The individual atoms are much more mobile than the crystalline structure here. They are probably moving way too fast to be captured on a time scale that the crystal grows.

Think in terms of trying to make video of a tree growing and expecting to see individual water droplets falling from the sky.

You can infer the droplets must be there (as the ground would otherwise dry out and the tree die) but it is not possible to have a video of tree growing sped up so that a human can see it growing but also show water droplets falling from the sky.

The video is at atomic resolution. The dots in the crystalline structure are individual atoms.

To do this in realtime video is bonkers.

It is bonkers. But the only way we can even have that video is because the molecules are being frozen in the crystalline lattice.

As far as I know there exists no recording of actual molecules in a liquid (except maybe for something extremely viscous).

The video reminds me this photo: https://petapixel.com/2010/10/27/first-ever-photograph-of-a-...

It shows first ever photo of a human being on a busy street.

But you would not be able to say so, because the street looks completely empty.

The only way we can even see the person is because they stood still for long enough to leave imprint. Other than this one person, everybody else on this busy street moved so fast there is no recognizable imprint of them left. Hence the street looks empty.

Gotcha, I misunderstood - thought you were referring to the scale of the raindrops relative to the tree, not the speed. The photo's a better analogy (and cool).

This much is clear. The post above answers the question of why we can't see atoms that are not part of the lattice.

My guess is that individual atoms entering and leaving the formation happens on a much faster timescale than the formation switching between quasi-stable states.

It's almost as-if there are mini phase-transitions between formation sizes, it's fascinating.

From the article it seems like the 'springing into existence' is a process called nucleation: https://en.wikipedia.org/wiki/Nucleation.

I think the trouble is that the time scale you've assumed for molecular reactions is off; atoms move and react much faster than you may expect. (In air at room temperature, an individual molecule will undergo billions(?) of collisions per second.) The atoms and molecules in this video are moving on the order of 200 m/s, and the image is only about 5 nm wide. To see atoms slowly forming together, you need to watch on the scale of picoseconds at a time, not the 20 ms - 40 ms per frame in this video.

A molecular dynamics simulation might give a better impression of the time scale: https://www.youtube.com/watch?v=FK29kxMSqH8&t=2m45s. (I'm not sure what volume is being simulated there. It's much larger than what's seen in the submission, which is about 18 water molecules wide if we assume water molecule size of 0.27 nm.) Note the time scale on the left. The whole simulation is 250 ps, so you have to watch it about 100 million times to match a single frame in the submission's video.

Both modes are possible, if I read it correctly (see the bottom half of the first figure). Pieces of the lattice can "spring" into existence very quickly-- that's a phase transition. But, once formed, it can also "grow" atom by atom.

The grid appears suddenly, but strips of atoms keep piling on like those magnet ball toys, and it looks like it grows too big for its containment.

Maybe it grows slower as it gains more surface?