A 12-hour COD reflux run lost 0.83 g from a 40 mL vial that had been punctured twice by an autosampler needle. The blank lost 0.04 g over the same period. Same oven, same batch of acid, same lot of vials. The only difference was two needle entries through a 9 mm PTFE-faced silicone septum, and that 0.79 g delta was enough to shift the dichromate titration by roughly 6 mg/L COD at the low end of the calibration. That is not a reagent problem. That is cod vial seal failure, and it shows up as evaporation and sample loss long before anyone suspects the vial.
Seal integrity is the part of vial selection that gets almost no attention until a batch of results drifts. I have spent years chasing these losses on bench runs and in the field, and the pattern is consistent: the leak is rarely a defective septum. It is a system of small, predictable leak paths that open when the cap is torqued wrong, when the septum takes a second puncture, or when the liner material takes a compression set at temperature. This article is about measuring those losses, finding the path, and fixing the hardware.
Every leak I have traced on a sealed cod vial falls into one of four paths. Knowing which one you have changes the fix, so measure before you swap parts.
Path 1: the needle track through the septum
A silicone septum does not heal. It closes, which is different. After a 22-gauge needle (0.72 mm OD) withdraws, the elastomer relaxes around the track but leaves a channel on the order of 0.05 to 0.15 mm across, depending on the silicone hardness and the dwell time of the needle. At 100 °C in a reflux block, that channel passes water vapor steadily. A single puncture on a fresh 9 mm septum typically holds to within 0.02 g over 12 hours. Two punctures in the same spot, or a puncture that lands within 1 mm of a previous one, can triple the loss rate. The track is the dominant path on any method that re-samples from the same vial.
Path 2: the cap-to-neck thread interface
The screw cap seals against the vial finish through the septum, not through the threads. But if the cap is under-torqued, the septum never develops full compression against the land, and vapor escapes around the outside of the septum and down the thread. This path is invisible on a visual inspection and only shows up when you torque-test. A cap that reads below 0.6 N·m on a 24 mm finish is a leak candidate. Above 1.2 N·m you start to deform the cap skirt and lose the seal the other way. The working window is narrow.
Path 3: compression set in the liner
This is the slow one. Silicone and PTFE-faced liners take a permanent deformation under sustained compression at temperature. A liner that seals perfectly at hour one can relax enough by hour eight to open a micro-gap at the land. The property is called compression set, and it is temperature- and time-dependent. A general-purpose silicone liner at 100 °C for 12 hours can take 15 to 25% set. A higher-grade silicone or a silicone/PTFE laminate holds under 10%. That difference is the whole game on long reflux runs.
Path 4: the septum-to-cap crimp or snap fit
On crimp caps, an incomplete crimp leaves the septum sitting proud of the seal ring, and vapor leaks between the septum edge and the aluminum. On snap caps, a cap that has been removed and re-seated loses retention force. I have measured snap caps that held 0.9 N retention on first fit and 0.4 N after three open/close cycles. Below 0.5 N, the septum lifts under internal pressure.
Re-seal behaviour after puncture is the number nobody publishes
Datasheets list septum thickness, durometer, and temperature rating. They do not list re-seal performance, because there is no standard test for it and no vendor wants to publish a number that makes their general-purpose liner look bad. So you have to generate the number yourself, and it is the single most useful spec you can hold for a COD method that samples more than once.
The test is simple. Take ten vials from the lot, cap them to your production torque, puncture each with the method needle, and weigh them at t=0, t=4h, t=8h, and t=12h at reflux temperature. Plot mass loss against time. A septum with good re-seal behaviour gives a flat line after the first hour, because the track closes and stays closed. A poor one gives a line with a steady slope, because the track stays open and vapor keeps passing. I use 0.05 g over 12 hours as the acceptance limit for a single-puncture vial at 100 °C. Anything above that and the septum is not fit for a re-sampled method.
Re-seal also depends on needle geometry. A blunt-ended or side-port needle (common on autosamplers) leaves a cleaner, smaller track than a sharp bevel needle, because it parts the elastomer instead of slicing it. If your method allows, specify a side-port needle and you will cut the track size by roughly a third. That is a cheaper fix than changing septum lots.
Compression set is a temperature-and-time budget, not a material label
The mistake I see most often is treating “silicone septum” as a single specification. It is not. Silicone covers a range of durometers (typically 40 to 60 Shore A), a range of cure chemistries, and a range of compression-set performance. Two septa that both read “silicone, 9 mm, 1 mm thick” can differ by a factor of three in set after 12 hours at 100 °C.
| Liner / septum type | Typical compression set, 12 h at 100 °C | Re-seal after 2 punctures | Max continuous temp | Best use |
|---|---|---|---|---|
| General silicone, 50 Shore A | 15 to 25% | Poor (steady loss) | 200 °C | Single-injection, short runs |
| High-grade silicone, 45 Shore A | 8 to 12% | Fair | 200 °C | Re-sampled methods under 8 h |
| Silicone/PTFE laminate | 5 to 10% | Good | 200 °C | Long reflux, multi-puncture |
| PTFE-faced silicone (one side) | 8 to 15% | Good on face, fair on edge | 200 °C | Solvent resistance plus re-seal |
| Butyl rubber | 10 to 20% | Fair | 120 °C | Low-temperature, low-cost |
Read that table as a starting point, not a guarantee. The set value moves with the actual block temperature and the actual run length. A method that holds 12 hours at 100 °C is a different problem from one that holds 2 hours at 150 °C, even though both are “hot”. Budget your liner choice against the worst case in your method, not the average.
I also check the septum against the vial finish before I trust any of this. A 2 mL vial uses a different finish and septum geometry than a 40 mL COD vial, and a liner that seals well on one will not necessarily seal on the other. Geometry matters as much as material.
Torque is the cheapest lever you have, and most labs ignore it
If you only do one thing after reading this, calibrate your capper. A hand capper set by feel produces a torque spread of 0.3 to 1.4 N·m across a single batch of 24 vials. That spread alone explains most of the lot-to-lot evaporation variance labs blame on septum quality.
The target for a 24 mm screw finish with a silicone/PTFE liner is 0.8 to 1.0 N·m. That is enough to develop full compression on the septum without deforming the cap skirt. Check it with a torque tester on the cap as it comes off the capper, not on the vial after it has been through the oven. Heat relaxes the cap, so a post-run reading is not the value that sealed the vial.
How to set and hold torque
- Set the capper to the low end of the window and measure ten consecutive caps. If any reading falls below 0.7 N·m, increase the setting one increment and re-measure.
- Record the capper setting and the measured torque on the batch record. A capper that drifts between operators is a leak source that never shows up in a CAPA because nobody measures it.
- Re-check torque at the start of every shift and after any capper changeover. The setting moves with cap lot and with ambient temperature.
- Do not re-torque caps after they have been through a run. The liner has already taken set; a second torque does not restore the seal, it just deforms the cap.
For methods that follow a regulated protocol, the torque value should be written into the SOP alongside the capping device, the way a method number is written into the procedure. U.S. EPA SW-846 test methods set the analytical requirement; your SOP sets the hardware that meets it. If you have not connected those two documents, the seal is only as good as the operator’s wrist.
Measuring evaporation loss gravimetrically over a run
You cannot fix what you do not measure, and a titration result is too far downstream to catch a seal problem early. The direct measurement is mass loss on an analytical balance. Here is the protocol I use, and it takes about an hour of bench time plus the run itself.
The gravimetric loss protocol
- Balance: 0.1 mg readability, calibrated, draft shield closed. A 0.01 g balance is not good enough; the losses you care about are in the 0.02 to 0.1 g range.
- Sample set: ten vials from the production lot, filled to the method volume with the method matrix (not water; the matrix matters for vapor pressure).
- Controls: three uncapped vials filled to the same volume (evaporation reference) and three capped vials that will not be punctured (seal reference).
- Weighing schedule: t=0 (immediately after capping), t=1h, t=4h, t=8h, t=12h. Weigh the whole vial, cap on, no wiping. Wipe only at t=0 to remove fill residue from the thread.
- Environment: run the weighings in the same room as the reflux block, or correct for the temperature difference. A vial weighed cold after a hot run pulls in moisture and reads heavy.
- Calculation: mass loss (%) = (m0 − mt) / (m0 − m_empty) × 100, where m_empty is the tare of an identical empty capped vial.
Plot the ten punctured vials as a band, not a single line. The spread tells you as much as the mean. A tight band with a low slope means the seal is consistent. A wide band means the capper or the septum lot is inconsistent, and you should fix that before you chase the mean.
What the numbers should look like
On a 40 mL COD vial at 100 °C for 12 hours, a good seal gives under 0.05 g total loss on a single puncture and under 0.10 g on two punctures. A marginal seal gives 0.10 to 0.25 g. A failing seal gives more than 0.25 g, and at that point the titration is already biased. If your method runs at 150 °C, multiply those limits by roughly 1.8, because vapor pressure rises steeply with temperature.
Convert the loss to a concentration effect before you decide it does not matter. On a 40 mL fill, 0.10 g of water loss is 0.25% of the sample mass. That sounds trivial until you remember the COD calculation divides by the sample volume, and a 0.25% volume change on a 50 mg/L result is 0.12 mg/L. It is small on a single vial. Across a batch where the capper drifted, it is a trend line that fails a QC chart.
Field case 1: the drift that was a capper, not a septum
A wastewater lab running 40 mL COD vials reported a slow upward drift in their low-range results over about six weeks. QC blanks were fine. The standard curve was fine. The drift only appeared on production samples. They had already switched septum lots twice with no change.
I asked them to run the gravimetric protocol on ten production vials. The mass-loss band was wide: 0.03 to 0.31 g over 12 hours. The mean was acceptable at 0.11 g, which is why nobody had flagged it. The spread was the tell. We measured capper torque on the same ten vials and found a range of 0.4 to 1.3 N·m. The capper had been adjusted by a new operator three weeks before the drift started, and the adjustment had moved the setting below the window.
The fix was a capper calibration to 0.9 N·m, a torque check at the start of each shift, and a re-run of the affected batch. The mass-loss band tightened to 0.02 to 0.06 g and the drift disappeared within one QC cycle. The septum lot was never the problem, and two lots had been discarded chasing it.
Field case 2: the second puncture that doubled the loss
A contract lab running a re-sampled COD method (two needle entries per vial, 8 hours apart) saw a 4% low bias on the second sample of each pair compared with the first. They suspected the autosampler. The autosampler was fine.
We ran the gravimetric protocol with a single-puncture control and a double-puncture test set on the same septum lot. Single puncture: 0.04 g mean loss at 12 hours. Double puncture, second entry 1 mm from the first: 0.19 g mean loss. The second track was landing close enough to the first that the two channels merged into one larger opening, and the liner was a general-purpose 50 Shore A silicone with 20% compression set at temperature.
The fix had two parts. First, we moved the autosampler needle position by 2 mm between the two injections so the tracks did not overlap. Second, we switched to a silicone/PTFE laminate septum with set under 10%. Double-puncture loss dropped to 0.07 g, and the second-sample bias disappeared. The 2 mm offset alone accounted for about half the improvement, which tells you geometry is as important as material.
A troubleshooting checklist for seal failure and evaporation loss
Work through this in order. Each step rules out a cause before you spend money on parts.
- Measure first. Run the gravimetric protocol on ten production vials. If total loss is under 0.05 g at 12 hours, the seal is not your problem and you should look elsewhere.
- Check the spread. If the band is wider than 0.05 g across ten vials, the capper or the septum lot is inconsistent. Torque-test the capper before touching the septum.
- Torque-test. Target 0.8 to 1.0 N·m on a 24 mm finish. Below 0.7 or above 1.2 N·m is a leak path. Recalibrate the capper and re-test.
- Count the punctures. If the method re-samples, check whether the second needle entry lands within 1 mm of the first. Offset it by at least 2 mm.
- Check the liner material. If the method runs above 80 °C for more than 6 hours, a general-purpose silicone is the wrong liner. Move to a laminate or high-grade silicone with set under 10%.
- Check the cap fit. On crimp caps, verify the crimp is fully seated. On snap caps, retire any cap that has been opened more than twice.
- Check the needle. A side-port needle leaves a smaller track than a bevel needle. If the method allows, specify side-port.
- Re-test after each fix. Change one variable at a time and re-run the gravimetric protocol. Otherwise you will not know which fix worked.
If you are also seeing coring debris or contamination peaks, that is a separate failure mode and a separate fix; do not conflate it with evaporation loss, because the two have different root causes and different remedies.
What to specify so the seal holds
The decision rule is short. For any COD method that runs above 80 °C for more than 6 hours, or that punctures the same vial more than once, specify a septum with compression set under 10% at the method temperature, cap to 0.8 to 1.0 N·m on a calibrated capper, and verify with a gravimetric loss of under 0.05 g per puncture over the run. For single-injection short runs under 80 °C, a general-purpose silicone septum is fine, and you should not pay for the laminate.
Put the septum specification, the torque value, and the gravimetric acceptance limit in the same SOP that carries the method number. A seal is only as good as the three numbers that define it, and if those numbers live in three different documents, the operator will pick the one that is easiest to hit. The vial is a method component. Treat the seal the same way you treat the calibration curve: measure it, set a limit, and hold it.
If you are building out the rest of the hardware around the vial, the same logic applies to the vial insert and the surrounding HPLC consumables, and it is worth checking the vendor’s published compression-set and re-seal data against your own gravimetric numbers. Vendors like Agilent, Thermo Fisher, and Restek publish useful technical notes on septa and sample handling; use them as a starting point, not a substitute for your own 12-hour mass-loss run. The run is the only number that reflects your method, your capper, and your oven.
References & further reading
Related resources from our lab consumables network:
- Cod Vial at codvial.com
- Vial Insert at hplcinsert.com
- 2Ml Vial at 2mlsamplevial.com
- Hplc Consumables at hplcconsumable.com
External references and standards: