Straight answers first. Unfilled PEEK holds about ±0.05 mm without special measures, and about ±0.02 mm on a named feature if it is annealed and measured at a stated temperature. Above roughly 100 mm, thermal expansion alone eats the tighter number. And the part that measured perfectly on Friday can be out of tolerance on Tuesday, because machining releases stress that was locked in when the rod was extruded.
This is the single most common reason a PEEK part fails inspection a week later, and it has nothing to do with the machine.
PEEK stock shapes are extruded or compression moulded, then cooled. Cooling leaves the outer skin and the core at different stresses. As long as the material stays whole, those stresses balance each other and the rod looks straight. Cut a pocket into one side, and you remove the material that was holding the balance — the remaining stresses pull the part, and it bows. The bow usually appears hours to days later, not during the cut, which is exactly what makes it dangerous: the part passes inspection at the machine and fails at the customer.
The countermeasure is annealing: heat the part above the glass transition (about 143 °C for PEEK) and below the melting point (about 343 °C), hold it, then cool it slowly — typically in steps, because a fast cool reintroduces the stress you just removed. Anneal the blank before the finishing cuts and again after roughing if the part is thin-walled. It costs a furnace cycle and it is the difference between a part that is right once and a part that is right for a year.
Numbers below are the practical machining envelope for unfilled PEEK, not a laboratory best case. The tighter number always carries a condition.
| Feature | Comfortable | Tight, with conditions | What the condition is |
|---|---|---|---|
| Outside diameter, turned | ±0.05 mm | ±0.02 mm | Annealed blank, finish pass light, measured at 20 °C |
| Bore | ±0.05 mm | ±0.025 mm | Reamed or bored, not drilled; check for taper |
| Thickness / length | ±0.05 mm | ±0.03 mm | Holds only under 100 mm; expansion takes over above that |
| Flatness on a 50 mm face | 0.05 mm | 0.02 mm | Requires annealing after roughing, plus a stress-relief pause |
| Position / true position | 0.10 mm | 0.05 mm | Single setup; re-clamping introduces error that no tolerance absorbs |
| Thin wall (under 1.5 mm) | Quote per part | — | Deflection during cutting is the limit, not the machine |
Two things sit outside any table. First, temperature: unfilled PEEK expands by roughly 47 µm per metre per °C, so a 200 mm part measured in a warm room and used in a cold one differs by more than the tolerance you specified. Write the measurement temperature on the drawing. Second, time: a tolerance that holds at the machine and not next week is not a tolerance. Ask for a measurement after the annealing cycle, or after 24 hours, and the number means something.
PEEK conducts heat about a thousand times worse than aluminium. The heat from cutting has nowhere to go except into the tool edge and the chip, so the failure mode is not a dull tool — it is smeared, gummy material welded to the flute.
| Operation | Surface speed | Feed | The thing that actually matters |
|---|---|---|---|
| Rough turning | 150–300 m/min | 0.1–0.25 mm/rev | Depth of cut generous; rubbing is what melts it |
| Finish turning | 200–400 m/min | 0.05–0.15 mm/rev | Sharp, polished insert with a positive rake |
| Rough milling | 200–400 m/min | 0.1–0.2 mm/tooth | Climb milling, high flute count, air blast |
| Drilling | 30–60 m/min | 0.05–0.15 mm/rev | Peck or through-tool air; back the drill out often |
| Reaming | 20–40 m/min | 0.1–0.3 mm/rev | Two thirds speed of drilling, never dry |
| Tapping | 10–20 m/min | — | Form taps beat cutting taps; undersize the hole |
These are published starting points for unfilled PEEK, not a record of the dial settings on our machines — those depend on the grade, the tool and the rigidity of the setup. Glass- and carbon-filled PEEK is abrasive and runs slower with carbide or diamond-coated tooling. Treat the table as a place to start and let the first cut tell you if the chip is coming off as a chip or smearing. If you are machining this yourself and want our first-article programme for comparison, ask for it with the quote.
The most expensive mistake is picking the cheapest material that survives the datasheet and ignoring what it does at temperature, under load, over a year.
| PEEK | POM (acetal) | PTFE | |
|---|---|---|---|
| Continuous service temperature | About 250 °C | About 90 °C | About 260 °C |
| Wear under load | Good, better filled | Good at room temperature | Poor — cold flows under steady load |
| Chemical resistance | Very wide | Poor against acids and fuels | The widest of any thermoplastic |
| Dimensional stability after machining | Good once annealed | Good; low moisture uptake | Poor — highest expansion of the three |
| Machinability | Fair; needs sharp tooling | The best of the three | Difficult; deforms rather than cuts |
| Relative cost of the blank | Highest | Low | Low to middle |
Read the table top-down for a part that runs hot, and bottom-up for a part that only has to be stiff and cheap. If the answer lands between two materials, the deciding question is usually the medium it sits in — a POM part in fuel fails at the seal, not at the load.
PEEK comes as rod, plate and tube in a fixed set of sizes. A 300 mm diameter part is a made-to-order blank with a lead time and a price, not something on a shelf. The quote says which.
We machine industrial grades. If the part goes into a body — or needs a master file, a validation pack or traceability to that standard — this is the wrong shop, and saying so early saves us both a month.
We can tell you what the material does dimensionally when cut. Long-term creep under load at temperature is a material property from the resin supplier, not something a machine shop measures. Ask them, not us.
Aluminium tools are for hundreds to low thousands. Past that, a steel tool at a production moulder is the correct answer, and we will point you at that instead of quoting something we cannot run well.
Grade, annealing and the measurement temperature are three decisions that cost almost nothing at quoting and a whole batch after delivery. Send the drawing and the duty; the rest is our problem.
Send a drawing for review