From pen to paper-based screening: an early-stage colorimetric µPAD with desktop-app readout for lead and nitrite

Industrial wastewater often contains several toxic contaminants, among which lead (Pb²⁺) and nitrite (NO₂ ^-) ions are regarded as significant contributors to human health risks. These are readily absorbed by humans <i>via</i> drinking and agriculture. This growing public health concerns demand urgent, on-site screening as it is a crucial step for monitoring industrial effluent hotspots in both rural and urban communities. Detection of these requires the use of complicated gold standard technology such as Atomic Absorption Spectroscopy, High Performance Liquid Chromatography and so on, which is challenging to establish at testing locations and much more difficult for laypersons to comprehend. Hence, the present work seeks to create a straightforward colorimetric paper sensor for the simultaneous screening of lead and nitrite ions in aqueous solutions, which combines facile fabrication methods with easy interpretation. The sensor utilises sodium rhodizonate and Griess reagent modified paper for the colorimetric readout. The readout can be obtained using a standard smartphone camera connected to the darkroom, simulating attachment to guarantee uniform lighting conditions. Further to counter the problems of reagent decomposition, a graphite-based reagent pencil has been fabricated that increases the solid-state stability of sodium rhodizonate on paper from 2 hours to more than 2 months under light-protected storage, independent of the nitrite assay components. Additionally, to fully integrate the sensor into an accessible detection system, a lightweight machine learning regression model coupled with a graphic user interface through desktop application has been used for semi-quantitative interpretation of the smartphone images. The model achieved an <i>R</i>-squared value of 0.95 and 0.96 for lead and nitrite ions, respectively. Colorimetric responses on the paper sensor were evaluated using samples collected from various water bodies and agricultural lands. The algorithm accurately predicted concentrations, achieving detection limits of 7 ppm for lead and 3 ppm for nitrite. The sensor's performance showed minimal influence from ambient temperature and humidity. Overall, a simple paper-based semi-quantitative approach for preliminary screening of common water pollutants at sites of highly contaminated water samples influenced by industrial activities could be demonstrated as a proof of concept.