Why an engine overheats, and how to find out which part let it

An overheating engine is not a fault. It is the consequence of one, and the question worth asking is never why is it hot — it is which of the four things the cooling system does has stopped happening.

The system has four separate jobs: move the coolant, reject its heat into the air, get air through the radiator, and hold the whole circuit under pressure. Any one of them can fail on its own, every one of them produces overheating, and each produces it with a different signature. That is what turns a list of causes into a diagnosis.

The article this is translated from listed twenty-five causes of overheating in no particular order, including a headwind and under-inflated tyres. Sorted into those four groups the same material becomes something you can work through in an afternoon — and three of the most common causes turn out to be missing from it altogether.

First: is it actually overheating?

The temperature gauge is not an instrument. Most modern clusters do not display the coolant temperature they are given. They display a damped, stepped approximation that sits at the centre mark across a band roughly twenty degrees wide, deliberately, so that ordinary variation does not alarm drivers. P2181 covers what that costs in the other direction — an engine running cold that nobody notices for years.

What it means here is worse: by the time the needle moves, the real temperature has been out of range for a while. The gauge is a late warning, not an early one.

So the evidence worth having is elsewhere.

• Read the actual number on a scan tool. 90–105 °C is normal on a modern engine, and more of them run at the top of that than the bottom. See reading live data for what else to watch alongside it.
• The heater going cold while the engine gets hot. This is one of the most reliable early signs there is, and it is nearly diagnostic on its own: the heater matrix usually sits high in the circuit, so it is the first thing to be starved when coolant is low or air is trapped. A heater that blows cold at idle and warm at speed says the same thing more quietly.
• Coolant disappearing with nothing on the driveway. An external leak makes a puddle. Coolant lost past a head gasket is burnt and leaves as steam, so no puddle is the more worrying finding rather than the reassuring one.
• Boiling or gurgling after you switch off. Once the pump stops, heat soaks out of the metal with nothing moving to carry it away. A system that boils on soak is already running much closer to its limit than the gauge admitted.

If it is overheating right now

• Heater to full hot, blower to maximum, air conditioning off. The heater matrix is a second radiator and running it flat out genuinely buys you a few degrees. It is unpleasant and it works.
• Stop, and switch the engine off. Every minute of running at temperature is doing damage that is measured in cylinder heads. P0217 is the code for this condition and the corpus is blunt about it, because warped heads and cracked blocks are what the other side of the decision looks like.
• Do not open the coolant cap. The circuit is held at over a bar and the coolant in it is well above 100 °C — it is only liquid because of the pressure. Releasing that cap flashes the contents to steam through the opening you are leaning over. Give it thirty minutes at least, and open it slowly to the first stop even then.
• Do not pour cold water into a hot engine. An aluminium head that is at 120 °C and suddenly meets cold water can crack, which converts a cheap fault into an engine.

The original does say — correctly — never to remove the pressure cap on a hot engine. It says it about twenty thousand characters in, after the antifreeze specification tables. Several thousand characters earlier, in the section on the radiator, it instructs the reader to keep the level within two centimetres of the filler neck, with no warning attached. The content was there; the ordering made it useless. That is worth naming, because it is the commonest way a technically correct article gets somebody hurt.

Reading the pattern

Before touching anything, note when it overheats. This one observation eliminates most of the list.

1. Circulation

The thermostat stuck closed is the classic, and it overheats fast — within a few minutes of a cold start, because the radiator is simply not in the circuit. The engine thermostat article covers it and the far commoner opposite failure, and explains why fitting a cooler one to "help" hides an overheating fault instead of fixing it.

The water pump has a failure worth knowing about because it is silent. Impellers — plastic on a great many engines, and steel ones that have corroded — can erode away or spin loose on the shaft. The pump does not leak, does not rumble, and does not stop turning. It simply stops pumping. Suspect it when everything else checks out and the top hose is hot while the bottom one is barely warm.

What drives the pump decides how much a leaking one costs. On a belt-driven pump it is a morning's work. On the many engines where the timing belt drives it, a seized pump strips the belt and the valves meet the pistons — which is why a weeping pump on those engines is not something to watch, and why the pump is replaced with the belt whether or not it has failed.

Belt tension matters only on the older arrangement. The original gives the rule of thumb — press with about a kilogram of force, expect 8–10 mm of deflection — and that is fine for a manually tensioned V-belt. On the automatic tensioner fitted to essentially every serpentine drive since the 1990s there is nothing to adjust, and deflection tells you about the tensioner rather than the belt.

An air lock is the cause the original's twenty-five-item list does not contain, and on a car that has recently had work done it is the first thing to suspect. Air collects at the high point, the pump cavitates, and flow drops to nothing while the gauge sits normal for a while because the sensor may be sitting in a pocket of air rather than coolant.

A collapsing bottom hose is the last one. The bottom hose is on the suction side of the pump, so a hose whose reinforcing spiral has failed pulls flat at high rpm and springs open again at idle — an engine that overheats on the motorway and is perfectly normal at the services.

2. Heat rejection

Blocked outside is the one everybody checks: flies, leaves and road film packed into the fin pack, or fins bent flat. Look through the radiator towards a light rather than at it.

Blocked inside is the one nobody checks. Years of plain water, or of coolant left long past its life, silt the narrow tubes with corrosion products. The signature is a radiator with cold patches across it once the engine is warm — an infrared thermometer across the face finds this in a minute and nothing else finds it at all.

Coolant that is not coolant belongs here too, and it is common enough to be worth its own section below.

Combustion gas in the coolant is the third of the causes the original omits. A failing head gasket pushes exhaust gas into the water jacket under enormous pressure; gas does not carry heat and it does not pump, so the engine overheats while the system pressurises, forces coolant out of the expansion tank, and the top hose goes rock hard almost immediately after a cold start. The test for this is cheap and definitive — a combustion leak tester holds a reagent above the coolant, draws the air off the top of the system through it, and the fluid changes colour if there is CO₂ in there. Do this test before agreeing to any cooling-system work on a car with unexplained coolant loss.

3. Airflow

This is the whole category the original leaves out. Its list mentions broken fan blades and a slipping belt, which describes a mechanical fan bolted to the water pump. On nearly every transverse-engined car built since the late 1980s the fan is electric, it is commanded by the engine module or a temperature switch, and its circuit — not its blades — is one of the commonest reasons a car overheats in traffic and behaves perfectly on the open road.

The corpus carries this ground in detail: P0480, P0481 and P0482 for the control circuits, P0483 for a fan the module has decided is not achieving what it was told to, P0484 for over-current and P0485 for the power and ground side. Check the relay before the fan motor — it is cheaper, it fails more often, and a fan that will not run tells you nothing about which of the two is at fault.

Two more that get missed. On a belt-driven fan with a viscous coupling, the coupling wears out and the fan freewheels — spin it by hand on a cold engine and it should have noticeable drag. And a missing or broken fan shroud costs more airflow than it looks like it should, because without it the fan pulls air round the radiator instead of through it.

4. Pressure

The cap is the cheapest part in the system and the least understood. A cooling system is pressurised for one reason: to stop the coolant boiling.

• A 50/50 glycol mix boils at about 108 °C at atmospheric pressure.
• A cap holding around 1.1 bar adds roughly 25 °C to that, so the same coolant stays liquid to about 130 °C.

That margin is the entire difference between an engine that is running hot and an engine that is boiling. A cap whose seal has hardened, or whose spring has tired, gives it all back — and because boiling produces vapour, and vapour transfers no heat and does not pump, a failed cap causes overheating that looks exactly like every other cause on this page. It is a five-minute test with a pressure tester and it should be done first, not last.

The same logic applies to any leak, however small. A system that cannot hold pressure has lost 25 °C of headroom before the level has dropped enough for anyone to notice.

Coolant: three things that are the opposite of what people assume

Pure antifreeze is worse than a mixture. Neat ethylene glycol freezes at about −13 °C. A 50/50 mix freezes at about −37 °C, and the best ratio — near 60/40 — protects to around −50 °C. Past that, adding more glycol makes the freezing point rise again, and heat transfer gets steadily worse the whole way, because water is far better at carrying heat than glycol is. Somebody topping up with neat antifreeze to be safe is making the car both more likely to freeze and more likely to overheat. The original states the freezing point of neat antifreeze correctly, which is more than most sources manage, and then never draws the conclusion.

Colour is not a specification. Green, pink, orange and blue mean whatever the blender chose. What matters is the inhibitor chemistry — IAT, OAT, HOAT, or a named manufacturer standard — and mixing incompatible ones drops the additive package out of solution as a gel that blocks the radiator and the heater matrix. Two red coolants can be entirely incompatible with each other. Read the specification on the bottle and match it to the handbook, never to the colour of what came out.

It is not seasonal. The original says to keep antifreeze in year-round and is right, but for the reason it gives second rather than first: the glycol is carrying a corrosion inhibitor package, and an engine full of plain water in July eats its own water pump, radiator and heater matrix from the inside. Freeze protection is the by-product people named it after.

One more thing the original does not say, and should have. Ethylene glycol is sweet, and it is lethal in quantities smaller than most people imagine — one or two teaspoons for a cat, three or four for a small dog, and it causes kidney failure in children who find a puddle of it. Do not leave it in an open container and do not let a leak sit on a driveway. Propylene-glycol coolants exist, are meaningfully less toxic, and are not sweet.

(The original calls the substance "ethyl glycol". It is ethylene glycol; ethyl glycol is a different chemical entirely.)

Refilling and bleeding

A cooling system will not fill itself, and an air lock introduced during a coolant change is one of the commonest reasons a car that was fine last week is overheating this week.

• Fill through the highest point in the circuit, which on most modern cars is the expansion tank and not the radiator. Some engines have a dedicated bleed screw on the thermostat housing or a heater hose; if yours has one, it exists because the system genuinely will not purge without it.
• Set the heater to full hot before you start. The matrix is a separate loop with its own high point, and a heater valve left closed traps air in it that comes back to haunt you a week later.
• Run the engine to temperature with the cap off, or with the tank's cap loose, until the thermostat opens and the level drops. It will drop suddenly, which is the thermostat opening and the radiator joining the circuit for the first time.
• Let it cool completely, then check the level again. One warm check proves nothing; air migrates as the system settles.
• Watch the top hose and the heater. A hot top hose and a hot heater at idle means the circuit is full. A hot engine with a cold heater means it is not, and running it in that state is how the head gasket becomes part of the story.

Two smaller corrections worth recording

The radiator is not made of copper. The original describes it as a copper tank, which was true of the cars it was written about and stopped being generally true decades ago. Almost every radiator now has an aluminium core crimped to moulded plastic end tanks — which matters practically rather than academically, because that crimped seam is where they leak, it cannot be soldered, and a leaking tank means a new radiator rather than a repaired one.

Its temperature figures contradict each other. In three places it gives the ideal engine temperature as 93 °C, as 73.8–82 °C, and as 75–95 °C, and the Fahrenheit conversion offered for the middle of those ("65 to 182") is not a conversion of anything. That is what stitching an article together from three textbooks looks like, and it is the reason the number to trust is the one on your own scan tool: 90–105 °C, measured, on the engine in front of you.

Video guides

Video, for the parts of this that are easier watched than read. The English-language ones come first.