A 96-position autosampler tray lost 0.42 mg per vial per 24 hours on a validation run last March, which sounds trivial until you multiply it out: 0.42 mg over 24 hours from a 1.2 mL fill is a 0.035% per-day loss, and over a 72-hour queue that is 0.1% of the injected mass gone before the needle ever touches the septum. On a 0.1% area normalization limit, that is the whole tolerance. The customer had blamed the HPLC, then the column, then the integrator. The actual cause was a 2ml vial seal failure at the crimp, not the cap. This article is about tracing evaporation and sample loss back to the vial, measuring it properly, and fixing it without changing the whole consumables line.

Every other method for judging seal integrity is a proxy. Visual inspection catches gross defects and nothing else. Dye ingress catches liquid leaks but misses vapour loss, which is the failure mode that actually matters for volatile analytes. Headspace pressure decay works on the production line but tells you almost nothing about a 40 °C autosampler tray over three days.

Gravimetric loss is direct. You weigh a filled, sealed vial, put it in the environment it will actually see, and weigh it again. The number you get is the sum of every leak path in the assembly. If the loss is above your threshold, the vial is failing, and you can then work backwards to which path is responsible.

The threshold depends on your analytes. For an aqueous sample with a non-volatile target, 0.5 mg/day is generous. For a methanolic extract with a volatile internal standard, 0.1 mg/day is the working ceiling I use, and I want to see it on a batch of 30 vials, not one.

How to run the measurement so the number means something

The procedure is simple but the details decide whether the result is real.

• Use a balance readable to 0.01 mg (a five-place analytical balance). A four-place balance cannot resolve daily loss on a single vial.
• Fill to a fixed, recorded volume. 1.0 mL is the standard working fill for a 2 mL vial; 1.2 mL leaves less headspace and loses less, so pick one and stick to it across the batch.
• Seal with the same crimper or screw capper you use in production, at the same setting. A hand crimper calibrated by feel is not reproducible.
• Tare the balance with a matched empty sealed vial, or weigh the whole tray in a fixed position to cancel drift.
• Condition at the temperature the tray actually runs at. Most autosamplers hold 4 to 40 °C; the 40 °C end is where seal failures show up first.
• Weigh at 0, 24, 48, and 72 hours. A single time point cannot distinguish a steady leak from a one-off loss during sealing.
• Run at least 30 vials per lot and report the mean and the 95th percentile, not just the mean. A lot with a good mean and a bad tail is a lot that will fail on position 78.

The output is a loss rate in mg/day per vial. Divide by the fill mass to get a percentage, and compare that to the stability acceptance criterion for your method. If your method allows 2% loss of a volatile standard over a 24-hour run, and your vial loses 0.3% per day, you have a 6x margin. If it loses 1.5% per day, you do not have a method, you have a race.

Where the vapour actually escapes: four leak paths, ranked by frequency

I have torn down enough failed vials to know that the leak is almost never where the operator thinks. Ranked by how often I find them:

1. The crimp-to-neck seal (crimped vials, aluminium seal)

This is the dominant failure mode on crimped 11 mm vials. The aluminium seal is deformed over the neck bead by the crimper jaws. If the crimper is set too loose, the seal sits on the bead without compressing the septum against the neck land, and vapour escapes around the circumference. If it is set too tight, the aluminium tears or the neck cracks.

The working parameter is crimp diameter. A standard 11 mm crimp seal on a 2 mL vial should measure 10.5 to 10.8 mm across the crimp skirt after crimping, measured with a crimp gauge, not by eye. Below 10.5 mm the seal is under-crimped; above 10.9 mm you are crushing the neck. Check the gauge on the first five vials of every shift and after every crimper jaw change. I have seen a single worn jaw produce a 0.6 mg/day loss on one side of a 96-position tray while the other side read 0.05 mg/day, and the operator never noticed because the tray averaged out.

2. The septum-to-land interface (both crimped and screw)

The septum is a disc of silicone or silicone/PTFE laminate, typically 1.0 mm thick for a standard 11 mm septum. It seals by compression against the neck land. The compression comes from the crimp (crimped vials) or from the cap torque (screw vials). If the compression is insufficient, or if the septum has taken a permanent set, the seal leaks even though nothing looks wrong.

This is where compression set enters. Silicone has excellent re-seal behaviour after puncture but mediocre compression set at elevated temperature. Hold a silicone septum at 40 °C under compression for 72 hours and it will not return fully to its original thickness when the compression is released. The loss of thickness is the loss of sealing force. A septum that measured 1.00 mm when new may measure 0.94 mm after a hot run, and that 0.06 mm is the difference between a seal and a slow bleed.

3. The puncture hole itself

A 22-gauge needle (0.72 mm OD) leaves a hole in a silicone septum. Silicone re-seals well; the hole closes to near-zero when the needle withdraws. But “near-zero” is not zero, and after multiple punctures the re-seal degrades. The standard claim is 10 to 20 punctures per septum for a silicone/PTFE laminate; in practice, I treat 5 punctures as the point where I re-measure loss, and I have seen loss rates triple between puncture 5 and puncture 10 on a hot tray. PTFE-faced septa re-seal less well than pure silicone because the PTFE layer does not flow; the silicone underneath does the sealing, and the PTFE face is there for chemical inertness, not for re-seal.

4. The cap-to-vial thread (screw vials)

Screw-thread 9 mm vials seal on a PTFE-lined or silicone-lined cap that compresses against the neck. The thread itself does not seal; the liner does. The failure mode is under-torque, which leaves the liner uncompressed, or over-torque, which deforms the liner and can crack the neck. The working torque for a 9 mm screw cap on a 2 mL vial is 0.6 to 0.9 N·m (roughly 5 to 8 inch-ounces). Below 0.5 N·m the liner does not seal; above 1.0 N·m you risk neck cracking and liner extrusion. A torque screwdriver is not optional if you care about this number.

Re-seal behaviour after puncture: what the datasheet does not tell you

The puncture-and-re-seal story is where a lot of seal failures hide, because the first injection is fine and the tenth is not.

When a needle punctures a silicone septum, the elastomer is displaced, not cut. On withdrawal, the elastic recovery closes the hole. The recovery is fast (milliseconds) but not complete. Residual strain remains, and the hole is a small conical channel rather than a perfect closure. Over repeated punctures in the same spot, the silicone fatigues and the channel widens. The loss rate through a single puncture is small; through a fatigued puncture it is not.

Three things make this worse:

• Needle gauge. A 22-gauge needle leaves a smaller hole than a 20-gauge. If your autosampler uses a 20-gauge (0.91 mm OD), expect faster re-seal degradation.
• Temperature. Silicone softens at elevated temperature, so the same puncture at 40 °C recovers more slowly and less completely than at 4 °C.
• Needle dwell time and speed. A slow needle withdrawal gives the elastomer time to recover; a fast one can tear the hole open.

I measure re-seal by running a vial through 1, 5, 10, and 20 punctures on a hot tray and weighing at each point. The curve is usually flat to 5 punctures, then rises. The number where it rises is the puncture budget for that septum and that tray temperature. For a standard silicone/PTFE septum at 40 °C with a 22-gauge needle, I use 5 punctures as the working limit and replace the vial after that. For a 4 °C tray, 10 is defensible. For a 20-gauge needle, cut both numbers in half.

Compression set is the slow failure that shows up as drift, not as a leak

Compression set does not usually produce a dramatic leak. It produces a slow, monotonic weight loss that starts after a few hours and accelerates over the run. That is why it is so often misdiagnosed as instrument drift. The chromatogram shows declining peak areas across the sequence, the operator blames the detector, and the actual cause is a septum that has taken a set and stopped sealing.

The mechanism: the septum is compressed between the cap and the neck land. At temperature, the elastomer relaxes and the sealing force drops. The sealing force is what keeps the vapour in. When it drops below the vapour pressure of the headspace, loss begins. The loss is not a leak in the sense of a hole; it is diffusion through and around a compromised seal.

The practical fix is material selection and temperature matching, not torque. A silicone septum is fine to about 40 °C for a 72-hour run. Above that, or for longer runs, use a laminate with a higher-temperature silicone or a butyl-based septum where chemical compatibility allows. Butyl has better compression set than silicone but worse re-seal and worse chemical resistance, so it is a trade, not an upgrade. Read the septum datasheet for the compression set figure at your temperature; if the supplier cannot give you a number, that is the number.

This is also where the difference between a generic vial and a specified one shows up. If you are sourcing through a 2ml vial supplier who can give you the septum material, thickness tolerance, and compression set data, you can predict the failure. If you cannot get those numbers, you are buying a lottery ticket. For the broader consumables picture around these vials, the HPLC consumables catalog and the chromatography supplier resources are worth reading before you lock a part number.

Two field cases: the symptom, the measurement, the fix

Case 1: the 72-hour tray that lost 8% of its internal standard

Symptom: A contract lab running a 72-hour sequence on an autosampler held at 40 °C saw the internal standard peak area fall 8% from the first calibration to the last sample. The method had a 3% acceptance criterion. The instrument had been serviced two weeks earlier.

Measurement: I ran a gravimetric loss test on the same lot of vials, filled to 1.0 mL with the same matrix, sealed with the lab’s crimper, and held at 40 °C. The mean loss was 0.38 mg/day, but the 95th percentile was 1.1 mg/day, and the crimp diameter on the high-loss vials measured 10.3 to 10.4 mm against a target of 10.5 to 10.8 mm. The crimper jaw had worn. The high-loss vials were all on the same side of the tray, which the lab had not noticed because they were looking at the sequence average.

Fix: Replaced the crimper jaw, re-gauged the first five vials of every shift, and re-ran the loss test. The 95th percentile dropped to 0.22 mg/day. The internal standard drift over 72 hours fell to 1.1%, inside the 3% criterion. The vials were not the problem; the crimp was. But the vial was where the loss showed up, which is why the lab had blamed the vial in the first place.

Case 2: the method that failed only on the last 20 positions

Symptom: A pharmaceutical QC lab running a 24-hour stability sequence on screw-thread 9 mm vials saw the last 20 positions fail the assay acceptance criterion. The first 76 positions were fine. The method had been validated on the same vials a year earlier.

Measurement: Gravimetric loss on the same lot, filled to 1.0 mL, capped with the lab’s torque screwdriver. Mean loss was 0.09 mg/day, well inside the limit. But when I measured the cap torque on the failing vials, it was 0.35 to 0.45 N·m, against a target of 0.6 to 0.9 N·m. The torque screwdriver had never been calibrated. The last 20 positions were the ones the analyst had capped at the end of a long shift, when the screwdriver had drifted and the analyst had stopped checking.

Fix: Calibrated the torque screwdriver against a torque tester, set it to 0.75 N·m, and added a torque check on the first and last vial of every batch. The loss test on the next lot showed a mean of 0.07 mg/day and a 95th percentile of 0.11 mg/day. The assay failures stopped. The lesson: a screw cap is only as good as the torque, and the torque is only as good as the calibration.

Decision table: what to change first when loss exceeds your limit

Work down this table in order. Each row is a diagnostic, not a guess.

What to put in the purchase specification so this stops happening

Most seal failures are specification failures. The vial was bought on price and dimensions, and nobody wrote down the septum material, the compression set, or the crimp target. Fix the specification and the field failures drop.

Put these on the PO:

• Septum material and thickness, with tolerance. A 1.0 mm silicone/PTFE laminate at ±0.05 mm is a specification; “silicone septum” is not.
• Compression set at your tray temperature, with a test method reference. If the supplier cannot supply it, that is a red flag.
• Crimp diameter target for crimped vials (10.5 to 10.8 mm) and torque target for screw vials (0.6 to 0.9 N·m), stated on the certificate.
• GPI thread finish for screw vials, so the cap and vial are dimensionally matched. A mismatched thread is a leak path that no torque will fix.
• Gravimetric loss limit at a stated temperature and fill volume, measured on a stated sample size. This is the only line on the spec sheet that predicts field behaviour.
• Lot traceability back to the septum lot, not just the vial lot. Septum lots vary more than glass lots.

If you are buying through an OEM vial channel, these numbers should be in the drawing, not in a conversation. And if you are validating a method against a regulatory framework, the seal integrity data belongs in the method validation package. The U.S. EPA SW-846 test methods and the vendor technical resources from Agilent, Thermo Fisher, and Restek all assume the sample is stable in the vial for the duration of the run. If your gravimetric data says it is not, the method is not validated regardless of what the chromatogram looks like.

The decision rule

If you cannot state the gravimetric loss rate of your 2ml vial at your tray temperature and fill volume, you do not know whether your method is stable over the run, and no amount of instrument servicing will fix a seal that is losing 0.4 mg/day. Measure the loss on 30 vials at 0, 24, and 72 hours; if the 95th percentile exceeds 0.2 mg/day at 40 °C, change the crimp or torque setting first, the septum material second, and the vial supplier last. The vial is usually the messenger, not the cause.

References & further reading

Related resources from our lab consumables network:

• 2Ml Vial at 2mlsamplevial.com
• Hplc Consumables at hplcconsumable.com
• Chromatography Supplier at chromatographysupplier.com
• Oem Vial at oemsamplevials.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