As I was walking to the train station yesterday I noticed that all the parked cars were covered in the same beautiful pattern of swirling fanlike crystals. The hoods and roofs and windshields looked like an artist had drawn all over them with a magical ice pen (does that sound cheesy? It’s exactly what it looked like). But I’m a science writer…I should know by now there’s no such thing as magic (sorry).

So I got in touch with one of our top crystal growth connoisseurs to find out what was really behind these beautiful patterns. Moneesh Upmanyu is an associate professor of mechanical and industrial engineering who uses computational techniques to understand the structures and properties of emerging materials. But he’s also pretty interested in the ancient, naturally occurring ones, too (in particular, rhododendrons!).

Here’s what he had to say about the crystalline car art I observed yesterday. “Frost on surfaces, commonly referred to as hoar frost, does not condense from the air,” he said. “Rather, it crystallizes from a thin liquid film of water on the surface.” This is usually only possible when a frosty day follows a rainy one. The phenomenon also requires three ingredients: A thin film of liquid water poised peacefully on a surface is the first. The second is a single particle of dust or other imperfection, like a pearl sitting unobtrusively in an oyster shell. Third, freezing cold temperatures.

The particle acts as a nucleus around which crystallization begins. As the water molecules in its immediate vicinity slowly drop in temperature, they begin to solidify. This layer of hoar frost is already patterned at a much smaller scale since the solidification doesn’t happen uniformly, Upmanyu said. It can grow in this branching manner due to the molecular-scale structure of the crystal. Because the water layer is so very thin, the growing frost front sucks up the liquid in its vicinity. It often begins to meander and curve as it tries to accumulate the liquid, which itself is still flowing.

Now, since the frosty regions are cooler, the water-saturated air above condenses and they thicken. This triggers further evaporation of the already thin liquid layer until it dries out, leaving dry, exposed patches of the metal roof, the glass windshield, between these frosty protrusions, Upmanyu said. Without any water, crystallization stops locally and we see the emergence of these fan-like frost patterns.

The evaporation of the exposed surface makes it much colder (just as we feel cold when water evaporates off our skin) that it triggers frost formation from the remaining liquid at a different location. This depends on how well the surface can conduct heat, and the process repeats until the entire surface is decorated. Overnight, the dry surfaces will accumulate some frost as well, but they will always be thin and the pattern will persist.

Upmanyu said he’s seen windshields where the whole thing was one giant set of concentric crystalline circles, instead of repeating fans like what I saw. “This is a sign of a very clean windshield,” he chuckled. Just a single particle of debris is enough to set it off into a symmetric growth pattern.

What happened on the cars is essentially the same way a snowflake forms up in the clouds, Upmanyu told me, although there are lots of things scientists are still scratching their heads about in both departments. As water molecules hang out in the slowly cooling air (it could take a whole day for the temperature to drop just a couple of degrees), the crystal growth occurs very slowly. We know that the super-slow growth pattern of water favors a six-sided crystal orientation, which just gets amplified again and again until you can actually see it on a macroscopic scale as a snowflake. Upmanyu said that if you cool water down in the lab more rapidly, you will see a much different pattern emerge.

“This is of course a simplified explanation,” Upmanyu said. “Pattern formation is almost always complex, but it’s beautiful to watch.” I couldn’t agree more