Nitrogen Oxides Analysis: Application of Dynamic Dilution to Instrument Calibration

Francesca Pennecchi, Istituto Nazionale di Ricerca Metrologica, Torino, Italy
Enrica Pessana, Istituto Nazionale Di Ricerca Metrologica, Torino, Italy
Francesca Rolle, Istituto Nazionale Di Ricerca Metrologica, Torino, Italy
Michela Sega, Istituto Nazionale Di Ricerca Metrologica, Torino, Italy
Pier Giorgio Spazzini, Istituto Nazionale Di Ricerca Metrologica, Torino, Italy

The European Directives prescribe the monitoring of nitrogen oxides (NO_X) by means of chemiluminescence as the reference method, which requires the use of proper reference gas mixtures for instrument calibration. This work presents the activity carried out at the Istituto Nazionale di Ricerca Metrologica (INRiM) concerning the use of dynamic dilution as a primary method, alternative to the gravimetric method, for the preparation of NO_X mixtures and their application to the calibration of analytical instruments.

Dynamic dilution allows preparing ready-to-use gas mixtures at low mass fractions by diluting a standard mixture (parent mixture) with a proper diluent gas, thus avoiding instability problems related to diluted mixtures of reactive gases contained in high pressure cylinders. At INRiM, two Mass Flow Controllers, previously calibrated by using primary flow standards, were employed for this purpose. The present work, starting from the classic model equation of the dynamic dilution, addresses the uncertainty budget for the molar fraction of the obtained mixtures taking into account contributions arising from 1) the flow of the parent mixture and of the dilution gas, 2) the molar fraction of the analyte gas in the parent mixture and 3) the impurities of the analyte gas in the diluent gas. Covariances between the input quantities to the model were also evaluated and taken into account.

The obtained NO_X reference mixtures were used to establish a response curve for the chemiluminescence analyser in the range of mass fractions suitable for the environmental monitoring. Weighted Total Least-Squares regression was chosen for fitting the curve to the experimental data: this technique is able to deal with uncertainties on (and possible with covariances among) the values of both the dependent and independent variables. We present some calibration cases and the corresponding validation study.