Snowflakes - 3D crystal lab and winter night

Grow a snow crystal in a cold chamber, change the weather, then send it falling into the winter night.

Switch to 2D

Presets

Playback

Weather in the cloud

Temperature and humidity decide the shape. Try -15 C and lots of vapour for a fern star.

Advanced (model knobs)

The raw numbers of the Gravner-Griffeath growth model.

Camera

Layers

Quality

Model checks

The science

Why six sides?

Ice crystals grow from water molecules that lock into a hexagonal lattice (ice Ih). Water vapour reaches the crystal by diffusion, dc/dt = D ∇²c, and sticks where the surface is "rough" enough. Because the lattice repeats every 60 degrees, the growth rules look the same along six directions, so the crystal has six-fold symmetry.

Why does every flake look different?

All six arms ride through the same cloud, so they feel the same temperature and humidity at the same moment and grow the same way. A different path through the cloud (the weather history graph) gives a different pattern. The turbulence slider changes that path; the seed picks a new one.

The model (Gravner-Griffeath, 2008)

Each hexagonal cell holds vapour d, boundary ice b and crystal ice c. Every step: (1) vapour diffuses to its 6 neighbours with the 7-point average d' = (d + Σdnb)/7; (2) cells touching the crystal freeze a fraction: b += (1-κ)d, c += κd; (3) a boundary cell joins the crystal if b ≥ β (1 or 2 ice neighbours), if b ≥ 1 (3 neighbours) or always (4+ neighbours); (4) the rest melts back by μ. A low threshold β lets tips run away and branch into dendrites; a high β keeps the shape a flat hexagon.

Temperature and humidity

The sliders move β and the vapour supply ρ in a way that follows the Nakaya morphology diagram qualitatively: thin plates near -2 C, prisms and needles near -5 to -8 C, plates and fern-like stars near -15 C when the air is very humid, columns again below about -22 C. The real diagram comes from lab measurements (Libbrecht, Rep. Prog. Phys. 2005); this model reproduces the trends, not the exact numbers.

Honest simplifications

  • The growth is planar. Real crystals also grow in thickness; the 3D view only extrudes the plate (thin for plates, long for columns).
  • One lattice cell is drawn as 10 micrometres and one step has no fixed duration, so sizes and masses are illustrative (mass uses a 917 kg/m³ ice density).
  • The temperature-to-parameter map is a smooth qualitative fit, not a fit to measurements.

Checks run in this page

(1) The diffusion kernel spreads a point of vapour with mean-square radius ⟨r²⟩ = 6n/7 after n steps (lattice units), measured on the grid. (2) Six-fold symmetry: rotating the ice by 60 degrees maps it onto itself (exactly 100 % with imperfection off; the update rule never depends on cell order). (3) Vapour mass is conserved by diffusion.

What to try

  • Press Fern dendrite, then Flat plate. Same ice, very different air.
  • Set the temperature to -15 C and drag the vapour slider from low to high: plate, then star, then fern.
  • Turn Cloud turbulence up and press Grow a new flake a few times: six identical arms, but each flake is new.
  • Switch on Imperfection to see what dust and noise do to the perfect symmetry.
  • Look at the mirror lines layer: the pattern repeats across all three lines.
Loading model...
Loading Blender model...
Radius grows with time
Which shape forms? (temperature vs vapour)
How filled is each distance from the centre?
The weather this flake lived through
Made with 🟠 Blender · Three.js