Keep liquid aliquots at or below a few percent of the reactor volume, and purge a gas syringe with three to five fill-and-vent cycles before you inject. The dominant sampling error is rarely the analyzer's resolution: it is the pressure drop, the air ingress and the change in headspace composition that your own extraction creates.
A sample changes the system it is measuring, and the change does not wash out. Three couplings matter in a photoreactor. Removing liquid lowers the liquid level, which changes the illuminated volume in a side-illuminated vessel and changes the headspace-to-liquid ratio in any vessel. Withdrawing gas lowers the total pressure, and every septum puncture is an opportunity for room air to enter and dilute what is left. And each extraction resets the dissolved-gas equilibrium that the run is following, so the reading you take is of a system that has just been disturbed.
None of this makes sampling optional; it means the protocol should be fixed before the first run and repeated exactly, so the perturbation cancels in a rate comparison. A protocol that changes mid-campaign produces a step that looks like a catalyst effect.
Take a few percent of the charge per sample and count the cumulative removal across the whole run, because the total is what changes the experiment. A single small aliquot is harmless; eight of them in a six-hour run are not, and the fall in apparent rate they produce is easy to misread as deactivation.
If the reactor is illuminated through a side wall, watch the liquid level as well as the volume. The illuminated path shortens as the level drops, so the photon flux absorbed by the suspension falls even when the lamp and the catalyst are unchanged. Front-window reactors with a fixed liquid depth avoid that particular artifact but still lose inventory, and the concentration you report refers to whatever remains. Where frequent sampling is unavoidable, a circulation loop that draws, measures and returns the liquid keeps the inventory constant and is worth the extra line volume.
Three to five purge cycles is arithmetic rather than folklore: each fill dilutes the previous contents by roughly the ratio of the drawn volume to the dead volume, so a small dead volume clears in three cycles and a large one does not clear in five. The table maps the common routes to the error each one introduces.
| Sampling route | What it changes | Dominant error | Control |
|---|---|---|---|
| Manual syringe from a septum port | Removes gas; the headspace pressure drops | Needle dead volume and air drawn in on venting | Three to five purge cycles, inject immediately, keep the handling time constant |
| Fixed-loop valve injection | Removes a fixed volume only if the loop is filled at system pressure | Under-filled loop, or a loop filled at a lower pressure than the reactor | Flush three to five loop volumes, fill at system pressure, then switch |
| Circulation loop to the analyzer | Removes nothing; adds line volume | Transit lag and adsorption on the line walls | Read only after the signal plateaus, and record the lag with the line volume |
| Headspace draw after a liquid sample | Lowers pressure a second time in the same step | Dilution by in-leakage while the vessel is below working pressure | Re-pressurize with the same gas before the next reading |
Read the table by column two: the routes differ less in what they measure than in what they change. A fixed loop at system pressure is the closest thing to a non-invasive gas sample; a manual syringe is the most flexible and the most dependent on technique. Pick one and keep it for a campaign; switching mid-series introduces a step larger than most catalyst effects.
Most disagreement between two injections of the same sample comes from four sources, and only one of them is the instrument.
Record time, volume, system pressure before and after, the number of purge cycles and the analyzer's response, then inject a duplicate every fifth sample. Duplicates that agree within the analyzer's repeatability say the protocol is under control; duplicates that disagree by more point at the sampling, and no amount of reactor work will fix them.
Keep the log with the run, not the instrument: whoever reads the rate curve later needs to know whether the dip at hour three was chemistry or a larger aliquot.
Three to five fill-and-vent cycles, judged against the dead volume of the syringe and needle rather than the syringe barrel alone. Each cycle should draw more than the dead volume, and the last fill should be taken slowly so that the sample is at the pressure you will record. If two consecutive injections of the same sample still disagree, add cycles before you blame the analyzer.
Not entirely, but you can keep the change small and constant. Take a few percent of the charge, re-pressurize with the same gas afterwards, and log the pressure before and after every draw. A loop that returns the liquid avoids the inventory loss but adds line volume, which is a different compromise, not no compromise.
Usually because the two were taken at different pressures, or the first sat in the syringe while the second was drawn, or the purge was incomplete. Check the pressure record first, then the handling time, then the number of cycles. If all three are identical and the disagreement persists, the syringe or the septum is leaking and should be replaced.
