What tolerance each molding process can actually hold on a vial body

A 2 mL glass vial that mics 11.62 mm at the body and 11.79 mm at the crimp shoulder will pass incoming inspection on a three-point sample and still jam in a 96-position autosampler on position 61. The average is fine. The ovality is not. I have watched that exact lot burn a Saturday of instrument time, and the root cause sat in the mold cooling circuit, not the glass chemistry. If you are specifying lab vials dimensional tolerance capability on a purchase order, the number you write down is only as good as the process behind it and the dimensions you chose to control.

Vendors quote “±0.1 mm” casually. That figure means nothing without a process attached. Here is what I have measured across supplier audits and first-article inspections over the years, and what I will actually accept on a drawing.

Process Realistic body OD tolerance Neck ID tolerance Height tolerance Where it breaks
Injection blow molding (glass, Type I) ±0.15 mm ±0.10 mm ±0.25 mm Parison temperature drift between cavities
Injection molding (cyclic olefin polymer) ±0.05 mm ±0.03 mm ±0.10 mm Gate freeze and pack pressure variation
Injection molding (polypropylene) ±0.08 mm ±0.05 mm ±0.15 mm Post-mold shrinkage over 24 to 48 hours
Extrusion blow molding (HDPE) ±0.30 mm ±0.25 mm ±0.50 mm Parison sag and wall thickness swing
Compression molding (thermoset cap) ±0.10 mm ±0.08 mm ±0.20 mm Flash line and cure cycle

The glass number is the one most buyers get wrong. They expect injection-molded precision from a blow-molded tube. You will not get ±0.03 mm on a glass body OD at production speed, and a supplier who promises it is either sampling the best cavity or measuring the wrong feature. Glass gives you excellent chemical inertness and a neck finish that holds a crimp, but its body dimensions drift with parison temperature and mold wear in ways a polymer molder can control more tightly.

Consider what the numbers mean at the tool. A glass parison leaves the gob at roughly 1100 °C and is blown against a mold held near 500 °C. The wall thickness at any point depends on how the parison stretched, and the stretch depends on the temperature gradient down the parison. A 15 °C difference between two cavities on the same IS machine moves body OD by 0.05 to 0.08 mm. That is why glass capability is a cooling story as much as a mold story. Polymer injection molding runs a melt at 230 to 260 °C into a tool at 40 to 80 °C, and the tolerance is governed by pack pressure and gate freeze. Hold pressure at ±2 bar on a cyclic olefin part holds body OD to ±0.03 mm; lose that control and you are at ±0.08 mm before lunch.

If your application is autosampler-driven and the gripper is unforgiving, a lab vial molded from cyclic olefin polymer at ±0.05 mm body OD is the more honest choice than fighting a glass tolerance the process cannot hold. That is a specification decision, not a brand preference.

Capability is a distribution, and Cpk is the only number that answers the question

A tolerance is a box. Capability is whether the process lives inside that box day after day. The measurement is Cpk, and for a vial feature I want to see 1.33 minimum and 1.67 preferred on anything that touches the instrument.

Here is how you get there without trusting a certificate. Pull 30 consecutive parts from one cavity across a full shift, not a hand-picked sample. Measure the critical feature with a tool that resolves to one-tenth of the tolerance (a 0.001 mm resolution micrometer for a ±0.05 mm feature). Compute mean and standard deviation. Cpk = min[(USL − mean), (mean − LSL)] / (3σ). If the supplier cannot show you the raw 30-piece data with cavity ID, you are looking at a marketing number.

Run the arithmetic once and it stops being abstract. Take a ±0.10 mm body OD tolerance on an 11.6 mm nominal, so USL 11.70 and LSL 11.50. A process centered at 11.60 with σ = 0.020 mm gives Cpk = 0.10 / (3 × 0.020) = 1.67. Same mean, σ = 0.025 mm gives Cpk = 1.33. Same mean, σ = 0.033 mm gives Cpk = 1.0, which means roughly 0.27% of parts fall outside the box, and on a 500,000-piece run that is about 1,350 out-of-spec vials. The mean did not move. The spread did. This is why I ask for σ, not just the average, and why a supplier who reports only an average has told me nothing about capability.

I insist on per-cavity Cpk, not a pooled figure. A 16-cavity mold with one hot runner zone running 4 °C cool will pool to a respectable 1.4 while cavity 11 sits at 0.9. That single cavity ships the jams. The Aijiren brand runs cavity-level capability studies as standard because pooled data hides exactly the failure you will experience on the line.

The measurement system has to be capable first

Before you believe any Cpk, run a gauge R&R. If the measurement system contributes more than 10% of the tolerance in repeatability and reproducibility, the capability study is measuring your caliper, not the vial. Optical comparators and vision systems beat hand mics on neck ID and ovality because operator feel introduces variation you cannot see in the number. I have thrown out more than one “capable” report that fell apart under a 30-piece gauge R&R. A hand micrometer on a 0.10 mm tolerance feature typically eats 20 to 30% of the tolerance in operator repeatability alone, which means the study cannot resolve a real shift until it is already a scrap event.

Which vial dimensions are genuinely hard to control

Not all features are equal. Some hold tolerance easily and some fight you every cycle. Knowing which is which tells you where to put your inspection money.

Ovality is harder than diameter

A vial can hit its average OD dead on and still be oval. Ovality is the difference between the maximum and minimum diameter on the same cross-section, and it is the feature that kills autosampler grippers and crimpers. I specify ovality ≤ 0.10 mm on a 11.6 mm body for instrument use. The cause is almost always asymmetric cooling: one side of the mold runs hotter, the part shrinks unevenly, and you get an egg where you drew a circle. A three-point micrometer will miss it. You need a two-point sweep or vision measurement at 90° intervals.

Ovality also has a nasty habit of being invisible at the gate. A 16-cavity tool with a single water manifold can run 3 to 5 °C hotter on the return side, and the cavities at the far end of the circuit inherit the imbalance. The parts mic within tolerance on a single-point gauge and fail on a sweep. I have seen a 0.09 mm ovality on a lot that read 11.60 mm average across 20 pieces. The average was a lie about the shape.

Neck ID and the crimp seat

The neck finish is where the septum seats and the crimp bites. Neck ID tolerance drives coring behavior and seal integrity. On a 11 mm crimp finish, I want neck ID held to ±0.10 mm and the crimp seat height to ±0.15 mm. The seat height is the sneaky one: if it drifts, the crimp either fails to compress the septa or over-compresses and thins it, and you get coring on the first injection. That is a dimensional failure wearing a septum costume.

Concentricity of neck to body

This is the dimension nobody puts on the drawing and everybody suffers from. If the neck axis is offset from the body axis by more than 0.15 mm, the vial sits crooked in the autosampler nest, the needle enters at an angle, and you get bent needles and partial injections. Measure it with a dial indicator on a rotating fixture or a vision system. Blow-molded glass is worse here than injection-molded polymer because the parison can shift in the mold. A 0.20 mm TIR on a glass vial is common enough that I treat it as the default risk unless the supplier proves otherwise with a 30-piece study.

Bottom thickness and flatness

A vial with a domed or dished bottom does not sit flat in the nest, and the needle depth changes by the amount of the dome. I hold bottom thickness to ±0.15 mm and flatness to 0.10 mm across the base. This matters most for VOA vial geometries where the sample volume and the needle-to-bottom clearance are both tight. A 0.15 mm dome on a vial with a 2 mm needle clearance is a 7.5% error in needle-to-bottom distance, and that is enough to change where the needle tip sits in the sample.

Two field cases where the tolerance number was right and the part was wrong

Case one: the ovality that passed inspection and failed the autosampler

Symptom: a customer running a 96-position autosampler reported intermittent “vial not found” errors, roughly 4 per 96-position tray, always in the same region of the tray.

Measurement that found it: we pulled 50 vials from the failing lot and measured OD at 0° and 90° with a two-point sweep instead of the single-point mic used at incoming. Average OD was 11.61 mm, inside the ±0.15 mm spec. Ovality ranged up to 0.19 mm on 9 of the 50. The gripper jaws on that autosampler open to a fixed 11.70 mm gap; an oval vial presented its long axis and jammed.

Fix: we added ovality as a controlled feature (≤ 0.10 mm) with vision inspection at the mold, traced it to a cooling circuit imbalance on two cavities, and rebalanced the water flow. The lot was replaced. The lesson: the spec said diameter, the failure was shape, and the incoming gauge could not see shape.

Case two: the neck seat that cored every septum

Symptom: a pharma QC lab saw coring on the first injection across an entire case of vials and blamed the septa. Septum lots were swapped twice with no change.

Measurement that found it: we measured crimp seat height on 20 vials and found a mean of 6.42 mm against a 6.20 mm nominal, a 0.22 mm shift with the crimper set to its standard anvil depth. The over-height seat meant the crimp over-compressed the septum, thinning it from 1.0 mm to about 0.7 mm at the needle track. A 0.7 mm silicone-faced septum cores at a fraction of the puncture count of a 1.0 mm one.

Fix: the molder had changed a neck insert without re-qualifying seat height. We required seat height to ±0.15 mm with a first-article on every insert change and moved the crimper anvil to match the corrected seat. Coring stopped. The septum was never the problem.

How to write a tolerance callout that a molder can actually meet

Most purchase orders say “dimensions per drawing” and leave the rest to hope. A working callout names the feature, the nominal, the tolerance, the measurement method, and the sampling frequency. Here is the checklist I put on every vial drawing that touches an instrument.

  • Body OD: nominal ±0.10 mm, measured by two-point sweep at 0° and 90°, ovality ≤ 0.10 mm, 5 pieces per cavity per lot.
  • Neck ID: nominal ±0.10 mm, pin gauge or vision, 5 per cavity.
  • Crimp seat height: nominal ±0.15 mm, optical comparator, 5 per cavity, re-qualify on any insert change.
  • Neck-to-body concentricity: ≤ 0.15 mm TIR, dial indicator on rotating fixture, 3 per cavity.
  • Height: nominal ±0.25 mm, caliper, 5 per cavity.
  • Bottom flatness: ≤ 0.10 mm, 3 per cavity.
  • Capability: Cpk ≥ 1.33 on OD, neck ID, and seat height, per cavity, 30-piece study on request.
  • Gauge R&R: ≤ 10% of tolerance on every controlled feature, documented.

Notice what is not on the list: a single blanket tolerance. Blanket tolerances are where capability goes to die, because the molder optimizes to the loosest feature and the tight feature drifts. Name each one. The other thing not on the list is a sampling frequency of “per lot” with no cavity breakdown. I want the cavity ID on every measurement record, because the moment a lot fails I need to know which cavity to look at, and a lot-level average tells me nothing about which of 16 cavities is the problem.

The standard references that anchor the numbers

You do not have to invent these tolerances. The neck finish dimensions trace to GPI (Glass Packaging Institute) finish specifications for 11 mm, 13 mm, and 15 mm crimp and screw-thread finishes, and the dimensional conventions align with ISO 8362 for injection containers and ISO 4796 for laboratory glassware. When a supplier’s drawing disagrees with the GPI finish spec on the crimp seat, the GPI spec wins, because that is what the crimper anvil was machined to.

For method-driven work, the container dimensions have to survive the method. EPA SW-846 methods assume a vial that seals reliably and holds its volume; a neck seat out of tolerance breaks that assumption silently. Instrument vendors publish the mechanical envelope their autosamplers accept, and those envelopes are tighter than most vial drawings. Agilent’s autosampler accessory documentation and Thermo Fisher’s sample handling specifications both state gripper and nest dimensions you can check your drawing against. Restek’s technical resources are useful for the practical side of what happens when a vial does not seat squarely.

Troubleshooting: reading a dimensional failure from the symptom

When a line starts failing, the symptom usually points at one feature. Use this to narrow it before you pull the drawing.

  • Autosampler jams in the same tray region: ovality or body OD. Measure two-point sweep, not single-point.
  • Bent or deflected needle: neck-to-body concentricity or neck ID. Check TIR first.
  • Coring on first injection: crimp seat height or neck ID, not the septum. Measure seat height against nominal.
  • Leaks after crimping: seat height and neck ID together; over-compression thins the septum and under-compression fails to seal.
  • Inconsistent injection volume: bottom flatness or height; the needle depth is changing.
  • Vial will not seat in the nest: body OD high side or bottom dome.
  • Cap torque drifts on screw-thread vials: thread major diameter and neck OD; check against GPI finish.

Cap torque deserves a number: on a 13 mm screw-thread finish I target 0.6 to 0.9 N·m for a hand-tight seal with a PTFE-faced liner, and I verify it with a torque tester on 10 pieces per lot. Out-of-tolerance thread diameter is the usual cause of a torque that will not repeat. If the torque swings more than 0.2 N·m across a lot, check the thread major diameter before you touch the capper.

What to actually put on the purchase order

Here is the decision rule I use. Specify the process first, because the process sets the tolerance floor. Glass blow molding gets ±0.15 mm on body OD and no tighter, and if that fails your autosampler, change the process to injection-molded polymer rather than demanding a tolerance glass cannot hold. Then specify per-feature tolerances with named measurement methods and sampling frequency, never a blanket callout. Require per-cavity Cpk ≥ 1.33 on body OD, neck ID, and crimp seat height, with the raw 30-piece data and a gauge R&R ≤ 10% attached to the first article. Re-qualify on any mold insert change, because that is where seat height and concentricity move without anyone touching the drawing. If a supplier cannot produce the per-cavity data and the gauge study, the tolerance on the paper is a wish, and the ovality you did not specify is the jam you will get.

References & further reading

Related resources from our lab consumables network:

External references and standards: