Overall water splitting must return H2:O2 of exactly 2.00 by mole. At a measured 100 µmol/h of H2, expect 50 µmol/h of O2 and accept 47.5–52.5 µmol/h, that is 2.00 ± 5 %. A ratio below 1.90 almost always means oxygen was consumed or never reached the detector.
Two water molecules give two hydrogen molecules and one oxygen molecule, so four electrons produce one O2. That factor of four converts an oxygen rate straight into an electron flux that can be compared with the hydrogen number and with the photon count. A reactor making 100 µmol/h of H2 turns over 200 µmol/h of electrons and should release 50 µmol/h of O2, or 1.6 mL/h at 25 °C and 1 atm.
Both products are measured independently: an error in light, catalyst mass or sample volume moves both numbers the same way and the ratio survives, while a gas-path fault moves only one and breaks it.
Deficits are usually chemical, not analytical. Work through these five before suspecting the detector.
| Oxygen sink | How it appears | Diagnostic | Mitigation |
|---|---|---|---|
| Reverse reaction H2 + ½O2 → H2O on the cocatalyst | Both products low, ratio drifting downward | Compare 15 min and 60 min rates; the drop scales with Pt loading | Lower the Pt loading, or use a CrOx or carbonate shell |
| O2 reduction by photogenerated electrons | O2 deficit with matching H2 excess | Run once without purging; the deficit grows with initial O2 | Purge below 1 µmol before illumination |
| Peroxide route (2H2O → H2O2 + H2) | Ratio near 1:0, oxygen essentially absent | Ti(IV) sulfate colour test or a peroxide strip | Report H2O2 as a product; the electron balance still closes |
| Oxygen still dissolved in the liquid | Ratio rises slowly over hours | Sample the liquid phase after the run | Extend the run, or strip with a circulating loop |
| Air ingress through a fitting | O2 excess plus nitrogen in the headspace | N2 peak after the run; pressure-hold test at 50 kPa | Replace septa; leak-test the loop, not just the vessel |
Chemistry removes oxygen and produces a deficit; plumbing adds it and produces an excess. The sign of the deviation tells you which half of the system to inspect.
The initial dissolved pool. Water equilibrated with air at 25 °C holds about 8.3 mg/L of dissolved oxygen, which is 0.26 mmol/L. A reactor charged with 100 mL of untreated water therefore starts with roughly 26 µmol of O2 already in the liquid — half an hour of production at a 50 µmol/h rate. Sparging with argon at 100 mL/min for 20 minutes, about twenty liquid volumes, takes the residual below 1 µmol.
The cost of a small leak. One millilitre of air holds 40.9 µmol of gas at ambient conditions, 21 % of it oxygen, so a single millilitre of ingress adds 8.6 µmol of O2 — equivalent to 17 µmol of hydrogen, or 17 % of an hour's production in the example above. A leak of 1 mL/min would swamp the measurement, which is why a pressure-hold test belongs in the method section.
The headspace concentration. Fifty micromoles of O2 in a 200 mL headspace is 250 ppmv, where the equilibrium dissolved concentration is only 0.3 µmol/L; once the initial pool is stripped, the gas phase carries essentially all the oxygen. CO2, roughly fifty times more soluble, behaves differently.
Thermal conductivity detection is the default, but the carrier gas decides whether it works at all. On a 5A molecular sieve at 40–50 °C, oxygen elutes before nitrogen with only a few tenths of a minute between them, and a column that has adsorbed water loses the separation entirely. Use argon as carrier for the oxygen channel: the conductivity difference is large and the peak is positive and well shaped. Helium gives a weak, sometimes inverted oxygen peak, and nitrogen carrier makes the measurement impossible because the carrier is the interference.
Because the two gases want different carriers, most laboratories run two channels: argon on a sieve column for O2 and N2, nitrogen or argon on a second column for H2. A 1 mL loop at 1 atm gives a TCD detection limit near 100 ppmv, or 4 nmol per injection — adequate for 50 µmol/h sampled every 15 minutes, not for the first ten minutes of a run.
Optical quenching sensors offer 0.01 % resolution and a 10–30 s response without a carrier gas, but they drift and depend on pressure and humidity, so use them as a cross-check rather than the sole measurement.
No. With methanol or triethanolamine present, holes are consumed by the agent rather than by water, so oxygen evolution stops or becomes a side reaction and the ratio is meaningless. The check applies only to overall water splitting without a sacrificial reagent.
Almost always air ingress, since a leak adds oxygen without adding hydrogen. Confirm by measuring nitrogen in the same injection: air is 78 % nitrogen, so an O2 excess accompanied by N2 is contamination. If nitrogen is absent, consider peroxide decomposition or oxidation of an organic impurity.
You can calculate it, but then you have tested nothing. The oxygen measurement is an independent check on the hydrogen number, the light measurement and the gas path; replacing it with arithmetic removes the only internal control in the experiment.
Long enough for the dissolved pool to equilibrate and the headspace to reach 200 ppmv of O2, typically 60–120 min at 50 µmol/h in a 200 mL headspace. Shorter runs measure the initial dissolved pool and dead volume, not the reaction.
