Ice crystals (and sugar and other crystals for that matter) can only form through a process called nucleation. What this means is that snow does not form spontaneously, but instead must have something — could be a speck of dust, could be another ice crystal — from which the crystal can start building on.
The classic “kitchen chemistry” experiment that demonstrates this is the one used to make rock candy — if you boil a totally saturated sugar water solution such that the sugar completely dissolves in the boiled state, and you let the boiled solution cool without disturbing the liquid, no crystals jump out. But, if you stick something into the water (e.g. a popsicle stick), the crystals immediately form on the stick. The stick acts as the nucleator, letting the sugar crystals build onto something.
Tara C. Smith of Aetiology (hat tip: A. Phan) points out that the ice crystals which make up snow often use bacteria as nucleators:
The authors [of the Science paper] were looking at ice nucleators (IN) in snowfall. According to the Science paper, those IN are frequently bacteria, including, as the author notes in the news interview, some pathogens of plants (such as Pseudomonas syringae). Apparently (unbeknownst to me), P. syringae is already used to make fake snow (link), so the fact that it can serve as a seed for precipitation isn’t new. However, the authors note just how important these biological nucleators (including P. syringae) appear to be in the atmosphere:
“The samples analyzed were collected during seasons and in locations (e.g., Antarctica) devoid of deciduous plants, making it likely that the biological IN we observed were transported from long distances and maintained their ice-nucleating activity in the atmosphere… our results indicate that these particles are widely dispersed in the atmosphere, and, if present in clouds, they may have an important role in the initiation of ice formation, especially when minimum cloud temperatures are relatively warm.”
Bacteria… is there anything they *can’t* do?