Uv-Vis Spectrum Analysis Of Heavy Metal Detection

UV-Vis spectroscopy is routinely used in the quantitative determination of solutions of transition metal ions and highly conjugated compounds. One of its applications is to find the molar absorptivity of different kind of heavy metals ions. Using molar absorptivity, concentration of heavy metals ions can be deduced based on the established calibration curve. This calibration curve first need to be constructed by repeatedly analyzed it UV-Vis spectrum analysis of the concerned ion with controlled concentration. The plot between known concentration and its corresponding absorptivity is used as a calibration curve to predict the unknown concentration of samples based on it absorptivity. The aim of this project is to study several different type of relationship between the Uv-vis spectroscopy absorptivity and heavy metals ion concentration. Conventionally, the relationship between absorptivity intensity variable and it corresponding concentration are all assume as linear. Heavy metals on the other hand show non-linear correlation and when the calibration curve wrongly adapts same construction method, it gives high errors. From three tested model fitted, experiment show that quadratic model is the most suited model to explain the relationship. Apart from the relationship between variable, this project also investigates a multipoint predictor to improve the calibration accuracy. We apply multivariate analysis to the several data extracted from various location of uv-vis spectrum and transform it to single most accurate predictor variable. Partial Least Square Regression (PLSR) analysis proven superior both in prediction of sample concentration and the degree of line fitting to the point compared to Principle Component Regression (PCR) analysis method. PLSR is capable in providing lowest error when linear regression calibration curve are required. Apart from those two factors, the project also uncovers several key factors in helping to improve the accuracy, one of the factor consist of the properties of raw data extraction decision. Different location of point extraction location and different number of point required vary by the transformation method used, but in most occasion, highest peak and second highest peak in spectrum show better potential for both single point raw data and multiple point extracted for preprocessing. Preprocessing step is another factor that supports the improvement of the calibration curve. Instead using the intensity of absorptivity in a wavelength spectrum, different between chosen two points is used as a raw data. Overall, the best calibration curve is the combination of preprocessing step and quadratic model fit.