The magmatic carbonate-sulfide occurrences discussed here are consistently hosted by alkaline volatile-rich ultramafic–mafic rocks, which commonly also display accessory P and Te minerals. The intimate association of C- and S-bearing minerals observed in a range of mantle rocks and in mineralized intrusions emplaced within the lower and mid crust may provide insights into the poorly known transport and concentration mechanisms of dense metal-rich sulfides in silicate magmas. We put forward the hypothesis that the observed textural relationships reflect a previously unrecognized process, which may enable the physical fluxing of volatiles and metals across the mantle-crust transition.
In much of the continental and oceanic crust, metals can be transported, (re)cycled, and (re)distributed through typical crustal processes, which almost always include a volatile component (e.g., H2O and CO2), present as hydrothermal fluids of varying compositions, fO2 and pH. Sulfur and C, largely in the form of carbonate, are common constituents in many hydrothermal ore deposits44,45 and, although C is clearly present as a volatile in crustal hydrothermal systems46, it is not generally considered to play a vital role in the mantle-to-crust magmatic transfer of metals, with the exception of rare metals including Nb, Ta, and rare-earth elements in carbonatite magmas.
Nevertheless, the mantle is considered to be an essential reservoir for the global C budget, representing a deep terrestrial reservoir containing CO2, carbonate, diamond and/or Fe-metal carbides47,48. In mafic and ultramafic magmas, the major volatile species are H2O, CO2, S, F, and Cl, whereas the metal inventory is dominated by chalcophile and highly siderophile elements (e.g., Ni, Cu, PGE, Au, and Te) in close association with S. In terms of their metal budgets and dominant fluid type(s), the composition of fluids/melts that exsolve from a magma is dependant on the depth (pressure) and the intial composition of the magma, which in turn is a function of the degree of partial melting and the nature of the mantle source3,49. The examples of mantle rocks that we show all indicate an intimate association between C- and S-bearing phases (Fig. 1). Isotopic evidence for the association of mantle-derived C with magmatic sulfide occurrences at varying lithospheric levels suggests that the parental melts of the host intrusions were derived from mantle source regions enriched in C (Fig. 4).
In the SCLM, metasomatism by melts and fluids derived from deeper regions of the mantle commonly leads to enrichment of C alongside other incompatible and volatile elements50. Carbonate melts exhibit lower viscosities than ambient silicate melts, with high wetting angles resulting in the ability to infiltrate silicate minerals and promote widespread lateral metasomatism51,52. We therefore concur with previous authors that C and other volatile components (S, Cl, F, and P), along with the Te in the mineralogical assemblages within the host magmas and mantle xenoliths documented here, result from metasomatic enrichment of the SCLM from a range of melts and fluids30,53,54.
In the deep lithosphere, there appears to be an intimate textural association between Ni-Cu-PGE-Te sulfide mineralization and carbonate. This observation supports the hypothesis that the presence of C may not just be an inherent source characteristic, but that it may also play a critical role in the physical transport and concentration of metal-rich sulfides in ascending magmas. This association appears to be common at varying lithospheric depths ranging from the upper mantle to the mid continental crust, although it is not seen in the upper crust. Some deposits may locally display evidence of carbonate alongside Ni-Cu-PGE sulfide mineralization due to localized assimilation of C-bearing crustal lithologies upon emplacement (e.g., Noril’sk55). but this is not the mantle-sourced C that we report from deeper in the lithosphere.
The question is whether the carbonate-sulfide association documented here is simply serendipitous, or if it reflects a C-driven physical mechanism to flux S and metals from the mantle into the lower continental crust, very much like water-dominated processes have been shown to play a fluxing role in the mid to upper crust12,16,14. To address this question, it is necessary to discuss the geochemical behavior of S- and C-bearing fluids in the mantle and in ascending silicate magmas.
The role of S in carrying metals in sulfide liquids is well known3. In general, S solubility in silicate melts increases with decreasing pressure56. If residual sulfide is present in the mantle during melting, as it would be for melts generated by <10% melting, then the S concentration of the silicate melt should be equal to the S concentration at sulfide saturation5. As a result, silicate magmas derived from relatively low degrees of partial melting of the lithospheric mantle may depart their source sulfide supersaturated and, under favorable conditions, remain sulfide supersaturated at the base of the continental crust (see below).
This is evidenced in exposed lower crustal intrusions in the Ivrea Zone, Italy, and the Seiland Complex, Norway21, which have abundant magmatic sulfides, and also from melt inclusions in Hawaiian basalts that indicate sulfide supersaturation at the mantle-crust boundary6. Furthermore, with increased depth sulfide is the dominant S species (over sulfate) at marginally higher fO2 conditions57, such that even though many magmas will be too oxidized to be sulfide supersaturated at upper crustal conditions58, lower crustal intrusions with similar composition may be supersaturated in sulfide. Regardless, mafic and ultramafic magmas emplaced into the upper crust will most likely be sulfide undersaturated, even if they have previously undergone sulfide saturation. For this reason, externally derived crustal S is considered critical in triggering supersaturation in most upper crustal magmatic systems59.
Conversely, the role of C in carrying metals is poorly understood. When dissolved in a silicate melt as CO2, C typically becomes less soluble with decreasing pressure60. If a silicate melt becomes supersaturated in CO2, any exsolved CO2 should behave as a supercritical fluid rather than a gas phase at temperatures >31 °C and pressures >75 bar (i.e., anything deeper than the uppermost crustal conditions). As a result, the contrasting pressure-dependent solubility of S and CO2 will dictate that if both S and CO2 are present in the same magmatic system, their physical state (e.g., sulfide liquid, S-undersaturated silicate melt, carbonate melt, CO2 supercritical fluid, etc.) will depend on depth and also be a function of the degree of melting and the initial composition of the magmas.
Increasing degrees of mantle melting dilute the concentrations of incompatible elements and volatiles including CO2 and S in melts. As such, melt saturation in CO2 and S, as well as separation of CO2-rich fluids and sulfide liquids in the mantle, are necessarily restricted to low-degree melting regimes (<10%), and therefore alkaline mafic/ultramafic melts. The recent experimental work of Chowdury and Dasgupta61 on the concentration of S at sulfide saturation in carbonate-rich silicate melts provides a potential theoretical framework in support of our hypothesis. Assimilation of silicate mantle wall rocks ubiquitously affect CO2-rich silicate magmas during their ascent through the lithospheric mantle62,63. This process results in an increase of SiO2 contents in alkaline mafic/ultramafic magmas and, therefore, lowers the solubility of CO2, which is inversely related to SiO2 concentrations60. At pressure ≥3.5 GPa (~100–110 km of depth), interaction of carbonate-rich melts and peridotite wall rocks (especially orthopyroxene) can drive out large amounts of CO2 from ascending melts and generate CO2-rich supercritical fluids64. A large drop in CO2 and related increase in SiO2 contents above 35–40% largely decrease the solubility of reduced S and promotes the formation of immiscible sulfide melts61 as well. What remains to be addressed is whether or not CO2-rich supercritical fluids and sulfide melts can remain physically connected during ascent once exsolved from their parental silicate magma.
Our proposed model is summarized below and in Fig. 5. In the lithospheric mantle, the carbonate-sulfide(-telluride-apatite) association identified in mantle xenoliths beneath Patagonia, South Africa, and Italy (Fig. 1), alongside examples from the Canary Islands24, Norway25, Australia26, and Scotland29, highlights the widespread link of carbonate melts and/or CO2-rich fluids associated with sulfide within the metasomatized domains of the lithospheric mantle. The variety of textures and mineralogy shown here (Fig. 1) and in the examples cited above reflects the heterogeneity of the mantle, though in all cases S is intimately associated with C. The common occurrence of sulfides included in diamonds23 substantiates this widespread link in the mantle.
Experimental studies by Woodland et al.65 have shown that silicate-carbonatitic melts in the mantle are able to dissolve and transport significant S; however, when mantle-derived magmas are sulfide supersaturated in the deepest portions of the lithosphere, chalcophile metals will be largely transported in sulfide droplets30,61,66. Following low-degree partial melting producing carbonate-rich alkaline melts, silica contamination would trigger both CO2 and sulfide supersaturation in the melt (Fig. 5b). However, supersaturation of sulfide liquids is not conducive to an efficient upward transfer of metals, as sulfides are dense and tend to coalesce into larger blebs, which would settle or break apart11. A mechanism is therefore required to overcome this density problem and facilitate the upward physical transport of dense metal-rich sulfide into the crust.
The physical and chemical form of C plays a key role in the efficient transport of sulfide, being most effective when present as a CO2 supercritical fluid phase compared to a carbonate melt, as CO2 can act as a physical buoyancy aid to sulfide droplets. Decarbonation of the CO2-rich silicate melt as a result of interaction with mantle wall rocks in the upper lithospheric mantle (<3.5 GPa) will exsolve CO262,64. At such depths, this will take the form of a CO2-rich supercritical fluid, where the low-density exsolved CO2 fluid phase has a density of ~1.2 g cm−3 at pressure of ~2 GPa62. The spread in carbonate C-O isotope values observed at Valmaggia supports the involvement of a CO2-rich hydrous fluid (Fig. 4), which could have exsolved from the related melt already at upper mantle depths. Its relatively low density, compared with the silicate magma, will contribute to increasing the inherent buoyancy of the melt, facilitating its rapid ascent and propagation through the Moho discontinuity.
The efficiency of CO2 to transport sulfide liquid will depend on a number of factors, which, by analogue, are all outlined by Yao and Mungall13 in the context of sulfide transport by water bubbles: the relative volumes and sizes of the volatile and sulfide phases in the compound droplets, and whether they reside in a melt or mush dominated regime. As such, one would expect that the more CO2-rich the melt is (a function of partial melting and source composition), the more efficient its capacity to transport sulfide droplets will be.
The strong wetting behavior of the CO2 fluid phase with the sulfide liquid will significantly increase the buoyancy of metal-rich sulfide liquid droplets. Even if the supercritical CO2 and sulfide melt are immiscible, they nonetheless wet each other (Fig. 5c), as reflected in the textural evidence from Valmaggia (Fig. 2). The relationship of calcite and sulfide shown in Fig. 2e strongly implies that after crystallization of the silicates, supercritical CO2 and sulfide were immiscible liquids wetting each other, with the lower density CO2 forming convex outer boundaries, which were preserved when the sulfide crystallized. The calcite subsequently crystallized from trapped supercritical CO2, as shown by the growth direction of the crystals from both the sulfide and silicate grain boundaries (Fig. 2e). Furthermore, the presence of Mn-Fe carbonate inclusions in the sulfides (Fig. 2e) implies trapping of a C-rich fluid or melt. The calcite and dolomite around the sulfide margins may be the result of reaction of the supercritical CO2 with Ca and Mg from the surrounding silicate melt, whereas the carbonate trapped in the sulfide may have gained Fe and Mn from the surrounding sulfide liquid.
Our proposed “sulfide buoyancy aid” process, operating from the metasomatized lithospheric mantle through to base of the continental crust, is analogous to the established mechanism where aqueous or saline vapor bubbles are suggested to “float” sulfide and/or magnetite at mid-upper crustal depths12,14,15,17. However, the critical difference is the deeper lithospheric window where this process operates, which provides a first order mechanism to fertilize the continental crust with mantle-derived chalcophile and siderophile metals. The exsolving CO2 fundamentally changes the physical properties of the ascending magmas, enhancing their bouyancy and catalysing the physical transport of the dense metal-rich sulfide cargoes entrained in mafic–ultramafic melts (Fig. 5a, c). To support this hypothesis, we note that Mungall et al.12 observed that at the high-pressure end of their experiments (2.5 kbar), the vapor bubbles were dominated by CO2, which evolve to H2O-dominant at lower pressures. We propose that Valmaggia represents a lower crustal equivalent of the compound model proposed by Mungall et al.12 for upper crustal systems, with CO2 being the dominant volatile phase as a supercritical fluid.
The very strong spatial relationship between carbonate and sulfide in the lower crustal example at Valmaggia is also present in places in the mid-crustal setting at Mordor and Sron Garbh, where there is also a significant amount of carbonate that is decoupled from the sulfide on a centimeter scale (Figs. 2 and 5d). This change in the C-S association may be due to separation of C and S, which is likely to occur at mid-upper crustal levels (Sron Garbh represents emplacement depths equivalent to at least >1.5 kbar18), where the CO2 supercritical fluid that initially fueled the sulfide transport in the silicate melts in the lower crust and mantle may have started to separate, or convert to CO2-H2O vapor12 by the time the system reaches the upper crustal levels (Fig. 5d).
Indeed, due to the inverse relationship between pressure and S solubility56, melts that are supersaturated at depth should start to resorb their sulfide on ascent. The depth at which complete resorption is attained will depend on the initial S concentration of the melt, but is poorly constrained due to a paucity of experimental data between 100 kPa and 1 GPa. However, it is clear that melts become increasingly sulfide undersaturated on ascent, to the point where it is likely that most mid to upper crustal magmas should be sulfide undersaturated. On ascent, sulfide droplets such as those shown in Fig. 5b-d would start to be resorbed (Fig. 5e)6. The dissolution of sulfide will return chalcophile metals and S into the silicate melts54 (Fig. 5e), which would then only be able to form upper crustal magmatic sulfide deposits if sulfide supersaturation is triggered again, e.g., by assimilation of crustal S59 or prolonged fractional crystallization.
We suggest that the mechanism of CO2-fueled sulfide mobilization not only plays an essential role in the transport of sulfides within the lower-mid crust, but is also critical in the fluxing of metals and sulfide from the mantle itself. We propose that CO2-rich supercritical fluids associated with alkaline mafic–ultramafic magmas enable the initial fluxing of metals and S from the mantle into the crust. Although the C-S association is not uncommon in the mantle and at lower crustal depths, it is rare in most upper crustal settings. The contrasting physio-chemical changes that C and S are subjected to due to decreasing pressure on ascent through the crust (i.e., increasing sulfide solubility vs decreasing CO2 solubility in silicate melts) means that whilst they may be intimately associated as sulfide melt and CO2-rich supercritical fluid (or perhaps sulfide-bearing carbonate melts, e.g., Kogarko et al.24,28) in the mantle and lower crust, they are likely to decouple in the upper crust (Fig. 3). This could be the result of a number of pressure-dependent factors, such as redissolution of the sulfide into the silicate melt or degassing of CO2.
The preservation of the intimate spatial C-S association appears to be lost with decreasing crustal depth. Although supercritical CO2 seems to be most critical for the transport of sulfides at mantle and lower crustal conditions, its solubility and low preservation potential increases the likelihood that H2O or other volatiles may overprint any originally C-driven textural signature and erase any geological record of its former occurrence. In such cases, other phases may appear to be the dominant volatiles preserved within upper crustal sulfide occurrences (e.g., hydrous silicate caps15), rather than CO2. The preservation of intimately associated sulfide-carbonate in the upper crust is thus rare and the opportunity for study inherently limited. Exceptions to this may be the upper crustal deposits such as Munali19, which has been noted to have carbonates associated with sulfides. More generally though, many upper crustal deposits may be the result of CO2-rich fluids acting as a sulfide buoyancy aid in the lower crust, but the process is untraceable due to either subsequent CO2-sulfide separation, or carbonate overprinting. What our data show are evidence of the fluxing process in action, representing sulfide transport along the lithospheric pathway from source to sink. In summary, we propose that C is a significant agent “in disguise” that may facilitate the transport of sulfides across the mantle-crust transition. We suggest that this may be a common but cryptic mechanism that operates in the deep lithosphere, which leaves very little (if any) footprint behind by the time magmas reach the uppermost crustal levels.
The presence of carbonates alongside sulfides in the SCLM is consistent with a metasomatic origin of these phases and associated metals. The remarkable textural relationships of carbonate as rims and clots alongside sulfide in the lower crust indicate that C, probably as a supercritical CO2-rich fluid, plays a critical role in aiding buoyancy and acting as a driver to propel sulfides up into and through the crust, in a similar way that aqueous and saline vapor bubbles have been proposed to do in the upper lithosphere. The carbonate-sulfide association may decouple at shallower levels, due to the inverse pressure-dependent solubility of S and CO2 in silicate melts, effectively erasing any clue about this important process in the upper crust. The analogy here would be that C acts as the propellant in the first fuel tank that detatches during the launch of a rocket into space. Indeed, it plays a vital role to the success of the departure of the rocket from the Earth’s surface (mantle) into the higher levels of the stratosphere (the lower crust). However, evidence of that short but crucially important first step is generally not recorded anywhere by the time the rocket exits the terrestrial atmosphere into space (the upper crust). As such, C acts as the crucial but covert agent in the physical flux of S and metals throughout the lithosphere.