Look closely at the center of a snow crystal. In many, or most, you will often not find a droplet center as we described in the previous post. Indeed, for the columnar crystals, you may never see a droplet center. As an example, look at the center of the dendrite crystal below (one of Mark's):

Why? Do these not also start on droplets?
Or are the droplets just too small?

This crystal, like all snow crystals did start on a frozen droplet, but the vapor deposited all around the frozen droplet (also called a 'droxtal'). The resulting growth can vary quite a bit, but roughly proceeds in a way first described in detail by the physicist Sir Charles Frank:

The crystal will still develop its two levels, top and bottom, just like we saw in the previous post, but in this process, it takes a little longer for the two levels to develop. The first stages involve the frozen droplet gaining flat faces (see steps (i) to (ii) below). Initially, as in (ii), there are more than 8 faces, but the fastest-growing faces soon vanish (they essentially grow themselves out of existence), leaving only the slower basal and prism faces in (iii):

The slower-growing faces win the game and are allowed to reveal themselves. With snow, we learn that sometimes slow wins.

In the above sequence, follow the red arrows for the time sequence. (From "(a)", I am borrowing some sketches I made for a publication in 2008. These differ slightly from those originally proposed by Frank.) The crystal sizes here are very small, about 1/10 to 1/2 the width of typical scalp hair (i.e., about 7 to 40 micro-meters). The axes on "(a)" show the crystallographic axes for ice, remnants of my paper.

You can see though that by stage (iii), the form of the original droplet has vanished. But let's continue, along the path to a recognizable branched crystal, roughly as described by Frank.

The crystal face grows outward by the spreading of layers. In the sketch above, the edge of these layers, called "steps", are shown by fine lines in the front view at right. The steps start near a corner, say at "A" or "C" in the sketch, and spread inward, closing in on the center. But soon a problem emerges: the steps are coming too fast and the center region cannot keep up. So a small pit emerges in the center of the face as you see at (b). On a symmetric hexagonal form like that above, this pit forms roughly at the same time on all six "prism" faces. (Pits may also form on the top and bottom faces, but for a crystal shape like that above, will not become very deep.) Here "SCR" stands for step-clumping region.

Typically, the rim of this pit expands as the crystal grows (c).

And grows. Eventually, the rim breaks through a crystal edge. Usually, it is the edge between two prism faces as shown below.

Notice that at stage (d), the crystal now has two levels, a top side and a bottom side. From this stage on, the two levels will compete with each other, with one eventually getting much larger than the other. This same process was shown in the previous post. But let's continue on and see how this develops.

Either one, or both levels may sprout branches. In the case (e) above, both have sprouted (note the notches that divide the just-sprouted branches). So far, both levels are developing at the same rate. But that cannot continue: the situation is unstable because if one level gets just a little bit ahead, it will stay ahead and increase its size difference with time. In fact, the smaller level will hardly grow at all - it is stuck in a region largely devoid of vapor excess.

Typically, it is the bottom level that gets ahead, as shown in (f) above because the crystal is falling down through the onrushing vapor-laden air. (The interior lines labeled in (f) above will be discussed another day.) Looking directly down on this crystal, it would look roughly like the following.

Consider that every crystal is a little bit different, but the basic features in (f) exist in the original picture at the top. I reproduce it below with a close-up of the center.

Some of these finer details will be discussed in a later post.
Finally, as to why this process sometimes occurs but sometimes the process of the previous post occurs instead remains a mystery. There are a few factors that will favor this process though: smaller initial droplets, and conditions of slower growth.

-- JN