Choose by absorption band and by how often you restart, not by wattage. Xenon gives a continuous spectrum, mercury concentrates output in UV lines, and metal halide adds visible-emitting additives to a mercury-type arc; for general photocatalysis a 300 W xenon source with 50 W of total optical power remains the flexible default.
The lamp is wrong if it emits where the catalyst does not absorb. Measure the UV-Vis diffuse reflectance spectrum first, read the absorption edge, and treat everything longer than that edge as heat rather than chemistry.
Emission outside the band is not neutral. It warms the reactor and can drive thermal side reactions you will later mistake for photocatalytic activity, and it makes photothermal claims untestable because the two contributions can no longer be separated.
Two consequences follow. A broadband source plus a band-pass filter is often a better match than a source whose native spectrum is narrow, because the filter is a documented and replaceable component. And the wasted band still has to be managed thermally. The broader survey of lamp families is in the existing lamps page.
Xenon, mercury and metal halide differ on four axes that matter operationally: spectral shape, restart behavior, how output degrades, and handling requirements. The table compares them on those axes rather than on headline wattage.
| Property | Xenon short arc | Mercury arc | Metal halide |
|---|---|---|---|
| Spectral character | Continuous from the UV through the near infrared, with some strong lines at the long-wavelength end | Discrete lines, with the strongest output in the UV and a few visible lines | Mercury lines plus added visible emission from metal halide additives |
| Output distribution | Broadband, relatively smooth; usable for action spectra | Concentrated in narrow bands; a large share of output sits where many catalysts do not absorb | Weighted toward the visible; better match to narrow-bandgap absorbers |
| Ignition and restart | Ignites from cold and can be restarted after shutdown | Requires a cool-down before it will restrike | Requires a cool-down before it will restrike |
| Behavior over life | Output falls and the arc position drifts, moving the spot | Output falls and the spectral balance shifts as the arc ages | Output falls and the color balance shifts as additives deplete |
| Dimming | Adjustable over a wide electrical range with modest spectral change | Limited; output and spectrum are coupled to arc current | Limited, with noticeable spectral shift at low current |
| Handling | Pressurized envelope, hot at the anode; handling discipline required | Contains mercury; disposal and breakage procedures required | Contains mercury; disposal and breakage procedures required |
| Best fit | General photocatalysis, action spectra, solar-related work | UV-driven reactions with a well-matched absorption band | Visible-light reactions needing high flux in a small area |
Read the table as constraints, not a scorecard. A mercury source wins only when the catalyst absorbs strongly at its lines; otherwise most of its output becomes heat in a small envelope. Metal halide is chosen when visible flux density matters more than spectral fidelity.
A source you cannot restart quickly constrains the experiment design, not just the convenience of the operator. Mercury and metal halide arcs must cool before they restrike, which inserts an unplanned dark interval every time a run is interrupted.
Three protocols feel it. Light-on and light-off cycles for background correction become slow and asymmetric. Chopped-light measurements need an external shutter, which adds its own timing error. A screening queue means leaving the lamp on between samples or accepting long waits.
Xenon short-arc sources restart without a cool-down, so one lamp can serve a screening queue and a stability run on the same day. If your protocol depends on switching, weigh a shutter against an electronically switched source; the trade-offs are in the LED and xenon article linked below.
Every family loses output long before it fails, so the useful question is how the loss shows up. Mercury and metal halide change spectral balance as the arc and its additives age, so a replacement lamp can deliver a different spectrum at the same setting. Xenon loses output more evenly but drifts in arc position, which moves the spot and silently changes the illuminated area.
Neither failure announces itself. Log irradiance at the sample plane and spot diameter at a fixed working distance, and replace on a measured decline rather than on a calendar. That record also protects you when a reviewer asks whether the light source was stable across a long campaign.
Four questions resolve most selections, and they should be asked in this order. Stop at the first one that gives a decisive answer.
If none of the four is decisive, the reaction is not light-limited in a way that the lamp family controls, and the cheaper decision is to standardize on the source your group already knows how to calibrate.
Xenon, because its output is continuous across the UV, visible and near infrared rather than concentrated in lines. It is not a solar simulator by itself: matching the air mass spectrum requires filtering and a calibrated reference, so treat a bare xenon source as solar-like rather than solar-equivalent.
Generally no. The ballast, ignition circuit, reflector geometry and cooling are matched to one lamp type, and a different arc size will not sit at the focal point the optics were designed around. Treat a lamp-family change as a source change, not a bulb change.
Yes. The ballast is matched to the lamp's electrical characteristics and to its ignition requirements, so a supply built for one family will not run the other correctly or safely. Confirm the supply and lamp are a matched pair before ordering spares.
