a critical exponent

In nonlinear PDEs, the strength of the nonlinearity is often tuned by a single number — a power p in a term like u^p. As you turn that dial, the equation's behaviour can change not gradually but suddenly: below a certain value of p solutions are tame, above it they misbehave. The exact value at which the behaviour flips is a critical exponent. It is one of the most important kinds of number in the field, because it draws a sharp line between two completely different worlds.

There are several distinct critical exponents, each marking a different threshold. The Fujita exponent p_F = 1 + 2/n (for the semilinear heat equation u_t = Laplacian u + u^p in n space dimensions) separates two blow-up regimes: for 1 < p <= p_F EVERY positive solution, however small, blows up in finite time, while for p > p_F small initial data can decay and exist globally. The Sobolev critical exponent 2* = 2n/(n-2) marks where the Sobolev embedding H^1 into L^q just barely fails to be compact — at exactly this power the variational methods that prove existence run into trouble (loss of compactness), and the famous Yamabe and prescribed-curvature problems sit right at it. The unifying idea is SCALING: each critical exponent is the power at which the nonlinear term and the principal (linear) term scale the same way under the rescaling u(x,t) -> lambda^a u(lambda x, lambda^2 t), so that neither dominates — that balance is precisely what makes the critical case delicate.

Why does it matter? Critical exponents are where the easy arguments break. Below critical (subcritical) you usually have good existence theory; above critical (supercritical) you expect blow-up or non-uniqueness; AT critical the standard estimates lose just enough room that the problem becomes genuinely hard and often open. Recognizing whether your nonlinearity is subcritical, critical or supercritical — usually by a scaling count — is the first thing an expert checks, because it predicts whether the methods you know will work at all.