Open Problems and the Navier-Stokes Millennium Prize

The one PDE worth a million dollars

Across this rung you have watched PDEs run the physical world: Euler and Navier-Stokes for fluids, the Schrodinger equation for quantum particles, Black-Scholes for option prices, Maxwell's equations for light. Most of them we cannot solve in closed form, yet we trust them, simulate them, and bet on them. This last guide asks a sharper question: for these equations, do solutions even exist for all time, and stay smooth — or can they blow up? For one of them the honest answer is we genuinely do not know, and the Clay Mathematics Institute will pay one million dollars to whoever settles it.

The star of the show is the Navier-Stokes Millennium Problem, one of the seven Millennium Prize problems announced in 2000. In words: take the incompressible Navier-Stokes equations in three space dimensions, start from any smooth, finite-energy velocity field, and let it flow forever. Does the solution stay smooth and finite for all time? Nobody has proved that it always does, and nobody has produced a single example where it fails. That gap — not a hard calculation but a missing theorem — is the prize.

Why a fluid equation might tear itself apart

To feel the danger, strip the physics down to one term. The Navier-Stokes momentum equation, schematically, reads u_t + (u·grad)u = nu Laplacian u - grad p. The viscous part nu Laplacian u is a friend: it is exactly the smoothing of a heat-type operator, draining sharp gradients and spreading energy out. The villain is the nonlinear self-transport (u·grad)u — velocity carrying velocity. That term can pile fluid faster than it can spread, concentrating the flow into ever-thinner, ever-faster filaments. The open question is whether viscosity always wins the race, or whether the nonlinearity can win in finite time and drive the velocity to infinity at a point: a finite-time blow-up.

This is not idle worry. Two facts make the question genuinely hard. First, the cousin equation — the inviscid Euler equations, with nu = 0 — is itself open for smooth blow-up in 3D, and there is mounting numerical and analytical evidence that some Euler flows do blow up. Second, the related scalar story you met earlier — a smooth profile in a conservation law steepening into a shock in finite time — proves that smooth data really can produce singularities in nonlinear PDEs. Navier-Stokes lives between those two: more regular than Euler because of viscosity, but in 3D nobody can prove viscosity is enough.

Weak solutions: a partial answer we already have

We are not empty-handed. Back in 1934 Jean Leray proved that 3D Navier-Stokes always has a weak solution for all time — a solution that satisfies the equation in the averaged, integrated sense you met when distributions let derivatives land on smooth test functions instead of on rough data. Leray's solutions live naturally in a Sobolev space: they carry finite energy and obey an energy inequality, but they are not yet known to be smooth, and they are not yet known to be unique. So existence-for-all-time is settled; it is regularity and uniqueness that remain open.

This pattern should feel familiar. Throughout this ladder, when a nonlinear equation refused to keep smooth solutions, the cure was to widen the meaning of solution and then re-impose order. For conservation laws, after a shock forms the weak solution is not unique until you add an entropy condition that picks the physically correct one (the Kruzhkov theorem then makes it unique). For first-order Hamilton-Jacobi equations the analogous selector is the viscosity solution. For Navier-Stokes the dream is the same — find the extra criterion that makes the weak solution unique and smooth — but that criterion, in 3D, has not been found.

Existence (Leray, 1934):  3D Navier-Stokes has a global weak solution.
Regularity (open):        Is that weak solution always smooth?
Uniqueness (open):        Is the global weak solution unique?

The Prize = prove (smooth & unique for all time)  OR  find a smooth
            finite-energy datum whose solution blows up.

Open problems are about questions, not just answers

Navier-Stokes is the famous one, but it is far from alone, and the surrounding open problems sharpen what "open" even means. Smooth blow-up for 3D Euler; sharp regularity criteria (we know that if the velocity stays bounded in certain Sobolev norms then it stays smooth — but we cannot prove the norm stays bounded); the long-time behaviour of dissipative systems and whether they settle onto a finite-dimensional global attractor; energy-supercritical wave and nonlinear Schrodinger equations, where focusing nonlinearity can concentrate energy and blow up. Each is a precise mathematical question hiding behind an everyday phenomenon — turbulence, a breaking wave, a collapsing laser pulse.

Here is the honest texture of the field, and the through-line of this whole ladder: the deepest obstacle is almost never "this integral is hard." It is well-posedness itself — does a unique, stable solution exist, and does it depend continuously on the data? Recall that running the heat equation backwards, or prescribing Cauchy data for Laplace's equation, is genuinely ill-posed: tiny changes in the data explode the answer, so no clever method can rescue them. Open problems like Navier-Stokes are, at heart, the same species of question — *we do not yet know which regularity class makes the 3D problem well-posed for all time.*

How people actually attack it

You will not win the prize by brute force, but the strategies are understandable, and you have seen the ingredients of every one. Here is the toolkit, in the order a working analyst reaches for it.

1. Find a conserved or decaying quantity (an a-priori estimate). Multiply the equation by u, integrate over space, and watch terms cancel. Just as conservation of energy controlled the wave equation, an energy bound is the first thing you try; the trouble is that in 3D the known bounds are too weak to forbid blow-up.
2. Hunt for a regularity criterion. Prove a conditional theorem of the form "if some norm stays finite up to time T, the solution is smooth up to T." These exist; the open part is showing the norm really does stay finite.
3. Probe with scaling and self-similar solutions. Navier-Stokes is invariant under a rescaling of space, time, and velocity; if blow-up happens it should look self-similar. Searching for (or ruling out) such profiles is a major line of attack.
4. Simulate at the edge of resolution. High-resolution computation can suggest where vorticity concentrates and whether it accelerates — though a finite grid can never prove a continuum singularity, it tells you where to look and what to conjecture.

Notice that nothing in this list is exotic to you now. Energy estimates, Sobolev norms, weak and viscosity solutions, scaling, the interplay of smoothing against nonlinear steepening, and the discipline of asking which data make a problem well-posed — these are the exact instruments you assembled climbing this ladder. The Millennium Problem is hard not because it uses ideas you have never met, but because the ideas you have met do not yet reach all the way. That is what a real frontier looks like.

Where this leaves you

You started this ladder unable to say what a partial derivative meant in an equation; you end it able to read the most famous unsolved problem in the subject and say exactly what is and is not known about it. That is not a small thing. The same three questions — existence, uniqueness, continuous dependence — that classified a quadratic discriminant B^2 - A C into elliptic, parabolic, and hyperbolic types are the very questions that stand, still open, at the research frontier. Mastery of PDEs is not knowing every answer; it is knowing precisely which question is being asked, and why it is hard.

And the frontier is not roped off. The conditional regularity theorems, the scaling arguments, the weak and viscosity solution frameworks — these are the working language of current research, and they are the language you now speak. If the Navier-Stokes problem still feels out of reach, that is not a sign you have learned too little; it is a measure of how genuinely hard the question is. You now stand on the same ground as the people trying to solve it.