Environmental Stress Cracking: The Part That Fails Weeks After It Ships

What ESC actually is

Environmental stress cracking is brittle failure of a plastic under a stress it would survive indefinitely on its own, in the presence of a chemical it would survive indefinitely on its own. Neither the stress nor the chemical is sufficient; together they are.

The mechanism is not chemical attack in the usual sense. The chemical does not dissolve the polymer or react with it. It adsorbs at the tip of a growing crack, lowers the surface energy needed to extend it, and lets the crack grow at a stress far below the bulk yield strength. The result is a brittle fracture with no visible deformation, often with a shiny or slightly crazed surface.

That is why it is so often misdiagnosed. A part that fails in service with a clean break and no stretching looks like an overload, and the investigation looks for a mechanical cause. The actual cause is a cleaning agent, a lubricant, a plasticiser migrating from a gasket, or the mould release agent that was never washed off.

Stress and chemical together

Time to failure falls steeply and predictably with stress. In a standard solvent at 20 percent of yield strength a part lasts about 2000 hours, at 30 percent 600 hours, at 40 percent 150, at 50 percent 40 and at 60 percent 12. Each additional ten percentage points of stress costs roughly half a decade of life.

Two consequences follow. The first is that a small reduction in stress buys a large increase in life: dropping from 40 to 30 percent takes the life from 150 hours to 600, a factor of four, for a modest design change. The second is that accelerated testing has to be interpreted carefully, because a test at 60 percent of yield that passes in 12 hours says almost nothing about a part running at 25 percent.

The chemical side has a threshold too. Below a certain concentration or above a certain molecular size, the agent cannot penetrate to the crack tip fast enough to matter, and the part behaves as if the chemical were not there. That is why a part can survive a splash and fail in a puddle, and why pooled liquid is far more damaging than the same volume spread thin.

Table 1 — Time to failure against stress in a standard solvent, on a log scale.

Which polymers resist

The single best predictor of ESC resistance is crystallinity. A semi-crystalline polymer has crystalline regions that a solvent cannot penetrate, so the crack has to find a path through the amorphous fraction, and that path is tortuous. An amorphous polymer has no such barrier, and the solvent diffuses freely.

Measured as the critical strain at which a bent strip cracks in a standard solvent, polyethylene survives 3.5 percent, polypropylene 2.5, ABS 0.6 and polycarbonate 0.4. That is nearly a factor of nine between PE and PC, and it is the reason a detergent bottle is HDPE and a lens is not.

Molecular weight matters almost as much within a grade. A higher molecular weight gives more tie molecules between crystallites and a longer crack path, so a high-molecular-weight HDPE resists ESC far better than a low-molecular-weight one of the same density. This is why pipe grades are specified on molecular weight as well as on density.

Table 2 — Critical strain in the same solvent, and what it means in service.

The four places it starts

ESC always starts at a stress concentration, and there are four that account for most field failures. The first is a sharp internal corner, where the local stress can be three to five times the nominal. The second is a weld line, where the material has already been weakened by the meeting of two flow fronts.

The third is a press fit or an over-tight screw boss. A boss that is 0.1 mm too small for the screw is under hoop stress for its whole life, and that stress is exactly what the solvent needs. The fourth is the gate area, where frozen-in orientation leaves a residual stress that survives annealing and shows up as a crack that starts at the gate and runs outward.

The common factor is that all four are invisible in a drawing review. A part can be dimensionally correct, meet its tolerance, and still have a 4 percent strain at a corner that nobody calculated. A stress analysis that shows the nominal stress is 10 MPa is not much help if the corner is running at 40.

Designing it out

The order of preference is to remove the chemical, then to reduce the stress, then to change the material. Removing the chemical is usually the cheapest and the most often skipped, because the chemical is often something the customer adds after the part is installed.

1. Find the agent. Cleaning fluids, thread lockers, lubricants, gasket plasticisers and mould release are the usual suspects.
2. Increase the corner radius to at least half the wall thickness, and preferably equal to it.
3. Reduce press fits to a strain below 0.5 percent, or use a clip instead of an interference fit.
4. Anneal the part to relax the frozen-in stress, particularly around gates and thick sections.
5. Change to a semi-crystalline grade, or a higher molecular weight within the same family.

Annealing is the most under-used of these. A two-hour anneal at 10 to 15 C below the heat deflection temperature can remove 60 to 80 percent of the moulded-in stress, and it costs oven time rather than tool changes. On a part that has already passed every other test, it is often the difference between a field failure and a long life.

Testing for it

There are three standard approaches and they answer different questions. A bent strip test puts a constant strain on a specimen, immerses it in the agent, and records the time to crack. It is fast, cheap and good for ranking materials, but it says little about a real part's stress state.

A constant load test hangs a weight on a specimen in the agent and measures time to failure. It gives the stress-life curve that the numbers in Table 1 come from, and it is the right test when the design stress is known. A full notch creep test on a notched specimen is the most severe and the most repeatable, and it is the one used for pipe grades.

For a real part, the most useful test is to expose a production sample to the actual agent at the actual temperature under the actual assembly load, and to run it for at least ten times the expected service exposure. That is slow, and it is the only test that catches the interaction between a real stress concentration and a real chemical.