A carbon-rich lithospheric mantle as a source for the large CO₂ emissions of Etna volcano (Italy)

Over geological time, volcanism represents the main input for atmospheric CO₂. The highly heterogeneous CO₂ flux of present-day volcanoes testifies to the presence of multiple sources, which must be identified to infer long-term variations in CO₂ fluxes. The subcontinental lithospheric mantle (SCLM) is an important geochemical reservoir that can sequester large amounts of CO₂ due to the infiltration of fluids and melts during carbonatite-like metasomatism. Carbon-rich domains within the SCLM can release CO₂ along with other volatiles (i.e., H₂O, S, Cl, and Fl), when tapped by volcanism. This process is well known in continental rift settings and is responsible for the significant CO₂ emission in the East African Rift, which contributes ~30% of global volcanic emissions (Foley and Fischer, 2017; Aiuppa et al., 2019; Muirhead et al., 2020). However, outside of rift systems, the role of the SCLM in Earth's CO₂ cycle is poorly known.

After the Nyiragongo-Nyamuragira volcanic system (East African Rift), the largest emitters of volcanic CO₂ are Popocatepétl volcano in Mexico (9345 t/day) and Etna volcano in Italy (9083 t/day). As such, Etna contributes to 10% of global volcanic emissions (Aiuppa et al., 2019). The volume of CO₂ released by Etna largely exceeds the amount that can be dissolved in the erupted magma (D’Alessandro et al., 1997; Ferlito, 2018; Aiuppa et al., 2019), and it is three times that of Kīlauea (Hawaii), despite emplacing just a quarter of the magma volume.

Etna is located on the edge of a subduction zone and its volcanism is related to mantle upwelling due to the rollback of the Ionian slab (Gvirtzman and Nur, 1999). As such, subduction-derived fluids or assimilation of Mesozoic carbonates have been proposed as a source of the emitted CO₂ (e.g., Frezzotti et al., 2009; Heap et al., 2013). Yet, the magmas of Etna show an intraplate affinity with a limited influence from the Ionian subduction (Schiano et al., 2001, Tonarini et al., 2001), and there is no evidence of carbonate assimilation (Tonarini et al., 2001). Thus, the excessive degassing of CO₂ remains enigmatic and possibly requires a deep source (D’Alessandro et al., 1997; Ferlito, 2018). Because Etna is located between oceanic and continental lithosphere, it is plausible that carbon-rich SCLM metasomatism influences the CO₂ degassing of Etna. To evaluate this possibility, we report high-precision concentrations of high field strength elements (HFSE) that have been demonstrated to be sensitive to carbonatite metasomatism (e.g., Green, 1995; Pfänder et al., 2012). Magmas from Etna were analyzed along with samples from neighboring volcanoes from different geodynamic settings, namely (1) Stromboli (Tommasini et al., 2007), which is characterized by a clear subduction signature; (2) Pantelleria (Avanzinelli et al., 2014), with pure asthenospheric affinity (i.e., no SCLM contribution); and (3) Monte Vulture (Avanzinelli et al., 2008), a hybrid volcano emplaced on a geodynamic setting similar to that of Etna.

Isotope dilution HFSE data along with radiogenic isotope data and trace element content are given in the Supplemental Material¹. The most striking feature of Etna lavas is their high Nb/Ta ratio (up to 26), which is not observed in mid-oceanic ridge basalt (MORB) and ocean-island basalt (OIB) (average of 15.9 ± 0.6; Pfänder et al., 2007). Other volcanoes of the region, such as Pantelleria and Stromboli, display lower Nb/Ta values (17–20), while Monte Vulture also shows high Nb/Ta (up to 22). The geochemical twins Nb and Ta are incompatible elements that show little or no relative fractionation during magmatic processes but are sensitive to subduction processes and carbonatite metasomatism (e.g., Green, 1995). As such, high Nb/Ta may derive from fluids that are released by the Ionian subduction, where residual Ti-rich phases like rutile can cause Nb-Ta fractionation (e.g., Green, 1995; Münker et al., 2004). On the other hand, carbonatite-like metasomatism of the SCLM can result in magmas with high Nb/Ta (Pfänder et al., 2012) as evidenced by the extreme Nb/Ta observed in carbonatites (up to 1000; Hoernle et al., 2002; Bizimis et al., 2003).

The OIB-like Nb/Ta of Pantelleria indicates that the high Nb/Ta is not a common feature of the regional asthenosphere, but affects only Etna and Monte Vulture. These two volcanoes share a peculiar tectonic setting, being on the opposite edges of the subducting Ionian slab. Monte Vulture shows a geochemical affinity between intraplate and arc magmatism (Avanzinelli et al., 2008; Caracausi et al., 2013). Etna has a geochemical signature similar to OIB (Viccaro and Zuccarello, 2017; Casetta et al., 2019), although influence by the Ionian subduction has been proposed for the younger alkaline products (Schiano et al., 2001; Tonarini et al., 2001). Etna has Ba/Ce and Th/La ratios similar to those of Pantelleria, OIBs, and MORBs but significantly lower than arc magmas such as Stromboli (Fig. 1). When considered together, Etna, Monte Vulture, and Stromboli display a broadly negative correlation in Ba/Ce and Th/La versus Nb/Ta space (Fig. 1), which argues against a subduction influence on the Nb/Ta ratios. At Etna, such a possibility is also excluded because the highest Nb/Ta values are observed in tholeiites, which are the oldest products of the volcano and devoid of subduction influence (Schiano et al., 2001; Tonarini et al., 2001). Because tholeiites require a larger degree of partial melting than alkali basalts (as also evident from Ba/Ce and Th/La; Fig. 1), a role of residual phases in generating the anomalous Nb/Ta is also unlikely. In any case, olivine, pyroxene, amphibole, garnet, and phlogopite (i.e., the mineral phases expected in the mantle source of Etna; Viccaro and Zuccarello, 2017; Casetta et al., 2019), show negligible or similar partition coefficients for Nb and Ta (Adam and Green, 2006), which ultimately eliminates the possibility that partial melting could significantly fractionate Nb from Ta.

Interaction with the lithospheric mantle can strongly influence the composition of intraplate magmas (Halliday et al., 1995; Pfänder et al., 2012). The possible role of the SCLM at Etna and Monte Vulture is supported by the negative correlation between Nb/Ta and K/K*, which is an index for assessing lithospheric mantle contributions (Halliday et al., 1995). Relative to MORBs and OIBs, carbonatites display remarkably higher Nb/Ta and lower Zr/Nb (Fig. 2). Etna and Monte Vulture are clearly distinguished from OIBs and MORBs since both ratios point toward the field of carbonatites. The lowest Zr/Nb recorded in tholeiites from Etna is at odds with their expected high degree of partial melting but consistent with a larger contribution of a carbonatite-like component. Therefore, the high Nb/Ta is best explained by carbonatite metasomatism, which is a widespread process in the lithospheric mantle (Pfänder et al., 2012; Foley and Fischer, 2017). The SCLM carbonatite-like metasomatism is also consistent with other geochemical features of Etna, such as the strong depletion in Rb (e.g., low Rb/Th; Fig. S1 in the Supplemental Material). Carbonatite melts are characterized by low Rb (and Cs) contents, with respect to silicate melts, where Rb is one of the most incompatible elements.

The high CO₂ degassing of Etna could derive from shallow level assimilation of crustal carbonates (e.g., Frezzotti et al., 2009; Heap et al., 2013), as suggested for other volcanoes such as Popocatépetl and Mount Merapi (Indonesia) (Schaaf et al., 2005; Deegan et al., 2010). These two volcanoes show clear evidence of carbonate assimilation such as carbonate xenoliths and shifts toward radiogenic ⁸⁷Sr/⁸⁶Sr (Schaaf et al., 2005; Deegan et al., 2010). At Etna, ⁸⁷Sr/⁸⁶Sr is significantly lower (0.7029–0.7036) than local Mesozoic carbonates (~0.707), and it correlates negatively with Nb/Ta (Fig. S2). Therefore, assimilation of crustal carbonate cannot explain the high Nb/Ta of Etna or, in general, its magma composition, as previously demonstrated by Tonarini et al. (2001). Thermally induced decarbonation of Mesozoic limestone may contribute to the CO₂ emission of Etna (Heap et al., 2013), but it would have no effect on the magma geochemistry. However, a major role for such a process was discarded on the basis of carbon and sulfur isotopes (D’Alessandro et al., 1997). Unlike crustal carbonates, mantle carbonatites are characterized by unradiogenic Sr (Hoernle et al., 2002; Bizimis et al., 2003). Thus, the high Nb/Ta (and low Zr/Nb and Rb/Th) of Etna and Monte Vulture reveals carbonatite-like metasomatism in the SCLM beneath southern Italy. Because mantle domains affected by carbonatite-like metasomatism can contain carbon on the order of few weight percent (Foley and Fischer, 2017), the SCLM represents a viable source for the anomalously high CO₂ fluxes recorded at Etna.

The carbon content of the SCLM can increase due to carbonatite-like metasomatism associated with the infiltration of deep melts that are possibly related to mantle plumes or paleo-subduction (e.g., Aulbach et al., 2017; Foley and Fischer, 2017). Carbonatite-like metasomatism in the SCLM has been suggested for the neighboring Hyblean plateau (Trua et al., 1998; Beccaluva et al., 1998) (a few tens of kilometers south of Etna), which taps a mantle source similar to that of Etna (Viccaro and Zuccarello, 2017; Casetta et al., 2019). Mantle xenoliths of the Hyblean plateau have ³He/⁴He ratios similar to those of Etna, which suggests a common origin of volatile elements (Frezzotti et al., 2009), and show evidence of carbonatite-like metasomatism as abundant CO₂ fluid inclusions (Sapienza and Scribano, 2000; Sapienza et al., 2005). Strikingly, magmas from the Hyblean plateau also have high Nb/Ta (Trua et al., 1998), which further stresses the connection between high Nb/Ta and carbonatite-like metasomatism. Notably, carbonatite metasomatism was also observed in mantle xenoliths from Vulture (Rosatelli et al., 2007)

Beyond their extreme Nb/Ta ratio, magmas from Etna and Monte Vulture display a peculiar tungsten signature, with W/U and W/Th lower than that of OIBs and MORBs (Fig. 3). As for Nb/Ta, Pantelleria and Stromboli have W/U and W/Th similar to typical mantle values (Fig. 3), which excludes fractionation of W from U and Th during mantle melting or subduction. The behavior of W during carbonatite metasomatism is unknown, but it is possible that the low W/U and W/Th of Etna originated in the SCLM. Indeed, Etna tholeiites have the lowest W/U and the highest Nb/Ta (Fig. 3B).

Fonseca et al. (2014) showed that a reduced mantle (fayalite-magnetite-quartz [FMQ] <−2) can contain higher amounts of W⁴⁺, which is significantly more compatible than W⁶⁺. Although carbonatites are relatively oxidized melts, most of the carbon in the SCLM is thought to be stored in reduced minerals, which include diamonds (Foley and Fischer, 2017). Therefore, reduced, carbon-rich SCLM domains may partially retain W during partial melting, which would result in melts with low W/U and W/Th. Mantle xenoliths from the Hyblean plateau support low oxygen fugacity (f_O2) conditions in the SCLM of Etna, as they contain reduced carbon mostly in the form of hydrocarbons but also as nanodiamonds (Simakov et al., 2015).

The previous observations can be combined with the peculiar tectonic setting of Etna and Monte Vulture to draw a coherent geodynamic model (Fig. 4). The roll back of the Ionian subducting plate (Gvirtzman and Nur, 1999) induces a corner flow of the mantle around the slab moving from the Hyblean domain toward Etna. This process may induce physical erosion of the bottom portion of the Hyblean continental lithosphere, which is the most susceptible to metasomatism (e.g., Foley and Fischer, 2017), bringing it beneath Etna. During partial melting, fragments of continental lithosphere, characterized by carbonatite-metasomatism and high Nb/Ta, can release large amounts of CO₂ and other volatiles (Fig. 4). A similar mechanism was proposed to explain the large CO₂ emission in the eastern rift of the East African Rift system, which involved advection of portions of the cratonic mantle roots toward thinner lithospheric domains (Muirhead et al., 2020).

Recently, Barreca et al. (2020) suggested the presence of a slab window in the subducting Ionian plate that allows portions of the shallow mantle wedge to flow in the opposite direction and toward Etna. Such a scenario could explain the presence of a slight subduction signature in the recent alkaline activity but not the high Nb/Ta of Etna magmas. In fact, the slab window opened after the tholeiitic activity of Etna (Barreca et al., 2020), which possibly explains the decreasing Nb/Ta in the younger alkaline basalts where the older carbonatite-like signature is partially overprinted. Accordingly, CO₂ and volatile fluxes at Etna during its tholeiitic phase were likely even higher than now, as suggested by the postulated increase in H₂O/CO₂ of the mantle source of Etna through time (Casetta et al., 2019). Indeed, relative to tholeiites, the more recent alkaline magmas are thought to derive from a mantle with higher H₂O/CO₂, as is expected from a contribution of water-rich subducting fluids that partially overprint the older carbonatite-like signature.

Monte Vulture also has elevated Nb/Ta (although not as high as at Etna) and has low Zr/Nb, W/Th, and W/U. Monte Vulture is located in a tectonic setting similar to that of Etna (on the eastern margin of the Ionian subduction; Fig. 4), and analogies linking the two volcanoes were suggested based on CO₂/³He ratios and He-Sr isotopes (Caracausi et al., 2013). Thus, mantle roots beneath the Apulia region may be also affected by carbonatite metasomatism.

HFSE concentrations reveal the involvement of a carbon-rich SCLM beneath Etna and Monte Vulture that represents the most likely source for the large CO₂ degassed at Etna. As such, HFSE can be used to study the role of carbonatite metasomatism and its effect on CO₂ degassing. This approach could potentially be applied to every volcano, in particular when mantle xenoliths are not available. Moreover, it can be used to reveal the occurrence of carbonatite metasomatism in magmas that erupted in the past, which provides a unique tool for identifying eruptions that might have released large amounts of CO₂ into the atmosphere, contributing to major climatic changes on Earth.

This work was supported by the European Commission through European Research Council grant 669666 “Infant Earth” to C. Münker and by the Italian Ministry for University and Research grants 20158A9CBM and 2015EC9PJ5 to S. Conticelli and R. Avanzinelli, respectively. We thank L. Francalanci for providing samples from Stromboli, R.O.C Fonseca C. Ballhaus and S. Tommasini for discussions, and M. Jansen, M. Kirchenbaur, C. Obert, J. Pakulla, J. Tusch, and F. Wombacher for help in the laboratory. The valuable help of Andrea Steglich in making Figure 4 is highly appreciated. Constructive reviews by editor G. Dickens and three anonymous reviewers greatly improved the manuscript.