Bernoulli Levitation: Floating a Ball on a Jet of Air

The demonstration, and the two things it hides

Aim a hair dryer straight up, set it on cold and high, and drop a ping-pong ball into the stream. The ball climbs to a preferred height and stays there, wobbling slightly. Nudge it sideways and it snaps back into the column of air. Tip the dryer 20–30° from vertical and the ball follows, hanging out over the edge of the table as if on an invisible tether. Swap the dryer for a leaf blower and a beach ball, and the same thing happens at a larger scale.

Two separate questions are buried in that simple picture. First, what holds the ball’s weight up against gravity? Second, and far more interesting, what centers the ball in the jet and lets it resist a sideways pull — even supply a force that has a horizontal component when you tilt the jet? The popular answer, “Bernoulli,” is only half right for the first question and needs real care to be right about the second. Getting the mechanism correct means separating momentum from pressure, and knowing where Bernoulli’s theorem is allowed to speak.

Holding the ball up: momentum, drag, and a jet that slows with height

The vertical support is fundamentally about momentum, not Bernoulli. A jet of air moving upward at speed v carries a momentum flux ρAv² through any cross-section of area A. When the ball sits in the stream, it deflects and slows that air, and the reaction is an upward force. In drag language, the ball feels F_D = ½ ρ v² C_d A_ball, with the sphere’s drag coefficient C_d ≈ 0.4–0.5 in this regime.

The ball settles where that upward drag exactly equals its weight. For a regulation ping-pong ball (40 mm, 2.7 g, weight ≈ 0.027 N, frontal area ≈ 1.26×10⁻³ m²), setting F_D = mg gives a local jet speed of about 8–9 m/s at the ball — comfortably below a hair dryer’s ~10–20 m/s nozzle exit.

Crucially, that height is stable, and this is what makes the ball hover rather than being blown to the ceiling. A round free jet has a “potential core” that persists for roughly 5–6 nozzle diameters; beyond it the centerline velocity decays roughly as U(x) ≈ 6 U₀ D/x — inversely with height, as the jet spreads and entrains still air. So if the ball drifts up, it enters slower air, drag drops below its weight, and it falls back; if it drifts down, it meets faster air, drag exceeds its weight, and it rises. The visible bobbing is a lightly damped oscillation about this drag–gravity balance point.

The clever part: why the ball self-centers

Now the horizontal stability — the property that actually surprises people. Suppose the ball drifts to the right, toward the edge of the jet. Instead of the jet slipping past on the left and abandoning the ball, the fast stream stays attached and wraps around the ball’s surface (this is the Coandă effect, below). Two equivalent descriptions of the resulting force both point the same way:

• Newton / momentum deflection. The ball, sitting off to the right, deflects the wrapped jet outward to the right. Turning that stream’s momentum rightward requires a rightward force on the air — and by the third law the air pushes the ball leftward, back toward the jet axis.
• Bernoulli / pressure gradient. On the jet-facing (left) side of the ball the attached air is moving fast, so by p + ½ρv² = const the surface pressure there is low; the quieter air on the outer (right) side is slower and at higher pressure. The net pressure force pushes the ball toward the low-pressure, high-speed side — again back into the jet.

Both readings describe one restoring force directed toward the jet’s center. It behaves like a soft spring: displace the ball and it accelerates back, overshoots, and rings down. The most convincing proof that this lateral force is real and substantial is the tilt test: incline the jet by an angle θ, and the ball hangs off-axis in a new equilibrium where the jet’s restoring/support force balances not just mg but its component mg sin θ across the stream. Vigorous jets manage θ up to roughly 30–45° before the ball falls out.

Bernoulli, Newton, and honest bookkeeping

It is worth being blunt about a widespread textbook error. Saying “the fast jet is at low pressure, so Bernoulli sucks the ball up” misapplies the theorem. Bernoulli’s equation holds along a single streamline in steady, inviscid, incompressible, essentially irrotational flow. You cannot legitimately compare the fast core of the jet with the still air of the room and conclude anything: those are different streamlines, and the boundary between jet and room is a turbulent shear layer where viscosity dissipates energy — exactly where Bernoulli’s energy conservation fails. The reliable ledger for the vertical support is momentum.

Where Bernoulli is valid is in the smoothly accelerating, attached flow hugging the ball’s surface on the upstream and jet-facing sides. There it correctly tracks the pressure drop that contributes to the self-centering force. This is the very same nuance that fuels the perennial “Bernoulli versus Newton” debate about how wings generate lift: both frameworks are correct when applied properly, because a pressure field and a momentum change are two accountings of one physical interaction. For the floating ball, the clean division is: momentum flux for lift against gravity, Coandă-mediated pressure asymmetry for lateral stability.

The Coandă effect: why the jet hugs the ball

The reason the stream refuses to peel away is the Coandă effect — the tendency of a fluid jet to follow a nearby convex surface. A free jet is turbulent and entrains surrounding air: its ragged edges mix with and drag along ambient fluid, so the region just outside the jet runs at slightly reduced pressure. Bring a curved surface up to one side of the jet and it cannot resupply air to the gap between jet and surface as fast as the jet sweeps it away; the pressure there drops, and the jet bends to lie against the surface. Wrapped around the ball, the jet stays attached over much of the sphere and re-attaches when the ball drifts, which is precisely what supplies the restoring geometry.

The effect is named for the Romanian aerodynamicist Henri Coandă, who noticed exhaust gases clinging to his aircraft’s fuselage around 1910 and later patented devices exploiting it in the 1930s; the term was coined by Theodore von Kármán. The same attachment is what lets a spoon back deflect a stream of tap water, what steers the airflow in a bladeless fan, and what “blown flap” aircraft use to keep flow attached at high wing angles.

Numbers, regimes, and how it breaks

With a hair dryer and a 40 mm ball, the Reynolds number is Re = vD/ν ≈ (10–20 m/s)(0.04 m)/(1.5×10⁻⁵ m²/s) ≈ 3–5×10⁴ — fully turbulent, which is why entrainment and mixing matter so much. That value sits comfortably below the sphere’s “drag crisis” near Re ≈ 3×10⁵, so the drag coefficient stays around 0.4–0.5 rather than collapsing. The ball also typically spins as it floats, because the shear it sits in is never perfectly symmetric; the rotation adds a modest Magnus force but is not the main stabilizer.

The trick has clear failure modes:

• Too heavy. If mg exceeds the maximum available drag even at the nozzle, no equilibrium height exists and the ball simply drops.
• Ball too big for the jet. If the sphere is much wider than the jet, the stream cannot wrap it; the Coandă attachment and the restoring force disappear and the ball is merely batted around.
• Jet too weak, narrow, or gusty. A puny or fluctuating jet cannot maintain steady attachment; the ball skips out sideways.
• Over-tilting. Past ~30–45° the required cross-stream force exceeds what the pressure asymmetry can supply and the ball falls out of the flow.

History, cousins, and real applications

The pressure–velocity relation at the heart of the demo comes from Daniel Bernoulli’s Hydrodynamica (1738), later written in its familiar differential form by Euler. The floating-ball demonstration itself is a physics-classroom and science-museum staple precisely because it packages momentum, Bernoulli, and Coandă entrainment into one glance.

The physics is not only a toy. Bernoulli grippers (also called Bernoulli or Coandă wafer chucks) lift delicate silicon wafers and solar cells without touching their faces: a radial jet of air spreads through the thin gap between the gripper face and the top of the plate, the fast flow drops the local pressure there, and atmospheric pressure below holds the plate up against a thin cushion of air, non-contact. Fluidic logic devices once used Coandă switching between attached-jet states to compute with no moving parts, and the same entrainment principle underlies air amplifiers, air-knife dryers, and bladeless fans. Closely related demonstrations include the ball held up inside an inverted funnel or a converging tube, where blowing harder pins the ball more tightly rather than blowing it out — the same low-pressure-follows-fast-flow logic. And it sits in a family of contact-free levitation methods — acoustic, diamagnetic, optical — each of which trades a different force and a different stabilizing trick for the same magic of an object hanging in mid-air.

How air-jet levitation compares with other ways to float an object without touching it

Method	Support force	What makes it stable	Typical object
Air jet (Bernoulli / Coandă)	Jet momentum flux and drag	Coandă wrap + surface pressure gradient	Ping-pong ball, beach ball
Acoustic levitation	Acoustic radiation pressure	Trapping at a standing-wave pressure node	mm droplets, small solids
Diamagnetic levitation	Induced diamagnetic repulsion	Earnshaw-stable for diamagnets	Pyrolytic graphite, water, a frog
Optical tweezers	Radiation-pressure gradient force	Intensity gradient of a focused laser	Microspheres, single cells
Aerodynamic (wind tunnel)	Aerodynamic drag / lift	Active feedback or a shaped flow field	Scale models, spheres