If three hours of irradiation give no peak above 5 ppm on a GC-TCD, suspect hardware before the catalyst. In a 500 mL headspace 1 µmol of H2 reads about 49 ppm, so a true blank means light, seal, sampling loop, carrier gas or detector failed — not that the photocatalyst is inert.
A blank is only meaningful against a detection limit. In a sealed 500 mL system at 25 °C and 1 atm the headspace holds 20.4 mmol of gas, so 1 µmol of hydrogen is 49 ppm, 10 µmol is 490 ppm. A TCD with a 1 mL loop resolves roughly 5–10 ppm, which sets the floor at about 0.1–0.2 µmol — an amount that 20 mg of an ordinary TiO2 sample produces in two to five minutes under 100 mW/cm2.
That arithmetic is the whole diagnostic. If a proven catalyst at a proven loading gives nothing after an hour, the experiment is producing less than one thousandth of its expected rate, which is a plumbing or optics fault far more often than a chemistry fault. Run a positive control — the same catalyst that worked last week, or a standard that evolves hydrogen on demand — before touching the sample.
| Stage | What fails | Fast test | Pass criterion |
|---|---|---|---|
| Light delivery | Shutter closed, lamp in standby, fibre disconnected, filter left in the path | Power meter at the inner window, not at the lamp exit | Within 10 % of the recorded irradiance in mW/cm2 |
| Reactor seal | Septum coring, greased joint, O-ring pinched by the clamp | Pressure-hold test at 1.2× working pressure, 30 min | Drift under 0.2 kPa/h after temperature equilibration |
| Gas path | Valve left closed, three-way port in the vent position, tube kinked | Flow-through check with a bubble meter at the outlet | Set flow ±5 % actually arriving |
| Sampling loop | Loop not purged, syringe dead volume, sample taken downstream of a trap | Two consecutive injections of the same standard | Peak areas agree within 3 % |
| Carrier and column | Cylinder empty, regulator closed, column disconnected at the nut | Inlet pressure gauge with the septum plugged | Pressure holds; no drift over 5 min |
| Detector | Filament off, attenuation set too high, calibration expired | Inject a 100 ppm H2 standard | Peak within ±5 % of its certified area |
First, the light never reaches the liquid. A power meter reading taken at the lamp port says nothing about the reactor window: a closed shutter, an unplugged liquid-light guide, a safety interlock that dropped out, or a 420 nm long-pass filter left from the previous experiment will each produce a perfectly clean blank. Measure at the inner window surface, with the reactor filled, every time.
Second, the septum. Repeated sampling with a 1 mm needle cores the silicone and leaves a channel that leaks at roughly the rate gas is produced. A reactor that holds pressure cold may still leak warm, because the septum softens at 40–60 °C. Replace septa every 20–30 injections and use a PTFE-faced septum for long runs.
Third, the sample never arrives. A 1 mL loop needs three to five loop volumes of purge, and a gas-tight syringe needs flushing twice before it is representative. Both failures look identical to zero production and both are caught by the duplicate-standard injection in the table.
Keep the sequence fixed. Improvised troubleshooting changes two variables at once and destroys the evidence you need to explain the result in a paper.
Genuine zero production exists. A missing sacrificial agent, a pH far from the point of zero charge, a catalyst that was calcined in the wrong atmosphere, or a suspension left overnight that has adsorbed all the reactant can each suppress hydrogen for hours. An induction period of 30–90 min is normal for some samples and is often mistaken for failure when the first sample is taken at 15 min.
The diagnostic that separates chemistry from hardware is the response to a deliberate change: doubling irradiance must increase the rate, and adding fresh catalyst must restore it. A signal that is exactly zero under both is not chemistry — zero is what a closed valve looks like.
With 20 mg of a working TiO2 at 100 mW/cm2 in 100 mL of 10 vol% methanol, expect roughly 2–10 µmol/h, which is 100–500 ppm in a 500 mL headspace after one hour and far above any TCD detection limit.
Because the amounts are small. 10 µmol of hydrogen in 500 mL changes total pressure by about 0.05 kPa, below the resolution of most gauges. Treat the chromatogram, not the gauge, as the primary measurement in lab-scale runs.
Yes, if air ingress is comparable to production. Air carries oxygen, which recombines with hydrogen on the catalyst, and nitrogen dilutes the sample. A 0.5 kPa/h leak is about 10 µmol/h of air, the same order as the hydrogen output of a small experiment.
Septa every 20–30 injections or every two weeks, whichever comes first; silicone tubing every six months, since it is permeable to hydrogen and hardens with heat. Log both dates next to the instrument.
