Statistical pattern analysis assisted selection of polymers for odor sensor array

Abstract

This paper demonstrates application and usefulness of multivariate statistical methods like principal component analysis (PCA) and hierarchical cluster analysis (HCA) for design and development of polymer coated acoustic wave (surface and bulk) vapor sensor array. Individual sensors in the array are functionalized with different polymers having varied selectivities towards chemical constituents of odor samples. In response to an odor sample a response pattern is generated analogous to olfactory epithelium in human nose. The pattern recognition algorithms extract vapor identification signatures similar to signal processing in biological nose. The successful operation of such an artificial nose system depends critically on the polymer coating materials that sensitize individual sensors so that the response patterns carry most discriminating information about vapor identities. The objective of this work is to show that the process of polymer selection and sensor array development can be made efficient by exploiting the existing vapor-polymer interaction databases through statistical data mining procedures and sensor array simulation for target vapor detection and identification. A prototype case study of polymer coated surface acoustic wave (SAW) sensor array for discrimination of human body odor is presented. The detection of 14 major volatile organic compounds (VOCs) emanating from different parts of human body (skin, breath, blood, armpit) has been targeted. The available information about vapor-polymer partition coefficients is analyzed by PCA and HCA methods to select most efficient polymers. Out of 22 commercial and synthesized potential polymers for this purpose we found that a subset of 3 polymers: alkyl-amino-pyridyl substituted polysiloxane (SXPYR), poly4-hexafluoroisopropanol-styrene (P4V) and poly ethylenimine (PEI), produces the most optimal discrimination between different odor constituents.