A 2 mL clear glass vial from a supplier we had qualified for three years came in at 11.72 mm neck OD on the incoming gauge, against a print callout of 11.60 ± 0.08 mm. That is 0.04 mm over the upper limit. The autosampler gripper still picked it up. The run failed anyway, because the crimp cap sat 0.3 mm high and the septum did not seat square over the bore. Twenty-two injections, then a bent needle and a lost sequence. The dimensional tolerance capability of a vial supplier is not a marketing number. It is the sum of the process, the gauge, and the control loop behind the print.
Most buyers ask a vial supplier for a certificate of analysis and read the pass/fail. That tells you nothing about capability. A Cpk of 0.7 will still pass a lot if you sample loosely. The question that matters is narrower: which dimensions can this supplier’s process actually hold, how is that proven, and which features will drift first when the mold gets hot or the tooling wears.
Glass forming and plastic molding are not the same tolerance problem, and a vial supplier that runs both will quote you a single number that hides the difference. I want the print broken out by feature and by process. Here is what I have seen hold in production, not in a sample room.
| Feature | Process | Typical held tolerance (mm) | Capability risk |
|---|---|---|---|
| Body OD | Glass tubular conversion (2 mL) | ± 0.10 | Low, driven by tube draw |
| Body OD | Injection-molded plastic | ± 0.05 | Low to moderate |
| Neck OD (crimp) | Glass | ± 0.08 | Moderate, tool wear |
| Neck ID / bore | Glass | ± 0.10 | Moderate |
| Overall height | Glass | ± 0.25 | Low |
| Bottom thickness | Glass | ± 0.15 | Moderate, wall flow |
| Concentricity (bore to OD) | Glass | 0.10 TIR | High |
| Thread pitch (screw) | Plastic | ± 0.08 | Moderate, shrink |
| Sealing surface flatness | Glass | 0.05 | High |
Those numbers are not universal. They are what a competent vial supplier running a controlled process should be able to demonstrate on a capability study. If a supplier quotes ± 0.02 mm on glass neck OD across a million-piece run, ask for the control chart, not the certificate. You will not get it, because the number is not real.
Plastic holds tighter dimensional tolerance than glass on most features because injection molding is a closed-die process with a repeatable cavity. Glass is formed from a molten gob or a redrawn tube, and the wall thickness varies around the circumference in ways the die cannot fully correct. That is why the concentricity callout is the one that bites. A glass vial can be perfectly on-size on OD and bore and still be 0.15 mm eccentric, and that eccentricity is what pushes the needle off-center.
How capability is actually measured, and how suppliers game it
Capability is Cpk, and Cpk is only as honest as the sample behind it. The formula is straightforward: the lesser of (USL minus mean) and (mean minus LSL), divided by three sigma. A Cpk of 1.33 means the process spread fits inside the tolerance band with a 25% cushion. That is the floor I want on any critical dimension. Below 1.0 and you are inspecting quality in, not building it in.
Where suppliers game it: they compute Cpk from a 30-piece sample pulled from one cavity on one shift at the start of a run, before the mold has reached thermal equilibrium. The number looks great. On day three, when the tool is at temperature and the cooling lines have scaled, the mean has shifted 0.05 mm and the real Cpk is 0.9. The certificate still says 1.4 because nobody resampled.
What I ask for instead:
- Sample size and stratification. Minimum 125 pieces across all cavities, across at least three shifts, with the tool at production temperature. Note the cavity number on every piece.
- Gauge R&R first. If the measurement system eats more than 10% of the tolerance band, the capability number is noise. A neck OD measured with a caliper by three operators will show you that fast.
- Control charts, not snapshots. X-bar and R charts over the run. I want to see the drift, not a single clean histogram.
- Per-cavity capability. A 16-cavity mold can have one cavity running 0.06 mm off the rest. Aggregate Cpk hides it. Per-cavity Ppk exposes it.
A supplier who cannot produce per-cavity data is either not measuring per cavity or not willing to show you the outlier. Either answer is a reason to keep looking.
The dimensions that are genuinely hard to control
Not every feature on the print is equally difficult, and the ones that fail are predictable. These are the four I watch.
Concentricity of bore to body OD
This is the hardest one on a glass vial and the one that causes the most autosampler grief. The bore is formed by a mandrel; the OD is formed by the outer die. If the mandrel is not perfectly centered in the die, or if the tube wall is uneven, the bore drifts off-axis. At the needle entry point 20 mm up the neck, a 0.10 mm eccentricity becomes a 0.10 mm lateral offset. Add needle straightness tolerance and gripper repeatability and you have a recipe for the bent-needle event. Measure it with a dial indicator on a mandrel, or optically, not with calipers. Calipers cannot see concentricity.
Sealing surface flatness at the neck rim
The crimp or screw cap seals on a narrow annular land at the top of the neck. If that land is not flat to within about 0.05 mm, the septum compresses unevenly. You get a seal that passes a leak test at ambient and fails at 40 °C in the autosampler tray, or fails after 12 hours of headspace equilibration. This feature is hard because it is formed at the very end of the glass-forming cycle, when the glass is cooling and the tool is releasing. A worn or contaminated rim tool shows up here first. Inspect it with an optical flat or a profilometer on a sample per cavity per shift.
Bottom thickness and heel geometry
For autosampler vials that sit in a tray and get pushed up by a spring or a lift, bottom thickness variation changes the insert depth and the residual volume. A 0.15 mm spread on bottom thickness moves the sample pickup point. For 2 mL vials used with a 250 µL insert, that is the difference between recovering 180 µL and 150 µL. The heel radius also matters: a sharp heel traps sample; a generous heel leaves dead volume. Both are mold-controlled and both drift with tool wear.
Thread pitch and start on screw vials
Injection-molded screw vials shrink after ejection, and the shrink is anisotropic because the glass-fiber or the wall section is not uniform. Pitch can drift 0.03 to 0.05 mm across a run, and more importantly the thread start can rotate a few degrees. That means the cap that torques to 1.5 N·m on one vial torques to 1.1 or 1.9 on the next. Under-torque leaks; over-torque cracks the cap or strips the thread. Specify a torque window and hold the supplier to it with a torque tester on a sample per shift.
Two cases where the tolerance number told us where to look
Case one: the drifting neck OD
A pharma customer reported intermittent autosampler pickup failures on a 2 mL glass vial, roughly 1 in 400. The instrument was blamed first. We pulled 200 vials from the suspect lot and measured neck OD on a calibrated optical comparator. Mean was 11.68 mm, sigma 0.03, against a print of 11.60 ± 0.08. Cpk came out at 0.89. The upper tail was clipping the gripper’s release tolerance. The fix was not the vial; it was the neck die, which had worn 0.06 mm over 400,000 shots and had not been replaced on schedule. We moved the supplier to a die-life PM at 300,000 shots and added a per-shift SPC check on neck OD at the press. Pickup failures dropped to under 1 in 20,000. The lesson: the tolerance was right, the control loop was missing.
Case two: the concentricity fail that passed every gauge
A different lot passed OD, height, and bore ID on calipers and still bent needles on a GC autosampler. We measured concentricity with a dial indicator on a mandrel and found 0.16 mm TIR on one cavity, against a 0.10 mm print. The caliper never saw it because calipers measure size, not axis position. The root cause was a mandrel that had shifted 0.04 mm in its holder after a tooling changeover and was never re-centered. The fix was a mandrel alignment check as part of every changeover, plus an optical concentricity check on five pieces per cavity per shift. This is the kind of defect that a dimensional tolerance capability audit catches and a certificate of analysis does not.
What to put in the specification so the tolerance is enforceable
If you want the supplier to hold a tolerance, the print has to say what the feature is, how it is measured, and what the acceptance rule is. Vague callouts get vague compliance. I use a short decision checklist on every vial specification:
- Datum scheme. Define the primary datum (usually the body OD or the base) and the secondary. Without datums, concentricity is unmeasurable and unenforceable.
- Measurement method per feature. Name the gauge. Neck OD by optical comparator or air gauge, not caliper. Concentricity by dial indicator on mandrel. Flatness by optical flat. This removes the argument at inspection.
- Capability requirement. Cpk ≥ 1.33 on critical dimensions, Ppk ≥ 1.33 on the qualification run, per cavity. State it.
- Sampling plan. Frequency, sample size, and the reaction plan when a point goes out of control. A control chart with no reaction plan is decoration.
- Gauge R&R threshold. Under 10% of tolerance for critical features, under 30% for non-critical. Reject the capability study if the gauge fails.
- Change control. Any tooling, resin, glass tube source, or process change requires re-qualification with a fresh capability study. No exceptions.
For reference standards, tie the print to a recognized dimensional framework where one exists, and to the relevant pharmacopeial container requirements where the product is regulated. The EPA SW-846 methods set container expectations for environmental work, and USP container chapters reach the vial on the regulated side. Neither replaces a dimensional print, but both constrain what the vial supplier can ship.
How molding process variables move the tolerance
If you want to understand why a lot drifts, understand what moves the mean. In injection molding, four variables dominate dimensional outcomes on a vial:
- Melt temperature. A 10 °C rise reduces viscosity and increases shrink, shifting neck OD down by roughly 0.02 to 0.04 mm. Suppliers who run hot to improve cycle time pay for it in dimensional drift.
- Pack pressure and hold time. Under-pack gives short shots and high shrink; over-pack gives flash and stress. Both move the thread and neck dimensions.
- Cooling time. Short cooling ejects a part that is still shrinking, and the shrink continues in the tray. This is the single most common cause of within-lot drift I have seen.
- Tool temperature uniformity. A mold with uneven cooling channels produces cavities that run at different effective temperatures, and therefore different dimensions. This is why per-cavity capability matters.
In glass conversion, the variables are tube draw tolerance, mandrel alignment, die wear, and the thermal profile through the forming station. Glass is more stable lot to lot than plastic once the process is dialed in, but it drifts more slowly and more silently, which is why the die-life PM schedule is the control that actually protects you.
For laboratories running method development, the practical consequence is that the vial is a variable in the method. If you are chasing retention time drift or peak shape changes, rule out the vial dimension before you touch the column. A supplier change mid-study introduces a new dimensional population. That is why I keep a retained sample of the previous lot and re-measure neck OD and concentricity whenever a new lot lands. For more on the analytical side of that work, lab chromatography resources and the technical notes from Restek cover the method variables; the vial dimension is the one most often left out.
Qualifying a new supplier on tolerance capability, not on price
The qualification sequence I use is short and it filters hard. First, request the dimensional print and the per-cavity capability study from the last 12 months. If they cannot produce it, stop. Second, request the gauge R&R for the two or three critical features. If the gauge eats more than 10% of the tolerance, the capability number is not trustworthy. Third, order three lots and measure the critical features yourself on a sample of 30 per lot, stratified across cavities if cavity numbers are marked. Fourth, run a torque test on screw vials and a crimp-height check on crimp vials at the specified 1.5 N·m target and confirm the window holds. Fifth, put the change-control clause in the contract and hold them to it.
If you are buying vials for HPLC work, the same discipline applies to the filters and membranes you pair with them. A HPLC filter with an out-of-tolerance housing dimension will not seat in the vial, and a PTFE membrane filter with inconsistent pore geometry will change your recovery. The dimensional discipline is the same across the consumables you buy. Agilent publishes dimensional and fit specifications for its GC columns and accessories that are worth reading as a model for how to write a print, and Thermo Fisher’s chromatography and sample handling documentation shows the same approach to fit and tolerance on the instrument side.
Troubleshooting checklist when a lot does not fit
When a vial fails to seat, seal, or pick up, work the dimensions in order before you blame the instrument:
- Neck OD out of band? Measure with the specified gauge, compare to the print, check the per-cavity data. If one cavity is the outlier, it is a tooling problem.
- Concentricity over 0.10 mm TIR? Dial indicator on mandrel. If it fails, the mandrel shifted or the tube wall is uneven. Reject the lot and request a mandrel alignment record.
- Sealing surface not flat? Optical flat. If the rim is worn, the rim tool needs replacement, not adjustment.
- Torque outside the window? Torque tester on 30 pieces. If the spread is over 0.4 N·m, the thread start is rotating, which points to shrink variation and a cooling problem.
- Bottom thickness spread over 0.15 mm? Section a sample. If the heel is sharp, the insert depth is wrong and your recovery is compromised.
Each of these is a dimensional question with a dimensional answer. None of them is solved by a certificate of analysis that says “conforms.”
The decision rule
Buy on demonstrated per-cavity capability, not on a quoted tolerance. Require a Cpk of at least 1.33 on every critical dimension, measured with a gauge that passes a 10% Gauge R&R, stratified across cavities and shifts, and backed by a change-control clause that forces re-qualification whenever the tool, the resin, or the glass tube source changes. If a vial supplier cannot show you the control chart behind the number, the number is not a capability; it is a hope. The dimensions that will fail first are concentricity, sealing-surface flatness, bottom thickness, and thread pitch, so put your inspection budget there and leave the easy features to the certificate.
References & further reading
Related resources from our lab consumables network:
- Vial Supplier at vialsupplier.com
- Lab Chromatography at labchromatography.com
- Hplc Filter at hplcfilters.com
- Ptfe Membrane Filter at ptfemembranefilter.com
External references and standards: