Vapor Discrimination With Single- and Multitransducer Arrays of Nanoparticle-Coated Chemiresistors and Resonators

This study explores whether arrays of vapor sensors assembled from two different types of transducers provide greater response diversity than arrays of a single transducer (ST) type. Calibrated sensitivities to five vapors of four chemiresistors (CRs) and four thickness-shear mode resonators (TSMR) coated with matching interface films of four different thiolate-monolayer-protected gold nanoparticles (MPNs) were considered. Test vapors consisted of toluene, nitromethane, 2-butanone,
$n$
-propanol, and
$n$
-octane. The pooled set of 40 vapor-sensor sensitivities was analyzed using principal components regression models in conjunction with Monte Carlo simulations to evaluate the classification performance with different levels of error superimposed on the sensor responses. Recognition rates (RR) were estimated for the individual vapors and their binary mixtures with virtual arrays consisting of all possible combinations of MPNs and transducer types. The best overall performance was obtained with a multitransducer (MT) array of
$n=4$
sensors, which provided average RRs of 99.7% for individual vapor discrimination and 74.7% for discrimination of the 10 binary mixtures from their components, both with 5% superimposed error. MT-array RR values did not improve for
$n>4$
. The corresponding average RRs for the all-CR and all-TSMR 4-sensor ST arrays were both
${\sim}97\%$
for the individual vapors and both
${\sim}69\%$
for the binary mixtures, respectively. Results demonstrate that MT arrays can provide modestly greater diversity than ST arrays of similar dimension.