AI-TALKS.ORG
INTERACTIVE LABS
Physics · Numerical modeling · Interactive learning
Explore the mathematics of moving fluids

The Navier–Stokes Lab

Stir a virtual fluid, inject colored dye, and investigate how viscosity reshapes vortices and flow. The lab solves an approximate two-dimensional incompressible-flow model directly in your browser, inspired by the Navier–Stokes equations.

2D · INCOMPRESSIBLE ∇·u ≈ 0MODEL TIME 0.00GRID 128 × 128
Drag to stir and inject dye · Click to add a plume · Flow wraps around the edgesSpace pause · R reset

How does flow energy change?

Mean kinetic energy vs model time
With no external forcing, viscosity and numerical diffusion typically reduce the kinetic energy. Stirring introduces new momentum.
The governing equations

Where viscosity enters

A simplified form of the incompressible Navier–Stokes system is

∂u/∂t + (u · ∇)u = −∇p + ν∇²u + f
∇ · u = 0

The velocity u is transported by the flow itself; pressure p helps enforce incompressibility, and viscosity ν smooths velocity gradients. Dragging the cursor applies an external force f and injects passive dye. The computation uses semi-Lagrangian advection, numerical diffusion, and a Fourier-based pressure projection, broadly following the real-time Stable Fluids method of Jos Stam.

The Millennium Problem

What this lab cannot prove

The celebrated Navier–Stokes existence and smoothness question concerns sufficiently general three-dimensional solutions. In 2026, an AI-generated blow-up construction using specially chosen external forcing attracted major attention, but its ultimate mathematical assessment and relation to unforced flows require care.

This is a 2D, finite-resolution educational approximation, not a test of a 3D singularity or a proof of the Millennium Problem. Two-dimensional flows have fundamentally different mathematical regularity properties. Numerical grids cannot establish whether exact continuum equations develop singularities.

The Taylor–Green preset is a standard exact-solution benchmark: its unforced 2D velocity amplitude decays proportionally to exp(−8π²νt) on the unit periodic square. Real-time semi-Lagrangian simulations exhibit additional numerical dissipation.

Clay Mathematics Institute ↗ · September 2026 announcement ↗