A customer sent back a lot of 11 mm PTFE-faced silicone septa last March with a single line of complaint: “coring on the 96th injection, three vials.” We pulled the retained samples, put 20 septa on the vision system, and found the center hole was running 1.62 mm ± 0.04 mm against a 1.50 mm nominal. The supplier’s process was holding ± 0.08 mm on that feature and calling it in spec. That is the whole argument in one number. Dimensional tolerance capability for vial septa is not a spec sheet claim; it is a measured Cpk on a specific feature, on a specific tool, and most suppliers cannot tell you what theirs is because they never measured it.
This article is about what each molding process can actually hold, how you verify it, and which features on a septum will bite you when the process drifts. If you are sourcing septa and you have been quoted “± 0.1 mm” as a blanket number with no process attached, you have been quoted nothing useful.
What tolerance can each molding process hold on a septum, and where does it break
Septa are molded by three routes in practice: compression molding of silicone sheet stock, transfer molding, and liquid silicone rubber (LSR) injection molding. Each has a different capability envelope, and the envelope is set by the physics of the material, not by the machine’s positioning accuracy. A press with 0.005 mm platen repeatability will still produce parts with 0.10 mm thickness spread if the compound charge weight varies by 0.5 g and the shut-off is worn.
Compression molding: ± 0.10 mm on thickness, worse on concentricity
Compression molding is what most of the volume septa on the market come from. Silicone compound is placed in a heated cavity, the platens close, and the material flows under pressure. The problem is flash and shut-off. Thickness on a 3.0 mm nominal septum is realistically held at ± 0.10 mm (a 6-sigma span of about 0.20 mm) on a well-run press with good compound lot control. Center-hole diameter is worse, often ± 0.12 mm, because the core pin shifts under lateral flow pressure and the tool wears at the gate. Concentricity of the hole to the outer diameter is the feature that quietly ruins autosampler needle alignment, and compression tooling regularly drifts to 0.15 mm TIR after 200,000 shots without rework.
The compound charge weight is the hidden variable. A 16-cavity compression tool running a 45 Shore A silicone needs each cavity charged within about 3% of target. On a 0.4 g charge that is ± 0.012 g, and most manual or semi-automatic charge systems cannot hold that. The cavities that get overcharged produce a thicker part with flash at the parting line; the cavities that get undercharged produce a thin part that will not seal. That is why a compression tool’s capability on thickness is a function of the charging system as much as the press.
Transfer molding: ± 0.05 mm, and the reason is the closed cavity
Transfer molding closes the cavity before material enters, so the core pin is supported on both ends and does not see flow-induced side load. That gets you to ± 0.05 mm on hole diameter and thickness on a maintained tool. It is slower and the tooling costs more, which is why you see it on higher-end LC-MS septa and on anything with a bonded PTFE film where the film has to stay flat. A transfer tool typically cycles at 45 to 90 seconds per shot against 90 to 180 seconds for compression, and the scrap rate on a well-run cell is under 2% versus 5 to 8% on compression.
LSR injection molding: ± 0.02 mm on the good features, and a different failure mode
Liquid silicone rubber injection holds the tightest tolerance because the material arrives as a two-part pumpable and the cavity fills at low viscosity under controlled pressure. A good LSR tool holds ± 0.02 mm on hole diameter and ± 0.03 mm on thickness. The catch is that LSR tools are expensive (a 16-cavity LSR tool with core pins runs $40,000 to $80,000 depending on the feature count), and the process is unforgiving of cold slugs and cure drift. When an LSR cell goes out of control it does not drift gently; it produces short shots and cure blush in bursts. You will see 40 good parts and then 6 that are 0.15 mm thin with a tacky surface.
| Process | Thickness (3.0 mm nominal) | Center hole diameter | Concentricity (TIR) | Realistic Cpk on hole dia. | Tool life before rework |
|---|---|---|---|---|---|
| Compression | ± 0.10 mm | ± 0.12 mm | 0.10–0.15 mm | 0.9–1.2 | 150k–250k shots |
| Transfer | ± 0.05 mm | ± 0.05 mm | 0.05–0.08 mm | 1.3–1.6 | 300k–500k shots |
| LSR injection | ± 0.03 mm | ± 0.02 mm | 0.02–0.04 mm | 1.6–2.0 | 500k+ shots with insert maintenance |
Those Cpk numbers are what I have seen on production tools with 30-piece samples per shift. They are not theoretical. If a supplier quotes you a tolerance without a Cpk behind it, ask for the last 90 days of SPC data on the hole diameter. Most will not have it.
How capability is actually measured on a septum, and why your supplier’s number is probably wrong
Capability (Cpk) is the distance from the process mean to the nearest spec limit, divided by three standard deviations. The formula is trivial. What matters is how you sample and what you measure.
A septum is a soft, compressible, three-dimensional part. Calipers are useless for hole diameter because the jaw compresses the material. You need either a vision system (backlit, 20x minimum, measuring the actual aperture) or a pin gauge set graduated in 0.01 mm. I use pin gauges for shop-floor checks because they are fast and the operator cannot fudge them, and vision for the capability study because it captures concentricity in the same pass.
The sampling rule that matters: 30 consecutive parts from a single cavity, per shift, per tool. Not 30 parts pulled from a bin. Multi-cavity tools produce cavity-to-cavity variation that bin sampling hides completely. A 16-cavity compression tool can have one cavity running 0.06 mm off the mean and you will never see it in a bin sample.
Then compute Cpk on the hole diameter, the thickness at the sealing land, and the concentricity. If your spec is 1.50 ± 0.05 mm and the process standard deviation is 0.025 mm, your Cpk is 0.67. That is a process that will produce out-of-spec parts on a routine basis and a supplier who tells you it is “in tolerance” is telling you about the sample, not the process.
One more thing: measure after cure and after any post-cure bake. Silicone continues to shrink for 24 to 72 hours after demolding. A capability study run on green parts will overstate the hole diameter by 0.01 to 0.03 mm depending on the compound and the cure system. Run the study on parts that have sat at least 72 hours at ambient.
There is also a gauge R&R problem that nobody talks about. Two operators with the same pin gauge set will disagree on a soft silicone hole by 0.01 to 0.02 mm because one pushes harder than the other. If your gauge R&R is above 30% of the tolerance, your Cpk number is noise. Run the study with one operator and one gauge, or use a vision system where the operator only places the part.
Which dimensions on a septum are hard to hold, and which are easy
Not all features are created equal. Some are trivial to hold and some are the reason your autosampler misses.
Easy: outer diameter and overall thickness
Outer diameter on an 11 mm crimp septum is a straight cavity dimension. Any of the three processes holds ± 0.05 mm on OD without drama. Overall thickness is similar, though compression molding is the weakest because of flash at the parting line. These are the dimensions suppliers love to quote because they are the ones they can hit.
Hard: center hole diameter, and it gets harder as the hole gets smaller
The center hole is where the needle goes. On a 1.0 mm hole, a 0.05 mm drift is a 5% change in aperture, and that is enough to change the puncture force by 15 to 20%. The core pin that forms the hole is a slender feature that sees lateral flow pressure, thermal cycling, and wear. On compression tools it deflects. On transfer tools it is supported but still wears at the tip. On LSR it is the most expensive part of the tool and the first thing to be replaced.
The practical consequence: if your method depends on a specific puncture force (and coring behavior does), the hole diameter is the dimension you must control with a Cpk above 1.33. Anything less and you are gambling on the lot.
Hard: concentricity of hole to OD, because it drives needle alignment
A septum with a perfectly sized hole that is 0.10 mm off-center will cause the autosampler needle to enter at an angle. Over a 100-injection sequence that angle produces asymmetric coring and, eventually, a plug of silicone in the needle. Concentricity is the feature nobody quotes and everybody should. Ask for it in TIR (total indicator reading). Anything above 0.08 mm on an 11 mm septum is a problem for a 0.63 mm OD needle.
Hard: the sealing land flatness on a PTFE-faced septum
When a PTFE film is bonded to the silicone, the flatness of the sealing land becomes a tolerance in its own right. A land that is not flat within 0.03 mm will leak at the crimp even with correct torque (I use 1.5 to 2.0 N·m on an 11 mm aluminum crimp, and the seal depends on that land being flat). Bonding process control, not molding, sets this. It is the feature that separates a good PTFE-faced septum from a bad one, and it is almost never on the drawing.
The PTFE film itself is a variable. A 0.05 mm film bonded with a 0.02 mm adhesive layer adds 0.07 mm to the stack, and if the film is not tensioned uniformly during bonding it will wave. A 0.05 mm wave across the sealing land is enough to open a leak path on a crimp that is otherwise correct. I have seen a supplier ship a lot where the film was bonded with 0.04 mm of adhesive in the center and 0.01 mm at the edge, and the leak rate on a headspace method went from 0.2% to 6%.
Two field cases where the tolerance number told the story
Case 1: the drifting hole that looked like a coring problem
A pharma QC lab running a 96-well headspace method started seeing silicone fragments in the GC inlet liner around injection 80 of every sequence. They blamed the needle. We pulled 30 septa from the current lot and measured the hole with a vision system: mean 1.58 mm, standard deviation 0.031 mm, spec 1.50 ± 0.05 mm. Cpk was 0.43. The tool had run 340,000 shots on a compression press and the core pin had worn oval. The fix was not a new needle; it was a tooling change. We moved them to a transfer-molded septum with a documented Cpk of 1.4 on the hole, and the fragment problem stopped on the next lot. The measurement that identified it was the standard deviation on 30 consecutive parts, not the mean.
Case 2: the thickness spread that leaked at the crimp
A cannabis testing lab was getting intermittent leaks on 2 mL HPLC vial closures. They had checked crimp torque and it was in range. We measured thickness on 50 septa from the lot with a drop gauge (0.001 mm resolution, 5 N preload): the range was 2.86 to 3.14 mm, a 0.28 mm spread against a 3.00 ± 0.10 mm spec. The thin parts were not sealing because the crimp could not compress them enough to fill the land. The cause was a compression tool with a worn shut-off and inconsistent compound charge weight. The fix was a compound lot control change and a tool refurbishment; the leak rate dropped from about 4% to under 0.3%. Thickness is the easiest dimension to measure and the one most often skipped.
A verification protocol you can run on incoming lots in under an hour
You do not need a metrology lab. You need a plan and the right gauges. Here is what I would put in a receiving inspection procedure for a critical septum.
- Sample size: 30 consecutive parts, per cavity if the supplier will tell you the cavity count. If they will not, that is your first finding.
- Hole diameter: pin gauge set, 0.01 mm increments. Record go/no-go at the spec limits and the actual pin that fits.
- Thickness: drop gauge with 5 N preload, measured at the sealing land, not the center.
- Concentricity: vision system or optical comparator, TIR, one measurement per part.
- Hardness: Shore A durometer, 5 readings per lot. A shift of more than 3 points from the reference lot means the compound or cure changed.
- Puncture force: a simple load cell fixture with a 0.63 mm OD needle at 100 mm/min. Record the peak force on the first and 10th puncture. A rise of more than 20% between first and 10th is a coring warning.
- Post-cure check: re-measure hole diameter on 5 parts after 72 hours at ambient. If it moves more than 0.02 mm, the cure is not stable.
For context on how these dimensions interact with the rest of the vial system, the vial septa reference covers the dimensional stack, and the autosampler vial page has the needle geometry that sets your concentricity requirement. If you are running headspace or GC methods, the vial and closure selection guidance is worth reading alongside this, and the EPA SW-846 method suite is where many of these container tolerances get written into the method itself. Vendor technical resources from Agilent, Thermo Fisher, and Restek all publish inlet and needle specifications that translate directly into septum tolerance requirements.
What to put in a purchase specification, and what to refuse
Here is the decision rule. Write a spec that names the feature, the nominal, the tolerance, and the required Cpk. Refuse any quote that gives you a single blanket tolerance across all dimensions. Refuse any supplier who cannot produce 30-part SPC data on the hole diameter. Refuse any tooling that has not been refurbished within its documented shot count.
A workable spec for a critical 11 mm crimp septum looks like this: hole diameter 1.50 ± 0.05 mm, Cpk ≥ 1.33; thickness 3.00 ± 0.08 mm, Cpk ≥ 1.33; concentricity ≤ 0.08 mm TIR; Shore A 45 ± 3; puncture force first-to-tenth ratio ≤ 1.20. Put the measurement method on the drawing (pin gauge vs. vision) because the method changes the number. Put the post-cure age on the drawing (72 hours minimum) because green parts measure differently.
The decision rule: if a supplier cannot show you Cpk above 1.33 on the hole diameter and the thickness, they cannot guarantee coring performance across a 100-injection sequence, and you should not put their septa on an autosampler that matters. Capability is the number. Everything else is a brochure.
References & further reading
Related resources from our lab consumables network:
- Vial Septa at vialsepta.com
- Hplc Vial at hplcgc.com
- Autosampler Vial at autosamplerhplc.com
- Vial at vials-filters.com
External references and standards:
