Navier–Stokes, Live — Can a Whirlpool Break the Math?

THE INTUITION

Follow the water — drop dye into a smooth, viscous flow and watch it fade.

FOLLOW THE WATER.

Imagine flicking specks of dye onto the surface. Each speck rides the water; its trail is the path the fluid took. The arrows are the velocity field — direction and speed at every point.

This is a Taylor–Green vortex, one of the few flows where Navier–Stokes can be solved by hand. Four counter-rotating cells, each decaying at exactly the same rate. It never breaks: this example stays smooth forever.

WHY DO THE SWIRLS SLOW DOWN?

Viscosity is the fluid's internal friction. With nobody stirring, the motion bleeds away — the peak speed falls as e^−2νt. Raise the friction and the same moment in time looks calmer.

FIELDu = sin x cos y · e^(−2νt)
v = −cos x sin y · e^(−2νt)

DOMAINPeriodic [0, 2π]² · viewed from above

TRACERSRunge–Kutta 4 integration

REGULARITYSmooth for all t — decays, never blows up

LIVE SIMULATIONSIM/01 · 2D TRACERS

PEAK SPEED 100% OF INITIAL · t = 0.0

INTERNAL FRICTION ν0.08

LESS SMOOTHINGMORE SMOOTHING

TIME0%

STARTLATE

Watch one dot. It circles its cell as the flow slows. Longer arrows mean faster motion. Drag the time slider to scrub; press play to let the fluid run.

THE PROBLEM

Does friction always win?

DOES FRICTION ALWAYS WIN?

In two dimensions it does — the answer has been known for decades. In three dimensions something new is possible: a vortex tube can be stretched, and stretching it thins it and spins it faster, the way a skater pulling in their arms spins faster.

So the race is on. Vortex stretching concentrates motion into a smaller region; viscosity spreads it back out. Does smoothing always keep up? Or can speed outrun every finite bound as some particular moment approaches?

WHAT "SINGULARITY" MEANS HERE

Unbounded speed would be a singularity — a point where the mathematical solution stops making sense. Turbulence alone is not that: a turbulent flow is messy but every speed in it is finite. And "smooth" is a statement about mathematical regularity, not calm water.

2DGlobal regularity proven — vorticity is only carried, never stretched

3DOpen since 1934 (Leray); a Clay Millennium Prize problem since 2000

THE ENGINEVortex stretching: (ω · ∇)u

THE BRAKEViscous diffusion: ν∇²ω

LIVE SIMULATIONSIM/02 · STRETCH vs SMOOTH

CORE SPEED 1.00× · CORE WIDTH 100%

STRETCHING RATE1.00

NONEVIOLENT

FRICTION ν1.00

THINSYRUP

A toy race, not the equations. A vortex tube is stretched along its axis (narrower, faster) while friction widens and slows it. Tilt the two knobs and watch which side wins: a bounded plateau, or a run toward the top.

THE CONSTRUCTION

A shrinking core. A rising speed.

A SHRINKING CORE. A RISING SPEED.

In the reported construction, fluid spirals in through the sides and is flung out along the axis. The intense core gets both narrower and shorter — but its width shrinks faster than its length, so motion is squeezed into an ever more slender thread.

The schematic keeps the proportions honest in spirit and exaggerated in scale. Follow a copper bead along a strand: it drifts outward along the axis even as the bright region it lives in collapses.

WHAT DO THE OUTER RINGS DO?

Where the core meets the surrounding flow, an imbalance grows. Surrounding pulses redistribute momentum to cancel it; what those corrections leave behind is a smooth external force — the f in the equation. Toggle the rings to see where they act; their shape here is schematic.

▸ THE CONSTRUCTION IN THE PAPER

LIVE SIMULATIONSIM/03 · 3D SCHEMATIC · DRAG TO LOOK

AXIAL STRETCHING ↑ · INWARD SPIRAL · 0% TOWARD SINGULAR TIME

TOWARD THE SINGULAR TIME0%

EARLYτ = 10⁻⁶

Not a numerical solution. Dashed circles mark the initial core size. The progression approaches the deadline logarithmically and stops at τ = 10⁻⁶. Strand colours identify strands, not speeds.

THE KEY DISTINCTION

Faster flow. Less core energy.

FASTER FLOW. LESS ENERGY.

This is the sentence that makes the result possible. Speed describes motion at a point. Kinetic energy depends on speed and on how much fluid is moving. If the fast region shrinks quickly enough, its energy can fall even while its speed rises — so unbounded speed and bounded energy are not in contradiction.

U ≍ τ^−1/2−h E_core ≍ τ^1/2−3h τ = T − t

With h = 0.005 and unit constants, as an illustrative example. As τ → 0 the first exponent is negative (speed explodes) and the second is positive (core energy vanishes).

READ THE SYMBOLS

τ	Time remaining before the singular time T. The slider moves it from 1 down to 10⁻⁶.
U	Peak speed in the core, relative to its starting value. Shown on a log scale.
E_core	Kinetic energy inside the core only — part of the fluid's total, which stays bounded.
h	A small positive parameter in the scaling; it is what tips energy downward.

LIVE READOUTSIM/04 · SCALING LAWS

SPEED (LOG) vs CORE ENERGY (LINEAR)

CORE SPEED1×INITIAL SPEED = 1×

CORE ENERGY100.0%INITIAL ENERGY = 100%

TIME REMAINING τ1.0e+0

EARLIERCLOSER TO THE SINGULAR TIME

Illustrative ratios. Speed is on a log scale and energy on a linear one. The percentages measure progress through this illustration, not elapsed physical time.

THE REPORTED RESULT

A counterexample — not a formula for every fluid.

A COUNTEREXAMPLE, NOT EVERY FLUID.

As reported on 8 September 2026: a smooth, externally forced flow that starts at rest and develops unbounded speed in finite time while its total energy stays bounded. The construction is claimed for every positive viscosity, addressing the Clay problem's breakdown alternatives (C) and (D).

Three things it is not. It is a forced result — the flow is stirred — so it says nothing about blowup with no external force. It is one flow, not a statement about fluids in general. And the animation on this page skips the startup from rest entirely: the proof, not the picture, carries the claim.

PRIMARY SOURCES

▸ THE ANNOUNCEMENT ▸ LEAN FORMALIZATION ▸ THE CLAY PROBLEM STATEMENT

This page does not independently verify the proof.

LIVE SIMULATIONSIM/05 · PEAK SPEED TRACE

|u|_max AGAINST TIME · FORCED FROM REST

Two traces. With forcing off, a fluid at rest stays at rest and a stirred one decays — bounded forever. With forcing on, the trace follows the reported scaling and climbs off the chart as t → T. Illustrative curves, not the paper's data.

READ THE EQUATION

Changing motion, carried motion, pressure, friction, and an outside push — all in balance.

TERM EXPLORERSELECT A TERM TO READ IT IN WORDS

—

Select a term above.

And one more line: ∇ · u = 0 — incompressibility. A moving parcel of fluid keeps its volume; squeeze it one way and it must expand another. It is what turns pressure into a constraint rather than a free variable.

◈ BEHIND THE VISUALS

SIM/01 integrates tracers through the exact Taylor–Green field u = sin x cos y e^−2νt, v = −cos x sin y e^−2νt, with pressure p = ¼(cos 2x + cos 2y) e^−4νt, on a periodic domain, by fourth-order Runge–Kutta. This is a genuine solution of the equations and it never blows up.

SIM/02 is a two-variable toy — core width and core speed under a stretching term and a diffusion term — built to show a race, not to solve Navier–Stokes. SIM/03 is a drawn schematic: it exaggerates the change in proportions, uses independent bead timing, approaches the deadline logarithmically and stops at τ = 10⁻⁶. It shows central inflow and axial outflow and omits the full profile's asymmetry, the distant radial outflow and the correction terms.

SIM/04 and SIM/05 plot the paper's stated scaling bounds U ≍ τ^−1/2−h and E_core ≍ τ^1/2−3h with h = 0.005 and unit constants, as an illustration and not the exact solution. Strand colours are decorative and do not reproduce any measured data.

CAN A LITTLE
WHIRLPOOL BREAK THE MATH?

◈ WHAT WAS REPORTED — AND WHAT THIS PAGE DOES NOT DO