A photoreactor purchase document only needs six numbers with tolerances: irradiated area in cm2, irradiance in mW/cm2 at a stated wavelength, working volume in mL, temperature range with stability in °C, pressure rating in MPa, and leak rate in kPa/min. Everything else is a preference.
"High intensity" has been quoted to mean 8 mW/cm2 and 800 mW/cm2 in the same tender round. "Gas tight" ranges from 5 %/h to 0.01 %/h, and the second one needs welded metal, not a greased joint. When the requirement is an adjective, the supplier picks the number, and the cheapest compliant bid is the one that reads the adjective most loosely. A number with a tolerance and a test method fixes the price instead of the argument.
The commercial consequence is real: going from a 20 cm2 to a 40 cm2 window doubles the photons needed at the same irradiance, which generally means a larger lamp, more heat to remove and a bigger chiller. Specifying area and irradiance separately is what makes those costs visible before the order.
| Parameter | Write this | Typical value | Verification at acceptance |
|---|---|---|---|
| Irradiated area | Clear aperture, ±5 % | 20 cm2 (50 × 40 mm) | Measure the lit rectangle on graph paper |
| Irradiance | mW/cm2 at a named line, at the window plane | 100 mW/cm2 at 365 nm | Calibrated meter, 9 points, ±10 % spread |
| Working volume | mL, with maximum headspace | 250–500 mL, headspace ≤10 % | Fill and weigh |
| Temperature | Range, stability, overshoot | 10–80 °C, ±1 °C, overshoot <2 °C | Step 25→60 °C, log 30 min |
| Pressure | Gauge MPa, at temperature | 0.1 MPa borosilicate, 0.5 MPa metal body | Hydrostatic hold 1 h |
| Leak rate | kPa/min at a stated gauge pressure | ≤0.5 kPa/min at 50 kPa, 30 min | Cap all ports, log the decay |
Two more belong in an annex if the work is quantitative: dead volume ≤5 % of working volume, and gas flow control of 10–200 mL/min at ±1 % of full scale. Neither is expensive to meet, and both are expensive to discover late.
Lamp power is a property of the lamp; irradiance at the window plane is a property of the system, and only the second one enters a rate law. A 300 W xenon source with a 365 nm bandpass filter typically delivers 80–150 mW/cm2 at 150–200 mm, but the same lamp behind a 10 mm water filter, a fibre bundle or a 20 % obscured mirror will not. So write: "100 ±10 mW/cm2 over the 365 ±10 nm band, measured at the inner window face with a calibrated radiometer, nine-point grid, maximum deviation 10 % from the mean."
Uniformity deserves its own clause. A 20 % hot spot at the centre is common with a short arc and no homogeniser, and it changes the apparent kinetics because the centre of the cell runs at a different rate from the edge. If the science needs flat illumination, say so and expect a light guide or a diffuser in the quote.
Window material follows wavelength, not habit. Borosilicate transmits above about 320 nm and is fine for 365 nm work; below 300 nm you need quartz, which passes from 200 nm and costs several times more. Ordering quartz "for safety" at 365 nm buys nothing but a larger bill and more fragile handling.
Seals are the usual weak point. Viton A suits 0–200 °C and vacuum; EPDM is better for aqueous work at moderate temperature; silicone permeates oxygen roughly two orders of magnitude faster than PTFE and will quietly ruin an anaerobic run. Specify the elastomer by name and, if oxygen ingress matters, ask for a permeability figure or a metal-to-metal seal. Wetted metal: 316L for anything with chloride traces or CO2 in water, 304 only for clean neutral systems.
Three tests catch most of what goes wrong, and all three are cheap. A 30 min pressure-decay test with every port capped, at the rated gauge pressure, logged rather than eyeballed. A nine-point irradiance map at the specified distance, with the meter certificate number recorded. A blank run under the specified light, temperature and purge for 2 h, which must return below three times the gas analyser detection limit — this one exposes residual contamination, leaks and dead zones in a single measurement.
Ask for the results as data, not as a sentence. A supplier who can produce the decay curve and the nine-point grid has a calibrated rig; one who cannot is quoting a catalogue.
Brand names, colours, screen size and the word "intelligent" do not belong in a specification, and each one narrows the field without improving the science. Conversely, do not leave out the boring items that generate support tickets: spare window gaskets, septum type and size, the fitting standard (GL14, 1/4 in Swagelok), and whether the lamp hour meter is resettable. These cost almost nothing to specify and a week to retrofit.
Both, with different roles. Irradiance at the window plane is the acceptance criterion because it enters the rate law; lamp power is a maintenance parameter used to predict replacement intervals. A 300 W lamp is a purchasing detail, 100 mW/cm2 at 365 nm is a scientific requirement.
≤0.5 kPa/min at 50 kPa gauge over 30 min for most photocatalytic work. Tighten to ≤0.05 kPa/min only if you measure products at ppm level in a sealed batch, because the tighter number forces welded metal and changes the price bracket.
Only below 300 nm. Borosilicate transmits above about 320 nm, so 365 nm LED and xenon work needs no quartz. If you plan 254 nm or 185 nm experiments, specify quartz and accept the higher cost and fragility.
Two pages: one table of six numbers with tolerances, one page of acceptance tests and annex items. Longer documents get read as negotiation, shorter ones get interpreted by the supplier.
