Conclusions

We have demonstrated a novel emulsion sufficiently refractive index matched and buoyancy matched that we can create three-dimensional image stacks of aggregated emulsions of significant (i.e. > 30%) volume fraction. Cross-flow emulsification has been shown to be a useful low-energy low-shear technique for generating delicate emulsions having a controllable and relatively constrained size distribution.

Polydisperse droplets have been located and their sizes estimated in three dimensions using a novel algorithm based on the Euclidean distance map. These droplet coordinates have then been used to construct individual droplet trajectories. Trajectories show deviations from simple vertical movement, and in particular suggest a modest degree of local network rearrangement.

Image stacks show dislocations linked to the presence of “swirls” in the sample. These swirls are of short duration (a few seconds) and are probably linked to transiting bubbles elsewhere in the system. Swirls, and explicit network deformations associated with bubbles, imply a rigid network structure. MRI, which we have demonstrated to a resolution of tens of microns, provides additional evidence for a rigid network. Network relaxation has been captured on a number of length and time scales.

Creaming studies obey a scaling that implies a universality of behaviour that is broken for low polymer formulations exhibiting collapse. The onset of this transient gel collapse gel has been explicitly captured via dynamic confocal microscopy. The collapse transition appears to be accompanied by large-scale collective network movement, with a transverse component that is accompanied by a vertical component that increases with time. Finally clusters go into “free-flight” and the network breaks up. The impression is of a mechanical disintegration into fragments, though the initiator for this behaviour remains unclear. The notion that a transient gel is quiescent prior to collapse we now know to be false, though on bulk scales the dynamics are very slow.

Brownian dynamics simulations, both monodisperse and polydisperse, and both with and without gravity, qualitatively reproduce the network structure seen in actual confocal images. An improved void analysis technique has been shown to be capable of quantifying the void structure in both simulated and real systems. Void analysis of emulsion image stacks at different times shows no discernible structural differences over and above noise levels.

Confocal imaging and related trajectory studies do not support the thesis of voids that grow and propagate in time and which, driven by gravity, give rise to the collapse of a transient gel state. Nor do Brownian dynamics simulations and void analysis at different times offer any particular support for this scenario. The partial recovery of the network in the wake of a transiting bubble, the explicit filling of a void artefact seen in the MRI, and the non-observation of a persistent void population from MRI all count against the fractures hypothesis, at least in the naïve way it was originally visualized.