Why does a ring of steel balls make wheels spin so freely?

Let a shaft turn directly in a hole and the two surfaces slide across each other: their textures scrape past one another, friction drags at the rotation, and energy leaves as heat and wear. Lubricated steel on steel slides with a friction coefficient near 0.1; dry, it can reach 0.45. The faster and harder the contact, the more energy is scraped away — the shaft heats, stiffens, and the hole wears loose.

A bearing’s whole idea fits in one sentence: replace sliding with rolling. Set a ring of balls between the shaft and its support, and the balls roll along two raceways. In ideal pure rolling, the point where a ball touches a raceway has zero relative speed — the textures no longer cross — and the friction coefficient falls to 0.001–0.002 (the order quoted in the SKF and NSK rolling-bearing catalogues), roughly one seventieth of the sliding value with this demo’s settings. A real rolling bearing still loses energy to rolling, sliding, lubrication and seal drag together (SKF’s friction model counts them separately): a bearing reduces friction, it does not abolish it.

Take apart the most ordinary deep-groove ball bearing and four parts appear: the outer ring, pressed into the support and stationary; the inner ring, fitted to the shaft and turning with it; a circle of rolling elements (balls) carrying load between the two raceways; and a cage keeping the balls apart. Load travels exactly along that chain: the shaft presses the inner ring, the ring presses a ball at their contact, the ball presses the outer ring, and the outer ring hands the force to the support.

Under a radial load the balls do not share evenly. Only those within ±90° of the load direction lie in the loaded zone, and the ball straight under the load works hardest: Stribeck’s classic result (around 1900) gives, for zero internal clearance, a busiest-ball load of about 4.37 times the reference Fr/Z (applied radial load divided by ball count) (this demonstration discretises nine balls and gets 4.39). As the inner ring turns, the loaded zone does not travel with it — the bright region stays aimed at the load while balls pass through it.

The cage is not the main path carrying shaft load to the support. At their mutual contact, neighbouring balls move their surfaces in opposite directions; without a cage they would press together and slide against each other, inviting sliding friction right back in. The cage’s pockets space and guide the balls, riding along with their orbit. To follow the kinematics of a single ball balancing both contacts, see the related ball bearing exploration.

The load’s direction decides which family to choose: a deep-groove ball bearing favours radial load and tolerates a little axial; a thrust bearing takes axial load but is crushed by radial; a tapered roller bearing carries both. Fit the wrong type and the bearing heats, binds, and can come apart.

Detailed controls

1. Slide vs roll. In the exploration, switch between “shaft slides in the hole” and “balls roll”, and watch the friction gauge: μ falls from 0.1 to 0.0015 while the dissipated watts drop with it. In the inset, the sliding surfaces’ textures cross each other; the rolling ball’s contact point holds still (ideal pure rolling, shown as a sketch).
2. How heavy a load? Drag the load slider. The contact rings brighten with each ball’s share, but the bright region stays around the load direction, and the busiest ball always faces the load line.
3. Remove the cage. Press “remove the cage”: the balls crowd together and red counter-rubbing marks appear between their contacts; restore it and they spread evenly again. The cage separates and guides the balls; the main load still passes through the balls and raceways.
4. Push from another direction. Turn the load direction from 0° to 90° and swap bearing types: a deep-groove ball bearing can carry axial load too, but this comparison assumes a lower axial limit, so its identical demonstration load turns the indicator red. That does not mean the family cannot carry axial load. Thrust bearings serve axial loads; tapered rollers can serve combined loads. These limits are teaching assumptions, not product ratings.

Guess first: which ball carries the most? Choose, then check — in this zero-clearance teaching model, divide the applied radial load into nine equal reference shares: the ball facing the load carries about 4.4 shares, while the topmost ball carries almost nothing.

Let ψ be a ball’s angle away from the load direction. With zero internal clearance, the Hertz point-contact relation gives Q(ψ) = Q_max·cos^3/2(ψ), with only |ψ| ≤ 90° loaded. Projecting every ball’s load along the load line and summing must equal the applied radial load F_r. Integrating the continuous distribution yields Q_max ≈ 4.37·F_r/Z (Z = number of balls); with Z = 9 the discrete sum here gives 4.39. This is the starting point of static load rating checks (ISO 76 defines ratings via the contact stress at the most heavily loaded element).

A rolling bearing still loses a little: material in the contact compresses and relaxes with micro-slip and hysteresis; the cage, ball-to-cage touch and the grease itself all resist. Hence μ ≈ 0.001 rather than 0. What rolling saves is not all friction — it trades whole surfaces sliding across each other for tiny local deformations, two orders of magnitude better. And a plain bearing with a full hydrodynamic oil film can also run at very low friction: bearing selection follows speed, load and duty; rolling does not always win.

A single skateboard wheel holds two small bearings side by side; desk fans, vacuum-cleaner motors, washing-machine drums, bicycle hubs and hard-drive spindles (many fluid-film) all spin quietly on them. Count the motors at home, then consider: many rotating parts use rolling bearings, while others use an oil film or another form of support to reduce friction.