Turing Diffusion: Reaction-Diffusion Patterns and Morphogenesis

Turing's 1952 morphogen idea

Alan Turing's 1952 paper "The Chemical Basis of Morphogenesis" tackled a deep puzzle: how does a nearly symmetric fertilised egg give rise to an organism with precise spatial structure? Turing proposed morphogens — diffusible chemicals whose concentration gradients carry positional information — and showed that a system of two interacting, diffusing morphogens can be unstable to small spatial perturbations, even though each process alone (diffusion, or uniform reaction) would be stable. This Turing instability is counterintuitive: diffusion is normally a smoothing process, yet here it creates structure instead of destroying it. The paper was largely ignored during Turing's lifetime and only recognised as foundational in the 1970s.

The activator-inhibitor mechanism

The general framework couples two concentrations, activator u and inhibitor v, through reaction-diffusion equations. The critical requirement is differential diffusion: the inhibitor must diffuse markedly faster than the activator. Where the activator is locally elevated, it amplifies itself (self-activation) while sending the faster inhibitor diffusing outward to suppress activation in the surrounding region — local activation, lateral inhibition — which sets a characteristic spacing between pattern elements. The Gierer-Meinhardt model (1972), built around hydra head/foot regeneration, supplies explicit kinetics with the nonlinear ratio u²/v that saturates the pattern at finite amplitude.

The Gray-Scott model and its parameter zoo

The Gray-Scott model (1984) models an autocatalytic reaction where U converts to V in the presence of V, while V slowly decays. Two parameters — feed rate F (replenishes U) and kill rate k (removes V) — control an extraordinarily rich phase diagram: (F,k)=(0.035, 0.065) gives spots, (0.037, 0.060) gives worms, (0.029, 0.057) gives spirals, (0.040, 0.060) gives labyrinthine stripes, and (0.025, 0.055) produces "mitosis" — localised spots that elongate and split, mimicking cell division.

∂u/∂t = Du·∇²u − u·v² + F·(1 − u)
∂v/∂t = Dv·∇²v + u·v² − (F + k)·v

F=0.035, k=0.065 → spots      F=0.025, k=0.055 → mitosis
F=0.037, k=0.060 → worms      F=0.029, k=0.057 → spirals

Confirmed in real animals

Murray (1988) showed that solving reaction-diffusion equations on domains of different shapes reproduces leopard spots, zebra stripes, and the way a thin cylindrical tail turns longitudinal stripes into rings. The clearest molecular confirmation came from Sick et al. (2006, Science), who identified WNT as the activator and DKK1 as the inhibitor spacing mouse hair follicles — perturbing DKK1 changed the spacing exactly as predicted. Nakamasu et al. (2009) later showed zebrafish melanophores and xanthophores satisfy the same local-activation/long-range-inhibition rule in stripe formation. The same mathematics also governs the Belousov-Zhabotinsky chemical reaction, sand dunes, and vegetation bands in arid ecosystems.

Frequently asked questions

Try it live

Everything above runs in your browser — open Turing Diffusion and tune the feed rate F and kill rate k to navigate Turing's morphogenesis landscape from spots to mazes to coral. Nothing is installed, nothing is uploaded.