Ice crystal size is the whole of ice cream texture

Ice cream is mostly ice, and that is the problem

A frozen custard at serving temperature is a composite: ice crystals, air cells, fat globules and an unfrozen syrup holding all of it together. A substantial fraction of the water in it is frozen solid, and the sensation of smoothness has almost nothing to do with the flavour and everything to do with the dimensions of those crystals. The tongue starts detecting ice as grittiness somewhere in the region of fifty micrometres, and a good ice cream keeps most of its crystals well below that.

So the technical objective of every method, from a hand-cranked bucket to an industrial continuous freezer, is the same. Make the ice crystals small, make lots of them rather than a few large ones, and prevent them from growing afterwards. Everything else — the fat, the sugar, the stabilisers, the air — is in service of that.

Nucleation and growth are separate events

Crystals form in two distinct stages. Nucleation is the appearance of a new crystal from liquid water; growth is the addition of further water molecules onto a crystal that already exists. The two compete for the same water, and whichever dominates decides the outcome. Many nuclei forming at once gives a great number of small crystals. Few nuclei with plenty of time to grow gives a small number of large ones, which is exactly the sandy texture nobody wants.

Rapid cooling favours nucleation, because it drives the mixture deep below its freezing point before crystals have a chance to establish themselves and grow. Slow cooling favours growth. This single fact explains why a domestic freezer set to a moderate temperature makes worse ice cream than a machine with a properly frozen bowl, and why professional equipment cools the mix hard and fast against a very cold surface.

Churning is scraping, not stirring

The paddle in an ice cream machine is often assumed to be there to beat in air, and it does that, but its more important job is mechanical. Ice forms first against the cold wall of the drum, and the dasher scrapes that layer off continuously and disperses it through the warmer mix in the middle. Each scraped fragment becomes a seed distributed through the body of the mixture, and the wall is cleared so that fresh liquid can freeze against it.

Two things follow. A mixture frozen without agitation forms a solid block with large crystals throughout, which is why still-frozen custard is unpleasant while the same base churned is not. And the shear itself matters: the dasher breaks crystals as well as moving them. This is also the mechanism behind the improvised method of freezing a mixture in a tray and beating it thoroughly every half hour, which genuinely works and is genuinely inferior, because the intervals allow growth between interventions.

Sugar and fat both interfere with crystals

Dissolved sugar depresses the freezing point of the water it is in, so at any given temperature a portion of the water remains liquid. That unfrozen syrup is what keeps ice cream scoopable rather than solid, and it also physically separates ice crystals from one another so they cannot fuse. Increase the sugar and more water stays liquid; too much and the mixture never firms up at all.

Fat does something different. Fat globules are not part of the ice at all — they sit between the crystals and the air cells, obstructing crystal growth and coating the tongue in a way that reads as richness and masks whatever ice is present. A partially coalesced fat network also gives the ice cream structural body, which is why a high-fat base tolerates a slightly coarser crystal structure than a lean one and still tastes smooth.

Storage is where good ice cream is usually lost

Whatever the crystals looked like at the end of churning, they will not stay that way. In a domestic freezer, temperature fluctuates as the compressor cycles and every time the door opens, and each small warming melts the smallest crystals — which have the highest surface energy and go first — while their water refreezes onto the larger ones. This is recrystallisation, and it is a one-way ratchet towards coarseness. It is exactly why homemade ice cream is best on the first day and disappointing by the fourth.

The countermeasures are unheroic and effective. Store the container at the back of the freezer rather than in the door, keep the surface covered so that water cannot sublimate from it and refreeze as crystals on top, and use a shallow wide container that cools through quickly rather than a deep tub with a warm centre. Stabilisers help too, by holding on to free water so it is less available for the growth stage. None of it stops the process; all of it slows it down.