A lot of 10 mL clear glass vials came off a line in 2022 with a 0.4 mm crimp skirt height spread across the pallet. The customer’s 1260 autosampler drew from the same lot for eleven days before anyone noticed the needle was riding 0.3 mm higher on one side of the tray. The fill volume was fine. The septa were fine. The reject was a crimp die that had drifted 0.15 mm on its depth stop after a tooling changeover that nobody recorded. That is the kind of thing a chromatography vial factory controls, or fails to control, on the line. This article maps what is actually being held to a setpoint at each station, and where the setpoint is only checked at the end of the run.

If you buy vials at volume, you already know the product. What you probably do not know is which of the twenty-odd parameters on a molding and finishing line are closed-loop and which are open-loop with a final gate. That distinction decides whether a lot drifts slowly or fails hard. It also decides what your supplier can tell you when you ask why a lot behaved differently from the one before it.

Glass forming sets the geometry, and the furnace is the only place it is corrected

A type I borosilicate vial starts as a tube. The tube is cut, the bottom is closed, the shoulder is formed, and the neck is calibrated. On a modern line that is a rotary machine with maybe 60 to 120 heads, each carrying one tube through a series of burners. The glass is at roughly 1100 to 1250 °C at the forming burners and drops to about 600 °C at the annealing lehr inlet.

The parameters that matter at this station are the tube outer diameter, the tube wall thickness, the cut length, the bottom-forming gas pressure, the neck-calibration mandrel diameter, and the lehr temperature profile. Of those, only the tube OD and the mandrel diameter are closed-loop in the ordinary sense. The tube OD is measured by a laser gauge on the incoming tube and the feed rate is trimmed to hold it. The mandrel is a fixed steel dimension; it does not drift, but it wears, and the wear is not detected until a neck gauge check.

What is actually closed-loop at the hot end

  • Tube OD: laser gauge, closed-loop on draw speed. Typical window ±0.10 mm on a 12 mm tube.
  • Lehr temperature: thermocouple closed-loop on the zone controllers. The setpoint is stable; the glass residence time is not, because it depends on line speed.
  • Bottom-forming gas pressure: usually open-loop, set by a regulator and checked by eye or by a downstream vision system. This is where a lot of concentricity variation comes from.
  • Neck mandrel: fixed dimension, checked at changeover. Wear is a slow drift that only a gauge catches.

The vision system at the end of the forming section is the first real gate. It checks for obvious defects (cracks, stones, chips, malformed bottoms) and rejects them. It does not measure the neck ID to 0.05 mm. That happens later, if at all. So a neck that is 0.06 mm undersized can pass the vision gate and only show up when a crimper cannot seat a seal evenly.

I have seen this exact failure. A customer reported that their 11 mm crimp seals were leaking intermittently on a headspace method. The vial neck OD measured 11.02 mm at the top and 10.94 mm at the bottom of the neck, a taper of 0.08 mm that should not exist. The mandrel on one head had worn. The line was running at 140 vials per minute and the neck gauge was a manual check every 30 minutes on a sample of five. That is an open-loop control with a sampling plan that cannot see a single-head defect. The fix was a mandrel replacement and a move to 100% inline neck gauging on that line, which the factory had been deferring for two years.

Annealing is where residual stress is either removed or baked in

Every glass vial has residual stress. The question is how much and where. The annealing lehr runs a temperature profile that brings the glass up to about 560 to 580 °C (the annealing point of borosilicate is around 565 °C), holds it, then cools it slowly enough that the stress does not re-form. The cooling rate in the last zone is the parameter that matters most. A lehr that cools too fast leaves stress in the shoulder and neck, which is exactly where a crimp or a screw cap applies load.

The standard test is a strain viewer, which is a polariscope that shows birefringence. USP <660> and the European Pharmacopoeia both specify limits, and the practical number is that a type I vial should show no more than a faint tint in the strain viewer, not the strong rainbow bands that indicate high stress. A vial with high residual stress will survive normal handling and fail at the crimp or, worse, crack in the autosampler needle guide after a few hundred injections.

The lehr is closed-loop on temperature and open-loop on stress

The zone temperatures are controlled. The belt speed is controlled. The stress itself is not measured inline; it is measured on a sample from the lehr exit, usually by a strain viewer and a manual read. So the factory is controlling the inputs that produce low stress, not the output. If a burner is running hot because a thermocouple is reading low, or if the belt speed was bumped to clear a jam, the stress goes up and the first evidence is a sample check or a customer complaint.

This is the single most common place where a line that looks fully instrumented is actually running open-loop. Ask your supplier what their lehr exit strain check frequency is, and whether it is per head or per line. A per-line check every two hours will miss a single-head problem for a long time.

Dimensional gauging is the gate, and it is not the same as control

After annealing, the vials go through a dimensional gate. On a good line this is a set of inline gauges: neck ID, neck OD, overall height, body OD, and sometimes bottom thickness. The gauges reject out-of-spec parts to a reject chute. This is the gate that catches what the hot end did not control.

The parameters and typical windows on a 2 mL HPLC vial are worth putting in a table, because the numbers are what your autosampler actually cares about.

Parameter Typical window Control type What it affects
Neck ID 6.00 ± 0.05 mm (11 mm crimp) Inline gauge, gate Crimp seal seating, septum compression
Neck OD 11.00 ± 0.10 mm Inline gauge, gate Crimp tool fit, seal concentricity
Overall height 32.00 ± 0.30 mm Inline gauge, gate Needle depth, tray seating
Body OD 12.00 ± 0.15 mm Inline gauge, gate Tray fit, autosampler gripper
Wall thickness 1.00 ± 0.10 mm Sample check Thermal mass, autosampler needle clearance
Residual stress Strain viewer, faint tint Sample check Crimp cracking, needle guide wear
Crimp skirt height 7.00 ± 0.20 mm after sealing Sample check (sealing line) Seal integrity, needle entry

Notice the split. The first four are inline gates. Wall thickness, residual stress, and crimp skirt height are sample checks. That means the factory is not controlling them in real time; it is verifying them periodically. If your method is sensitive to wall thickness (and headspace methods are, because thermal mass changes the equilibration time), you need to know that this parameter is not on a closed loop.

The wash and depyrogenation tunnel is a temperature and residence-time problem

For vials sold as low-particulate or for methods where carryover matters, the wash and depyrogenation tunnel is a controlled station. The parameters are wash water conductivity (usually < 1 µS/cm for the final rinse), detergent concentration, rinse temperature, and the depyrogenation zone temperature and residence time. The standard for depyrogenation is a 3-log reduction in endotoxin, which the industry demonstrates with a 250 °C zone and a residence time that gives an FH value sufficient for the validation. Many lines run 300 to 350 °C for a shorter time.

The tunnel is closed-loop on temperature. It is open-loop on particle count, which is checked on a sample by a liquid particle counter after the tunnel. A vial that comes out with a particle count above the limit is rejected at the sample gate, not caught inline. This is why the wash chemistry matters so much: if the final rinse conductivity drifts above 1 µS/cm, the particle count goes up, and you find out at the sample check.

I dealt with a line where the final rinse conductivity was reading 0.6 µS/cm on the inline meter but the particle counts on the sample check were running at 15 to 20 particles/mL at ≥ 10 µm, against a limit of 10. The inline meter was downstream of the rinse manifold and the conductivity probe had a dead leg. The water in the dead leg was stale and low-conductivity; the water hitting the vials was not. The fix was relocating the probe to the manifold outlet and adding a flow-through cell. That is a control-loop placement problem, not a chemistry problem, and it is the kind of thing that only shows up when the sample check disagrees with the inline reading.

Septa and caps are a separate line with its own setpoints

The septum is where a lot of the performance lives, and it is manufactured on a different line with different controls. A silicone septum with a PTFE face is molded or punched, and the parameters are the silicone durometer, the PTFE thickness, the slit or pre-slit geometry, and the extraction history. The durometer is typically 45 to 55 Shore A for a standard 11 mm septum. The PTFE face is usually 0.05 to 0.13 mm thick.

The puncture count is the number that matters to your autosampler. A standard silicone/PTFE septum will take roughly 50 to 100 punctures before coring becomes a real risk, depending on needle gauge and bevel. A pre-slit septum takes fewer, because the slit is already a stress concentrator. If your method runs 96 injections from one vial, you are near the edge and you should be using a septum rated for it, or changing vials mid-run. The HPLC septa page on our sister site has the durometer and puncture data by product line.

What the sealing line controls

If the vial is sold pre-capped or crimped, the sealing line has its own setpoints: crimp jaw pressure, crimp depth, and cap torque for screw-thread vials. A screw cap on a 9 mm thread is typically torqued to 0.6 to 1.0 N·m; over-torque distorts the septum and under-torque leaks. The crimp depth on an 11 mm aluminum seal is set to give a skirt height of about 7.0 mm, and the tolerance is tight because the autosampler needle has to pass through the septum without hitting the aluminum. A crimp that is 0.3 mm too deep will pull the septum down and change the needle penetration depth.

The sealing line is closed-loop on crimp jaw pressure and open-loop on skirt height, which is checked on a sample. That is the same pattern as the rest of the line: the machine holds the input, the output is verified periodically. The 2022 case at the top of this article was a crimp depth stop that drifted after a tooling changeover. The line was holding jaw pressure perfectly. The depth stop was a mechanical set screw that nobody re-verified.

Where the closed-loop ends and the final gate begins

Here is the honest map. On a well-run chromatography vial line, the parameters that are genuinely closed-loop are: tube OD, lehr zone temperatures, wash rinse conductivity (if the probe is placed correctly), and crimp jaw pressure. Everything else is either a fixed dimension that wears, a sample check, or a final gate.

The final gate is the outgoing inspection: a sample from the lot is measured for dimensions, checked in a strain viewer, particle-counted, and functionally tested on a representative autosampler. This is where the lot is either released or held. It is a gate, not a control. If the gate catches a problem, the lot is rejected; the line is not corrected until someone investigates.

That distinction is why two lots from the same factory can behave differently. The closed-loop parameters were held on both. The open-loop parameters drifted on one. The final gate caught it on one and not the other, because the sample did not happen to include the bad head.

Two field cases that show the difference between control and gate

Case 1: the drifting neck that passed the gate

Symptom: intermittent crimp seal leaks on a headspace GC method, roughly 3% of vials, no pattern by tray position.

Measurement: neck OD measured at three heights on 50 vials from the suspect lot. Top 11.02 mm, middle 10.98 mm, bottom 10.94 mm. The taper was 0.08 mm. The inline gauge was measuring neck OD at a single height, at the top. The gate passed every vial because the gauge only saw the top.

Fix: mandrel replacement on the worn head, plus a second gauge at the bottom of the neck. The line had been running the single-height gauge for years because the original validation only required one measurement point. The customer’s leaks stopped after the tooling change.

Case 2: the rinse conductivity that read low

Symptom: particle counts on the sample check running 15 to 20 particles/mL at ≥ 10 µm against a 10 limit, on a lot that was supposed to be low-particulate.

Measurement: inline conductivity 0.6 µS/cm, sample-check particle count high. A grab sample from the rinse manifold outlet read 1.4 µS/cm. The inline probe was in a dead leg downstream of the manifold.

Fix: probe relocated to a flow-through cell at the manifold outlet, dead leg removed, and the lot re-washed. The line had been passing its inline check for months while the sample check was the only thing catching the problem.

A checklist for auditing a supplier’s line controls

If you are qualifying a supplier or investigating a lot that behaved differently, ask these questions. The answers tell you whether the line is controlled or just gated.

  • Which dimensional parameters are measured inline, and at how many points? (Neck OD at one height is not the same as neck OD at three.)
  • Is the lehr exit strain check per head or per line, and how often?
  • Where is the rinse conductivity probe, and is it in a flow-through cell or a dead leg?
  • What is the crimp depth stop, and how is it verified after a tooling changeover?
  • What is the sample size and frequency on the outgoing gate, and does it stratify by machine head?
  • What is the septum durometer and PTFE thickness, and what puncture count is it rated for?

For the dimensional side, the chromatography vial product pages list the nominal dimensions and tolerances per catalog number. For the instrument side, Agilent publishes autosampler tray and needle geometry in its GC columns and accessories documentation, and Thermo Fisher’s chromatography and sample handling pages carry the same for their autosamplers. If you are setting up a method that is sensitive to wall thickness or residual stress, the HPLC vial reference material and the chromatography products index are worth a look before you write the spec. For method-level context on why the vial geometry matters to the injection, EPA SW-846 and the Restek technical resources library both have worked examples.

The decision rule

Buy from a line that can tell you which parameters are closed-loop and which are gated, and that stratifies its outgoing sample by machine head. If a supplier cannot answer the lehr strain check frequency and the crimp depth verification after changeover, treat every lot as unverified until you have run your own dimensional and functional check. The closed-loop parameters will be fine. The open-loop ones are where the drift lives, and the gate only catches it if the sample happens to include the bad head.

References & further reading

Related resources from our lab consumables network:

External references and standards: