Engine run-in

Why does a new engine run more smoothly after its first thousand miles than it did leaving the factory?

A new engine is the most precisely made object most people own — every bore measured, every clearance held to microns. And yet it is at its roughest the day it leaves the factory: more fuel used, more power lost to friction, more metal in its first oil change than in the next ten together. A thousand miles later it is measurably better.

Wear is supposed to be damage. Here a controlled dose looks like an improvement. What got better, and why could the factory not simply make it that way?

Start with what "flat" means on a machined surface. A honed cylinder bore looks like a mirror, but under a profilometer it is a landscape of ridges and valleys with peaks of order a micrometre high. Two such surfaces pressed together do not meet across their apparent area; they meet where peak touches peak.

How little is that? A classic result is worth deriving. A contacting peak carries load by yielding plastically until it has spread over enough area to support it, so real contact area is roughly load divided by the material's hardness. Steel's hardness is of the order of a few thousand megapascals (recalled, order-of-magnitude). If nominal contact pressure between a piston ring and a bore is of order one megapascal, the ratio of real to apparent contact is about one in a few thousand. The whole load rides on a thousandth of the surface you can see — so while the average pressure is mild, the local pressure at each peak sits at the yield limit by construction.

Now bring in the oil. A ring separates from its counterface by a hydrodynamic film, and whether it separates fully depends on how film thickness compares with the roughness of both surfaces. Tribologists use that ratio — lambda — as a regime marker: below about one, peaks still touch through the oil; above about three, the surfaces are separated and wear essentially stops (recalled convention). A new surface has tall peaks, so lambda is low even when the oil is doing all it can.

Here is the step the whole thing turns on. Wear removes material preferentially where contact pressure is highest, which is the tops of the peaks. It does not touch the valleys. So the surface does not simply get thinner, it gets flatter: the tall asperities are truncated while the honing valleys survive as oil reservoirs. Combined roughness drops, lambda rises, and pressure at the remaining peaks falls because the load is shared over many more of them — so the wear rate falls too. This is self-limiting: the process removes the very thing driving it. And it is specific — the ring is not becoming a better ring in the abstract, it is becoming the shape of this bore, and the bore the shape of this ring.

There is a failure mode on the other side of the same mechanism. Push too hard too early — high load, high speed, cold oil — and the peaks do not merely deform, they weld to their opposite numbers and tear away. That is scuffing, and it makes the surface worse. The instruction to load a new engine moderately but not gently is that balance: enough pressure to truncate the peaks, not enough to weld them.

Three readings deserve ruling out. The first is that the metal "hardens" with use. There is real work hardening in the topmost layer, but it cannot be the explanation, because it claims something about bulk material while the improvement is a change in contact geometry — and, decisively, it is undone the moment you change which surfaces are mating. The second is that the rings wear the bore into shape; mostly the traffic runs the other way, the softer ring conforming to the harder bore, and which part wore is beside the point. What matters is that the pair became mutually conformal. The third is that this is just infant mortality in a machine population — it is not, as the next-door comparison below shows.

The falsification test the account passes is the swap. If the improvement is the geometry of one specific mating pair, a fully run-in ring placed in a different, equally run-in bore should lose it — and it does: friction and blow-by rise again until the new pair has conformed. Had the metal improved, the benefit would have travelled with the part. That is also why a rebuild fitting new rings requires the bore to be re-honed. A second test runs the other way: plateau honing, a finish that truncates the peaks at the factory and leaves the valleys, shortens the run-in period — what "the surfaces only needed their peaks taken off" predicts, and what "the metal needs to toughen up" does not.

Some of this is contested. Manufacturers increasingly say no special procedure is needed, and they are not merely being convenient: better machining and plateau honing really have moved much of the conformity into the factory. Rebuilders and racing engine builders insist a load schedule in the first hours still measurably changes the ring-bore seal and eventual oil consumption. Both have an interest — the manufacturer in a customer who drives away without a warranty argument, the specialist builder in a service worth paying for — and honestly, what is disputed is the size of the remaining effect, not its direction.

Where this sits next to its neighbours: bathtub-failure-curve.md covers the same early period and reaches an opposite conclusion, and the two part at whether anything about an individual improves. That piece shows infant mortality is a selection effect: units born broken are removed, and no single machine got better. Run-in is the genuine exception — a real maturation in one machine, because a surface changed shape. Both are true at once in a new engine, some being culled while every one conforms.

The load-bearing claim is that the improvement is a change in surface geometry of a specific pair, not in bulk material properties. If a run-in ring kept its benefit in a strange bore, the account would have to be rebuilt around the material.