Cryptand/metal ion coated piezoelectric quartz crystal sensors with artificial back propagation neural network analysis for nitrogen dioxide and carbon monoxide

Abstract

Titanium(IV)/cryptand22 and Zinc(II)/cryptand22-coated piezoelectric (PZ) quartz crystal detection systems with home-made computer interfaces for data acquisition and processing were developed and successfully applied for the detection of nitrogen dioxide (NO 2) and carbon monoxide (CO), respectively. Among the 20 metal ion/cryptand22 complexes, e.g. Mn 2+/cryptand22, Ti 4+/cryptand22, Zn 2+/cryptand22, Al 3+/cryptand22, Cr 3+/cryptand22, Cu 2+/cryptand22 and Ru 3+/cryptand22, Ti 4+/cryptand22 and Zn 2+/cryptand22 were found to be the best adsorbents to adsorb the NO 2 and CO in the air respectively. The Ti 4+/cryptand22 coated PZ crystal NO 2 sensor exhibited a good linear response to the concentration of NO 2 with excellent sensitivity (9.10 Hz/ppm), detection limit (0.16 ppm NO 2) and reproducibility with relative standard deviation (RSD) of 2.09%. The Zn 2+/cryptand22 coated CO sensor also showed a good sensitivity of approximately 1.6 Hz/ppm to CO with a detection limit of 0.91 ppm and reproducibility with RSD of 2.15%. The coating load effects on the frequency responses of these cryptand22/metal ion coated PZ crystal NO 2 and CO sensors were also investigated. The frequency responses of the PZ crystal sensors increased with the coating load. The cryptand22/metal ion coated PZ crystal sensors were also successfully applied as gas chromatographic (GC) detectors to detect NO 2, CO and some organic compounds, e.g. propyl amine and propanoic acid. The Ti 4+/cryptand22 coated PZ crystal GC detector compared well with the thermal conductivity detector (TCD) for NO 2 vapor which TCD did not response. An artificial back propagation neural (BPN) network with a two-channel PZ crystal flow sensor was prepared and applied to recognize and distinguish each component (NO 2 and CO) in a mixed NO 2/CO sample and it showed quite good distinction of NO 2 and CO in the mixed gas sample. The concentrations of NO 2 and CO were also computed by multivariate linear regression analysis (MLR).