Vapor recognition with small arrays of polymer-coated microsensors. A comprehensive analysis

A comprehensive analysis of vapor recognition as a function of the number o f sensors in a vapor-sensor array is presented. Responses to 16 organic vap ors collected from Six polymer-coated surface acoustic wave (SAW) sensors w ere used in Monte Carlo simulations coupled with pattern recognition analys es to derive statistical estimates of vapor recognition rates as a function of the number of sensors in the array (less than or equal to 6), the polym er sensor coatings employed, and the number and concentration of vapors bei ng analyzed, Results indicate that as few as two sensors can recognize indi vidual vapors from a set of 16 possibilities with <6% average recognition e rror, as long as the vapor concentrations are >5 x LOD for the array. At lo wer concentrations, a minimum of three sensors is required, but arrays of 3 -6 sensors provide comparable results. Analyses also revealed that individu al-vapor recognition hinges more on the similarity of the vapor response pa tterns than on the total number of possible vapors considered. Vapor mixtur es were also analyzed for specific 2-, 3-, 4-, 5-, and B-vapor subsets wher e all possible combinations of vapors within each subset were considered si multaneously. Excellent recognition rates were obtainable for mixtures of u p to four vapors using the same number of sensors as vapors in the subset. Lower recognition rates were generally observed for mixtures: that included structurally homologous vapors. Acceptable recognition rates could not be obtained for the 5- and 6-vapor subsets examined, due, apparently, to the l arge number of vapor combinations considered (i.e., 31 and 63, respectively ). importantly, increasing the number of sensors tin the array did not impr ove performance significantly for any of the mixture analyses, suggesting t hat for SAW sensors and other sensors whose responses rely on-equilibrium v apor-polymer partitioning, large arrays are hot necessary for accurate vapo r recognition and quantification.