The isotopic signature of CO₂ emitted by arc volcanoes is controlled by subducted sediments, where oxidation of graphite and isotopic buffering by carbonates play a key role. In our previous work (Tumiati et al., 2022), we demonstrated that the δ¹³C of CO₂ is dictated by the CO₂/CaCO₃ ratio rather than the bulk organic/inorganic carbon fraction. However, our model required refinement to account for very low or very high graphite/carbonate ratios, which, despite appearing extreme, are actually the most common in subducted marine sediments and their metamoprhic equivalents. At these ratios, the amount of CO₂ produced by graphite oxidation deviates from our previous model predictions and, in particular, it is no longer constant but instead depends on the graphite/carbonate ratio. Here, we present new high-pressure, high-temperature experiments to explore these conditions. To reproduce carbon sources in subducted sediments, we used a mixture of cultivated calcareous algae and green algae to set a defined Corganic/Cinorganic ratio, equivalent to graphite/CaCO₃ under experimental conditions of P=3 GPa, T=700°C and fO₂=FMQ. Our results show that at sedimentary graphite/carbonate ratios either close to 0.01, representative of calcareous sediments (limestones), or close to 0.99, representative of anoxic sediments (black shales), isotopic buffering is not fully captured by our previous model, requiring a correction factor. We introduce a non-linear mathematical model that adjusts the XCO₂ predicted by GCOH fluid models for graphite-saturated systems, given by:

Fcorr = 0.0216 · tan(0.0308 · wt.%graphite – 1.5325) + 0.5550

where wt.%graphite represents the weight percent of graphite, with the remaining fraction consisting of CaCO₃. To assess the impact of variable graphite/carbonate ratios on carbon mobility and isotopic exchange, we validate our model against thermodynamic predictions from the Deep Earth Water (DEW) model. This allows us to refine our understanding of subducted carbon cycling and arc emissions.

Tumiati S. et al. (2022) DOI: 10.1038/s41467-022-30421-5.