Extracting information from noise spectra of chemical sensors: single sensor electronic noses and tongues L.B. Kish *,1 , R. Vajtai, C.G. Granqvist The A Ê ngstro Èm Laboratory, Department of Materials Science, Uppsala University, P.O. Box 534, SE-75121 Uppsala, Sweden Received 10 April 2000; received in revised form 20 June 2000; accepted 22 June 2000 Abstract Electronic noses and tongues can utilize noise data taken at the output of a chemical sensor. It is shown that even one single sensor may be suf®cient for realizing an electronic nose or tongue. # 2000 Elsevier Science B.V. All rights reserved. Keywords: Sensor principle; Electronic nose; Electronic tongue; Excess noise spectroscopy 1. Introduction Intensive research has been going on during the past several years to use chemical and biological sensor [1±5] elements to develop systems known as electronic noses and electronic tongues [6±10]. They consist of several sensor elements and a pattern recognition unit comprising data acquisition and usually a neural network software. The neural network, or a similar pattern recognition tool, is necessary due to the nonlinear character of the sensor system, which `learns' to interpret data during a calibration process. There are several important practical issues, such as ways to decrease the number of necessary sensors, ascer- taining suf®cient sensitivity, obtaining reproducibility, diminishing the need for frequent calibration, and establish- ing the most suitable pattern recognition technique. This paper addresses the need for multiple sensors and demon- strates that measurements of conductivity noise in a sensor can lower the required number of sensors to, in principle, only one. Conductivity noise in conducting polymer sensors was studied by Bruschi and coworkers [11,12] who pointed out that such data provided information of interest for sensing. The authors demonstrated that the conductance noise spec- trum is a sensitive measure of the chemical environment. The exact origin of the microscopic noise component, due to the chemical environment is presently unclear. It can cer- tainly be associated with ¯uctuations of the carrier mobility and density, due to concentration ¯uctuation and motion of chemical fragments, originating from the chemical environ- ment (ambient gas or liquid). The resultant resistance ¯uc- tuations provide an ac signal with wide frequency bandwidth, which obviously carries more information than the dc resistance, that is normally recorded in a sensor. In addition to being able to detect multiple species, the noise technique enhances the sensitivity and selectivity of the sensor. 2. The new nose/tongue principle We propose to use the random temporal ¯uctuations (noise) of the measured physical quantity in arti®cial noses and tongues, see Fig. 1. We would like to emphasize that the generality of the principle, makes it of secondary importance only to consider the actual sensor device [1±5] (®lm, tran- sistor, optical surface, etc.) and the corresponding physical quantity (voltage, current, resistance, capacitance, light intensity, angle, etc.) applied for sensing. The mathematical expressions given below hold for any measured quantity which is used for chemical sensing. However, for the sake of simplicity [13], we use here the term resistance to represent the pertinent physical quantity. Assuming then a resistive sensor exposed to a chemical environment, we propose the use of the resistance noise dR(t) of the sensor resistance R instead of using the induced static change dR of its mean value. The exact origin of resistance ¯uctuations [14±20] in solids is a longstanding unsolved problem [15±17]. Different results have been interpreted in different ways, and even basic issues are Sensors and Actuators B 71 (2000) 55±59 * Corresponding author. Tel.: 46-18-471-7232; fax: 46-18-500-131. E-mail address: laszlo.kish@angstrom.uu.se (L.B. Kish). 1 Until 1999, his name was L.B. Kiss. 0925-4005/00/$ ± see front matter # 2000 Elsevier Science B.V. All rights reserved. PII:S0925-4005(00)00586-4