between Rb/Sr (and SiO2, inset) and crustal thickness in central and south western America.
figure 3

The plots are based on a compilation of 7,916 magmatic rocks (see Supplementary Dataset 2), and the crustal thickness is calculated from the latitude and longitude coordinates of the samples and the Crust 1.0 model (http://igppweb.ucsd.edu/g̃abi/crust1.html). Red curves are the maximum probability function of the data (using the method described in Fig. 2, with sliding windows of 10 km), and blue curves are the running median. Both statistical methods give similar results, supporting the reliability of our approach.

Full size image

The increase in Rb/Sr and SiO2 with crustal thickness could be due to smaller degrees of melting under thicker lithospheric lids13; however, the increase in Rb/Sr tends to be associated with a decrease in Sr/Nd typically attributed to residual plagioclase (Supplementary Fig. 1). Thus, the increase in Rb/Sr (and SiO2) seems to be linked to increased differentiation in areas of thicker continental crust14. In detail, the rate of increase of Rb/Sr with crustal thickness may have been different earlier in Earth history. Nonetheless, these data indicate that the progressive increase in the juvenile continental crust Rb/Sr ratios (and the extrapolated SiO2) from 3 Ga to 1 Ga (Fig. 2) may reflect at least the relative thickening of the continental crust through time.

Using the relationships in Figs 2 and 3, the juvenile crust Rb/Sr ratios were converted into average thickness of the juvenile crust (Fig. 4, purple curve). The estimated average thickness of new continental crust increases from ∼20 km at 3 Ga to ∼40 km by 1 Ga, and then decreases to ∼30 km towards the present. The relationship based on subduction-related magmas (Fig. 3) may not be relevant for crust formed in other geodynamical settings (Fig. 4, purple dashed curve before 3 Ga), but a number of studies have inferred that the crust was thinner when the mantle temperatures were higher in the early Archaean15,16,17. The data also reaffirm that what is meant here by ‘juvenile or ‘new’ continental crust is the composition of mantle-derived magma that crystallized in the crust after differentiation, rather than the composition of magmas that initially cross the Moho18. The implication is that the volume of residual material returned back into the mantle through delamination also increases with crustal thickness, perhaps because garnet is more readily developed resulting in an increase in density19.

Figure 4: Variation in the thickness of juvenile continental crust through time, calculated from the relationships in Figs 2 and 3.
figure 4

Timing for the onset of plate tectonics is from ref. 5.

Full size image

The time period around 3 Ga is interpreted as a key transition period in Earth’s evolution. Before this period, new continental crust was dominantly mafic and is inferred to have mostly formed in pre-plate tectonic settings20. The onset of plate tectonics and the development of subduction zones around 3 Ga (ref. 5) seem to have led to the development of thicker continental crust (Fig. 4). The high volume of continental crust established by 3 Ga (that is, 60–70%; refs 21, 22) in turn implies that large volumes of crust have been recycled back into the mantle since the onset of subduction. High rates of recycling from ∼3 Ga may reflect the more mafic composition, and greater density of the crust generated in the Archaean. There is a striking increase in both the estimated crustal thickness and the global increase of the sedimentary component in the magmatic record reflected in higher δ18O isotopes since the late Archaean22,23. The inferred increase in crustal thickness would have been associated with greater plate strength and the development of higher-relief mountain belts that would have resulted in increased erosion/weathering and the recycling of the sedimentary material in subduction zones24,25,26. Enhanced erosion/weathering is also implied by the increase in Sr isotope ratios in sea water from 3 to 2 Ga (ref. 27), and invoked for variations in the sulphur28 and the phosphorus29,30 fluxes in the ocean.

From ∼1 Ga ago, the Rb/Sr of the juvenile continental crust, and therefore its average thickness, seem to have decreased towards the present day (Figs 2 and 4). These data suggest that the crust may have reached its maximum thickness at the time of the assembly of the Rodinia supercontinent and the Grenville orogeny. Assuming that crustal thickness is a proxy for crustal volume, the inferred reduction in crustal thickness since ∼1 Ga indicates that for the first time in the history of the Earth crust destruction rates may have exceeded the rates at which new crust was generated26.

Methods

Code availability.

The code used to generate Fig. 2 can be accessed in the Supplementary Information.