Estimation of the limit of detection in semiconductor gas sensors through linearized calibration models

Javier Burgués; Juan Manuel Jiménez-Soto; Santiago Marco

doi:10.1016/j.aca.2018.01.062

Estimation of the limit of detection in semiconductor gas sensors through linearized calibration models

Anal Chim Acta. 2018 Jul 12:1013:13-25. doi: 10.1016/j.aca.2018.01.062. Epub 2018 Feb 7.

Authors

Javier Burgués¹, Juan Manuel Jiménez-Soto², Santiago Marco³

Affiliations

¹ Department of Electronics and Biomedical Engineering, Universitat de Barcelona, Marti i Franqués 1, 08028, Barcelona, Spain; Signal and Information Processing for Sensing Systems, Institute for Bioengineering of Catalonia (IBEC), The Barcelona Institute of Science and Technology, Baldiri Reixac 10-12, 08028, Barcelona, Spain. Electronic address: jburgues@ibecbarcelona.eu.
² Signal and Information Processing for Sensing Systems, Institute for Bioengineering of Catalonia (IBEC), The Barcelona Institute of Science and Technology, Baldiri Reixac 10-12, 08028, Barcelona, Spain.
³ Department of Electronics and Biomedical Engineering, Universitat de Barcelona, Marti i Franqués 1, 08028, Barcelona, Spain; Signal and Information Processing for Sensing Systems, Institute for Bioengineering of Catalonia (IBEC), The Barcelona Institute of Science and Technology, Baldiri Reixac 10-12, 08028, Barcelona, Spain.

PMID: 29501088
DOI: 10.1016/j.aca.2018.01.062

Abstract

The limit of detection (LOD) is a key figure of merit in chemical sensing. However, the estimation of this figure of merit is hindered by the non-linear calibration curve characteristic of semiconductor gas sensor technologies such as, metal oxide (MOX), gasFETs or thermoelectric sensors. Additionally, chemical sensors suffer from cross-sensitivities and temporal stability problems. The application of the International Union of Pure and Applied Chemistry (IUPAC) recommendations for univariate LOD estimation in non-linear semiconductor gas sensors is not straightforward due to the strong statistical requirements of the IUPAC methodology (linearity, homoscedasticity, normality). Here, we propose a methodological approach to LOD estimation through linearized calibration models. As an example, the methodology is applied to the detection of low concentrations of carbon monoxide using MOX gas sensors in a scenario where the main source of error is the presence of uncontrolled levels of humidity.

Keywords: Humidity interference; Limit of detection; Metal-oxide sensors; Non-linear; Semiconductor gas sensors; Temperature modulation.