Quantum efficiency divides product electrons by incident photons; Faradaic efficiency divides product electrons by electrons pushed through the circuit. One photoanode run at 1.2 mA/cm2 under 100 mW/cm2 can score 91.5 % FE, 3.7 % AQY and 1.35 % STH at the same time, and all three are correct.
Both metrics answer “how much of the input became product”, but they measure different inputs. Quantum efficiency asks how well the catalyst converts available light, so the denominator is a photon count. Faradaic efficiency asks how selectively the circuit's charge becomes the product you want, so the denominator is a charge count. The numerator is the same in both cases: moles of product multiplied by the electrons each molecule requires.
That is why the two numbers can differ by a factor of thirty without either being wrong. A photoanode that absorbs weakly at 400 nm can convert 92 % of its collected charge into oxygen while using under 4 % of the photons striking it. Reporting only one of them hides whichever half of the problem is worse.
| Metric | Numerator | Denominator | Unit | Question it answers |
|---|---|---|---|---|
| AQY (apparent quantum yield) | n × mol product | Incident photons in a stated band | % | How much of the delivered light becomes product |
| IQE / APCE | n × mol product | Absorbed photons | % | How good the material is, with light absorption removed |
| IPCE | Collected electrons | Incident monochromatic photons | % | Wavelength-resolved response of a photoelectrode |
| FE (Faradaic) | n × mol product | Charge passed through the external circuit | % | Selectivity of charge use, only defined with a circuit |
| STH | Chemical energy stored | Incident solar power, 100 mW/cm2 | % | Bias-free solar fuel performance |
| Energy efficiency | ΔG or ΔH × product rate | Optical power in | % | Thermodynamic accounting across a full device |
The electron count n is fixed by the half-reaction and is the most common silent error: 2 for H2 and CO, 2 for formate, 4 for O2, 6 for methanol, 8 for methane, 12 for ethylene. Use 2 for O2 and the calculated efficiency doubles; the arithmetic still looks tidy, which is exactly why it survives to print.
Take a photoanode of 1 cm2 geometric area under 100 mW/cm2 of 365 nm light, held at 1.23 V versus RHE for two hours. Chronoamperometry integrates 8.64 C, which is 8.95 × 10-5 mol of electrons. GC finds 41 µmol of H2, and the incident photon flux over that area is 1.10 mmol/h.
| Metric | Calculation | Result |
|---|---|---|
| FE | 41 µmol H2 × 2 = 82 µmol e-, against 89.5 µmol e- passed | 91.5 % |
| AQY | 82 µmol e- in 2 h, against 2.20 mmol incident photons | 3.7 % |
| STH | 1.2 mA/cm2 × 1.23 V × 0.915 / 100 mW/cm2 | 1.35 % |
Now put the same chemistry in a slurry reactor with no circuit: 50 µmol/h of H2 under 150 mW/cm2 of 365 nm light over 19.6 cm2. The photon flux is 32.3 mmol/h, the product represents 100 µmol/h of electrons, and AQY is 0.31 %. There is no circuit, so there is no FE. Any “Faradaic efficiency” quoted for that experiment is a relabelled quantum yield, and a reviewer who reads the methods will say so.
One more discipline helps both numbers: report the measurement uncertainty. A calibrated radiometer contributes 5–7 %, GC calibration 2–3 %, and charge integration under 1 %. Quoting AQY as 3.71 % from inputs that carry 7 % uncertainty is a claim the data cannot support; 3.7 ± 0.3 % is.
Yes, routinely, and in the same experiment. FE compares product to charge, so a photoelectrode with good selectivity but poor light absorption reaches high FE at low AQY. The 91.5 % FE and 3.7 % AQY in the worked example come from one two-hour run.
Almost always an accounting error rather than a physical one. The usual causes are a GC calibration that has drifted, hydrogen crossing the membrane, oxygen leaking in and being counted as product, or a second product that was never quantified. Re-run the calibration and close the mass balance before reporting.
IPCE is a quantum efficiency measured one wavelength at a time, with the photoelectrode held at a fixed bias, and it counts collected electrons rather than product molecules. For a selective electrode the two agree within a few percent; where they diverge, the missing charge is going to a product you did not measure.
Both, with their conditions. Give AQY at a stated wavelength band and conversion level, and give FE with the charge passed and the cell configuration. A single efficiency figure without its denominator invites the reader to assume the most favourable interpretation.