Development of multivariate calibration method for simultaneous determination of nickel, lead and zinc in tap water

Conventional spectrophotometric methods for simultaneous determination of nickel, lead and zinc in forms of complexes with a reagent is not feasible due to the overlap of their absorption spectra. A multivariate calibration method was used to overcome this problem. In this study, the calibration model was constructed based on absorption spectra of 30 mixture standards in the range from 490 to 600 nm. Factors influencing experimental results such as amount of reagents, pH, and color development time were optimized. The standard calibration ranges for determination of nickel, lead and zinc were found at 0.5-5 ppm. The method was applied for determination of these ions in tap water samples at ppm level, with recoveries (and RSD) of nickel, lead and zinc were 103.3 % (3.0 %), 74.9 % (11.5 %) and 104.6 % (4.6 %), respectively.


INTRODUCTION
Nickel, lead and zinc co-exist in many samples. Several techniques such as XFS [1], polarography [2], AAS [3], ICP-OES [4] were used to determine these metal ions. Among these techniques, spectrophotometry was commonly used in thanks to its low instrumental investment and the ease of performance. However, direct determination of the metal ions without prior separation is impossible due to the spectral overlap. Real samples having complicated components usually have overlapped absorption spectra which decreases accuracy. This research depicted a method for simultaneous determination of nickel, lead and zinc by combining spectrophotometry with multivariate calibration technique to increase the speed and accuracy of the analysis. Complexes of these ions with xylenol orange (XO) in synthetic and tap water samples were analyzed by spectrophotometry without any prior treatment.
Achieved data was treated by partial least square method (PLS) to build a correlation vector to predict concentration of each component in the samples from their spectral signals. The data were in the form of matrices. Rows of X and Y matrices were absorbance at every wavelength of a standard mixture and the corresponding concentrations of the ions, respectively [5]. These two matrices can be described as equations (1) and (2).
where T, P, and E are X-scores, X-loadings, and X-residuals, respectively. And U, Q, and F are the same coefficients for Y matrix. T and U were correlated as in equation (3) [ bf is the regression coefficient for the f latent variable. After replacing uf in Y, the new equation . To calculate the concentration of each components of new samples, new scores T* replace T in the equation (4) as followed: The more detailed calibration set was built (or more standard mixtures were used), the higher its predictive ability would have.

Materials and equipment
Lead, nickel and zinc standard mixture solutions were prepared from their individual stock solutions which were purchased from Merck (Germany). 1000 ppm stock solutions of these metals were prepared in 0.5 % nitric acid and stored at 7 o C; from which working solutions were prepared daily with distilled water. Nitric acid, acetic acid, sodium acetate, and xylenol orange (XO) from Guangdong Guanghua Co.
(China) were of analytical grade.
A Shimadzu AA-6650 Atomic Absorption Spectrophotometer equipped with an Auto Sampler ASC-6100. Standard solutions were used to confirm the content of metals in samples.

A
Shimadzu UV-1800 UV-VIS Spectrophotometer controlled by UV Probe software and a 1.00 cm glass cells was used in this study. pH of buffers were adjusted by Schott Instrument Lab850 pH meter and the data were treated by SIMCA P-11 software (Umetrics, Sweden).

Spectrophotometric condition optimization
Prior to multicalibration study the conditions for complex formation were investigated. Times for full color development, pH and concentration ratio of metal ions to XO were optimized. Complexes of three analytes and XO were used for all investigations of pH, reaction time, and ration of M:XO.  A calibration set consisting of 30 standard mixtures were prepared varying the metal concentrations from 0.5-5.0 ppm. We encoded the lowest concentration in each component as 1, the middle as 2 and the highest as 3, in the order of Ni 2+ , Pb 2+ and Zn 2+ (Table 1). In order to validate the developed method, 9 samples of these ions were prepared in distilled water ( Table 2).

Atomic Absorption Spectrometry (AAS) confirmation
The concentrations of metal ions of the mixtures were confirmed by AAS. The absorbance was measured at 232, 217 and 213.9 nm for nickel, lead and zinc, respectively.

Spectrophotometric analysis conditions
Spectra of the M-XO complexes show that maximal absorptions were observed at 582, 570 and 561 nm for Ni-XO, Pb-XO and Zn-XO, respectively ( Fig. 2-I). In addition, Pb-XO complex had the lowest sensitivity while Ni-XO had the highest one.
In the investigation of pH, reaction time and concentration ratio, measurements of the mixture were carried out at 582 nm. It was found that pH of 5.5, reaction time of 55 min and the concentration ratio M:XO = 1:2 were the best conditions regarding sensitivity and repeatability. Unfortunately, it is impossible to have this ratio constant for samples with unknown M 2+ levels.
Therefore from now on a fixed amount of XO was used for complexation and this resulted in the ratio varied from 1:2-1:20 depending on the M 2+ level.

Calibration model
Partial least square (PLS) was used to treat the spectrophotometric data. It should be noted that only spectra in the range from 490-600 nm were used for this work since XO absorbed strongly at lower wavelength and very low signals of M-XO at wavelength higher than 600 nm.
The score plot offered by PLS indicated the distribution of the standard mixtures (Fig. 3). It is obvious that Ni 2+ had the strongest influence on the distribution and samples with increasing level of Ni 2+ were distributed along t1 from left to right, while samples with increasing Zn 2+ levels were separated by t2 from bottom to top. The situation of Pb 2+ was the same as Zn 2+ but to a lesser extent. In other words, the influence order of the ions on the spectra of their mixtures decreased from Ni 2+ , Zn 2+ to Pb 2+ . A more careful look on the score plot one can easily found that at higher concentration of Ni 2+ and Zn 2+ , the much weaker influence of Pb 2+ level in differentiating the samples as emphasized by open circles in the score plot, especially at high level of nickel and zinc. It illustrated that the stronger absorption intensities were, the more significant model effect of analytes was. These observations could lead to higher errors of predicted Pb 2+ level by the model.
The model fitness, R 2 , and prediction ability, Q 2 , were of 0.979 and 0.958, respectively showing that in general the model was of good quality and can be used to predict the ion levels [6].

Application to water samples prepared in pure water and in tap water
The predictive ability of model was tested using 9 samples prepared in distilled water ( Table   2). It illustrated that the proposed procedure was successfully applied for the assay of nickel, lead and zinc simultaneously in synthetic samples at ppm level. The average RSDs were acceptable, 3.0, 11.5, and 4.6 % for nickel, lead and zinc, respectively. The satisfactory recoveries were achieved with Ni 2+ and Zn 2+ . However, low level of Pb 2+ in synthetic samples still had high errors and uncertainty due to the low sensitivity of its complex in comparison to the other two ions.

CONCLUSIONS
The method for simultaneous determination of nickel, lead and zinc ions based on their complexes with xylenol orange was developed. By using spectrophotometric method combined partial least square as a multivariate calibration technique, it was possible to obtain a model from absorbance signals. Absorption spectra in the 490 to 600 nm range of 30 different standards of which concentrations found from 0.5 to 5 ppm were used to create a model. The method was applied for determination of nickel, lead and zinc ions in synthetic samples without carrying out costly and time consuming for sample treatment. However the requirement of equal sensitivity for all components is the limitation of the method.