A printed bearing bracket spends three months in a wet workshop and then cracks overnight, while nobody has touched it. A PETG bracket holding a bottle of cleaner goes cloudy and then crumbles at the screw hole. A nylon clip that was perfect in the summer grows a fuzzy white bloom and snaps in the winter. None of these failures were mechanical. The load never increased, the part was never dropped, and the screws were never overtightened. Something in the environment got inside the polymer and changed it, and the change took weeks to become visible.
There is a reason this subject gets skipped. Most 3D printing advice talks about strength as a single number you can improve with wall count, orientation and infill, and all of that advice is correct and still incomplete. Strength is a starting condition. Service life is a process. A part can be enormously strong today and functionally dead in six months because of a slow chemical reaction that no print setting addresses.
This article is about that second number. We will go through the four different ways the environment attacks a printed part, why they produce such similar-looking failures, which filaments actually survive which chemicals, and what you can change at the design stage to buy years rather than weeks.
Why every failure starts in the same place: the amorphous problem
Almost all printable thermoplastics are amorphous or semi-crystalline polymers, and the ones that dominate our shelves are amorphous: PLA, ABS, ASA, PETG, polycarbonate, and the acrylic and styrenic photopolymers used in resin printing. Amorphous polymers have no ordered crystal regions holding the chains together. They are held together by chain entanglement and by relatively weak secondary bonds, and there is empty free volume distributed between the chains.
That free volume is what makes the material printable in the first place. It is also what makes it vulnerable, because small molecules can move into it. When a solvent, an oil, a cleaner, or even water enters that free volume, three things happen at once and in a predictable order. The polymer swells, because the absorbed molecules push the chains apart. The glass transition temperature of the plasticised surface layer drops, so a surface that was rigid at 25 °C becomes rubbery at 25 °C. And the mechanical strength of that softened layer falls, not gradually but steeply.
Notice that all three effects are concentrated at the surface. Absorption into a thick part is slow, and for most real chemicals at room temperature only the outer fraction of a millimetre is affected in the first weeks. This is good news and bad news. The bulk of the part stays strong, but the surface layer that carries the highest stress in bending and at every notch is exactly the layer that has been softened. That is why chemical attack so often announces itself as a crack at a stress concentration rather than as a uniform loss of strength.
The four mechanisms that actually kill parts
It helps to separate the four things that go wrong, because they need different fixes and they look alike on a failed part.
| Mechanism | What it does | Time to show | Signature on the failed part |
|---|---|---|---|
| Chemical attack | Breaks or rearranges polymer chains; strength and mass drop permanently | Days to weeks | Chalky or crazed surface, colour shift, part remains weak even when dry |
| Solvent absorption | Plasticises and swells the surface layer without breaking chains | Hours to days | Soft or tacky skin, dimensional growth, recovers partly when dried |
| Environmental stress cracking | A mild chemical plus sustained tensile stress opens cracks far below yield | Weeks to months | Brittle split at a notch, no visible chemical damage elsewhere |
| Hydrolysis | Water attacks the chain backbone, especially in polyesters and polyamides | Weeks to months | Loss of toughness, then brittle snap; often worst in warm or alkaline water |
The fourth one deserves a note because it explains a very common disappointment. PLA and PETG are polyesters, and nylon is a polyamide. All three have a backbone that water can attack, though at very different rates and under very different conditions. Warm water accelerates hydrolysis dramatically, and alkaline conditions accelerate it further. A PLA part that shrugs off a cold rinse can lose a third of its impact strength after a few weeks of standing in warm soapy water.
Environmental stress cracking: the failure nobody designs for
If you learn one thing from this article, make it this one. Environmental stress cracking, usually abbreviated ESC, is the mechanism behind most of the mysterious late failures in printed functional parts, and it is the mechanism that design decisions influence most directly.
The classic demonstration is simple enough that you can repeat it in an afternoon. Take two identical strips of the same filament, load both in three-point bending so the outer surface is in tension, and put one in air and one in a mild solvent such as isopropyl alcohol or a household cleaner. The dry strip holds its load for a very long time. The wet strip cracks at a stress well below its yield strength, often within minutes, with a sharp brittle split and no visible melting or softening at the fracture face.
Three conditions must all be present for ESC to occur. There must be a chemical that the polymer can absorb. There must be a sustained tensile stress, and sustained matters more than large. And there must be time. Remove any one of the three and the part survives.
That gives you three places to intervene, and the third one is the most underused. Absorbed chemical you control by choosing the material and by keeping the part out of contact with the wrong things. Sustained tensile stress you control by design: preload from screws, interference fits that are too tight, snap fits that are permanently deflected, parts that are bent as they are installed, and the residual stress left behind by printing itself. Time you control only indirectly, by lowering the other two.
The stress side is where printed parts are unusually exposed. Every print carries residual stress from differential cooling, and every layer interface is a plane along which that stress can open a crack once the surface has been plasticised. A part that would tolerate a mild cleaner at zero stress can fail in a week when it is also carrying a constant 8 MPa of preload from a screw threaded directly into plastic. Related mechanisms in creep, fatigue and preload compound the effect: the stress does not even need to be steady if the part is cyclically loaded.
Geometry concentrates the chemistry
Stress concentration and chemical attack multiply each other rather than adding. A sharp internal corner multiplies the local stress by a factor of two to three. It also traps solution, because a sharp crevice holds liquid against the surface where evaporation is impossible. And the crack tip that forms in the corner has fresh polymer exposed at every increment of growth, so the attack never exhausts itself at the surface. This is why the failures appear at the same five locations over and over: the screw hole, the base of a boss, the inside of a snap hook, the transition from a thick section to a thin one, and the root of a printed thread.
Material by material: what each filament actually tolerates
The table below is a practical ranking, not a chemical compatibility chart for a spec sheet. It describes what you should expect from a well-printed, dry, well-bonded part in intermittent contact at room temperature. Continuous immersion, elevated temperature, or both will move every entry downwards, and the third column tells you how badly.
| Material | What it handles well | What it does not | Sustained-load risk |
|---|---|---|---|
| PLA | Dry indoor air, brief water contact, dilute acids at room temperature | Alkalis and warm water (hydrolysis), chlorinated solvents, acetone, long outdoor exposure | Very high - creeps and hydrolyses at once |
| PETG | Water, alcohols, dilute acids, most household cleaners in brief contact | Strong solvents, long immersion in alkaline cleaners, some adhesive solvents | Moderate - softens and crazes under load |
| ABS | Dilute acids and alkalis, water, most oils | Acetone, ketones, esters, chlorinated solvents, UV | Moderate - ESC sensitive at notches |
| ASA | Everything ABS handles, plus outdoor UV and weathering | The same solvents as ABS | Moderate |
| PA6-CF | Oils, fuels, greases, most organic solvents | Strong acids, phenols, hot water, alkalis; moisture uptake is significant | Low to moderate once dry |
| PET-CF | Broad chemical range, low moisture uptake, good at moderate temperature | Strong alkalis, some chlorinated solvents at temperature | Low |
| PPS-CF | Nearly everything at room temperature, including fuels, acids, alkalis and most solvents | Very few common chemicals; cost and print temperature are the real limits | Very low |
| TPU 95A | Oils, water, alcohols, most mild cleaners | Strong solvents, ketones, concentrated acids; hydrolysis in hot water | Low, but it deforms rather than cracks |
Two entries in that table deserve expansion. Polypropylene, which is genuinely excellent against solvents and acids, is largely absent from desktop printing because it warps spectacularly and sticks to almost nothing, so the practical solvent-resistant options are PPS and, to a lesser degree, PET-CF. And nylon's reputation for chemical resistance is real but conditional: dry PA6-CF resists fuels and oils better than anything else on a typical shelf, yet it absorbs moisture continuously and loses stiffness as it does, which is a slow environmental failure that looks like a design failure. Polycarbonate sits in a similar conditional position: chemically reasonable, but hydrolytically fragile if it is printed wet.
Resin is a different category, not a stronger filament
Photopolymer parts behave differently enough to warrant their own warning. A cured resin is a crosslinked network, which means it does not dissolve the way a thermoplastic does. Where a thermoplastic swells and softens, a resin tends to be attacked at the surface, and the damage shows up as chalking, colour change, or a loss of surface integrity. The real vulnerability of resin is mechanical: most standard resins are strong in compression and brittle in tension, so they crack rather than deform, and a chemically softened surface on a brittle part has nowhere to redistribute stress.
This is why resin parts that live in contact with anything aggressive are usually specified as engineering impact or high temperature grades rather than standard resin. It is also why the solvent story for resin is genuinely different from the filament story, and we covered that separately in resin prints in solvents, oils and caustics.
The specific threats nobody puts on the list
Most of the failed printed parts in circulation were not killed by an industrial chemical. They were killed by something ordinary, and that ordinariness is why the failure was never diagnosed.
| Threat | What it attacks | Practical mitigation |
|---|---|---|
| Isopropyl alcohol, repeatedly | Plasticises and crazes PETG and polycarbonate; attacks some resins | Wipe, do not soak; use the lowest concentration that works |
| Acetone and generic "super glue" solvents | Dissolves ABS and ASA on contact; attacks resin surfaces | Use dedicated plastics adhesives; keep solvent vapours away |
| Alkaline degreasers and dish soap in warm water | Hydrolyses PLA and PETG; stains nylon | Rinse cold, dry immediately, do not leave parts soaking |
| Chlorine from pools and cleaning bleach | Attacks nylons and polyamides severely | Do not put printed parts in chlorinated water permanently |
| Sunlight and heat together | UV embrittles ABS and PETG; heat accelerates every other mechanism | ASA or ASA-CF outdoors, and shade is worth more than any filament |
| Oils, greases and fuel | Mostly benign to nylon and PPS; swells TPU slowly; softens PLA | Nylon or PPS for anything in permanent contact with lubricant |
| Thread sealants and anaerobic adhesives | Some formulations attack polycarbonate and styrenics at a stressed thread root | Choose sealant compatible with the plastic, or isolate with a metal insert |
That last row is worth dwelling on because it combines all three ESC conditions in one place: a chemical, a thread root that concentrates stress, and a permanent preload. A printed threaded boss that survives for a year in dry air can crack in a month when the same thread is assembled with an aggressive sealant. If the joint matters, put a metal insert in it and read creepage, inserts and potting for the surrounding geometry.
Wall thickness and surface finish are chemical variables
Here is the design lever most people miss. If attack is confined to the surface, then the depth of material you can afford to lose is a design parameter, and the way you print the part determines how much of it is exposed.
A part printed with four walls and sparse infill does not present a solid barrier to a liquid. Liquid enters through any pore at a layer interface, and once it is inside, the internal surfaces of the infill are exposed with no way for the liquid to evaporate out. The part then behaves like a sponge holding the chemical against the polymer for months. This is a mechanism that no amount of external coating fixes, because the reservoir is internal.
The practical rules follow from that. Parts that will see chemicals should be printed with enough perimeter walls that the wall is essentially solid, and with a higher infill percentage than a purely structural part would need. Fewer, thicker walls beat many thin ones, and the reasoning is the same as in the discussion of load paths: the wall is the part.
Why a glossy surface is not a protective surface
Print orientation changes which surface the chemical meets. A part printed so that its functional face is the top surface has smooth, ironed-looking exteriors and a dense skin. A part printed so that the same face is a support interface has a rough, porous surface with exposed layer edges, and that surface holds liquid by capillary action while presenting many more entry points. If the part sees chemicals, flip the orientation even at the cost of print time. The same reasoning that makes surface finish matter for geometric accuracy applies directly to service life.
Design rules that buy years, not weeks
These are the changes that consistently separate parts that survive from parts that crack at the screw hole. They are all geometry, and none of them require a different material.
- Round every internal corner. A radius of half the local wall thickness removes the stress concentration that ESC needs. A sharp corner is a crack starter with a built-in liquid trap.
- Never let a thread be the only fastening. Use a through-bolt with a nut, or a heat-set insert, so the plastic is in compression rather than carrying the preload in tension at a thread root.
- Design snap fits to be relaxed in service. A hook that stays deflected for years is a part that is already being stress-cracked, and the chemical only decides which week it fails.
- Give gaskets a real groove with lateral constraint. A seal that is squeezed in an unconfined gap extrudes and loses contact pressure; the geometry for that is covered in sealing printed parts and in stopping capillary leaks.
- Keep chemical contact away from joints. Two parts bolted together form a crevice, and a crevice holds solution. If the joint must be in the wet zone, seal it rather than relying on the clamp.
- Leave drainage and vent holes. A blind pocket on a chemical-exposed part will fill and stay full. A 2 mm drain costs nothing and eliminates a permanent reservoir.
- Print dry, and dry properly. Moisture in the filament is hydrolysis before the part is even in service. The mechanism is the same as why PLA forgives you and PETG does not, and a dryer is a chemical-resistance tool as much as a print-quality tool.
Testing it yourself, before the part is in service
You do not need a lab to find out whether a design will survive. You need three cheap tests and the discipline to run them before you commit to a print farm run of twenty parts.
The first is a soak test on a witness coupon. Print five small bars in the same material, orientation and settings as the real part, measure and weigh them, and immerse them in the actual chemical at the actual service temperature. Pull one out at 24 hours, one at a week, one at a month. Measure mass gain, dimensions and a simple bend strength, and look at the surface. A part that gains more than about one percent in mass in a week is absorbing seriously.
The second is a stressed soak, which is the one that catches ESC. Load a coupon in constant bending with a simple jig so that the outer surface is at roughly a third of the material's yield stress, and soak it in the same chemical alongside the unloaded coupons. If the loaded coupon cracks while the unloaded one does not, you have found an ESC problem, and no amount of additional wall thickness will fix it. You need to change the stress or the chemical.
The third is a real-part test in the real environment, and its only requirement is that it runs long enough. Two weeks is not long enough to prove anything about a mechanism that operates over months. Build the test into a place where it will be seen: a spare part installed next to the working part, checked monthly, with the date written on it in paint marker.
Reading a failed part correctly
The failure surface tells you which mechanism was at work, and reading it correctly is the difference between fixing the design and reprinting the same mistake.
- A clean brittle split at a notch with no surface discolouration usually means ESC. The chemical was mild and the stress did the work.
- A chalky, matte or colour-shifted surface around the failure means direct chemical attack, and the material choice was wrong for the environment.
- A soft, swollen or tacky region means absorption, and the part may still be dimensionally out of tolerance even if it has not cracked.
- A fuzzy white bloom on nylon is plasticiser and low molecular weight material migrating to the surface, and it usually means the part has absorbed enough water to be dimensionally unreliable, not just chemically attacked.
- An interface split that follows a layer line with no chemical damage anywhere means the print was weak to begin with, and the environment only finished the job.
When the honest answer is to stop printing the part
There is a category of application where the correct engineering decision is not a better filament. Continuous immersion in strong solvents at elevated temperature, permanent contact with fuels under pressure, parts that must hold a seal against an aggressive chemical for years, and anything where a failure presents a real safety or environmental hazard should not be a printed polymer part. A printed prototype proves the geometry and the fit, and the production part is machined from PPS, PVDF, or metal.
That is not a defeat. Printed parts earn their place in chemical environments in the applications where the exposure is intermittent, the stress is low, or the part is cheap enough to be a consumable. A jig that gets wiped with solvent and used for a year, a bracket in a splash zone, a drying-box fitting, a cover that never touches the chemical at all: those are printed-part problems, and the rules above make them reliable.
The one-page checklist
- Name the chemical and the temperature, in writing, before choosing a material. "It gets wet" is not a specification.
- Decide whether the contact is splashed, intermittent or continuous. Only continuous contact demands the top of the material list.
- Reduce sustained tensile stress wherever you can. No printed thread carries a preload in a chemical zone.
- Round internal corners to at least half the wall thickness.
- Print walls solid and infill dense enough that internal volume is not a liquid reservoir.
- Orient the part so functional faces are dense top surfaces, not support interfaces.
- Dry the filament before printing, and check that it stayed dry. See how to dry PLA properly, and note that a heated drying cabinet removes the guesswork.
- Run a soak test and a stressed soak test on coupons before committing to a batch.
- Install one sacrificial witness part and inspect it monthly, with the date on it.
- If the exposure is continuous, hot, and safety-relevant, change the material class rather than the design.
None of this is exotic. It is the ordinary discipline of asking what the environment will do over the life of the part rather than only what the load does on the day it is installed. Printed parts are perfectly capable of surviving hostile environments, but only when the design treats the environment as a load. You already know how to design for a load. This is the same skill pointed at a slower adversary.