Pick the wavelength from the band gap, not from preference: the ceiling is 1240/Eg. Anatase TiO2 (3.2 eV) needs 388 nm or shorter, so 365 nm UV is mandatory; g-C3N4 (2.7 eV, 459 nm) works at 420–450 nm, and BiVO4 (2.4 eV, 517 nm) at 455–470 nm. Light beyond the absorption edge is wasted at any irradiance.
A semiconductor absorbs only photons whose energy exceeds its band gap, and the conversion is λ(nm) = 1240 / Eg(eV). That line decides whether you buy a UV or a visible source, and sets the wavelength at which apparent quantum yield becomes meaningful. Sub-gap absorption from dopants or defects exists but is weak, so it does not justify a 450 nm source for a 3.2 eV material.
| Photocatalyst | Band gap (eV) | Absorption edge (nm) | Practical source |
|---|---|---|---|
| Anatase TiO2 | 3.2 | 388 | 365 nm UV LED or 365 nm mercury line |
| Rutile TiO2 | 3.0 | 413 | 385–405 nm LED |
| g-C3N4 | 2.7 | 459 | 420–455 nm LED |
| BiVO4 | 2.4 | 517 | 455–470 nm LED |
| CdS | 2.3 | 539 | 470–520 nm LED |
Three options dominate. Low-pressure mercury lamps emit almost entirely at 254 nm, far above the band gap of most oxides, and generate ozone in air; medium-pressure lamps give a line spectrum at 254, 313, 365 and 436 nm, most of it in the lines you did not want. UV LEDs at 365, 385 or 395 nm have a full width at half maximum of 10–15 nm and dimmable output: a 100 W array typically reaches 100–300 mW/cm² over a 50 × 50 mm area.
Band-pass filtering a xenon lamp is the least efficient route: a 10 nm band-pass passes only 3–5% of the lamp’s radiant output, so a 300 W source yields a few mW/cm² in the passband. It suits wavelength scans, not photon-hungry rate tests.
Visible-light photocatalysis usually means 420, 450 or 470 nm LEDs, or a xenon lamp with a 420 nm long-pass filter that throws away 60–70% of the radiant power. Two problems follow. Narrow-gap materials often have low absorption coefficients in the visible, so light penetrates deep into a slurry and the apparent rate depends on catalyst loading and optical path. And the irradiance needed is higher: 100–300 mW/cm² at 450 nm is common against 20–100 mW/cm² at 365 nm, because each 450 nm photon carries 2.76 eV against 3.40 eV at 365 nm.
Ambient light matters more in the visible. Laboratory lighting contributes roughly 0.5–1 mW/cm² at the reactor, which is 2–5% of a 20 mW/cm² experiment. Run the dark control with the room lighting in its normal state, or shield the reactor.
Apparent quantum yield is molecules converted per incident photon, so the photon count has to come from the actual spectrum. One watt of 365 nm light is 3.06 µmol photons per second; one watt of 450 nm light is 3.77 µmol/s. Quote the photon flux in µmol/s or µEinstein/h, state the wavelength and bandwidth, and state whether the number is based on incident or absorbed photons. A 365 nm and a 450 nm AQY on the same material answer different questions and should never share a comparison column without that context.
UV work forces material choices that visible work does not. Borosilicate transmits to about 320 nm, so 365 nm is acceptable but 254 nm is not — that needs quartz. FKM and EPDM seals harden and crack under 365 nm after a few hundred hours; silicone and perfluoroelastomer parts last longer. Any plastic tubing in the light path will yellow. Visible setups are easier on materials but harder on heat: a 100 W LED array dumps 60–70 W as heat into a small head, and uncorrected that shifts the emission peak by 2–3 nm and costs 10–15% of output within an hour.
Only after modifying the material. Pristine anatase absorbs less than 5% of a 450 nm beam, because 450 nm corresponds to 2.76 eV against a 3.2 eV gap. Visible activity in the literature comes from doping, oxygen vacancies, dye sensitisation or heterojunctions — not from the lamp. If you change nothing but the source, the rate drop is the honest result.
It is UVA, but it sits at the edge for TiO2: 395 nm equals 3.14 eV, just below the 3.2 eV gap, so absorption is weak and confined to defect states. Use 365 nm when the material is anatase, and reserve 385–405 nm for rutile or for narrow-gap systems where the edge is comfortably longer.
365 nm, with the bandwidth stated, because it is above the band gap, available from both LEDs and mercury lines, and comparable with published values. Report the incident photon flux in µmol/s, the illuminated area, and the fraction of light absorbed by the suspension, otherwise the number cannot be reproduced.
Usually only the window and the seals. Quartz is required below 320 nm and is harmless above it, so a quartz-window reactor covers both. What does change is thermal load and beam uniformity: visible LED arrays are larger and less collimated, so check irradiance at the bed centre and edge separately before comparing with an earlier UV dataset.
