The Coanda Effect: Why Flow Sticks to a Curved Surface

The mechanism: entrainment sets up a low-pressure pocket

A turbulent free jet does not travel as a clean column. Its shear layer drags surrounding fluid along with it — entrainment — pulling ambient fluid into the jet and spreading it. A round jet entrains mass at roughly ṁ(x)/ṁ₀ ≈ 0.32·x/d (x = distance downstream, d = nozzle diameter), so it can double or triple its mass flow within a few diameters.

Now place a curved wall just to one side of the nozzle. On the wall side, the jet still tries to entrain fluid, but the surface blocks fresh fluid from feeding the gap. The trapped region can only be replenished slowly, so its pressure drops below ambient. That transverse pressure difference — higher on the open side, lower against the wall — pushes the jet sideways until it touches the surface. Once attached, a stable low-pressure zone hugs the wall and the jet rides the contour. The classic contrast: remove the wall and the jet flies straight; add the convex wall and it wraps around, deflecting bodily by 30°, 90°, even ~180°.

Why the jet turns: the balance that keeps it on the curve

Following a curved wall means the jet is constantly accelerating centripetally, so something must supply the inward force. That force is the radial pressure gradient. For a thin jet of density ρ and speed u following a wall of local radius R, force balance across the jet requires:

∂p/∂r = ρ·u²/R

Pressure must rise moving outward from the wall (across the jet, radius increasing) — equivalently, pressure at the wall is lower than ambient by roughly Δp ≈ ρ·u²·t/R for jet thickness t. That sub-ambient wall pressure is exactly what glues the jet down and simultaneously provides the centripetal force to bend it.

Numbers make it concrete: air (ρ ≈ 1.2 kg/m³) at u = 100 m/s over R = 0.05 m gives ρu²/R ≈ 240,000 Pa/m. Across a t = 3 mm jet that is a wall suction of order Δp ≈ 720 Pa — small versus atmospheric (101 kPa) but more than enough to hold a thin sheet against the surface. Tighten R and the required suction rises; make R too small and the jet cannot supply enough turning force, so it separates.

The failure mode: separation when curvature is too aggressive

The effect is not unconditional — the pitfall is assuming a jet will follow any curve. Attachment persists only while the boundary layer has enough momentum to survive the adverse pressure gradient around the far side of the curve, where pressure recovers back toward ambient. When curvature is too sharp (R too small relative to jet thickness), or the jet is too slow, or the surface too rough, the flow separates, snaps off the wall, and the deflection collapses — often abruptly and with hysteresis (it re-attaches at a different condition than it detaches).

• Thickness ratio t/R: a practical rule of thumb is that reliable attachment needs t/R below roughly 0.5; thin jets on gentle curves turn best.
• Reynolds number: turbulent jets (Re > ~10⁴) entrain more vigorously and attach more strongly than laminar jets.
• Compressibility: at high subsonic/supersonic speeds, shock cells and expansion fans over the curve disrupt the neat pressure balance and can force early separation.

This bistable attach/detach behavior is a feature in fluidics: a small control jet can flip the main jet from one wall to another, giving a no-moving-parts switch or amplifier.

Where it is used: real machines and dates

Aviation. Henri Coandă described jet attachment after a 1910 experiment and patented flow-deflection devices; his US patent 2,052,869 issued 1936. The physics underpins blown flaps and circulation control: engine bleed air is blown tangentially over a rounded trailing edge so the flow stays attached far past a normal flap angle, moving the rear stagnation point and boosting circulation. The Antonov An-72/An-74 deliberately positions its engines to blow exhaust over the wing upper surface (upper-surface blowing) for short-takeoff performance; the Boeing YC-14 (1976) used the same trick. NOTAR helicopters (MD 500/600 series) replace the tail rotor with a Coandă slot along the tail boom that deflects downwash to produce anti-torque thrust.

Consumer and industrial. Dyson's Air Multiplier bladeless fan blows a thin jet over a curved airfoil-section ring; the attached jet entrains room air and multiplies total flow by roughly 15× over the ~1 L/s of primary jet. Coandă nozzles are used for drying, air knives, air-conditioning air spread, and thrust-vectoring exhausts (e.g., fluidic control tested on the BAE MAGMA UAV, 2019). In process plants, wall-attachment fluidic elements act as logic gates and flow sensors with no moving parts.

A worked example: sizing a Coandă thrust-deflection slot

Suppose you want a wall jet to turn 60° around a cylindrical trailing edge of radius R = 25 mm using an air slot of height t = 2 mm, supply pressure giving jet speed u = 120 m/s.

• Centripetal demand: ρu²/R = 1.2 × 120² ÷ 0.025 ≈ 691,000 Pa/m.
• Wall suction across the jet: Δp ≈ (ρu²/R)·t ≈ 691,000 × 0.002 ≈ 1,380 Pa below ambient — comfortably achievable, so attachment is plausible.
• Thickness ratio: t/R = 2 ÷ 25 = 0.08, well under the ~0.5 guideline → good margin against separation.
• Momentum coefficient (the blowing metric used in circulation control): C_µ = ṁ·u / (q·S). For meaningful lift or deflection you typically need C_µ in the 0.01–0.1 range; below ~0.005 the jet barely alters the base flow, above ~0.3 you get diminishing returns and heavy bleed penalties.

Trade-off: raising u increases turning authority but scales pumping power with u³, and a tighter R turns more per unit length but pushes you toward the separation cliff. Good designs live in the sweet spot of moderate t/R and C_µ.

Coandă vs. a plain deflector, and a common misconception

Compared with a solid vane or splitter that physically redirects flow, a Coandă surface deflects the jet with pressure alone and can turn it much further around a smooth surface without the flow slamming into a plate and losing energy to a stagnation region. That is why Coandă deflection can be remarkably efficient — and why a tiny control jet can steer a large main jet (fluidic gain).

The misconception: people often say lift 'comes from the Coandă effect' or that the effect is simply Bernoulli — 'fast flow over a curve makes low pressure.' Both are sloppy. The Coandă effect is specifically about attachment (entrainment starving a gap of fluid); Bernoulli then describes pressure once you already know the streamline pattern. Attachment enables high circulation on a blown wing, but the lift itself is the reaction to net downward momentum imparted to the air — Newton and the circulation, not a magic 'sticking' force. Conflating the two leads engineers to over-credit surface curvature and forget that momentum and entrainment, not the wall shape alone, do the work.

Coanda / circulation control vs. conventional mechanical flap for wing lift augmentation

Attribute	Coanda blown surface (circulation control)	Conventional hinged flap
Lift mechanism	Tangential jet delays separation & moves stagnation point → high circulation	Increases camber/area, geometric deflection of flow
Max lift coeff. CL	Up to 5–9 with strong blowing	Typically 2.5–3.5 (slotted/Fowler)
Moving parts	Few/none (fixed rounded trailing edge + valves)	Tracks, hinges, actuators, linkages
Energy cost	Needs bleed air/compressor; blowing power penalty	Actuation only; parasitic drag when deployed
Failure mode	Jet separation if blowing lost → abrupt lift loss	Jam or asymmetric deploy; graceful on partial fail