Anal Bioanal Chem (2003) 375 : 443–449 DOI 10.1007/s00216-002-1695-y
O R I G I N A L PA P E R
S. L. C. Ferreira · W. N. L. dos Santos · M. A. Bezerra · V. A. Lemos · J. M. Bosque-Sendra
Use of factorial design and Doehlert matrix for multivariate optimisation of an on-line preconcentration system for lead determination by flame atomic absorption spectrometry Received: 2 July 2002 / Revised: 6 November 2002 / Accepted: 8 November 2002 / Published online: 29 January 2003 © Springer-Verlag 2003
Abstract A system for on-line preconcentration and determination of lead by flame atomic absorption spectrometry (FAAS) was proposed. It was based on the sorption of lead(II) ions on a minicolumn of polyurethane foam loaded with 2-(2-thiazolylazo)-5-dimethylaminophenol (TAM). The optimisation step was carried out using twolevel full factorial and Doehlert designs for the determination of the optimum conditions for lead preconcentration. The proposed procedure allowed the determination of lead with a detection limit of 2.2 µg L–1, and a precision, calculated as relative standard deviation (RSD), of 2.4 and 6.8 for a lead concentration of 50.0 and 10.0 µg L–1, respectively. A preconcentration factor of 45 and a sampling frequency of 27 samples per hour were obtained. The recovery achieved for lead determination in the presence of several cations demonstrated that this procedure has enough selectivity for analysis of environmental samples. The validation was carried out by analysis of certified reference material. This procedure was applied to lead determination in natural food. Keywords Lead determination · Doehlert matrix · On-line preconcentration · Polyurethane foam · FAAS
S. L. C. Ferreira (✉) · W. N. L. dos Santos · M. A. Bezerra · V. A. Lemos Instituto de Química, Grupo de Pesquisa em Química Analítica, Campus Universitário de Ondina, Universidade Federal da Bahia, 40170-290 Salvador, Bahia, Brasil e-mail:
[email protected] M. A. Bezerra · V. A. Lemos Departamento de Química e Exatas, Campus de Jequié, Universidade Estadual do Sudoeste da Bahia, 45200-190 Jequié, Bahia, Brasil J. M. Bosque-Sendra Department of Analytical Chemistry, Faculty of Sciences, University of Granada, Fuentenueva s/n, 18071 Granada, Spain
Introduction The optimisation of analytical procedures by multivariate techniques [1, 2, 3, 4, 5, 6] is faster, more economical and effective than the traditional “one-at-a-time” way because it makes it possible to understand circumstances that are not explained by the traditional approach, that is if there are interactions between the factors that influence the analytical response [7]. These advantages have been adequately reported in the literature [8, 9]. However, few preconcentration procedures have been optimised using experimental designs, namely an on-line preconcentration system for zinc determination by using inductively coupled plasma optical emission spectrometry (ICP–OES) [10], methods for simultaneous solvent extraction of several metals [11, 12] and a preconcentration procedure for molybdenum determination in seawater samples by (ICP– OES) [13]. Multivariate optimisation in analytical procedures generally implies two step: a) a preliminary evaluation, using factorial designs, in order to select the significant factors in the analytical procedure and b) an appropriate estimation of the real functional relationship (response function) between the analytical response and significant factors is obtained and then, the optimum values of the levels of these factors could be calculated. Methods to optimise analytical procedures can be classified in two groups: those that use an individual analysis of data to find the best experimental conditions of the system (i.e. Simplex optimisation) and methods or designs for response surface methodology (RSM) [2]. Simplex optimisation is a sequential procedure based upon a k-factor first-order model although the response function is not obtained [7]. RSM provides more and better information than the first-order designs because it not only determines the influence of the factors but also enables one to obtain the response function. This methodology makes use of the method of steepest ascent [5], a combination of first-order designs and central composite designs, or different second-order designs (i.e. Box–Behnken or Doehlert designs) [14] to obtain the response function. Doehlert designs offer advan-
444
tages in relation to more frequently used designs such as central composite or Box–Behnken. They need fewer experiments, which are more efficient and can moved through the experimental field, and points obtained from a previous design could be used in later designs when they are adjacent [15]. In addition, the quantity of the levels related to each factor can be selected in order to obtain more information about significant or problematic factors [16]. In this paper, factorial design and Doehlert matrix were used for optimisation of the experimental variables of an on-line preconcentration system for lead determination in natural food samples by using flame atomic absorption spectrometry (FAAS). It was based on solid-phase extraction of lead(II) ions as 2-(2-thiazolylazo)-5-dimethylaminophenol [17] (TAM) complexes onto polyurethane foam. In our research group, polyurethane foam has been used as a sorbent in on-line preconcentration systems for the determination of zinc [18], cadmium [19] and lead [20].
Experimental Instrumentation A Varian Model SpectrAA 220 (Mulgrave, Victoria, Australia) flame atomic absorption spectrometer was used for the analysis. The lead hollow cathode lamp was run under the conditions suggested by the manufacturer (current 4.0 mA). Also, the wavelength (283.3 nm), the bandwidth of the slit (0.5 nm) and burner height (13.5 mm) had conventional values. The flame composition was acetylene (flow rate 2.0 L min–1) and air (flow rate 13.5 L min–1). Nebulizer flow rates were 5.0–6.0 mL min–1. An Alitea C-6 XV (Stockholm, Sweden) peristaltic pump furnished with Tygon tubes was used to propel all solutions. A Rheodyne 5041 (Cotati, California, USA) model four-way manual valve was used to select preconcentration/elution steps. All connections were made using fittings, unions and tees made of plastic and PEEK materials. The manifold was built up with PTFE tube of 0.5-mm bore.
terwards, the column was washed with a 10% (w/v) sodium hydroxide solution until the purple effluent became colourless to remove the excess of TAM. Then, the column was washed with 5% (v/v) hydrochloric acid solution and deionised water, respectively, using the same flow rate. Washing with nitric acid was necessary to prevent any metal contamination. Sample preparation For relevant details the reader is referred to the article by Ferreira et al. [21]. The certified reference material Citrus Leaves NIST 1572 and leaf samples were decomposed according to the following procedure: 0.4 g of material was treated with 4.0 mL of nitric acid 1:1 (v/v) and kept overnight in a Teflon vessel. Afterwards, the Teflon vessel was closed and put into a pressurized digestion system. The thermal heating was carried out in a stove at 110 °C for 24 h. After cooling to room temperature these solutions were adjusted to pH 7.7 with a 10% (w/v) sodium hydroxide solution and a Tris buffer solution. The volume was then made up to 100 mL. On-line preconcentration system The diagram of the on-line preconcentration system is shown in Fig. 1. The flow system was carried out using two peristaltic pumps fitted with Tygon tubes, a four-way valve and a minicolumn packed with polyurethane foam loaded with TAM, coupled to a flame atomic absorption spectrometer (FAAS). A sample solution (S) containing between 0.0 and 150.0 µg L–1 of lead was kept at pH 7.7 with a Tris buffer solution and was pumped at 7.0 mL min–1 through a minicolumn that retained the cation. Lead(II) ions are retained by chemical sorption as the TAM complex and the remaining solution was discharged (W). By switching the injection valve a stream of 1.90 mol L–1 hydrochloric acid (E) that flows at 5.5 mL min–1 displaces the lead ions. This eluate was taken direct to the nebulizer-burner system of the flame atomic absorption spectrometer. Analytical signals (integrated absorbance) were measured as peak height by using instrument software. It was read three times and averaged. The minicolumn was not reconditioned at the end of each cycle, since sample was buffered before preconcentration. This strategy has
Reagents All reagents were of analytical grade unless otherwise stated. Ultrapure water was obtained from an EASYpure RF (Barnstedt, Dubuque, IA, USA). Nitric and hydrochloric acid were of Suprapur quality (Merck). Laboratory glassware was kept overnight in 10% nitric acid solution. Before use the glassware was rinsed with deionised water and dried in a dust-free environment. Lead solution (30.0 µg L–1) was prepared by diluting a 1,000 µg L–1 lead solution (Merck) with a 1% hydrochloric acid solution. TAM solution (0.05%) was prepared by dissolution of 0.10 g of 2-(2-thiazolylazo)-5-dimethylaminophenol in 200 mL of ethanol (Merck). Tris buffer solution (pH 7.7) was prepared by dissolving 12.10 g of tris(hydroxymethyl)aminomethane (Merck) in 1,000 mL of deionised water and pH was adjusted with hydrochloric acid. Column preparation For relevant details the reader is referred to the article by Lemos and Ferreira [20]. The 0.05% (w/v) TAM solution was percolated through a laboratory-made cylindrical minicolumn with 4.50-cm length and internal diameter of 4.0 mm containing about 0.10 g of polyurethane foam at a flow rate of 5.17 mL min–1 for 15 min. Af-
Fig. 1 System of on-line preconcentration. Schematic diagram of the flow system used for the preconcentration and determination of lead by FAAS. P1 (5.80 mL min–1); P2 (5.5 mL min–1); P peristaltic pump; C polyurethane foam/TAM minicolumn; V four-way valve; FAAS flame atomic absorption spectrometer; W waste; A four-way valve in the preconcentration step; B four-way valve in the elution position
445 been recommended [22, 23] to simplify the on-line system (avoiding an additional line with buffer solution) and to increase the concentration efficiency (CE). This strategy causes a small loss of analyte, but it does not deteriorate the preconcentration system, since this loss should always be reproducible for both standards and samples. Achieved sampling rate was 27 samples per hour for a preconcentration time of 120 s and elution time of 15 s. During all optimisation process the flow system was operated in a volume-base mode, but the procedure proposed was based in a time-base mode. Optimisation strategy The optimisation step was carried out using a two-level full factorial and Doehlert designs [24]. All the experiments were carried out in duplicate, with a random order, using 10.0 mL of lead working solution containing 30.0 µg L–1. Five variables (sampling flow rate, elution flow rate, elution concentration, buffer concentration and pH) were regarded as factors. An important aspect in the optimisation procedure is the choice of the most suitable analytical response. Therefore, in this case, the factorial design was evaluated in two forms: first using the analytical signal (absorbance) and also using the analytical parameter “sensibility efficiency” in order to compared the analytical data obtained. It was then possible to choose the best analytical response to carry out the estimation of the response function. The experimental data were processed by using the Statistic computer program [25]. Sensibility efficiency The parameter sensibility efficiency (SE) was proposed in a previous paper [26] and is defined as the analytical signal obtained for an on-line enrichment system for a preconcentration time of 1 min. It can be used to demonstrate that the concentration efficiency [23] increases with increasing sampling flow rate for this system. This parameter can be calculated from the equation SE= AS/t, where AS is the analytical signal and t is the preconcentration time in minutes. Considering that t=sample volume(SV)/sampling flow rate (SR): SE =(AS × SR) SV (1) This demonstrates the relationship between the sensibility efficiency and the sampling flow rate. This parameter can be useful for comparing the efficiency (as analytical signal) of two different analytical systems or different experimental conditions of the same system. Lagrange’s criterion The Lagrange’s criterion [10, 13] is used to check the existence of a critical point in the second-order function (Y=α+βA+δB+εA2+ γB2+κAB) estimated with the Doehlert matrix and it is based on the calculation of the Hessian determinant of the response function: 2 H ( A, B) = δ 2 Y δ A 2 δ 2 Y δ B 2 − δ 2 Y δ Aδ B (2) The critical point (ao,bo) is maximum if H(ao,bo) >0 and δ2Y/ δA2(ao,bo) 0 and δ2Y/δA2(ao,bo) >0. A point saddle exists if H(ao,bo)
