2.6. Data Analysis

For hydrogen sulfide, the correction method described by Feilberg et al. [16] was applied. This was necessary because the proton transfer reaction that takes place in the drift tube of the PTR-MS is only energetically possible for VOCs with a proton affinity greater than that of water (691 kJ mol⁻¹). Because hydrogen sulfide only slightly exceeds this value (705 kJ mol⁻¹), a humidity-dependent backward reaction of protonated hydrogen sulfide becomes significant. A power function was found as the best fit for the conditions applied in this study (R² > 0.9822).

For the carboxylic acid, 4-methylphenol, and trimethylamine in particular, a delay in the signal rise and fall-off time was observed when these were measured directly with the PTR-MS system. A delay of adsorptive compounds in the PTR-MS system has previously been reported [19,20,21,22]. To eliminate the effect on the odorant breakthrough curves, which was caused by the PTR-MS system, all compounds were measured directly with quadruple determination by PTR-MS. For these measurements, an automatic three-way valve, which was connected to—and controlled by—the PTR-MS instrument, was added for immediate compound dosage to estimate the true response time of the PTR-MS signal. For each compound, five different dilution steps were included, and the compound concentrations were kept at the same level as in the olfactometer experiments.

The data obtained for the signal response and tail-off could be modelled with an adequate fit by means of a non-linear least squares regression with a sigmoid curve and an exponential decay function of the type given in Equations (1) and (2), respectively:

where C_s is the concentration signal measured by PTR-MS, C_exp is the inlet odorant concentration, t is the time and B is a fitted parameter characterizing the slope of the curve and, thus, the time before the signal plateau at C_exp or the background level is reached. Table 1 lists the B-parameters fitted for the data and the average model fit.

In the following analysis, these models were used to estimate the expected PTR-MS signal equivalent to a certain time and inlet concentration.