In a 96-hour headspace run last March, a 20 mL vial with a 20 mm crimp seal lost 11.4 mg over the sequence. The same lot of vials with a different cap lost 1.8 mg. Same instrument, same method, same analyst. The difference was not the glass. It was the seal, and the loss showed up as a 6% drop in the internal standard response by the last time point. That is the kind of failure I trace back to the chromatography supplier more often than any other consumable issue, and it almost never announces itself until the data is already suspect.
Evaporation loss and seal failure are the slowest, quietest defects in the sample path. They do not throw an error. They do not spike a chromatogram. They shift your mass balance, they drift your standard response, and they cost you a re-run you cannot explain. This article is about the mechanisms: where the leak paths are, how a septum behaves after the needle leaves, how compression set accumulates, and how to measure the whole thing gravimetrically so you have a number instead of a suspicion.
A crimp or screw-top vial is not one seal. It is a stack of interfaces, and every interface is a potential leak path. When I get a loss complaint, I walk these four in order because the probability drops fast as you go down the list.
Path 1: the septum-to-neck land
This is the annular contact between the underside of the septum and the flat rim of the vial neck. It is the primary seal, and it depends on compression. A 20 mm crimp seal on a standard 20 mm neck needs the aluminum to deform the septum against that land with enough force to close the micro-gaps in the elastomer surface. If the crimp is loose, or the neck rim has a chip or a mold line, vapor walks straight out under the skirt. I have seen a 0.3 mm neck-height variation across a lot produce a 4x spread in 24-hour loss between vials in the same box.
Path 2: the cap-to-septum interface (screw caps)
On screw-thread vials the cap torque sets the compression. Under-torque and the septum never seats. Over-torque and you deform the septum, thin it at the center, and create a different leak. The window is narrow. For a typical 9 mm short-thread screw cap with a silicone/PTFE septum, I target 0.9 to 1.2 N·m and verify with a torque tester, not by feel. Hand-tight is not a specification.
Path 3: through the septum body
This is permeation plus puncture. Silicone is the most permeable common septum material. A 1.0 mm silicone septum on a 20 mm vial will pass measurable water vapor and solvent over a multi-day run even with a perfect crimp. PTFE-faced septa slow permeation but the face is thin (typically 0.05 to 0.13 mm) and does nothing for the puncture channel underneath.
Path 4: the puncture channel itself
Every needle creates a channel. Whether it closes is the whole question, and it deserves its own section because it is where most people misdiagnose the problem.
Re-seal is a material property, not a given
A septum is asked to do something remarkable: hold pressure, accept a needle, and close behind it. The closure is driven by the elastomer’s elastic recovery. Silicone recovers fast and almost completely, which is why it is the default for autosampler septa. But recovery is time- and temperature-dependent, and it degrades with every puncture.
I test re-seal the way I test anything else: puncture, pressurize, measure. A practical bench method is to fill vials with a volatile solvent, cap them, puncture N times with the actual needle gauge in use, and weigh at fixed intervals. A fresh silicone septum punctured once with a 23-gauge needle typically re-seals to better than 95% of its original seal within seconds at room temperature. By 50 punctures in the same spot, recovery can fall below 80%, and the loss rate climbs accordingly. Puncture in a new spot each time and the numbers hold much longer.
Needle gauge and bevel matter more than people think
A 26-gauge needle (0.46 mm OD) leaves a channel roughly half the diameter of an 18-gauge (1.27 mm OD). The channel area scales with the square, so the 18-gauge leaves about 7.6x the open area. A side-port or conical bevel cuts a cleaner slit that closes better than a square-cut bevel, which tends to core and tear. If your method allows a smaller gauge, take it. If you are running a 22-gauge for viscosity reasons, plan on shorter septum life and more frequent replacement.
Temperature changes the recovery curve
Silicone softens as it warms. At 40 °C headspace incubation, recovery is slower and less complete than at 25 °C. At 150 °C in a GC inlet, a standard silicone septum will bleed and deform within a day. That is why inlet septa are a different product (often a molded silicone with a PTFE face and a higher crosslink density) rated for a specific inlet temperature. Using a vial septum in an inlet is a classic cross-contamination of parts, and it fails predictably.
Compression set: the slow drift nobody measures
Compression set is the permanent deformation an elastomer takes after being held under load. It is the reason a septum that sealed perfectly on day one leaks on day thirty. The material takes a set, the recovery force drops, and the seal relaxes even though nothing moved.
The standard measure is ASTM D395, method B: compress a specimen 25% for 22 hours at a set temperature, release, and measure the residual deformation. A good vial septum shows set under 20% at 70 °C. A cheap one shows 35% or worse. That difference is invisible at the bench but shows up as a vial that will not hold vacuum after a week in a tray.
How set shows up in real runs
If you are running long sequences (overnight autosampler queues, multi-day stability studies), the vials sit capped for hours before injection. A septum with high compression set relaxes during that hold. The first injection in the queue is fine. The fortieth, twelve hours later, has lost volatile analyte through a relaxed seal. The symptom is a time-dependent response decline that tracks queue position, not analyte concentration. I have chased that exact pattern twice, and both times it was set, not the instrument.
Storage and shelf life
Elastomer septa have a shelf life. Stored in a hot warehouse, silicone continues to cure and can take a partial set from its own packaging. I keep septa below 25 °C, in the original sealed bag, and I do not use stock older than the supplier’s stated shelf life (commonly 2 to 3 years from manufacture). If a lot has been on the shelf for four years, the set numbers will not match the datasheet. Ask your chromatography supplier for the manufacture date, not just the expiry.
Measuring loss gravimetrically: a method you can run today
You cannot fix what you cannot measure. The good news is that evaporation loss is easy to quantify with a balance, a timer, and a controlled environment. This is the test I run whenever a customer reports drift and cannot tell me whether it is the vial, the cap, or the method.
The setup
- A balance readable to 0.1 mg (0.01 mg if you can get one), calibrated, on a vibration-isolated bench.
- A temperature-controlled space at 25 ± 1 °C and stable humidity. A drafty bench ruins the run.
- Twelve vials minimum per condition: six filled and capped, six as controls.
- A volatile solvent that matches your method (methanol, acetonitrile, or the actual sample matrix).
- A log sheet with time points at 0, 1, 4, 8, 24, 48, and 72 hours.
The procedure
Fill each vial to the same nominal volume (I use 10.0 mL in a 20 mL vial so there is headspace to evaporate into). Cap or crimp to the specification you are evaluating. Weigh immediately and record as t=0. Weigh at each time point, always in the same order, and record the mass to 0.1 mg. Subtract the t=0 mass from each subsequent reading. The difference is cumulative loss.
Express the result two ways: absolute loss in mg, and loss rate in mg per hour per vial. The rate is the number that lets you compare septa, caps, and crimp settings. A well-sealed 20 mL crimp vial with a silicone/PTFE septum running methanol typically holds loss under 0.5 mg per 24 hours at 25 °C. If you are seeing 5 mg or more, you have a seal problem, not a method problem.
What the numbers tell you
A flat loss curve that steps up after a puncture event points to re-seal. A steadily rising rate points to compression set or a permeation-dominant septum. A high loss from t=0, before any needle touches the vial, points to the crimp or the neck land. Run the test with and without a puncture to separate the two. That single comparison has saved me more diagnostic time than any other.
| Condition | Typical 24 h loss (mg) | Likely cause if exceeded |
|---|---|---|
| Crimp, silicone/PTFE, no puncture | < 0.5 | Crimp force, neck rim defect |
| Crimp, silicone/PTFE, 1 puncture | < 0.8 | Poor re-seal, needle gauge |
| Screw cap, 1.0 N·m, no puncture | < 0.7 | Under/over torque, thread damage |
| Screw cap, hand-tight | 1.5 to 4.0 | Insufficient compression |
| Bare silicone, 72 h hold | 3.0 to 8.0 | Permeation through septum body |
These are working numbers from my own bench, not a guarantee. Your solvent, temperature, and headspace will move them. The point is to establish your own baseline and compare lots against it.
Two field cases where the seal was the whole story
Case 1: the queue-position drift
A lab running an overnight autosampler queue of 60 injections saw the internal standard response fall 9% from injection 1 to injection 60. The instrument was clean, the column was new, and the standard was fresh. I asked them to weigh the vials at the start and end of the queue. The vials at the back of the tray had lost 6 to 9 mg more than the front vials over the 14-hour hold. The septa were a low-cost silicone with no PTFE face, and the caps were hand-tightened. The fix was a switch to a PTFE-faced septum and a torque driver set to 1.0 N·m. The queue-position drift disappeared on the next run. The root cause was compression set plus under-torque, and it took a balance to prove it.
Case 2: the crimp that looked fine
A different site had a 20 mL headspace method with a 12% response drop over 48 hours. The crimps looked uniform by eye. I measured neck height on 30 vials with a caliper and found a spread from 15.9 to 16.4 mm across the lot, well outside the 16.0 ± 0.1 mm I expect from a controlled mold. The taller necks were not getting full crimp compression. The supplier had changed mold tooling mid-lot. We rejected the lot, moved to a supplier who could show neck-height data per lot, and the loss fell to under 1 mg per 24 hours. The lesson: ask for dimensional data on the neck, not just the septum.
The supplier-side controls that actually predict seal performance
Most seal failures trace to a decision made before the vial reached your bench. When I qualify a chromatography supplier for vials and closures, I ask for four things that separate a controlled product from a commodity one.
Lot-level dimensional data on the neck and cap
Neck height, neck OD, and thread pitch (for screw vials, the GPI 9 mm or GPI 13-425 standard) need to be reported per lot, not per catalog. A supplier who cannot produce a neck-height histogram is not controlling the variable that drives your seal. This is the same discipline that applies to OEM vial sourcing, where the mold is the specification.
Septum material certification
Ask for the elastomer formulation, durometer (typically 45 to 55 Shore A for vial septa), and compression set data to ASTM D395. A datasheet that lists only “silicone” tells you nothing about set or permeation. The good suppliers publish the set number. The vial closures you choose should come with that data attached to the lot.
Crimp and torque specifications with a tolerance
If your supplier sells crimpers or torque tools, they should state the target and the window. If they sell neither, ask what crimp force their seal was validated at. A seal validated at one crimp setting and used at another is an untested combination.
Independent verification
For high-value methods, I send a sample of each new lot to an analytical testing lab for a gravimetric loss screen before the lot enters production. It costs less than a single failed run. The published methods for sample handling and container performance in U.S. EPA SW-846 and the general guidance in Restek’s technical resources give you a defensible framework for the test.
Troubleshooting checklist: from symptom to seal
When a run drifts or a mass balance fails, work this list in order. It is faster than guessing.
- Weigh the vials before and after the run. If loss exceeds 0.5 mg per 24 hours on a sealed, unpunctured vial, the problem is the crimp or the neck, not the method.
- Compare punctured vs unpunctured. A step change after puncture is re-seal. A steady rise is set or permeation.
- Check queue position. Loss that tracks position is a hold-time seal problem, usually set or under-torque.
- Measure neck height on 30 vials. Anything outside ±0.1 mm on a nominal 16.0 mm neck is a lot problem.
- Verify torque with a driver. Target 0.9 to 1.2 N·m for 9 mm screw caps. Do not trust hand-tight.
- Confirm septum material and set data. No ASTM D395 number, no qualification.
- Check storage temperature and age. Septa over shelf life or stored hot will not meet the datasheet.
- Match needle gauge to the septum. Smaller gauge, better re-seal. Side-port bevel over square-cut.
For instrument-side context on how septa and liners interact with the inlet, the vendor technical libraries from Agilent and Thermo Fisher are worth a read, but they will not replace your own gravimetric baseline. Run the test on your bench with your solvent.
What to specify on the purchase order
The decision rule is simple: a vial and closure is only qualified when you have a gravimetric loss number for it under your method conditions, and that number is under 0.5 mg per 24 hours for an unpunctured sealed vial. Everything else follows from that.
On the PO, specify the septum material and durometer, the compression set limit (under 20% at 70 °C, ASTM D395 method B), the neck height and tolerance, the cap torque window, the needle gauge the seal was validated with, and the manufacture date. Require lot-level dimensional data and reject lots that fall outside it. If your current chromatography supplier cannot provide those numbers, that is the finding. The seal failure is a symptom; the missing specification is the cause. Fix the specification and the drift stops.
References & further reading
Related resources from our lab consumables network:
- Chromatography Supplier at chromatographysupplier.com
- Oem Vial at oemsamplevials.com
- Vial Closures at vialclosures.com
- Analytical Testing at labanalyticaltesting.com
External references and standards: