Under atmospheric conditions water is ice when the temperature is below zero Celcius... but only if one waits long enough, sometimes a very long time indeed.  On realistic time scales for atmospheric clouds, water is very often found in a metastable, supercooled state, and requires an external catalyst for ice formation.  But how those catalyzing ice nuclei work, at a fundamental level, is not well understood, despite a large collection of empirical observations.  Our laboratory has contributed some of those observations, mainly focused on several puzzles related to ice nucleation.  For example, what is special about an ice nucleus that contacts a supercooled water drop compared to being immersed in the water drop?  Simple experiments show that nucleation rates increase by orders of magnitude when the nucleus is at the water surface. 

There has been some suggestion that the nucleus-water-air "triple line" may play a role, and our recent experiments are illuminating the caveats. In our experiments we record the freezing of water droplets on idealized substrates at very high speed, so that we can determine the location of the nucleation events. This random freezing process is repeated many times to build a statistical ensemble indicating possible preference for the triple line. On atomically smooth substrates no triple line preference is observed, but on substrates with imposed texture we can cause nucleation at the triple line; specifically, the spatial frequency of the texture apparently plays a role, with patterns having roughness scales below 100 nm being effective. Might surface roughness therefore be a property of atmospheric aerosol particles that regulates their ice nucleating efficiency? That intriguing question will only be answered with more experiments.