Lot 24-0611 of 2 mL clear glass vials passed everything on the certificate. On the third injection of a 96-position sequence the autosampler flagged a pickup error, and the run aborted at position 41. We pulled the lot and measured it: neck OD averaged 10.94 mm against a nominal 11.00 mm, a shift of 60 µm from the previous lot. That is inside the drawing tolerance and outside what the gripper on that instrument would accept. Nothing on the paperwork predicted it, and no single vial looked wrong. The drift only showed up when 96 of them were handled in sequence, which is exactly the condition the certificate never tests.
That is the problem with hplc autosampler vial lot-to-lot variation and incoming inspection: the parameters that move between lots are rarely the ones printed on the CoA, and the ones that move are the ones that interact with a mechanical interface. This article is about which parameters actually drift, how far, and the 20-minute test that catches the drift before you load a sequence.
Glass forming is a thermal process with a moving setpoint. Gob weight, mold temperature, and cooling air all wander through a shift, and the vial that comes out carries that wander in its dimensions. Over 15 years of incoming inspection I have watched the same five parameters move, in roughly this order of frequency.
Neck OD and the crimp shelf
This is the one that bites first. The 11 mm crimp neck has a tolerance band of about ±0.2 mm on most drawings, but autosampler grippers and tray nests are set much tighter than the drawing. A gripper that opens to 11.15 mm will grip a 10.98 mm neck and drop a 10.90 mm neck. Lot-to-lot neck OD drift of 50 to 80 µm is normal from a forming line that has not been requalified, and it is invisible on a certificate that reports “11 mm” as a nominal.
Body OD and concentricity
The 2 mL vial sits in a 12 × 32 mm envelope, and the tray nest is cut to about 12.1 mm. Body OD drift of 0.1 mm turns a free-dropping vial into a sticking one. Concentricity matters more than OD for the same reason: a vial that is round but off-center by 0.15 mm will bind in a nest even when its maximum diameter is in spec. You cannot see this by eye and you cannot catch it by weighing.
Septum thickness and durometer
Silicone and silicone/PTFE septa are molded, and mold wear shows up as thickness drift. A septum that runs 0.95 mm instead of 1.00 mm changes the compression the crimp cap applies, and compression sets the seal. The same 5 µm shift changes how the needle enters: a thin septum coring at a different point in the stroke. Durometer drift of 3 to 5 Shore A is common between lots from the same mold, and it changes puncture force by 15 to 20%.
Crimp cap skirt height and wall thickness
Aluminum caps are stamped, and stamping dies wear. Skirt height drift of 0.1 mm changes how much the crimper has to deform the cap to reach a seal. A cap that is 0.1 mm short crimps to a loose seal at the same crimper setting; a cap that is 0.1 mm tall crimps to a tight seal and cracks the neck. This is the parameter most often blamed on the crimper when the crimper is fine.
Glass hydrolytic class and extractables
This one does not show up as a mechanical failure. It shows up as a ghost peak at 8.2 minutes that was not there in the previous lot. Borosilicate composition drift, or a change in the annealing lehr, changes the surface chemistry enough to move silanol activity. USP <660> and the hydrolytic resistance test in USP <661> catch the gross failures; they do not catch the 20% shift in silanol that produces a small ghost peak in a sensitive method. For that you need a blank run, which is the second half of the test below.
The certificate tests the lot, not the interface
Every CoA I have seen reports dimensions as nominal with a tolerance, and reports them from a sample of 10 to 30 vials pulled at the end of the forming run. That sample is fine for what it measures. It does not measure the interface. It does not measure how the vial behaves in a gripper, in a tray nest, or under a needle at 30 mm/s.
The gap between tested and nominal is the whole problem, and it is worth being blunt about it: a certificate that says “neck OD 11.0 ± 0.2 mm” tells you the lot is inside a band that is four times wider than the band your instrument cares about. If your gripper needs 11.00 ± 0.05 mm, the certificate cannot confirm or deny that. You have to measure it yourself, on the interface, at the resolution the interface cares about.
This is also why a lot can pass on one instrument and fail on another. The certificate is instrument-agnostic. The failure is not. If you run the same vial on an Agilent 7693 and a Thermo Fisher TriPlus, the two grippers have different opening tolerances and the same lot will behave differently on each. Your incoming test has to be run on the instrument that is most sensitive, or on a fixture that mimics it.
A 20-minute incoming test that catches drift before the sequence starts
The test has four parts and uses equipment most labs already have. It does not replace a full dimensional audit, and it is not a substitute for the supplier’s own SPC. It is a gate: pass, release to stock; fail, hold the lot and escalate. Run it on every incoming lot of 2 mL vials and on every new lot of caps and septa.
Part 1: dimensional spot check on 30 vials
Use a digital caliper with 0.01 mm resolution, or better a bench micrometer. Measure neck OD at two points 90° apart on each of 30 vials pulled from three positions in the case (top, middle, bottom). Record the mean and the range. A lot whose neck OD range exceeds 0.08 mm across 30 vials is drifting, even if every individual value is inside the drawing tolerance. The range is the early warning; the mean is the lagging indicator.
Do the same for body OD and for cap skirt height. Thirty vials, three positions, two measurements each. It takes about eight minutes with a caliper and a printed log sheet.
Part 2: the drop test
Load 30 vials into an empty tray and let them drop from the nest height under gravity. Count how many do not seat. Any sticking vial is a concentricity or OD problem. One in 30 is noise; three in 30 is a lot you should not release. This is the cheapest interface test there is and it catches the failures the caliper misses.
Part 3: the puncture test
On a spare autosampler or a bench fixture, puncture 20 septa with the needle gauge your method uses and count coring. A 22-gauge needle should give you zero cores in 20 punctures on a good lot. Two or more cores in 20 means the septum lot has a durometer or thickness problem. This is also where you catch the lot that will bend needles at position 300 of a 500-injection run. If you want the deeper version of this test, the puncture count and needle geometry interaction is covered in the coring article on this site.
Part 4: the blank run
Fill three vials from the new lot with your mobile phase, cap them, and run them as blanks under your method. Compare the chromatogram at 210 nm against the blank from the previous released lot. A new peak above 0.5 mAU that was not in the previous blank, or a shift in an existing peak of more than 10%, is a glass surface change. Hold the lot.
Decision table: what to do with the numbers
The thresholds below are the ones I use for 2 mL crimp vials on a standard 12 × 32 mm tray. They are starting points, not universal limits; tighten them if your method is sensitive to the parameter.
| Parameter | Nominal | Release | Hold and investigate | Reject |
|---|---|---|---|---|
| Neck OD (30-vial range) | 11.00 mm | ≤ 0.06 mm | 0.06 to 0.10 mm | > 0.10 mm |
| Body OD (30-vial range) | 12.00 mm | ≤ 0.08 mm | 0.08 to 0.12 mm | > 0.12 mm |
| Cap skirt height (30-vial range) | 6.00 mm | ≤ 0.08 mm | 0.08 to 0.12 mm | > 0.12 mm |
| Drop test (30 vials, not seated) | 0 | 0 | 1 | ≥ 2 |
| Cores in 20 punctures | 0 | 0 | 1 | ≥ 2 |
| Blank peak at 210 nm | baseline | < 0.3 mAU | 0.3 to 0.5 mAU | > 0.5 mAU or new peak |
Two things about this table. First, the release column is deliberately tighter than the drawing tolerance. If you release to the drawing, you are releasing to a band the instrument cannot resolve. Second, the hold column is not a rejection; it is a trigger to run the full test again on a second sample and to call the supplier. A lot that lands in the hold band twice is a lot you should not put into a validated method.
Field case 1: the lot that passed the caliper and failed the drop test
Symptom. A 96-position sequence on a Thermo autosampler aborted at position 63 with a pickup error. The previous lot had run clean for three months. The new lot’s neck OD measured 10.97 mm mean, range 0.05 mm, well inside release. The caliper said the lot was fine.
Measurement that identified it. The drop test: 4 of 30 vials did not seat. We then measured concentricity on the failing vials with a dial indicator on a V-block and found 0.18 mm runout on the body, against 0.06 mm on the previous lot. The neck was fine. The body was not round, and the nest was binding on the high side.
Fix. We held the lot, ran the drop test on a second sample, got 3 of 30 again, and returned it. The supplier’s forming line had changed a mold without requalifying the concentricity. We added the drop test to the incoming gate permanently. It now runs before the caliper, because it is faster and it catches the failure the caliper misses.
Field case 2: the ghost peak that tracked the glass lot
Symptom. A method running a 5 mM phosphate mobile phase at pH 6.8 developed a peak at 8.2 minutes that grew from 0.2 mAU to 1.4 mAU over two weeks. The peak appeared only on vials from one case, and only on the bottom row.
Measurement that identified it. Blank runs on three vials from the suspect case and three from the previous lot, same mobile phase, same instrument. The suspect lot gave a 1.1 mAU peak at 8.2 minutes; the previous lot gave 0.15 mAU. We then ran the same blanks in borosilicate vials from a third supplier and got 0.1 mAU. The peak was the glass, not the method.
Fix. The supplier had changed annealing lehr temperature by 15 °C to increase throughput. The change moved surface silanol activity enough to extract at pH 6.8. We did not reject the lot outright; we moved that method to a different vial lot and put the blank run into the incoming gate so the next change would be caught before it reached a sequence. The general lesson, that ghost peaks follow a release curve rather than a leak curve, is covered separately on this site.
Where the incoming test sits in the chain
The test is cheap, and it is not the whole answer. It catches drift at the lot boundary. It does not catch drift inside a lot, and it does not catch the change that happens when a supplier switches a mold mid-run. For that you need the supplier’s own SPC data, which is a commercial conversation and not this article.
What the test does do is move the detection point from position 63 of a sequence to the loading dock. That is the whole value. A failed lot caught at the dock costs you a return and a phone call. A failed lot caught at position 63 costs you a sequence, a re-run, and the sample you cannot get back.
If you are setting this up for the first time, start with the drop test and the blank run. They need no caliper, no fixture, and no training beyond a printed procedure. Add the dimensional spot check once you have a log sheet and someone who owns it. The full four-part test takes 20 minutes per lot and pays for itself the first time it holds a lot that would have aborted a run.
For the consumables side of this, the HPLC autosampler vial range and the sample vial specifications are where the nominal numbers come from; check them against your instrument’s gripper tolerance before you set your release limits. If you are sourcing across a chromatography products catalog, ask for the lot-level dimensional range, not the nominal. And if you want to see how a supplier documents its forming-line control, Aijiren publishes its SPC approach, which is a useful benchmark for the questions to ask. For method-level context, the U.S. EPA SW-846 methods and the Restek technical resources both assume the vial is a controlled input, which is the assumption this test is designed to verify.
The decision rule is simple: no lot enters a validated method until it has passed the drop test, the 20-puncture core count, and the blank run against the previous released lot. If any of the three fails, hold the lot and do not run it. The caliper is the fourth check, not the first, because the failures that abort sequences are the ones the caliper cannot see.
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:
