Scaling and criticality: why dimension matters

Many PDEs are unchanged by a rescaling of space and time. If \( u \) solves the heat equation, so does \( u(\lambda x, \lambda^2t) \). If \( (u, p) \) solves the Navier-Stokes equations \( u_t + (u\cdot\nabla)u = \nu\Delta u - \nabla p \), \( \nabla\cdot u = 0 \) in \( \mathbb R^3 \), then so does \[ u_\lambda(x,t) = \lambda\,u(\lambda x, \lambda^2t), \qquad p_\lambda(x,t) = \lambda^2p(\lambda x, \lambda^2t), \] because every term (\( \partial_tu \), \( (u\cdot\nabla)u \), \( \Delta u \), \( \nabla p \)) picks up the same factor \( \lambda^3 \). Large \( \lambda \) zooms in on small scales and short times.

A norm is critical if it is unchanged by the scaling, subcritical if it shrinks when we zoom out and grows when we zoom in (\( \lambda \to \infty \)), and supercritical if it does the opposite. For the velocity at a fixed time, a change of variables gives \( \|u_\lambda(\cdot,t)\|_{L^p(\mathbb R^3)} = \lambda^{1 - 3/p}\|u(\cdot,\lambda^2t)\|_{L^p} \), so \( L^3 \) is critical: the example solves \( 1 - 3/p = 0 \). Similarly \( \dot H^{1/2} \) is critical. The energy, \( \|u\|_{L^2}^2 \), scales like \( \lambda^{2-3} = \lambda^{-1} \), and the dissipation \( \int_0^\infty\|\nabla u\|_{L^2}^2\,dt \) scales the same way.

Here is the heuristic that organises the whole problem. A possible singularity is a concentration at small scales, and zooming in on it means \( \lambda \to \infty \). A quantity we control that grows under this zoom (subcritical) makes a concentrated singularity cost more and more of it, so it forbids blow-up. A critical quantity sits on the borderline and can be enough with extra structure: that is the two-dimensional situation, where the energy scales like \( \lambda^{2-2} = 1 \) and the vorticity obeys a maximum principle. In three dimensions the energy scales like \( \lambda^{-1} \): a singularity can be arbitrarily concentrated at arbitrarily small energy cost. Every known global bound is supercritical, and closing that gap is the Navier-Stokes problem.

Picture it: a magnifying glass over a developing singularity. In two dimensions the energy seen through the glass stays the same at every magnification, and it caps what can happen. In three, the more you magnify, the less energy is needed to fill the view.

Think it: conditional regularity results sit exactly on the critical line. The Ladyzhenskaya-Prodi-Serrin condition says a Leray-Hopf weak solution in \( L^q_tL^p_x \) with \( \frac2q + \frac3p = 1 \), \( 3 < p \le \infty \), is smooth; the scaling of that norm is \( \lambda^{1 - 2/q - 3/p} = \lambda^0 \). Escauriaza, Seregin and Šverák reached the endpoint \( L^\infty_tL^3_x \). Scaling will not solve the problem, but it tells you at a glance whether an idea has a chance: an argument that only uses supercritical information cannot work without a new structural insight.