the Turing instability
/ TOOR-ing /
Here is one of the most counterintuitive facts in all of applied mathematics, due to Alan Turing in 1952. Diffusion is the great smoother: leave a system alone and diffusion erases bumps, flattening everything toward uniformity. Yet Turing showed that diffusion can do the OPPOSITE — it can take a perfectly uniform, stable chemical mixture and make it spontaneously break up into spots and stripes. This is how a leopard might get its spots and a zebra its stripes: not by a blueprint, but by a pair of chemicals diffusing at different rates.
The setup is a two-component reaction-diffusion system, u_t = D_u Laplacian u + f(u, v), v_t = D_v Laplacian v + g(u, v), with a uniform steady state (u*, v*) where f = g = 0. Suppose that WITHOUT diffusion this steady state is stable — perturb the well-mixed chemicals and they return to equilibrium. The Turing miracle is that ADDING diffusion can destabilize it, provided one species (the inhibitor) diffuses much faster than the other (the activator). The mechanism: the activator promotes itself and the inhibitor locally; but because the inhibitor spreads far and fast while the activator stays put, you get 'short-range activation, long-range inhibition'. A small clump of activator boosts itself locally and seeds a surrounding ring of inhibition, which fixes a characteristic spacing. The mathematics: linearize, look for modes e^(i k x), and find that for a band of nonzero wavenumbers k the growth rate becomes positive — those patterns grow, selecting a finite wavelength.
Why is this profound? It overturns the intuition that diffusion only smooths, and gives a purely physical, parameter-driven explanation for biological pattern WITHOUT a designer. The required ingredient — a large diffusion ratio D_v / D_u between inhibitor and activator — is a genuine, testable condition, and it is the reason real Turing patterns were hard to find in chemistry for forty years (most molecules diffuse at similar rates). Honest caveat: Turing's mechanism is one route to pattern, elegant and real, but not the only one, and showing it actually operates in a given organism is subtle.
In an activator-inhibitor pair, suppose the inhibitor v diffuses ten times faster than the activator u (D_v / D_u = 10) and the kinetics are balanced so the well-mixed state is stable. Linearizing reveals a band of wavenumbers k whose growth rate becomes positive: perturbations of those scales grow while longer and shorter ones decay, so the system settles into a regular pattern with a fixed spacing — spots if the domain is two-dimensional.
Fast inhibitor + slow activator: diffusion makes patterns instead of erasing them.
A single reaction-diffusion equation can NEVER show a Turing instability — you need at least two coupled species with sufficiently different diffusion rates. With equal diffusion rates the mechanism fails entirely; the differential diffusion is the whole point.