A lot of 1,000 crimp seals left the line at 4.6 mm skirt height instead of the 4.9 mm the drawing called for, and nobody caught it until a customer’s autosampler started dropping vials at position 47 of a 96-well tray. The press had been running fine. The tooling had been running fine. What moved was the aluminum coil thickness, from 0.010 in to 0.0093 in, and the press stroke was set by a mechanical stop that assumed a nominal gauge. That is the kind of thing a chromatography supplier factory either controls on the line or finds out from a complaint. This article is about which of those two it is, station by station, for the parts that touch your sample: vials, caps, septa, ferrules, and the molded and machined pieces that hold a column in a GC or an HPLC.

I have run and audited enough of these lines to say plainly: most chromatography supplier factories control fewer parameters in closed loop than their quality certificates imply. The certificates are real. The control is often downstream inspection dressed up as process control. Knowing which stations are actually servoed and which are only gated at the end tells you what kind of failure you will eventually see, and how fast.

Where a chromatography supplier factory’s control actually sits: closed loop versus end-of-line gate

Every consumable line has two kinds of control. Closed-loop control means a sensor reads the actual output and the machine corrects in real time, cycle after cycle. End-of-line gate means the part is made, then measured, then accepted or rejected. The first prevents drift. The second only catches drift after it has already produced scrap, and only if the sampling happens to land on the bad parts.

Here is how the stations on a typical vial-and-closure line split out. This is the map I use when I audit a new source, and it is the map I would hand a procurement manager who wants to know what they are actually buying.

Station What is controlled Control type Typical spec and tolerance What fails if uncontrolled
Glass tube forming Outer diameter, wall thickness, camber Closed loop (laser OD gauge on the draw) 12 x 32 mm, OD ±0.15 mm, camber <0.25 mm over 32 mm Autosampler gripper mispick, needle strike
Neck and thread forming Thread pitch, neck ID, finish height Closed loop on the mold, open on wear GPI 8-425 or 13-425 finish, neck ID ±0.1 mm Cap cross-threading, septa not seating flat
Septa punching Punch force, die clearance, sheet tension Open loop (force set, not measured per part) PTFE-faced silicone, 0.9 to 1.3 mm thick, 45 to 55 Shore A Coring, slivering, injector needle drag
Cap molding Skirt height, crimp ring geometry, resin lot Closed loop on melt pressure, open on dimension Skirt 4.9 mm ±0.1 mm, ring ID ±0.05 mm Poor crimp seal, vial drop, evaporation
Assembly and insertion Septa seating depth, cap-to-septa concentricity Open loop (pneumatic press, set by eye) Seating depth ±0.2 mm Leak path, needle coring on first puncture
Crimp (for pre-crimped closures) Crimp jaw force, jaw wear Closed loop on force, open on jaw geometry 8 to 12 in-lb equivalent jaw force, jaw wear limit 0.05 mm Loose or crushed seals, seal height drift
Final dimensional and leak gate Sampled dimensions, dye leak, puncture count End-of-line gate AQL per lot, dye leak at 0.5 bar, puncture to 50 cycles Anything the upstream stations let through

Notice how much sits in the open-loop column. That is normal, and it is the part most buyers never ask about. The line is not uncontrolled, but the control is often a setpoint a technician dialed in once and a gauge nobody re-verifies between shifts.

The glass draw is the one station that is almost always genuinely closed loop, and that is why tubes rarely fail

On a modern vial line the glass tube is formed by a draw or a blow, and the outer diameter is read by a laser micrometer that feeds back to the forming head. This is real closed-loop control. The gauge samples continuously, the controller adjusts air or draw speed, and the tube comes out within ±0.15 mm of nominal almost every time. When I see a dimensional problem on a 12 x 32 mm vial, it is rarely the OD. It is camber or wall thickness, and those are the parameters the laser does not read.

Camber is the bow in the tube over its length. A tube can be perfectly round and still banana-shaped. The spec I write into purchase orders is camber under 0.25 mm over 32 mm, because above that the vial starts to bind in a 96-well tray and the autosampler gripper either mispicks or drops it. Wall thickness matters for a different reason: too thin and the vial crushes under crimp force, too thick and the internal volume shrinks enough to change headspace. Neither is closed-loop on most lines. They are checked on a sample at the end.

If you are sourcing vials, the question to ask is not “is your OD controlled” (it is) but “how do you control camber and wall thickness, and how often.” A factory that answers with a sampling frequency and a gauge ID is telling you the truth. One that answers “we control everything” is telling you it has not thought about it. The OEM vial side of this business lives or dies on exactly these two numbers, and it is worth reading a supplier’s process sheet before you trust a certificate.

Septa punching is where the factory’s control is weakest and your chromatography pays for it

I will be blunt: septa punching is usually the least controlled station on the line, and it is the one that ruins runs. A punch press takes a sheet of silicone with a PTFE face and stamps out discs. The variables that matter are punch force, die clearance, sheet tension, and die sharpness. On most lines, force is set by a regulator and never measured per part. Die clearance is fixed by the tool. Sheet tension is whatever the operator set when loading the roll. Die sharpness is on a maintenance schedule that gets extended when production is behind.

When die clearance drifts past roughly 5% of sheet thickness, or the punch dulls, you get coring: a small plug of septum is cut free and pushed into the vial or into the needle. You also get slivering, where a thin curl of silicone hangs off the puncture edge and either blocks the needle or sheds into the sample. Neither shows up in a dimensional check. Both show up as ghost peaks, a blocked needle, or a mysterious loss of injection volume.

The failure mode I deal with most is a septum that passes a visual inspection and fails on the third puncture. The first puncture is clean. The second drags. By the third, the needle has carved a channel and the seal no longer holds headspace pressure. A septum rated for 50 punctures on paper is often good for 10 to 15 in practice if the punch was dull. The rating is a material property, not a process guarantee. This is why the vial closures you buy should come with a puncture-count test method, not just a number.

Field case: the ghost peak that was a punch problem

A contract lab running EPA method SW-846 8260 volatile organics kept seeing a small unidentified peak at roughly the retention time of a common internal standard, but only on the third injection from each vial. The instrument was fine. The column was new. The measurement that identified it was a simple one: they crimped 20 vials, injected each three times, and weighed the septa before and after. The mass loss was 0.4 mg per septum on the third puncture, against 0.02 mg on the first. That is silicone being carved off the puncture edge and carried into the inlet. The fix was a supplier change to a line running a fresh die with a documented clearance of 3% of sheet thickness and a punch force logged per shift. The ghost peak disappeared. The lesson is that the septum is a punched part, and a punched part is only as good as its die.

Crimp geometry is set by tooling, and tooling wear is the parameter nobody watches

For pre-crimped closures, the crimp station is where the cap is rolled onto the vial. The jaw force is usually closed-loop, which sounds reassuring, but the geometry of the jaws themselves is not. As jaws wear, the same force produces a different seal. The wear limit I write into a spec is 0.05 mm on the jaw profile, checked with a profile projector every 50,000 crimps. Most factories check it when a customer complains.

The symptom of jaw wear is a slow drift in seal height, from a nominal 4.9 mm toward 4.6 or 5.2 mm. A seal at 4.6 mm is loose enough that the autosampler can lift the cap off with the vial. A seal at 5.2 mm is over-crimped, and the septum is compressed past its elastic limit, so it leaks after a few punctures. Neither is visible on the line unless someone is measuring seal height on a sample, and on many lines that measurement is a go/no-go gauge rather than a number.

The torque equivalent matters for screw caps. A 9 mm screw cap on a 12 x 32 mm vial wants roughly 8 to 12 in-lb of application torque, and the factory that assembles pre-capped vials should be able to tell you the torque its capper is set to and how it verifies it. A capper running at 15 in-lb will crack the cap or deform the septum. One at 5 in-lb will back off in transit. Ask for the number. If the answer is “the machine is set correctly,” you have your answer about the control.

What the factory controls in closed loop versus what it only checks at the end: a decision checklist

When I qualify a new chromatography supplier, I ask for a station-by-station control map and then I verify three of the claims on the floor. Here is the checklist, with the numbers I use.

  • Glass OD and camber: ask which gauge reads it and whether the reading feeds back to the forming head. Expect a laser micrometer and a camber spec under 0.25 mm. Verify by measuring 30 vials from one lot.
  • Wall thickness: ask the sampling frequency. If it is once per shift on five parts, that is a gate, not control. That is acceptable if the glass draw is stable, but you should know.
  • Septa punch force and die clearance: ask for the clearance as a percentage of sheet thickness (target 3 to 5%) and the die change interval. This is the single most predictive question on the list.
  • Cap skirt height: ask for the tolerance (±0.1 mm) and the gauge. A factory that measures it with calipers on a sample is gating; one with an in-line vision system is controlling.
  • Assembly seating depth: ask how septa are seated (±0.2 mm). Pneumatic press with a mechanical stop is fine if the stop is verified. “By hand” is not.
  • Crimp jaw wear: ask for the wear limit and the check interval. 0.05 mm every 50,000 crimps is a defensible answer.
  • Final leak and puncture gate: ask the AQL, the dye leak pressure (0.5 bar), and the puncture count method. This is the backstop, not the control.

If a factory can answer six of these seven with a number and a gauge, it is running a real process. If it answers two, it is running an inspection operation with a press attached. The chromatography supplier you choose is, in practice, the set of control decisions its line makes, whether or not anyone wrote them down.

Ferrules and column hardware: the machined parts where tolerance is the whole product

Ferrules and column end fittings are a different line from vials, and the control question is different. A graphite or graphite-polyimide ferrule is a machined or molded part with a bore that has to match the column OD within a few hundredths of a millimeter. A 1/16 in ferrule with a bore that is 0.05 mm oversized will not grip the column, and you get a leak at 0.1 MPa that shows up as a wandering retention time. A bore that is undersized will crush the polyimide coating on the column and create a dead volume at the connection.

The stations here are turning or molding, then a bore gauge, then a fit test. The fit test is the interesting one. A good factory runs every ferrule over a go/no-go pin at the bore tolerance, and the pin is checked against a master every shift. A weaker factory samples. The difference shows up in your chromatogram as peak shape on early-eluting compounds, because a ferrule that does not seat concentrically creates a tiny swept volume right at the column inlet.

Temperature rating is the other parameter. A graphite ferrule is good to about 400 °C in a GC inlet, a polyimide-blend ferrule to about 350 °C, and a PTFE-based seal much lower. If the factory cannot tell you which resin lot went into which ferrule, you cannot predict the upper temperature limit of the lot you receive. I have seen a batch of ferrules bleed at 320 °C that should have been stable to 350, and the cause was a resin substitution the factory made to cover a shortage. The line controlled the bore and did not control the resin. That is a control gap, and it is invisible until you run a high-temperature method.

Field case: the retention time drift that was a ferrule bore

A food-safety lab running pesticide residues on a GC-MS/MS saw retention times on the earliest eluting compounds drift by 0.04 min over a week, then snap back after a new column install. The column was not the problem. The measurement that found it was a leak check at the inlet: 0.15 MPa decay over 30 seconds, against a spec of 0.05. The ferrule had been re-tightened four times and had deformed. The bore had been machined at the top of its tolerance, and repeated tightening crushed it off-center. The fix was a ferrule with a bore tolerance of ±0.02 mm and a documented torque of 1/4 turn past finger-tight, plus a rule to replace rather than re-tighten after two uses. Retention time stability came back to ±0.005 min. The line had controlled the bore at the station but not the torque at the bench, and the bench is where the failure happened.

What “OEM quality” means on the line, and what it does not

Buyers ask for OEM quality as if it were a grade. It is not. It is a set of control decisions that a specific factory made, and it transfers only if the factory transfers the control, not the drawing. A vial made to an OEM drawing on a line with a dull punch and no camber gauge is not an OEM vial. A vial made to the same drawing on a line with a laser OD loop, a camber gauge, and a die change interval is. The drawing is necessary and nowhere near sufficient.

This is why I treat the audit as the product. When I look at a line, I want to see the gauge IDs, the setpoint logs, the die change records, and the last three corrective actions. I want to see a punch force log with a signature on it. I want to see the crimp jaw wear check on the maintenance board. Those documents are the control. The certificate of analysis is a summary of the gate, and the gate is the weaker of the two control types.

If you are buying columns and consumables from a major brand, the control is usually there because the brand audits its own contract manufacturers against a process spec. Agilent publishes its GC column and accessory specifications and holds its suppliers to them. Thermo Fisher’s sample handling portfolio carries similar process documentation. Restek’s technical resources are a useful benchmark for what a well-controlled consumable should be able to state about itself. If you are buying from an unbranded source, you are the auditor. The analytical testing you do on incoming lots is the only gate you actually own.

How to verify control without a factory visit: three tests you can run on incoming lots

Most procurement managers cannot get to the factory. You can still separate a controlled line from a gated one with three tests on the lot you receive.

  • The camber and drop test. Take 50 vials, run them through your autosampler tray, and count mispicks. More than one in 50 points to a camber or OD problem upstream, and it is the cheapest test you will ever run.
  • The three-puncture mass loss test. Weigh 20 septa, puncture each three times with the actual needle gauge you use, and reweigh. Mass loss above 0.1 mg per septum on the third puncture means the punch or die is worn. This is the test that catches the failure the certificate misses.
  • The seal height and leak test. Measure seal height on 30 crimped closures with a micrometer. A spread wider than 0.2 mm means the crimp station is drifting. Then dye-leak at 0.5 bar and count failures. Any failure is a process problem, not a shipping problem.

Run these on the first three lots from any new source and keep the numbers. A line under control will hold them lot to lot. A line that is only gating will drift, and you will see it in the data before you see it in a failed run.

The decision rule

Buy from the factory that can name, for each station that touches your sample, whether the parameter is closed-loop or end-of-line gated, and give you the number, the gauge, and the check interval. If a supplier can do that for septa punch clearance, cap skirt height, and crimp jaw wear, the rest of the line is probably under the same discipline. If it cannot, treat every certificate as a sample of one and budget for incoming testing, because the control you are relying on is the gate, and the gate only catches what it happens to measure.

References & further reading

Related resources from our lab consumables network:

External references and standards: