Where the vapour actually escapes: four leak paths ranked by how often I find

A 96-vial sequence on a GC headspace tray lost 0.42 mg per vial per hour across an 18-hour run, and the analyst caught it only because the internal standard peak area had drifted 6.8% low by injection 71. The septa were the right durometer on paper. The crimper had been set to a 12 mm die the month before, and nobody re-checked the anvil gap after the die swap. That single mechanical detail is where most vial closures seal failure, evaporation and sample loss begins, and it almost never shows up as a visible leak. It shows up as a slow weight drift you can measure if you know where to look.

This is a field guide to diagnosing that drift. I am assuming you already run validated methods, know your vial format, and have a closure specification in hand. What follows is how to find the leak path, why a punctured septum never re-seals to its original value, how compression set quietly consumes your seal, and how to prove any of it with a balance and a clock.

Where the vapour actually escapes: four leak paths ranked by how often I find them

Every closure failure I have traced in fifteen years resolves to one of four paths. They have different signatures, different fixes, and very different costs.

Path 1: the crimp or thread interface (most common by far)

On a crimp vial, the aluminium seal must compress the septum against the glass land with enough radial and axial force to close the gap. A properly set 11 mm crimper produces a skirt that rolls under the bead with even pressure around the full circumference. When the anvil gap is too wide, you get a seal that looks crimped and holds liquid but leaks vapour. The failure is silent because the closure is still mechanically attached.

On screw-thread vials the equivalent failure is under-torque. A 9 mm short-thread vial with a PTFE/silicone septum needs roughly 1.5 to 2.0 in-lb to compress the liner into the land. Below about 1.2 in-lb the seal is marginal at room temperature and fails outright at 40 °C. Above 2.5 in-lb you start to deform the cap and, on some designs, extrude the septum.

Path 2: through the septum body (permeation and pinholes)

Silicone permeates water vapour and many small organics far faster than a butyl or PTFE-faced laminate. A 3 mm silicone disc on a 20 mL headspace vial will lose measurable mass to the atmosphere over a multi-hour run even with a perfect crimp. This is not a defect. It is the material doing what its datasheet says. If your method cannot tolerate it, you need a PTFE-faced septum or a laminate, not a thicker silicone one.

Path 3: the septum-to-cap bond

In bonded closures (common in 11 mm snap and some 13 mm crimp formats) the elastomer is glued or mechanically captured in the cap. Heat cycling and solvent attack can release that bond, and the septum lifts slightly under internal pressure. You will not see it until you pull the cap and find the disc sitting proud of its seat.

Path 4: the glass land itself

A chipped or out-of-round vial mouth will leak no matter what closure you put on it. This is rare in new glass but common in vials that have been through a wash cycle or a rough autosampler. Check the land with a loupe on any lot that suddenly underperforms.

Re-seal behaviour after puncture: what a needle actually does to your closure

This is the part most labs get wrong. A septum does not “self-heal” in any meaningful sense. It closes, but it closes to a different state than it started in.

A 23-gauge needle (0.64 mm OD) pushed through a 1.5 mm silicone/PTFE septum leaves a channel that, after withdrawal, relaxes to somewhere between 0.1 and 0.3 mm depending on durometer and how long the needle sat. The PTFE face does not flow back at all. It is a slit, not a hole, and it stays a slit. Every subsequent puncture widens it slightly and degrades the surrounding elastomer.

Practical numbers from my own bench testing on 11 mm crimp vials with 1.5 mm silicone/PTFE septa at 25 °C:

• First puncture: mass loss on a 10 mL aqueous fill over 24 h was under 1 mg, effectively the same as an unpunctured control.
• Tenth puncture: loss rose to roughly 4 to 6 mg over 24 h. Still fine for most methods.
• Thirtieth puncture: loss was 15 to 25 mg over 24 h, and the variance between vials widened sharply.

The takeaway: track puncture count per septum, and do not reuse a septum across a long autosampler queue just because the vial is clean. If your method has a queue that re-injects from the same vial more than about ten times, you are already in the regime where re-seal behaviour matters. This is one reason I recommend single-use closures for anything quantitative, and why the closure specification on vial closures should state a puncture rating rather than just a septum thickness.

Why the second puncture is worse than the first

The first needle pass creates a stress concentration at the slit tip. The second pass propagates that slit further into the elastomer rather than opening a fresh channel. By the fifth or sixth puncture the slit can run most of the way through the disc. If you are seeing a step change in loss after a specific injection number, this is the mechanism. The fix is not a thicker septum. It is fewer punctures per closure or a move to a pre-slit closure designed for the queue length.

Compression set: the slow failure that eats your seal over a run

Compression set is the permanent deformation an elastomer takes after being held under load at temperature. A septum that sits compressed at 40 °C for 18 hours will not return fully to its original thickness when the load is removed. It has taken a set. If your method holds vials at temperature and then cools them before injection, the closure is now looser than it was at the start.

The number that matters is the set value. A good silicone septum at 40 °C over 24 h should show under 15% set (ASTM D395, method B, 25% deflection). A cheap peroxide-cured silicone can hit 30% or more. Once set exceeds about 20%, the residual sealing force at the crimp interface drops below what is needed to hold against even modest internal pressure.

This is why a closure that passes a leak test on day one can fail on day three of a stability study. The seal did not “go bad.” It relaxed. I have seen this most often on headspace methods that hold the oven at 60 to 80 °C for extended equilibration. If your method does that, specify a platinum-cured silicone septum with a documented set value and check it against the run duration.

Temperature ratings you can actually rely on

Most closure datasheets quote a maximum continuous temperature. That number is usually the point at which the elastomer begins to degrade, not the point at which it still seals. For a practical working limit, derate by 20 °C. A septum rated to 200 °C continuous is a reliable seal to about 180 °C, and its compression set behaviour at that temperature over a multi-hour hold is what you should be asking the supplier for, not the headline rating.

Measuring loss gravimetrically: the only test that tells you the truth

You cannot see a vapour leak. You can weigh it. A gravimetric loss test is the definitive field check, and it takes an afternoon to set up.

The method

1. Fill a representative set of vials (n ≥ 10) with a volatile solvent or an aqueous standard to a known mass. Record the fill mass to 0.1 mg on a five-place balance.
2. Cap or crimp using the exact tooling and torque you use in production. Include a control set capped by a second operator or a second crimper.
3. Weigh the whole vial assembly (vial plus closure plus contents) and record. Do not open it again.
4. Place the vials under your actual method conditions: autosampler tray temperature, oven hold, whatever the run does.
5. Re-weigh at fixed intervals: 1 h, 4 h, 8 h, 24 h. The balance reading is the whole assembly, so any loss is vapour that left through the closure.
6. Plot mass loss per vial against time. A well-sealed vial loses a few tenths of a milligram over 24 h. A leaking one loses tens of milligrams, and the curve bends upward as the seal degrades.

Run this test whenever you change closure lot, crimper, or cap supplier. It is the cheapest early-warning system you have. If you want an independent read, a contract lab can run the same protocol alongside your own data, and the comparison is worth the money the first time. Analytical testing services that specialise in container-closure work will report loss rates with the same gravimetric method against a USP <661> reference, which gives you a defensible number for a deviation report. See analytical testing for the scope of what such a lab covers.

Reading the curve

A flat line at low loss means the closure is doing its job. A line that starts flat and then rises after 6 to 8 hours is compression set or a re-seal failure kicking in. A line that is steep from the start is a mechanical leak at the crimp or thread. The shape of the curve tells you which fix to apply, which is more useful than a single end-point number.

Field case 1: the 6.8% internal standard drift on a headspace queue

Symptom: A 96-vial headspace GC run showed the internal standard area drifting low from about injection 60 onward, ending 6.8% below the first injection. The analyst initially blamed the column.

Measurement: I ran the gravimetric test on the same lot of vials and closures under the tray conditions. The loss curve was flat for the first 5 hours and then climbed steeply, reaching 0.42 mg per vial per hour by hour 12. The control set, crimped on a second crimper with a verified anvil gap, stayed under 0.05 mg per vial per hour.

What it was: The production crimper had been fitted with a 12 mm die during a changeover and the anvil gap had never been re-set for the 11 mm closure. The skirt was rolling short, so the septum was under-compressed by roughly 0.15 mm. At room temperature that was marginal. At the 80 °C headspace oven it failed. The drift started when the queue reached the vials that had sat longest before injection.

Fix: Re-set the anvil gap to the die manufacturer’s spec, re-qualify with the gravimetric test, and add a crimp-height check (target 5.5 ± 0.2 mm for this closure) to the start-of-run checklist. Drift went to under 0.5% across the full queue.

Field case 2: the screw-thread cap that passed at 25 °C and failed at 40 °C

Symptom: A stability study on a volatile analyte showed recovery falling 12% by day 14, but only on the 40 °C arm. The 25 °C arm was clean.

Measurement: Gravimetric loss on 9 mm short-thread vials with PTFE/silicone liners, torqued by hand to what felt “tight.” Measured torque with a cap torque tester was 0.9 to 1.1 in-lb, below the 1.5 in-lb minimum for that liner. At 25 °C the loss was under 0.3 mg over 24 h. At 40 °C it was 2.1 mg over 24 h and rising.

What it was: Under-torque plus thermal expansion of the cap and liner. At 40 °C the differential expansion opened the seal interface just enough to leak. The hand-torque habit was the root cause, and it was inconsistent between operators (measured range 0.7 to 1.6 in-lb across three analysts).

Fix: Move to a calibrated torque driver set to 1.8 in-lb, retrain, and add a torque spot-check to the study setup. Recovery at 40 °C held within 2% through day 28. The closure was never the problem. The application was.

Troubleshooting checklist: match the symptom to the leak path

Use this when a run drifts and you suspect the closure. Work from the top down; the first three rows catch most cases.

Specifying a closure that will not lose your sample

When you write the specification, ask for four numbers beyond the usual dimensions. First, the compression set value at your method temperature over your run duration. Second, a puncture rating (how many punctures before loss exceeds a stated threshold). Third, the permeation rate for your analyte class through the septum material. Fourth, the recommended torque or crimp height with a tolerance.

If a supplier cannot give you the first and second, they are selling you a part, not a seal. The closures that hold up in my experience are the ones where the datasheet states a set value under ASTM D395 and a puncture count under a defined needle gauge. For HPLC work where the vial sits on an autosampler for hours, the same logic applies to the chromatography vial and its closure as a system, not as separate line items. And if you are running a method with a long queue, the HPLC vial closure you choose should be rated for the queue length, not just the injection volume.

For reference methods, the EPA SW-846 series is a useful cross-check on how the regulatory methods handle container and holding-time requirements, and it is worth reading the sample-handling sections against your own SOP. See U.S. EPA SW-846 for the current method set. Vendor technical libraries from Agilent, Thermo Fisher, and Restek also publish closure and septum compatibility tables that are more honest about permeation than most product datasheets.

The decision rule

If a run drifts, do not replace the column first. Run the gravimetric loss test on the actual lot and tooling under the actual method conditions, and read the curve shape before you touch anything else. A flat-then-rising curve means compression set or re-seal failure; a steep-from-start curve means a mechanical leak at the crimp or thread. Fix the tooling and torque before you change the closure, and specify compression set and puncture rating as numbers, not adjectives. The closure is a seal with a measurable loss rate, and once you start measuring it, the failures stop being mysterious.

References & further reading

Related resources from our lab consumables network:

• Vial Closures at vialclosures.com
• Analytical Testing at labanalyticaltesting.com
• Chromatography Vial at chromvials.com
• Hplc Vial at hplcvialfactory.com

External references and standards:

• Agilent chromatography consumables and columns
• Thermo Fisher chromatography and sample handling
• U.S. EPA SW-846 test methods for waste and materials
• Restek technical resources for chromatography