This is a diffusion-limited aggregation (DLA) model: individual solute particles enter the fluid at a ring around the growing crystal and take a random walk, cell to cell, until they either wander off (returning to solution) or land next to an already-crystallized cell. When that happens they attach with a probability instead of automatically — the same rule real crystal growth follows, where not every collision with the interface results in a bond:
p_stick = clamp[ (0.95 − T/120) · (0.55 + 0.45·S/100) · aniso(θ), 0.04, 0.95 ]
Higher supersaturation (S) pushes more particles into the fluid per second and raises the attachment odds — a stronger concentration gradient drives faster growth. Higher temperature (T) lowers the sticking probability, standing in for increased re-dissolution / surface mobility at the interface, which is why hot, dilute solutions tend to grow slower and more compact crystals than cold, saturated ones.
The facet anisotropy slider biases attachment toward six crystallographic directions 60° apart (aniso(θ) above). At 0% every direction is equally likely and the crystal grows as a smooth, compact disk. Near 100%, growth strongly favors those six directions and starves the gaps between them of new bonds, which is exactly how faceted dendritic branches emerge from an isotropic random walk.
- Bright core → dim edge — cell brightness encodes the order cells attached in, oldest (core) brightest.
- Cyan dots — solute particles still diffusing through the fluid, not yet part of the lattice.
- Multiple seeds — several nucleation sites compete for the same dissolved solute and can merge into one crystal.