A 2 mL clear glass vial that mics at 11.98 mm on the outside diameter can still jam a 12-position autosampler tray if the same lot measures 12.06 mm at the shoulder, and that is the call I got on a Tuesday morning after 340 injections had run clean. The instrument was fine. The tray was fine. The chromatography vial lot had drifted 0.04 mm on a dimension nobody was charting. That is the whole argument for treating chromatography vial dimensional tolerance capability as a process property, not a certificate value. A number on a CoA tells you what one sample measured. Capability tells you what the next 50,000 pieces will do, and those are not the same thing.
Three processes make the clear 12 x 32 mm vial you load into an autosampler: tubular glass converting (the dominant route), injection molding of resin vials, and injection blow molding for the small share of plastic formats. Each has a different tolerance floor, and the floor is set by physics, not by how carefully the supplier inspects.
Tubular converting starts from a purchased glass tube, usually borosilicate 33 expansion or 51 expansion, with a wall that already varies. The tube is cut, one end is fire-formed into a shoulder and neck, and the finish is calibrated. The bottom is closed in a separate station. Because you are reforming an existing tube rather than filling a mold cavity, the incoming tube tolerance propagates straight through. A tube that arrives at ±0.15 mm on OD cannot produce a finished vial better than about ±0.12 mm on body OD, and in practice you land closer to ±0.10 mm after the forming heat relaxes some of the variation. Neck ID and the crimp finish are where converting earns its keep: a well-run line holds neck ID to ±0.10 mm and the crimp bead to ±0.08 mm, which is why glass still owns the GC crimp market.
Injection molding of resin vials (polypropylene, sometimes COC) gives you a mold cavity that repeats. Cavity-to-cavity variation is the enemy here, not piece-to-piece within a cavity. A 16-cavity tool can hold ±0.05 mm on body OD within a single cavity and still spread ±0.12 mm across the tool if the hot runner is not balanced. That cross-cavity spread is the number that bites you, because the tray sees all 16 cavities mixed in one bag.
Injection blow molding sits in between and is mostly used for larger or specialty formats. It holds ±0.15 to ±0.25 mm on body dimensions and is not competitive for autosampler work where the tray pocket is 12.10 mm.
Working tolerance by process, body OD and neck ID
| Process | Body OD capability (Cpk ≥ 1.33) | Neck ID capability | Bottom thickness | Where it fails first |
|---|---|---|---|---|
| Tubular glass converting | ±0.10 mm | ±0.10 mm | ±0.15 mm | Incoming tube OD drift |
| Injection molded PP | ±0.05 mm within cavity, ±0.12 mm across tool | ±0.04 mm | ±0.05 mm | Hot runner imbalance |
| Injection blow molded | ±0.20 mm | ±0.15 mm | ±0.20 mm | Parison temperature |
Read that table as a floor, not a promise. Any supplier can hit it on a good day. The question is whether they hold it on a Tuesday in August when the plant ambient is 31 °C and the annealing lehr is running 8 °C hot.
Capability is a distribution, and most vial CoAs are a single point
When I audit a vial supplier I ask for two things before I look at a single certificate: the control chart for body OD over the last 90 days, and the sample size behind each point. A CoA that says "12.00 mm ± 0.10" with no distribution is a claim. A control chart with 30 subgroups of n=5 tells me whether the process is centered, whether it is drifting, and whether the spread is stable. Those are three different failure modes and they need three different fixes.
Capability is measured as Cpk against the print tolerance, and the formula is simple enough to do on the back of a run sheet: Cpk = min[(USL − mean), (mean − LSL)] / (3σ). For a 12.00 mm nominal body OD with a ±0.10 mm print, USL is 12.10 and LSL is 11.90. If the process mean sits at 12.01 and σ is 0.025 mm, Cpk is (12.10 − 12.01)/(3 × 0.025) = 1.20. That is a marginal process. It will pass incoming inspection most days and still produce a tray jam every few lots. If the same process runs at σ = 0.018 mm, Cpk climbs to 1.67 and the jam rate falls off a cliff.
The number I actually want from a supplier is Cpk ≥ 1.33 on body OD, neck ID, and crimp bead, and Ppk ≥ 1.33 on the same dimensions from a 30-lot rolling window. Cpk is short-term capability (within-subgroup). Ppk is long-term performance (total variation including lot-to-lot drift). A supplier who quotes Cpk 2.0 and Ppk 1.1 is telling you the process is capable when it is in control and loses control between lots. That is exactly the supplier who shipped me the 12.06 mm shoulder.
Why Cpk on a single dimension is not enough
Two vials can both pass body OD and still fit differently in the tray, because the tray contacts the vial at the shoulder, not the body. If body OD is in spec and shoulder diameter runs 0.05 mm over, the vial sits high and the needle strikes the septum off-center. That is why I ask for capability on a feature stack, not a list of independent dimensions. The stack that matters for autosampler fit is: shoulder diameter, body OD, and the vertical distance from the bottom to the shoulder datum. Get all three in capability and the tray stops being a variable.
The dimensions that refuse to stay put
Some dimensions are easy. Body OD on a converted glass vial is a turning operation and turning is boringly repeatable. Others fight you every shift. These are the four I have watched drift on otherwise good lines.
Neck ID and the crimp finish
Neck ID sets whether the septum seats flat and whether the crimp cap grabs. On a 11 mm crimp finish the neck ID target is typically 6.0 mm ± 0.10. The failure mode is not a wide scatter, it is a slow taper: the forming tool wears, neck ID creeps up 0.02 mm per 100,000 pieces, and after 400,000 pieces you are at 6.08 and the septum no longer compresses enough to seal. The symptom is not a leak you can see. It is evaporation loss showing up as a 3 to 5 % area drop on your early-eluting peaks after 18 hours in the autosampler. The fix is tool change at a defined piece count, not at failure. I set that count from the wear rate, not from the print.
Bottom thickness and the taper
Bottom thickness on a converted vial runs 0.9 to 1.1 mm nominal. The hard part is the taper: the bottom is not flat, it has a slight concave or convex profile that the fire-forming station controls. A convex bottom (crowned) means the vial rocks in the tray and the needle depth varies by 0.2 mm run to run. That shows up as injection-to-injection RSD on peak area climbing from 0.8 % to 2.5 % with no change to the method. I have chased that one for a week before mic'ing the bottom.
Concentricity of neck to body
This is the dimension nobody charts and everybody should. Neck-to-body concentricity on a converted vial is typically held to 0.10 mm TIR. When it drifts to 0.20 mm, the vial still fits the tray, still takes a cap, and still injects, but the needle enters the septum at an angle. Over 500 injections that angle shaves the septum and you get coring, and coring shows up as a phantom peak at the retention time of your analyte's neighbor. The measurement that identifies it is a simple dial indicator runout check on 30 vials from the lot, not a chromatogram.
Wall thickness variation around the circumference
Glass tube is never perfectly round. Wall thickness on a 1.0 mm nominal wall can vary ±0.08 mm around the circumference of a single vial. That variation is invisible until you heat the vial, at which point the thin side expands faster and the vial can crack at the shoulder during a 40 °C autosampler run. It also changes the optical path if you are doing any photometric work. The control is incoming tube inspection at the converter, and the only way to know your supplier does it is to ask for the tube CoA, not just the vial CoA.
Two lots that passed the print and failed the bench
Case one: the shoulder that jammed a tray. A contract lab running a 12-position autosampler on a Thermo Vanquish called me after 340 clean injections and then a hard fault. The tray error flagged position 7. I pulled 20 vials from the lot and mic'd them: body OD was 12.00 to 12.02 mm, dead center. Shoulder diameter measured 12.06 to 12.09 mm against a 12.10 mm tray pocket. The print only specified body OD and overall height. The supplier had never been asked to control shoulder diameter, so they never did. The fix was a print revision adding shoulder diameter at 12.00 ± 0.08 mm and a capability requirement of Cpk ≥ 1.33 on that feature. The next lot ran 40,000 injections with no tray fault. The lesson is that the print defines what the supplier controls, and a print that ignores a contacting feature is a print that guarantees a future jam. This is the same class of problem the HPLC vial world sees on the 9 mm screw-thread side, where thread pitch tolerance decides whether the cap bottoms out.
Case two: the crimp that leaked at 18 hours. A GC lab running EPA method work under SW-846 saw a 4 % loss on a volatile analyte between the 0-hour and 18-hour standards. They blamed the septa and switched brands twice. I asked for the neck ID data on the vial lot. It ran 6.06 to 6.09 mm against a 6.00 ± 0.10 print, so it was technically in spec but hugging the upper limit. The crimping tool was set for a 6.00 mm neck and could not compress the cap enough on a 6.09 mm neck. The measurement that identified it was a 30-vial neck ID check with a pin gauge, not a leak test. The fix was two-part: tighten the neck ID print to 6.00 ± 0.06 mm and re-set the crimper to 1.2 N·m with a 6.05 mm setting gauge. Loss dropped to 0.6 %. The HPLC septa were never the problem, which is usually the case.
What to put on the print, and what to demand in the capability report
The print is the contract. If a dimension is not on it, the supplier has no obligation to control it, and a capable process that is not asked to be capable will drift. Here is the checklist I use when I write or audit a vial print for autosampler work.
- Body OD: 12.00 ± 0.10 mm, Cpk ≥ 1.33, Ppk ≥ 1.33 over 30 lots.
- Shoulder diameter: 12.00 ± 0.08 mm, Cpk ≥ 1.33. This is the tray contact feature and it is the one most prints omit.
- Neck ID (11 mm crimp): 6.00 ± 0.06 mm, Cpk ≥ 1.33. Tighter than the generic 6.00 ± 0.10 because the crimper needs margin.
- Neck-to-body concentricity: 0.10 mm TIR max, 100 % gauged or sampled at n=30 per lot.
- Bottom thickness: 1.00 ± 0.10 mm, with a flatness callout of 0.05 mm to kill the rock.
- Overall height: 32.00 ± 0.20 mm. Generous, because height is rarely the failure mode and over-tightening it drives cost with no benefit.
- Glass expansion coefficient: 33 expansion borosilicate, USP Type I, with the tube CoA attached to the lot.
- Capability report: control charts for the four critical dimensions, 90-day window, sample size stated per point. A CoA without a chart is a claim.
On the plastic side the same logic applies but the numbers shift. Injection molded PP vials hold tighter body OD but the tool has cavities, so the capability report must be per cavity, not pooled. A pooled Cpk of 1.5 on a 16-cavity tool can hide one cavity running at Cpk 0.8. Ask for the cavity breakdown. If they cannot produce it, they are not running SPC, they are running hope.
Incoming inspection that actually catches drift
You cannot inspect quality into a vial, but you can catch a drifting lot before it reaches the autosampler. My incoming protocol for a new lot is 30 vials, three measurements each: body OD with a digital micrometer at 0.001 mm resolution, neck ID with a pin gauge set, and concentricity with a dial indicator on a V-block. That is 15 minutes of work and it catches the shoulder and neck problems that a visual check misses. The chromatography products supply chain is deep enough that a bad lot is usually a bad supplier, not a bad industry, and 15 minutes at goods-in tells you which one you have.
The decision rule
Buy on capability, not on certificate. A vial supplier who cannot show you a 90-day control chart with Cpk and Ppk on body OD, shoulder diameter, neck ID, and concentricity is selling you a print, not a process, and the print will be met on the day they sample and missed on the day you run. Set the four critical dimensions on the print, require Cpk ≥ 1.33 and Ppk ≥ 1.33 over 30 lots, demand the per-cavity breakdown on molded parts, and do the 30-vial goods-in check yourself. The instrument vendors publish their tray and needle geometry for a reason; the Aijiren consumables documentation and the Thermo Fisher sample handling specs both give you the pocket and septum dimensions to design the print against, and the Restek technical library has enough method detail to tell you which dimensions your specific application will punish. If a supplier's capability data does not survive that comparison, the lot will not survive the tray.
References & further reading
Related resources from our lab consumables network:
- Chromatography Vial at chromvials.com
- Hplc Vial at hplcvialfactory.com
- Chromatography Products at chromatographyforsale.com
- Hplc Septa at hplcsepta.com
External references and standards:
