ranilic acid [27] or chlorate [28] are susceptible to inter ... electrochemical reduction of chlorate in an acidic medi um. ..... The fact that iron, copper, lead and zinc.
ЖУРНАЛ АНАЛИТИЧЕСКОЙ ХИМИИ, 2010, том 65, № 5, с. 532–537
ОРИГИНАЛЬНЫЕ СТАТЬИ УДК 543
CATALYTIC ADSORPTIVE STRIPPING VOLTAMMETRY DETERMINATION OF ULTRA TRACE AMOUNT OF TUNGSTEN USING FACTORIAL DESIGN FOR OPTIMIZATION © 2010 г. K. Zarei, M. Atabati, R. Shoari School of Chemistry, Damghan University of Basic Sciences Damghan, Iran Received 12.08.2008; in final form 31.08.2009
A highly sensitive procedure is presented for the determination of ultratrace concentration of tungsten by catalytic adsorptive stripping voltammetry. The method is based on adsorptive accumulation of the tungstenpyrocatechol vi olet complex onto a hanging mercury drop electrode, followed by reduction of the adsorbed species by voltammetric scan using differential pulse modulation. The reduction current is enhanced catalytically by chlorate. The influence of variables was completely studied by factorial design analysis. Optimum analytical conditions for the determina tion of tungsten were established. Tungsten can be determined in the range 0.06–12.0 ng/mL with a limit of detec tion of 0.02 ng/mL. The influence of potentially interfering ions on the determination of tungsten was studied. The procedure was applied to the determination of tungsten in one sandwich polyoxometalate and some synthetic sam ples similar to alloy compounds with satisfactory results.
Tungstate was found to correct hyperglycemia in ani mal models of diabetes when administered in drinking fluid and represented a potential medicine for treatment of diabetes [1]. However, tungsten as a trace element, is toxic to peo ple and animals, as 5 μg/kg of tungsten led to the death of animal embryos [2]. On the other hand, the accurate and precise determination of tungsten, which exists in steel and alloys, can alter their strength, toughness, and abra sion – resistance, which is extremely important in the in dustry [3]. In recent years, the techniques of adsorptive cathodic stripping voltammetry (ACSV) and catalytic adsorptive stripping voltammetry (CASV) have been developing con siderably [4–14]. The ACSV technique is based on ad sorptive accumulation of the desired species (or its com plex) on a hanging mercury drop electrode (HMDE) fol lowed by electrochemical reduction of the adsorbed species. In CASV, the reduction current can be enhanced greatly if the reduction product of a voltammetric scan is chemically reoxidized in the presence of an oxidant on a time scale that is fast relative to the scan rate. The detec tion limit of CASV is very low, and the method has excel lent selectivity, sensitivity, accuracy and low cost of instru mentation. Thus it is very suitable for ultratrace analysis. Various methods have been developed for the determi nation of tungsten including spectrophotometry [2, 15– 18], flame atomic absorption spectrometry (FAAS) [19], inductively coupled plasma atomic emission spectrome try (ICPAES) [1, 20–21], inductively coupled plasma mass spectrometry (ICPMS) [22], Xray fluorescence (XRF) [23], neutron activation analysis (NAA) [24]. The dependence on a nuclear reactor and radiochemical lab
oratories limit the applicability of NAA and XRF to rou tine analysis. Furthermore, these methods, however, have inadequate sensitivity and excessively high determination limits or involve troublesome and time consuming pre concentration steps such as co precipitation or extraction. Several papers have been published to date on the de termination of tungsten by stripping voltammetry. Mala khova et al. measured 0.028 mass% of tungsten in carbon steel by anodic stripping voltammetry at an ultratrace graphite electrode with antipyrine or 4dimetylaminoan tipyrine as a complexing agent [25]. This method, howev er, requires the laborious preparation of the graphite elec trode prior to each measurement in order to achieve re producible results. Shumilova et al. proposed adsorptive stripping voltammetry at a carbon paste electrode modi fied with 8quinolinethiol and dimethyl sulfoxide, but the method provided neither adequate sensitivity nor suffi ciently low determination limit [26]. Highly sensitive methods based on the catalytic reduction peak of chlo ranilic acid [27] or chlorate [28] are susceptible to inter ference from many foreign elements. The new method described in this paper is based on the adsorptive accumulation of tungsten (W) – pyrocate chol violet (PCV) complex and its catalytic effect on the electrochemical reduction of chlorate in an acidic medi um. It is very sensitive and precise for determination of low levels of tungsten and is free from interference of com mon interfering ions such as iron. The influence of variables on the peak current was studied by factorial design—a powerful method for de signing an experiment. The theory of full and fraction fac torial design has been widely described elsewhere [29].
532
CATALYTIC ADSORPTIVE STRIPPING VOLTAMMETRY DETERMINATION
ЖУРНАЛ АНАЛИТИЧЕСКОЙ ХИМИИ
том 65
Table 1. Selected levels for each factor Level H2SO4, M
PCV, M
KClO3, M
Accumulation potential E, V
–1
0. 010
6.0 × 10–7
0.10
0.20
0
0.045
1.3 × 10–6
0.25
0.05
+1
0.080
2.0 × 10–6
0.40
–0.10
Data were processed with a specially written MAT LAB program (the math works Inc., version 7.0). The op timum values of parameters were calculated with MAT LAB programming. RESULTS AND DISCUSSION Preliminary investigations. Tungsten can form a com plex with the PCV. The PCV or pyrocatechol sulfonaph thalein is a dye in the triphenylmethane series, and it is a sensitive reagent for the tungsten. Fig. 1 shows differential pulse voltammogram for W(VI)PCV in 0.04 M H2SO4. As can be seen in Fig. 1b, W(VI)PCV complex is reduced at –470 mV vs. Ag/AgCl reference electrode. In cyclic voltammogram, Addition of 0.4 M of potas sium chlorate causes an increase of the tungsten reduction current in comparison to the W(VI)PCV complex (Fig. 2b), where in the absence of chlorate the reduction current is rather small (Fig. 2a), this suggesting a catalytic – process. Also analysis of a W(VI)PCVClO 3 system (with high amount of chlorate) at various scan rates (with 2500 b 2000 1500
ΔI, nA
EXPERIMENTAL Reagents. All solutions were prepared with doubly dis tilled water. Chemicals used were of analytical grade and were purchased from E. Merck. A solution of 1.0 × 10–3 M of PCV was prepared by dis solving 1.92 × 10–2 g of the reagent in water in a 50 mL vol umetric flask. A solution of 1.48 × 10–4 M PCV was pre pared by dilution of the stock solution with water. A 1000 μg/mL solution of W(VI) was prepared by dis solving 0.1794 g of sodium tungstate dihydrate in water and diluting to 100 mL; more dilute solutions were pre pared by diluting this solution with water. Potassium chlorate solution (0.5 M) was prepared by dissolving 15.318 g of KClO3 in water in a 250 mL volu metric flask. Sulfuric acid solution was prepared by dissolving 5.5 mL of conc. sulfuric acid in water and diluting to 100 mL. Apparatus. Experiments were conducted by using a PAR (Princeton Applied Research) model 394 Polaro graphic analyzer equipped with a model 303A electrode system (EG&G). A conventional three electrode system, comprising a medium–sized hanging mercuric drop elec trode, with a surface area of 1.8 mm2, a platinum wire counter electrode and an Ag/AgCl (in saturated with KCl) reference electrode was used in all experiments. The re ported potentials were referred to the Ag/AgCl electrode. Solutions were deoxygenated with high purity nitrogen for 4.0 min prior to each experiment preformed under a ni trogen atmosphere. Procedure. The sample solution (10.0 mL), contain ing 1.48 μM PCV, 0.4 M KClO3, 0.04 M H2SO4 and ap propriate volumes of W(VI) was transferred into the volta mmetric cell. After that, the resulting solution was diluted to the mark with water. In 2.0 min after addition of W(VI), the stirrer was switched on and the solution was purged with nitrogen gas for 4.0 min. The accumulation potential (–0.025 V) was applied to a fresh HMDE for 5 min whilst stirring the solution. Following the accumulation period, the stirrer was stopped and after 5 s the voltammogram was recorded by applying a negativegoing differential pulse scan from 0.0 to –0.9 V versus the Ag/AgCl refer ence electrode under the following instrumental condi tions: pulse height, 50 mV; scan rate, 33.3 mV/s. The peak current for tungsten was measured at about –470 mV and recorded as a function of tungsten concentration. A blank solution without tungsten was used to obtain the blank peak current. The experiment was designed by full factorial design [22]. We think that the effect of acid, PCV and chlorate concentrations and accumulation potential on the re sponse (peak current) is two orders. Therefore, three lev els for each factor were selected (Table 1), and 90 exper iments (34 + 9 replicates) were run (factorial levels in replications: acid concentration, –1; chlorate concen tration –1; PCV concentration and accumulation po tential vary at –1, 0 and +1 levels).
533
№5
1000 500 a 0 0
–0.2
–0.4
–0.6
–0.8
–1.0
E, V
Fig. 1. Differential pulse voltammograms for 0.04 M H2SO4, 1.48 × 10–6 M PCV, 0.4 M chlorate, pulse height 20 mV and 5 min accumulation at –0.025 V. (a): without tungsten, (b) with 50 ng/mL of W(VI). 2010
534
ZAREI и др.
1000
500 d 400
800 ΔI, nA
c
ΔI, nA
600
300 b 200
400 100
b
a
200 0 –0.30 –0.35 –0.40 –0.45 –0.50 –0.55–0.60 E, V
a 0 –0.2
–0.3
–0.4
–0.5 –0.6 E, V
–0.7 –0.8 Fig. 3. Differential pulse voltammograms of W(VI)PCV
Fig. 2. Cyclic volltammograms of 6.0 ng/mL W(VI) in the 0.04 M H2SO4 containing 1.48 × 10–6 M PCV (a) without chlorate (b) with 0.4 M chlorate. Accumulation potential –0.025 V, accumulation time 5 min, scan rate 33.3 mV/s.
ClO 3 , recorded without an accumulation period (cur ve a), and after 2 min (curve b), 3 min (curve c) and 5 min (curve d) of accumulation at –0.025 V.
out adsorption time) shows a linear relationship between Ip/υ1/2 and υ–1 (Ip is cathodic current, nA, and υ is scan rate, mV/s) with a regression coefficient of 0.9990 (and slope of 4190) involving mechanism of EC'. On the basis – of the above results, the catalytic mechanism of ClO 3 in the medium containing W(VI) and PCV is a typical elec trochemical – chemical process. On the other hand, an Ep–log (scan rate) analysis of a series of linear scan voltammograms of (W(VI)PCV – ClO 3 ) gave a slope of 95 mV indicating a one–electron transfer reduction process [30]. A possible reaction mechanism is that W(V) (which is generated at the electrode surface upon the reduction of pyrocatechol violet complex of W(VI)) reduces chlorate whilst the thus generated W(VI) is rereduced at the elec trode to W(V), thus contributing repeatedly to the reduc tion current. Fig. 3 shows differential pulse voltammograms of – W(VI)PCVClO 3 , recorded without an accumulation period (curve a), after 2 (curve b), 3 (curve c) and 5 min (curve d) accumulation at –0.025 V. This increase in re sponse shows that W(VI)PCV has high adsorptive char acteristics at the mercury electrode. On the other hand, there is a linear relationship between Ip and υ (with 5 min accumulation period) that confirms high adsorptive char acteristics of W(VI)PCV complex. Factorial design and optimization. In this experiment, the effects of chemical parameters such as acid, pyrocate chol violet and chlorate concentrations and instrumental
parameters such as accumulation potential and the inter action of these parameters on peak current by full factorial design were studied. A preliminary study shows that with increasing accu mulation time to 5 min peak current (ΔIp) increases rap idly, and after that it increases slowly. Therefore, we have considered to choose 5 min as accumulation time. Peak current (ΔIp) also increases with increasing of scan rate, therefore, we have selected constant value of 33.3 mV/s as the scan rate. In making the model, the response (peak current) was written as a function of acid, PCV, chlorate concentra tions and accumulation potential and all possible interac tions. The coefficients of these parameters were obtained by multiple leastsquares regression. For each parameter and interaction, the parameter coefficient, the standard error, the t value for the null hypothesis (H0), and the cor responding P value were calculated (Table 2). A program written in MATLAB was used to perform the calculations. The null hypothesis states that the value of the parameter coefficient is zero. The P value is the probability that a pa rameter coefficient can be zero. If the P value for each pa rameter is greater than 0.05, the parameter has no signifi cant effect in the model (confidence limit 95%), and can be eliminated (the value of the parameter coefficient is taken to be zero). But the model was kept hierarchical; i.e., a factor with a P value greater than 0.05, but with the higher order of this factor having a P value less than 0.05, was not removed. The analysis of variance (ANOVA) of the model is shown in Table 3. The F value of the model (regression)
–
ЖУРНАЛ АНАЛИТИЧЕСКОЙ ХИМИИ
том 65
№5
2010
CATALYTIC ADSORPTIVE STRIPPING VOLTAMMETRY DETERMINATION
describes that the regression at the confidence limit of 99.9% (p = 0.0001) is significant, and the model has a cor relation coefficient 0.9825. The root mean square error is 5.7288, which corresponds to a relative error of 7.1% in the measured peak currents. There is no lack of fit in the confidence limit 95% (calculated Flack of fit is less than the critical value). In Fig. 4 residuals are plotted versus num ber of experiments, which predicts that we have hemosce dastic error in our experiment. In Fig. 5, the current cal culated from the model is plotted versus the measured current. The correlation coefficient for the plot is 0.9825, which indicates good performance for the model. From the model, a program in MATLAB was written to calculate the optimum value for the each parameter. Within the confidence limit of 95%, the model shows three significant secondorder interactions, three signifi cant thirdorder interactions, and four significant qua dratic effects. As shown is Table 2, chlorate concentration has a significant effect on the response. The later signifi cant effect is the PCV concentration. The effect of accu mulation potential is smaller than acid, PCV and chlorate concentrations. In quadratic effects, the effect of acid concentration is significant because this parameter has important effect on the reaction, and the negative sign of the coefficient implies that the peak current has the max imum value with respect to this parameter. In secondor der interaction, the interaction of PCV concentration with chlorate concentration is very important, as for low chlorate concentrations the variation of current with re spect to different PCV concentrations is low, but for high er chlorate concentrations these variations are high. Among other second order interactions, the interaction of acid concentration with chlorate concentration and PCV concentration are in the second and third position, re spectively. Between thirdorder interactions, the interac tion of second power of PCV concentration with chlorate concentration is important. The optimum values for acid, PCV, chlorate con centrations and accumulation potential calculated with MATLAB programming are 0.04, 1.48 × 10–6, 0.4 M and –0.025 V, respectively. Figures of merit. Calibration curve, detection limit and reproducibility of the method. Under the optimum condi tions achieved by factorial design analysis, two regions for calibration were obtained. The first calibration curve was obtained in the range of 0.06–0.20 ng/mL and the second in the range of 0.20–12.0 ng/mL. The equations of the calibration curves in two regions are: ΔIp = 13.37 + 2830.62(W(VI))(ng/mL) (R2 = 0.9603, n = 7) and ΔIp = 487.28 + 401.48(W(VI))(ng/mL) (R2 = 0.9972, n = 8). A detection limit (cLOD = 3sb/m, where sb is the stan dard deviation for 5 replicates determination of the blank signal and m is the slope of the calibration curve) of 0.02 ng/mL for W(VI) was obtained. The relative standard ЖУРНАЛ АНАЛИТИЧЕСКОЙ ХИМИИ
том 65
№5
535
Table 2. Selected parameters Parameter Intercept
Parameter value
SE t for H0
P
114.48
2.05
55.94 0.0001
Acid
7.97
0.85
9.41 0.0001
PCV
13.39
0.85
15.81 0.0001
29.98
1.53
19.54 0.0001
–5.03
0.79
–6.33 0.0001
–
ClO 3 E
2
–27.74
1.39 –19.96 0.0001
2
–13.64
1.39
–9.81 0.0001
( ClO 3 )
–7.24
1.58
–4.59 0.0001
E2
–8.56
1.58
–5.43 0.0001
5.50
1.03
5.35 0.0001
Acid × ClO 3
–
–6.18
1.01
–6.12 0.0001
–
7.23
1.01
7.16 0.0001
–
–4.75
1.66
–2.87 0.0021
–
–8.53
1.66
–5.15 0.0001
–
4.68
1.23
3.82 0.0001
(Acid)
(PCV)
– 2
Acid × PCV
PCV × ClO 3
ClO 3 × (Acid)2 ClO 3 × ( PCV)2 ClO 3 × PCV × Acid
Table 3. The analysis of variance (ANOVA) of the model Source
df
Sum of squares
Total
89
8.4530 × 104
Treatment (regression)
14
8.1609 × 104 149.6767
Residual
75
2.9209 × 103
Lack of fit
66
2.6459 × 103
Pure error
9
275
R2
0.9654
RMSE
5.7288
Mean
80.44
C.V.
7.12
2010
F
1.3120
ZAREI и др. Calculated current ⎯Measured current
536 100
50
0
–50
–100 0
20
40
60 80 100 120 Mesured current, nA
140
160
Fig. 4. Residuals versus measured current.
160 y = 0.9654x + 2.7797 R2 = 0.9654
Calculated current, nA
140 120 100 80 60 40 20 0 0
20
40
60 80 100 120 Mesured current, nA
140
160
Fig. 5. The current calculated from the model versus the measured current.
deviations for five replicate measurement of 2 and 4 ng/mL are 5.8 and 5.2%, respectively. Interference study. The effects of more than 38 ions on the determination of 2.0 ng/mL of W(VI) were stud ied with the optimized conditions described above. The tolerance limit is defined as the foreignion concentra tion causing an error smaller than 3.0% for the determi nation of 2.0 ng/mL of tungsten. The results are present ed in Table 4. The results indicate that many of ions do not interfere. Only Mo(VI) interferes at a level similar to that of tungsten. The fact that iron, copper, lead and zinc do not interfere is of particular significance, considering their major interference in most of the previously report ed systems. Real sample analysis. The proposed method has been successfully applied to the determination of tung sten in the synthetic samples of compositions analogous to some alloys [31]. The results are shown in Table 5.
This method has also been applied for the determina tion of tungsten in a sandwich polyoxometalate (NH4)10[P2W18Cd4(H2O)2O68] ⋅ 27H2O). For this pur pose, 0.0169 g of the polyoxometalate were dissolved in 10.0 mL of 20% NaOH, boiled for 10.0 min, neutral ized, diluted to 100 mL and used for the determination of tungsten by standard addition method. The data ob tained for samples spiked with tungsten have shown good recoveries (Table 6). *** The present study demonstrates that catalytic strip ping voltammetric determination of tungsten in the pres ence of PCV and chlorate is excellent for determination of ng/mL concentrations of tungsten in real samples, be cause of high sensitivity, large dynamic range, simplicity
ЖУРНАЛ АНАЛИТИЧЕСКОЙ ХИМИИ
том 65
№5
2010
CATALYTIC ADSORPTIVE STRIPPING VOLTAMMETRY DETERMINATION Table 4. Maximum tolerable concentration of interfering species (optimum reagentconcentration and 5 min accu mulation at –0.025 mV, W(VI) concentration 2.0 ng/mL) Species
Tolerance limit (w/w)
Hg2+, K+, Fe3+, Fe2+, Ba2+, Na+, 2–
+
2–
Cl–, Cr 2 O 4 , NH 4 , CO 3 , 2–
Rh(III), Mg2+, Ca2+, Zn2+, SO 3 , –
2–
50000*
Pb(II), NO 3 , Ce(IV), C 2 O 4 , Mn2+, U(VI), V(V), Cu2+,Os(VIII), 2–
SO 4 , Co2+ –
40000
IO 3
2–
CN–, S 2 O 3 , Pd2+, Si(IV)
12500
As(III)
5000 2–
Sn2+, C 2 O 4
1250
Ru(III)
200
Mo(VI)
1
* Maximum concentration tested.
Table 5. Determination of tungsten in alloys sample Alloy I II III IV
Real amount, % Found amount, % 6.40 4.75 6.00 2.00
5.4 × 0.9 4.1 × 0.4 5.4 × 0.4 1.9 × 0.2
Composition of alloys: I) C: 1.28%, Co: 8.5%, Cr: 4.2%, Mo: 5%, V: 3.1%, Fe: 71.52%; II) C: 0.4%, Cr: 5.25%, Fe: 87%, Mn: 0.25%, Mo: 1.35%, Si: 1%; III) C: 0.5%, Co: 15%, Cr: 28%, Fe: 12.5%, Mn: 1%, Ni: 35%, Si: 2%; IV) C: 1.45%, Cr: 14%, Fe: 78.55%, V: 4%.
Table 6. Determination of tungsten in polyoxometalate sample W added, µg/mL Total W found, µg/mL Recovery of spike, % *– 50.0 100.0
105.5 159.5 200.0
– 108.0 94.5
* Real amount of tungsten in primary sample is 100.0 µg/mL, as determined by Xray spectrometry.
and rapidity, which favourably differs from previously re ported systems. ACKNOWLEGEMENTS The authors acknowledge the Research Council of Damghan University of Basic Science for the support of this work. ЖУРНАЛ АНАЛИТИЧЕСКОЙ ХИМИИ
том 65
№5
537
REFERENCES 1. Patrick, P., Sophie, L.L. and Gerard, C., Anal. Chim. Acta, 2000, vol. 405, p. 221. 2. Li, S., Deng, N., Zheng, F. and Huang, Y., Talanta, 2003, vol. 60, p. 1097. 3. Tanaka, T., Sakairi, Y. and Ishiyama, T., Anal. Sci., 2001, vol. 17, p. 1949. 4. Ensafi, A.A. and Khaloo, S., Talanta, 2005, vol. 65, p. 781. 5. Ensafi, A.A., Khayamian, T. and Khaloo, S., Anal. Chim. Acta, 2004, vol. 505, p. 201. 6. de Andrade, J.C., de Almeira, A.M., Aleixo, L.M.; Cosciane, A.R. and Abreu, M.F., Anal. Chim. Acta, 2003, vol. 487, p. 243. 7. Quentel, F. and Mirceski, V., Electroanalysis, 2004, vol. 16, p. 1690. 8. van den Berg, C.M.G., Anal. Chim. Acta, 1991, vol. 250, p. 265. 9. Ensafi, A.A. and Zarei, K., Talanta, 2000, vol. 52, p. 435. 10. Ensafi, A.A., Rahimi Mansour, H. and Zarei, K., Fres enius J. Anal. Chem., 1999, vol. 363, p. 646. 11. Safavi, A. and Shams, E., Anal. Chim. Acta, 1999, vol. 385, p. 265. 12. Gao, Z. and Siow, K.S., Talanta, 1996, vol. 43, p. 727. 13. Vega, M. and van den Berg, C.M.G., Anal. Chim. Acta, 1994, vol. 293, p.19. 14. Ensafi, A.A., Khayamian, T. and Atabati, M., Talanta, 2003, vol. 59, p. 727. 15. Stefan, L. and Slawomir, B., J. Alloys Compounds, 2000, vol. 303/304, p. 132. 16. Chattopadhyay, P. and Mistry, P., J. Fresenius, Anal. Chem., 1998, vol. 360, p. 38. 17. Tsiganok, L.P., Vaculich, A.N., Vishnikin, A.B. and Koltsova, E.G., Talanta, 2005, vol. 65, p. 267. 18. Lubal, P., Koprivova, H., Sedo, O., Havel, J., Lis, S. and But, S., Talanta, 2006, vol. 69, p. 800. 19. Malhotra, R.K., Murthy, D.S.R., Srinivasan, G. and Satyanarayana, K., At. Spectrosc., 1987, vol. 8, p. 161. 20. Padmasubashini, V., Ganguly, M.K., Satyanarayana, K. and Malhotra, R.K., Talanta, 1999, vol. 50, p. 669. 21. Agrawal, Y.K. and Sharma, K.R., Talanta, 2005, vol. 67, p. 112. 22. Sohrin, Y., Iwamoto, S. and Akiyama, S., Anal. Chim. Acta, 1998, vol. 363, p. 11. 23. Smolniakov, V.I., XRay Spectrom., 1994, vol. 4, p. 183. 24. Sims, K.W.W. and Gladney, E.S., Anal. Chim. Acta, 1991, vol. 251, p. 297. 25. Malakhova, N.A., Popkova, G.N., Wittmann, G., Kalnichevskaia, L.N. and Brainina, Kh.Z., Electroanaly sis, 1996, vol. 8, p. 375. 26. Shumilova, M.A., Trubachev, A.V. and Kurbatov, D.I., Russ. J. Anal. Chem., 1997, vol. 52, p. 753. 27. Zarebski, J. and Henze, G., Chem. Anal., [Warsaw] 1997, vol. 42, p. 15. 28. Wang, J. and Lu, J., Talanta, 1992, vol. 39, p. 801. 29. Deming, S.N. and Morgan, S.L., Experimental Design: A Chemometrics Approach, 2nd ed. Elsevier, Amsterdam, 1993. 30. Bard, A.J. and Faulkner, L.R., Electrochemical methods, Wiley, New York, 1980. 31. http://www.matweb.com/ 2010
