An optoelectronic nose for the detection of toxic gases

Abstract

We have developed a simple colorimetric sensor array that detects a wide range of volatile analytes and then applied it to the detection of toxic gases. The sensor consists of a disposable array of cross-responsive nanoporous pigments with colours that are changed by diverse chemical interactions with analytes. Although no single chemically responsive pigment is specific for any one analyte, the pattern of colour change for the array is a unique molecular fingerprint. Clear differentiation among 19 different toxic industrial chemicals (TICs) within two minutes of exposure at concentrations immediately dangerous to life or health were demonstrated. Based on the colour change of the array, quantification of each analyte was accomplished easily, and excellent detection limits were achieved, generally below the permissible exposure limits. Different TICs were identified readily using a standard chemometric approach (hierarchical clustering analysis), with no misclassifications over 140 trials.

Fig. 1. The colorimetric sensor array (CSA) consists of 36 different chemically responsive pigments that have been printed directly on a polyethylene terephthalate (PET) film

a An image of the array from an ordinary flatbed scanner with the different pigment classes labeled. b. Molecular structures of examples from each dye class are shown; in order to absorb visible light, dyes are inherently nanoscale, as noted by the scale bar. The 36 dyes were selected empirically based on the quality of their color response to a representative selection of chemically diverse analytes.

Fig. 2. Color change profiles of representative toxic industrial chemicals (TICs) at their IDLH (immediately dangerous to life or health) concentration after 2 min of exposure

The IDLH concentrations are listed under each analyte in ppm. A full digital database is provided in the Supplementary Table S3. For display purposes, the color range of these difference maps are expanded from 4 to 8 bits per color (RGB range of 4–19 expanded to 0–255).

Fig. 3. The effect of concentration on array response to NH_3, Cl₂, and SO₂

The color change profiles of these three representative TICs are shown over roughly a hundred-fold range of concentration. The observed limits of detection (LOD) are well below the PEL (permissible exposure limit). For display purposes, the color range of these difference maps are expanded from 4 to 8 bits per color (RGB range of 4–19 expanded to 0–255), except for SO₂ at 1 ppm (RGB range of 2–7).

Fig. 4. Response time of the array

Total Euclidean distance of the array is plotted versus time for five representative TICs at their IDLH concentrations; the average of seven trials is shown. The Euclidean distance is simply the total length of the 108-dimensional color difference vector, i.e., the total array response.

Fig. 5. Reversibility of colorimetric array response

a, SO₂ exposure of the array from N₂ to the IDLH (100 ppm) level, and then repeatedly from the IDLH to PEL (5 ppm) and back. b, Cl₂ exposure of the array from N₂ to the IDLH (10 ppm) level, and then repeatedly from the IDLH to PEL (1 ppm) and back. Data was acquired every min. The response times in this data reflect the dead volume in the gas mixing system and do not represent the intrinsic response of the array (which is shown in Fig. 4). The error bars shown are the standard deviation of triplicate trials.

Fig. 6. Hierarchical cluster analysis (HCA) for 19 TICs at IDLH concentrations and a control

HCA is a routine model-free statistical classification method based on Euclidean distance–. In these experiments, the Euclidean distances are defined by the color difference vectors of each experimental trial in the full 108-dimensional space made up of the changes in red, green, and blue values of the 36 nanoporous pigments in the sensor array. As shown, all experiments were run in septuplicate: no confusions or errors in classification were observed in 140 trials. The IDLH concentrations of each analyte are shown in ppm. The HCA used minimum variance (i.e., “Ward’s Method”) for clustering.