Melt degassing by bubble nucleation and growth is a driving mechanism of magma ascent. Therefore, decompression experiments with hydrous silicate melts were used to investigate the onset and the dynamics of H2O degassing. Nominally H2O-undersaturated trachytic Campi Flegrei and phonolitic Vesuvius melts representative for the magma compositions of the Campi Flegrei volcanic system were decompressed at a super-liquidus temperature of 1050 °C from 200 MPa to final pressures (Pfinal) of 100, 75, and 60 MPa using continuous decompression rates of 0.024 and 0.17 MPa/s. Experiments started from either massive glass cylinders or glass powder to demonstrate the influence of the starting material on melt degassing. Glass powder can be used to shorten the equilibration time (teq) prior to decompression for dissolution of H2O in the melt. The decompressed samples were quenched and compared in terms of bubble number density (NV), porosity, and residual H2O content in the melt.
Decompression of all glass cylinder samples led to homogeneous bubble nucleation with high NV of ~105 mm−3. The supersaturation pressures for homogeneous bubble nucleation were estimated to be <76 MPa for the trachytic and <70 MPa for the phonolitic melt. In contrast to glass cylinders, the usage of glass powder equilibrated for 24 h before decompression prevented homogeneous bubble nucleation during decompression. We suggest that trapped air in the powder pore space resulted in the formation of tiny H2O-N2 bubbles throughout the samples prior to decompression. Degassing of these glass powder samples was facilitated by diffusive growth of these pre-existing bubbles and thus did not require significant H2O supersaturation of the melt. This is evidenced by several orders of magnitude lower NV and lower residual H2O contents at correspondingly higher porosities compared to the glass cylinder samples. However, a significant extension of teq to 96 h in the glass powder experiments led to degassing results comparable to the glass cylinder samples. This effect is probably due to Ostwald ripening, coalescence, and the ascent of the pre-existing bubbles during the extended teq prior to decompression.
The NV of the glass cylinder samples were used to test the applicability of the vesiculation model provided by Toramaru (2006). For the applied decompression rates, the experimental NV are up to 5 orders of magnitude higher than the values predicted by the model. This may be mainly attributed to the usage of the macroscopic surface tension and the total H2O diffusivity in the model to describe the molecular process of bubble nucleation. A significant increase in modeled NV can be achieved by application of a reduced surface tension in combination with the lower diffusivity of network formers as a limiting parameter for the formation of a bubble nucleus.
This study demonstrates that the investigation of homogeneous bubble nucleation necessitates an optimized experimental protocol. We strongly recommend performing experiments with massive glass cylinders as starting material. The timescale of decompression is a limiting parameter and must be short enough to minimize the opportunity for a reduction of NV by bubble coalescence. Considering our comparably high NV, the samples of many previous experimental studies that were used to calibrate models for homogeneous bubble nucleation were probably subject to significant NV reduction. Newly derived data from optimized experiments will require improved models for homogeneous bubble nucleation during magma ascent.