Identificador: TDX:773
Autores: Comas Lou, Enric
Resumen:
Analytical measurements and the instruments that generate them can be classified regarding the number of data that are obtained when a sample is measured. When a matrix of response is obtained, it is known as second-order data.In this thesis, second-order data were used, obtained from a high performance liquid chromatography (HPLC) couple with a diode array detector (DAD). This instrument is quite common in the analytical laboratories. However, the concentration of the analytes of interest is normally found without using all the measured data. The spectral model only is used to identify the analytes of for verifying the peak purity, whereas the area or the height of the peak is used to quantify using univariate calibration. This is a very useful strategy. However, the measured response must be selective to the analyte of interest.When environmental pollutants were analyzed, like water samples, it is no so easy to get selective measurements. When the responses are not selective, the analyte on interest can still be quantified by using second-order calibration methods, which are the methods that use second-order data.This thesis is based on the study of the properties of the second-order calibration method Generalized Rank Annihilation Method (GRAM).This method was developed in the mid eighties and has very attractive properties:1) To determine the concentration of the analyte of interest in a test sample, it is only necessary one calibration sample or standard.2) Selective measurements are not necessary, implying the reduce of the separation time.Despite these advantages, GRAM has some limitations which make that it is not applied routinely. The objectives of the thesis are to study the advantages and limitations of GRAM and improve the negative points in order to apply GRAM routinely.To use GRAM the experimental data must accomplish some mathematical requirements: (i) the measured response must be result of the addition due to the different analytes in the peak and (ii) the response of the analyte must be proportional in the different samples: the analyte of interest must elute at the same retention time both in the calibration and in the test sample. When these conditions are not met, the GRAM predictions are biased.Mathematical algorithms have been developed to overcome such difficulties. An algorithm to align chromatographic peaks has been developed, based on curve resolution method (Iterative Target Transformation Factor Analysis, ITTFA). In HPLCDAD systems is quite often that the peaks of the analyte of interest elute at different retention time in the calibration and in the test sample. Even the differences are not big (few seconds), they can be enough to make the GRAM results incorrect.GRAM is a factor based calibration method, and the number of factors has to be introduced as an input to build a GRAM method. A graphical criterion has been selected to determine the number the number of factors, which is base on the use of a parameter of the GRAM algorithm (á).Finally, a criterion to detect outlying samples has been developed, which is based on the Net Analyte Signal (NAS).All the above commented were applied to real cases. Specifically to the analysis of aromatic sulfonates and polar pollutants in water form river samples and waste water plants. We were able to show the applicability of GRAM and to compare GRAM with other second-order calibration methods, such as PARAFAC i MCR-ALS. We found that the three methods provided comparable results.
Repositori URV