How I stopped ruining tensile specimens

My first month of tensile data was unusable and I blamed the films. It was the specimen preparation.

The symptom

Roughly a third of specimens failed at or very near the grip rather than in the gauge section. A grip failure tells you about your clamping, not about your material, so those runs have to be discarded — and when it is a third of them, the surviving data is biased toward whichever specimens happened to be cut well.

What was going wrong

Three things, in order of how much they mattered:

1. Scissor-cut edges. I was cutting with a scalpel against a steel rule. Any nick in the edge of the gauge section is a stress concentrator, and on a 90 µm film a nick you cannot see with the naked eye is enough.
2. Uneven thickness. Solution-cast films are thicker at the edge of the dish. I was cutting specimens wherever there was room, so nominally identical specimens differed by 15 µm or more.
3. Grip pressure. Too much and you crush and pre-damage the film at the grip line; too little and it slips, which looks like a long false elongation at the start of the curve.

What works now

I had a die cutter made to the ASTM D882 strip geometry, which removes the edge-quality problem entirely. I map thickness across each cast film with a digital micrometer at nine points and only cut specimens from the central region where thickness is within ±5 µm. And I put a strip of masking tape on the film at each grip line, which spreads the clamping load enough to stop the crush failures.

Grip failures are now well under one in ten. The remaining ones are almost always films I should have rejected on thickness before cutting.

None of this is novel — it is standard practice that nobody had written down for our lab. That is largely why I wrote it down.