An unmodified graphene foam chemical sensor based on SVM for discrimination of chemical molecules with broad selectivity

Abstract

Compared to conventional chemical sensors, this paper presented a chemical sensor system with broad selectivity for a variety of molecules without any surface modification. The system consisted of an unmodified graphene foam as sensing element, an electrical resistance time domain detection system and a Support Vector Machine (SVM) identification system. The chemical sensor adopted 3D graphene foam to increase the reaction area and improve the sensitivity for detecting target molecules. The electrical resistance time domain detection system was constructed to record the graphene resistance curve in real time with different molecules. Based on the diverse shapes of the electrical resistance curves, SVM was used to extract features of each resistance curve and discriminate the corresponding molecules via pattern recognition of each resistance curve without any graphene modification. As validation experiments, six kinds of chemical molecules (chloroform, acetone, ether, toluene, ethyl benzene and methanol) have been tested. The discrimination accuracy for each molecule could be above 98% which showed a broad selectivity for a variety of molecules. Furthermore, through theoretical calculation with the first principle, we concluded that different band structures of the graphene caused by different molecules were the mechanism for the graphene chemical sensor system to discriminate chemical molecules with selectivity. This work may present a new strategy for research and application for graphene chemical sensors.