Steel wheel on steel rail is already one of the lowest-friction contacts in everyday engineering — a rolling resistance coefficient around 0.001 to 0.002. That's roughly 100x less than rubber on asphalt. So what happens if we take that last sliver of friction away entirely? The answer isn't just "trains go faster." It's that trains become fundamentally uncontrollable in ways that are surprisingly instructive.

The curve problem comes first. Trains don't steer. They rely on conical wheel profiles and the flange-rail interface to navigate curves. The wheel tread is slightly tapered — about 1:20 on standard rail — so when a wheelset drifts laterally, the effective rolling radius changes, creating a self-centering restoring force. But this restoring force is entirely friction-dependent. The differential in rolling radius only works because the contact patch can transmit a lateral creep force. Remove friction, and the wheelset has zero lateral guidance from the tread geometry. The flange becomes your only constraint — and flanges are a last resort, not a steering mechanism.

Let's calculate when the flange fails. On a standard curve (say, radius R = 500 m, common for mainline rail), the centripetal acceleration needed is v²/R. The flange contact can transmit a lateral force, but without friction, the only thing keeping the wheel on the rail is the mechanical interference of flange against rail head. Standard rail head width is about 73 mm, and the flange extends roughly 28 mm below the tread surface with a thickness of about 25 mm.

A loaded freight car has a mass of about 130,000 kg on 4 axles, so roughly 32,500 kg per wheelset. The centripetal force needed at velocity v is:

F = mv²/R = 32,500 × v² / 500 = 65 × v² (Newtons)

The flange-to-rail contact is essentially a steel-on-steel bearing surface, but without friction it's just a normal force. The rail can withstand enormous compressive loads — rail steel has a yield strength around 900 MPa. The flange contact area is roughly 15 mm × 10 mm = 150 mm² = 1.5 × 10⁻⁴ m². That gives a maximum lateral force before plastic deformation of about:

F_max = 900 × 10⁶ × 1.5 × 10⁻⁴ = 135,000 N

Setting 65v² = 135,000, we get v = √2077 ≈ 46 m/s, or about 165 km/h. Above that speed on a 500 m curve with no friction, the flange begins yielding. It doesn't shatter instantly — it plastically deforms, climbing the rail, and within seconds the wheel mounts over the rail head. Derailment.

But the real disaster is stopping. Without friction, you have zero tractive braking force. Disc brakes, tread brakes — all useless. Regenerative braking still works (the motor is magnetically coupled to the axle, not the rail), but even modern high-speed trains get only about 50-60% of their braking from regenerative systems. The rest is friction brakes. An emergency stop from 300 km/h typically needs about 3 km. With only regenerative braking, you're looking at 5-6 km, and at low speeds where regenerative braking is weakest, you'd be coasting almost indefinitely.

The deeper lesson: Superconducting maglev trains (like the Chuo Shinkansen) actually did eliminate rail friction — and they solved the curve problem with U-shaped guideways that physically contain the vehicle, and braking with electromagnetic track brakes. They essentially reinvented every system that conventional rail takes for granted from friction. The guideway walls act as the "flange" with enormous contact area, and eddy-current brakes work without contact at all.

Friction isn't a tax on rail efficiency. It's the entire operating principle disguised as a nuisance.