POM / Acetal

Published · Researched 2026-09-21

The gear plastic. POM — polyoxymethylene, acetal, DuPont's Delrin — is what machined gears, bearings, bushings, and sliding mechanisms are made of: the lowest friction of any common thermoplastic, outstanding wear resistance, high stiffness, and dimensional stability that holds tolerances. In FDM it's a specialist filament with a reputation problem: the parts it makes are unmatched, and getting them to stick to the build plate is the kind of fight people write forum posts about for weeks. One experienced practitioner on the Bambu forums, having printed POM parts up to 150 mm across, calls it "the strongest, toughest, best layer adhesion material you can print with" — once dialed in. The key phrase is "once dialed in." Most people never get there, and the material has no interest in meeting you halfway.

What it is, in two sentences: POM is a highly crystalline engineering thermoplastic prized for low friction, wear resistance, and stiffness — the material you'd CNC a gear from. As a filament it's a micro-niche: essentially two to three active brands worldwide, thin documentation, and a print process with zero tolerance for improvisation.

What's genuinely good is the part you hold in your hand when it works. The friction numbers tell the story: POM's coefficient of friction against itself is roughly a third of nylon's and a tenth of ABS's (per reprap practitioner discussion), which is why gears, pulleys, linear bushings, and wear pads printed in POM actually behave like their machined counterparts instead of grinding themselves to dust. It is, per that same experienced Bambu-forum practitioner, the strongest and toughest printable material with the best layer adhesion — stronger praise than anything in the nylon or PC files earned. Dimensional stability is excellent once the part is down: it doesn't creep, doesn't absorb meaningful moisture, and holds its shape in service. The price is almost comically low for what it is — Gizmo Dorks POM still sells at $26.95/kg on Amazon, a price that has roughly halved since 2017. And crucially, the adhesion problem has a real answer now. The Bambu forum thread cracked it: POM sticks to G10/Garolite/Printbite beds — epoxy-fiberglass board — because bisphenol epoxies and acetals can bond chemically at the surface. The recipe is specific: scuff the G10 with fine abrasive, clean with acetone, bed at 110 °C, first layer at 260 °C, subsequent layers 250–255 °C, filament dried hard (70 °C in a food dehydrator), and otherwise just run a generic PA/PC profile. That's the first credible, repeatable POM workflow this project has found, and it comes from someone printing at 150 mm scale — not 20 mm cubes.

What's actually wrong starts with the bed and doesn't really end. Adhesion is everything and almost nothing works. Glue stick, PEI, glass, standard surfaces: all no. The fallback lore — blue painter's tape with hairspray — gives mixed results depending on the tape and hairspray brands, which is another way of saying it's not a solution. Cellulose/paper platforms (resume paper glued to borosilicate with Elmer's) work for small parts. Delrin tape works but becomes a one-shot "zeroth layer" that peels off the bed with any part bigger than ~20 mm. Only the G10/Garolite route has practitioner evidence at real scale, and it requires buying a dedicated bed. Then there's warping: POM is highly crystalline and shrinks "about the same as raw PA, which is a lot." Large parts contract hard, and the Prusa forum's verdict is blunt — for the warping and poor layer adhesion, you want a very hot enclosure or you want a different material. Safety is a real line item, not a footnote: overheated or degraded POM releases formaldehyde. Retailer guidance draws the line at 230 °C — do not keep POM heated above it — and ventilation is a requirement, not a suggestion. At proper print temps the release is minimal, but the margin between "printing" and "degrading" is narrower than on any other common filament. Availability is the third problem: two to three active brands globally (Gizmo Dorks in the US, Grilon3 in Argentina/LatAm), thin documentation, and Grilon3's pricing still unverified. You're buying into a micro-niche with no competitive pressure and no safety net. And a quiet note from the Prusa forums worth keeping: POM+GF and POM+CF filled grades exist, but the filler makes the already-fragile layer adhesion worse — don't reach for them expecting an easier print.

Pass 3 additions (2026-09-21): a correction on the temperature conflict. This review cites both a practitioner's 250–260 °C first-layer recipe on G10 and retailer guidance to never exceed 230 °C — those are contradictory sources, not one rule. The practitioner's 150 mm-scale G10 recipe comes with a hot chamber, aggressive drying, and ventilation; the 230 °C ceiling is retailer caution about formaldehyde from degraded POM. The honest framing: the process window between "sticking" and "degrading" is genuinely narrow, and you choose your risk between the two pieces of guidance. Fresh evidence adds a scale limit to the Delrin-tape workaround (works only around 20 mm; larger parts peel the tape off the bed), one more adhesive datapoint (wood glue on a 120 °C glass bed, printing around 205 °C), and Prusa-forum confirmation that GF/CF grades reduce warp but worsen the already-fragile layer adhesion — several owners concluded PCTG or machining was the rational move.

Who should skip it: anyone without a 110 °C-capable bed and an enclosure; anyone unwilling to buy or make a G10/Garolite build plate (this is close to a hard requirement, not a nice-to-have); anyone printing gears bigger than a coin without shrinkage compensation dialed in; anyone who needs the part quickly — POM is a tuning project before it's a printing project. Also skip it if "low friction" is a nice-to-have rather than the point: the Prusa forums' escape hatch is PCTG — easy to print, excellent layer adhesion, fairly tough, low abrasion, genuinely slippery-ish — for when you want slippery without the ordeal. And if the part could be CNC-machined from Delrin rod instead, the forums are blunt: machine it. FDM POM is for geometries machining can't do.

The social pulse is thin but unusually high-signal. Eight keyword queries across the big Facebook groups returned nothing — POM has no social-media constituency, no unboxing culture, no TikTok moment. Threads search returned nothing at all. What exists is concentrated practitioner knowledge: the Bambu forum thread with the 150 mm G10 recipe is the single most valuable POM document on the public internet, the reprap archives preserve a decade of failed adhesion experiments (Delrin tape, masonite, wood glue — the archaeology of desperation), and the Prusa forums host the pragmatic "just use PCTG" counter-movement. The shape of the debate is notable: nobody argues POM is easy, nobody argues it's unnecessary. The consensus is that it's the best material for the job and the worst material to get there — and that the people who succeed all converge on the same G10-and-hot recipe. That's a healthier signal than hype would be.

Value call, observed 2026-09-20: Gizmo Dorks POM at $26.95/kg ($25.60 Subscribe & Save) is the default US buy and has held that price for the better part of a year — absurdly cheap for an engineering filament, which is part of POM's strange economics: the plastic is cheap, the process is expensive. Desertcart Israel lists the black 1 kg at ₪215 incl. duties, in stock. Grilon3 (Argentina) pricing remains unverified, but it's the brand with the best public printing guidance — worth the email if you're in LatAm. The honest total cost: a $27 spool plus a G10 bed, a dehydrator you probably already own, ventilation, and a weekend of tuning. The spool is the cheapest part of printing POM.

One more honest note on longevity: the practitioners who stuck with POM report parts that outlast their PLA and PETG equivalents by embarrassing margins — gears that ran for years where printed nylon gears wore out. That's the payoff the whole ordeal buys. But it also means POM's failure mode is front-loaded: if the print survives the bed, the cooling, and the first assembly, it tends to survive everything after. Most filaments fail in service. POM fails on the build plate or not at all.

The tradeoff: POM makes the best low-friction functional parts in FDM and demands the most punishing process to get them. G10 bed, 110 °C, 260 °C first layer, dry it like you mean it, ventilate — and stay within your chosen temperature guidance, since the 230 °C retailer ceiling and the 250–260 °C practitioner recipe genuinely conflict. Do that and you get gears that behave like gears. Skip any of it and you get a very expensive lesson in crystallinity.