Simultaneous determination of Zn(II), Cd(II) and Hg(II) in water

The use of artificial neural networks (ANNs) as a tool for simultaneous determination of Zn(II), Cd(II) and Hg(II) in water has been investigated, by employing an optode based on the 2-(5-bromo-2-pyridylazo)-5-(diethylamino)phenol (Br-PADAP) complex immobilised on Amberlite XAD-4 resin.

Studies were performed with binary and ternary mixtures of Zn(II), Cd(II) and Hg(II) ions in the 0–3.0 mg l⁻¹ range (prepared in pH 8.0 NH₄Cl solution), in order to avoid sensing phase saturation. A 0.1 mol l⁻¹ HCl solution was employed to regenerate the optode after each measurement. Reflectance measurements were recorded in the 400–800 nm region after pumping metal ion solution for 1, 1.5, 2.0, 2.5 and 3.0 min. Spectra were smoothed by employing a weighted sliding average of three reflectance intensities (weights of 1:3:1) and normalised to the reflectance intensities of the immobilised reagent in the NH₄Cl solution. A feed-forward ANN with error back-propagation training algorithm was employed for data treatment. The ANN was initially fed through seven inputs neurons, corresponding to reflectance intensities at 558, 568, 583, 590, 605, 615 and 623 nm, and optimised with respect to the number of hidden neurons, momentum and learning rate. Binary mixtures of Zn(II) and Cd(II), and Hg(II) and Cd(II) were employed to evaluate the capability of prediction by the ANN. Measurements were carried out in triplicate leading to standard deviations (expressed in mg l⁻¹) better than 0.29 for Cd(II), 0.38 for Hg(II) and 0.35 for Zn(II) and absolute errors better than 0.14 for Cd(II), 0.31 for Hg(II) and 0.36 for Zn(II). Similar results were obtained when solution mixtures containing all three metal ions were employed in the studies.

These results indicate that ANN can be employed for simultaneous determination of metal ions in water. However, attention must be given to the leaching and to the saturation of the sensing phase, which limit the lifetime and analytical range of the sensor, respectively.