A comprehensive analysis of vapor recognition as afunction of the number of sensors in a vapor-sensor arrayis presented. Responses to 16 organic vapors collectedfrom six polymer-coated surface acoustic wave (SAW)sensors were used in Monte Carlo simulations coupledwith pattern recognition analyses to derive statisticalestimates of vapor recognition rates as a function of thenumber of sensors in the array (
6), the polymer sensorcoatings employed, and the number and concentration ofvapors being analyzed. Results indicate that as few as twosensors can recognize individual vapors from a set of 16possibilities with <6% average recognition error, as longas the vapor concentrations are >5 × LOD for the array.At lower concentrations, a minimum of three sensors isrequired, but arrays of 3-6 sensors provide comparableresults. Analyses also revealed that individual-vaporrecognition hinges more on the similarity of the vaporresponse patterns than on the total number of possiblevapors considered. Vapor mixtures were also analyzed forspecific 2-, 3-, 4-, 5-, and 6-vapor subsets where allpossible combinations of vapors within each subset wereconsidered simultaneously. Excellent recognition rateswere obtainable for mixtures of up to four vapors usingthe same number of sensors as vapors in the subset.Lower recognition rates were generally observed formixtures that included structurally homologous vapors.Acceptable recognition rates could not be obtained for the5- and 6-vapor subsets examined, due, apparently, to thelarge number of vapor combinations considered (i.e., 31and 63, respectively). Importantly, increasing the numberof sensors in the array did not improve performancesignificantly for any of the mixture analyses, suggestingthat for SAW sensors and other sensors whose responsesrely on equilibrium vapor-polymer partitioning, largearrays are not necessary for accurate vapor recognitionand quantification.