Gold (III) paper sensor with machine learning for field arsenic detection.

A paper-based sensor doped with Au (III) has been developed as a non-toxic, stable, and cost-effective alternative to mercuric bromide reagent in field-deployable arsenic detection kits. The sensor requires only microgram quantities of gold to detect arsenic concentrations in the range of 2-100 µg/L, with dilution applied for higher concentrations. Real water samples were successfully analysed by incorporating optimized reagents to minimize potential interferences. The detection process is simple, rapid, sensitive, and reliable, making it suitable for large-scale water quality monitoring. The detection limit of the method is significantly lower than the WHO permissible value of 10 µg/L. Reducing the gold loading on the strip enhanced sensitivity at lower arsenic concentrations. To enhance real-time, on-site assessment, machine learning algorithms were applied to RGB values extracted from the colorimetric response of the sensor, enabling accurate, data-driven prediction of arsenic concentrations directly in the field. Of the machine learning models investigated, the neural network model achieved the highest performance. This integrated approach provides a scalable and environmentally friendly solution for safeguarding drinking water quality, particularly in regions where arsenic contamination poses a significant public health challenge.