Mechanisms of bubble coalescence in silicic magmas

Abstract

Bubble coalescence is an important process that strongly affects magmatic degassing. Without coalescence, bubbles remain isolated from one another in the melt, severely limiting gas release. Despite this fact, very little has been done to identify coalescence mechanisms from textures of magmatic rocks or to quantify the dynamics of bubble coalescence in melts. In this paper, we present a systematic study of bubble-coalescence mechanisms and dynamics in natural and experimentally produced bubbly rhyolite magma. We have used a combination of natural observations aided by high-resolution X-ray computed tomography, petrological experiments, and physical models to identify different types of bubble–bubble interaction that lead to coalescence on the timescales of magma ascent and eruption. Our observations and calculations suggest that bubbles most efficiently coalesce when inter-bubble melt walls thin by stretching rather than by melt drainage from between converging bubble walls. Orders of magnitude are more rapid than melt drainage, bubble wall stretching produces walls thin enough that inter-bubble pressure gradients may cause the melt wall to dimple, further enhancing coalescence. To put these results into volcanogical context, we have identified magma ascent conditions where each coalescence mechanism should act, and discuss the physical conditions for preserving coalescence structures in natural pumice. The timescales we propose could improve volcanic eruption models, which currently do not account for bubble coalescence. Although we do not address the effect of shear strain on bubble coalescence, the processes discussed here may operate in several different eruption regimes, including vesiculation of lava domes, post-fragmentation frothing of vulcanian bombs, and bubbling of pyroclasts in conduits.

Appendix

Tomography methods

We implemented a number of image processing steps in order to reduce the noise inherent in the μ-CT scans. We performed most of the noise reduction with ImageJ anisotropic diffusion filter and then converted the denoised grayscale images to binary format. We characterized the geometry of the IBFs both qualitatively, by examining the general form of the interfaces throughout given tomographic volumes and by measuring the wavelengths, amplitudes, and thicknesses of the glass wall at the point of maximum deflection between the bubbles. These three-dimensional characteristics were measured on the binary image stacks with Blob3D software, which allows the user to manually select the individual bubbles for geometric characterization. We measured the volume of the dimpled walls using Blob3D, again with the plane tool, which separates the deformed region from the rest of the spherical bubble. These volume measurements slightly overestimate the true volume of the dimple, as the planar section will truncate part of the curve defining the surface of the spherical bubble. The error associated with this separation routine is minor, however, and we estimate that to be less than 0.1 % of the dimple volume and likely 0.01 % of the total bubble volume.

Determination of maximum and average bubble growth rates from VSDs

The decompression experiments of Burgisser and Gardner (2005) were performed on prehydrated and prevesiculated rhyolite glasses. In order to create materials suitable for decompression experiments, they applied a pretreatment involving rapid decompression of the hydrous rhyolite (825 °C and equilibrated at 160 MPa) to a pressure of 100 MPa. This rapid decompression step generated a population of small bubbles that would grow in subsequent experiments according to the applied decompression regimen. We characterized the initial size distribution of this “dwell” population of bubbles with 3D X-ray μ-Tomography, the results of which are presented in Fig. 3. Changes to the bubble population during the different decompression histories were determined by comparing the VSDs in decompression experiments at different rates. Specifically, we measured the difference in the modal and maximum bubble size between the dwell population of the respective decompression runs (Fig. 4). This difference, divided by the experiment duration, yielded the average and maximum bubble growth rates. The resulting rates were used to estimate the magnitude of viscous stress that might arise if two neighboring bubbles grow at different rates.