Where vapor actually escapes from a capped reagent bottle

A 96-hour HPLC sequence on a 4 L mobile phase reservoir lost 38 g over the run, and the retention times on the late-eluting peaks had drifted 0.04 min by the last injection. The method was fine. The column was fine. The problem was the cap on the reservoir: a two-port GL45 screw cap with a PTFE-faced silicone septum that had been punctured 14 times by a 1/16 in. OD solvent line and never re-seated. That is a reagent bottle seal failure, and it will show up as evaporation and sample loss long before anyone suspects the bottle.

Most labs chase this kind of drift through the pump, the mobile phase pH, or the column temperature. I have done it too. But seal integrity on the reservoir, the volumetric flask, and the autosampler supply bottle is a measurable, testable property, and once you start weighing bottles instead of guessing, the leak paths become obvious. This piece is about the mechanisms: where vapor escapes, what happens to an elastomer after it is punctured, how compression set quietly kills a seal, and how to quantify loss gravimetrically so you can prove the fix worked.

A screw cap is not a single seal. It is a stack of interfaces, and each one is a potential leak path. On a standard GL45 bottle you have, from the inside out: the wetted face of the liner against the glass land, the liner against the inside of the cap, the cap thread against the bottle thread, and the cap skirt against the neck shoulder. Vapor can leave through any of them, and through any penetration in the liner itself.

The four leak paths I check first

• Liner-to-glass land. The sealing land on a GL45 neck is roughly 3 mm wide. A liner that is 1 mm undersized, or a cap torqued to 1.2 N·m instead of 2.0 N·m, leaves a micro-gap. Methanol at 20 °C has a vapor pressure around 97 mmHg; it will find a 20 µm gap.
• Thread bypass. Coarse bottle threads are not designed to seal. They locate. If the liner is compromised, the thread is the escape route. A cap that feels tight is not evidence of a seal.
• Septum puncture. Every needle or tube entry through a silicone or silicone/PTFE septum creates a permanent channel. A 22-gauge needle leaves a hole roughly 0.7 mm across that does not close completely.
• Cap body porosity and liner delamination. Cheaper phenolic caps can be porous at the molding gate. PTFE-faced liners bonded to silicone can delaminate at the puncture site after repeated autoclaving, and once the face lifts, the silicone underneath is a sponge for solvent vapor.

The order matters. I check the liner face against the glass first because it is the most common failure and the easiest to fix. If you are running volatile mobile phases (acetonitrile, methanol, THF, hexane), the liner is doing almost all the work. The glass does nothing to stop vapor; it only stops liquid. If you want to understand the bottle as a system rather than a single part, the anatomy breakdown at reagent bottle is a reasonable starting reference for the components, but the seal behavior is what you have to test on your own bench.

Re-seal behavior after puncture is the number nobody records

Here is the thing most labs miss. A septum is not rated by whether it seals. It is rated by how well it re-seals after the needle is withdrawn, and that is a function of elastomer hardness, crosslink density, and the number of prior punctures. A fresh silicone septum (Shore A 50) will close a 22-gauge hole to within a few micrometers on the first withdrawal. By puncture 20 or 30, the same septum has a visible dimple and a channel that leaks measurably.

I ran a simple bench test years ago that changed how I specify septa for reservoir caps. Using a 4 L bottle filled with 3 L of methanol, capped with a GL45 cap carrying a single-port silicone/PTFE septum, I weighed the bottle at t=0 and every 24 h for 10 days at 25 °C, with no tubing inserted. Then I repeated with a 1/16 in. OD PTFE tube inserted through the septum and left in place. The results were not subtle.

Two things jump out. First, a tube left inserted is far worse than 20 needle punctures, because the tube holds the channel open and gives vapor a continuous path. Second, harder elastomer (Shore A 60) re-seals better than softer (Shore A 50), because it recovers its shape faster. If you are running a multi-day method with a solvent line in the bottle, a septum is the wrong interface. Use a cap with a fixed port and an O-ring compression fitting, or a cap with a dedicated tube port that seals on the tube OD with a ferrule, not on a punctured elastomer.

What re-seal actually means at the bench

Re-seal is elastic recovery over time. A silicone septum compressed by a needle recovers roughly 80% of its deformation within 60 seconds and the rest over hours. The residual channel is what leaks. If your method draws from the bottle every few minutes, the septum never gets a full recovery window, and the leak rate is effectively the sum of all the unhealed channels. This is why a bottle that passes a static leak check in the morning can lose 30 g by the end of a shift.

Compression set is why a cap that sealed last month leaks this month

Compression set is the permanent deformation left in an elastomer after it has been squeezed for a long time at temperature. A cap liner that sits torqued at 2.0 N·m on a GL45 bottle for six months at 25 °C takes a set of 10 to 20% of its original thickness. When you remove the cap and reinstall it, the liner no longer has the springback to fill the same gap, and the seal is weaker even though the torque is identical.

The standard reference is ASTM D395, method B, which measures set after 22 h at 70 °C or 24 h at 23 °C under 25% deflection. For silicone, expect 15 to 25% set under those conditions. For PTFE-faced silicone, the PTFE face does not take a set but the silicone backing does, so the effective seal thickness shrinks. For a solid PTFE liner, there is essentially no set, but PTFE creeps under load and cold-flows, so you get a different problem: the liner thins and the torque drops.

The practical compression set rule

Replace cap liners on a schedule, not on failure. For bottle caps in daily use with volatile solvents, I change liners every 3 months or every 50 punctures, whichever comes first. For long-term storage bottles (standards, reference solutions) that are opened monthly, annual replacement is enough. If your bottle sits on a hot plate or in a 40 °C incubator, halve those intervals. Heat accelerates set, and set is the mechanism.

Torque matters here too. A GL45 cap on a 45 mm neck wants 1.8 to 2.2 N·m. Below 1.5 N·m the liner does not fully compress against the land; above 2.5 N·m you deform the liner and accelerate set. Use a torque wrench with a 45 mm cap adapter, or a calibrated cap wrench. Hand-tight is not a specification. I have seen a lab lose 4% of a hexane fill over a weekend because every cap on the shelf was hand-tightened to roughly 1.0 N·m.

Measuring evaporation loss gravimetrically over a run

You cannot fix what you do not measure, and the measurement here is a balance, a clock, and a log sheet. The protocol is simple enough that any lab can run it.

The 7-day gravimetric protocol

1. Fill the bottle to the stated working volume with the actual solvent or mobile phase you use. Record the exact mass to 0.01 g on a balance rated for the load.
2. Cap to the specified torque (1.8 to 2.2 N·m for GL45). If the method uses a tube through the cap, install it exactly as it runs.
3. Store at the temperature the bottle sees in service, not room temperature if the bottle lives in a 30 °C cabinet.
4. Weigh at t=0, 8 h, 24 h, 72 h, 168 h. Record mass, ambient temperature, and barometric pressure. Do not open the bottle between weighings.
5. Compute loss rate in g/day and as % of fill per day. Flag anything above 0.2% per day for volatile solvents (methanol, acetonitrile, hexane) or above 0.05% per day for aqueous phases.

A sealed GL45 bottle with a fresh silicone/PTFE liner should lose well under 0.05% per day with methanol. If yours is losing 0.3% per day, the leak is real and it is almost always the liner or the torque.

What the numbers tell you

Plot the mass against time. A straight line means a constant leak rate, which points to a steady-state path (thread bypass or a tube port). A curve that steepens means the leak is growing, which points to compression set or liner delamination. A step change after a specific event (a needle entry, an autoclave cycle, a cap swap) tells you exactly which interface failed. This is the same discipline you would apply to a drift problem in chromatography, and it is worth reading how the instrument vendors frame sample handling and loss, for example Thermo Fisher’s chromatography and sample handling resources, because the bottle is part of the sample path whether the method treats it that way or not.

Two field cases where the bottle was the drift

Case 1: the 38 g reservoir and the 0.04 min drift

The symptom was late-peak retention drift over a 96-hour sequence. The measurement that identified it was a gravimetric check: the 4 L reservoir lost 38 g over the run, and a control bottle with a fresh liner lost 4 g over the same period. The difference was a two-port cap with a punctured septum and a solvent line left in place. The fix was a cap with a fixed 1/8 in. port and a ferrule seal on the line, plus a fresh liner on the second port. Loss dropped to 3 g over 96 h, and the drift disappeared. The lesson: a continuous tube through a septum is a permanent leak, and no amount of method tuning fixes it.

Case 2: the standard that drifted 2.1% over three weeks

A calibration standard in a 1 L amber bottle, stored at 25 °C and opened weekly, drifted 2.1% low over three weeks. The bottle was not suspected because the standard was made fresh monthly. A gravimetric check showed 0.9 g/week loss, which on a 1 L fill is 0.09% per week, too small to explain 2.1%. The real cause was compression set on a liner that had been in service 14 months: the cap was re-torqued to the same 2.0 N·m, but the liner had taken a 22% set, so the seal was weaker and the bottle was breathing with each temperature cycle. Replacing the liner and switching to a PTFE-faced silicone with a Shore A 60 backing dropped loss to 0.2 g/week, and the standard held within 0.3% over the next month. The lesson: set is invisible on a torque wrench, and a liner has a service life.

If you are running volatile standards through an autosampler, the same seal logic applies to the vial, and it is worth comparing the cap and septum specs on the HPLC autosampler vial side, because a leaking vial cap produces the same drift signature as a leaking reservoir and gets misdiagnosed as an injection problem far more often than it should. For a broader view of where the bottle sits in the sample path, the consumables references at Restek and the method framework in EPA SW-846 are useful for thinking about how sample handling and preservation interact with container choice.

Troubleshooting checklist for seal failure and evaporation loss

Work through this in order. The first three items catch the majority of cases.

• Weigh the bottle. If you have not measured loss gravimetrically, you do not have a diagnosis. Run the 7-day protocol above.
• Check torque. GL45 caps at 1.8 to 2.2 N·m. Below 1.5 N·m is a leak; above 2.5 N·m is set acceleration.
• Inspect the liner face. Look for a ring-shaped compression mark, dimples from punctures, and any lifting of the PTFE face. A liner with a visible dimple is done.
• Count punctures. More than 20 on a silicone septum means replace it. More than 50 on any septum means replace it regardless of appearance.
• Check the tube interface. A tube through a septum is a permanent leak. Move to a ferrule-sealed port or an O-ring compression fitting.
• Check the neck land. A chipped or out-of-round GL45 land will not seal with any liner. Inspect with a loupe and reject the bottle.
• Check storage temperature. Every 10 °C above 25 °C roughly doubles the set rate and the vapor pressure. A bottle in a 40 °C cabinet loses faster than the same bottle on the bench.
• Log liner install dates. If you cannot say when the liner went in, it is past due.

Selection criteria that actually matter

When you specify a bottle for volatile service, the decision comes down to a few concrete parameters. Neck finish: GL45 for standard 500 mL to 4 L reservoirs, with a liner land that is machined flat, not molded with a parting line. Liner: PTFE-faced silicone with a Shore A 60 backing for puncture service, solid PTFE for long-term storage where creep is acceptable. Cap: PP or PPS with a defined torque rating, and fixed ports with ferrule or O-ring seals if any tube is involved. Temperature rating: silicone liners to 200 °C, PTFE to 260 °C, but the cap body is usually the limit at 120 to 140 °C. For a deeper look at how bottle dimensions and neck geometry affect fit, the fit-problem framing on sample vial pricing and specification pages is a useful cross-check, and the chromatography consumables catalog at chromatography products is where I would start for matched cap and liner sets.

The decision rule

If a bottle loses more than 0.2% of fill per day with a volatile solvent, or more than 0.05% per day with an aqueous phase, it has a seal failure and you should stop using it for that method until the liner, torque, and any tube interface have been replaced and re-verified gravimetrically. Replace liners every 3 months or 50 punctures, torque GL45 caps to 1.8 to 2.2 N·m, and never leave a tube through a punctured septum if the bottle sits in service for more than a shift. Those four numbers will catch almost every evaporation and sample loss problem that gets blamed on the instrument.

References & further reading

Related resources from our lab consumables network:

• Reagent Bottle at reagentbottleforsale.com
• Hplc Autosampler Vial at hplcautosamplervials.com
• Sample Vial at samplevialprice.com
• Chromatography Products at chromatographyselling.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