Photoelectrochemical (PEC) work fails on three source properties before it fails on anything else: output that drifts during a scan, a beam smaller than the electrode, and a chopping response slower than the measurement. Each produces a curve that looks like photophysics. Verify all three over the length of a complete scan before recording a single J–V curve.
A slow monotonic drift in the source is indistinguishable from a trend in the electrode, and PEC scans are slow enough to collect it. The signature is a repeat scan on the same electrode that starts and ends at different photocurrents, with the difference tracking time since ignition rather than applied potential.
Three effects act on the same timescale: the arc or LED output still settling after start-up, the cooling circuit coming to equilibrium, and the cell itself warming. Only the first is a lamp problem, but all three appear in the data as the same slope. Separate them by running a photodiode in place of the electrode — any drift that remains belongs to the source or the optics, not to the sample.
Current density is only meaningful when the illuminated area is known, and an underfilled electrode makes it unknowable. If the beam covers the center of a larger sample, the current is generated in a region you did not measure; dividing by the geometric area of the electrode then understates the real current density, and moving the cell a few millimeters changes the result.
The opposite error is easier to miss. A beam wider than the electrode spills onto the contact, the edge and the electrolyte, adding a photocurrent that has nothing to do with the material. Either way the fix is the same: define the illuminated area with a mask, measure the beam at the electrode plane, and record both numbers in the method.
Beam shape matters as much as size. A gaussian profile concentrates flux in the center, so the reported current density depends on where the electrode sits in the beam; a flat-top profile does not. Re-measure the profile whenever the working distance changes, because the spot size moves with it.
A chopped measurement is valid only when the chopping period is long enough for the photoresponse to settle and short enough that the electrode does not drift between the on and off states. Interface capacitance and series resistance set a time constant that belongs to the cell, not to the shutter, and a shutter faster than the cell produces a transient that is easy to integrate wrongly.
Two checks settle it. Double the chopping period; if the on-state current changes, the original period was too short. Then look at the raw photocurrent on an oscilloscope instead of the processed value — the shape of the decay tells you whether you are reading a steady state or a capacitance.
The table maps each requirement to the artefact it prevents, so an anomaly in the data can be traced back to a specific property of the source.
| Requirement | What satisfies it | Artefact if violated | Check before the scan |
|---|---|---|---|
| Output stability | Warm-up complete, cooling at equilibrium, constant lamp current | Repeat scan drifts; apparent photocorrosion | Photodiode run for the full scan duration |
| Beam coverage | Beam at least as large as the masked electrode | Current density computed on the wrong area | Beam profile at the electrode plane |
| Beam uniformity | Flat-top profile across the electrode | Result depends on cell position | Translate the cell and re-measure |
| Collimation | Low divergence over the working distance | Flux falls as the cell is moved | Flux measured at two distances |
| Chopping response | Shutter faster than the cell time constant | Distorted on and off transients | Double the period and compare |
| Spectral definition | Filter or monochromator with a known bandpass | Incident photon-to-current efficiency (IPCE) inflated by stray light | Measure with and without a long-pass filter |
Read the table from the artefact column backwards. Most reported PEC anomalies — a photocurrent that decays over minutes, a current density that changes when the cell is remounted, an IPCE that implies absorption the film does not have — resolve to one of these rows rather than to the material.
The source only has to be stable, uniform over the electrode and fast enough to chop; the spectrum matters only when a specific reference is claimed. A 300 W class xenon source with 150–300 W adjustment and 50 W of total optical power covers most white-light photocurrent work, and a 400 W source with 100 W of total optical power leaves more margin once the beam is expanded over a large electrode or passed through a monochromator.
For IPCE the constraint changes. The monochromatic beam is weak, so stray light and beam stability dominate the uncertainty, and the illuminated area has to be defined as carefully as in a white-light scan. A source with a stable arc and a well-defined output port is worth more here than one with extra power.
Two habits make the source auditable. Record the flux at the electrode plane with a calibrated meter before every session, and note the lamp hours and the working distance in the same line of the log. When a photocurrent later fails to reproduce, those three values are the first thing to check.
Yes for white-light photocurrent, provided the spectrum and the flux at the electrode plane are reported. Unfiltered output includes ultraviolet that the electrolyte may absorb before it reaches the semiconductor, so the flux measured at the window is not the flux the electrode receives.
Either the sample is changing or the source is. Repeat the hold with a photodiode in place of the electrode: a flat diode response means the decay belongs to the electrode, while a matching decay means it belongs to the lamp, the shutter or the cooling.
At least as large as the masked illuminated area, and ideally with a flat profile across it. A spot smaller than the electrode leaves part of the sample dark, and a spot much larger wastes flux on the contact and the cell body.
