Rising background, poor repeatability and distorted peaks are contamination signals, and they usually appear in that order. A system that still passes a leak check at ≤0.1 μmol/h measured as oxygen can hold enough residue in its lines to shift a hydrogen reading, so diagnose the path before you blame the catalyst.
Contamination announces itself in the blank, not in the sample, and that is the difference that matters. A deactivated catalyst produces less product at a stable background, while a contaminated system produces a drifting background that changes what the same catalyst appears to do. Run a catalyst-free blank before concluding anything about the material.
| Signal | What contamination looks like | What deactivation looks like | Test that separates them |
|---|---|---|---|
| Rising background | Signal appears without catalyst and grows across the run | Background flat, product rate falls | Catalyst-free blank under the same light and purge schedule |
| Poor repeatability | Scatter between identical runs widens and the order of runs matters | Repeatable decline from run to run | Two consecutive blanks, a reference sample, then a blank again |
| Distorted peaks | Tailing, split peaks, or a shoulder that appears only after a long run | Peak shape unchanged, area smaller | Direct injection of a standard through the same path |
| Response to cleaning | Signal drops after a flush and returns over hours | No change after a flush | Flush one segment at a time and re-measure |
The fourth row settles most arguments. If a flush changes the number, the number was coming from the path. If a flush changes nothing and the rate is still low, the catalyst or the light is the place to look, and the light deserves a check before the catalyst does, because a hazed window and a loaded cooling filter both reduce rate without touching the chemistry.
Clean the gas path from the sample end backward, so residue is never pushed into a component you have already cleaned. Isolate and remove the reactor, then work through tubing, valves and the sampling loop in that order, replacing anything that cannot be cleaned. Tubing is the cheapest part of the system and the most likely to hold a film, so replace it rather than trying to flush a long narrow run.
Verification is a blank run against the detector's own floor. With a trace gas analyzer whose detection limit is 0.05 μmol for hydrogen and 0.0005 μmol for methane and carbon monoxide, a clean path should return a background well below the smallest peak you intend to quantify, and it should stay flat for the length of a normal run. A background that is small but rising is not clean; it is a slow release from somewhere in the path.
Clean the liquid path in the direction the sample travels, and give the dead volumes real attention. A water-splitting system with 65 mL of line volume holds a lot of liquid in places that a rinse passes straight through, and a rinse that only wets the walls removes almost nothing. Recirculate, hold, then drain, and repeat with fresh solvent.
Sequence the solvents by compatibility rather than by strength: water first for aqueous residues, then a solvent that dissolves the organic deposit, then water again to remove the solvent, and finally the reaction medium so the line is conditioned before the run. Between steps, drain fully; a mixed solvent left in the line can precipitate what the previous step dissolved. Avoid detergent, which leaves a surfactant film that outgasses into the headspace and shows up in gas measurements long after the rinse.
Catalyst fines are a special case. They settle in low points and in valve seats, and they are invisible in a clear line until they resuspend and produce a spike. Flush with the flow direction, then against it if the design allows, and inspect any transparent section against the light. If fines keep appearing, the problem is upstream in the sampling or separation step rather than in the line itself.
A path is clean when a blank run returns a flat background and two repeats agree within your repeatability band. Do not accept a single low reading as proof; contamination that is slowly desorbing produces a low first measurement and a rising second one, which is easy to miss if you only take one.
Record the blank alongside the run so the background can be compared across campaigns. A background that creeps upward over months is a maintenance signal, not a detector quirk, and it usually traces to a septum, a trap or a length of tubing that has been in service too long. Keeping the trend visible turns a mysterious loss of sensitivity into a scheduled replacement.
Most contamination events come from three habits: reusing septa and ferrules, sharing tubing between different reaction systems, and leaving sample in the line between campaigns. None of them is expensive to change. Dedicate a tubing set per chemistry, replace consumables on a count rather than on failure, and flush the line before the system is shut down rather than after it is started up again.
Keep the reactor itself in the same routine. Residue on the window and on the inner wall behaves like contamination in the data, because it changes the light reaching the sample, so the window check belongs in the same maintenance pass as the line flush. Two checks, one visit, and the next campaign starts from a known state.
Run a catalyst-free blank under the same light, purge and sampling schedule. If the background rises without catalyst present, the signal is coming from the system. If the blank is flat and the product rate still falls with the same catalyst, the material or the illumination is the place to look next.
No. Gas lines are cleaned by purge and replacement, and the risk is moisture and desorption from elastomers. Liquid lines are cleaned by sequential rinses, and the risk is trapped dead volume and solvent residue. Applying one procedure to both usually leaves the other kind of contamination behind.
Until the blank is flat, not for a fixed time. Purge in stages and measure between them: the background should stop falling and hold steady for the length of a normal run. A fixed duration assumes the deposit is gone, which is exactly what the measurement is there to establish.
