Heavy metal contamination poses significant environmental and public health risks. Conventional detection methods, such as atomic absorption spectrometry, high-performance liquid chromatography, and mass spectrometry coupled with electrochemical and fluorescence detection, are highly sensitive but costly and require extensive sample preparation. On the contrary, electrochemical sensors with anodic stripping square wave voltammetry (SWV) detection techniques offer a rapid, on-site, and cost-effective alternative; however, their performance is hindered not only by the low electrical conductivity and high charge transfer resistance of the electrodes and poor electrocatalyst coatings but also by operating the sensor in suboptimal SWV conditions. In this study, first, we investigated the effect of electrochemical polishing (ECP) on the general electrochemical behavior of carbon screen-printed electrodes (cSPEs) for detecting three heavy metals: cadmium (Cd²⁺), lead (Pb²⁺), and arsenic (As³⁺), using various electroanalytical, imaging, and statistical techniques. Cyclic voltammetry revealed a 41±1.2% increase in electroactive area and a 51±1.6% decrease in potential separation in ECP-treated electrodes, indicating a significant improvement in electrode kinetics (n=3). Electrochemical impedance spectroscopy further confirmed an 88±2% reduction in charge transfer resistance, demonstrating enhanced electrical conductivity. By suitably modifying the ECP-treated cSPEs with bismuth, molybdenum disulfide, reduced graphene, and iron oxide nanoparticles, we demonstrate high sensitivity values of 6.6±0.1, 2.9±0.1, and 6.5±0.1 μA ppb¹ cm² for Cd²⁺, Pb²⁺, and As³⁺, respectively. Secondly, we investigated the effects of pH, SWV parameters, nanocomposite formulations and loadings, as well as pre-concentration steps on various sensor metrics. By applying the Fast Causal Inference algorithm, we identified the causal structure between the SWV frequency (5-50 Hz), step potential (5-70 mV), and amplitude (1-100 mV) and the sensor’s limit of detection, sensitivity, and full width at half maximum. The results show that frequency is correlated with both the limit of detection and the full width at half maximum while remaining independent of sensitivity. In contrast, amplitude is correlated only with sensitivity. Additionally, step potential is independent of all observed variables. Finally, we validated the sensor function in spiked surface water samples (i.e., 1, 2, 5, 10, 20, and 30 ppb heavy metal concentrations) containing interferents such as zinc and copper, achieving sub-ppb detection limits. The proposed work showcases an effective strategy for creating and implementing affordable, portable sensors for extensive on-site environmental water monitoring and safety applications.
