Use batch below about 500 mL and 20 runs per day; switch to flow once you need over 1 g/day of product or a residence time held within ±10 %. A 100 mL flow module at 10 mL/min processes 14 L per day against 0.4 L for the same volume run as a 2 h batch.
Photons do not care about volume, they care about path length. In a 0.5 g/L TiO2 slurry, more than 90 % of 365 nm light is absorbed within the first 2–3 mm, so a 50 mm batch cell spends most of its volume in the dark: the illuminated fraction is roughly 5 %, and stirring only mixes dark slurry into the lit layer.
A flow channel of 1–3 mm puts the whole reacting volume inside the absorption depth. The comparison that matters is volumetric rate of photon absorption, in mW/cm³ of reactor volume, not incident irradiance in mW/cm² at the window. Two rigs with the same 100 mW/cm² lamp can differ by a factor of ten in that number purely because of geometry.
This is also the reason batch has a ceiling. Widen a batch cell to raise throughput and the dark fraction grows; the usual fix is an annular geometry with the lamp inside, which recovers path length but multiplies glassware and cleaning effort.
| Metric | Batch | Continuous flow |
|---|---|---|
| Optical path | 20–60 mm, one lit face or an annulus | 1–3 mm channel or 0.1–0.3 mm film on a plate |
| Residence time control | Set by run length; ±1 min repeatable | Set by V/Q; ±5 % with a calibrated pump |
| Throughput at 100 mL | 0.3–0.5 L/day with cleaning between runs | 14 L/day at 10 mL/min, continuous |
| Residence time distribution | All molecules see the same time | Axial dispersion; 8–16 modules in series narrow it to ±10 % |
| Catalyst form | Slurry 0.2–2 g/L, easy to change | Immobilised film or 100–300 µm packed bed; slurry needs a settler |
| Gas-liquid mass transfer | Dependent on sparger and stirrer, kLa 0.01–0.05 s−1 | Segmented flow or falling film, kLa 0.1–1 s−1 |
| Scale-up route | Size-up, blocked by light penetration past 100 mm | Number-up, 8–16 identical modules |
| Typical failure mode | Hot spot at the window, settling slurry | Channel fouling, bubble trapping, pump pulsation |
Three things, and the first is usually the real reason a group switches.
Reproducibility. A batch run has a warm-up: the lamp needs 10–20 min to stabilise, the slurry needs to reach adsorption equilibrium, and dissolved oxygen drifts. Every run repeats that transient, which is why run-to-run scatter of ±10 % is normal. A flow rig reaches steady state once and then holds it for hours; sampling at 30, 60 and 90 min gives three points on one steady state instead of three separate runs.
Parameter space per gram of catalyst. Scanning residence time from 2 to 30 min in batch means five runs and five workups. In flow it means changing the pump setting, waiting three residence times, and sampling — roughly 40 min for the whole series, on 50 mg of catalyst.
Safety and pressure. Small-bore flow hardware holds a few millilitres at pressure; a 1 L batch autoclave holds a litre. For CO2 hydrogenation at 2 MPa the inventory difference is the whole risk assessment.
Do not convert for its own sake. Batch remains the better choice when the catalyst settles fast (particles above about 100 µm block narrow channels), when the reaction needs hours not minutes, when you have 5 mL of a precious substrate, when solids form and would plug a 1 mm channel, and when the experiment is a screen of twenty catalysts where changing the powder is the main activity. Screening first in 20–50 mL batch vials, then confirming the winner in flow, is a sensible division of labour.
The crossover is usually economic rather than scientific: the moment a run's product is needed in grams, the labour of repeated batch workups exceeds the cost of building the loop.
Answer five questions before specifying hardware. Is the required product mass above 1 g/day? Is the target residence time under 60 min? Is the optical path required below 5 mm to absorb the light? Will the same conditions be run more than 20 times? Can the catalyst be immobilised or kept in suspension without settling? Four or more yes answers point to flow; two or fewer point to batch; a split answer points to a batch screen followed by a flow confirmation.
Per pass, usually yes, because residence time is short and the outlet is a mixed sample. Final conversion can match batch by recirculating the outlet through the loop, which is a batch in disguise but keeps the thin-channel optics. Compare on space-time yield, in mol per litre of reactor per hour, not on single-pass conversion.
Yes, but the particles must stay suspended: 0.2–0.5 g/L in channels above 1 mm, with flow fast enough to exceed the settling velocity — roughly 0.5 mm/s for 5 µm anatase aggregates. Most groups immobilise the catalyst instead: a 1–5 µm film on a plate, or a 100–300 µm packed bed.
Ten, run in parallel rather than in series, each with its own lamp and pump channel. Series connection raises residence time and pressure drop, not throughput, and one fouled module then contaminates the whole train.
Only above a threshold. Below about 20 runs per week the labour of cleaning and re-plumbing dominates and batch is cheaper; above that, flow wins because a flush takes 5–10 min against 30–60 min for dismantling, cleaning and refilling a batch cell.
