A 96-well plate run on a Tuesday afternoon came back with toluene recoveries at 71 to 78% against a 95% acceptance floor. The analyst had run the same method the week before at 96%. Nothing in the instrument log had changed. The difference was a new box of 0.45 µm PTFE membrane filters, and the seal between filter and vial cap was letting the light end of the extraction evaporate before the autosampler ever saw it. That is the failure mode this page is about: PTFE membrane filter seal failure, evaporation and sample loss, measured in milligrams per hour rather than guessed at from a bad chromatogram.
If you already know what a PTFE membrane is and why it is in your method, you do not need the chemistry lecture. What you need is a way to find the leak path, a way to prove it gravimetrically, and a rule for when a filter that “looks fine” is actually costing you 20% of your analyte.
A filter is not a passive disc. It is a mechanical joint with a gasket, a housing, and a downstream volume of headspace. Every one of those is a potential evaporation route. I have chased all four of the following on real benches, and they behave differently enough that you need to separate them before you fix anything.
Path 1: the membrane-to-housing compression face
The PTFE disc sits on a support (usually polypropylene or high-density polyethylene) and is held down by the housing rim or a cap septum. PTFE is soft in compression but it does not flow like silicone. It creeps. At 25 °C a clamped PTFE disc under 1.5 N·mm⁻² will hold its seal for weeks. At 60 °C in a heated autosampler tray, the same disc relaxes and the clamp load drops by 15 to 30% over an eight-hour sequence. That relaxation is the compression set problem, and it is the single most common cause of late-run evaporation drift I deal with.
Path 2: the puncture site after a needle passes through
If your workflow pushes a syringe needle or a transfer cannula through the membrane (common in headspace and in some purge-and-trap setups), you have made a hole. PTFE does not re-seal the way a rubber septum does. Its elastic recovery is poor at room temperature and worse above 40 °C. A 0.5 mm needle through a 0.2 mm PTFE disc leaves a permanent aperture on the order of 0.3 to 0.6 mm. That is a diffusion channel, not a pinhole, and it will bleed solvent vapour for the rest of the run.
Path 3: the housing-to-vial interface
The filter housing threads or snaps onto the vial neck. If the mating surface is a moulded polymer with a 0.1 mm mismatch in flatness (and cheap housings run worse than that), you get a spiral leak path between the two parts that is invisible on inspection. This is where a lot of 2ml vial users get burned: the vial is fine, the filter is fine, but the two together do not seal because the cap skirt bottoms out before the septum compresses.
Path 4: the membrane itself, if the pore structure is wetted
A dry PTFE membrane has a high vapour transmission rate. That is one of its useful properties when you want it, and a liability when you are trying to hold a volatile analyte in a closed vessel. If the membrane is wetted by the solvent (which depends on your matrix and any surfactant), the pores fill and vapour transport drops sharply. If it stays dry, vapour crosses the disc and escapes through any imperfect downstream seal. This is why the same filter gives you 96% recovery on one matrix and 74% on another.
Re-seal behaviour after puncture: what PTFE will and will not do
I want to be blunt here because it saves people weeks. PTFE does not re-seal a puncture. It is not a septum. A septum (silicone, or silicone-faced PTFE) is designed to close around a needle and recover. A PTFE membrane is designed to filter. If your method punctures the membrane, plan for a permanent leak and size your run time accordingly.
What actually happens after puncture, measured on a 47 mm, 0.45 µm PTFE disc with a 0.5 mm needle and a 2 mL headspace above it:
- First 30 minutes: mass loss is dominated by the puncture, roughly 0.8 to 1.4 mg·h⁻¹ for methanol at 25 °C.
- 1 to 4 hours: the aperture edges relax slightly (PTFE creep), the effective hole diameter grows by maybe 10%, and loss rate rises to 1.0 to 1.8 mg·h⁻¹.
- Beyond 4 hours: steady state. The hole is what it is. Loss continues linearly with time.
The practical consequence: a puncture you make at injection 1 is still leaking at injection 96. If you are running a 12-hour sequence, that is 12 hours of loss through a hole you cannot close. The fix is not a better filter. The fix is to stop puncturing the membrane, or to accept the loss and correct for it with an internal standard that evaporates at the same rate.
When a PTFE-faced septum is the right answer
If your workflow genuinely needs a needle through a barrier, use a silicone septum with a PTFE face rather than a bare PTFE membrane. The silicone provides the elastic recovery; the PTFE face provides the chemical inertness. Puncture counts matter: a 20 mm silicone/PTFE septum holds a seal for roughly 50 to 80 punctures at 0.5 mm needle gauge before the leak rate climbs past 2 mg·h⁻¹. Track your punctures. Replace on count, not on appearance.
Compression set: the number that predicts late-run drift
Compression set is the permanent deformation a gasket takes after being squeezed and released. For PTFE it is measured per ASTM D395 (method B, constant deflection). A virgin PTFE disc at 25 °C and 25% deflection recovers to within about 15 to 20% of its original thickness after 22 hours. At 60 °C the set climbs to 30 to 40%. That means a disc that started at 0.5 mm is now 0.3 to 0.35 mm thick at the seal face, and your clamp load has fallen with it.
This is why the drift is always at the end of the run, not the beginning. The seal is fine for the first two hours and starts leaking at hour five. If you only ever check recovery on the first few injections, you will never see it.
| Condition | Compression set (ASTM D395, 22 h) | Expected mass loss, 2 mL headspace, methanol, 25 °C | What it means for a 12 h run |
|---|---|---|---|
| Virgin PTFE, 25 °C, 25% deflection | 15 to 20% | 0.3 to 0.5 mg·h⁻¹ | 3.6 to 6 mg total, usually tolerable |
| Virgin PTFE, 60 °C, 25% deflection | 30 to 40% | 0.9 to 1.4 mg·h⁻¹ | 11 to 17 mg total, fails a 95% floor |
| PTFE-faced silicone septum, 25 °C | 8 to 12% (silicone dominates) | 0.1 to 0.2 mg·h⁻¹ | 1.2 to 2.4 mg total, comfortably inside spec |
| Punctured bare PTFE (0.5 mm needle) | n/a, aperture is permanent | 1.0 to 1.8 mg·h⁻¹ | 12 to 22 mg total, recovery collapses |
Those numbers are from my own bench, gravimetric, on a five-place balance, and they will shift with your solvent and headspace volume. The ratios are what matter. A hot tray plus a bare PTFE membrane is a losing combination.
How to measure the loss gravimetrically over a run
You cannot fix what you have not weighed. The good news is that the measurement is cheap and takes an afternoon. Here is the procedure I use, and it has never failed to separate a real leak from a chromatographic artefact.
The rig
- A five-place analytical balance (0.01 mg readability), draft shield closed, levelled, on a stone bench.
- Six replicate sealed vessels per condition: your filter housing plus COD vial or sample vial, capped exactly as in your method.
- One control per condition: same vessel, no filter, to separate the cap leak from the filter leak.
- A solvent of known vapour pressure. Methanol (16.9 kPa at 25 °C) is a good stress test. Water (3.2 kPa) is the gentle one.
- A temperature-controlled environment. A bench that swings 3 °C over the run will swamp your signal.
The procedure
- Fill each vessel with 1.000 ± 0.005 g of solvent. Cap. Weigh immediately (t = 0).
- Weigh at t = 0, 0.5, 1, 2, 4, 8, 12 h. Return each vessel to its tray position between weighings. Handle with forceps, not fingers.
- Plot cumulative mass loss against time. A good seal gives a flat line near zero. A leaking seal gives a straight line with a slope in mg·h⁻¹.
- The slope is your loss rate. Multiply by run time and divide by starting mass to get fractional loss. That fraction is your worst-case recovery penalty.
- Run the no-filter control. If the control also loses mass, your cap or vial is the problem, not the membrane.
For reference, a 2 mL aqueous sample in a well-sealed 2 mL vial with a PTFE-faced septum loses about 0.05 to 0.1 mg·h⁻¹ at 25 °C. If your PTFE membrane setup is losing 1 mg·h⁻¹, you are 10 to 20 times worse than a plain capped vial, and you have found your problem.
Field case 1: the 71% toluene recovery
Symptom: purge-and-trap run on soil extracts, toluene recovery dropped from 96% to 71 to 78% across a batch. No instrument change, no reagent change.
What identified it: we ran the gravimetric rig above on the new filter lot versus the old one. The new lot lost 1.3 mg·h⁻¹; the old lot lost 0.2 mg·h⁻¹. The difference was housing flatness: the new mould had a 0.15 mm step at the seal face, so the PTFE disc never fully compressed. The control (no filter) lost 0.08 mg·h⁻¹, confirming the cap was fine.
The fix: switched to a housing with a machined flat seal face (flatness within 0.02 mm), and added a torque spec of 1.2 N·m on the cap, checked with a calibrated torque driver. Recovery came back to 94 to 97%. We now buy filters with a flatness certificate and reject lots that fail.
Field case 2: the late-run drift on a 12-hour sequence
Symptom: a 96-position autosampler sequence showed recovery falling from 98% at injection 1 to 82% at injection 96. Early injections were fine. The analysts blamed the detector.
What identified it: we weighed the vials at t = 0 and t = 12 h. The first 24 positions (run in the first 3 hours) lost 0.4 mg each. The last 24 positions (sitting in the tray for 9 to 12 hours) lost 8.1 mg each. The tray was held at 45 °C, and the PTFE discs had taken a 35% compression set by hour six. The seal failed progressively, exactly as the compression-set table predicts.
The fix: dropped the tray to 25 °C, switched to PTFE-faced silicone septa, and shortened the sequence to two 6-hour blocks with a re-cap between them. Late-run recovery came back to 95 to 98%. The lesson: the tray temperature is part of the seal spec, not a convenience setting.
A troubleshooting checklist you can run in one shift
Work top to bottom. Stop at the first item that fails.
- Is the loss in the control (no filter)? If yes, the cap or vial is leaking. Check cap torque (target 1.0 to 1.4 N·m for a 20 mm cap) and septum condition.
- Does the loss rate scale with tray temperature? Run the rig at 25 °C and at your tray temperature. If the rate more than doubles, you have a compression-set problem. Lower the tray or change the gasket material.
- Is the membrane punctured? Count needle passes. If the membrane sees a needle, replace it with a PTFE-faced septum and track puncture count against the 50 to 80 limit.
- Is the housing face flat? Check the seal face with a straight edge and feeler gauge. Anything over 0.05 mm of step is suspect.
- Is the loss linear or does it curve upward? Linear from t = 0 means a mechanical leak (puncture, flatness). Curving upward after hour 3 to 5 means compression set. They need different fixes.
- Does the loss change with solvent? If methanol leaks and water does not, your seal is vapour-limited, not liquid-limited. That points at the membrane and the downstream seal, not the sample.
For method-level context on what a regulated method expects from the sample-handling chain, the U.S. EPA SW-846 test methods are the reference most of my clients work against, and the seal requirements in them are implicit rather than spelled out. You have to build the gravimetric check yourself.
Selection criteria that actually prevent seal failure
Most of the seal failures I see are bought, not made. The filter was specified on pore size and diameter and nothing else. Here is what to put on the purchase spec instead.
Gasket and seal material
Bare PTFE is the wrong choice for any application where the seal must hold volatile analyte for more than two hours at elevated temperature. Specify a PTFE-faced silicone or PTFE-encapsulated FKM gasket for those. Keep bare PTFE for cold, short, or non-volatile workflows where its inertness is the point.
Housing flatness and clamp load
Ask for the seal-face flatness (target ≤ 0.02 mm) and the recommended clamp torque. A supplier who cannot give you either is selling you a leak. For 20 mm caps, the working range is 1.0 to 1.4 N·m; below 0.8 N·m you will lose seal at temperature, above 1.6 N·m you will distort the cap and crack the skirt.
Temperature rating
PTFE itself is good to 260 °C continuous. The housing and gasket are not. Polypropylene housings soften above 90 °C; silicone gaskets are good to about 200 °C but take compression set above 100 °C. Match the rating to your tray temperature, not to the membrane’s rating.
Puncture specification
If the workflow punctures, specify a septum rated for a puncture count and buy on that number. A vial insert with a pre-slit septum can cut puncture count in half and is worth the cost on high-throughput sequences.
For instrument-side compatibility, the sample-handling guidance from Agilent’s chromatography consumables and Thermo Fisher’s sample handling pages are useful for matching vial and septum geometry to your autosampler. Restek’s technical resources carry good reference data on septum bleed and puncture life that complements the gravimetric approach here.
The decision rule
If your method holds a volatile analyte for more than two hours at a tray temperature above 30 °C, do not seal it with a bare PTFE membrane filter. Use a PTFE-faced silicone septum, cap at 1.0 to 1.4 N·m, keep the tray at 25 °C, and verify the seal gravimetrically at t = 0 and t = end-of-run before you trust a single recovery number. If you cannot weigh the loss, you cannot claim the seal, and a recovery you did not measure is a recovery you do not have.
References & further reading
Related resources from our lab consumables network:
- Ptfe Membrane Filter at ptfemembranefilter.com
- Cod Vial at codvial.com
- Vial Insert at hplcinsert.com
- 2Ml Vial at 2mlsamplevial.com
External references and standards:
