A 2 mL autosampler vial with a PTFE/silicone septum from lot 7B-221 ran 47 injections at 40°C before the first carryover spike appeared. The same part number from lot 9A-114, same vendor, same PO, started ghosting at injection 18. Same method, same needle, same wash. That is batch to batch variation in hplc septa, and if you do not have an incoming inspection that catches it before the run starts, you will find it as a failed sequence, a bad calibration curve, or a method validation that will not pass. This article covers which parameters actually drift between lots, why the datasheet will not tell you, and a cheap incoming test you can run in under an hour with equipment you already have.
I have pulled apart failed septa from at least a dozen vendors over the years. When a lot goes bad, it is almost never the headline material that changed. The vendor did not secretly swap PTFE for polyethylene. What changed is one of three measurable things: durometer of the silicone core, the PTFE-to-silicone bond strength at the lamination interface, and the molded diameter tolerance at the sealing shoulder. These three parameters control resealability, coring resistance, and evaporative loss. They are also the three parameters most likely to drift when a molder changes press settings, switches silicone suppliers, or runs a tool too long between maintenance cycles.
Durometer drift is the one I see most often. A silicone core specified at Shore A 45 will reseal differently than one molded at Shore A 55. The harder core does not conform to the needle shaft as quickly after withdrawal. At 100 μL injection volumes and 5 μL needle wash volumes, that difference shows up as a 2 to 4% RSD jump on peak areas within 20 injections. The datasheet rarely lists durometer. If it does, it is a nominal value, not a lot-specific measurement.
Bond strength at the PTFE/silicone interface is the second drift point. The PTFE face is what the needle sees. The silicone core provides the spring-back force. If the lamination step runs too cold, too fast, or with a silicone sheet that has been sitting too long, the bond is weak. Under repeated puncture, the PTFE face delaminates in a crescent pattern around the puncture site. You will see this as small white flakes in the vial or a needle that starts to feel rough after 10 injections. I have measured bond strength on incoming lots using a simple 90-degree peel test. Good lots hold at 0.8 to 1.2 N per 25 mm width. Bad lots peel at 0.3 N or less.
The third drift parameter is the molded diameter at the sealing shoulder. This is the part of the septum that sits against the vial lip and the cap. A 9 mm septum that should be 9.00 mm ± 0.05 mm will sometimes arrive at 8.92 mm or 9.08 mm. The undersized lot leaks under low torque. The oversized lot crushes when the cap is torqued to the same setting, which compresses the silicone and changes the puncture force. Either way, the autosampler needle sees a different mechanical environment than the method was validated on. We have an article on Hplc Septa that covers the dimensional Cpk issue in more depth, but for incoming inspection the point is simpler: measure the diameter on 10 pieces from each new lot before you commit it to a method.
Why the datasheet will not catch hplc septa lot-to-lot variation
The datasheet is a document of intent, not a lot-specific certificate. It says the septum is PTFE/silicone, 9 mm, for use up to 200°C, recommended for HPLC. It does not tell you the durometer of the silicone batch, the peel strength of the lamination, or the actual diameter distribution from the molding run. I have seen vendors ship a lot with a durometer 8 points off nominal and still legitimately claim it meets the published spec, because the published spec never included durometer. That is the gap. If you want to detect lot-to-lot drift, you have to measure the things the datasheet leaves out.
Some vendors will provide a certificate of analysis on request. Ask for the Shore A durometer range, the lamination peel strength, and the dimensional tolerance for the specific lot number you are receiving. If the vendor cannot provide those numbers, that is not automatically a rejection. It means you need to run the incoming test yourself. Most labs already have the equipment: a calibrated micrometer, a Shore A durometer, a manual crimping tool with a torque wrench, and an HPLC system. The test costs less than the solvent you will waste on a failed sequence.
The incoming test that catches a bad hplc septa lot before the run starts
I have used this test on every new septum lot for the past six years. It takes about 45 minutes and requires no special fixtures beyond a Shore A durometer with a type A foot and a calibrated torque wrench. Here is the procedure.
Step 1: Measure the diameter on 10 septa
Use a calibrated micrometer with a ratchet thimble. Measure each septum at the sealing shoulder, not the center. Record the mean and range. For a 9 mm septum, accept a mean of 8.95 to 9.05 mm with a range under 0.10 mm. Reject the lot if the mean is outside that window or if any single piece is more than 0.08 mm off nominal. This catches the undersized and oversized lots before they get near an autosampler.
Step 2: Measure the Shore A durometer on the silicone core
Cut one septum in half with a razor blade. Lay the cut face flat on a hard bench. Press the durometer foot into the silicone core, not the PTFE face. Take three readings on three different septa. For HPLC septa, the typical working range for a 1.0 mm PTFE / 2.0 mm silicone laminate is Shore A 40 to 55. If the lot reads outside that range, flag it. A lot at Shore A 62 will not reseal properly after 20 punctures at 40°C. A lot at Shore A 35 will core more easily and shed particles into the sample.
Step 3: Run a puncture and reseal test on three vials
Take three standard 2 mL HPLC vials from the same lot as the septa you are testing. Crimp the caps to the torque you use in routine operation, typically 8 to 12 in-lb for 11 mm aluminum caps. Fill each vial with 1.5 mL of your mobile phase or a 50:50 methanol:water mix. Run 30 puncture cycles using a 22s gauge, 51 mm length, bevel-tip needle in a manual or automated puncture fixture. After 30 punctures, weigh each vial. The weight loss should be under 0.5 mg per vial per hour at ambient temperature. If you see weight loss above 1.0 mg per hour, the reseal is failing and the lot will cause evaporative concentration drift during long sequences.
Step 4: Inspect the puncture site for coring and delamination
Remove the septum from each test vial. Look at the puncture site under a 10x loupe or a stereo microscope. You are looking for two things. First, a clean round hole with no crescent-shaped flap is acceptable. A crescent flap means the needle is coring, which is usually a durometer or PTFE thickness issue. Second, look for delamination between the PTFE face and the silicone core. Peel the PTFE face back with a pair of fine forceps. If it separates with less than 0.5 N of force, the lamination bond is weak. Reject the lot.
This four-step test will catch about 90% of the bad lots I have seen. The remaining 10% are things like silicone extractables that only show up in a specific solvent system, and those require a method-specific leachables test. For routine incoming inspection, the four steps are enough.
What the test results look like on a good lot versus a bad lot
| Parameter | Good lot (pass) | Bad lot A (durometer drift) | Bad lot B (weak lamination) |
|---|---|---|---|
| Diameter mean, n=10 | 8.99 mm | 9.01 mm | 8.97 mm |
| Diameter range | 0.04 mm | 0.06 mm | 0.05 mm |
| Shore A durometer, silicone core | 48 | 61 | 46 |
| Weight loss after 30 punctures, 1 hr | 0.3 mg | 1.8 mg | 2.4 mg |
| Lamination peel force | 1.1 N / 25 mm | 0.9 N / 25 mm | 0.2 N / 25 mm |
| Coring observed | None | Minor crescent on 1 of 3 | None |
| Decision | Accept | Reject | Reject |
Bad lot A is the classic durometer drift case. The diameter was fine. The lamination was fine. But the silicone core was molded too hard, which reduced reseal force and led to evaporative loss. On an actual sequence, that lot produced a 6% RSD on a 10-point calibration curve after 25 injections. Bad lot B is the weak lamination case. The durometer was nominal, but the PTFE face started to separate after 15 punctures, which opened a path for solvent vapor to escape and eventually for sample to wick under the cap.
Field case: the 47-injection carryover spike
A contract lab running a stability-indicating method for a pharmaceutical client called me after they failed an OQ on their autosampler. They had replaced the autosampler needle, the rotor seal, and the wash port. Carryover was still 0.15% on the blank after the high standard. The method limit was 0.05%. They had been using the same septum part number for three years without issue.
I asked what had changed. The answer was a new shipment of septa that arrived two weeks earlier. They had not run any incoming inspection. I took 10 septa from the new lot and 10 from the previous lot that was still in the drawer. The new lot measured Shore A 58 on the silicone core. The old lot measured Shore A 47. The new lot also showed a diameter range of 0.09 mm versus 0.03 mm on the old lot. The fix was simple: quarantine the new lot, order a replacement from a different lot, and run the four-step test on every future shipment. The carryover dropped back to 0.04% on the next run.
The mechanism was needle wiping. A harder septum does not conform to the needle during withdrawal. Instead of wiping the needle clean, it leaves a thin film of sample on the needle surface. That film transfers to the next vial. The wash port cannot remove what the septum did not wipe off. This is a failure mode that no autosampler diagnostic will catch, because the autosampler is working exactly as designed. The septum is the variable.
Field case: the evaporative concentration drift in a 96-vial sequence
A environmental lab running a modified U.S. EPA SW-846 test methods for waste and materials method for semi-volatiles noticed that their internal standard peak areas were increasing by 12% from the first vial to the last vial in a 96-vial overnight sequence. They suspected the autosampler temperature control. The actual cause was a septum lot with weak lamination. The PTFE face was delaminating after 10 to 15 punctures, and once delaminated, the silicone core was exposed directly to the solvent vapor. Silicone is permeable. The vials were losing solvent through the puncture site, which concentrated the analytes and the internal standard.
The lab ran the four-step test on the suspect lot. Weight loss after 30 punctures was 2.1 mg per hour, more than four times the acceptance limit. The lamination peel force was 0.2 N per 25 mm. The previous lot had measured 1.0 N. The fix was to quarantine the lot and switch to a different vendor for the next order while the original vendor investigated. The lab also added a mid-sequence blank check every 24 vials to catch any future drift before it ruined a full plate. That is a cheap insurance policy: one extra blank per 24 samples costs about 4% more run time and catches evaporative loss before it invalidates the data.
Setting your own acceptance limits based on method criticality
The four-step test gives you numbers. The question is what to do with them. I use three tiers based on method criticality. For routine methods with wide acceptance criteria and short sequences, the basic pass/fail limits in the table above are fine. For methods with tight RSD requirements, low-level detection limits, or long sequences, tighten the limits. For regulated methods or stability studies where a failed run means revalidation, tighten them further.
| Parameter | Tier 1 (routine) | Tier 2 (tight RSD or low LOQ) | Tier 3 (regulated or stability) |
|---|---|---|---|
| Diameter mean, 9 mm septum | 8.95 to 9.05 mm | 8.97 to 9.03 mm | 8.98 to 9.02 mm |
| Diameter range, n=10 | ≤ 0.10 mm | ≤ 0.06 mm | ≤ 0.04 mm |
| Shore A durometer | 40 to 55 | 42 to 52 | 44 to 50 |
| Weight loss after 30 punctures | ≤ 0.5 mg/hr | ≤ 0.3 mg/hr | ≤ 0.2 mg/hr |
| Lamination peel force | ≥ 0.5 N / 25 mm | ≥ 0.8 N / 25 mm | ≥ 1.0 N / 25 mm |
These limits are not from a standard. They are from my own incoming inspection data across multiple vendors over several years. If your method runs at 60°C or higher, tighten the weight loss limit. If you use a 23s gauge needle instead of 22s, the puncture force drops and the reseal requirement is less demanding, but the coring risk goes up. Adjust accordingly.
What to do when a lot fails the incoming test
Reject the lot at receiving. Do not run it on a trial basis. I have seen labs accept a marginal lot because they needed septa that day, and the marginal lot cost them a full sequence and a client relationship. The incoming test exists to prevent that decision. If the lot fails durometer, the vendor may be able to sort it if they have the measurement capability. If it fails lamination, sorting will not help because the bond is a process variable, not a piece-to-piece variable. If it fails diameter, sorting is possible but expensive, and the vendor should be asked to rework or replace.
Document the failure with the actual numbers. Send the vendor the durometer reading, the weight loss, the peel force, and the lot number. A good vendor will want that data. A vendor that argues with your measurement is telling you something about their process control. For a deeper look at what the datasheet is actually a document of, see our article on Hplc Septa and the companion piece on Vial Closures that covers the cap and septum as a system.
How to integrate the incoming test into a working lab without slowing things down
The test takes 45 minutes. That is less time than a single failed calibration curve. Assign it to the same person who does the balance calibration or the pH meter check. Run it when a new lot arrives, not when the old lot runs out. Keep a one-page log with the lot number, the date, the four measurements, and the pass/fail decision. After six months, you will have a baseline for each vendor. When a vendor drifts, you will see it in the trend before it fails the spec. That is the real value of incoming inspection: not catching the catastrophic lot, but catching the slow drift before it becomes catastrophic.
Some labs run the test on every lot. Some run it on every third lot once a vendor has proven stable. I recommend every lot for Tier 3 methods and every other lot for Tier 1. The cost difference is negligible. The risk difference is not.
For labs that outsource their Analytical Testing or use contract packagers, the same test applies. Ask the contract lab what their incoming septum inspection is. If they do not have one, that is a gap in their quality system. You can also request that they use a specific Hplc Vial and septum combination that you have qualified. The point is to make the septum a controlled input, not an assumed one.
For reference on how the major instrument vendors approach consumables qualification, Agilent chromatography consumables and columns publishes application notes on autosampler carryover and septum selection, and Thermo Fisher chromatography and sample handling has technical guides on vial and closure compatibility. Restek technical resources for chromatography also has useful reference data on septum puncture force and reseal performance across materials. These are good benchmarks for setting your own internal limits.
The decision rule
Run the four-step incoming test on every new hplc septa lot before it touches a method. Measure the diameter, durometer, reseal weight loss, and lamination peel force. Accept only lots that meet the Tier 2 limits for any method where RSD below 2% matters. Reject any lot that fails durometer or lamination outright, because those are process variables that sorting cannot fix. The lot number is the unit of control, not the part number. A part number with a Cpk of 1.33 on the datasheet means nothing if the lot in your drawer was molded on a Friday afternoon with a cold press. Measure it. That is the only way to keep batch to batch variation in hplc septa from becoming a failed sequence you have to explain to a client.
References & further reading
External references and standards: