A 38 mm PP closure lot came back from the customer with an out-of-round reading of 0.42 mm on the inside skirt diameter. Our drawing called for 38.05 ± 0.10 mm. The parts had passed every dimensional check on our own CMM at incoming. The difference was not the part. It was the measurement: we were checking at 20 °C on a contact gauge with a 2.5 N measuring force, and the customer was checking with an air gauge at 23 °C. The closure had been molded at 0.15 mm under nominal and was relaxing upward in the two weeks between shipment and their incoming inspection. That single lot is why I now treat vial closures dimensional tolerance capability as a process property with a time axis, not a number on a drawing.
Most tolerance arguments in this product category are really arguments about which process made the part, which dimension is being measured, and when. Get those three aligned and the arguments stop. This piece is about the third one that nobody writes down: what the molding method can actually hold, how you prove it, and which features on a vial closure stay stubborn no matter how good your press is.
Vial closures are not one manufacturing process. A 13 mm butyl septum, a 20 mm aluminum crimp seal, and a 38 mm polypropylene screw cap come off three different machines with three different capability profiles. If your supplier quotes a single tolerance band across the whole catalog, they are either averaging or guessing.
Here is what I have measured across suppliers and across our own presses. These are process capability (Cpk) values at 1.33 or better, meaning the process holds the tolerance band with room to spare, not the best single part anyone ever measured.
| Process | Typical dimension | Tolerance band held at Cpk ≥1.33 | Where it drifts |
|---|---|---|---|
| Injection molding, PP/HDPE cap | Inside skirt diameter | ±0.08 mm | Cavity-to-cavity imbalance, mold temperature gradient |
| Injection molding, PP/HDPE cap | Thread pitch diameter (GPI 400) | ±0.10 mm | Core wear, ejection marks on thread flank |
| Injection molding, PP/HDPE cap | Top thickness | ±0.05 mm | Pack pressure variation, gate freeze timing |
| Compression molding, butyl/silicone septum | Thickness | ±0.10 mm | Compound viscosity batch, cure exotherm |
| Compression molding, butyl/silicone septum | Diameter | ±0.15 mm | Flash trim, mold wear at parting line |
| Transfer molding, butyl septum | Thickness | ±0.08 mm | Pot pressure, cure time at thick section |
| Aluminum shell stamping + knurl | Outer diameter | ±0.10 mm | Tool wear on the knurl roll, strip thickness |
| Aluminum shell, tear-off skirt depth | Skirt length | ±0.20 mm | Die clearance, springback after forming |
Read that table as a floor, not a ceiling. A well-run injection cell on a 16-cavity tool can hold ±0.05 mm on skirt diameter for a year. A poorly run one on the same tool will hold ±0.15 mm and call it “within drawing” because the drawing was written at ±0.20 mm. The tolerance you get is the tolerance you specified, minus whatever the supplier quietly negotiated into the print.
One thing I insist on now: the process capability band on the print has to be tighter than the functional requirement, not equal to it. If the crimp seal needs the closure to sit within 0.15 mm of nominal to seal against a 20 mm vial finish, write the print at ±0.10 mm. Otherwise you are buying parts that are “in spec” and still leak. The vial closures catalog is only as good as the prints behind it.
Capability is a statistic, not a certificate
Anyone can print “Cpk 1.67” on a datasheet. The question is what population they sampled, over what time, and how they handled within-cavity versus between-cavity variation. Here is how I want to see it done.
Sample design that actually measures the process
A capability study on a 16-cavity tool needs at least 30 consecutive shots per cavity, not 30 parts pulled from one shot. Pulling 30 parts from one shot measures the tool, not the process. You want the shot-to-shot variation (pack pressure, melt temperature, cooling time) in the data. That means 480 parts minimum, run over at least two shifts, ideally across a material lot change.
Measure the same feature with a gauge that has 10:1 resolution against the tolerance. If the tolerance is ±0.10 mm, your gauge needs to resolve to 0.02 mm. A caliper at 0.01 mm resolution sounds fine until you realize the operator’s thumb force adds 0.03 mm of scatter. Use an air gauge or a non-contact optical comparator for anything under ±0.10 mm, and control the room to 20 ± 2 °C. Polypropylene moves about 0.15 mm per 10 °C on a 38 mm diameter. That is bigger than half the tolerance band.
Reading the numbers
Compute Cpk per cavity, then compute it again on the pooled data. If the pooled Cpk is much worse than the worst cavity, you have a cavity-to-cavity offset problem, which is a tool or hot-runner issue, not a process window issue. If the pooled Cpk is close to the worst cavity, the variation is common-cause and you need to widen the window or accept the band.
I also want Ppk on the first production lot, because Cpk on a stable process tells you what it should do and Ppk tells you what it did. A supplier who only reports Cpk from a golden run is telling you about a Tuesday in 2021. This is the same logic the analytical side applies to method validation, and it is worth reading how analytical testing labs handle capability and uncertainty, because the statistics transfer directly.
The four dimensions that will not stay put
Some features on a vial closure are easy. Skirt outside diameter on a round cap is a turning exercise. Top thickness on an injection-molded cap is a pack pressure exercise. The hard ones are hard for physical reasons, and knowing why saves you a year of arguing with a supplier who cannot fix them.
1. Inside skirt diameter and roundness
This is the dimension that decides whether the cap threads onto the vial without cross-threading. It is also the one that fights you hardest, because the part is a thin-walled cylinder that shrinks unevenly as it cools. A 38 mm PP cap with a 1.2 mm wall will be measurably out-of-round within 24 hours of molding, and the roundness will keep moving for a week as the polymer relaxes. If you measure at the press you get one number, and at the customer’s dock two weeks later you get another.
The fix is not tighter molding. It is post-mold conditioning: hold the parts at 23 °C for 48 hours before final gauging, and specify the roundness check at that point, not at the press. I have seen a 0.25 mm out-of-round reading drop to 0.08 mm just by letting the parts sit. If your supplier ships straight from the press, you are buying the press number and the customer is measuring the relaxed number.
2. Thread pitch diameter and flank angle
The GPI 400 thread standard (the 38-400, 24-400, 20-400 family) specifies pitch diameter and thread form, but the tolerance on a molded PP thread is wider than most people assume. Core wear is the silent killer: a thread core that has run 500,000 shots will have worn 0.02 to 0.04 mm on the flank, and the pitch diameter drifts with it. If the supplier does not track core life against a master gauge, the thread will drift out of the band over a production campaign and nobody notices until a capping line starts rejecting.
Ask for the core inspection interval. Anything longer than every 100,000 shots on a high-cavitation tool is optimistic. The chromatography consumables supply chain runs these tools hard, and core maintenance is where the good suppliers separate from the ones who just replace a tool when it fails.
3. Septum thickness and compression set
On a compression-molded butyl septum, thickness is the dimension that controls seal force, and it is also the one most affected by compound batch. A 3.0 mm nominal septum held at ±0.10 mm is achievable, but only if the compound Mooney viscosity is controlled batch to batch. I have watched a septum lot run 0.18 mm thick on the high side because the compound arrived stiffer and the press could not close the same way. The cure exotherm in a 3 mm section is also significant: the center cures later than the skin, and if the press opens too early the part relaxes and the thickness drifts on the low side.
The measurement that catches this is compression set after 22 hours at 100 °C, not just thickness. A septum that gauges in tolerance but has 40% compression set will leak after a week on the shelf. Ask for the set number on the certificate, and cross-check it against the EPA SW-846 method requirements for your application if you are running environmental samples.
4. Shell-to-septum concentricity on crimp seals
On an aluminum crimp seal, the septum has to sit concentric with the shell to within about 0.15 mm for the crimper to seat it evenly. This is a two-part assembly tolerance: the shell inner diameter plus the septum outer diameter plus the assembly fixture. It is where stamping variation and molding variation compound. A shell that is 0.10 mm oval and a septum that is 0.10 mm large gives you a 0.20 mm interference on one side and a gap on the other. The crimp looks fine and leaks at the gap.
The control is an assembly go/no-go gauge that checks concentricity after crimping, not before. I have seen suppliers check the shell and the septum separately, both in tolerance, and ship seals that fail a helium leak test at 1 × 10⁻⁶ mbar·L/s. The HPLC vial side of this problem is well documented; the GC crimp seal side is worse because the aluminum work-hardens and the crimp force has to be dialed in per shell lot.
Two field cases where the measurement, not the part, was wrong
Case 1: the drifting skirt diameter
Symptom: a customer running a 20 mm aluminum crimp line reported 3 to 5% capping rejects, all on one lot of closures, with the cap sitting high. Measurement that identified it: we pulled 50 parts and measured inside skirt diameter with an air gauge at 20 °C, then again after 72 hours at 23 °C. The as-molded parts read 20.02 mm mean with 0.06 mm standard deviation. The 72-hour parts read 20.11 mm mean with 0.09 mm standard deviation. The parts were relaxing upward past the upper limit. Fix: the molder was packing the cavity too hard to hit the low side of the band at the press, and the relaxation carried it out the top. We moved the target to nominal, reduced pack pressure by 8%, and added a 48-hour hold before final gauge. Rejects dropped to 0.4%.
Case 2: the concentricity gap on crimp seals
Symptom: a pharmaceutical customer’s helium leak test showed 0.2% failures on 20 mm tear-off seals, all from one shell lot. Measurement that identified it: we sectioned 20 failed seals and measured shell wall thickness at four quadrants. The mean was 0.28 mm with a 0.06 mm range across quadrants, against a 0.30 ± 0.02 mm spec. The shell was oval from a worn knurl roll, and the septum was centered in the oval, leaving a thin wall on two sides. Fix: replaced the knurl roll, added an ovality check (max minus min diameter, limit 0.08 mm) at shell incoming, and moved the septum assembly to a self-centering fixture. Leak failures went to zero over the next three lots.
Neither of these was a molding tolerance failure in the narrow sense. Both were failures of the measurement plan to capture the dimension that mattered at the time it mattered. That is the pattern I see most often, and it is why I now write the gauge plan before I write the purchase order.
A dimensional acceptance checklist you can hand to a supplier
This is the list I send with any RFQ for a new closure. It is not a sampling plan (that is a separate document). It is the capability and measurement agreement that has to be in place before the first lot ships.
- Feature list with functional rationale: every dimension on the print tied to a function (seal force, thread engagement, crimp concentricity). If a dimension has no function, remove it or widen it.
- Tolerance band at Cpk ≥1.33 on the process, stated per cavity, from a 30-shot-per-cavity study over two shifts.
- Gauge R&R under 10% of the tolerance band for each critical feature, with the gauge type and resolution named. Air gauge or optical for anything under ±0.10 mm.
- Conditioning before final gauge: 48 hours at 23 ± 2 °C for PP and HDPE parts, 24 hours for elastomer septa. State it on the inspection plan.
- Core and tool wear limits: thread core inspected against a master gauge every 100,000 shots; knurl roll ovality checked every 50,000 strokes; replacement trigger at 0.04 mm flank wear.
- Material lot traceability on compound Mooney viscosity and on PP melt flow index, with the acceptance window on the certificate.
- Assembly concentricity gauge on crimp seals, go/no-go, 100% or AQL-based, with the limit at 0.15 mm TIR.
- First-article inspection to ASQ/ISO 2859-1 Level II on the first three production lots, with the actual Cpk and Ppk reported, not just pass/fail.
If a supplier cannot produce the 30-shot-per-cavity data, that is your answer. They either do not have the tool instrumented or they do not want you to see the cavity spread. Either way, you are buying a promise. For reference on how the instrument side handles this, the Agilent GC consumables documentation and the Thermo Fisher chromatography technical notes both publish dimensional and lot-release data at a level of detail that closure suppliers should be matching. The Restek technical resources are also worth reading for how they handle lot-to-lot variation on septa and liners, because they treat it as a capability number, not a marketing claim.
The decision rule for a new closure program
Specify the functional tolerance first, then demand a process capability study that proves the chosen molding method holds it at Cpk ≥1.33 across all cavities and across a material lot change. If the method cannot hold it, change the method or change the design, not the print. And gauge every critical dimension after conditioning, at 23 °C, with a gauge that resolves to one-tenth of the band, because the number you measure at the press is not the number your customer will measure at the dock. That single rule has saved more lots than any press upgrade I have ever signed off on.
References & further reading
Related resources from our lab consumables network:
- Vial Closures at vialsclosures.com
- Analytical Testing at testinganalytical.com
- Hplc Vial at vialforhplc.com
- Chromatography Consumables at chromatographyconsumables.com
External references and standards: