Pressurise the assembled reactor to 1.2× working pressure with dry N2, equilibrate 30 min, then require a decay no faster than 0.2 kPa/h over 4–6 h for gas-phase work — about 0.56 mbar·L/s in a 1 L system. A 1 K temperature drift moves pressure 0.34 kPa and will mask the result.
Pressure decay is quoted in kPa/h or %/h, vacuum work uses mbar·L/s, and specification sheets use Pa·m³/s. Converting a decay into a gas flow needs the system volume, so the same 0.2 kPa/h admits ten times more gas in a 10 L cabinet than in a 1 L cell.
The bridge is the ideal gas law, n = ΔP·V/(R·T). For 1.0 L at 298 K, 0.2 kPa/h is 81 µmol/h of air ingress. That matters because a 500 mL water-splitting system typically produces 100–500 µmol/h of H2: a leak that passes the pressure test can still admit air at a rate comparable to the product, and the oxygen recombines on the catalyst and depresses the apparent rate.
In a sealed constant-volume vessel, pressure scales with absolute temperature — at 101.3 kPa and 298 K, 0.34 kPa per kelvin. A 1 K drift over the test equals 1.7 h of a 0.2 kPa/h budget, and a cell carried from a 22 °C bench into a 25 °C hood shows 1 kPa of apparent leak in minutes, then plateaus. Ambient barometry adds 1–3 kPa over a 12 h weather cycle, larger than the whole budget.
Three countermeasures make the test meaningful: jacket the cell at ±0.1 °C, log temperature beside pressure, and reference a differential gauge to a sealed dummy volume instead of using an absolute gauge.
| Method | What is measured | Sensitivity | Duration | Best used for |
|---|---|---|---|---|
| Immersion / bubble test | Bubbles from a pressurised joint | ~1×10−3 mbar·L/s | 1–2 min | Screening one suspect joint |
| Positive pressure decay | ΔP with N2 at 1.2× working | 0.1 kPa/h | 1–6 h | Assembled reactor, whole system |
| Vacuum decay | Rise below 1 kPa absolute | 0.01 kPa/h | 8–12 h | Glass vacuum manifold |
| Helium sniffer | He partial pressure at a probe | 1×10−6 mbar·L/s | 5–20 min | Locating a leak, not rating it |
| O2 or Ar tracer with GC | Peak growth in a blank run | ~100 ppm | 6 h | Qualifying a rig for trace work |
The ordering matters: the bubble test locates, pressure decay rates, and only the tracer test measures the quantity that perturbs a photocatalytic experiment. Running the first two alone is why rigs that hold pressure still show an unexplained O2 peak.
| Application | Metric | Acceptance | Test length |
|---|---|---|---|
| Preparative / screening | N2 decay at 1.2× working | ≤1.0 kPa/h | 1 h after 30 min equilibration |
| Gas-phase H2 and O2 | Decay plus O2 in a 6 h blank | ≤0.2 kPa/h and O2 ≤0.1 % | 6 h |
| Trace CO / CH4 at ppm | He leak rate, vacuum mode | ≤1×10−5 mbar·L/s | 15 min scan |
| Pressure-swing cell | Decay at 1.5× maximum working | ≤0.5 %/h of reading | 4 h, shield in place |
Tie the threshold to the smallest signal you must resolve: a leak is acceptable when its molar ingress stays below 1 % of the lowest rate you intend to report. For a 20 µmol/h target that means 0.05 kPa/h in 1 L, which no pressure gauge will confirm — go straight to the tracer test.
Ranked by how often they explain a failed test: septa (a 10 mm silicone septum permeates air continuously and is spent after about 50 punctures — fit PTFE-faced or butyl septa, or a valve port); compression fittings (1¼ turns past finger-tight for metal ferrules; one flat short is the commonest assembly error); ground-glass joints (grease-free joints seep at 10−1 to 1 mbar·L/s without a PTFE sleeve and clip); window O-rings (torque in a star pattern, replace after 20–30 thermal cycles); and PTFE valve tips, where a 0.1 mm particle score gives a permanent 10−2 mbar·L/s path. Test empty and dry: water vapour in a wet rig swamps the decay signal.
It depends on the smallest rate you must report. For preparative work, ≤1.0 kPa/h at 1.2× working pressure is adequate. For µmol-level H2 and O2, require ≤0.2 kPa/h plus O2 below 0.1 % after a 6 h blank, which corresponds to roughly 80 µmol/h of air ingress in 1 L.
30 min of equilibration plus at least 4 h of measurement when the limit is 0.2 kPa/h and the gauge resolves 0.1 kPa. Four hours accumulate 0.8 kPa, eight times the resolution. Shorter tests cannot separate a leak from thermal drift.
Either the vessel reached thermal equilibrium, or a soft seal deformed and closed. The first explains most curves that flatten within an hour. The second is a real defect: repeat at a different pressure and the plateau moves.
No. Test empty and dry, because slurry blocks valves and lamp heat adds a strong transient. Once the empty system passes, do one 6 h blank with the reaction medium in place, which validates the seal under real conditions.
