Episode 69 · The Machine's Theorems, part 1 of 20 (Arc 1: Prerequisites)

What a PDE Is: Equations That Remember Their Neighbours

Sankalp Thakur · 59 minutes

A flow-assurance engineer's CFD solver blows up at two in the morning, and she asks a retired applied mathematician whether that is numerics or physics. Through their weekly calls, the episode builds the basics needed for the Navier-Stokes arc. ODEs remember their own past, while PDEs remember their neighbours. Transport moves shapes along, and its nonlinear form (Burgers) forms shocks in finite time. Heat smooths, and gives the square-root-of-time diffusion law. Waves echo, and Laplace's equation describes equilibrium. The episode covers Hadamard's three conditions (existence, uniqueness, continuous dependence) and why inverse problems are ill-posed. Dimensional analysis leads to the Navier-Stokes scaling symmetry, u of x and t going to lambda u of lambda x, lambda squared t. The heat equation's maximum principle has no analogue for 3D fluid speed. The energy identity shows that bounded energy does not mean bounded speed. The toy ODE y-dot equals y-squared blows up at time one over y-nought. The episode also shows how to tell numerical blowup (CFL violations, checkerboarding, energy creation) from genuine mathematical singularities. Farah Al-Mansoori and Augustin Lévêque are fictional characters. All quotations from real people come from the sources listed in the verification ledger.

Listen to the show on your preferred podcast platform.