Driving force. Supersaturation σ = c/ceq(T) − 1 with c in g of anhydrous solute per 100 g water and a fitted solubility curve ceq = c20 ek(T−20) (rounded textbook values). Cooling a saturated solution raises σ; the crystal then consumes the excess.
Face growth (Burton-Cabrera-Frank). Spiral growth at screw dislocations gives Ri = vi C σ2 tanh(σ1/σ): parabolic (∝σ2) at low σ, linear at high σ. Each face family has a relative factor vi. The normal distance of every plane advances, dhi/dt = Ri, and the crystal is the convex region {x : ni·x ≤ hi} (vertices from plane triples). Frank-Chernov: the slowest faces survive and hi/hj → vi/vj, which sets the habit.
Mass balance. Volume from the divergence theorem V = (1/3) Σ hiAi; each gram of crystal removes w g solute and (1−w) g water (alums are dodecahydrates). Evaporation removes water. Above the metastable limit (σ > 0.15) spontaneous nucleation sheds a snow of fines, and growth gets cloudy (kinetic roughening).
The 2D view. The picture is the exact mid-section of the convex habit: every growth plane i is the line n_i . x = h_i in the cut plane, and the section polygon comes from clipping a large square by each half-plane. Each ring is a past outline coloured by the supersaturation at that moment (blue low, green metastable, orange labile), so rings read like growth zoning in a real crystal. The lower-left chart is the solubility curve with the jar path on it.
Honest simplifications. Well-mixed solution (no diffusion boundary layer or convection), temperature uniform, BCF constants chosen for realistic mm/day rates rather than measured per mineral, ideal activities, no inclusions/twinning/dislocation limits; dissolution is a linear retreat of all planes; hidden faces keep advancing at their own rate. Solubility parameters are rounded literature values. Do not taste lab chemicals: only table salt, alum and KDP are real household-lab safe-ish crystals, and children need an adult.
Independent checks (run on load):
| running… |