Where the line starts: glass tube forming and the tolerances that get locked

Last March a lot of 300 µL glass inserts failed our incoming dimensional check at a rate of 4 in 200: the cone tip sat 0.45 mm off center, enough that on a 126-well autosampler the needle grazed the wall on the outer columns and bent at the 40th injection. The factory’s control chart showed nothing wrong. That gap between what the factory measures and what the customer measures is the whole story of insert manufacturing. This article maps the line station by station and separates the parameters that are closed-loop controlled from the ones that are only inspected at the end, because that distinction decides where your risk actually lives.

Where the line starts: glass tube forming and the tolerances that get locked in first

A vial insert is drawn or molded from borosilicate glass, usually 1.0 mm or 0.9 mm wall, on a forming line that runs at several hundred parts per minute. The first control point is the outer diameter of the tube stock, held to ±0.05 mm on a 6.0 mm nominal. That number matters because everything downstream is positioned relative to the OD.

On a modern line the forming station runs closed-loop on OD via laser micrometer feedback to the draw speed. Wall thickness is a different story. Most lines check wall thickness by sampling every 30 minutes with a dial indicator or an optical comparator, not continuously. If your application depends on wall uniformity (and coring behavior does), you are relying on a sampled parameter, not a controlled one. I have seen wall run 0.82 mm to 1.04 mm within a single lot that passed every incoming certificate.

The forming station also sets the bottom geometry. Flat-bottom inserts and conical inserts come off different tooling, and the cone angle (typically 15° to 20° included) is set by the mold or the flame-forming fixture. Once the tool is cut, that angle is fixed. No downstream station corrects it.

Annealing and the residual stress nobody measures until the insert cracks in transit

Glass that is formed fast carries residual stress. The annealing lehr runs at 560 °C to 580 °C for borosilicate 3.3, with a controlled cooling ramp. The parameter that matters is not the peak temperature but the cooling rate through the strain point, and this is where I see the widest gap between what is controlled and what is checked.

Large factories run in-line polariscope inspection on annealed ware. Smaller ones do not. If your supplier cannot tell you the lehr profile or show you a polariscope image, assume residual stress is uncontrolled. The failure mode shows up later: inserts that survive incoming inspection but crack at the cone during the first autoclave cycle or during shipping in a cold truck. A stress-relieved insert will pass a three-cycle 121 °C autoclave test without cracking. An under-annealed one often will not.

Silanization: the coating station that decides whether your analyte sticks to the wall

Most inserts sold for HPLC and LC-MS work are silanized to reduce surface silanols and analyte adsorption. The coating station is a vapor-phase or solution-phase process, and the controlled parameters are silane concentration, dwell time, and cure temperature. A well-run line holds cure at 120 °C to 150 °C for 30 to 60 minutes and verifies the coating by contact angle or by a dye test on a sample from each batch.

Here is the honest part: coating uniformity inside a 5 mm ID tube is hard to verify. Most factories verify the outside of the glass and a destructive sample, not every part. If you are running basic analytes at low concentration and seeing tailing that follows the insert lot rather than the column, the silanization station is the first place to look. Ask for the contact angle data and the cure profile. If the answer is “we use a proprietary process,” move on.

Tip forming and the centering step that decides whether the needle survives

The cone tip is where most field failures originate, and it is the station where closed-loop control is most often absent. The tip is formed by flame drawing or by a separate grinding step, and the critical parameters are tip concentricity, tip wall thickness, and the depth of the internal cone.

Concentricity is the one that bites. On the line, concentricity is usually checked by a vision system on a sample every hour. That is a check, not a control. A drift in the forming fixture can put the tip 0.3 to 0.5 mm off center for a stretch of production before the hourly sample catches it. If your autosampler needle path has less than 0.5 mm clearance to the insert wall (and many do), that drift is a bent needle waiting to happen.

This is the parameter I would demand continuous or near-continuous monitoring on. A vision system that measures every part at the tip station is not expensive relative to the cost of a bent needle on a 96-well run.

Cleaning and depyrogenation: the station that is almost always validated, rarely monitored

Inserts destined for LC-MS or bioanalysis go through a wash and a depyrogenation tunnel. The tunnel runs at 250 °C for at least 30 minutes to achieve a 3-log endotoxin reduction, and the temperature profile is validated once and then monitored by thermocouple on a schedule. The wash chemistry (usually a detergent followed by DI water rinse and a final solvent rinse) is where contamination risk lives.

What the factory controls: rinse water conductivity (typically held below 1 µS/cm), detergent concentration, and final rinse volume. What it often only checks at the end: extractables by LC-MS on a sample per lot. If you are running trace analysis, ask for the extractables data and the rinse water conductivity log. A factory that cannot produce both is telling you something.

Packaging and the last station where a good part becomes a bad one

Inserts are packed into trays, bags, or nests. The packaging station introduces two failure modes: particulate from the bag or tray material, and physical damage from inserts rubbing against each other. Static control matters here. A line running without ionizing bars will pick up glass fines on the inserts, and those fines show up as spurious peaks in your blank.

Most factories check particulate on a sample per lot by rinsing and counting. Very few monitor the packaging line continuously. If you see a particulate spike that correlates with a lot change but not with the glass itself, look at the bag supplier and the static control on the packaging line.

What is closed-loop controlled versus what is only inspected at the end

This is the table I wish every supplier would publish. It is the difference between a parameter the factory is actively steering and one it is merely sampling.

Read that table with your application in mind. If your failure mode is needle damage, the sampled concentricity check is your exposure. If your failure mode is adsorption, the sampled coating check is your exposure. If your failure mode is cracking in transit, the residual stress check (or its absence) is your exposure. A general vial insert purchase order that does not name these parameters is buying whatever the line happened to produce that day.

Field case one: the off-center tip that passed every certificate

Symptom: a contract lab running a 96-well LC-MS method on a 2 mL vial format reported bent needles on the outer two columns of the autosampler, roughly every 40th injection, with no change in method or column.

Measurement: we pulled 200 inserts from the suspect lot and measured tip concentricity on an optical comparator. The lot certificate said ≤ 0.2 mm TIR. Our measurement found a bimodal distribution: 180 parts at 0.12 to 0.18 mm, and 20 parts at 0.38 to 0.46 mm. The hourly vision sample had caught the good mode. The drift happened between samples.

Fix: the factory moved concentricity measurement to every part at the tip station and added an automatic reject gate. The lab also switched to a 2 mL vial with a wider-mouth insert nest to buy needle clearance. Bent needles stopped. The lesson is that a sampled check on a drifting fixture is a coin flip, and the customer pays for the tails.

Field case two: the silanization lot that tailed everything

Symptom: a bioanalytical lab running a basic drug panel saw peak tailing (USP tailing factor) jump from 1.3 to 2.4 on a single insert lot, with the same column and the same mobile phase. The column was replaced twice with no change.

Measurement: we ran a dye test on inserts from the suspect lot and the previous lot. The suspect lot showed patchy coating on the internal cone, visible as uneven dye uptake. Contact angle on the outside was within spec. The internal surface was not.

Fix: the factory traced it to a cure oven zone that had drifted 18 °C low on one side of the tunnel. The oven thermocouple was reading the other zone. They added a second thermocouple per zone and a daily dye test on the internal cone. Tailing returned to 1.3. The general point: coating uniformity inside the tube is the parameter most likely to be assumed rather than measured, and it is exactly the one your analyte sees.

How to audit a line without a two-day factory tour

You do not need to see every station. You need to see four things, and you can ask for them in a single email.

• The control chart for tip concentricity over the last 90 days. If it is a sample chart, ask how often the sample is taken and what the reject gate does. If there is no reject gate, you are buying the tails.
• The lehr profile and a polariscope image from the last three lots. No image, no stress control.
• The silanization cure profile and a dye test image of the internal cone. Outside contact angle alone is not enough.
• The rinse water conductivity log and the extractables report. If the factory cannot produce these, your trace analysis is running on faith.

I would also ask for the packaging line’s static control setup. Ionizing bars on the packaging line are cheap and they eliminate a whole class of particulate complaints. A factory that has them is thinking about the customer’s blank; one that does not is thinking about throughput.

Where OEM quality actually gets decided

The line does not produce a single quality level. It produces a distribution, and the factory’s control strategy decides how wide that distribution is and where the tails fall. Closed-loop stations (OD, lehr temperature, cure temperature, rinse conductivity) hold their distributions tight. Sampled stations (wall thickness, concentricity, coating uniformity, particulate) hold them only as tight as the sampling frequency allows, which is to say they drift between samples and the customer finds the drift.

If you are sourcing inserts through a distributor, the HPLC consumables catalog will list the insert by volume and glass type and say nothing about the control strategy. That is fine for general work. For methods where a bent needle or a tailing peak costs you a run, you need the control strategy in writing. A chromatography supplier that cannot get you the four audit items above is a reseller, not a partner, and the difference shows up on your autosampler.

The regulatory frame is worth knowing here. EPA SW-846 methods and the sample handling guidance from instrument vendors like Agilent and Thermo Fisher assume the consumable is fit for purpose; they do not specify the insert’s control strategy. That is your job at the purchase order stage. Restek’s technical resources are a good starting point for the extractables and adsorption questions you should be asking the factory.

Troubleshooting checklist: match the symptom to the station

When an insert-related problem shows up, the fastest path is to map the symptom to the station that controls it. Use this as a starting point, not a substitute for measurement.

• Bent or blunted needles on outer wells: tip concentricity (tip forming station). Measure TIR on 50 parts from the lot.
• Peak tailing that follows the lot: silanization uniformity (coating station). Run the dye test on the internal cone.
• Cracks at the cone after autoclave or cold shipping: residual stress (annealing station). Request polariscope images.
• Particulate spikes in blanks: packaging line static control or rinse water quality. Check the conductivity log and the ionizing bars.
• Volume shortfall at low draw: bottom geometry and cone angle (forming station). Verify the cone angle against the drawing.
• Adsorption of basic analytes: coating cure (silanization station). Ask for the cure profile and contact angle.

Each of these maps to a station, and each station has a control mode. If the control mode is “sampled,” your lot is a sample too. Buy accordingly.

The decision rule is simple: for any method where a bent needle, a tailing peak, or a particulate spike costs you a run, require the factory to name the control mode for tip concentricity, coating uniformity, and residual stress in writing, and require a reject gate on concentricity. If the factory cannot name them, the line is producing a distribution you have not seen, and you will meet its tails on the autosampler.

References & further reading

Related resources from our lab consumables network:

• Vial Insert at hplcinsert.com
• 2Ml Vial at 2mlsamplevial.com
• Hplc Consumables at hplcconsumable.com
• Chromatography Supplier at chromatographysupplier.com

External references and standards:

• Agilent chromatography consumables and columns
• Thermo Fisher chromatography and sample handling
• U.S. EPA SW-846 test methods for waste and materials
• Restek technical resources for chromatography