An engine's cooling system is usually described as protection: it stops the metal melting. That is true and it is not the whole story. Every joule the cooling system carries away was bought with fuel and will never become torque. The radiator is a running account of everything the engine failed to use.
Where the heat actually goes
Burn petrol in a cylinder and the energy released divides three ways. Some becomes work on the piston. Some leaves down the exhaust pipe as hot gas. And some conducts straight into the metal surrounding the flame — the piston crown, the cylinder walls, the head, the valves — from where the coolant collects it and the radiator throws it into the air.
That third path is the awkward one. Exhaust heat at least did some work on the way out, and a turbocharger can recover a further slice of it. Heat conducted into the cylinder wall did nothing at all. It left the gas before the gas had finished pushing, so the pressure during the power stroke is lower than it would otherwise have been, and the piston is pushed less hard for the rest of its travel.
This gives a designer an unusual objective. Not "remove heat efficiently" but "have less heat to remove", by keeping it in the working gas long enough for the gas to do its job. Every kelvin the combustion chamber surface does not steal is pressure still acting on the piston.
Coatings that are bad at conducting
The tool for this came from jet engines. A thermal barrier coating is a thin ceramic layer sprayed onto a metal component to slow heat flowing into it. The standard material is yttria-stabilised zirconia — zirconium dioxide with a few per cent of yttrium oxide added, which holds the crystal structure stable through the temperature swings that would otherwise crack it apart.
What makes zirconia useful is a conductivity roughly two orders of magnitude below the aluminium or steel underneath. A layer a fraction of a millimetre thick therefore supports a large temperature drop across itself: hot face near the gas temperature, cool face near the metal's. The metal lives at a temperature it can survive while the gas sees a surface much hotter than bare metal could ever be allowed to get.
In a gas turbine that buys turbine inlet temperature, and turbine inlet temperature is very nearly the whole game — the thermodynamic ceiling on a gas turbine's efficiency is set by how hot it dares run the gas entering the turbine. In a piston engine the payoff is different in character. Keeping heat in the gas raises cylinder pressure during expansion and cuts the load on the cooling system, and it keeps the exhaust hotter, which matters more than it sounds: catalytic converters do almost nothing until they are warm, so most of a modern car's emissions are produced in the first minute after a cold start.
Why steadier is better than cooler
The other half of the argument is about survival rather than efficiency, and it is about temperature swings rather than temperature. A component that heats and cools repeatedly expands and contracts, and if one part of it is hotter than another, the cool part restrains the hot part's expansion. That restraint is a stress, and it appears every cycle whether or not the engine is under load. Enough cycles and a crack starts, usually at whichever corner or hole concentrates the stress most. This is thermal fatigue, and it is a fatigue failure in the ordinary sense: driven by the number of cycles and the size of the swing, not by any single overload.
Which is why a coating that reduces the swing can extend life even where it does not much reduce the average temperature. It is also why the modern approach to cooling is to control temperature rather than minimise it — running different regions of the engine at deliberately different, steady temperatures, hotter where combustion needs it and cooler where the oil film has to survive, instead of flooding everything with coolant and accepting whatever gradients result.
Where the fuel's energy actually ends up is worth seeing as a budget rather than a list — the efficiency demo splits it into work, exhaust and coolant and lets you move the operating point.
Go deeper: why the coating has to be slightly wrong on purposefor engineers
Zirconia and the metal beneath it expand at different rates, so every heating cycle tries to shear the coating off its substrate. A perfectly dense, perfectly bonded ceramic layer would be the worst possible design: with nowhere for the strain to go, it would crack through and spall.
Real thermal barrier coatings are therefore deposited with deliberate porosity, and often with microcracks running through their thickness. The pores and cracks let the layer accommodate strain by opening and closing slightly, in the way a dry-stone wall accommodates ground movement that would crack a mortared one. The porosity lowers the conductivity further as a bonus, since gas-filled voids conduct badly.
There is usually an intermediate bond coat as well, a metallic alloy layer whose expansion sits between the ceramic's and the substrate's, and which oxidises in a controlled way to form the adhesion the ceramic relies on. When a thermal barrier coating fails, it is most often that oxide layer growing too thick and letting go — not the ceramic itself giving up.
Related: Why Engine Timing Is Measured in Degrees, Not Seconds