Application of partial least squares multivariate calibration for the determination of mixtures of carbaryl and thiabendazole in waters by transmitted solid phase spectrophosphorimetry L. F. Capitán-Vallvey, Mahmoud Kalil A. Deheidel, I. de Orbe and R. Avidad* Department of Analytical Chemistry, University of Granada, Granada 18071, Spain. E-mail: lcapitan@goliat.ugr.es Received 7th October 1998, Accepted 5th November 1998 Mixtures of the pesticides carbaryl and thiabendazole were studied using solid phase spectrophosphorimetry at room temperature with Whatman No. 4 filter-paper as solid support and Pb(ii) as enhancer of the phosphorescence. The transmitted phosphorescence was used as an analytical signal and its measurement was performed by placing the paper containing the sample between two quartz plates. In this way, the use of a dry inert gas flow to avoid the quenching of atmospheric agents (oxygen and moisture) is not needed. The influence of several experimental parameters (e.g., pH, nature of the enhancers and solid supports, gate and delay times) on the phosphorescent emission of both chemicals was also studied.As a result, a new method for the simultaneous determination of these pesticides is proposed, using partial least squares (PLS-1) multivariate calibration. The statistical parameters of the optimised matrix, are included as a table. The applicable concentration ranges were from 0.16 to 1.20 mg l21 for thiabendazole and from 0.50 to 4.00 mg l21 for carbaryl.The method was applied to the determination of both pesticides in water, with recoveries between 93.4 and 105.6% for carbaryl and between 90.0 and 105.0% for thiabendazole. HPLC was used as a reference method. Introduction One of the current requirements in analytical chemistry is to detect and quantify smaller amounts of chemical contaminants in samples of diverse natures, with low cost and time consumption. Hence selective, sensitive, rapid and inexpensive methods of analysis, using readily available, conventional instrumentation, must be developed.Room temperature phosphorimetry is a technique that allows the development of analytical methods that combine these qualities because, together with the natural sensitivity and selectivity of luminescent methods, it can reduce costs through the use of inexpensive materials and reagents, and provides short analysis times. The methodology employed in room temperature phosphorescence (RTP) is based on the measurement of the phosphorescence intensity in a solid phase,1–5 or in solution if the analyte forms an inclusion complex (cyclodextrins),6,7 or micellar-stabilised media with non-polar molecules containing a polar group.8–10 Obviously, the use of paper as a solid support diminishes the cost of the analysis, but it has the disadvantage of showing a background signal produced by the substrate, which can limit its application in the detection of trace analytes.There are several ways to diminish the background signal of the paper used as a solid support, the best known being chemical pre-treatment, irradiation with ultraviolet light or extraction with water in a Soxhlet apparatus.11–15 With these treatments of the paper used as support, the background signal can be 6% of its initial value, but the time of analysis increases and the precision of the measurements can decrease. Additionally, a factor influencing strongly the phosphorescence of the sample and decreasing the intensity of the emitted signal, both in solution and in the solid phase, is the quenching effect produced by environmental moisture and oxygen. This effect has been avoided in the past by the addition of sulfite, in the case of measurements in solution,16 or by using a flow of a dry, inert gas (usually N2 or He) in the compartment of the spectrometer when the measurement is performed in the solid phase.17 However, it is clear that the use of a flow of inert gas and/or pre-treatment of the solid support to avoid background signals increases the time and cost of the analysis.With the aim of improving the experimental procedures, using paper as a solid support, we studied the phosphorescent behaviour of phosphors, modifying the experimental procedure that is usually used in phosphorimetry. This modification consists of the measurement of the diffuse transmitted phosphorescence emitted from the analyte, placing the paper containing the sample between two quartz plates.In this way, a flow of inert gas and pre-treatment of the paper used as a solid support are not needed and the analysis is consequently faster. A second objective was to test the suitability of multivariate calibration methods for the determination of mixtures of analytes in the solid phase whose phosphorescence spectra overlap substantially. The phosphors studied here were two pesticides widely used in agriculture, carbaryl (CBL) and thiabendazole (TBZ).The former has been described as a phosphorescent chemical,1,2,9,15,18 but we did not find any references to the phosphorescence emission of TBZ. These chemicals show similar spectra, and to resolve the considerable spectral overlap we used the partial least squares (PLS) multivariate calibration method, developed and applied by Wold et al.19–21 The main advantages of this statistical method are the speed of the handling of results and the determination of the analytes without previous separation. PLS methods have been used in several studies for the determination of different analytes (e.g., pesticides, sulfonamides and flavour enhancers) 22–29 in solution, although the main application of the PLS methods has been for calibration in near-infrared and nuclear magnetic resonance spectroscopy in the solid phase.30 In this work, in addition to the study mentioned above of the phosphorescent behaviour of CBL and TBZ, we applied the PLS method in solid phase spectrophosphorimetry to test its suitability for this methodology.As a result of this study, a new method for the determination of CBL and TBZ in mixtures is proposed. The method combines the advantages of solid phase Analyst, 1999, 124, 49–53 49spectrophosphorimetry (selectivity and sensitivity) and of multivariate calibration (speed and no previous separation of components). It is also inexpensive, because only common reagents and materials and conventional instrumentation are needed.Experimental Apparatus The measurements of phosphorescence were performed with a Perkin-Elmer (Norwalk, CT, USA) LS-50 luminescence spectrometer, working in the phosphorescence mode, and equipped with a Hammamatsu R289 photomultiplier, two Monk– Gillieson F/3 monochromators and a xenon discharge lamp (with a power equivalent to 20 kW during 8 ms), and interfaced to a PS/230-386 microcomputer.The spectrometer was checked daily using a P1 standard (12.5 3 12.5 3 45 mm) containing europium(iii) thenoyltrifluoroacetonate dissolved in a transparent matrix of poly(methyl methacrylate), supplied by Perkin- Elmer. Other apparatus and laboratory materials were a Crison (Barcelona, Spain) digital pH meter with a combined glass– saturated calomel electrode, a micropipette Biohit Proline (Helsinki, Finaland) microtip 10 ml, a Selecta (Barcelona, Spain) ultrasonic bath, a Braun Silencio 1600 hair-dryer (maximum power 2000 W) and a 250 W infrared heat lamp.Two rectangular (45 3 12 3 1 mm) Hellma Suprasil (M�ullheim, Baden, Germany) quartz plates were also used to perform the measurements of the transmitted RTP. Software programs used for the treatment of the spectral data were Grams/286 Software Package Version 1.0, Add Application PLS Plus Version 2.1 (Galactic Industries, Salem, MA, USA) and Data Leader Software Package (Beckman, Fullerton, CA, USA; 1987). Reagents Stock standard solutions of TBZ (Dr.Ehrenstorfer, Ausburg, Germany) and CBL (Riedel-de Haën, Hannover, Germany) were prepared in ethanol at a concentration of 100.0 mg l21. Working standard solutions were prepared by appropriate dilution with reverse osmosis quality water (obtained using a Milli-Ro 12 plus Milli-Q Station from Millipore, Bedford, MA, USA). Stock standard solutions of saturated Pb(ii), 1 m Tl(, 1 m Ag(i) and 1 m KI were prepared from lead acetate, AgNO3, TlNO3 and KI solid salts (Merck, Darmstadt, Germany), respectively.A buffer solution of the required pH (4.0) was prepared from 1 m sodium acetate and 1 m acetic acid solutions (Merck). All reagents were of analytical reagent grade unless stated otherwise. Phosphorescence measurements Diffuse transmitted phosphorescence spectra were recorded between 400 and 570 nm, the plane of the sample forming two angles of 45° with the excitation and emission beams (Fig. 1).31 These spectra were obtained with a delay time td = 0.1 ms and a gate time tg = 12 ms. The excitation and emission slits were 2.5 and 5.0 nm, respectively, and the scan speed was 240 nm min21. The solid support containing the sample was placed in the holder of the spectrometer shown in Fig. 1. Procedure On a 50 312 mm strip of Whatman (Maidstone, Kent, UK) No. 4 filter-paper soaked in acetic acid–sodium acetate buffer solution (pH 4.0) and then dried, 3 ml of saturated Pb(CH3COO)2 solution and 3 ml of sample solution containing between 0.5 and 3.6 ng of TBZ and between 1.5 and 12.0 ng of CBL were spotted with the aid of a micropipette.The paper containing the reagents was dried for 3 min by means of a hot air stream from the hair-dryer placed 7 cm above the sample at 600 W power. Next, the solid support with the sample was placed between the quartz plates and the assembly was inserted in the sample compartment of the instrument. The phosphorescence spectra were obtained as described in the previous section.Results and discussion Spectral characteristics CBL and TBZ in the solid phase gave the emission spectra shown in Fig. 2. It can be seen that the spectra of both analytes Fig. 1 Placing of the sample and quartz plates in the holder of the spectrometer. Fig. 2 Phosphorescence spectra of TBZ and CBL fixed in Whatman No. 4 filter-paper. (a) Blank; (b) CBL at 3.8 mg ml21; (c) TBZ at 1.0 mg ml21; (d) CBL (2.8 mg ml21) and TBZ (0.8 mg ml21) mixed. 50 Analyst, 1999, 124, 49–53overlap greatly and they are clearly different from the blank spectrum. The peak wavelength in the emission spectrum of the TBZ is 472 nm (lex = 303 nm) and for CBL there are two peaks in the emission spectrum at 486 and 520 nm (lex = 282 nm). This overlap hinders the simultaneous determination of the two chemicals by conventional spectrophosphorimetric methods, but the problem can be resolved by using multivariate calibration after the optimisation of the experimental parameters that influence the phosphorescence intensity emitted by the phosphors.The experimental parameters, individually optimised for each pesticide, were as follows: pH, volume of analyte, nature and volumes of buffer and heavy ion solutions, delay and gate times, nature of the solid support and drying time of the sample. Whatman No. 4 and Albet No. 1305 filter-papers and Whatman P-81 and Whatman DE-81 ionic exchange papers were checked as solid supports.It was found that the greatest difference between the signals produced by the sample and the respective blank were obtained when Whatman No. 4 filterpaper was used as the solid support. Next, in order to minimize the background signal produced by the solid support, several strips of Whatman No. 4 paper were subjected to extraction with reverse osmosis quality water in a Soxhlet column for 24 h, after which they were dried and irradiated with ultraviolet light for 12 h.The blanks prepared with these pre-treated supports presented a background signal three times lower than the blanks prepared with the paper without pre-treatment; however, the net phosphorescence intensity (sample signal 2 blank signal) obtained by using pre-treated papers was around 14% lower with papers without pre-treatment. As a consequence, Whatman No. 4 filter-paper without pre-treatment was used as the solid support in subsequent experiments, giving a better analytical signal than with pre-treated paper, and in a shorter time.Second, the influence of the heavy ion usually used as an enhancer of phosphorescence intensity was tested, using Pb(ii) as the acetate salt, Tl(i) and Ag(i) as the nitrate salts and KI at different concentration levels. The best net phosphorescence intensity was obtained using saturated Pb(ii) solution as the enhancer for TBZ and 1 m Tl(i) or saturated Pb(ii) solution for CBL. As a consequence, Pb(ii) was selected as the enhancer of phosphorescence intensity for both analytes.Next, different volumes of saturated Pb(ii) solution, ranging between 1.0 and 5.0 ml, were tested in order to find the optimum volume. It was found that 3.0 ml produced the best net phosphorescence intensity (NPI). Hydrochloric acid and sodium hydroxide solutions of different concentrations were used to test the influence of pH on the phosphorescence emitted by the analytes. As can be seen in Fig. 3, the phosphorescence intensity emitted by CBL remains constant at pH values between 2.0 and 8.0, decreasing rapidly at lower or higher pH values. The decrease at pH > 8.0 is due to the hydrolysis of CBL, producing the derivative 1-naphthol, which does not show phosphorescence under the present experimental conditions. For TBZ, pH has no influence on the phosphorescence intensity in the range 1.0–10.0, and only in a very basic medium (pH � 12.0) is a decrease noticeable. As a consequence of these results, although the pH did not appreciably influence the phosphorescence intensity of either chemical, standards and samples were measured at pH 4.0 to ensure that the CBL was not hydrolysed during the preparation of the sample.A 1 m acetic acid–sodium acetate buffer solution (pH 4.0) was used for this purpose. The influence of the drying time of the samples on the phosphorescence intensity was studied using a 250 W infrared heat lamp and a hair-dryer operated at 600 W.Different samples, all prepared in the same way, were placed at different distances and for different time intervals. The best results were obtained when the hair-dryer was placed 7 cm above the sample for 3 min. The sample volume spotted on the solid support also influenced the phosphorescence intensity emitted by the analytes. To test this influence, different volumes of standard solution, ranging between 0.5 and 5.0 ml, and containing 1.0 mg ml21 of each analyte were tested.It was observed that the phosphorescence intensity increased when the sample volume was increased from 0.5 to 3.0 ml, and decreased very slowly when the sample volume was higher than 4.0 ml. As a consequence, 3.0 ml was adopted as the working volume in subsequent experiments. The influence of instrumental parameters such as delay and gate times, size of the excitation and emission slits and scan speed was also studied. It was found that the last parameter did not influence appreciably the phosphorescence intensity emitted by the phosphors but, as expected, td and tg exerted a notable influence.To optimise these parameters, different values of td (between 0.1 and 0.5 ms) and tg (between 4.0 and 14.0 ms) were checked, and it was found that 0.1 and 12.0 ms, respectively, produced the maximum phosphorescence intensity for both analytes. The optimum sizes of the excitation and emission slits were 2.5 and 5.0 nm, respectively.Finally, the quenching effect produced by the presence of atmospheric moisture and oxygen was studied using four samples, each containing 1.0 mg ml21 of each analyte. The phosphorescence intensities were measured at timed intervals, between 1 and 30 min, counted from the preparation of the samples. These samples were measured with and without quartz plates and with and without a dry N2 flow. The samples measured using quartz plates (with or without an N2 flow) showed phosphorescence intensities higher than those without quartz plates (Fig. 4). Hence if the phosphorescence intensity of the sample is measured before 8 min after its preparation, an N2 flow is not needed, thus simplifying the measurement process. Application of the PLS-1 model Optimising the data. A training set of 12 samples, randomly selected, was prepared to obtain the calibration matrix using the experimental data obtained fm single and binary mixtures of both pesticides at low, medium and high concentration levels.Table 1 gives the concentrations of each pesticide in each sample of the set. These concentrations were chosen avoiding correlation between the different samples, because this can produce underfitting in the PLS models. The spectra of these samples, obtained under the previously optimised experimental conditions described above, were recorded from 400 to 570 nm, with an interval of 0.5 nm between consecutive points, resulting in 341 experimental points per spectrum.PLS algorithms consist of two steps, calibration and prediction. In the calibration step we assumed that the Fig. 3 Influence of pH on the phosphorescence intensities of (a) CBL and (b) TBZ. Analyst, 1999, 124, 49–53 51concentration of the different analytes is related to the experimental measurements by the equation ck = R bk + e where ck is the vector of concentrations of analyte k in the N samples of the calibration set, R is the matrix of the instrumental measurements of the N samples in the different channels, e is the vector of the residuals of the concentrations that not are fitted in the model and the coefficients vector bk is obtained in the calibration step from the expression bk = R+ ck where R+ is the pseudo-inverse matrix of the matrix R.In PLS models, R+ is obtained from the decomposition of the original matrix R taking into account the information contained in the concentration vector ck of the calibration set.In the prediction step the concentration of the analytes (k = 1 for CBL, k = 2 for TBZ) are obtained from the matrix expression ck = rT bk + ek where rT is the vector of the instrumental measurements of the sample when this is measured in the j wavelength channels. Here, b coefficients were estimated by PLS-1 and the decomposition and regression of the matrix were performed separately for each component,28 taking into account the collinearity between the different wavelength channels. In our case, to select the optimum number of factors, the crossvalidation method32 was used and, as the training set was formed with 12 spectra, the calibration was performed on 11 of them, thus predicting the concentration of the excluded standard.The process was repeated 12 times (one for each standard) and the predicted and known concentrations were compared. The fitness of the PLS model was calculated by the prediction error sum of squares (PRESS), applied each time that a new factor was added to the model, using the F-statistic as a significance test.Applying the Haaland and Thomas criterion,33 seven factors (half of the standards + 1) were accepted as the maximum number of initial factors and the optimum number of factors was calculated for the first value of PRESS whose Fratio probability fell below 0.75. In this way, three factors were selected as optimum. These three factors should correspond to two analytes plus the background variation of the paper used as the solid support.Table 2 gives the estimated values of residual mean standard deviation absolute (RMSD): RMSD c c N i i i N = - = Â ( � )2 1 as an indicative value of the average error in the determination of each component and the values of R2: R c c c c i i i N i i i N 2 2 1 2 1 = - - = = Â Â ( � ) ( ) indicating that the data are fitted to a straight line. In order to determine the potential interferences produced for organic species and foreign ions usually present in waters, a systematic study of the effects produced by these chemicals was carried out.Standard solutions containing 1.0 mg l21 of CBL and 0.8 mg l21 of TBZ were spotted with the potentially Fig. 4 Influence of time on the phosphorescence intensity emitted by the analytes under different conditions. (A) Carbaryl and (B) thiabendazole. In both cases: (a) with quartz plates and N2 flow; (b) with quartz plates and without N2 flow; (c) with N2 flow and without quartz plates; (d) without quartz plates or N2.Table 1 Composition of the training set Standard CBL/mg ml21 TBZ/mg ml21 1 0.0 0.5 2 0.0 1.1 3 3.5 0.0 4 1.0 0.0 5 2.2 0.7 6 3.8 0.3 7 0.8 1.0 8 2.8 0.8 9 1.4 0.2 10 1.9 0.6 11 3.1 0.4 12 0.4 0.5 Table 2 Statistical parameters of the optimised matrix using the PLS-1 model Pesticide No. of factors RMSDa R2 b CBL 3 0.22713 0.96912 TBZ 3 0.028233 0.99315 a Residual mean standard deviation. b Square of the correlation coefficient.Table 3 Effects of foreign ions or organic species on the determination of CBL (1.0 mg l21) and TBZ (mg l21) Ions or species Tolerance level/ mg l21 Humic acids, dichlone, captan, atrazine, morestan, hexametazone, folpet, carbendazime 10.0 SO4 22, Cl2 , Ca(ii), Mg(ii), Cu(ii) 10.0 Al(iii), Be(ii), NO32 5.0 Fe(iii), PO4 32, warfarin 2.0 52 Analyst, 1999, 124, 49–53interfering species at a 10.0 mg l21 concentration and the concentrations of the analytes (CBL and TBZ) were determined using the PLS-1 method under the conditions established above. If interference occurred the concentration level of the foreign species was reduced until the error produced did not exceed ±5.0% in the determination of either of the two analytes.As can be seen in Table 3, the greatest interference was produced by warfarin and Fe(iii). Application of the model to real samples. In order to test the accuracy and applicability of the method, the optimised matrix obtained by the PLS-1 model was applied to the analysis of real samples of different kinds of water.Because the samples of waters did not contain pesticides (or the concentration levels were lower than the detection limit of the method), a recovery study was carried out. These samples (which were different from the 12 samples used to obtain the calibration matrix) were also analysed by using HPLC as a reference method.34 Table 4 gives the results obtained with the two methods.These data are the average values from three measurements of each sample by PLS-1 and HPLC. The results obtained by the two methods were compared statistically and the values of P are included in Table 4. Conclusions The use of the quartz plates during the measurement of the phosphorescence of the analytes improves the analytical process because a flow of inert gas is not needed. The PLS-1 method can be applied for the statistical treatment of the experimental spectrophosphorimetric data obtained from mixtures of phosphors in the solid phase.The use of a readily available filter-paper as a solid support allows for an inexpensive method, the main advantages of which are the sensitivity and selectivity that derive from the phosphorescence technique and the simplicity and speed that derive from the application of PLS-1. In all cases examined the results obtained by PLS-1 and HPLC were similar, as proved by the applied test. However, we are conscious of the need to carry out further work to demonstrate the applicability of the method to other samples under different conditions, and this is our intention.References 1 J. J. Vannelli and E. M. Schulman, Anal. Chem.,1984, 56, 1030. 2 S. Y. Su, E. Asafu-Adjaye and S. Ocak, Analyst, 1984, 109, 1019. 3 R. Q. Aucélio and A. D. Campiglia, Anal. Chim. Acta, 1995, 309, 345. 4 T. Vo-Dinh, E. Lue Yen and J. D. Winefordner, Anal. Chem., 1976, 48, 1186. 5 M. C.García-Alvarez Coque, G. Ramis Ramos, A. M. O’Reilly and J. D. Winefordner, Anal. Chim. Acta, 1988, 204, 247. 6 S. Scypinski and L. J. Cline Love, Anal. Chem., 1984, 56, 322. 7 S. Scypinski and L. J. Cline Love, Anal. Chem., 1984, 56, 331. 8 W. J. Jin and C. S. Liu, Anal. Chem. 1993, 65, 863. 9 Y. S. Wei, W. J. Jin, R. H. Zhu, C. S. Liu and S. S. Zhang, Talanta, 1994, 41, 1617. 10 S. Panadero, A. Gómez-Henz and D. Pérez-Bendito, Anal. Chem., 1994, 66, 919. 11 R. A. Paynter, S.L. Wellons and J. D. Winefordner, Anal. Chem., 1974, 46, 736. 12 R. P. Bateh and J. D. Winefordner, Talanta, 1982, 29, 713. 13 S. Su, D. L. Bolton and J. D. Winefordner, Chem. Biomed. Environ. Instrum., 1983, 12, 55. 14 D. L. McAleese and R. B. Dunlap, Anal. Chem., 1984, 56, 600. 15 A. D. Campiglia and C. G. de Lima, Anal. Chem., 1987, 59, 2822. 16 M. E. Díaz-García and A. Sanz-Medel, AnChem., 1986, 58, 1436. 17 M. C. Simeone, A. Gioia and A. D. Campiglia, Anal. Chim. Acta., 1994, 287, 89. 18 C. G. De Lima, M. M. Andino and J. D. Winefordner, Anal. Chem., 1986, 58, 2867. 19 H. Wold, in Research Papers in Statistics, ed. F. David, Wiley, New York, 1966, pp. 411–444. 20 S. Wold, in Food Research and Data Analysis, ed. H. Martens and H. Russwurm, Applied Science, London 1983. 21 S. Wold, M. Martens and H. Wold, in The Multivariate Calibration Problem in Chemistry by PLS, ed. A. Ruhe and B. Kagstrom, Springer, Heidelberg, 1983, pp. 286–289. 22 I. Durán-Merás, A.Muñoz de la Peña, A. Espinosa-Mansilla and F. Salinas, Analyst, 1993, 118, 807. 23 A. Espinosa-Mansilla, F. Salinas and I. de Orbe, Fresenius’ J. Anal. Chem., 1996, 354, 245. 24 A. Espinosa-Mansilla, F. Salinas and A. Zamoro, Analyst, 1994, 119, 1183. 25 A. Espinosa-Mansilla, A. Muñoz de la Peña and F. Salinas, Anal. Chim. Acta, 1993, 276, 141. 26 R. G. Brereton, Analyst, 1997, 122, 1521. 27 P. W. Araujo, D. A. Cirovic and R. G. Brereton, Analyst, 1996, 121, 581. 28 F.R. Burden, R. G. Brereton and P. T. Walsh, Analyst, 1997, 122, 1015. 29 C. Demir and R. G. Brereton, Analyst, 1998, 123, 181. 30 H. Martens and T. Næs, Multivariate Calibration, Wiley, Chichester, 1994. 31 L. F. Capitán-Vallvey, F. Ojeda, M. del Olmo, R. Avidad, A. Navalón and T. Vo-Dinh, Appl. Spectrosc., 1998, 52, 101. 32 M. Stone, J. R. Stat. Soc. B, 1974, 36, 111. 33 D. M. Haaland and E. V. Thomas, Anal. Chem., 1988, 60, 1193. 34 M. V. Dilna, R. E. Hall, J. D. Shamis and S. A. Whitlock, J. Chromatogr., 1984, 283, 383. Paper 8/07825D Table 4 Recovery study of mixtures of TBZ and CBL in tap water samples spotted with one or two components CBLb TBZb Sample No.a Taken/ mg l21 PLS-1/ mg l21 RSD (%) HPLC/ mg l21 RSD (%) P (%)c Taken/ mg l21 PLS-1/ mg l21 RSD (%) HPLC/ mg l21 RSD (%) P (%)c 1 0.00 (0.01) — — — — 1.00 0.93 3.1 0.96 4.5 18.4 2 3.00 3.15 2.0 3.10 3.0 27.4 0.50 0.45 4.7 0.47 3.2 8.0 3 2.50 2.64 2.5 2.59 3.4 21.1 0.60 0.63 2.9 0.60 4.9 8.7 4 3.50 3.41 2.3 3.48 3.4 24.2 0.40 0.40 2.9 0.41 3.9 19.0 5 1.00 1.09 3.2 1.08 4.6 64.0 0.60 0.59 2.5 0.61 5.3 29.5 6 0.00 (0.02) — — — — 0.40 0.39 4.1 0.40 3.1 16.1 7 1.00 1.01 3.5 0.98 5.1 20.8 0.40 0.39 4.8 0.41 2.7 32.0 8 3.80 3.55 6.0 3.60 4.1 60.2 0.00 (0.02) — — — — 9 2.00 1.97 7.1 2.00 4.5 64.8 0.50 0.48 5.5 0.49 3.7 41.8 10 1.50 1.41 3.4 1.44 4.1 20.4 0.30 0.30 3.1 0.32 5.2 6.6 11 1.80 1.72 4.6 1.75 5.2 31.2 0.00 (0.04) — — — — 12 2.50 2.41 1.9 2.44 2.4 23.2 0.85 0.87 2.0 0.86 3.9 32.0 a Samples 1–3, tap water, Granada City; samples 4–6, Genil river water (Granada); samples 7–9, mineral water from Lanjarón (Granada); samples 10–12, mineral water from Fontvella (Gerona). b The values in parenthesis are those measured by the method when this compound is not present. c P value of the comparison test. Analyst, 1999, 124, 49–53 53