Pouring the agar…
🧪 Petri dish controls
2D GPU solver with the F-k map →
The science

Model. Gray-Scott reaction-diffusion: ut = Du∇²u − uv² + F(1−u), vt = Dv∇²v + uv² − (F+k)v. The autocatalytic step U + 2V → 3V makes v from u; F supplies fresh u and flushes the dish, k removes v. Tap or click the agar to drop dye (v). The height and the blue-violet colour of the agar show v.

Why spots and stripes. Turing (1952): two substances that diffuse at different speeds (Du = 2 Dv here) can turn a uniform mixture into patterns. In this model most patterns spread from a finite-size seed invading the empty state u = 1, v = 0, which is stable to small noise, so the linear Turing band alone does not explain them.

Numerics. Explicit Euler with the 9-point isotropic Laplacian (weights 0.2 / 0.05 / -1), a 128 x 128 grid with a round no-flux dish wall, grid units dx = dt = 1, Du = 0.2097, Dv = 0.105; stable while the diffusion number stays below the explicit limit. It is the CPU float32 path of the very same solver (gscore.js) that drives the 2D page on the GPU; the 2D page cross-checks the two paths.

Honest simplifications. The relief is exaggerated and only mirrors the catalyst v; real Gray-Scott experiments use gel reactors (Lee et al. 1993), not agar plates, and a real petri dish has no feed flow. Grid units are not tied to millimetres, and the bench, dish and lid are an illustration of the scene, not part of the model.

Independent checks (run on load, the same core drives the 2D page):

running…

-

Steps-
Dye area-
Mean u / v-
Pattern scale-
Uniform states-
Orbit with a drag; tap the agar to drop dye.
Made with 🟠 Blender · Three.js