A 200-piece lot of 250 µL glass inserts failed incoming inspection at a contract lab because the outside diameter at the shoulder measured 5.94 mm on 14 pieces against a 6.00 ± 0.05 mm print. The inserts still dropped into the 2 mL vial. They also sat 0.6 mm low, and the autosampler needle on a fixed-height Agilent 7693A tower punched the shoulder instead of the well on every tenth or eleventh injection. That is what vial insert dimensional tolerance capability actually means in production: not whether the part fits, but whether the process that made it can hold the dimension that decides whether the needle reaches the sample.
A print tolerance tells you what the designer will accept. Capability tells you what the mold, the glass former, or the machining center will deliver shift after shift. Those are different numbers, and confusing them is the most common reason a supplier ships conforming parts that fail in the field.
Capability is measured with Cpk, and the formula is not negotiable: Cpk = min[(USL − mean), (mean − LSL)] / (3σ). For a 6.00 ± 0.05 mm OD with a process mean of 6.01 mm and a standard deviation of 0.008 mm, the Cpk is (6.05 − 6.01) / (3 × 0.008) = 1.67. That is a capable process. Move the mean to 6.03 mm with the same σ and Cpk drops to 0.83. The parts still pass the print. The needle still misses.
I run capability studies on 30 consecutive pieces minimum, ideally 50, sampled across a full shift including a tool change if there is one. Thirty pieces gives you a σ estimate with roughly ±13% uncertainty at 95% confidence. Fifty tightens it to about ±10%. Anything under 30 and you are measuring the sample, not the process. I have seen suppliers quote Cpk from a 10-piece “capability study” run on a single cavity of an eight-cavity tool. That is not a capability study. That is a sales document.
For vial inserts specifically, the practical floor is Cpk ≥ 1.33 on every functional dimension, and Cpk ≥ 1.67 on the two dimensions that drive needle geometry: the well depth and the shoulder OD. Anything below 1.33 on those two and you are buying a lottery ticket with each lot.
What each manufacturing route can actually hold
Vial inserts are made three ways, and the achievable tolerance differs by an order of magnitude between them. Knowing which process made your insert tells you what tolerance to demand before you ever see a certificate.
| Process | Typical OD tolerance | Typical well depth tolerance | Typical wall tolerance | Realistic Cpk ceiling |
|---|---|---|---|---|
| Injection-molded PP | ± 0.05 mm | ± 0.10 mm | ± 0.04 mm | 1.5 to 1.8 |
| Injection-molded COC/COP | ± 0.03 mm | ± 0.06 mm | ± 0.025 mm | 1.8 to 2.2 |
| Glass tube forming (converted) | ± 0.15 mm | ± 0.25 mm | ± 0.08 mm | 0.9 to 1.3 |
| Machined from rod (PTFE, PEEK) | ± 0.02 mm | ± 0.03 mm | ± 0.02 mm | 2.0 to 2.5 |
Read that table carefully. A converted glass insert cannot hold ±0.05 mm on OD. The forming process runs hot, the tube relaxes on cooling, and the mandrel wears. If your print calls for ±0.05 mm on a glass insert, you have written a specification that no supplier can meet, and every supplier will either decline the job or ship marginal parts and hope. The right move is to write the tolerance the process can hold, then control the dimension that actually matters with a gauge.
Injection-molded polypropylene is the workhorse for 150 µL to 300 µL inserts. A well-maintained eight-cavity tool in a 90-ton press holds ±0.05 mm on OD comfortably. The moment you add a hot runner, a slide, or a thin-wall section under 0.4 mm, capability drops. I have seen a 250 µL insert with a 0.35 mm wall lose 0.4 of a Cpk point on wall thickness because the cavity fill was unbalanced. The fix was a gate change, not a tighter print.
COC and COP cost more and hold tighter tolerances, and they are the right call when the insert has to survive −80 °C storage or when low extractables matter. Machined PTFE is for special cases only, and the surface finish is the giveaway: a machined insert has visible tool marks and a slightly domed floor unless the toolpath was dressed. For a general survey of the consumable side, the HPLC consumables category is worth a look, but the tolerance story is a molding story.
The four dimensions that are genuinely hard to control
Not every dimension on a vial insert print is difficult. Most are easy. Four are not, and those four are where capability studies fail.
Well depth at the conical floor
The depth from the top rim to the bottom of the well is the dimension that decides how much sample the needle can reach. In a molded insert, the floor is formed by the core pin, and the core pin tip wears. A 0.1 mm wear on the pin tip is a 0.1 mm shallower well, and the needle now bottoms out 0.1 mm higher. On a 250 µL insert with a nominal 32.5 mm height and a 29.0 mm well depth, a 0.10 mm drift is a 0.34% loss of well volume, which sounds trivial until the sample is 50 µL and the needle is set to aspirate 40 µL from a 1.0 mm above-bottom position. That 0.1 mm is now 2.5% of the aspiration depth.
The control is a pin-height check every 5,000 shots and a well-depth gauge on the first-off and last-off of every shift. I use a depth gauge with a 1.5 mm ball tip, not a flat anvil, because the conical floor will read differently against a flat anvil depending on where the tip lands.
Shoulder outside diameter
The shoulder is the step where the insert’s OD transitions from the body to the neck. On a 6.00 mm body with a 4.0 mm neck, the shoulder is a 1.0 mm radial step. That step is where the sampler’s gripper or the vial seat contacts the insert. A 0.05 mm oversize on the shoulder OD and the insert will not seat flush in a 2 mL vial; a 0.05 mm undersize and it rattles. Either way, the vertical position of the well relative to the vial top changes, and the needle height reference is gone.
This is the dimension that failed in the opening case. The mold had a worn shoulder insert, and the drift was 0.06 mm over 40,000 shots. The supplier’s print tolerance was ±0.05 mm, so 14 pieces out of 200 were technically out of spec, but the supplier’s own inspection plan sampled 5 pieces per lot and missed it. The fix was a shoulder insert replacement and a switch to 100% go/no-go gauging on the shoulder OD for this customer.
Concentricity of the well to the OD
Concentricity is the quiet killer. A well that is 0.08 mm off-center relative to the OD means the needle, which is positioned relative to the vial, is now 0.08 mm off the well center. On a 4.0 mm well ID with a 0.6 mm needle, you have 1.7 mm of radial clearance, so 0.08 mm is survivable. On a 2.0 mm well ID with a 0.5 mm needle, you have 0.75 mm of clearance, and 0.08 mm eats 11% of it. Add a slightly bent needle and a vial that is 0.1 mm off-center in the tray and you are at the wall.
Concentricity is hard to control because it depends on the alignment of the core pin to the cavity, which is set by the tool’s guide pins and the press’s platen parallelism. A tool that has been run for 500,000 shots will drift on concentricity before it drifts on OD. Measure it with a pin gauge and a dial indicator, or optically on a vision system. Demand Cpk ≥ 1.33 on concentricity for any insert with a well ID under 3.0 mm.
Wall thickness uniformity
Wall thickness drives two things: the insert’s crush resistance and its thermal mass. A 250 µL insert with a nominal 0.45 mm wall that varies from 0.38 to 0.52 mm around the circumference will have a cold spot and a hot spot, and in a heated sampler tray that means a temperature gradient across the sample. For most assays this does not matter. For a method that runs at 4 °C in the autosampler, it does. The variation comes from core-pin deflection under injection pressure, and the fix is a stiffer core or a lower fill pressure, both of which are tooling decisions.
How to run a capability study that survives an audit
The study has to be documented well enough that a customer’s quality auditor can follow it. Here is the sequence I use, and it takes about four hours per part number.
- Define the CTQ dimensions first. For a vial insert, that is well depth, shoulder OD, body OD, concentricity, and wall thickness. Everything else is a reference dimension.
- Sample 50 consecutive pieces from a single cavity, then repeat for each cavity. Do not pool cavities until you have looked at each one. A pooled Cpk of 1.5 can hide one cavity at 0.7.
- Measure with a gauge that matches the print. Depth with a ball-tip gauge, OD with a micrometer at three axial positions, concentricity on a vision system. Record the gauge and its calibration date.
- Calculate mean, σ, and Cpk per cavity. Report the worst cavity, not the average. The worst cavity is what ships.
- Run the study across a tool change if the tool has more than one cavity. A tool that holds Cpk 1.8 on cavity 1 and 0.9 on cavity 6 is a tool that needs a gate balance, not a new print.
- Re-run every 6 months or 100,000 shots, whichever comes first. Tool wear is a slow drift, and a study from two years ago describes a tool that no longer exists.
If a supplier cannot produce a per-cavity capability report with the raw measurements attached, treat the certificate as a marketing document. I have rejected three suppliers on this basis in the last five years, and two of them came back with real data within a month once they understood the ask.
Field case: the 0.6 mm seating error that looked like a needle problem
A pharmaceutical QC lab running a 150 µL insert in a 2 mL vial reported intermittent “no sample” errors on a Thermo Fisher TriPlus RSH. The errors clustered on one tray position. The lab’s first move was to replace the needle, then the syringe, then the tray. None of it helped.
The measurement that identified it was a simple one: we took 20 inserts from the suspect lot and measured the height from the vial top to the insert rim with a depth gauge, with the insert seated in a clean 2 mL vial. The nominal is 1.2 mm. The suspect lot read 1.8 mm on 9 of 20 pieces. The insert was sitting 0.6 mm high because the shoulder OD was 0.12 mm oversize and the insert was hanging on the vial’s shoulder instead of dropping to the seat. The needle was programmed to aspirate 1.0 mm above the well floor, and the 0.6 mm error put the needle tip above the sample.
The fix was a lot rejection and a print change: shoulder OD tightened from ±0.08 mm to ±0.04 mm, with 100% go/no-go gauging at the supplier. The lab also added a seated-height check to incoming inspection, which is a 30-second test per lot and would have caught the problem on day one. The 2 mL vial seat geometry is the reference, not the insert alone, and that is the lesson.
Field case: the glass insert that could not hold ±0.05 mm
A contract lab specified a converted glass insert at 5.00 ± 0.05 mm OD to fit a custom tray. Two suppliers quoted, one accepted, and the first three lots ran at 5.02 to 5.09 mm. The fourth lot ran at 5.11 mm on 30% of pieces. The supplier’s explanation was that the forming mandrel had worn, and that ±0.05 mm was “at the edge” of what the process could hold.
They were right. We ran a capability study on 50 pieces from the fourth lot: mean 5.09 mm, σ 0.021 mm, Cpk = (5.05 − 5.09) / (3 × 0.021) = −0.63. A negative Cpk means the process mean is outside the specification. The process was not marginal; it was incapable by design.
The fix was a print change to 5.00 ± 0.12 mm with a 100% OD gauge at the supplier, plus a tray modification to accept the wider tolerance band. The lab spent about $4,000 on the tray change and avoided a $40,000 re-qualification of the autosampler method. The lesson is that a tolerance has to be matched to the process, and the process has to be named on the print. A glass insert and a molded insert are not interchangeable specifications.
A decision checklist for incoming lots
Use this on the first lot from any new supplier, and on every lot where the supplier has changed tooling, resin, or cavity count. It takes about 45 minutes and it catches most of what a certificate misses.
- Verify the capability report is per-cavity and covers at least 30 pieces per cavity, with raw data attached.
- Measure well depth on 10 pieces with a ball-tip gauge. Compare to the print and to the supplier’s reported mean. A shift of more than 0.05 mm from the reported mean is a red flag.
- Measure shoulder OD on 10 pieces with a micrometer at three axial positions. Calculate the range within a single piece; more than 0.03 mm means the part is out of round.
- Check seated height in a clean 2 mL vial on 10 pieces. Nominal is 1.2 mm for most 150 to 300 µL inserts. More than 0.2 mm above nominal means the insert is hanging, not seating.
- Check concentricity on 5 pieces with a pin gauge and dial indicator, or optically. More than 0.08 mm on a well under 3.0 mm ID is a fail.
- Confirm the resin or glass grade against the certificate of analysis. A switch from virgin PP to regrind will show up as a color shift and a 0.02 to 0.04 mm increase in σ.
If a lot fails any of these, quarantine it and request a replacement capability study before accepting the next shipment. Do not accept a “concession” on a dimension that drives needle geometry. The concession is a promise that the next lot will be better, and it usually is not. For broader reference material on method and container requirements, the U.S. EPA SW-846 methods and the Restek technical library are useful for the analytical side, but the tolerance question is answered by the tool, not the method.
The decision rule I use
Demand Cpk ≥ 1.33 on every functional dimension and Cpk ≥ 1.67 on well depth and shoulder OD, verified per cavity on at least 30 pieces, and match the print tolerance to the process: ±0.05 mm for molded PP, ±0.03 mm for COC/COP, ±0.15 mm for converted glass, ±0.02 mm for machined. If a supplier cannot show per-cavity data, or if the print tolerance is tighter than the process can hold, the part is not capable regardless of what the certificate says. Buy from the supplier who shows you the raw numbers, and gauge the shoulder OD and seated height on every incoming lot. The vial insert that passes those two checks is the one that will not cost you a re-run at 2 a.m.
References & further reading
Related resources from our lab consumables network:
- Vial Insert at hplcinsert.com
- 2Ml Vial at 2mlsamplevial.com
- Hplc Consumables at hplcconsumable.com
- Chromatography Supplier at chromatographysupplier.com
External references and standards: