Detecting Chemical Hazards with Temperature-Programmed Microsensors: Overcoming Complex Analytical Problems with Multidimensional Databases

Abstract

Complex analytical problems, such as detecting trace quantities of hazardous chemicals in challenging environments, require solutions that most effectively extract relevant information about a sample's composition. This review presents a chemiresistive microarray–based approach to identifying targets that combines temperature-programmed elements capable of rapidly generating analytically rich data sets with statistical pattern recognition algorithms for extracting multivariate chemical fingerprints. We describe the chemical-microsensor platform and discuss its ability to generate orthogonal data through materials selection and temperature programming. Visual inspection of data sets reveals device selectivity, but statistical analyses are required to perform more complex identification tasks. Finally, we discuss recent advances in both devices and algorithms necessary to deal with practical issues involved in long-term deployment. These issues include identification and correction of signal drift, challenges surrounding real-time unsupervised operation, repeatable device manufacturability, and hierarchical classification schemes designed to deduce the chemical composition of untrained analyte species.