Workers have proposed that population peaks in the U-Pb zircon record which correspond to intervals of supercontinent assembly are an artifact of preservation9,31,32. The preservation potential for early subduction-related (early assembly) and later extensional-related (breakup) magmatism is conceptually lower than for collision-related magmatism (i.e. late assembly). Roberts (2012)4 suggested a similar phenomenon affects the Hf record. However, here we propose that the global Hf record is less sensitive to selective preservation, and thus consideration of such a preservational phenomenon is arguably less crucial for interpretation of Hf isotope datasets.
The zircon U-Pb record is based on the frequency of observations. Yet, selective preservation may erode the signal by reducing the number of analyses at a given age. In comparison, the Hf record is arguably less sensitive to the count of preserved analyses, and, depending on the statistical interrogation of the Hf dataset, may be more sensitive to their collective excursion within Hf isotope space. The pattern of low preservation potential affects those parts of a supercontinent cycle with a dominance of juvenile material. During these intervals, selective removal of some of the juvenile analyses through erosion may move the average of that group towards more evolved values. However due to the distribution of data, this process is unlikely to move the average to the type of values seen in those parts of the cycle with a dominance of evolved material (which also have the best preservation potential). Hence, the net result of any selective analysis removal from the dataset may be to dilute the juvenile part of the signal towards a global crustal balance at the median, but to still maintain a pattern of juvenile and evolved excursions through time (Fig. 1B). In addition, we find there is no correlation between the number of analyses per bin, and ΔDM95, which is consistent with a lack of significant preservation bias in the global Hf signal (Fig. 4).
Assembly of the Supercontinents
The sensitivity of the three ΔDM (50, 95 and 99) signals to the assembly of each of the supercontinents is shown in Fig. 5. Final assembly of Columbia between 1900–1780 Ma is arguably best recorded both by the juvenile 95% and by the median signals, since they both show excursions towards more explicitly evolved values with progression of assembly (Fig. 5G). Immediately prior to Columbia assembly, the Hf signals show a pronounced negative excursion between 2070–1970 Ma, suggestive of an extended period of continental reworking. This evolved excursion may support the existence of a Palaeo to Meso-Proterozoic supercontinent (e.g., Sclavia and/or Superia)33. The positive excursion that follows at 1900 Ma may reflect its breakup, as suggested by a peak in the abundance of passive margins27.
Inspection suggests differences in the median ΔDM50 and juvenile (ΔDM95 and ΔDM99) records with respect to the assembly of both Pangaea and Rodinia (Fig. 5A,C). Specifically, the median signal appears insensitive to the final assembly of Pangaea (300–250 Ma), in that it does not record a shift to evolved values. This observation has previously been interpreted as reflecting a balance between juvenile and evolved material14. However, our analysis suggests that final Pangaea assembly is recorded in both the juvenile Hf signals through the magnitude of their excursion towards more explicitly evolved values, especially through the 95% signal (Fig. 5A). Compared to Pangaea, Rodinia has a much more protracted interval of assembly (250 versus 50 Myr). The juvenile 95% Hf signal traces two pronounced excursions towards more evolved values, a pattern not observed in the median signal (Fig. 5E).
The Case of Gondwana
We observe a trend in the gross deviation of Hf isotopic excursions with successive supercontinent cycles (i.e. that recorded by ΔDM’). This trend is suggestive of a signal increasingly biased towards more radiogenic and evolved values over time until and including Gondwana assembly, which of all the supercontinents exhibits the most evolved Hf signal (Fig. 2B). The observed differences between Rodinia, Gondwana and Pangaea assembly may be informed through understanding the relationship between the juvenile and median signals, as shown in Fig. 5(B,D,F,H). The arithmetic difference between the juvenile and median part of the signal at a point in time represents the spread of data points toward more evolved values, and is a measure of the heterogeneity of global magma volumes at any period in time. This is best described through the de-trended difference between the 95% and 50% Hf binned ratios (Fig. 5B,D,F,H).
During final Gondwana assembly, there is an accord between all three ΔDM signals (Fig. 5C). However, there is a positive spread between the juvenile ΔDM95 and median signal (Fig. 5D), in contrast to that seen for the assembly of Pangaea (Fig. 5B). Murphy and Nance34 suggested that observed differences in Nd model ages between Pangaea and Gondwana, reflected different modes of assembly: extroversion (Gondwana) versus introversion (Pangaea). Extroversion describes a supercontinent that reassembles through consumption and reworking of older oceanic material, and is consistent with the spread in Hf data for Gondwana which indicates the availability of more evolved material over global melt volumes.
Gondwana is the supercontinent that exhibits the most evolved signal during assembly. Differences in the Hf record associated with Gondwana final assembly in comparison with that of Rodinia have been proposed as reflecting extensive reworking of Palaeoproterozoic and older crust11, specifically within the Pan-African orogeny35. While our data are consistent with this model, an alternate explanation for the significantly more evolved Gondwana assembly signal may be that it reflects a secular change in subduction style since the breakup of Rodinia. Oceanic crust formed since the Neoproterozoic is some 2–3 times thinner than that formed in the Mesoproterozoic and older36. Further, there is a suggestion that plates grew stepwise to a pentagonal configuration controlled by geoid highs37, widening the ocean basins. The combined effect of this process is colder, thinner, and less buoyant oceanic crust entering subduction zones during Gondwana assembly, leading to steeper, and colder, subduction, marked by the appearance of lawsonite eclogites in the rock record at this time38. This increase in subduction angle means enhanced subduction-erosion, ultimately leading to greater crustal recycling (Fig. 6). Furthermore, it is possible that steeper subduction may favour the process of cratonic lithospheric delamination, i.e. the loss of its cratonic root, such as that experienced by the Saharan metacraton during the Pan African Orogeny39. Such lithospheric delamination may then enhance reworking of the older cratonic crust, which in turn may accentuate the predominance of evolved Hf values as observed for final assembly of Gondwana. We thus suggest an increase in crustal recycling associated with steeper subduction may be observed in the pronounced Gondwana Hf excursion.
Pangaea would also experience this more modern-style steep subduction geometry, and the primacy of subduction-erosion as a crustal reworking process over new crust addition via arc magmatism in the Phaneorozic has been proposed40. However, the data show that the final assembly of Pangaea exhibits a relatively minor evolved Hf excursion compared to both Gondwana and Rodinia, and there are two possible contributions to this. Firstly, its relatively minor evolved Hf excursion may reflect its final assembly by dominantly introversion processes34. Secondly, the Central Asian Orogenic Belt (CAOB), the most major orogenic event during the Phanerozoic41, has been estimated to have an unusually high proportion of juvenile addition. Greater than 50% of the 500–100 Ma magmatism associated with the CAOB comprises a juvenile component of over 60%42, and this may contribute to an overall less evolved Hf signal for Pangaea assembly.
The periodic assembly and dispersal of continental fragments, referred to as the supercontinent cycle, bear close relation to the evolution of mantle convection and plate tectonics. Supercontinent formation involves complex processes of “introversion” (closure of interior oceans), “extroversion” (closure of exterior oceans), or a combination of these processes in uniting dispersed continental fragments. Recent developments in numerical modeling and advancements in computation techniques enable us to simulate Earth's mantle convection with drifting continents under realistic convection vigor and rheology in Earth-like geometry (i.e., 3D spherical-shell). We report a numerical simulation of 3D mantle convection, incorporating drifting deformable continents, to evaluate supercontinent processes in a realistic mantle convection regime. Our results show that supercontinents are assembled by a combination of introversion and extroversion processes. Small-scale thermal heterogeneity dominates deep mantle convection during the supercontinent cycle, although large-scale upwelling plumes intermittently originate under the drifting continents and/or the supercontinent.