In this study, we report the design and analytical validation of a glyphosate nanosensor based on the modulation of upconversion emission from NaYF4:Yb3+/Er3+ nanoparticles. The sensing mechanism relies on a rapid and straightforward interaction between Cu2+ ions and the upconverting nanoparticles (UCNPs), which occurs selectively in the presence of glyphosate. The functional groups of the herbicide enable simultaneous coordination with both Cu2+ and the UCNPs, triggering a selective resonant energy transfer mechanism that distinguishes glyphosate from other organophosphates, including its own metabolites. This is transduced in specific variations of the upconversion emission signals as well as the light scattering of the samples, both of which correlate with glyphosate concentration. To translate this sensing capability into an on-site detecting tool, we developed a convolutional neural network (CNN) embedded into a portable optoelectronic device capable of robust data analysis and accurate quantification of glyphosate concentrations from emission and light-scattering images captured by a compact camera. The CNN-generated results are transmitted via Bluetooth to a mobile application for real-time visualization and monitoring. The integrated system achieved a coefficient of determination R2 = 0.987, demonstrating high predictive accuracy in measurements completed in just 5 s. Each measurement involves capturing numerous images under white LED illumination and then under 975 nm laser excitation followed by automated data integration and analysis by the neural network to yield a glyphosate concentration value. A minimum detectable concentration of 12.5 ppm was achieved, within a linear response range up to 3125 ppm, underscoring the simplicity and practicality of the detection process. These results highlight the potential of this technology for rapid, reliable, and scalable monitoring of glyphosate in soil and water under field-relevant conditions.
Molina-González et al. (Fri,) studied this question.