Calcium isotope compositions as a means to trace carbonate recycling.

Marine carbonate, an important CO2 reservoir, is continuously sent to the Earth's deep interior at subduction zones, forming an essential part of the global carbon cycle. The pros and cons of using calcium isotope compositions to trace marine carbonates recycled into the mantle are discussed in this Perspective.

Marine carbonates, such as sedimentary carbonates and secondary carbonates formed during marine alteration processes, are returned to the mantle at subduction zones, and their fates regulate the global carbon cycle. Some carbonates are destroyed at the subduction zones, contributing to arc volcanism [1,2]. However, some subducted carbonates may be sampled by intraplate volcanism via mantle plumes or upwelling in big mantle wedges or rifts (e.g. [3,4]). Geochemists are exploring novel geochemical and isotopic tools that can be used to trace the recycled carbonate signatures in both mantle derived rocks (basalts) and mantle rocks (peridotites; see contributions in this issue).

Marine carbonates are mostly CaCO3 with ∼40% CaO while the Earth's mantle typically has only 3.5% CaO. If subducted carbonates have a different Ca isotope composition, measured as δ44/40Ca (= [(44Ca/40Ca)sample/(44Ca/40Ca)standard – 1] × 1000), compared to the mantle [Fig. 1], then δ44/40Ca can be a ‘smoking gun’ for tracing recycled carbonates. This idea led Huang et al. [3] to use δ44/40Ca to estimate the contribution of recycled carbonate in the mantle source of Hawaiian shield lavas.

Figure 1.

Histogram of δ44/40Ca in carbonates and silicates. Although carbonates have a lower average than silicates (0.61 ± 0.36 [1σ] vs. 0.89 ± 0.22 [1σ]), the δ44/40Ca distributions in carbonates and silicates largely overlap. The estimated mantle values range from 0.94 to 1.05 (see Antonelli and Simon [9] for a summary), which are close to the average silicate value. We note that the highest and the lowest δ44/40Ca in silicates are found in mantle rocks: United States Geological Survey (USGS) standard samples DTS-1, -2 (Twin Sisters dunite) and Fe-rich peridotitic xenoliths from the North China Craton [14], respectively. As a consequence, the average silicate δ44/40Ca value is close to, if not representative of, mantle value. Data are from the compilations in refs [11] and [16].

Recently, Amsellem et al. [5] reported low δ44/40Ca values (∼0.6 lower than the mantle value) for a global sampling of carbonatites (up to 3 Ga). They argued that the low δ44/40Ca reflects contributions from surface carbonates recycled into the mantle sources of these carbonatites. However, a subsequent study found that global carbonatites, after screening for alteration, have the same δ44/40Ca range as that shown in silicates [6]. This result is consistent with the finding that peridotites metasomatized by carbonate-rich melt have δ44/40Ca similar to that in unmetasomatized peridotites [7]. In contrast, Banerjee et al. [8] found that the majority of ancient carbonatites (>300 Ma) have mantle-like δ44/40Ca, but some young carbonatites (<300 Ma) have significantly low δ44/40Ca value (∼0.5 lower than mantle value).

Both carbonates and silicates have large δ44/40Ca variations, and they largely overlap with each other (Fig. 1). The mantle value has been estimated based on δ44/40Ca in mid-ocean ridge basalts (MORBs), peridotites and chondrites, ranging from 0.94 to 1.05 (see Antonelli and Simon [9] for a summary). This value is close to the average of silicates, 0.89 ± 0.22 (1σ, Fig. 1). It is evident that recycled carbonates can have higher, similar or lower δ44/40Ca compared to the mantle. If the subducted carbonates have mantle-like δ44/40Ca values, Ca isotope composition cannot distinguish between carbonate and mantle contributions. That is, δ44/40Ca cannot be used to trace carbonate recycling in this case. However, it is unlikely that all subducted carbonates have mantle-like δ44/40Ca values, and δ44/40Ca can be used with other geochemical and isotopic parameters to trace recycled carbonates. This is the approach that we used to identify recycled carbonate in the Hawaiian mantle plume. Specifically, Huang et al. [3] found ∼0.3 variation in δ44/40Ca in a group of Hawaiian shield lavas. The δ44/40Ca variation is correlated with Sr/Nb and 87Sr/86Sr, which are also tracers of marine carbonates. These correlations are best explained by adding up to 4% recycled carbonate into the Hawaiian mantle plume. In this case, the other two carbonate-sensitive parameters, Sr/Nb and 87Sr/86Sr, are used to constrain the δ44/40Ca characteristics of the recycled carbonate sampled by Hawaiian mantle plume. However, the contribution of recycled carbonate estimated using only Sr/Nb and 87Sr/86Sr relies on the assumed concentrations of Sr and Nb in the recycled carbonates, which can vary up to several magnitudes. The relative contribution of recycled carbonate is best constrained by δ44/40Ca using mass balance, because Ca is a major element in carbonates and its content does not vary too much.

Although δ44/40Ca has the potential to trace carbonate recycling, several issues must be better addressed/understood before making full use of δ44/40Ca. First, there are considerable inter-laboratory differences among published δ44/40Ca data. For example, published δ44/40Ca values in the BHVO-1 and BHVO-2 standards range from 0.73 to 0.96 [10], much larger than the typical analytical uncertainty of ±0.05. As discussed above, three recent studies on global carbonatites [5,6,8] yielded very different results. Sun et al. [6] suggested that the lower δ44/40Ca in Amsellem et al. [5] might result from the matrix effect of rare-earth elements (REEs) that may not be effectively removed during their chemical separation of Ca. However, not all carbonatites studied by Amsellem et al. [5] have high REE contents, and their measured carbonatite δ44/40Ca does not correlate with REE content. Therefore, the exact reason for such large inter-laboratory difference remains unclear. Nevertheless, the reported large inter-laboratory difference in published Ca isotope studies must be understood and addressed before making full use of δ44/40Ca data to trace recycled carbonates.

Second, partial melting effects on δ44/40Ca should be better addressed. Using a combination of δ44/40Ca measurements and ab initio calculation results, Zhang et al. [11] showed that partial melting of spinel peridotite introduces only a minor effect on δ44/40Ca in melts, which is smaller than typical analytical uncertainty. Instead, a large δ44/40Ca effect, of up to 0.3, can be produced in melting residues. Wang et al. [12] found that garnets have higher δ44/40Ca, by 0.1 to 0.4, than their coexisting clinopyroxenes in a series of eclogites and garnet-peridotites from the Dabie-Sulu orogen, China, consistent with the ab initio calculation results [13]. Wang et al. [12] also inferred that jadeite-rich clinopyroxene, which is typically stable under high pressure, tends to be enriched in heavy Ca isotopes. Consequently, it is expected that partial melts of garnet peridotites and garnet pyroxenites have lower δ44/40Ca, as observed in low-Mg adakitic rocks [12]. However, the exact δ44/40Ca effects during partial melting of garnet-bearing lithology are not well constrained. This is because there are no available mineral-melt Ca isotope fractionation factors under high pressures (>3 Ga) within the garnet stability field. The low-pressure mineral-melt Ca isotope fractionation factors inferred by Zhang et al. [11] should not be directly applied to higher pressures, because the structures of minerals and melt respond very differently to pressure. Similarly, Ca isotopic effects during partial dissolution and breakdown of carbonates, and isotopic exchange among carbonate and silica minerals at subduction zones, are also not well constrained due to the lack of reliable mineral-fluid Ca isotope fractionation factors. Mineral-melt/fluid Ca isotope fractionation factors under high pressures must be determined experimentally and/or by ab initio calculations before making full use of δ44/40Ca data to trace recycled carbonates.

Last, kinetic processes can also fractionate Ca isotopes. Zhao et al. [14] found very low δ44/40Ca, down to −0.1, in some peridotite xenoliths from the North China Craton, and they attributed the low δ44/40Ca to the melt-rock reaction that occurred at the base of the lithospheric mantle. Specifically, lighter Ca isotopes are preferentially incorporated into clinopyroxene during the melt-mantle reaction. Antonelli et al. [15] reported low δ44/40Ca in both natural and experimentally grown plagioclases, and they attributed the low δ44/40Ca in plagioclases to kinetic effects during magmatic plagioclase crystallization. Additional geochemical and isotopic tracers must be used together with δ44/40Ca to distinguish between recycled carbonates, metasomatized mantle and plagioclase cumulates.

In summary, because of the large δ44/40Ca variation in carbonates compared to the mantle, and the large CaO concentration difference between marine carbonates and the mantle, Ca isotope composition is a potentially powerful tool for tracing the recycling of carbonates into the mantle, especially when used in combination with other geochemical and isotopic tracers for carbonates (see contributions in this Special Topic). However, there remain some issues that should be addressed: (i) large inter-laboratory differences in Ca isotopic measurements, and (ii) a better understanding of Ca isotopic effects during igneous, metasomatic and metamorphic processes, including under high pressures, using analytical, experimental and theoretical approaches.