Fate of bubble clusters rising in a quiescent liquid

We use experiments to study the evolution of bubble clusters in a swarm of freely rising,
deformable bubbles. A new machine learning-aided algorithm allows us to identify and track
bubbles in clusters and measure the cluster lifetimes. The results indicate that contamination in
the carrier liquid can enhance the formation of bubble clusters and prolong the cluster lifetimes.
The mean bubble rise velocities conditioned on the bubble cluster size are also explored, and we
find a positive correlation between the cluster size and the rise speed of the bubbles in the cluster,
with clustered bubbles rising up to 20% faster than unclustered bubbles.