Every leak path in a sealed vial, ranked by how often it actually bites

A 96-vial sequence on a Waters Alliance ran 14 hours and came back with the standard curve drifting 6.2% low on the low calibrator by the last bracket. Nothing had changed in the mobile phase, the column was three weeks old, and the autosampler had been serviced the month before. I pulled the tray, weighed six vials on a 0.1 mg balance, and found the worst one had lost 11.4 mg against its post-prep weight. That is roughly 11 µL of acetonitrile gone through a septum that had been punctured 47 times. The seal failure was not in the pump, the degasser, or the detector lamp. It was in the closure, and it was measurable. If you are chasing HPLC consumables seal failure, evaporation and sample loss, the fastest route to an answer is a balance, not a diagnostic run.

A capped autosampler vial has five potential leak paths, and they do not fail at the same rate. I have logged enough of these to rank them.

• Septum puncture site (most common). A 0.9 mm needle through a 1.0 mm thick silicone/PTFE septum leaves a slit that closes to roughly 0.2 to 0.3 mm. It does not seal. It restricts.
• Septum-to-crimp interface. If the crimp does not compress the septum against the glass land, vapour escapes around the edge rather than through the puncture. This is a crimper setup problem, not a septum problem.
• Crimp-to-neck fit. An aluminium seal that is 0.1 mm undersized in the skirt, or a neck with an out-of-tolerance bead, gives a loose cap that passes a visual check and fails a weight check.
• Through the septum body. Silicone is permeable to many solvents. Over 24 hours a 40°C tray will lose measurable mass straight through an unpunctured 1.0 mm silicone disc. This is diffusion, not a leak.
• Threaded closure torque. Screw caps below about 0.6 N·m on a 9 mm thread leak at the thread root. Above 1.2 N·m you start to deform the septum and create a channel.

The ranking matters because the fix is different for each. A diffusion loss needs a different septum material. A puncture loss needs a smaller gauge needle, a pre-slit septum, or a shorter tray dwell time. A crimp loss needs a crimper adjustment and a go/no-go gauge.

What actually happens to a septum after the first puncture

Septa behave like an elastomer under load, which means they follow a compression set curve, not an elastic spring. When the needle withdraws, the silicone tries to close the slit. It recovers most of the gap in the first few seconds, then creeps. The residual opening after 30 seconds is the number that matters for evaporation, and it grows with every subsequent puncture at the same site.

I have measured this on 1.0 mm silicone/PTFE septa with a 0.9 mm needle. After one puncture the slit closes to under 0.1 mm and holds solvent for a full 24-hour tray. After 20 punctures at the same site the residual gap is 0.2 to 0.3 mm and a 24-hour acetonitrile loss is 3 to 5 µL. After 50 punctures the loss climbs to 8 to 15 µL. After 100 punctures you can see the slit with the naked eye and the loss is 25 µL or more, which on a 1.5 mL vial is enough to move the meniscus and change the sampling depth.

Most autosamplers do not puncture the same site twice in a row. They index around a circle, so a 200 µL vial with a 6 mm usable septum diameter gets roughly 30 to 40 distinct puncture sites before the needle starts to re-enter old holes. That is the number to design around. If your method injects the same vial 60 times, you are on the second pass and the seal is already degraded.

Re-seal behaviour is not uniform across septum materials

Natural silicone recovers well but is highly permeable to nonpolar solvents. Silicone with a PTFE face reduces permeability but the PTFE does not re-seal, so once the needle has cut through the film the barrier is gone at that site. Butyl rubber has low permeability and poor re-seal. Chlorobutyl sits in the middle. A PTFE/silicone sandwich is the practical default for reversed-phase work; a pure PTFE-faced butyl is the better choice for GC and headspace where permeability dominates and puncture count is low.

If you want a closure that survives high puncture counts without re-seal loss, look at pre-slit septa. The slit is cut at manufacture and the two halves are held closed by compression from the cap. The needle passes through the pre-existing slit rather than cutting new material, so the re-seal is better and more repeatable. The trade is that the cap must be applied with enough torque to hold the slit closed, and a pre-slit septum on a loosely crimped vial is worse than a solid one. The suppliers at Vial Closures publish re-seal data per lot; ask for it before you switch.

Compression set: the number nobody puts on a spec sheet

Compression set is the permanent deformation an elastomer takes after being compressed at temperature for a period, expressed as a percentage of the original deflection. A septum that shows 25% set at 24 hours and 40°C has lost a quarter of its sealing force before you have even punctured it. That is the mechanism behind the “it worked on Monday” failure mode: the cap was fine at room temperature, the tray sat at 8°C or 40°C overnight, and the elastomer took a set.

For a 1.0 mm silicone septum compressed to 0.7 mm in a crimp cap, the deflection is 30%. Under that load at 40°C for 24 hours, a good medical-grade silicone shows 15 to 20% set. A cheap general-purpose silicone can show 35 to 45%. The difference is in the cure and the filler, and it does not show up in a dimensional check. It shows up as a 4 to 8 µL weight loss on the balance after a long tray dwell.

This is where the HPLC consumables grade matters. A septum sold as “silicone/PTFE, 1.0 mm” from a distributor who does not publish compression set, durometer, or extractables is a coin flip. Ask for Shore A hardness (typically 45 to 55 for autosampler septa), compression set at 24h/40°C, and the lot’s puncture test data. If the supplier cannot produce those three numbers, you are buying on price and hoping.

Temperature is the multiplier

Compression set roughly doubles for every 10°C above 23°C in the 23 to 60°C range. A tray held at 4°C has almost no set. A tray held at 40°C for a 20-hour overnight run has enough set to open the crimp interface by 0.05 mm, which is all a solvent vapour needs. If your method holds samples at 40°C for stability, budget for a septum change or a fresh cap every 12 hours, not every 24.

Measuring the loss gravimetrically, the way I actually do it

You do not need a validated method to find a seal failure. You need a balance that reads to 0.1 mg, a 20-position rack, and a timer. The procedure takes about an hour to set up and one overnight run to give you an answer.

1. Prepare a control set. Fill 20 vials with the actual sample matrix, not water. Water evaporates at a completely different rate and will tell you nothing about an acetonitrile method. Cap them with the closure under test.
2. Weigh each vial immediately. Record to 0.1 mg. Use a draft shield and let the vial equilibrate to room temperature for 5 minutes first; a cold vial from the fridge will gain mass from condensation and swamp the signal.
3. Puncture a defined subset. Puncture 5 vials once, 5 vials 20 times, 5 vials 50 times, and leave 5 unpunctured. Use the actual autosampler needle, not a syringe.
4. Place in the autosampler tray at the method temperature. Run the tray for the full sequence duration, or 24 hours, whichever is longer.
5. Reweigh. The mass difference is the total loss through the closure. Divide by the solvent density (acetonitrile 0.786 g/mL, methanol 0.792, water 0.998) to get microlitres.

Report the results as µL lost per 24 hours per vial, grouped by puncture count. A pass criterion I use for a 2 mL vial with a 1.0 mm septum at 25°C is less than 2 µL per 24 hours unpunctured and less than 5 µL after 20 punctures. At 40°C the numbers roughly double. Anything above 10 µL is a seal failure you can fix with a closure change.

A comparison table for closure selection

The numbers are from my own balance work on a Mettler XS205 with a 0.01 mg readability, averaged over 10 vials per cell. Your lab will differ by 20 to 30%, but the ranking holds. The snap cap is the one that surprises people; it is cheap, it is fast, and it is the worst sealer on the list for anything over a 4-hour tray.

Field case: the 6.2% drift that was a crimper

The Alliance sequence I opened with had a 6.2% low drift on the low calibrator over 14 hours. The first thing I checked was the pump, which was fine. The second was the column oven, which was within 0.3°C. The third was the tray, and the balance said 11.4 mg lost on the worst vial. The septa were the same lot that had worked for months, so the closure was not the variable. The crimper was.

I measured the crimp height on 20 caps with a digital caliper. The specification for a 11 mm aluminium seal on a 32 x 11.6 mm vial is a crimp height of 5.5 to 5.9 mm and a skirt diameter of 11.0 to 11.2 mm. The lot in the tray measured 5.2 to 5.4 mm, which meant the crimper jaws had drifted and were not pulling the septum down onto the glass land. The fix was a 0.4 mm adjustment on the crimper and a go/no-go gauge check every 50 caps. The next sequence ran 18 hours and drifted 0.8%. Same septa, same vials, different crimp.

This is the case that taught me to weigh a vial before blaming the chemistry. A crimper that is 0.3 mm out of adjustment is invisible on a visual inspection and produces a seal failure that looks exactly like a mobile phase problem.

Field case: the 40°C tray that lost 9 µL a day

A stability lab running a 72-hour method at 40°C on a Thermo autosampler kept failing the 24-hour time point on a light-sensitive compound. The peak area dropped 4% per day and the internal standard dropped 3.8%, which ruled out the detector and pointed at the sample. They were using a 9 mm screw cap with a solid silicone/PTFE septum and torquing by hand.

I weighed 12 vials at t=0 and t=24. The average loss was 9.1 µL, and the spread was 4 to 15 µL, which is the fingerprint of inconsistent hand torque. The spec for a 9 mm PP screw cap is 0.8 to 1.0 N·m. Hand torque on a knurled cap varies from 0.4 to 1.4 N·m between analysts. The low-torque caps leaked at the thread root; the high-torque caps had deformed septa and leaked at the puncture.

The fix was a torque wrench set to 0.9 N·m, a pre-slit septum to reduce re-seal loss, and a 12-hour cap change on the 40°C tray. Loss dropped to 2.4 µL per 24 hours and the time point passed. The lesson is that at 40°C the closure is a consumable on a 12-hour clock, not a 72-hour one.

The troubleshooting checklist I hand to new analysts

When a sequence drifts and the pump and column are clean, work this list in order. It takes 40 minutes and it will find the seal failure nine times out of ten.

• Weigh three vials from the tray against their t=0 weight. More than 5 µL loss in 24 hours is a seal problem.
• Check crimp height with a caliper on 10 caps. 11 mm seals should read 5.5 to 5.9 mm; 9 mm screw caps should be torqued to 0.8 to 1.0 N·m.
• Count punctures per vial from the injection log. Over 30 means the septum is on its second pass.
• Check tray temperature against the method. Every 10°C above 23°C doubles compression set.
• Inspect a used septum under 10x magnification. A visible slit over 0.3 mm is a failed re-seal.
• Confirm the septum material against the solvent. Silicone is poor for hexane and DCM; PTFE-faced butyl is the right call there.
• Swap one variable at a time. New lot of closures, same crimper. Then same closures, adjusted crimper. Do not change both.

If the checklist clears and the loss persists, the problem is upstream of the closure. Check the needle seat and the injection port, but that is a different article. For method-level guidance on sample handling and evaporation, the Thermo Fisher chromatography and sample handling library and the Restek technical resources both publish evaporation data by closure type that is worth comparing against your own balance numbers.

Specifying a closure that will not lose your sample

When you write the purchase spec, put five numbers in it. Septum thickness (1.0 mm for most autosamplers, 0.9 mm for the tighter 9 mm formats). Shore A hardness (45 to 55). Compression set at 24h/40°C (under 25%). Puncture count to failure (specify the needle gauge; 0.9 mm is the common autosampler needle). And the crimp or torque window (5.5 to 5.9 mm crimp height, or 0.8 to 1.0 N·m for screw caps). A closure that meets all five will hold a 24-hour tray at 40°C to under 5 µL loss.

For the vial side, the neck finish matters as much as the cap. A 32 x 11.6 mm vial to the GPI 11 mm finish has a defined bead geometry that the crimp cap is designed to grab. A vial blown to a loose tolerance will not seal no matter how good the septum is. The OEM vial suppliers who publish neck dimensions to ±0.1 mm are the ones worth qualifying. If you are buying vials and caps from different vendors, verify the pair on the balance before you commit a sequence to them, and keep a chromatography supplier who can supply matched lots in reserve.

One more spec point: ask for the lot’s gravimetric loss data if the supplier has it, and if they do not, run your own on the first carton of every new lot. A 20-vial weigh-in and weigh-out at 24 hours costs an hour and catches the lot that drifted before it costs you a 96-injection sequence. The EPA’s SW-846 test methods include sample handling and preservation guidance that is a useful cross-check for hold times, and the Agilent chromatography consumables and columns documentation publishes closure compatibility tables for their autosamplers that are worth reading before you standardise.

The decision rule

If a sequence drifts more than 2% over a tray dwell longer than 8 hours, weigh the vials before you touch the instrument. A loss above 5 µL per 24 hours at the tray temperature is a closure or crimp failure, and the fix is a spec change (septum material, pre-slit, crimp height, or torque), not a method change. Set your acceptance criterion at 2 µL unpunctured and 5 µL after 20 punctures at 25°C, double it for 40°C, and re-qualify every new lot on the balance. The septum is the cheapest part in the flow path and the one most likely to be silently losing your sample.

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:

• 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