Melt mixing causes negative correlation of trace element enrichment and CO2 content prior to an Icelandic eruption

- We present trace element and volatile data for 110 Icelandic melt inclusions.
- Melt mixing and fractional crystallisation explain trace element variability well.
- Inclusion incompatible trace element enrichment correlates negatively with CO2 content.
- Inclusion CO2 contents are controlled by both melt mixing and CO2 exsolution.
- Depleted melt inclusions may record CO2-supersaturated compositions.

Major elements, trace elements and volatiles were measured in 110 olivine-hosted melt inclusions from the subglacial Skuggafjöll eruption in the Eastern Volcanic Zone of Iceland. Variations in melt inclusion trace element concentrations can be accounted for by incomplete mixing of diverse mantle parental melts accompanied by variable extents of fractional crystallisation. Binary mixing between an incompatible trace element-enriched and depleted melts provides a good fit to observed variations in trace element ratios such as Ce/Y. Surprisingly, the CO2 contents of melt inclusions correlate negatively with their degree of trace element enrichment. Depleted, low-Ce/Y inclusions with ∼1200 ppm CO2 have high CO2/Nb contents (∼400), suggesting that melts experienced little or no CO2 exsolution before inclusion entrapment. Enriched, high-Ce/Y inclusions contain ∼300 ppm CO2, have low CO2/Nb (contents 50-100) and melts are likely to have exsolved much of their original CO2 contents prior to inclusion entrapment. The negative correlation between CO2 content and trace element enrichment may arise either from the more efficient exsolution of CO2 from enriched melts, or from the intrusion of CO2-supersaturated depleted melts into enriched melts that had already exsolved much of their original CO2 contents. Some inclusions have lower CO2 contents than predicted from binary mixing models, which suggests that at least some CO2 exsolution occurred concurrently with mixing. Enriched inclusions record entrapment pressures of ∼0.5 kbar. These pressures probably correspond to the depth of mixing. Higher pressures recorded in depleted inclusions may have resulted from the development of CO2 supersaturation during ascent from storage at ≥1.5 kbar. The presence of CO2 supersaturation in melt inclusions has the potential to constrain timescales of melt inclusion entrapment.