The Million-Dollar Question About Water: Can a Fluid Reach Infinite Speed?

The equations that describe how water flows were written down in the 1820s. We still don't know if they always give sensible answers. That gap—between a 200-year-old equation and a question no one can answer—is one of the seven Millennium Prize Problems, worth a million dollars. Here is exactly what the question is, and why it resists everyone.

The equation and the fear

The Navier-Stokes equations describe the motion of a fluid—water, air, blood. Give them a starting state (the fluid's velocity at every point right now) and they tell you how the flow evolves the next moment, and the moment after, forever. They are Newton's law, force equals mass times acceleration, rewritten for a continuous fluid instead of a solid ball.

The million-dollar question: if you start with a smooth flow, does it stay smooth forever?

"Smooth" has a precise meaning. The velocity of the fluid is finite everywhere, and so are its rate of change and its curvature—no sudden jumps, no infinite spikes. A smooth flow can still be turbulent, chaotic, violent. But at every point the speed is a finite number.

The failure case is a singularity (or "blowup"): a moment where the equations predict the speed at some point reaches infinity in finite time. Not fast—infinite. If that can happen, the equations have broken themselves. No real fluid moves at infinite speed, and past that instant the mathematics can no longer say what happens next. So the real question is: does a model we trust to describe reality secretly contain an instant where it stops making sense?

Why this is genuinely hard, not just unsolved

You might think: just run the equations and watch. Two things block you.

First, the equation contains a fight between two terms.

• The nonlinear term describes fluid pushing on fluid—fast bits dragging slow bits, motion concentrating into ever-tighter whirls. This term wants to build a singularity. It funnels the energy of a large region into a shrinking one, and the tighter it packs, the harder it pushes. That feedback is what could, in principle, drive speed to infinity.
• The viscosity term is internal friction, the fluid's stickiness. It spreads concentrated energy out and resists sharp features. It fights the singularity.

Whether a flow blows up is the outcome of this war. Here is the quantitative heart of it. Both terms scale with size, but at different rates. Squeeze the active region by a factor of ten and the nonlinear term's strength grows faster than viscosity's ability to smooth it. So the crucial question becomes: as the flow concentrates into smaller and smaller regions, does the nonlinear term eventually outrun viscosity and run away to infinity—or does viscosity's smoothing catch up in time? In gentle flows viscosity clearly wins. In the worst imaginable flows, no one has proven either side must win.

Second, our tools only work at low intensity. The main tool is energy accounting: total kinetic energy in the fluid can only decrease (viscosity drains it), so it stays bounded. That bound is enough to control a gentle flow forever, and any flow for a short time. But energy is a total—summed over the whole fluid. A singularity is local: energy piling into a vanishing speck. A finite total says nothing about whether an infinite spike is forming at a single point. Exactly when the flow turns violent enough to matter, the accounting that reassured us goes silent. That silence is the open problem.

The move Buckmaster and collaborators made

For decades the field tried to prove blowup can never happen. Recent work goes the other way: it hunts for a singularity on purpose and tries to build one by hand.

The strategy has a name worth carrying: self-similar blowup. Suppose a fluid tearing itself apart does so in a scale-repeating way. Zoom into the exploding region, rescale the picture—shrink the ruler and speed up the clock in a matched ratio—and it looks identical to before: the same shape of flow, just smaller, faster, more intense. Like a whirlpool whose every close-up is a copy of the whole, all the way down to a point.

Why guess this particular shape? Because it is precisely the structure the scaling fight predicts. If the nonlinear term keeps winning by the same margin at every scale, the flow reproduces itself at every scale—self-similarity is what a sustained runaway would look like.

And it converts an impossible task into a tractable one. Instead of tracking an infinitely complex flow through the instant of catastrophe—a moving-target question—you ask: is there a single fixed profile that reproduces itself under rescaling? That is a static question, about one shape rather than an evolving movie. Static questions can sometimes be answered, sometimes by computer: search numerically for a self-similar profile, then prove rigorously that an exact one exists near the computed guess.

The breakthroughs found such profiles for close cousins of the full equations—simplified fluid models, and versions without viscosity (the Euler equations, fluid with no internal friction), where the nonlinear term has no opponent and the machinery pushes through. The full Navier-Stokes equations with viscosity remain open; viscosity's smoothing is exactly the term that resists these constructions. But the enterprise shifted from "prove nothing bad ever happens" to "construct the bad thing explicitly." An impossible-to-verify absence became a buildable object.

What you should take from this

Two portable ideas, both good far beyond fluids.

One: the hard problem is almost never the equation—it's the range. Navier-Stokes is fully known. What's unknown is whether its good behavior at low intensity survives to high intensity, because the controlling quantity (total energy) is blind to what happens locally at the extremes. Huge stretches of hard science and economics have this exact shape: a model provably fine in the tame regime, the danger living entirely in the extreme regime, and a canyon between where the reassuring accounting quietly stops applying. When someone says a system is "well understood," ask: understood across its whole range, or only where it's gentle? Then ask which quantity is doing the reassuring, and whether that quantity can see the failure mode. Often it can't.

Two: to prove something can happen, try to build it. When "prove it never occurs" stalls for decades, flip to "construct one instance." Constructing forces you to name the exact shape of the catastrophe, and a named shape is something you can compute, test, and pin down. This is a general research reflex—in mathematics, engineering, strategy, security. If you can't rule out the disaster, try to engineer it in miniature. Either you succeed and now know its anatomy, or you fail in a way that reveals exactly what protects the system.

The million dollars is still unclaimed. The more useful thing—knowing where in a trusted model the danger hides, and how to hunt it—is already in hand.

Distilled from Hacker News