A 96-position LC-MS run lost 11.4% of its internal standard peak area between injection 4 and injection 78, and the gravimetric check on a duplicate plate put the average well loss at 3.8 mg over 19 hours at 8 °C. The method had been validated. The column was new. The autosampler was serviced. What had changed was a lot of 9 mm screw-thread vials with a different septa vendor, and the seal failure was invisible until we weighed the plate. That is the shape of most vial seal failure, evaporation and sample loss investigations I have run: the symptom shows up as drift or a missing low-boiler peak, and the cause lives in the closure interface, not in the instrument.
This is a troubleshooting article. It assumes you already know what a vial is, that your autosampler has a needle, and that your method has a run time. What it covers is where the seal actually leaks, how a septum behaves after it has been punctured, why compression set decides whether the second injection matches the first, and how to put a number on the loss instead of arguing about it.
A crimp or screw vial is a compression seal, not a gasket joint. The septum is squeezed between the glass land and the cap, and the only thing keeping solvent inside is the elastic recovery of the elastomer against that land. Every leak path I have traced in twenty years falls into one of four categories, and they are not equally common.
Path 1: insufficient compression at the cap-to-glass land
This is the big one. On a 11 mm crimp vial, the aluminum seal has to be deformed so the skirt grips under the bead and the septum is compressed roughly 30 to 40% of its nominal thickness. A 1.0 mm silicone/PTFE septum at 40% compression is being squeezed to about 0.6 mm. If your crimper is set too loose, you get maybe 15% compression, the septum never reaches its sealing stress, and the vial holds pressure for an hour and then bleeds. On a 9 mm screw vial the equivalent failure is under-torquing. I target 0.9 to 1.2 N·m on a 9 mm screw cap with a silicone/PTFE septum, checked with a torque screwdriver, and I have watched labs run the same caps at 0.4 N·m because nobody ever calibrated the capper.
Path 2: over-compression and land damage
The opposite error is just as real. Crank a crimper down and you cut the septum, deform the glass land, or crack the neck. I have seen crimpers set so tight that the aluminum skirt splits at the knurl, which is a slow leak you cannot see until the run drifts. Over-torquing a screw cap past about 1.5 N·m starts to lift the septum off the land at the thread entry and can chip the finish. The seal has an optimum, and it is narrower than most people assume.
Path 3: the septum-to-needle interface after puncture
Every injection leaves a hole. The question is whether the elastomer closes it. This is covered in detail below because it is the path most people misdiagnose as a cap problem.
Path 4: permeation through the septum body and the headspace above the liquid
Even a perfectly seated septum is a polymer membrane. Small, non-polar molecules permeate silicone at a measurable rate. Acetonitrile and methanol are the usual suspects, and the rate scales with temperature and with the concentration gradient across the septum. This is why a vial that holds fine at 4 °C for a week loses 2% a day at 25 °C. If you are running a headspace vial at 80 °C for 30 minutes, permeation is a design consideration, not a defect.
Re-seal behaviour after puncture is the number nobody specifies
Here is the mechanism. A needle punches through the septum and displaces elastomer radially. When the needle withdraws, the elastomer recovers toward its original geometry, but recovery is never complete. The puncture leaves a residual channel whose diameter depends on needle gauge, septum thickness, elastomer hardness, and how long the needle dwelled. A 22-gauge needle (0.72 mm OD) through a 1.0 mm silicone septum recovers to a channel of roughly 0.1 to 0.2 mm within seconds. A 26-gauge (0.46 mm) leaves less. A blunt or burred needle leaves more, because it tears rather than parts the elastomer.
The practical consequence: the first injection from a vial is usually fine, and the loss shows up on later injections from the same vial. If your autosampler does one injection per vial and discards it, puncture re-seal barely matters. If you are doing a 6-injection calibration curve from the same vial, or a long sequence that re-samples a standard, it matters a lot. This is the reason some methods specify a fresh vial per injection and others do not, and it is worth checking which your method actually assumes.
Puncture count: what the septum will actually take
I use these working limits for silicone/PTFE septa, and I treat them as ceilings, not targets:
- 1.0 mm silicone/PTFE, 22-gauge needle: 3 to 5 clean punctures before the channel is measurable by gravimetric loss. Beyond 5, expect drift.
- 1.3 mm silicone/PTFE, 22-gauge: 5 to 8 punctures, with the caveat that thicker septa need more compression to seal and are more sensitive to crimper setup.
- PTFE-faced silicone, pre-slit: designed for high puncture counts, but the slit is a permanent leak path once the vial is pressurized. Good for autosamplers with a fixed needle, poor for anything that needs a pressure seal.
- Butyl rubber: excellent re-seal, poor solvent resistance, and it will swell in acetonitrile. Not a general-purpose choice.
If your method re-samples a standard more than five times, the septum is the wrong component for the job, regardless of what the vendor’s datasheet says. Use a fresh vial or move to a larger-volume standard that is sampled less often. The GC vial selection logic is the same: match the septum to the puncture count, not to the price.
Compression set is why the second run does not match the first
Compression set is the permanent deformation an elastomer retains after being compressed and released. It is the reason a septum that sealed perfectly on day one can leak on day three without anything else changing. When you crimp or torque a cap, the septum is held at 30 to 40% compression. Over hours at autosampler temperature, the elastomer relaxes. The sealing stress decays. A silicone septum that started at 1.5 MPa sealing stress may be at 1.0 MPa after 24 hours, and a screw cap that was torqued to 1.0 N·m may have backed off to 0.7 N·m because the septum took a set and the cap lost preload.
This is why gravimetric loss is not linear. The first few hours look fine, then the slope steepens. If you only weigh at the start and end of a 24-hour run, you will average a leak that was actually concentrated in the last third of the run. Weigh at intervals.
The temperature dependence
Compression set accelerates with temperature. At 4 °C in a cooled autosampler tray, set is slow and a well-sealed vial holds for days. At 25 °C on the bench, set is faster. At 40 °C in a warm room, or in a vial heater, set can be substantial within hours. The same septum that gives you a 0.5% loss over 24 hours at 4 °C can give you 3% at 30 °C. This is not a manufacturing defect; it is the material doing what elastomers do. The fix is to shorten the time the vial sits at temperature, not to demand a septum that does not exist.
Measuring loss gravimetrically: the procedure that settles arguments
You cannot troubleshoot what you cannot measure. The gravimetric method is simple and it is the only one I trust for seal integrity, because it captures permeation, leak paths, and re-seal loss in a single number. Here is the procedure I use.
Materials and setup
- Analytical balance readable to 0.1 mg, calibrated, on a vibration-isolated bench.
- Matched vials and caps from the lot under test, plus a reference lot you trust.
- A thermostatted environment at the temperature your method actually sees (4 °C, 25 °C, or the autosampler tray temperature).
- Solvent matched to your mobile phase, filled to a defined volume (I use 1.0 mL in a 2.0 mL vial, leaving a defined headspace).
The run
Weigh each capped vial to 0.1 mg immediately after capping. Record the time. Weigh again at 1, 4, 8, 12, and 24 hours. Do at least six replicates per lot. Calculate loss as a percentage of the initial solvent mass, and plot it against time. A well-sealed vial at 4 °C loses less than 0.3% over 24 hours. A vial with a marginal crimp loses 1 to 2%. A vial with a failed seal loses 5% or more. The curve shape tells you which failure mode you have: a straight line is permeation, a curve that steepens is compression set or a channel opening up, and a step change is a physical leak that started at a specific time.
What the numbers mean
Here is a decision table I use to interpret gravimetric results on 2.0 mL vials with 1.0 mL fill and silicone/PTFE septa:
| 24-hour loss at 4 °C | Interpretation | Action |
|---|---|---|
| < 0.3% | Seal is sound | Accept lot |
| 0.3 to 1.0% | Marginal crimp or torque, or permeation through a thin septum | Re-check crimper/torque settings, verify septum thickness |
| 1.0 to 3.0% | Compression set or under-compression | Increase compression to 35%, re-test; check capper calibration |
| > 3.0% | Leak path present | Reject lot; inspect cap and glass land; check for cracked neck |
Run the same table at 25 °C and the thresholds roughly double, which is expected. That is not a reason to relax the limits; it is a reason to keep your autosampler tray cool.
Field case 1: the drift that was a crimper, not a column
A pharmaceutical QC lab running a 60-injection sequence on an HPLC saw the internal standard area fall 8% across the run, with the low-boiler peak falling faster than the high-boiler peak. The differential was the clue: if the column were failing, all peaks would fall together. The selective loss of the volatile component pointed at evaporation. Gravimetric testing on the suspect lot showed 2.6% loss over 18 hours at 22 °C, against 0.4% for the reference lot. The crimper had been serviced and the anvil height had drifted by 0.2 mm, which reduced compression from about 38% to about 22%. Re-setting the crimper and re-testing brought the lot to 0.5% loss and the run drift to under 1%. The column was fine the whole time. The lesson: when the loss is selective by boiling point, suspect the seal, not the separation.
Field case 2: the screw cap that backed off
A cannabis testing lab running terpene analysis on a GC vial with a 9 mm screw cap saw a 12% drop in the monoterpene response over a 12-hour sequence, with the effect growing after injection 30. Gravimetric testing showed the loss was not linear: 0.2% at 4 hours, 1.1% at 8 hours, 3.4% at 12 hours. That curve shape is compression set. The caps had been torqued by hand to an unknown value, and a torque check showed a range from 0.3 to 1.1 N·m across operators. Switching to a calibrated capper set to 1.0 N·m and using a 1.3 mm septum brought the 12-hour loss to 0.6% and the sequence drift to 2%. The fix was not a different vial; it was a torque value and a calibrated tool.
The troubleshooting checklist, in the order I run it
When a run shows drift or a volatile component is disappearing, work this list in order. Stop when you find the cause.
- Is the loss selective by boiling point? If yes, it is evaporation, not the column. Move to seal checks.
- Does the drift start after a specific injection number? If it starts after injection 4 or 5, suspect puncture re-seal. If it starts later and grows, suspect compression set.
- Gravimetric test the suspect lot against a reference lot at the actual tray temperature. Use the table above to interpret.
- Check crimper anvil height or capper torque with a gauge. Target 30 to 40% septum compression, or 0.9 to 1.2 N·m on a 9 mm screw cap.
- Check septum thickness with a micrometer on a sample of ten. A 1.0 mm nominal septum that measures 0.85 mm will under-compress at your normal setting.
- Check the glass land for chips or cracks under magnification. A chipped land is a leak path no crimper can fix.
- Check needle condition for burrs. A burred needle tears the septum and defeats re-seal.
- Confirm the septum material matches the solvent. Silicone swells in non-polar solvents; PTFE-faced silicone is the general-purpose choice, butyl is for aqueous and short-chain alcohols only.
For method-level context on container performance requirements, the U.S. EPA SW-846 test methods set container performance expectations that a vial has to meet, and vendor technical resources from Agilent, Thermo Fisher, and Restek publish septum and cap compatibility data worth checking against your solvent list. None of that replaces a gravimetric test on your own lot at your own temperature.
The decision rule for seal integrity
If your method re-samples a vial more than five times, or holds a volatile analyte for more than 12 hours at room temperature, specify a septum and closure that pass a gravimetric test of under 0.5% loss over 24 hours at your actual tray temperature, and control the crimp or torque with a calibrated tool rather than by feel. If your method does one injection per vial and runs within a few hours, a standard silicone/PTFE septum at 35% compression will do the job, and the money is better spent on capper calibration than on a premium septum. The vial is not the variable you can leave to the supply closet; the seal is a measured parameter, and it either passes the gravimetric test or it does not.
References & further reading
Related resources from our lab consumables network:
- Vial at vials-filters.com
- Headspace Vial at headspacevialsupplier.com
- Gc Vial at gcvialfactory.com
- Reagent Bottle at reagentbottleforsale.com
External references and standards: