A 96-well autosampler tray sat overnight at 8 °C and came back 0.9 mg lighter per vial on the analytical balance the next morning. That is 0.9 µL of mostly water and acetonitrile gone from a 1.5 mL fill in 16 hours. The run itself was fine. The sequence that ran the next afternoon drifted 4.1% in peak area against the bracketing standards, and the only thing that had changed was the time between capping and injection. That is what HPLC vial seal failure looks like when nobody is watching: not a catastrophic leak, but a slow, silent loss of the sample you thought was sealed.
I have chased this failure mode across three labs and two manufacturing lines, and the pattern is consistent. The vial is almost never the sole culprit, but it is almost always part of the path. This article is about where the sample actually escapes, what happens to a septum after you puncture it, how compression set eats your seal over a run, and how to measure the loss gravimetrically so you stop guessing and start quantifying.
Where the sample actually leaves: four leak paths, ranked by how often I see them
When a customer calls about evaporation loss, I ask them to cap a vial, weigh it, and weigh it again in four hours. If the mass drops, the question becomes which path. There are four, and they are not equally common.
Path 1: the septum-to-neck interface (the big one)
This is the seal between the underside of the septum and the flat land of the vial neck. It is a face seal, held by the compression of the cap. If the cap torque is low, or the neck land is not flat, or the septum is too hard, the seal is incomplete before the needle ever touches it. A 11 mm crimp vial with a 1.0 mm thick silicone/PTFE septum needs the aluminum crimp to deform the skirt and pull the septum down onto the land with roughly 2.5 to 3.5 N·m of applied crimping force (not cap torque in the screw sense; crimp force is set on the crimper). If your crimper is set to 2.0 N·m, you have a leak path you cannot see.
Path 2: the puncture hole after the needle withdraws
Silicone is elastic, but it is not perfectly elastic. After a 0.5 mm OD needle passes through a 1.0 mm silicone layer, the hole does not close to zero. It closes to something like 0.1 to 0.2 mm in a good septum, and to 0.3 mm or more in a poor one. That residual hole is a diffusion path. It is small, but over an 18-hour tray it matters. This is the mechanism people miss when they blame the cap.
Path 3: the cap-to-neck thread or crimp skirt
On screw vials (9 mm short thread, GPI 9 mm), the seal is made by the septum being compressed between the cap’s internal shoulder and the neck land. If the cap is under-torqued, the shoulder does not load the septum. If it is over-torqued, the septum extrudes or the cap cracks. The window is narrow: I recommend 1.2 to 1.6 N·m for 9 mm screw caps, checked with a torque tester on the benchtop, not by feel.
Path 4: through the septum material itself
Silicone is permeable to water vapor and to small organics. A 1.0 mm silicone septum passes water vapor at a measurable rate even with no puncture. PTFE-faced septa cut this dramatically because the PTFE film is the barrier; the silicone is only the spring. If you are running aqueous mobile phases and long trays, a pure silicone septum is the wrong choice, full stop.
Re-seal behaviour after puncture: what the needle actually does to the septum
A septum is not a self-healing membrane. It is a compressed elastomer that stores elastic energy and releases it when the needle withdraws. The re-seal quality depends on four things: the elastomer, the thickness, the needle gauge, and how many times you puncture it.
Puncture count is a real specification, not marketing
I treat puncture count as a hard limit. A 1.0 mm silicone/PTFE septum in a 9 mm cap will typically re-seal acceptably for 3 to 5 punctures with a 0.5 mm needle. By puncture 8, the residual hole is large enough that I can measure weight loss on a 1.5 mL fill within 6 hours at ambient. If your method injects the same vial more than three times (common in stability studies and in some EPA SW-846 prep workflows), you are outside the reliable re-seal window and should be using a fresh vial per injection or a septum rated for higher puncture counts.
Needle gauge changes everything
A 0.5 mm OD needle leaves a smaller hole than a 0.8 mm OD needle, and the difference in re-seal is not linear. I have measured roughly 2x the residual hole area going from 0.5 to 0.8 mm on the same septum. If you are running a method that needs a larger bore for viscous samples, budget for more evaporation and shorter tray life. The needle is part of the seal system whether your method development document says so or not.
Pre-slit septa: a trade, not a free win
Pre-slit septa reduce coring and need less needle force, but the slit is a permanent opening. They re-seal worse than a good unslit septum after the first puncture because there is no intact elastomer to close against. I use pre-slit only when the method demands low needle force (fragile autosampler, high-throughput) and the tray is short. For overnight trays, unslit is the better seal.
Compression set: the slow failure nobody measures until the run drifts
Compression set is the permanent deformation a septum takes after being compressed for a period of time at temperature. It is the reason a vial that sealed perfectly on day one leaks on day three. Silicone has good compression set resistance at room temperature but degrades above 40 °C. If your autosampler tray is chilled to 4 to 8 °C, compression set is low and the seal holds. If your sample prep involves a 60 °C water bath or a heated tray, the septum takes a set and does not fully recover when it cools.
What compression set looks like in numbers
A good silicone septum under 25% compression at 23 °C for 22 hours should show compression set under 20%. Above 30%, the septum has lost enough spring that the cap cannot maintain the face seal, and you get a slow leak at the neck land. I have seen septa from low-cost suppliers hit 35 to 40% set after a single heated prep cycle. That is a seal that is already compromised before the run starts.
Why the cap matters as much as the septum
The cap provides the compression. A crimp cap that deforms plastically (cheap aluminum, thin skirt) loses its grip over time. A screw cap with a soft plastic shoulder creeps under load. If you are buying caps and septa separately, you own the interface risk. Buy them as an assembled closure from a supplier who controls both, and verify the assembly torque or crimp force on receipt. The vial closures you choose are half the seal, and the septum is the other half.
Measuring evaporation loss gravimetrically: the method I actually use
You cannot fix what you cannot measure. Here is the protocol I run when a customer reports drift and suspects the vial.
Step 1: prepare matched vials
Take 12 vials from the suspect lot and 12 from a known-good lot. Fill each with 1.0 mL of the same solvent blend (use your actual mobile phase, not water; water evaporates differently). Cap with the same closure type. Record the tare mass of each empty vial and cap on a balance readable to 0.1 mg.
Step 2: weigh immediately after capping
Weigh each filled, capped vial within 2 minutes of capping. Record the mass. This is your time-zero. Do not skip the tare; the vial-to-vial mass variation is larger than the evaporation loss you are trying to measure.
Step 3: store under your real conditions
Put the tray in the autosampler at your actual tray temperature (4 °C, 8 °C, or ambient) for the duration of a real run. If your run is 18 hours, measure at 18 hours. Do not shortcut this to 1 hour; the loss curve is not linear early on.
Step 4: reweigh and calculate
Reweigh each vial. Subtract the time-zero mass. Divide by the fill mass. That is your percent loss. For a 1.0 mL aqueous fill at 8 °C over 18 hours, I expect under 0.3% from a good closure and 0.8 to 1.5% from a marginal one. If you see more than 1.5%, the seal is failing and the drift you are chasing is real.
Step 5: separate the paths
To isolate the puncture path, run a matched set that is never punctured. The difference between punctured and unpunctured loss is your re-seal penalty. To isolate the neck seal, run a set with the cap over-torqued and a set under-torqued. The loss curve will tell you which path dominates. This is the only way I have found to stop arguing about whether the vial or the cap is at fault.
Field case 1: the 8 °C tray that lost 1.2% in 14 hours
Symptom: A stability lab running a 14-hour overnight sequence at 8 °C saw peak area drift of 3.8% across the tray, with the last injections lowest. Bracketing standards were within 1%.
Measurement: Gravimetric check on 12 vials from the run showed 1.2% mass loss over 14 hours. The unpunctured control set lost 0.4%. The punctured set lost the rest. The septum was a 1.0 mm silicone-only septum (no PTFE face) in a 9 mm screw cap torqued by hand.
Fix: Switched to a 1.0 mm silicone/PTFE septum in the same cap, set the cap torque to 1.4 N·m with a torque tester, and moved to a fresh vial every 3 injections. Loss dropped to 0.25% over 14 hours and drift fell under 1%. The PTFE face was the single biggest contributor; the torque setting was second.
Field case 2: the heated prep that sealed on day one and leaked on day three
Symptom: A method with a 60 °C derivatization step in the vial showed no loss on day one but 2.1% loss by day three on the same capped vials. The lab had assumed the caps were fine because the first day was clean.
Measurement: Compression set test on the suspect septa: 38% set after one 60 °C cycle. The cap was a thin-skirt aluminum crimp that had also relaxed. The neck land on a sample of vials showed 0.05 mm of flatness variation, enough to break the face seal once the septum stopped springing back.
Fix: Moved the derivatization out of the vial into a sealed reaction block, then transferred to a fresh 9 mm screw vial with a 1.0 mm silicone/PTFE septum and a cap torqued to 1.5 N·m. Loss dropped to 0.2% over three days. The lesson: any heated step inside the vial is a compression set event, and the seal you had on day one is not the seal you have on day three.
Decision checklist: what to specify and what to verify on receipt
Use this as a purchasing and incoming-inspection checklist. The numbers are the ones I have found hold up across suppliers.
| Parameter | Specify | Verify on receipt |
|---|---|---|
| Septum material | Silicone/PTFE faced for aqueous or long trays | PTFE film present, bonded, no delamination |
| Septum thickness | 1.0 mm for 9 mm and 11 mm vials | Micrometer check, ±0.05 mm |
| Compression set | <20% at 23 °C, 22 h, 25% compression | Supplier data sheet, or in-house test on 5 pieces |
| Puncture rating | 3 to 5 punctures with 0.5 mm needle | Gravimetric loss after 5 punctures <0.5% in 6 h |
| Cap torque (9 mm screw) | 1.2 to 1.6 N·m | Torque tester on 10 caps per lot |
| Crimp force (11 mm) | 2.5 to 3.5 N·m on crimper | Crimp height gauge, 10 per lot |
| Neck land flatness | <0.03 mm variation | Optical or contact check on sample |
| Evaporation loss (18 h, 8 °C, 1.0 mL) | <0.3% | Gravimetric protocol above, 12 vials |
If a supplier cannot give you compression set and puncture data, that is your answer. The HPLC septa are the part of the closure that does the sealing work, and the data should come with the part, not after the complaint. For broader method context, the Agilent consumables documentation and the Restek technical library both publish useful closure and septum guidance that lines up with what I measure in the lab.
Operational precautions that actually move the number
These are the changes I make first, in order of impact per unit of effort.
- Use a PTFE-faced septum for any tray longer than 8 hours or any aqueous fill. The PTFE film is the barrier; the silicone is the spring. This alone cut loss by more than half in both field cases above.
- Set cap torque with a torque tester, not by feel. Hand-torqued 9 mm caps vary by ±0.6 N·m across operators. That is the difference between a seal and a leak.
- Replace the vial every 3 injections. The re-seal penalty after puncture 5 is measurable and grows. If your method injects more, budget for it in the drift allowance.
- Keep the tray at 4 to 8 °C, not ambient. Every 10 °C rise roughly doubles the vapor loss rate through the septum. If you cannot chill, shorten the tray.
- Do not store capped vials for days and assume day-one seal. Compression set is cumulative. Re-cap with a fresh closure if the vial sits more than 24 hours.
- Buy vial, septum, and cap as an assembled closure from one supplier. The interface is where the leaks live, and splitting the purchase splits the accountability.
For the vial side of the equation, the flatness of the neck land and the concentricity of the bore are the two dimensions that determine whether the septum can seal at all. A HPLC vial with a neck land that varies by more than 0.03 mm will leak even with a perfect septum and correct torque. If you are sourcing across multiple suppliers, the chromatography products catalogs and the Thermo Fisher sample handling references are useful for cross-checking dimensional specs before you commit a lot.
The decision rule
If your gravimetric loss exceeds 0.3% over an 18-hour tray at 8 °C, treat it as a seal failure and work the four paths in order: PTFE-faced septum first, torque or crimp force second, puncture count third, neck land flatness fourth. If loss is under 0.3%, the drift you are chasing is somewhere else and the vial is not your problem. Measure before you replace, and replace the whole closure assembly, not just the part you suspect.
References & further reading
Related resources from our lab consumables network:
- Hplc Vial at hplcvialfactory.com
- Chromatography Products at chromatographyforsale.com
- Hplc Septa at hplcsepta.com
- Vial Closures at vialsclosures.com
External references and standards:
