Calibrate with a 5-point curve spanning 0.1–1.0× your highest expected peak, accept R² ≥ 0.999 and a response-factor RSD ≤ 2 %, then inject a mid-point check standard every 10 samples with a ±3 % tolerance. A 1 mL loop at 100 ppm holds 4.1 nmol, so build the curve in moles, not in millimetres of peak height.
Peak area is proportional to the moles passing the detector, and that number depends on loop volume, pressure and temperature. Fix the arithmetic once: at 101.3 kPa and 298 K, one millilitre of ideal gas is 40.9 µmol, so a 1 mL loop filled with 100 ppm H2 delivers 4.09 nmol and a 1 % standard delivers 409 nmol. Convert every level to nanomoles before fitting, and the response factor transfers between loop sizes and instruments.
The conversion also exposes the largest routine error in this analysis: reactor pressure. A headspace sampled at 120 kPa holds 19 % more moles per millilitre than the same volume at 101 kPa, and a calibration run at atmospheric pressure silently under-reports the rate by that amount. Either sample through a back-pressure regulator at the calibration pressure, or correct each injection with n = PV/RT using the pressure logged at the sampling valve.
A curve is only meaningful inside the detector's linear range, and photocatalytic work sits at the difficult end: hydrogen and CO from a fresh catalyst are often at tens of ppm, near the noise floor of a thermal conductivity detector.
| Detector | Carrier | Species | LOD (1 mL loop) | Linear range |
|---|---|---|---|---|
| TCD | Ar | H2, He, O2, N2, CO2 | 10–50 ppm | 104 |
| TCD | He | CO2, N2, O2 | 50–200 ppm | 104 |
| FID | N2 | CH4, light hydrocarbons | 0.1–1 ppm | 106–107 |
| FID + methanizer | N2 | CO and CO2 as CH4 | 0.5–2 ppm | 105 |
| PDHID | He | H2, O2, CO | 20–100 ppb | 105 |
| Quadrupole MS | any | all, with overlap | 1–10 ppm | 104–105 |
Use the table backwards to size the experiment. A 1 µmol/h hydrogen rate into a sealed 300 mL headspace adds 3.33 µmol/L per hour, which is 81 ppm/h at 25 °C. A TCD at 20 ppm LOD resolves that in about 15 min; an FID is irrelevant because it cannot see H2 at all. At 0.05 µmol/h the hourly increment is 4 ppm, and a TCD curve will never be trustworthy — switch to PDHID or irradiate longer.
| Check | Acceptance | Frequency | If it fails |
|---|---|---|---|
| Curve linearity, 5 levels | R² ≥ 0.999, residuals ≤ 2 % | After column, septum or filament change; monthly | Re-fit without the top level, or dilute |
| Response-factor RSD | ≤ 2 % across levels | With every full curve | Check loop fill time and pressure |
| Mid-point check standard | Within ±3 % of curve | Every 10 injections, start and end of day | Re-run the curve before reporting |
| Retention time drift | ≤ 1 % per sequence | Continuous | Leak check, then re-equilibrate |
| Blank, carrier only | Below LOD, no ghost peaks | Daily | Bake out loop and column, new septum |
| Peak asymmetry | 0.9–1.2 at 10 % height | Weekly | Repack or trim the inlet liner |
Three habits prevent most bad curves. Flush the loop with at least five loop volumes before injection — 5 mL at 20 mL/min is 15 s, and cutting it short gives low areas that look like detector drift. Keep the same carrier for calibration and analysis: a curve built on Ar does not transfer to He, because TCD response depends on the conductivity difference between carrier and sample. And remove water before the column, not after — a Nafion dryer is fine for H2, O2, CH4 and CO, but it adsorbs part of the CO2 you are quantifying, so validate the loss with a wet and dry standard pair.
A full 5-point curve monthly and after any hardware change; a mid-point check standard every 10 injections, or at the start and end of each day. If the check drifts past 3 %, re-run the curve rather than applying a correction factor.
Only for one species at fixed flow, column temperature and injection volume, and never where peaks broaden. Height responds to flow and temperature in ways area does not, so a height-based curve can report 5–10 % change when the room warms up.
Usually a small inlet leak, which lowers loop pressure and the moles injected, or a changed split ratio if the septum nut was over-tightened. Re-run the mid-point standard first; if it recovers after 10 min of equilibration, the cause was thermal, not chemical.
Yes for a TCD, where response factors differ by more than an order of magnitude between H2, O2, N2 and CO2. For FID with a methanizer, one methane curve covers CO and CO2 after reduction, but the methanizer conversion must be verified with a CO2 standard instead of assumed to be 100 %.
