FluidLab

Airfoil simulation, tilt it and watch

A NACA 0012 section in a lattice-Boltzmann stream. Raise the angle of attack and the suction on the upper surface grows, the wake bends, and eventually the flow lets go of the surface.

Drag Cd
—
Lift Cl
—
L / D
—
Cd (momentum check)
—
Stability
—

Where lift comes from

A symmetric section at zero angle produces no lift: the flow is the same above and below, and any circulation it might carry is cancelled. Tilt it, and the stagnation point moves down the leading edge. Air has to travel around the sharp trailing edge, which is impossible in an inviscid fluid — the Kutta condition forces the circulation that makes the flow leave the trailing edge smoothly.

That circulation shows up as a speed difference: faster over the upper surface, slower underneath. By Bernoulli, faster means lower pressure, so the section is pushed up. In the vorticity view you can see the same thing as a layer of opposite-signed vorticity clinging to each surface.

What to watch as you raise the angle

small α

The wake is thin

Flow follows both surfaces, the stagnation streamline sits near the nose, and lift climbs roughly linearly with angle — the textbook regime.

moderate α

The suction peak sharpens

Vorticity concentrates near the leading edge on the upper surface. Lift keeps rising but the favourable-pressure region gets shorter and the wake starts to widen.

large α

Separation

The upper-surface layer detaches. A broad, low-pressure wake replaces the thin sheet, lift stops growing and drag jumps — stall, in slow motion.

Why the numbers here are trends, not answers

The default Reynolds number is a few hundred (based on chord). At that scale the boundary layer islaminar and separates at the slightest adverse pressure gradient, often forming a separation bubble that reattaches further aft. Real wings at flight Reynolds numbers are turbulent; this solver has no turbulence model to put that back.

Add a two-dimensional domain, a coarse grid and a section thickness of 12%, and the honest claim is this: the shapes and trends are informative, the absolute coefficients are not. The page prints them because watching them move as you drag the slider is the point.

If you want numbers you can defend, the same solver will get you much closer on thecylinder page, where the geometry is a circle and the reference values are unambiguous.

FAQ

Are the lift numbers accurate enough to design with?

No. At this Reynolds number (a few hundred) the boundary layer is laminar and separates easily, there is no turbulence model, and the domain is two-dimensional. Use the page to see how lift rises, saturates and collapses with angle — not to quote a Cl value.

Why does lift drop instead of rising forever?

Past a critical angle the boundary layer can no longer follow the upper surface. It separates, the suction peak collapses and the wake thickens — stall. In this low-Reynolds regime it often appears as a laminar separation bubble rather than a clean break.

What does the momentum-balance number mean here?

It is a second estimate of drag computed only from the flow field across two cross-sections, with no surface integration. When the two drag values disagree a lot, the wake has not settled or the airfoil is too coarse for the grid.

Why is the angle limited to −5°…20°?

Outside that range the section is fully stalled or the leading edge is so badly resolved that the numbers stop meaning anything. The slider is deliberately narrower than the solver can technically run.