A 2 mL clear glass vial lot measured 11.98 to 12.09 mm across the crimp bead on our optical comparator, against a 12.00 ± 0.10 mm drawing. That is a 110 µm spread, and it is inside print. The run still failed on a Thermo autosampler at position 74, because the cap skirt sat 0.4 mm proud of the carrier plate and the gripper could not close. The vial met every dimension on the certificate. The stack did not. That is the gap between a dimensional tolerance and a dimensional tolerance capability, and it is the only thing worth arguing about when you buy vials for an HPLC autosampler vial program at production volume.
Most datasheets give you a nominal value and a plus/minus. Almost none give you a process capability index, a sample size, or the dimension that actually drives the failure. This article is about what each molding and forming process can hold, how to measure whether it can hold it repeatedly, and which dimensions on a 2 mL vial are genuinely hard to control. If you already know what the vial does inside the instrument, you can skip ahead. The question here is narrower: can your supplier hold the tolerance, lot after lot, at the rate they ship?
A tolerance is a promise about one part. A capability index is a promise about the population. When a molder quotes you Cpk 1.33 on neck OD, they are saying the process mean sits at least four standard deviations from each specification limit. That is a 64 ppm defect rate on that dimension if the process is centered and stable. Cpk 1.67 gets you to about 0.6 ppm. Cpk 1.00 means roughly 0.27% of parts fall outside print, which on a 5,000-piece lot is around 13 vials. You will not find 13 bad vials in a 30-piece AQL sample. You will find them when the autosampler tray jams at 3 a.m.
Here is the part that matters more than the number itself. Cpk is only valid if the process was in statistical control when it was calculated. A molder can hand you a beautiful Cpk 1.8 on a dimension that drifted 0.05 mm over the three days they sampled. The index hides the drift because the standard deviation is small. Ask for the control chart, not just the index. If they cannot produce an X-bar and R chart with at least 25 subgroups of 5, the Cpk is a snapshot, not a capability.
The three numbers a capability report should contain
- Ppk versus Cpk. Ppk uses the overall standard deviation including between-lot shifts. Cpk uses within-subgroup variation. A supplier reporting only Cpk on a multi-cavity mold is hiding cavity-to-cavity variation. Demand both. A gap larger than about 0.3 between them means the process is drifting between lots.
- Sample size and frequency. A capability study on 30 parts from one cavity is worthless. You want at least 125 pieces spanning all cavities, taken across at least three setups or shifts.
- The measurement system. Gauge R&R below 10% of tolerance. If the supplier measures neck OD with a caliper at 0.02 mm resolution against a 0.10 mm tolerance, the gauge eats 20% of the tolerance band before the process even starts. Optical comparators and air gauges are the right tools here, not digital calipers.
I have watched a supplier report Cpk 1.5 on a dimension where the gauge itself contributed 35% of the observed variation. The real process capability was closer to 0.9. The parts were fine on paper and marginal in the tray. If you are sourcing through a distributor rather than the molder, ask who owns the capability data. A distributor reselling under their own label often cannot get it. That is one reason I prefer working with a manufacturer that runs its own forming line, like the operations behind Aijiren, where the capability study and the mold live in the same building.
Glass forming versus plastic molding: which process holds which tolerance
Autosampler vials are glass, and the neck and body come off a forming line, not an injection molding machine. That distinction drives everything. Glass forming is a hot process with a gob of molten borosilicate at around 1,200 °C going into a blank mold, then a blow mold. Thermal shrinkage during cooling is not perfectly uniform. The result is that glass holds some dimensions tightly and others loosely, and the loose ones are usually the ones that matter for the autosampler.
Plastic components on the same vial, the cap and the septum, are a different story. Injection-molded polypropylene caps can hold ±0.05 mm on critical features all day at Cpk 1.67. Compression-molded silicone/PTFE septa hold thickness to about ±0.08 mm. So the tolerance stack on a capped vial is dominated by the glass, because the glass is the loosest link and the cap has to sit on it.
| Feature | Typical process | Realistic tolerance (Cpk ≥ 1.33) | Hard to control? |
|---|---|---|---|
| Body OD (12 mm nominal) | Glass blow mold | ±0.10 mm | Moderate |
| Neck OD (crimp finish) | Glass blow mold | ±0.12 mm | High |
| Overall height (32 mm) | Glass form + cut | ±0.25 mm | Moderate |
| Bottom thickness | Glass form | ±0.15 mm | High |
| Verticality / camber | Glass cooling | 0.10 mm max | Very high |
| Cap skirt OD | Injection mold | ±0.05 mm | Low |
| Septum thickness (1.0 mm) | Compression mold | ±0.08 mm | Low |
If a glass vial datasheet quotes ±0.05 mm on neck OD, either the tolerance is a sorting limit (they measure every part and reject), or the number is aspirational. Sorting is real and legitimate, but it costs money and it does not fix the process. A sorted lot has a capability of whatever the sort gate allows, and the next lot starts over. Ask which it is.
The four dimensions that actually fail on the autosampler
Not every dimension on the drawing is equally dangerous. In fifteen years of running these lines and chasing failures in customer labs, four dimensions account for most of the tray jams and needle misses I have seen. They are the ones to demand capability data on.
Neck OD and crimp bead diameter
The crimp bead is where the aluminum seal grabs. On a 11 mm crimp finish, the bead diameter typically runs 11.00 ± 0.10 mm. If it runs large, the seal will not seat and you get a loose cap. If it runs small, the crimper over-compresses and cracks the neck or deforms the septum. The window is narrow because the aluminum seal has its own tolerance and the two stack. I have seen a 0.15 mm shift in bead diameter turn a 90% seal-pass rate into 60% on the same crimper with no adjustment.
Verticality (camber)
This is the dimension nobody puts on the datasheet and everybody should. A vial that leans 0.15 mm over its 32 mm height will not sit flat in the tray well. The gripper picks it at an angle, the needle enters off-center, and after a few hundred injections you have a bent needle and a coring complaint that gets blamed on the septum. Camber comes from uneven cooling. It is the hardest glass dimension to control because it depends on the thermal profile of the lehr, not the mold.
Bottom thickness and concentricity
The autosampler needle has to reach within about 1 mm of the bottom to get the last of the sample. If the bottom is thick on one side, the needle either hits glass or leaves sample behind. A 0.2 mm bottom thickness variation across a lot is enough to change your residual volume from 50 µL to 120 µL. That is a real analytical problem when your method is validated to a 100 µL injection.
Neck ID and septum seating depth
The neck ID controls how the septum seats and how much of it the needle sees. Too tight and the septum buckles during crimping. Too loose and it rocks. The seat depth variation shows up as a change in the effective puncture zone, which is why two lots of the same vial can give you different coring behavior on the same needle.
If you are buying vials against a method that specifies a particular HPLC autosampler vial, these are the four dimensions to put in your incoming specification. The rest can follow the drawing.
How to run a capability audit on a vial supplier in one afternoon
You do not need a metrology lab to find out whether a supplier can hold tolerance. You need a comparator, a few hundred parts, and a plan. Here is the checklist I use when qualifying a new source.
- Pull 125 pieces across at least five cavities and three production days. If they cannot tell you which cavity made which vial, stop. Cavity traceability is the baseline.
- Measure neck OD, bead diameter, camber, and bottom thickness on an optical comparator or a vision system, not a caliper. Record every value.
- Calculate Cpk and Ppk for each dimension against the drawing tolerance. Reject anything below Cpk 1.33 on the four critical dimensions.
- Check the control chart for the same period. Look for trends, shifts, and any point outside the control limits. A process with a trend is not capable, it is lucky.
- Run a gauge R&R on their measurement method. If it exceeds 10% of tolerance, their numbers are noise.
- Do a fit test. Cap 100 vials with your crimper at a set torque, then measure seal pull-off force. You want a tight distribution, not just a passing mean.
That is a half-day of work at the supplier and it tells you more than any certificate of analysis. A CoA reports the values on the parts they sampled. A capability audit reports what the process will do on the parts they ship next month.
Field case: the camber that passed every dimensional check
A pharma customer running a Waters autosampler reported intermittent needle bends, roughly one every 800 injections, on a validated method. The vial supplier’s CoA was clean. Neck OD, height, and bead diameter all measured in the middle of tolerance. The customer had already replaced the septum twice and blamed coring.
We pulled 200 vials from the lot and measured camber on a comparator with a V-block and a dial indicator. The distribution was bimodal. About 12% of the vials leaned 0.14 to 0.18 mm; the rest were under 0.06 mm. The lean correlated with position on the lehr, which correlated with a cooling-air nozzle that had drifted. The vials that leaned sat at an angle in the tray well, the gripper picked them off-axis, and the needle entered at a 0.3° tilt. Over hundreds of injections the needle fatigued and bent.
The fix was on the supplier’s side: recalibrate the lehr air profile and add a camber check at 100% via a vision gauge. The customer’s fix was a camber limit of 0.10 mm max in the incoming spec. The point is that no dimensional check on the drawing would have caught it, because camber was not on the drawing. It is now.
Field case: the neck OD that drifted over a three-week run
A different customer, a contract lab running EPA methods, started seeing crimp seal failures on about 4% of vials from a single lot. The seals popped off during the run, which is a hard failure because it exposes the sample to air and the run is invalid. The supplier’s capability report showed Cpk 1.6 on neck OD.
We asked for the control chart. The chart showed a slow upward drift over three weeks, from a mean of 11.02 mm to 11.09 mm. The Cpk was calculated on the first week’s data, when the process was centered. By week three the mean was 0.06 mm high, the upper spec limit was 11.10 mm, and the process was running against the wall. The drift traced to a worn blow-mold insert that was changing the neck geometry as it eroded. The mold had been in service for 180,000 shots against a 150,000-shot replacement interval.
The fix was a mold change and a re-baselined control chart with a preventive replacement schedule at 120,000 shots. The lesson is that a capability index is a point-in-time measurement. Without the chart, you are buying a number that was true last quarter. This is exactly the kind of thing a distributor cannot see, which is why I prefer to buy where the mold and the chart are in the same plant. If you are comparing sources, look at how they handle their own sample vial production data before you look at price.
Where the tolerance stack actually breaks: the cap-to-vial interface
Here is the mechanism most people miss. The autosampler does not grip the vial. It grips the cap. The gripper jaws close on the cap skirt, and the cap sits on the neck bead. So the dimension that determines whether the vial is picked cleanly is the assembled cap OD, not the vial OD. And the assembled cap OD is the sum of the neck OD, the cap wall thickness, and the crimp compression.
If the neck OD runs +0.08 mm and the cap wall runs +0.04 mm, the assembled cap OD runs +0.12 mm over nominal, and on a gripper with a 0.15 mm clearance you are 0.03 mm from a jam. That is why a vial that passes on instrument A fails on instrument B: the grippers have different clearances. Agilent and Thermo autosamplers do not use identical gripper geometry, and the tolerance stack that clears one may not clear the other. You can read the instrument-side specs in the Agilent consumables documentation and the Thermo Fisher sample handling literature, and the clearances are not the same. Neither vendor publishes a gripper clearance number, which is why you have to test on your own instrument.
The practical fix is to control the assembled cap OD, not just the vial neck OD. Crimp a sample of vials with your production crimper at your production torque and measure the assembled OD. That is the number your autosampler sees. A supplier who only measures the bare glass is measuring half the stack.
The specification I would write into a purchase order
If you are writing the incoming spec for a 2 mL clear crimp vial, here is what I would put in it, and why each line earns its place.
- Neck OD: 11.00 ± 0.10 mm, Cpk ≥ 1.33, Ppk ≥ 1.00. Both indices, because the gap between them tells you about drift.
- Camber: 0.10 mm max, 100% inspected. Not a Cpk dimension, a hard limit. Camber is the silent killer.
- Bottom thickness: 0.85 ± 0.15 mm, Cpk ≥ 1.33. Drives residual volume.
- Assembled cap OD after crimp at 1.2 N·m: within 0.15 mm of nominal. This is the number the gripper sees.
- Capability data delivered with every lot: control chart, sample size, cavity traceability, gauge R&R.
- Mold maintenance records on request. Shot count against replacement interval. If they will not share it, that is your answer.
The torque value matters because it is the one variable you control. Too little and the seal leaks. Too much and you deform the neck or crack the glass. I run crimpers at 1.2 N·m on 11 mm aluminum seals and check pull-off force on a sample every shift. The crimper is a variable in the tolerance stack as much as the mold is.
What to do when a lot is marginal
You will occasionally get a lot that measures inside print but with a distribution that is shifted toward one limit. The instinct is to accept it because it passes. Do not. A marginal lot is a capability warning. Here is the decision rule I use.
If Ppk is below 1.00 on any of the four critical dimensions, reject the lot regardless of whether individual parts pass. If Ppk is between 1.00 and 1.33, accept only with a 100% inspection gate on that dimension and a written corrective action from the supplier. If Ppk is above 1.33, accept and move on. The reason is simple: a lot with Ppk 0.9 has a real tail of out-of-spec parts that your AQL sample will not catch, and the failure mode is a jam at 3 a.m. in a run you cannot repeat.
This is the same logic that applies across the whole consumables category. A chromatography products supplier who can show you capability data on their vials is usually showing you the same discipline on their columns and fittings. The ones who cannot are selling you a price. For anything method-critical, the capability data is worth more than the certificate. You can cross-check the analytical expectations against method guidance from U.S. EPA SW-846 and the technical notes at Restek, both of which assume the vial is not a variable. Your job is to make sure it is not.
The decision rule is this: buy on demonstrated process capability on the four critical dimensions, not on nominal tolerance and not on price. A vial that holds ±0.10 mm at Cpk 1.5 costs a few cents more per piece than one that holds ±0.10 mm at Cpk 0.9, and it will save you a failed run, a bent needle, and a 3 a.m. phone call. Ask for the control chart. If the supplier cannot produce it, the tolerance is a wish, not a capability, and you should source elsewhere.
References & further reading
Related resources from our lab consumables network:
- Hplc Autosampler Vial at hplcautosamplervials.com
- Sample Vial at samplevialprice.com
- Chromatography Products at chromatographyselling.com
- Aijiren at aijirentech.com
External references and standards:
