Most xenon arc instability is electrical, not optical. With a 300 W lamp, log output for 60 seconds after warm-up: peak-to-peak fluctuation above 2–3%, a visible 100 Hz flicker, or a spot that wanders more than 1 mm all point to supply ripple, loose lamp leads or a failing ignition module. Change the bulb only after those are ruled out.
Three symptoms get called “flicker” and they have different causes. Separate them before touching anything, because the fix for one worsens the others.
| Symptom | Most likely cause | First measurement |
|---|---|---|
| Steady 100 Hz flicker, visible by eye | DC supply ripple, aged filter capacitor | Ripple on the lamp supply, accept under 1% |
| Random dropouts every few minutes | Loose lamp lead, oxidised contact clamp | Clamp torque, contact temperature |
| Spot wanders slowly over 10–20 minutes | Convection, fan airflow, arc wander in an aged bulb | Spot position on target paper, bulb hours |
| Hard ignition, repeated restrikes | Ignition module, worn electrodes, low mains voltage | Mains voltage at the socket under load |
| Output falls 20–30% within one session | Electrode erosion, reflector haze, filter heating | Irradiance in mW/cm² against baseline |
| Bright band or dark notch inside the spot | Arc off the reflector axis, tilted reflector | Spot photograph at the window plane |
Xenon short-arc lamps run at low voltage and high current, so the supply and its wiring dominate arc stability. Work through these in order:
If the electrical path is clean, the arc itself is moving. Short-arc lamps depend on forced-air cooling: a blocked filter, a slow fan, or an ambient temperature above 35 °C destabilises the discharge. Convection matters too — a lamp specified for vertical operation that runs horizontally will wander, and the effect grows with power.
Electrode erosion is the slow version of the same problem. By 500–1000 hours the arc attachment point has moved, output has typically dropped 20–30%, and the spot no longer sits where it did when the lamp was new. That is not a fault; it is the end of service life, and no amount of re-focusing recovers the missing photons.
Photocatalytic rates scale with absorbed photons, so irradiance noise passes straight into the rate. A 3% peak-to-peak ripple becomes roughly 3% scatter in hydrogen or CO production, and the same 3% appears in apparent quantum yield, where the photon count is the denominator. Resolving a 10% activity difference at that noise floor needs triplicates: averaging three 1-hour runs cuts random scatter by about a factor of 1.7, but it cannot remove a systematic 100 Hz modulation correlated with your sampling interval.
The practical rule: stabilise the arc before you optimise the catalyst. An experiment that cannot repeat its own baseline within 2% will not detect a real 8% improvement.
Because the DC supply is not fully smoothing the rectified mains. A 50 Hz supply rectified in full wave produces a 100 Hz component, and a degraded filter capacitor lets it reach the arc. Measure ripple at the supply output; below 1% is acceptable, above 2–3% explains the flicker.
Yes, but usually as slow wander or hard ignition rather than fast flicker. A bulb near the end of its 500–1000 hour life has an eroded cathode, so the arc attachment point moves and the spot drifts. Fast periodic flicker is almost always electrical.
Up to about 1% peak-to-peak is generally tolerable. Between 2% and 3%, document it as a limitation and run triplicates. Above 5%, apparent quantum yield numbers are not defensible, because the photon count in the denominator is uncertain by the same amount.
Track hours and irradiance together, and replace at rated life or when reactor-plane irradiance has fallen 20% below the installation value — whichever comes first. “Still lit” is not a criterion; a lamp at 70% output silently changes every rate you measure.
