Resistive Sensor Array for Selective Zn(II) Ion Detection From a Mixed Solution Using Machine Learning Techniques

Advancing industrialization has led to an exponential degradation of consumable water, and hence, there is a direct need for an efficient monitoring system. In this study, we demonstrated hierarchical, nanostructured Ni2O3- and Cu0.05Ni1.95O3-based resistive bisensor arrays for selective detection of Zn(II) ions in solution. The sensors showed a maximum response of 4.12 times and 6.34 times for Cu(II) and Zn(II) ions individually when exposed to Ni2O3 (device 1) and Cu0.05Ni1.95O3 (device 2) receptor layers, respectively. The limit of detection (LOD) was estimated as ~6 and ~4.6 ppb with the fast response times of 2.4 and 2.8 s, respectively, for the two devices. The devices showed excellent repeatability with a maximum response variation of 7.22% and 8.24% for 160 ppm of the two ions over a period of 180 days. Though the sensors performed well for individual ions, however, both the sensors failed to detect the individual ions when exposed to a mixed solution. Hence, a multimodal fusion approach coupled to random forest algorithm was used to identify and quantify Zn(II) ions from a mixture, which showed an estimation error of ~4% and an
${R}^{{2}}$
score of 0.91. An “adsorption energy-assisted sensing model” has been proposed to explain the unusual behavior of the sensors in the presence of a mixed solution. It supports that the proposed array is significantly more efficient than its conventional market counterparts when it comes to long-term water quality monitoring.