Replace seals on duty, not on the calendar: a seal that has been thermally cycled at 1.6 MPa is finished long before one that has held ambient liquid through a dozen runs at the same reading. Replace on evidence — permanent compression set, swelling, or a leak that returns after re-torque — and choose the material by solvent and temperature before anything else.
Seals fail by four mechanisms, and three of them leave no visible damage until the leak starts. Compression set is the first: the elastomer takes a permanent new shape and stops pushing back, so the joint loses preload even though the hardware has not moved. Chemical swell is the second: the seal absorbs solvent, grows, and either extrudes into the gap or tears on the next opening. Thermal cycling is the third, and it is the one that quietly consumes life on every heat-up. The fourth, mechanical damage during installation, is the only one you can see before it costs you a run.
That order matters for diagnosis. A seal that leaks at pressure but holds at ambient has usually taken set. A seal that swells and jams in its groove has a solvent compatibility problem, and replacing it with the same material will reproduce the failure. A seal that weeps only at temperature is often the wrong grade for the upper end of the cycle.
Choose the material from the media, the temperature and the pressure, in that order, and treat the seal datasheet as the authority rather than habit. The table gives the working logic for the common options in laboratory photoreactors.
| Material | Where it fits | Where it fails | Practical note |
|---|---|---|---|
| Fluoroelastomer (FKM) | Aqueous systems, dilute acids and bases, moderate temperature gas work | Polar solvents, amines, hot water above its service range | Good default for liquid-phase reactors; verify against your solvent list |
| Perfluoroelastomer (FFKM) | Aggressive solvents, oxidizing media, high-temperature gas service | Not a general answer to mechanical damage; costs more and is softer | Use where FKM swells; confirm the specific grade for your medium |
| PTFE and PTFE-encapsulated | Broad chemical resistance, static joints, low-friction faces | Cold flow under load, poor recovery after thermal cycling | Needs a defined groove and higher preload; not for repeated opening |
| Silicone | Wide temperature band, dry gas, optical and window seals | Non-polar solvents, which swell it badly | Acceptable for window seats, poor for solvent contact |
| EPDM | Water, steam, dilute aqueous chemistry | Mineral oils and most hydrocarbon solvents | Cheap and effective for water-only loops |
Read the table as a first filter, not a specification. Two seals of the same nominal material can behave differently by grade, and the only reliable test is immersion of a spare in the actual medium at the actual temperature for the length of a run. Do that once per new solvent and keep the result with the reactor log, because it will save you the same test next year.
Most leaks that appear immediately after a service are installation faults, not material faults, and they repeat because the same person makes the same move. The common ones are worth listing explicitly.
Clean the groove with a lint-free wipe and a solvent that evaporates completely, then let it dry. A seal fitted onto a wet groove traps solvent, which either attacks the elastomer or outgasses into the reactor later and shows up as an unexplained background in the gas analysis.
Set the interval from the number of pressure and temperature cycles the seal sees, and shorten it for solvent contact, because those are the two loads that consume life. A seal in a static ambient liquid reactor may last many campaigns; the same seal in a vessel that is heated to 300 °C and cooled repeatedly will not, even though the calendar looks identical. Write the cycle count on the maintenance card, not just the date.
In gas-tight systems the acceptance test is quantitative. A water-splitting rig with 65 mL of line volume is expected to hold a leak rate at or below 0.1 μmol/h measured as oxygen, and a joint that has just been resealed should be proven against that number, not against the absence of a visible bubble. Pressure-hold tests find gross leaks; they do not find the slow ingress that ruins a hydrogen measurement over a long run.
Work from the outside in and stop at the first failure, because opening a joint that was already sealing introduces a new risk. Start with a pressure hold on the assembled system and watch the gauge over a period long enough to distinguish a real drop from thermal equilibration. If it fails, isolate sections with valves and repeat; the segment that fails contains the leak. Only then break the joint, and inspect the seal before you touch the flange, since the deformation pattern tells you which fault you are looking at.
When the leak sits at a heated flange, repeat the test at temperature. A seal that holds cold and weeps hot is a material or preload problem, and re-torquing it cold will not fix it. Record the outcome, including which seal was fitted and how many cycles it had seen, so the next replacement decision is based on evidence rather than on memory.
Base it on cycles and media rather than on months. Replace when the seal has taken visible set, when it has swollen in service, or when a joint that was re-torqued leaks again within a run. For vessels that see repeated heat cycles, replacement is a scheduled consumable; for static ambient liquid work, inspection at each opening is usually enough.
No. PTFE tape is a thread sealant for tapered fittings, and it does nothing for a face seal. On a flange or a window seat it migrates under load, leaves fragments in the reactor, and produces a joint that holds once and then creeps. Use the specified seal in the specified groove.
Because the seal and the flange expand by different amounts, so preload changes as the joint reaches temperature. Either the material is wrong for the upper temperature, the groove has no room for expansion, or the bolts were torqued cold with no allowance for differential growth. Test at temperature before you trust the assembly.
