Use GC-TCD for the number you report and an H2 sensor for the time axis. A 1 mL loop at 100 ppm carries 4.1 nmol, and a TCD at a 10 ppm limit of detection resolves 0.41 nmol per injection — enough for 1 µmol/h accumulating in a 300 mL headspace within 15 min. Sensors update every 30–60 s but drift 2–5 % per month.
A gas chromatograph answers a different question from a sensor. The GC reports how many moles passed through a fixed loop at 14:20; the sensor reports a partial pressure right now. Convert one into the other, or the two will never agree. The figures below are typical for instruments used on photocatalytic rigs — check your own datasheet, because detection limits and ranges move by an order of magnitude between models.
| Method | What it actually measures | LOD (1 mL loop) | Update interval | Main interference |
|---|---|---|---|---|
| GC-TCD, Ar carrier | Moles in a fixed volume at the sampling valve | 10–50 ppm | 3–15 min | Responds to every gas; a He carrier inverts the H2 peak |
| GC-PDHID | The same, with a discharge ionisation detector | 20–100 ppb | 3–15 min | Halocarbons and co-eluting permanent gases |
| Electrochemical sensor, ppm range | Partial pressure at a diffusion barrier | 1–10 ppm | 30–60 s (T90) | CO and sulfur compounds poison the electrode; humidity shifts sensitivity |
| Pd thin-film or MEMS sensor | Resistance change from hydrogen uptake | 10–100 ppm | 10–30 s | Saturates above roughly 1 %; CO and H2S degrade the film |
| Quadrupole MS | Ion current at m/z 2 | 1–10 ppm after calibration | 1–10 s | Mass 2 cannot separate H2 from deuterium fragments, which breaks 2H tracing |
The distinction matters for reviewers. A sensor trace is evidence that hydrogen appeared at a certain time; only a calibrated injection is evidence of how much. Papers that report a rate from a sensor alone need an independent standard, which is why the cheapest defensible setup is a sensor plus a shared GC rather than either one on its own.
One micromole per hour into a sealed 300 mL headspace adds 3.33 µmol/L per hour, which is 82 ppm/h at 25 °C and 101 kPa (molar volume 24.47 L/mol). Run the same arithmetic before choosing hardware, because a detector that cannot resolve the hourly increment cannot report a rate no matter how well it is calibrated.
| H2 rate | Increment in 300 mL | Method that can actually resolve it |
|---|---|---|
| 0.05 µmol/h | 4.1 ppm/h | PDHID, or extend the run to 6–8 h so the accumulated signal clears the TCD floor |
| 1 µmol/h | 82 ppm/h | TCD with a 15 min sampling interval |
| 10 µmol/h | 815 ppm/h | TCD comfortable; a ppm-range sensor enters its span |
| 100 µmol/h | 8,150 ppm/h (0.8 %) | Sensor for the time axis, GC for the absolute value, checked hourly |
For screening, yes; for publication, no. A sensor has no traceable link to moles, so it cannot establish a rate without an independent calibration. Use it to find the induction period and the steady state, then quantify those points by GC.
Usually water vapour and loop pressure. A headspace saturated at 40 °C dilutes hydrogen by 7.3 %, and a loop filled at 101 kPa after a reactor run at 120 kPa loses another 16 %. Check those two before suspecting the catalyst.
In the loop, downstream of the reactor and upstream of any condenser. In the headspace it reads a local concentration that depends on where you mounted it, and stirring rate alone can move the reading by 10–20 %.
Run the same experiment in the dark and with the catalyst omitted. A true photocatalytic signal disappears in both controls. For water splitting, confirm the H2:O2 ratio is close to 2:1 — a large excess of hydrogen usually means a reduced sacrificial agent or a leak drawing air in, which suppresses the oxygen reading.
