Mixing happens in two stages. First, a flow stretches and folds the scalar field — the concentration of dye, the temperature distribution, whatever is being mixed. The stretching creates thin filaments with large gradients. Second, molecular diffusion acts on those gradients, smoothing the filaments and completing the homogenization. Without diffusion, the filaments get thinner forever but never merge. Without flow, diffusion alone is slow — it operates at the molecular timescale, which is far too long for practical mixing. The standard story is that both stages matter and the mixing rate depends on both.
Liss and Luan (arXiv:2603.09238, March 2026) prove that for parallel shear flows, the mixing rate does not depend on diffusion at all. The rate is uniform in diffusivity: the same sharp mixing rate holds whether diffusion is strong, weak, or arbitrarily close to zero. The flow alone sets the pace.
The mechanism is the geometry of the shear. A parallel shear flow has critical points — positions where the shear velocity has zero derivative. Near these points, the flow stretches the scalar field at rates determined by the shear profile's curvature. The filaments thin at a rate set by the flow. As long as the shear has finitely many critical points, this stretching dominates the mixing dynamics regardless of the molecular diffusivity. Diffusion finishes each filament, but the rate at which filaments are created and thinned is independent of how fast they blur.
The proofs use two different perspectives. The first represents the advection-diffusion equation stochastically — particles follow the shear flow plus Brownian noise — and shows by integration by parts that the mixing estimate is independent of the noise amplitude. The second treats the system as a dynamical system and shows that the flow geometry controls the mixing rate even in the zero-diffusivity limit. Both arrive at the same conclusion: the mixing rate is a property of the flow, not of the fluid.
The physical implication is clean. In shear-dominated mixing, diffusion is irrelevant to the rate. It is necessary — without it, mixing never completes — but it is not rate-limiting. The flow creates the conditions; diffusion exploits them. And the conditions are created at the same speed regardless of how fast they are exploited. The bottleneck was never the molecules. It was the geometry.
Liss and Luan, "Uniform-in-diffusivity mixing by shear flows: stochastic and dynamical perspectives," arXiv:2603.09238 (March 2026).