A 96-vial sequence on a split/splitless GC inlet lost 11.4% of its internal standard peak area between injection 1 and injection 96, with no drift in the analyte-to-IS ratio and no change in retention time. The column was fine. The inlet liner was fresh. The autosampler syringe had 400 injections on it. The cause was a batch of 11 mm septa with a 0.9 mm molded center that had been sitting on a shelf for 19 months: they sealed on the first puncture and leaked on the next 95. That is a vial septa seal failure, and it shows up as evaporation and sample loss long before anyone suspects the cap.
What follows is how I trace that loss, what the leak paths actually are, why a septum that re-seals perfectly on injection 1 can fail on injection 40, and how to measure the loss gravimetrically so you stop guessing. Coring debris and its contamination peaks belong to a different article; here the only question is whether the seal holds.
Where the sample actually escapes: the four leak paths on a crimp or screw vial
Every septa leak I have chased resolves into one of four paths. They have different signatures, and confusing them wastes weeks.
Path 1: the puncture itself
A 23-gauge autosampler needle is roughly 0.64 mm OD. It punches a hole that the elastomer must close by elastic recovery. On a good silicone/PTFE septum, recovery is fast (sub-second) and the residual channel is under 0.05 mm. On a degraded or under-cured septum, the channel stays open at 0.1 to 0.2 mm and stays open. That is enough to lose a volatile analyte at 4 to 8 °C in the autosampler tray over a two-hour run.
Path 2: the septa-to-cap interface
This is the one people miss. The septum sits in the cap well and is compressed by the crimp or the screw torque. If the septum OD is undersized by more than about 0.2 mm relative to the cap well, or if the well has flash from a worn mold, the seal is at the perimeter, not the face. You get a slow annular leak that is invisible to the eye and constant in rate.
Path 3: the cap-to-vial interface
For crimp vials, this is the crimp itself. A crimp that is too loose (below about 2.5 N pull-off force on a 11 mm aluminum seal) lets the septum lift under inlet pressure. Too tight and you crush the septum and induce compression set before the first injection.
Path 4: permeation through the elastomer
Real but small. Silicone permeates water vapor and small hydrocarbons at measurable rates; PTFE-faced septa block the face but not the edge. For most HPLC work, permeation is negligible next to paths 1 to 3. For headspace and long-stored GC extracts, it matters more than people admit.
If you want a baseline on the components themselves, the vial septa reference set is where I start before I touch the instrument.
Re-seal behaviour after puncture is a property of the compound, not the durometer
The single most useful thing I learned in fifteen years of this is that re-seal is not a hardness number. Two septa at 45 Shore A can differ by a factor of five in re-seal performance. What controls it is the compression set of the cure system and the elastic memory of the crosslink density.
Here is what actually happens on puncture. The needle displaces elastomer radially. When it withdraws, the material recovers toward its original geometry. If the cure is complete and the crosslink density is in the right window, recovery is 95%+ within a second and the residual channel is sub-0.05 mm. If the cure is incomplete (under-cured), the polymer chains have not fully networked and recovery is slow and partial. If the cure is overdone, the material is brittle and the puncture leaves a permanent slit.
The practical test is not a durometer reading. It is a puncture-and-hold test: insert a needle, hold 5 seconds, withdraw, and measure the residual channel under a 10x loupe or a profile projector. On a good lot, the channel closes to under 0.05 mm in under 2 seconds. On a bad lot, it stays at 0.15 mm and you can see it.
I also track re-seal cycles: how many punctures a septum takes before the residual channel exceeds 0.1 mm. For a standard 11 mm silicone/PTFE septum, that number is typically 3 to 5 punctures at the same site. For a pre-slit septum, it is effectively 1 (the slit is already there). For a high-quality molded-center septum, I have seen 8 to 10 before failure. That number is the one to ask your supplier for, and most will not have it.
Why the same septum that held on injection 1 leaks on injection 40
Because the puncture is not the only thing that happens. Each injection adds a cycle of compression and decompression at the cap interface. Over 40 injections, a septum with poor compression set takes a permanent deformation of 0.1 to 0.3 mm at the seal face, and the cap-to-septum seal goes from face contact to edge contact. The loss rate jumps. This is why a run can look clean for the first hour and drift in the second.
Compression set is the number that predicts evaporation loss, and almost nobody measures it
Compression set is the permanent deformation remaining after a compressive load is removed. ASTM D395 covers it. For a vial septum, the relevant condition is 70 hours at 100 °C under 25% deflection, and a good silicone septum comes back to under 20% set. A cheap one comes back at 35 to 45%, which means it has lost a third of its sealing force before it ever sees a sample.
Why does that matter for evaporation? Because the sealing force is what holds the septum against the cap well and the vial lip. Once the set exceeds roughly 30%, the residual force at the seal face drops below what is needed to resist the vapor pressure of the headspace, and you get a slow, steady loss that looks exactly like a sample that is degrading.
I keep a simple rule: if the compression set is over 30%, the septum is a storage liability, not a seal. If it is under 20%, it will hold through a 100-injection sequence at 4 °C with under 2% loss of a volatile IS. Anything in between is a coin flip and should be qualified on your own method before it goes into production.
Temperature and the set curve
Compression set is temperature-dependent, and the curve is not linear. At 4 °C (autosampler tray), set is minimal over a run. At 40 °C (headspace oven or a warm room), set accumulates faster. At 100 °C (the ASTM condition), you see the true material property. If your method runs the tray cold, you can get away with a 25 to 30% set septum for a single sequence. If your method stores extracts overnight in the tray, you cannot.
How to measure evaporation loss gravimetrically over a run
This is the part that separates a real diagnosis from a hunch. You can measure the loss directly, and you should, because the instrument will not tell you.
The basic method
- Fill 12 vials with a volatile solvent mix (I use 50:50 methanol:water with 0.1% formic acid, or a 100 µg/mL acetone in water standard for GC work).
- Cap with the septum under test, using the same crimp or torque as production.
- Weigh each vial to 0.1 mg on an analytical balance. Record the tare.
- Place in the autosampler tray at the method temperature (4 °C or ambient).
- Weigh at t = 0, 1, 2, 4, 8, and 24 hours. Do not open the vials.
- Plot mass loss versus time. The slope is your evaporation rate in mg/hour/vial.
A good crimp seal on a 2 mL vial with a 11 mm silicone/PTFE septum loses under 0.5 mg over 24 hours at 4 °C. A failing seal loses 3 to 15 mg over the same period. The difference is not subtle. If you see more than 2 mg over 24 hours, you have a seal problem, full stop.
What the curve shape tells you
- Linear loss from t=0: a perimeter leak (path 2 or 3). The seal never held.
- Flat then rising: compression set taking hold. The seal held, then relaxed. Classic for a poor cure system.
- Step at each injection: puncture recovery is failing. The loss happens at the needle, not the cap.
- High initial loss then flat: usually a wet septum or residual solvent in the elastomer, not a leak. Bake it out and re-test.
I run this test on every new lot before it goes into a validated method. It takes a day and a balance, and it has caught three bad lots in five years that would otherwise have cost me a re-run of a 96-sample sequence.
Cross-check with the chromatogram
Gravimetric loss is the ground truth, but the chromatogram is the audit trail. A seal failure shows up as a downward drift in the internal standard peak area with a stable analyte/IS ratio. If the ratio drifts too, you have a different problem (degradation, adsorption, or injection volume). If only the IS drifts, suspect the seal. That single diagnostic has saved me from replacing a syringe more than once.
Field case 1: the 19-month-old lot that passed incoming inspection
Symptom: a 96-vial GC sequence lost 11.4% of the IS peak area over the run. Retention times were stable. The analyte/IS ratio was flat.
Measurement: I pulled 12 vials from the same lot and ran the gravimetric test. Loss was 6.8 mg over 24 hours at 4 °C, against a spec of under 0.5 mg. Compression set on a retained sample came back at 41%. The lot had been manufactured 19 months earlier and stored in a warehouse that hit 38 °C in summer.
Fix: switched to a lot with a documented compression set under 20% and a date code under 12 months old. Re-ran the gravimetric test: 0.3 mg over 24 hours. The 96-vial sequence came back to within 1.2% IS drift. The lesson is that incoming inspection on a septum has to include compression set and a date code, not just a visual and a dimension check.
Field case 2: the crimp that was 0.4 mm too deep
Symptom: a headspace method on a 20 mL vial showed 8 to 12% loss of a volatile analyte between duplicate vials, but only on the right-hand side of the tray.
Measurement: I measured crimp height with a crimp gauge on 20 vials from each side of the tray. The right side averaged 0.4 mm deeper than the left. The crimper jaw had worn on one side. Deep crimps crushed the septum at the seal face, induced set before the run, and the seal relaxed in the oven.
Fix: replaced the crimper jaw, set the crimp height to the manufacturer’s spec (for a 20 mm crimp, typically 5.5 to 5.7 mm finished height, checked with a go/no-go gauge), and re-qualified. Duplicate RSD dropped from 9% to 1.8%.
For headspace and GC work, the crimp is part of the seal, and the HPLC vial and autosampler vial selection has to be matched to the crimper, not the other way around.
Torque and crimp settings: the numbers that hold a seal
For screw-thread vials, the seal is a function of torque, and the window is narrower than most people think. Below the low end, the septum lifts under inlet pressure. Above the high end, you crush the septum and induce set. Both fail, and they fail differently.
| Vial type | Seal method | Target setting | Failure below | Failure above |
|---|---|---|---|---|
| 2 mL screw, 9 mm | Torque | 0.8 to 1.2 N·m | Lifts under pressure, linear loss | Crushed septum, set at seal face |
| 2 mL crimp, 11 mm | Crimp height | 5.5 to 5.7 mm finished | Loose seal, perimeter leak | Crushed septum, set before run |
| 20 mL headspace, 20 mm | Crimp height | Per manufacturer gauge | Loose seal, headspace loss | Crushed septum, oven relaxation |
| 4 mL screw, 13 mm | Torque | 1.0 to 1.5 N·m | Lifts under pressure | Set at seal face |
These are starting points, not gospel. Verify with a torque tester and a crimp gauge on your own hardware. A calibrated torque screwdriver costs less than one re-run of a failed sequence.
A troubleshooting checklist for seal failure and evaporation loss
When a run drifts, work this list in order. It is ordered by frequency, not by ease.
- Gravimetric loss test on the lot in use. Under 0.5 mg/24 h at 4 °C is good. Over 2 mg is a fail.
- Compression set on a retained sample. Under 20% is good. Over 30% is a storage liability.
- Puncture recovery under 10x loupe. Residual channel under 0.05 mm within 2 seconds is good.
- Crimp height or torque on 10 vials from the run. Check against the gauge or torque spec.
- Septum OD vs cap well. Undersize by more than 0.2 mm is a perimeter leak risk.
- Lot date code. Over 12 months old is a risk; over 24 months is a fail for volatile work.
- Storage temperature. A warehouse that hits 38 °C will set a septum before it ships.
- Tray temperature. A 4 °C tray buys you time; a 25 °C tray does not.
- IS drift vs analyte/IS ratio. IS-only drift points to the seal. Ratio drift points elsewhere.
If you want the broader context on method requirements and how a septum fits into a regulated workflow, the U.S. EPA SW-846 test methods and the Restek technical resources are where I send new analysts. For hardware and consumable compatibility, the Agilent chromatography consumables and Thermo Fisher chromatography and sample handling references are the practical starting points.
What I specify, and what I refuse to specify
For volatile work on a 2 mL vial, I specify a 11 mm silicone/PTFE septum with a molded center, compression set under 20% per ASTM D395, and a lot date code under 12 months old. I specify a crimp height of 5.5 to 5.7 mm on a go/no-go gauge, or a torque of 0.8 to 1.2 N·m on a screw cap. I specify a gravimetric loss test on every new lot before it enters a validated method. I refuse to specify a septum on durometer alone, and I refuse to accept a lot without a date code.
The decision rule is simple: if you cannot measure the loss gravimetrically and you cannot document the compression set, you do not know whether the seal holds, and the run will tell you eventually. Measure the loss, check the set, and match the crimp or torque to the vial. Everything else is a guess dressed up as a specification.
For the vial and filter side of the same workflow, the vial and filter reference is a useful cross-check when you are qualifying a new lot.
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:
