A 24-vial sequence on a headspace autosampler came back with the internal standard area down 18% on the last six vials and a retention time shift of 0.04 min. The method was fine. The column was three weeks old and well within its use count. The problem was a box of 9 mm screw caps whose PTFE-faced silicone septa had been stored in a warm room since the previous summer, and the seal had taken a permanent set. That is the failure mode this article is about: chromatography products seal failure, evaporation and sample loss traced to the consumable, not the instrument.
Most labs chase this kind of drift through the inlet, the liner, the column, and the detector. They replace parts in that order, and they spend a week doing it. The seal is usually the last thing anyone checks because the vial looks fine. It is not fine. It has been quietly losing 2 to 6 microlitres of solvent per vial per hour, and the analyte concentration in the remaining liquid has been climbing the whole time.
Evaporation loss is not a theory. It is a mass balance you can run on a four-place balance in about 90 minutes. I do this on every new lot of septa before it goes into a validated method, and I have caught three bad lots in the last two years that way.
Gravimetric loss test, the version that actually works
Take ten vials from the lot. Fill each with 1.000 mL of HPLC-grade methanol using a positive-displacement pipette, not an air-displacement one, because you want the fill volume to be reproducible to 1 mg. Cap them with the septa and caps under test, torque to the manufacturer’s spec (for 9 mm screw thread that is typically 0.5 to 0.7 N·m; for 11 mm crimp, the crimper setting matters more than torque). Weigh the whole set immediately on a balance readable to 0.1 mg. Record the time. Then put them on the autosampler tray at the method’s actual tray temperature and leave them for the full run length. Weigh again at the end, same order, same balance, same draught conditions.
Divide the mass loss by the number of hours and by the number of vials. You now have a loss rate in mg/vial/hour. For methanol at 25 °C, anything above 1.5 mg/vial/hour is a fail for a method with a 12-hour tray residence. Good septa on a properly torqued screw cap run 0.2 to 0.6 mg/vial/hour. Silicone-only septa with no PTFE face and a loose cap run 4 to 9 mg/vial/hour, and that is where your 18% internal standard drop comes from.
Why the gravimetric number beats the peak area number
Peak area drift tells you something changed. It does not tell you what. A 3% loss of methanol from a 1 mL fill concentrates everything in the vial by roughly 3%, so your internal standard area should go up, not down, unless the standard itself is volatile. When the area goes down, you are usually looking at a leak path that lets air in and lets the more volatile component out preferentially, and the ratio shifts in a way that looks like instrument drift. Weigh the vials. It takes an hour and it ends the argument.
Four leak paths, and only one of them is the septum face
When a vial loses mass, the septum face gets blamed. In my experience the face is the culprit maybe 40% of the time. Here are the paths, in the order I check them.
1. The cap-to-neck thread interface
A screw cap seals on the top of the neck, not on the threads. If the cap is under-torqued, the septum is not compressed against the neck rim and vapour escapes around the outside. If it is over-torqued past about 0.9 N·m, the cap skirt can crack at the stress riser and you get a slow leak you cannot see. I have seen a whole rack of 11 mm snap caps pop off under thermal cycling because the snap feature was moulded at the low end of tolerance. Check the cap, not just the septum.
2. The septum-to-cap seat
The septum sits in a recess in the cap. If the septum OD is undersized by 0.2 mm, it does not fill the recess and the cap cannot compress it evenly. This is a dimensional fit problem, and it is why I buy septa and caps as a matched set from one source rather than mixing. Sites that mix brands to save a few cents per vial are the ones that call me about drift.
3. The needle puncture itself
Every puncture leaves a hole. The hole does not close perfectly. This is the re-seal problem, and it deserves its own section below.
4. The septum-to-neck contact ring
This is the one most people miss. The septum is compressed against the neck rim, a ring maybe 0.4 mm wide. If the neck rim has a moulding flash or a nick, the seal is incomplete no matter how good the septum is. Hold a vial up to a light and look at the rim. If you can see a burr, that vial will leak. I reject whole cases over this.
Re-seal after puncture is a material property, not a hope
A septum is asked to do something physically unreasonable: be soft enough to seal against a needle, stiff enough to close the hole when the needle withdraws, and chemically inert to everything in the vial. Silicone does the first and third well and the second poorly. That is why the PTFE face exists, and why the choice of septum construction is the single biggest lever on evaporation loss after the first puncture.
What the puncture actually does
A 23-gauge needle is 0.64 mm OD. It pushes a clean hole through the silicone. When it withdraws, the elastomer tries to close the hole by elastic recovery. Pure silicone recovers maybe 70 to 85% of the hole diameter within a few seconds, which leaves a residual channel of 0.1 to 0.2 mm. That channel is a leak. A PTFE-faced septum has a thin fluoropolymer film on the sample side that is slit by the needle and does not recover, but the silicone behind it does, and the combination seals better than either alone because the PTFE acts as a low-permeability barrier over the recovering silicone.
This is why PTFE-faced silicone is the default for anything with more than one puncture, and why silicone-only septa are only appropriate for single-injection vials that go straight from prep to tray to needle. If your method reinjects or your autosampler does a wash cycle with the needle still in the vial, silicone-only is the wrong choice and you will see it in the gravimetric test.
Puncture count versus loss rate
I ran a bench test last year on three septum types, 1 mL methanol fill, 25 °C, weighing after each puncture. The numbers below are averaged over ten vials per type. They are the kind of numbers you should generate for your own lots, because they vary by manufacturer.
| Septum type | Loss after 1 puncture (mg/h) | After 5 punctures | After 10 punctures | Practical puncture limit |
|---|---|---|---|---|
| PTFE-faced silicone, 1.0 mm | 0.3 | 0.5 | 0.9 | 8 to 10 |
| Silicone-only, 1.0 mm | 0.6 | 1.8 | 4.1 | 3 to 4 |
| PTFE-faced silicone, 0.5 mm | 0.4 | 0.9 | 2.2 | 5 to 6 |
| Butyl rubber, 1.0 mm | 0.2 | 0.4 | 1.1 | 6 to 8 (but poor solvent resistance) |
Notice the thickness effect. A 0.5 mm septum looks like a cost saving and behaves like a 5-puncture part. If your method punctures more than five times, buy 1.0 mm and stop arguing about the price difference, which is usually under two cents per vial.
Compression set is why a septum that passed incoming inspection fails in week three
Compression set is the permanent deformation a elastomer takes when it is held under compression for a long time at temperature. A septum is compressed from the moment the cap is torqued until the vial is discarded. If the method holds vials on a tray at 40 °C for 8 hours, the septum is under compression at 40 °C for 8 hours, every day, for the life of the lot. Over weeks, the elastomer relaxes. The seal force drops. The leak rate climbs.
The number that matters
Compression set is measured per ASTM D395 Method B: compress the sample 25%, hold at temperature for 22 hours, release, measure the residual deformation after 30 minutes. A good silicone septum for chromatography returns to within 15 to 20% of its original thickness. A cheap one returns to within 35%, which means it has lost a third of its sealing force permanently. That is the lot that fails in week three.
Ask your supplier for the compression set number at the tray temperature your method uses, not at room temperature. A septum rated at 15% set at 25 °C can be 30% at 40 °C. The datasheet usually gives the room-temperature figure because it is the flattering one.
Storage is part of the seal
Septa have a shelf life and it is shorter than the box says if the box sits on a windowsill. Silicone is prone to ozone attack and UV degradation, both of which raise compression set. Store sealed, dark, below 25 °C, and use within the manufacturer’s date code. I have seen a lot that was two years past its date code and had a compression set of 41%. It was sold at a discount. It cost the lab three weeks of drift investigation.
Field case 1: the 18% internal standard drop
Symptom: Headspace GC method, 24-vial sequence, internal standard area fell 18% across the run, retention time shifted 0.04 min, no change in peak shape.
Measurement: Gravimetric loss test on the current lot of 9 mm screw caps. Loss rate was 3.8 mg/vial/hour at 40 °C tray temperature, against a method specification of 1.0. The septa were PTFE-faced silicone but 0.5 mm thick, and the caps had been torqued by hand to an unknown value. Compression set measured at 40 °C was 34%.
Fix: Switched to 1.0 mm PTFE-faced silicone septa from the same supplier as the caps, torqued to 0.6 N·m with a preset torque driver, and moved the septum storage to a cool dark cabinet. Loss rate dropped to 0.4 mg/vial/hour. Internal standard RSD across the sequence went from 11% to 1.8%. The method was not changed.
Field case 2: the crimp cap that looked fine
Symptom: A GC-FID method on 2 mL crimp vials showed a slow baseline rise over a 6-hour tray residence and a 7% drop in the least volatile analyte. The crimper was the usual hand crimper, and operators swore it was set correctly.
Measurement: We weighed 20 crimped vials with 1 mL of hexane, left them on the tray for 6 hours, and reweighed. Eight of the 20 lost more than 5 mg. The other 12 lost under 1 mg. The difference was the crimp height. We measured the crimp with a crimp gauge and found the failing vials were at 5.6 mm against a target of 5.2 mm. The crimper jaw had worn and was not pulling the cap down far enough to compress the septum against the neck rim.
Fix: Replaced the crimper jaw, set the crimp height with a gauge, and added a 1-in-20 crimp height check to the shift start-up. Loss rate across the tray dropped below 0.5 mg/vial/hour. The baseline rise disappeared.
A troubleshooting checklist you can run in one shift
When you suspect seal failure, run these in order. Each one takes under 30 minutes and each one either clears a suspect or confirms it.
- Weigh ten vials before and after the run. Loss above 1.5 mg/vial/hour at tray temperature is a fail. This is the only test that separates a seal problem from an instrument problem.
- Check cap torque with a preset driver. 9 mm screw: 0.5 to 0.7 N·m. 11 mm crimp: crimp height 5.2 mm, measured with a gauge, not by feel.
- Inspect the neck rim under a light. Any flash or nick is a reject. Check ten vials from the case.
- Confirm septum thickness and construction. 1.0 mm PTFE-faced silicone for multi-puncture methods. 0.5 mm only for single-injection.
- Ask for the compression set at tray temperature. Above 25% is a fail for a method with more than 4 hours of tray residence.
- Check the date code and storage. Past date code or stored above 25 °C in light: reject the lot.
- Match septum and cap from one source. Mixed brands are the most common cause of dimensional misfit.
If all seven pass and you still lose mass, the problem is not the consumable. Look at the tray temperature, the autosampler wash solvent, and the vial fill volume. A 0.5 mL fill in a 2 mL vial has five times the headspace of a 1.5 mL fill and loses mass faster because the vapour space is larger. Fill volume is a seal variable too.
What to specify when you buy, and what to reject at the dock
The fix for seal failure is a purchase specification, not a troubleshooting procedure. Write these into the PO and hold suppliers to them. I buy through chromatography products sources that publish the numbers, and I reject lots that arrive without them.
Specify 1.0 mm PTFE-faced silicone septa, compression set below 20% at the method tray temperature per ASTM D395 Method B, supplied with caps from the same mould family so the septum recess matches. Specify 9 mm screw caps torqued to 0.6 N·m or 11 mm crimp caps at 5.2 mm crimp height. Specify a maximum gravimetric loss rate of 1.0 mg/vial/hour at tray temperature for a 1 mL methanol fill over 6 hours, and require the supplier to state the test method. If they cannot state it, they have not measured it.
For the source side, the major vendors publish septum specifications and material data: Agilent’s GC columns and accessories pages list septum thickness and construction, Thermo Fisher’s chromatography and sample handling pages cover caps and septa together, and Restek’s technical resources have good application notes on septa selection. For methods that fall under regulatory scope, the U.S. EPA SW-846 test methods give the holding-time and preservation requirements that your seal has to meet, and those requirements are where the gravimetric number gets its pass/fail threshold.
For matched vial, cap, and septum sets, the practical route is a single supplier who moulds both. Vials with caps is the category to search when you want the set rather than the parts, and vial supplier sources are where I check dimensional tolerances before I commit a method to a lot. On the manufacturing side, Aijiren publishes septum and cap specifications that include compression set, which is more than most catalogs do.
The decision rule
If a run loses more than 1.5 mg per vial per hour on a gravimetric test at tray temperature, the consumable is the problem and no amount of inlet maintenance will fix it. Replace the septum and cap as a matched set, specify 1.0 mm PTFE-faced silicone with compression set under 20% at your tray temperature, control torque to 0.6 N·m or crimp height to 5.2 mm, and re-run the gravimetric test on every new lot before it enters a validated method. The number on the balance is the number that matters, and it is the only one that tells you whether the seal or the instrument is losing your sample.
References & further reading
Related resources from our lab consumables network:
- Chromatography Products at chromatographyselling.com
- Aijiren at aijirentech.com
- Vials With Caps at vialswithcaps.com
- Vial Supplier at vialsupplier.com
External references and standards: