A customer sent back six lots of 9 mm screw-thread vials because their autosampler was throwing 0.4% RSD on a 2 µL injection that had run at 0.08% for two years on the previous supplier. The vials looked identical. The septa were the same 1 mm silicone/PTFE. The caps torqued to the same 1.2 N·m. What changed was the neck bore: our molding tool had drifted from 6.48 mm to 6.61 mm over 340,000 shots, and the resulting increase in headspace dead volume was enough to widen the vapor plug and smear the injection band. That is a dimensional tolerance capability problem, not a materials problem, and it is where most OEM quality disputes in HPLC consumables actually live.
This article is about tolerance capability in HPLC consumables production: what each manufacturing process can actually hold, how to measure capability rather than assume it, and which dimensions on which parts are the ones that fight back. If you are sourcing vials, caps, ferrules, or column hardware and you have ever been told “we hold ±0.1 mm,” this is the conversation you need to have with your supplier before the next tool wears out.
Process capability is a number, and most consumable suppliers cannot produce it
When a supplier says they hold a tolerance, ask for the Cpk. If they cannot send a histogram with a sample size and a date, they are quoting a drawing, not a process. A drawing tolerance is a permission. Capability is a measurement.
For HPLC consumables, capability is expressed the same way it is in any other molding operation: Cpk = min[(USL − μ)/(3σ), (μ − LSL)/(3σ)], where USL and LSL are the upper and lower specification limits and σ is the within-lot standard deviation from a stable process. A Cpk of 1.33 means the process is 4σ from the nearest limit and will produce roughly 63 ppm out of tolerance if the mean stays put. A Cpk of 1.67 is 5σ and about 0.6 ppm. Most automotive and medical molding contracts specify 1.33 minimum, 1.67 for critical features. Most chromatography consumable purchase orders specify nothing at all, which is why you get drift.
Two things go wrong when you ask for Cpk on a consumable:
- They report Ppk instead. Ppk is overall capability including between-lot shifts. It is a useful number, but it hides tool wear. If you want to know whether the tool will still be in capability at shot 500,000, you need Cpk measured per lot, plotted over shot count, and you need the tool wear slope.
- They measure the wrong feature. The critical dimension on a screw-thread vial is not the outside diameter of the neck. It is the sealing land diameter, the flat annular surface the septum compresses against. It is a 0.6 mm wide face on a 6.5 mm bore, and it is the one dimension that determines whether the cap seals at 1.2 N·m or leaks at 0.8.
What each process can actually hold, in microns
Here is the table I use when a customer asks why a molded vial costs more than an extruded one. These are the capability ranges I have seen from stable, in-control processes running PP, COC, or PFA on production tools. They are not theoretical best-case; they are what you can expect from a supplier who is actually measuring.
| Feature / process | Typical achievable tolerance (mm) | Cpk at ±0.05 mm | Notes |
|---|---|---|---|
| Injection-molded PP vial, neck OD | ±0.08 | ~0.9 | Draft angle and gate location dominate |
| Injection-molded PP vial, sealing land flatness | ±0.03 | ~1.5 | Needs a hardened insert, not a core pin |
| Injection-molded COC vial, bore ID | ±0.04 | ~1.7 | COC shrinks 0.5–0.7%; PP shrinks 1.2–2.0% |
| Extruded PTFE tubing, ID | ±0.10 | ~0.4 | Wall thickness variation is the limit |
| Machined PEEK ferrule, cone angle | ±0.5° | n/a | Lathe runout is the enemy, not the tool |
| Machined stainless column end fitting, thread | 6g/6H per ISO 965 | n/a | Thread class, not a tolerance |
| Die-cut silicone septum, thickness | ±0.05 | ~1.2 | Compression set dominates at temperature |
| Die-cut silicone septum, diameter | ±0.15 | ~0.4 | Web stretch in the die |
Read that table carefully. The die-cut septum diameter is the worst capability on the list, and it is the dimension most often specified on a purchase order as if it were critical. It is not. The septum diameter can be ±0.15 mm and still seal perfectly because the cap compresses it against the land. What matters is the land diameter and flatness on the vial and the septum thickness. If your supplier is spending metrology budget on septum OD and not on vial land flatness, they are measuring the wrong thing, and your injections will drift.
The three dimensions that fight back
Every consumable has a handful of features that are hard to hold. These are the ones I have watched go out of capability across multiple suppliers, and they are the ones you should put on an incoming inspection plan even if you buy from a distributor.
1. The sealing land on a screw-thread vial
The land is the flat annular face at the top of the neck, typically 0.5 to 0.7 mm wide, that the septum compresses against. Two things go wrong. First, mold flash at the parting line creates a 10 to 30 µm ridge that the septum cannot conform to, and you get a leak path that only shows up under vacuum (headspace) or under pressure (HPLC). Second, core pin wear changes the land diameter over shot count, and because the cap torque is fixed, the compression ratio changes. A 6.48 mm land and a 6.61 mm land with the same cap and septum produce different seal pressures. That is the drift that killed the customer’s RSD.
How to control it: specify land flatness at ≤10 µm, land width at 0.55 mm ±0.05 mm, and require a per-lot Cpk report on land diameter with a minimum of 1.33. Ask for the tool wear plot. A good molder will have it.
2. The ferrule cone angle and bore concentricity
A PEEK or stainless ferrule is a small part with a big job: it has to swage onto the column or tubing OD and hold 400 bar without leaking. The critical dimensions are the cone angle (typically 10° or 1/16 in. per the fitting standard), the bore ID, and the concentricity between them. Concentricity is the one that fails quietly. If the bore is eccentric by more than 15 µm, the ferrule swages asymmetrically, and you get a leak that only appears above 250 bar or after 200 injections when the PEEK has crept. Machined ferrules from a lathe with 5 µm runout hold this. Ferrules from a lathe with 25 µm runout do not, and you cannot see the difference with calipers.
Control: specify concentricity ≤15 µm TIR and cone angle ±0.5°. Verify with an optical comparator on a sample of 10 per lot, not with a pin gauge.
3. Septum thickness and compression set
A 1 mm silicone septum in a 9 mm screw cap is compressed to roughly 0.7 mm at 1.2 N·m torque. That is a compression ratio of about 30%, which is right for silicone. The problem is not the as-molded thickness (die-cutting holds ±0.05 mm easily). The problem is compression set after the first puncture. A septum that has been punctured 50 times at 40 °C will take a permanent set of 15 to 25% and will not reseal to the same force. If your autosampler reuses septa across a sequence, the tolerance that matters is not the dimensional tolerance, it is the puncture-to-failure count, and that is a material and cure property, not a molding one. Ask for the compression set data per ASTM D395 Method B at 70 °C for 22 h. Good silicone is under 20%. Cheap silicone is over 35%.
For the closures themselves, the cap thread engagement and the torque-to-seal curve are the practical control. A vial closures supplier who can send you a torque-to-leak curve (torque in N·m on the x-axis, leak rate on the y-axis) is telling you they understand the dimension that matters. One who sends a drawing is not.
How to measure capability without a metrology lab
You do not need a CMM to check a vial. You need three instruments and a sampling plan.
- Optical comparator or vision system for land width, land flatness, and ferrule concentricity. A 10x comparator with a 0.001 mm stage is enough. Measure 10 parts per lot, three lots per quarter, and plot the mean against shot count.
- Air gauge or bore gauge for vial bore ID and ferrule bore ID. Air gauging resolves 1 µm and is fast enough for 100% inspection if you need it.
- Torque tester with a leak fixture for the assembled cap-vial-septum stack. Torque to 1.2 N·m, pressurize to 2 bar with air, submerge in water, count bubbles. A good stack holds zero bubbles for 60 s. A marginal stack leaks one bubble every 10 s. That is the difference between a Cpk of 1.3 and a Cpk of 0.8.
If you are buying OEM and private-label consumables, put the capability requirement in the contract. Specify the feature, the tolerance, the Cpk minimum, the sample size, the measurement method, and the reporting frequency. A supplier who cannot meet that is a supplier who will drift on you at shot 200,000, and you will find out from your customers, not from them. The HPLC consumables market has enough suppliers now that you can demand this and get it.
Two field cases where capability, not material, was the root cause
Case 1: the drifting headspace RSD
Symptom: A contract lab running EPA Method 8260 by headspace GC-MS saw RSD on a 2 µL injection climb from 0.08% to 0.4% over six weeks, then plateau. No change in column, liner, or method.
Measurement: We pulled 30 vials from three lots and measured land diameter on an optical comparator. Lot A (received 8 weeks prior) was 6.48 mm ±0.02, Cpk 1.6. Lot C (received 2 weeks prior) was 6.59 mm ±0.04, Cpk 0.7. The tool had run 340,000 shots without a core pin change, and the molder had not plotted wear.
Fix: The molder replaced the core pin insert and added a per-lot Cpk report on land diameter with a 1.33 minimum and a wear limit at 250,000 shots. RSD returned to 0.09% on the next lot. The lab now measures land diameter on every incoming lot with a comparator, 10 parts, and rejects below Cpk 1.33.
Case 2: the ferrule that leaked at 280 bar
Symptom: An analytical lab running a 300 bar method on a 2.1 mm column saw intermittent leaks at the column inlet fitting, always after 150 to 200 injections on a PEEK ferrule, always at the same fitting. New ferrules fixed it for another 150 injections.
Measurement: We sectioned 10 used ferrules and measured bore-to-cone concentricity on an optical comparator. Failed ferrules were 22 to 28 µm TIR. Good ferrules from a different lot were 8 to 12 µm TIR. The supplier’s lathe had a worn collet, and the runout was invisible to a pin gauge.
Fix: The supplier replaced the collet and added a 100% optical concentricity check at 15 µm TIR on the ferrule line. The lab switched to a supplier who could provide the TIR data per lot. Leak rate went to zero over the next 400 injections. The lesson: for machined fittings, concentricity is the tolerance that matters, and it is the one no one specifies.
A decision checklist for specifying tolerance capability
When you write the next purchase order or OEM agreement, put these clauses in it. They are specific enough that a supplier either has the data or does not, and you find out before the tool wears.
- Feature list: Name the critical dimensions explicitly. For a screw vial: land diameter, land width, land flatness, bore ID, neck OD. For a ferrule: cone angle, bore ID, concentricity. For a septum: thickness, durometer (Shore A ±3), compression set per ASTM D395.
- Tolerance and Cpk: State the tolerance and the Cpk minimum (1.33 for critical, 1.67 for safety-critical). State the sample size (n ≥ 30 per lot) and the measurement method (comparator, air gauge, torque-leak fixture).
- Tool wear plan: Require a wear plot and a defined shot-count limit for core pin or insert replacement. Ask for the last 12 months of Cpk data by lot.
- Change control: Any tool, resin lot, or process change requires a first-article inspection with a full capability report before shipment. No exceptions.
- Incoming inspection: Define your own AQL. For critical dimensions, use a tightened plan (e.g., AQL 0.65, general inspection level II per ISO 2859-1) and measure the critical features, not just the cosmetic ones.
If you are sourcing across multiple suppliers, the chromatography supplier network is a useful cross-check on what is achievable, and the major instrument vendors publish their own consumable specifications that you can use as a benchmark: Agilent’s chromatography consumables and columns documentation and Thermo Fisher’s chromatography and sample handling specs are both worth reading before you set your own limits. For method-driven work, the EPA SW-846 test methods define the container performance you have to meet, and Restek’s technical resources have practical guidance on fitting and sealing that maps to the dimensional features discussed here. For OEM programs, the OEM vial specification sheets are a good template for how to write a capability clause that a molder can actually sign.
The decision rule
If a supplier cannot give you a per-lot Cpk on the sealing land diameter, the ferrule concentricity, and the septum compression set, they are selling you a drawing and hoping the tool holds. Specify Cpk 1.33 minimum on those three features, put a shot-count wear limit in the contract, and measure the land diameter yourself on every incoming lot with a comparator. The rest of the dimensions will follow. The three that will not follow on their own are the three that decide whether your RSD stays at 0.08% or drifts to 0.4%, and by the time you see the drift in the data, the tool has already been out of capability for six weeks.
References & further reading
Related resources from our lab consumables network:
- Hplc Consumables at hplcconsumable.com
- Chromatography Supplier at chromatographysupplier.com
- Oem Vial at oemsamplevials.com
- Vial Closures at vialclosures.com
External references and standards: