Air that touches a wall sticks to it, so a thin layer next to the wall is slowed down. Its thickness grows with distance x, and with the Reynolds number Re_x = U x / nu (nu = 1.5e-5 m2/s for air). This page marches the real momentum-integral equation along the wall for any outer velocity Ue(x):
von Karman: d(theta)/dx = Cf/2 - (H + 2) (theta/Ue) dUe/dx
Blasius (flat, laminar): delta99 = 4.91 x / sqrt(Re_x), theta = 0.664 x / sqrt(Re_x), Cf = 0.664 / sqrt(Re_x)
Thwaites (laminar): theta^2 Ue^6 = 0.45 nu * integral(Ue^5 dx), lambda = (theta^2/nu) dUe/dx
separation when lambda = -0.09 (H = 3.55)
Transition: Re_theta >= 163 + exp(6.91 - Tu[%]) (Abu-Ghannam & Shaw, zero gradient)
Turbulent: Head entrainment + Ludwieg-Tillmann Cf = 0.246 * 10^(-0.678 H) * Re_theta^-0.268
separation when Cf -> 0
- Why it separates - in a slowing flow (adverse pressure gradient, dUe/dx < 0) the slow fluid near the wall cannot climb the pressure hill: the wall shear falls to zero and the layer lifts off. A turbulent layer mixes in fast fluid from outside, so it survives a much stronger gradient - that is why golf balls are dimpled and aircraft wings carry vortex generators.
- Trip wire - forces transition at x = 0.06 m. Compare "Diffuser, laminar stalls" with "Diffuser, tripped holds".
- Checks - the Blasius constants come from a separate RK4 shooting solve of f''' + f f''/2 = 0; the Howarth case (Ue = U (1 - x/L)) has an exact separation point at x/L = 0.1199 from the literature.
- Honest simplifications - 2D steady mean flow, edge velocity prescribed (no feedback of the thick layer on the pressure), no separation bubble or reattachment: once the layer separates the march stops. The displayed vertical scale is greatly exaggerated.