A customer sent back a lot of 480,000 vials last March because their crimper started rejecting caps at a rate of 1 in 340. The reject criterion was simple: the seal sat proud by more than 0.15 mm over the neck shoulder. We pulled 30 retained samples and measured neck OD on a vision system. The lot ran 0.28 mm below nominal on the outside diameter of the neck, which put the cap skirt past its interference band. The root cause was not the crimper. It was a worn core pin in cavity 14 that had been running since the previous mold service. That is the kind of failure the phrase 2ml vial dimensional tolerance capability is supposed to prevent, and it only works if you know which dimensions your process can actually hold and which ones you are only hoping to hold.

This article is about capability, not conformance. A certificate of analysis tells you a lot passed. It does not tell you how much margin is left before the next lot fails. Those are different questions, and only one of them predicts a recall.

Most incoming inspection on a 2ml vial is attribute-based: go/no-go on a handful of dimensions, sign the sheet, release the lot. That tells you nothing about how close the process is running to the limit. A lot can pass at 100% and still be one shift away from a 4% reject rate.

Capability is expressed as Cp and CpK. Cp is the ratio of the tolerance width to six sigma of the process. CpK accounts for centering. For a 2 mL vial, the practical targets I work to are:

• CpK ≥ 1.33 on critical-to-function dimensions (neck OD, bore ID, height, seal land flatness). This is the floor for a stable process.
• CpK ≥ 1.67 on anything that touches the autosampler or the crimper. Neck OD and crimp ring height sit here.
• CpK ≥ 1.0 is acceptable only on cosmetic dimensions where a failure does not stop the line.

The number that matters more than the target is the sample size behind it. A CpK calculated from 30 parts is a point estimate with a confidence interval wide enough to hide a bad cavity. I want at least 125 parts per lot for a routine CpK, and a full 32-cavity study at PPAP with 5 parts per cavity minimum. Anything less and you are measuring noise.

One more thing about CpK on molded glass or plastic: it is only valid if the process is in statistical control. Run an X-bar/R chart first. If the mean is drifting, CpK is meaningless because the distribution is not stationary. I have seen suppliers quote CpK of 1.8 on a dimension that was walking 0.02 mm per 10,000 shots because the tool was cooling unevenly. The chart caught it. The CpK did not.

What each molding process can actually hold on the neck, bore, and height

The single biggest determinant of tolerance capability is the forming process, not the tool. A good tool in the wrong process will not hold what the drawing asks. Here is what I have measured across suppliers over the last decade, expressed as the tightest tolerance I will accept on a production drawing without a capability study to back it:

Read that table as a starting negotiation, not a spec. The numbers assume a tool built to the tight end, a machine with closed-loop shot control, and a resin lot with a tight melt flow index. Drop any one of those and the achievable tolerance loosens by 30 to 50%.

For glass, the neck is the hard part. Tubing conversion gives you a good bore because the ID comes from the cane, but the neck and crimp ring are formed hot and the tool wear shows up there first. A glass neck that starts at ±0.12 mm will be at ±0.20 mm after 200,000 pieces if the forming jaws are not dressed. Ask your supplier for the jaw maintenance interval in shots, not months.

For plastic, the bore is the hard part on COC and COP because of shrinkage anisotropy. The resin shrinks more along the flow direction than across it, and the bore sits in a region where the flow front has already turned. You get an ovality of 0.03 to 0.05 mm on a nominal 6.0 mm bore unless the gate is placed carefully and the cooling is balanced. Ovality is not on most drawings. It should be, because it is what limits autosampler needle clearance.

The four dimensions that eat your yield, and the ones that don’t

Not every dimension on a 2 mL vial drawing deserves the same attention. Four of them drive almost all field failures. The rest are either cosmetic or self-correcting.

Neck outside diameter and crimp ring height

This is the pair that determines whether a crimp seals. A standard 11 mm aluminum crimp cap expects a neck OD in a band of roughly 10.95 to 11.05 mm at the crimp ring, and the ring height must be within about ±0.10 mm of nominal or the cap skirt either bottoms out or leaves a gap. GPI 11 mm is the reference; Agilent’s consumables documentation and the cap drawings from any major supplier will confirm the interference band. If your neck OD capability is ±0.10 mm and your crimper is set for nominal, you have zero margin. I want the neck OD capability at ±0.06 mm so the crimper has room.

Bore inside diameter and ovality

The bore has to clear a 0.6 mm or 0.5 mm autosampler needle without scraping the septum on the way in. A 6.0 mm nominal bore with 0.05 mm ovality is fine. A 5.9 mm bore with 0.05 mm ovality is not, because the minor axis is 5.85 mm and the needle guide geometry starts to bind. Draw the bore at 6.0 mm ±0.08 mm and call out ovality separately at 0.04 mm max.

Overall height and bottom thickness

Height matters for two reasons: autosampler gripper clearance and the draw volume. A 32 mm vial that runs 32.4 mm will not sit flush in every tray. Bottom thickness matters more than most people think because it sets the residual volume. A 0.2 mm variation in bottom thickness is 0.2 mm of dead volume, which on a 6.0 mm bore is about 5.7 µL. On a 1.5 mL draw that is 0.4% of your sample sitting below the needle tip. Not fatal, but it shows up in low-volume methods.

Seal land flatness

This is the one that gets ignored and should not be. The seal land is the annular face the septum sits on. If it is not flat within 0.05 mm total indicator reading (TIR), the septum deforms unevenly and you get a leak path that only shows up under pressure. Flatness is a tooling issue, not a process issue, and it degrades with tool wear. A seal land that starts at 0.03 mm TIR will be at 0.08 mm after 500,000 shots if the cavity insert is not replaced.

How to run a capability study that predicts field failures

The mistake I see most often is a capability study run on the wrong sample. A supplier pulls 30 parts from the end of a run, measures them on a CMM, and reports CpK of 1.6. That sample is biased three ways: it is from one cavity, one resin lot, and one shift. It tells you nothing about the cavity-to-cavity spread that will actually bite you.

Here is the protocol I use:

• Sample every cavity. On a 32-cavity tool, pull 5 consecutive shots and measure all 32 cavities. That is 160 parts. You are looking for the spread between the best and worst cavity, not the mean.
• Measure the feature the way the customer uses it. Neck OD should be measured with a contact gauge at the crimp ring, not with calipers across the whole neck. Calipers read the largest diameter and hide ovality.
• Run the study at production speed. A capability study at 60% of cycle time will show better numbers than the process delivers. I have seen a 0.04 mm difference in neck OD between a slow trial and full rate because the cooling time was cut.
• Repeat across three resin lots. Shrinkage varies with melt flow index. A resin lot at the high end of the MFI band will run 0.02 to 0.03 mm smaller than one at the low end. If your capability study only covers one lot, you are measuring that lot, not the process.

Then plot the cavity means on a control chart. If cavity 14 is 0.15 mm off the others, you have a worn core pin, a blocked cooling channel, or a gate that is partially plugged. That is the finding that saves you a recall, and it does not show up in a 30-piece study.

Two field cases where the dimension was not the one on the drawing

Case 1: the crimp that leaked only on the third capping station

Symptom: a customer running three crimping stations saw leak failures only on station 3, at a rate of about 2 per 1,000. The vials passed incoming inspection. The caps passed incoming inspection. The crimper was within its force spec.

Measurement: we measured neck OD at three heights on the crimp ring (top, middle, bottom) on 50 vials from the same lot. The mean was fine. The taper was not. The neck was 0.04 mm smaller at the bottom of the ring than at the top. That taper was within the OD tolerance band, so it passed. But station 3 had a slightly worn crimping jaw that applied force lower on the ring, and the taper meant the jaw was closing on a smaller diameter than the other two stations. The seal was marginal, and station 3 pushed it over.

Fix: we added a taper callout to the drawing (0.02 mm max over the ring height) and replaced the station 3 jaw. The taper callout is now standard on every 2 mL vial drawing we release. It costs nothing at the tool and catches a failure mode that OD alone cannot.

Case 2: the autosampler that missed every 40th vial

Symptom: a lab running a 200-vial sequence on an autosampler saw the needle fail to pierce on roughly 1 in 40 vials. The needle would bend or the vial would lift. The lab blamed the septum.

Measurement: we measured seal land flatness on 200 vials from the lot and found a bimodal distribution. About 5% of vials had a seal land TIR of 0.09 to 0.12 mm; the rest were under 0.04 mm. The high-TIR vials were all from one cavity. The cavity insert had a burr on the sealing face that held the septum off-center, and the needle was hitting the raised edge of the septum instead of the center.

Fix: cavity insert replaced, and we added a 100% inline vision check on seal land flatness for that product family. The check runs at 400 parts per minute and rejects at 0.06 mm TIR. That is the kind of inspection that pays for itself in one prevented service call.

The measurement system is part of the tolerance

You cannot hold ±0.06 mm on neck OD if your gauge repeats to ±0.03 mm. Before you accept a capability number from a supplier, ask for a gauge repeatability and reproducibility (GR&R) study on the critical dimensions. The rule I use: GR&R must be under 10% of the tolerance for the dimension to be usable, and under 30% to be acceptable at all. A supplier quoting CpK of 1.5 on a dimension with 25% GR&R is quoting a number that is mostly gauge noise.

The common failure here is measuring neck OD with calipers. Calipers have a resolution of 0.01 mm but a repeatability closer to 0.02 mm in the hands of a production operator, and they read the maximum diameter across the jaw, which hides ovality and taper. Use a blade micrometer or a vision system. For bore ID, use pin gauges or an air gauge, not a telescoping gauge. For seal land flatness, use a flatness gauge or a profilometer, not a visual check.

If you are sourcing across multiple suppliers, insist on the same measurement method. I have seen two suppliers report neck OD within 0.02 mm of each other on paper while their parts differed by 0.08 mm in the crimper. The difference was the gauge, not the part. A shared measurement protocol is worth more than a tighter drawing.

What to put on the drawing, and what to leave to the process

A 2 mL vial drawing with 40 dimensions is a drawing that will be ignored. I keep it to the features that matter and let the rest run to a general tolerance. Here is the decision checklist I use when releasing a drawing for a new OEM vial program:

• Neck OD at the crimp ring: ±0.06 mm, with a taper callout of 0.02 mm max. Reference GPI 11 mm.
• Crimp ring height: ±0.08 mm. This is what the cap skirt engages.
• Bore ID: 6.0 mm ±0.08 mm, with ovality 0.04 mm max.
• Overall height: ±0.15 mm. Tighter than this is usually not worth the tool cost.
• Seal land flatness: 0.05 mm TIR. This is a tooling callout, not a process one.
• Bottom thickness: ±0.10 mm. Sets residual volume.
• Concentricity of bore to neck OD: 0.10 mm. This is what keeps the needle centered.
• General tolerance: ISO 2768-m for everything else.

Everything not on that list should be a general tolerance and should not be inspected at receiving. Inspecting 40 dimensions on a 2 mL vial is how you get a supplier who ships good parts and a receiving dock that rejects them for a cosmetic flash line.

One more drawing note: call out the resin or glass type and the shrinkage allowance you expect. For PP, shrinkage runs 1.2 to 2.0% depending on wall thickness and cooling. If the tool is cut for 1.5% and the resin runs at 1.8%, every dimension shrinks by 0.3% and your neck OD capability disappears. The drawing should state the shrinkage basis so the toolmaker and the molder are working from the same number.

The decision rule for tolerance capability

Before you release a 2 mL vial program, demand a 32-cavity capability study with 5 parts per cavity, run at production speed across three resin lots, with a GR&R under 10% of tolerance on neck OD, bore ID, and seal land flatness. Accept the lot only if CpK is at or above 1.33 on those three dimensions and 1.67 on neck OD and crimp ring height, and if the cavity-to-cavity spread is under 0.04 mm on neck OD. If the supplier cannot produce that study, the tolerance on the drawing is a wish, not a capability, and the first crimper reject you see will be the one you did not predict. For the wider consumables picture, HPLC consumables sourcing follows the same logic, and the chromatography supplier you choose should be able to hand you the cavity data without a fight. If they cannot, the drawing is doing work the process should be doing, and that is the wrong place for it.

References & further reading

Related resources from our lab consumables network:

• 2Ml Vial at 2mlsamplevial.com
• Hplc Consumables at hplcconsumable.com
• Chromatography Supplier at chromatographysupplier.com
• Oem Vial at oemsamplevials.com

External references and standards:

• Agilent chromatography consumables and columns
• Thermo Fisher chromatography and sample handling
• U.S. EPA SW-846 test methods for waste and materials
• Restek technical resources for chromatography