A 24-cavity mold ran a 500,000-piece lot of 9 mm screw-thread vials with the neck ID drifting from 6.42 mm to 6.58 mm across the run, and the customer’s septa push-in force dropped from 18 N to 9 N by the last third of the lot. Nobody caught it on the shop floor because every individual part measured inside the drawing. The problem was capability, not conformance. That gap between “in spec” and “in control” is what cod vial dimensional tolerance capability actually means, and it is the number that decides whether your autosampler runs clean for a year or starts coring septa in month four.
I have spent enough years on the molding floor and enough on the QC bench to know that most suppliers quote you a tolerance and almost none quote you a Cpk. Those are different promises. A tolerance is a wall. Capability is how far the process stands from the wall and how much it wanders. This article is about the second number, which dimensions on a cod vial are hard to hold, and which molding process can actually hold what.
When a datasheet says the neck ID is 6.50 mm ± 0.10 mm, that is a conformance statement. It tells you a part either passes or fails a gauge. It says nothing about the distribution of parts inside that band. A process running at the center with a tight spread and a process running against one wall with a wide spread can both ship 100 percent “in tolerance” and behave completely differently in your lab.
Capability is measured with Cpk, and for a two-sided tolerance it is the smaller of (USL minus mean) and (mean minus LSL), divided by three sigma. For a one-sided callout like a minimum wall thickness, you use Cpk against the single limit. The convention I hold suppliers to: Cpk ≥ 1.33 for a mature process, ≥ 1.67 for a critical dimension like the septum seat or the neck ID that governs seal force, and anything below 1.0 means the process is generating defects you have not found yet. A Cpk of 1.0 means roughly 2,700 parts per million fall outside the limit. On a 500,000-piece lot that is over 1,300 bad vials you are shipping and hoping the customer’s incoming inspection catches.
I ask for a Ppk number too, not just Cpk. Cpk uses within-subgroup variation, Ppk uses overall variation including shifts between shifts and between mold cavities. If a supplier quotes Cpk 1.8 and Ppk 1.2, the process is capable on a good day and drifting on a bad one. That spread between the two numbers is the tell.
Reading a capability study on a multi-cavity tool
A 24-cavity mold is 24 processes wearing one coat. Cavity-to-cavity variation is often larger than run-to-run variation, and a supplier who pools all cavities into one Cpk hides it. I require a per-cavity breakdown. If cavity 7 runs 0.04 mm high on neck ID while the rest sit at center, a pooled Cpk of 1.5 masks that cavity 7 alone is at 0.9. That single cavity will produce a stream of loose-seal vials forever until someone reworks the tool.
- Minimum 30 consecutive parts per cavity, ideally 50, sampled across a full shift.
- Report mean, standard deviation, and Cpk per cavity, not pooled.
- Flag any cavity whose mean sits more than 0.5 sigma off the tool average.
- Re-run the study after any tool maintenance. A polished cavity changes its fill.
Which dimensions on a cod vial are genuinely hard to control
Not all features are equal. Some fall out of the mold at Cpk 2.0 without effort. Others fight you every cycle. Here is where the difficulty actually lives, ranked by how often I have seen them go out of control.
Neck ID and septum seat depth
The neck ID sets the interference fit on the septum and therefore the seal. It is hard because it is a small diameter on a tall, thin feature, and the core pin that forms it is subject to thermal growth and wear. A 0.05 mm drift in neck ID moves septa retention force noticeably. This is the dimension I watch hardest, and the one most tied to whether your cod vial seals on the autosampler.
Wall thickness and concentricity
Wall thickness on the body governs both crush resistance and how evenly the part cools. Concentricity (TIR, total indicator runout) between the neck bore and the body OD is what determines whether the vial sits straight in a 12-position tray. A TIR over 0.15 mm and the vial wobbles enough to mis-seat in a fixed tray. This is a mold-alignment and core-centering problem, and it is unforgiving.
Thread form and start position
For screw-thread vials the thread pitch and the angular start position of the thread are hard to hold because they depend on the core unscrewing cleanly every cycle. A thread that starts a few degrees off means the cap cross-threads on the first turn. GPI 13-425 and GPI 18-400 thread standards define the form; holding the start position within a few degrees across a million cycles is the hard part.
Height and bottom flatness
Overall height is comparatively easy. Bottom flatness matters for autosampler grippers and for the vial standing stable. A concave bottom of more than 0.10 mm makes the vial rock. This one is usually a cooling and ejection issue, not a mold issue.
What tolerance each molding process can actually hold
The process choice sets a ceiling on capability before you ever talk about the tool. Here is the honest range I have seen across production, not the marketing number. These are for the critical neck and wall features on a vial in the 8 to 12 mm OD class, in virgin PP or PET.
| Process | Typical neck ID tolerance held | Realistic Cpk on neck ID | Wall thickness tolerance | Notes |
|---|---|---|---|---|
| Injection blow molding (IBM) | ± 0.08 mm | 1.1 to 1.4 | ± 0.06 mm | Neck is injection-formed, body is blown. Good neck control, weaker wall control. |
| Extrusion blow molding (EBM) | ± 0.15 mm | 0.8 to 1.1 | ± 0.10 mm | Cheapest, worst dimensional control. Fine for non-critical closures only. |
| Injection molding (single stage) | ± 0.05 mm | 1.5 to 2.0 | ± 0.04 mm | Best dimensional control. Higher tool cost, slower cycle. |
| Injection stretch blow (ISBM, 2-stage) | ± 0.06 mm on neck | 1.3 to 1.7 | ± 0.05 mm | Neck formed by injection, body stretched. Strong all-round choice for vials. |
The pattern is clear. If the neck is formed by injection (IBM, ISBM, straight injection), you get a tight neck. If the neck is formed by blowing (EBM), you do not. For any vial where the septum seal matters, EBM is the wrong process, full stop. The dimensional stack that has to fit at the autosampler does not forgive a loose neck.
I will add one practical note. The tolerance a process “can hold” assumes a tool built to support it and a machine running a stable cycle. A great process on a tired tool holds nothing. The HPLC consumables world is full of parts that were specified well and molded badly.
Where the process actually loses control: three mechanisms
Tolerances drift for reasons, and they cluster. Here are the three I see most on vial tools, and how each shows up in the data.
Thermal drift across the run
Mold temperature zones wander over a long run. As the core pin heats, the neck ID shrinks. A run that starts at 6.48 mm can end at 6.40 mm without a single alarm, because each part is still inside ± 0.10. The capability study catches this as a downward trend line, not a scatter. Fix is tighter mold temperature control, usually ± 1 °C on the neck core, plus a mid-run first-article check.
Cavity-to-cavity imbalance
Fill imbalance across cavities shows as one or two cavities running off-center on wall thickness. It comes from gate wear or a hot-runner tip that has drifted. The tell is a bimodal histogram when you plot all cavities together. Fix is a gate inspection and re-balance, not a process tweak.
Material lot variation
Shrinkage depends on the resin. A lot of PP with a slightly different melt flow index shrinks differently and the whole distribution moves. Suppliers who do not quarantine and re-qualify resin lots will hand you a capability study that was true for last month’s material. I ask for the resin lot on the CoA and match it to the study date.
Two field cases where capability, not conformance, was the failure
Case one: the septa that stopped sealing. A contract lab reported that septa were backing out of vials during a 40-vial autosampler run. Push-in force on a sample of 50 vials averaged 11 N against a spec minimum of 15 N, but every vial passed the incoming neck-ID gauge at ± 0.10 mm. The measurement that identified it: a per-cavity neck-ID study showing cavity 14 running at 6.58 mm, near the upper limit, with a Cpk of 0.7. The tool had a worn core pin on that cavity. The fix was a core pin replacement and a re-qualification study. Push-in force came back to 19 N and stayed there. The lesson: the gauge passed the part, the capability study failed the process.
Case two: the vials that wobbled in the tray. A customer running a fixed 12-position tray reported intermittent gripper faults, roughly 1 in 200 vials. Neck ID was fine. The measurement that found it: TIR between neck bore and body OD on 200 vials, which showed a subset at 0.22 mm runout against a 0.15 mm limit. Root cause was a mold alignment issue on one half of the tool after a maintenance cycle. The fix was a mold re-alignment and a new TIR capability study at Cpk 1.5. Gripper faults dropped to zero.
Both cases share a shape. The part was in tolerance. The process was not capable. Only a capability study separated the two.
A capability acceptance checklist for a cod vial supplier
This is what I put in front of a new supplier before I release a PO. It is short on purpose. A supplier who cannot answer these is not ready to quote a critical vial.
- Per-cavity capability study on neck ID, septum seat depth, and TIR, minimum 30 parts per cavity, Cpk ≥ 1.33 with critical dimensions at ≥ 1.67.
- Both Cpk and Ppk reported, with the gap between them under 0.4. A wider gap means the process shifts between shifts or lots.
- Process declaration naming the molding method (IBM, ISBM, injection) and the tolerance the tool was built to hold.
- Resin lot traceability tying the study to a specific material lot, plus a re-qualification trigger on any resin change.
- Mold maintenance log showing core pin and gate inspection intervals, and a re-study after every maintenance event.
- Gauge R&R on the neck-ID and TIR gauges under 10 percent, because a capability number from a shaky gauge is fiction.
If a supplier pushes back on the per-cavity requirement, that is the answer. Pooled capability is a way to hide a bad cavity. The 2 mL vial and insert assemblies that feed the same autosampler have the same problem, and the same fix.
How the tool and the process have to agree
Capability is a joint property of tool and machine. A tool built to ± 0.05 mm on a machine running a sloppy cycle holds ± 0.12. The reverse is also true. When you specify a vial, specify both.
Concretely, for a critical cod vial I want a tool built to hold the drawing at 60 percent of the tolerance band, so the process has room to drift and still make Cpk. I want the machine cycle locked with documented parameters, and I want the mold temperature on the neck core controlled to ± 1 °C. I want a first-article and a mid-run check, not just end-of-run. And I want the supplier to tell me which dimensions they consider hard to hold, because a supplier who claims every dimension is easy is telling you they have not measured carefully.
The dimensional relationship between the vial, the septum, and the insert is where this all cashes out. A vial insert that does not seat because the neck ID drifted is a capability failure upstream, not an insert problem. The chromatography consumables supply chain rewards suppliers who can show the number. Cross-check what the instrument makers publish on consumable fit tolerances, for example the fit specifications in Agilent’s consumables documentation and the sample-handling guidance from Thermo Fisher, and you will see the same logic: the interface dimension is the one that matters, and it needs a capability number behind it.
For regulated work, the method governs what the vial has to do, and the capability study is how you prove the vial can keep doing it. The U.S. EPA SW-846 test methods and the technical notes from Restek both assume the vial is a controlled component, not a commodity. A supplier’s Cpk is the evidence that it is.
Troubleshooting a capability number that will not hold
When a supplier’s study comes back weak, work the problem in this order. It saves time.
- Check the gauge first. A Gauge R&R over 10 percent invalidates the study. Re-measure before blaming the mold.
- Split by cavity. If one or two cavities are off-center, it is a tool wear or gate issue, not a process issue.
- Plot against time. A trend line means thermal drift. A step change means a material lot or a parameter edit.
- Check the resin lot. Match the study date to the CoA. A lot change with no re-study is a red flag.
- Inspect the core pin. Neck ID drift almost always traces to core pin wear or thermal growth.
- Re-study after the fix. A fix without a new capability number is a guess.
The decision rule
Do not accept a cod vial on a tolerance statement alone. Require a per-cavity capability study with Cpk ≥ 1.33 on neck ID, septum seat depth, and TIR, both Cpk and Ppk reported with a gap under 0.4, a declared molding process, and a re-qualification trigger tied to resin lot and tool maintenance. If the supplier cannot produce that number, the tolerance on the datasheet is a wall with no floor under it, and your autosampler will find the gap before your incoming inspection does.
References & further reading
Related resources from our lab consumables network:
- Cod Vial at codvial.com
- Vial Insert at hplcinsert.com
- 2Ml Vial at 2mlsamplevial.com
- Hplc Consumables at hplcconsumable.com
External references and standards: