The auditor picked lot PT-2409-117, a 47 mm, 0.45 µm PTFE membrane, and asked for the calibration certificate of the balance that weighed the 240 g of polymer going into that batch. The lab produced a certificate. It was for a different balance, in a different building, and the calibration had expired 11 days before the extrusion run. That single broken link turned an otherwise clean ISO 17025 assessment into a finding, and the finding was not about the membrane’s performance. The membrane passed every test. The paperwork chain had a hole in it. That is what ISO 17025 traceability for a PTFE membrane filter really means: not whether the filter works, but whether you can prove, link by link, where its numbers came from.
Most labs I visit can trace a volumetric flask to a national standard without thinking. Ask them to trace a consumable, and the chain falls apart somewhere between the purchase order and the polymer lot. This article is about what an unbroken traceability chain looks like for a PTFE membrane filter, and specifically about the one link that most labs cannot produce when the assessor is sitting across the table.
The traceability chain for a consumable has six links, and most labs only hold four
Traceability under ISO 17025 is not a single document. It is a chain of custody for a measurement result. For a disposable membrane filter, the chain runs from the balance that weighed the raw polymer all the way to the balance in your lab that weighed the residue. I count six links. When I audit a supplier or a lab, I walk them in order.
- Link 1, raw material lot identity. The polymer resin lot number, the certificate of analysis from the resin maker, and the date it entered the supplier’s inventory.
- Link 2, process measurement instruments. The balances, thermocouples, pressure gauges, and thickness gauges used during extrusion, stretching, and sintering, each with a current calibration certificate traceable to a national metrology institute.
- Link 3, in-process test data tied to the lot. Bubble point, thickness, and gravimetric pore-size data recorded against the specific lot, not a generic product datasheet.
- Link 4, the finished-goods certificate of analysis. A document that references the lot and states the measured values, not just the specification limits.
- Link 5, the custody record. Shipping, storage, and receipt, including any temperature excursion log between the supplier’s dock and your shelf.
- Link 6, your in-house verification. The acceptance test you ran when the lot arrived, on your own calibrated instrument, recorded against the lot number.
In my experience, most labs hold links 4, 5, and 6 solidly. Link 1 is often present but vague. Links 2 and 3 are where the chain breaks, and link 2 is the one that breaks hardest.
The link most labs cannot produce is the calibration behind the supplier’s in-process numbers
Here is the failure I see over and over. An assessor asks: “The certificate says this lot has a minimum bubble point of 1.4 bar in isopropanol. What instrument measured that, and when was it last calibrated?” The lab manager hands over the supplier’s CoA, which states the value but says nothing about the instrument. The lab emails the supplier. The supplier sends a generic quality manual and a statement that “all instruments are calibrated annually.” That is not traceability. That is an assertion.
ISO 17025 clause 6.5 requires that measurement results be metrologically traceable to the SI, through an unbroken chain of calibrations each contributing to the measurement uncertainty. A supplier’s internal bubble point gauge, if it is calibrated against a transfer standard that is itself calibrated by an accredited lab, satisfies this. If it is calibrated “in house” against a reference gauge of unknown provenance, it does not. The distinction is invisible on the CoA. You have to ask for the calibration certificate of the bubble point reference standard, and you have to check that the calibrating body is accredited to ISO 17025 for pressure.
I have seen a supplier produce a certificate for a pressure gauge calibrated by a body whose scope did not include pressure, only temperature. The certificate looked official. It was worthless for this purpose. That is the link most labs cannot produce.
What to actually request from the supplier
When I qualify a membrane supplier now, I ask for a specific document set, and I ask for it before the first order, not during an audit. The request is short.
- The calibration certificate for the bubble point reference standard, showing the calibrating lab’s ISO 17025 accreditation number and the pressure range covered.
- The calibration certificate for the analytical balance used in gravimetric pore-size and thickness work, with the uncertainty stated.
- The calibration certificate for the thickness gauge, since a 0.45 µm membrane with a ±5 µm thickness tolerance is only meaningful if the gauge resolves to 1 µm.
- A statement of the measurement uncertainty for each in-process test, not just the pass/fail result.
A supplier who cannot produce these is not necessarily making a bad membrane. But they are asking you to accept their numbers on faith, and under ISO 17025 you cannot.
Why the certificate of analysis is not the same as a traceability record
This trips up more labs than any other single point. A CoA states that the lot met a specification. A traceability record states how you know. They are different documents with different purposes, and a CoA that does not reference the instruments behind its numbers is a claim, not evidence.
I have a rule I apply when reviewing a new supplier’s paperwork. If the CoA could have been written before the lot was made, by filling in a template, it is not traceability. A genuine record has lot-specific numbers in it: an actual measured bubble point of 1.52 bar rather than a checkmark against a 1.4 bar minimum, an actual thickness of 51 µm rather than “within tolerance.” The presence of real measured values is a rough proxy for whether the supplier is running the tests or just certifying them.
The same logic applies downstream. When you receive the lot and run your own acceptance test, that result is link 6. It is only traceable if your own instrument is calibrated and the result is recorded against the lot number, not just filed in a notebook by date. A PTFE membrane filter that arrives with a thin CoA and no calibration backup is a filter you are accepting on trust. That is a defensible position for a research lab. It is not defensible for a lab issuing accredited results.
Acceptance testing on receipt is the link you control, so make it count
You cannot fix a supplier’s broken link 2. You can make link 6 strong enough that the assessor sees you did not simply accept the shipment. The receipt test is where you demonstrate control. Here is the checklist I use, with the numbers that matter.
| Receipt test | Method | Acceptance criterion | Instrument and calibration |
|---|---|---|---|
| Bubble point | Isopropanol wet, per ASTM F316 | ≥ 1.4 bar for 0.45 µm; ≥ 2.4 bar for 0.22 µm | Calibrated pressure gauge, ±0.5% full scale |
| Thickness | Micrometer, 5 points per disc | Within ±5 µm of nominal; RSD ≤ 3% | Gauge resolving to 1 µm, calibrated annually |
| Diameter | Caliper, 3 orientations | 47.0 ± 0.5 mm | Caliper, 0.02 mm resolution |
| Gravimetric extractables | Weigh dried disc before and after extraction | ≤ 0.5 mg per 47 mm disc | Balance, 0.01 mg readability, calibrated |
| Visual | Backlit inspection | No pinholes, no creases, no edge tears | None required |
The extractables test is the one most labs skip and the one that most often catches a drifting lot. A 47 mm disc of 0.45 µm PTFE typically extracts well under 0.5 mg, and a lot that suddenly reads 1.2 mg is telling you something changed in the sintering step. That is a traceability signal, not just a quality signal.
Record the instrument, not just the result
When you log the acceptance test, log the balance ID and the calibration due date alongside the result. An assessor who sees a bubble point of 1.51 bar with no instrument reference will ask for it. If the reference is already in the record, the question never comes. I have watched labs save themselves an hour of scrambling by putting the instrument serial number in the same row as the measurement.
Two field cases where the chain broke, and how the measurement found it
Case one: the bubble point that drifted 0.3 bar across three lots
A contract lab running EPA Method 3510C separatory extractions noticed that recovery of a semi-volatile surrogate dropped from 92% to 81% over six weeks. Nothing in the method had changed. The measurement that identified the cause was a routine receipt bubble point on three consecutive lots of 47 mm, 0.45 µm membrane. The values were 1.48, 1.31, and 1.19 bar. The specification minimum was 1.4 bar. The third lot was out of spec on arrival and had been accepted because the lab was only checking the CoA, not running the test.
The fix was twofold. First, the lab started running the receipt bubble point on every lot, which caught the drift at lot two instead of lot three. Second, and more relevant here, the lab asked the supplier for the calibration certificate behind the bubble point reference standard. The certificate was three years old and the gauge had a stated drift of 0.02 bar per year. The supplier’s in-process numbers had been reading high by roughly 0.06 bar, which is why lots that were actually marginal had been passing. The traceability link was not just missing. It was actively misleading. That is the worst kind of broken chain, because it looks intact until you measure against it.
Case two: the balance that weighed the wrong thing
An environmental lab doing gravimetric extractables on PTFE discs found that one lot read 0.9 mg per disc against a spec of 0.5 mg. The lab rejected the lot and asked the supplier to investigate. The supplier’s investigation found that the extractables balance had been calibrated, but the calibration was performed at 200 g, and the discs weighed about 0.4 g. The balance was linear to 0.02 mg at 200 g but had a 0.06 mg offset at the low end of its range. The offset was within the balance’s stated specification but not within the uncertainty the supplier had assumed for the test.
The fix was to specify a low-range calibration point for the extractables balance, at 1 g rather than 200 g, and to state the measurement uncertainty at that point in the CoA. The lot was re-tested and passed at 0.4 mg. The membrane was fine. The traceability chain had a link that was calibrated to the wrong range, which is a subtle failure that a pass/fail CoA would never reveal.
Both cases share a pattern. The membrane was not the problem. The measurement behind the membrane’s numbers was. That is why ISO 17025 traceability for a PTFE membrane filter is really a question about instruments, not about polymer.
How to structure the traceability record so an assessor finds it in under five minutes
The best traceability chain in the world is useless if the assessor cannot follow it. I have seen labs with excellent records lose points because the documents were scattered across three systems and a filing cabinet. Structure the record so the path is obvious.
Attach the supplier’s calibration certificates to the lot record at receipt, not in a separate quality folder. When the assessor pulls lot PT-2409-117, the calibration certificates for the bubble point reference and the balance should be one click away, not a request to the supplier and a three-day wait. The same applies to your own acceptance test: the instrument ID, the calibration due date, and the raw measurement belong in the same record as the result.
For labs running HPLC or GC methods that use the membrane as a sample-prep step, the traceability of the filter feeds directly into the traceability of the result. If you are filtering into a 2 mL vial and the method requires the filter to be traceable, then the vial lot and the filter lot both belong in the run record. A COD vial used in a colorimetric method has the same requirement, and a vial insert for small-volume injections is no different. The consumable chain is only as strong as its weakest documented link, and the filter is usually the link that gets the least attention because it is the cheapest item in the chain.
A practical document set for the lot file
This is what I put in a lot file now, in this order, so the assessor walks the chain without asking.
- Purchase order and receipt date, with the lot number on the packing slip.
- Supplier CoA with measured values, not just pass/fail.
- Supplier calibration certificates for the bubble point reference and the analytical balance, with accreditation numbers visible.
- Your receipt acceptance test, with instrument ID and calibration due date recorded against each measurement.
- Storage condition log between receipt and first use, if the method requires it.
Five documents. If all five are present and current, the chain is unbroken and the assessor moves on. If any one is missing, that is the finding.
The decision rule: buy the traceability, not just the membrane
When you specify a PTFE membrane filter for an accredited method, the specification is not just 47 mm, 0.45 µm, and a bubble point minimum. The specification includes the supplier’s ability to produce a calibration certificate for the instrument behind every number on the CoA, traceable to an accredited calibration body, with the measurement uncertainty stated. If the supplier cannot produce that document on request, the membrane is not qualified for accredited work, regardless of how well it performs. Ask for the certificate before the first order. If it arrives within a day and the accreditation scope covers pressure, you have a supplier you can build a chain around. If it arrives after three emails and a phone call, you have found the link that will break during your next assessment, and you have found it early enough to do something about it.
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:
