Control and report the bulk liquid temperature, not the chiller setpoint. At 100 mW/cm² over 20 cm² a photoreactor absorbs roughly 0.4–0.6 W, which lifts 100 mL of water 2–3 °C with a 35 min time constant — invisible in a 10 min run. Hold the bulk at ±0.5 °C; 1 K of drift moves the rate 5–7 %.
A xenon lamp is not a cold light source. Roughly half the radiant power reaching the window sits above 700 nm, and water absorbs that band strongly enough that a 10 mm path removes a quarter of it. Photons that do drive chemistry add little: at a quantum yield of a few percent, only a few percent of absorbed photon energy is stored in products, and the rest relaxes as heat within picoseconds.
| Spectral band | Share at window | Main absorber | Effect |
|---|---|---|---|
| 250–400 nm | 5–10 % | Catalyst, quartz | Photochemistry; negligible bulk heating |
| 400–700 nm | 30–40 % | Catalyst, coloured solutes | Product formation plus local heating |
| 700–1100 nm | 30–40 % | Water (975 nm overtone), glass | Dominant heat load in aqueous slurry |
| 1100–2500 nm | 10–20 % | Water beyond the first millimetre | Heating with no photochemical benefit |
A worked balance for a slurry cell: 2.0 W incident on 20 cm², about 1.0 W in the near-infrared, 25 % of that absorbed in the first centimetre (0.25 W), plus 0.15–0.25 W of visible light absorbed by 0.5 g/L catalyst. Absorbed power P ≈ 0.5 W. With 0.02 m² of wetted glass and a natural-convection coefficient near 10 W/(m²·K), the loss conductance is hA ≈ 0.2 W/K, so the steady-state rise is ΔT = P/hA ≈ 2.5 K, with time constant τ = mcp/hA = 419 J/K ÷ 0.2 W/K ≈ 2100 s. Three time constants to 95 % is 105 min — longer than most kinetics runs.
For an apparent activation energy of 40–50 kJ/mol near 298 K, the rate constant rises 5–7 % per kelvin. A 3 K uncontrolled drift therefore inflates the rate 16–22 % between the first and last sample of a 90 min batch, which is routinely reported as an "induction period" at the start or as deactivation at the end.
Temperature also moves what you measure. Henry solubility falls about 2 %/K for O2, 2.5 %/K for CO2 and 1.5 %/K for H2 near 25 °C, so a 3 K rise pushes dissolved CO2 into the headspace and changes the concentration you assumed constant. Water vapour pressure climbs from 3.17 kPa at 25 °C to 3.69 kPa at 28 °C, a 17 % increase that shows up as condensation in cold transfer lines and a wandering TCD baseline.
A bare thermocouple junction sitting in the beam reads 1–3 K high at 100 mW/cm² because it absorbs radiation itself. The bias scales with irradiance, so it looks like a plausible number while being systematically wrong.
| Probe arrangement | Accuracy in use | Response (63 %) | Where it fails |
|---|---|---|---|
| Chiller or jacket setpoint | ±0.5 K on coolant | 5–10 min | Measures the loop, not the suspension |
| Bare 0.5 mm thermocouple in beam | 1–3 K high | 1–3 s | Radiative heating of the junction |
| Sheathed 1 mm probe, shielded, in bulk | ±0.3 K | 5–20 s | Slow in viscous slurries |
| Pt100 in a wall well | 1–3 K low at 100 mL | 10–30 s | Reads wall, not bulk |
| IR thermometer on window | ±2 K | <1 s | Reads wet surface, emissivity unknown |
Keep one probe, in one position, for every run in a series. Change probe or position mid-campaign and the temperature step is indistinguishable from a real effect on rate.
1. Cool the bulk, not the jacket. A jacket at 15 °C with slurry at 25 °C removes heat only as fast as the wall conducts; at 0.2 W/K you need a 2.5 K wall-to-bulk difference to shed 0.5 W. Set the chiller low enough to hold the bulk, and log the bulk.
2. Cut the infrared before it enters. A 10 cm water filter or a KG-type heat-absorbing filter removes most of the 700–1100 nm band and typically halves the heat load. Report the transmitted spectrum and the irradiance measured after the filter, since cutting the infrared changes the photon count behind any quantum-yield figure.
3. Shorten the optical path. A 1–3 mm flat-plate channel holds far less water per unit area, which is one reason thin-layer reactors drift less than a 50 mm beaker.
4. Pre-equilibrate under illumination. Start the clock only after the bulk has stayed within 0.2 K for 10 min — for the cell above, 45–60 min of pre-irradiation. That removes the light-on transient instead of fitting it.
Give bulk temperature at the first and last sample, the coolant setpoint and flow, and the measured ΔT. State probe type, sheath diameter and position relative to the beam. Say whether an infrared filter was used, and give irradiance at the window at operating temperature, not at room temperature — a warm meter head reads a few percent low. Reviewers asking about "temperature control" want to know whether the compared runs sat at the same temperature; a setpoint alone does not answer that.
At 100 mW/cm² with no cooling, a 100 mL aqueous slurry typically settles 2–3 K above ambient within an hour and keeps creeping for a second hour. At photothermal irradiance of 1–5 W/cm² the same cell can exceed 60 °C, and cooling becomes a design requirement.
For rate comparisons at the 10 % level, yes. For 5 % reproducibility, no: 1 K is 5–7 % of the rate constant, so control to ±0.5 K and report the spread you actually achieved.
Only through the photon count, not the chemistry, provided the filter removes no photons the catalyst absorbs. Integrate photon flux over the absorption band of the catalyst after the filter, not over the full lamp output.
The jacket measures coolant, and heat must cross the wall to reach it. A 0.5 W load through a 0.2 W/K wall needs a 2.5 K gradient, so the bulk sits above setpoint by design. If the offset drifts mid-run, light output or stirring has changed.
