Machining Ultem (PEI): Stress Cracking, Tolerances, and Where It Beats PEEK

Ultem is the material customers reach for when a part has to survive heat, carry a load, pass a flame test and still cost less than PEEK. It machines cleanly and holds tight tolerances. It also has one failure mode that catches shops out repeatedly: parts that leave the building perfect and crack in the customer's hands a few weeks later, because of what they were washed in.

Here's what Ultem actually is, how we approach it on the shop floor, and how to decide when it's the right call.

What is Ultem?

Ultem is a brand name — originally General Electric's, now SABIC's — for polyetherimide, abbreviated PEI. It's an amber, semi-transparent, amorphous thermoplastic, and that word does more work than any other in this article.

Amorphous means the polymer chains have no crystalline order. Compared with a semi-crystalline plastic like PEEK or acetal, that gives Ultem three characteristics that shape everything downstream:

• It's dimensionally predictable. No crystallization means far less of the post-machining movement that catches people out on semi-crystalline stock.
• It's transparent. Amber-tinted rather than water-clear — not an optical-grade material — but you can see through it, which matters when a part has to double as an inspection window.
• It's more vulnerable to solvents and stress cracking. An amorphous polymer has no crystalline regions to resist chemical penetration. Ultem is unusually good for an amorphous resin — its maker says so explicitly, and its everyday chemical resistance is broad — but it has a short list of specific enemies, and that list is the reason for the coolant section below.

The unfilled grade — Ultem 1000 — is the one most machined parts are cut from. Glass-filled grades (Ultem 2100, 2200, 2300, at 10/20/30% glass) are stiffer and more dimensionally stable, at the cost of transparency and tool life.

Ultem 1000 properties at a glance

Values below are from Ensinger's TECAPEI natural (SABIC Ultem 1000 series) stock-shape datasheet, cross-checked against SABIC's own ULTEM 1000 resin sheet. The basis is not the same for every row, so it's marked — the mechanical rows come from extruded rod and plate, which is what a machined part is actually cut from; the thermal, electrical and flammability rows are injection-moulded or public-source data, exactly as the supplier footnotes them.

Treat these as typical published values, not a specification. Grade, fill, stock form and section size all move the numbers, so work from the supplier's datasheet for the exact stock on a critical part. Note that the tensile figure above is at break, on extruded stock — a yield-basis or moulding-resin figure for the same material runs lower, in the 15,000–16,000 psi region, which is why two datasheets for "Ultem 1000" can look like they disagree when they don't.

Three rows need a caveat rather than a footnote.

• Tg and the service temperature are measuring different things, which is why they're so far apart. Tg (~217 °C) is where the polymer softens; the ~170 °C figure is a UL 746B relative temperature index — an estimate of where it still holds its properties over a long service life. Design to the service number, not to Tg.
• The dielectric strength is the best-case figure, and two separate things pull it down. 830 V/mil is measured in air on a thin specimen. Move the same 1/16" specimen into oil and it reads 709 V/mil; go to a 1/8" section in oil and it is roughly 500 V/mil. Medium and thickness each cost you. If you're sizing a machined standoff or insulator, get the figure for your actual section in your actual medium — do not design to 830.
• The UL 94 rating is thickness-dependent, so a flame callout without a thickness means nothing. Unfilled Ultem 1000 is V-0 at 0.75 mm and above and 5VA at 3 mm, but only V-2 at 0.4 mm. What it does have is a rating that comes from the base polymer itself — no flame-retardant additive package — which is a large part of why Ultem ends up in aircraft interiors and electrical housings. Unfilled Ultem 1000 also passes FAR 25.853 at 3 mm.

The thing that actually ruins Ultem parts: stress cracking

This is the section to read twice.

Ultem is notch-sensitive and prone to environmental stress cracking. Two ingredients have to be present together: internal or applied stress, and a chemical the polymer doesn't like. Machining supplies the stress. The shop supplies the chemical, usually without meaning to.

Partially halogenated solvents will actually dissolve it — methylene chloride is the one that catches people out. Fully halogenated solvents are largely fine, which is why blanket statements about "chlorinated solvents" are only half right. The shop-floor rule still has to be stricter than the chemistry, though: nothing chlorinated, because closely related solvents sit on opposite sides of that line and nobody wants to be checking which one is in the parts washer.

And the one most likely to be sitting on your bench is the ketone rag. In the resin maker's own compatibility table, methyl ethyl ketone at room temperature ruptures a specimen held at just 0.25% strain in two hours — as fast as anything on that chart — and acetone is grouped in the same family. Aromatic-hydrocarbon-based cutting fluids attack it too, though straight aliphatics like heptane and naphtha are actually the supplier-recommended degreasers. Strong caustics are worth avoiding as well; the same guide tells you not to clean Ultem with them, though its own table only shows them biting once the solution is hot.

A part cut with the wrong coolant, or degreased in the wrong solvent, can measure perfectly and pass inspection, then develop crazing or cracks days or weeks later as residual stress works against a chemically weakened surface. The failure shows up at the customer, not at the machine — which is exactly why it keeps happening.

Three rules follow from that, and none of them are optional:

• Use a non-aromatic, water-soluble coolant, a spray mist, or compressed air. Nothing chlorinated. No ketones — no acetone or MEK wipe-downs. No strong caustics in the parts washer. This applies to the cleaning and degreasing steps just as much as the cutting — and it is worth saying out loud on the shop floor, because a ketone wipe passes the "is it chlorinated? is it petroleum-based?" test that most people actually run.
• Stress-relieve. Annealing before finish machining, and often again after, relieves the stress the cutter puts in. Follow the stock supplier's published schedule for the grade and section thickness — the right temperature and soak time depend on both, and guessing at either is how you warp a part instead of relieving it.
• Design out the notches. Sharp internal corners, abrupt section changes and undersized radii are where a notch-sensitive material starts its crack. Generous radii cost nothing at the print stage.

What tolerances you can realistically hold

Ultem machines well. It's rigid, it doesn't gum up, and being amorphous it doesn't move the way semi-crystalline stock does after the cut. Machined-plastic tolerance is best expressed proportionally — on the order of ±0.001" per inch of dimension — since a 1" part and a 12" part aren't the same problem. On well-planned Ultem setups we routinely hold ±0.002", and ±0.001" on selected critical features when the part is designed for it.

What limits you is heat and stress, not the material's machinability:

• Rough, relieve, then finish. Take the bulk off, let the part normalize or anneal, then cut the critical dimensions. Skipping the middle step is the most common reason a part is in spec at the machine and out of spec on the CMM.
• Keep it cool and clamp lightly. Ultem's thermal expansion is several times steel's. A hot cut grows the dimension you just measured, and heavy clamping distorts thin walls.
• Sharp, positive-rake carbide. Dull edges rub and generate the heat you're trying to avoid. Clear chips aggressively so they don't recut.

Glass-filled Ultem grades are stiffer and more stable but abrasive — plan on faster tool wear and expect to need coolant on drilling operations.

Ultem or PEEK?

They get compared constantly, and the honest answer is that they're not interchangeable.

Choose Ultem when the part needs high temperature and flame or smoke performance and electrical insulation, when you want transparency, when dimensional predictability matters more than the widest possible chemical window — or when the budget won't carry PEEK. It's the usual choice for aircraft interior components, electrical and semiconductor hardware, and steam-sterilizable medical parts.

That last one comes with a condition worth stating plainly. Ultem is genuinely hydrolytically stable — the resin maker's own product guide shows excellent tensile-stress retention after 10,000 hours in 100 °C water and rates the material for repeated autoclaving. But the same guide's chemical-compatibility table shows Ultem 1000 rupturing in 100 °C steam within 216 hours at 0.5% applied strain, while a specimen held at 0.25% strain is untouched after 21 days. Steam service is not a property of the material alone; it is a property of the material and how much stress is left in your part. A machined autoclave component has to be properly stress-relieved and designed to low strain, or the hydrolytic stability on the datasheet is not the stability you will get.

Choose PEEK when service temperature runs above Ultem's ~338 °F long-term rating, when the part sees aggressive solvents or hydrocarbons, or when it needs higher impact strength and wear resistance under load. On like-for-like extruded stock, PEEK's notched Izod runs roughly 0.90 ft-lb/in against Ultem's 0.60. PEEK is semi-crystalline, and that's precisely what widens its chemical window: Ultem's own chemical resistance is broad — it is specifically the partially halogenated hydrocarbons, and stress cracking under load, that it doesn't tolerate. PEEK tolerates both.

If the part lives at moderate temperature and sees mild chemistry, neither is the right answer — Delrin or nylon will do the job for a fraction of the cost. Our material comparison table puts the numbers side by side.

The takeaway

Ultem is a well-behaved material to cut and an easy one to ruin after the cut. Get the coolant and cleaning chemistry right, relieve the stress, design in radii, and it will hold tight tolerances in a part that survives heat, flame and voltage at a real cost saving over PEEK.

If you have an Ultem print — or you're weighing Ultem against PEEK and want a straight answer rather than a datasheet — our team has been CNC machining engineering plastics since 1968, and material selection is most of what we get asked about. Request a quote and send us the drawing.