Fresenius J Anal Chem (2001) 369 : 684–689
© Springer-Verlag 2001
S P E C I A L I S S U E PA P E R
Alberto Chisvert · Amparo Salvador · Maria-Carmen Pascual-Martí · Joan G. March
Efficient flow injection and sequential injection methods for spectrophotometric determination of oxybenzone in sunscreens based on reaction with Ni(II) Received: 27 July 2000 / Revised: 22 September 2000 / Accepted: 25 September 2000
Abstract Spectrophotometric determination of a widely used UV-filter, such as oxybenzone, is proposed. The method is based on the complexation reaction between oxybenzone and Ni(II) in ammoniacal medium. The stoichiometry of the reaction, established by the Job method, was 1:1. Reaction conditions were studied and the experimental parameters were optimized, for both flow injection (FI) and sequential injection (SI) determinations, with comparative purposes. Sunscreen formulations containing oxybenzone were analyzed by the proposed methods and results compared with those obtained by HPLC. Data show that both FI and SI procedures provide accurate and precise results. The ruggedness, sensitivity and LOD are adequate to the analysis requirements. The sample frequency obtained by FI is three-fold higher than that of SI analysis. SI is less reagent-consuming than FI.
1 Introduction The large molar extinction coefficient of oxybenzone in the UVA and UVB ranges makes it a powerful UV filter suitable for sunscreens. However, some sensitizing and phototoxic reactions that have been described for formulations containing oxybenzone [1] make analytical determination necessary. Moreover, determination of oxybenzone in pharmaceutical or cosmetical samples is compulsory in order to ensure that the maximum authorized level
Awarded a Poster Prize on the occasion of the Euroanalysis XI Conference, Lisbon, Sept. 4 to 8, 2000 A. Chisvert · A. Salvador () · M. C. Pascual-Martí Departamento de Química Analítica, Facultad de Química, Universitat de València, Doctor Moliner 50, 46100-Burjassot, Valencia, Spain e-mail:
[email protected] J. G. March Departamento de Química, Facultad de Ciencias, Universitat de les Illes Balears, Carretera de Valldemosa km 7.5, 07071-Palma de Mallorca, Spain
is respected. This level is 10% according to European legislation [2], whereas the US Food and Drug Administration [3] authorizes only 6%. Different techniques such as high pressure liquid chromatography [4–12], gas chromatography [13–15] and nuclear magnetic resonance [16] have been used to determine oxybenzone. A colorimetric reaction between nickel and oxybenzone is proposed here in order to determine oxybenzone in sunscreens. Continuous flow injection (FI) analysis [17, 18] has been widely applied in most analytical techniques. Ruzicka and Hansen have presented an overview [19] of flow injection analysis covering: continuous and discontinuous flow-injection systems, hyphenated systems and mapping of chemical processes, flow injection and chromatography and future possibilities. Sequential injection (SI) analysis is based on the same principles as FI and is an alternative under development [20]. This technique substantially decreases reagent consumption and thus the waste generated. In addition, devices based on SI yield robust, stable systems that are suitable for routine monitoring [21] and allow, as does FI, in-line sample preparations [22], kinetic determinations [23–24], and other strategies. Recent reviews [25, 26] show a growing interest in SI methodologies as an alternative approach to process analytical chemistry and describe the advances in SI [27]. FI and SI methodologies should be compared in order to select the better strategy for a concrete determination. However, few articles have been published in which the two methodologies are compared [28]. FI and SI determinations of oxybenzone in sunscreen formulations, based on the proposed colorimetric reaction, were carried out here with comparative purposes. Results were in good agreement with those obtained by an high performance liquid chromatography (HPLC) procedure.
685 Ethanol 96% cosmetic grade (Guinama, Valencia, Spain), ammonium hydroxide (25% NH3, d = 0.910 g/mL) (Probus, Badalona, Spain) and nickel nitrate hexahydrated (Probus, Badalona, Spain) were used for the FI and SI determinations. 2,2′-Dihydroxy-4,4′-dimethoxybenzophenone (Aldrich, Barcelona, Spain) was used as internal standard in the HPLC procedure. Tetrahydrofurane (THF) (Scharlau, Barcelona, Spain) and acetic acid (HAc) (Panreac, Barcelona, Spain) were also used as solvents in this procedure. 2.3 Procedures 2.3.1 Reference method
Fig. 1 a The flow injection system, b The sequential injection system. (1) 0.02 M Ni(II) in 20:80 water:ethanol solution; (2) 1 M NH3 in 20:80 water:ethanol solution; (3) peristaltic pump; (4) connector; (5) six channels injection valve with 200 µL loop; (6) reaction coil; (7) flow cell; (8) ethanol; (9) autoburette with 10 mL syringe; (10) holding coil; (11) eight-channels selector valve; (12) 0.01 M Ni(II) in 20 : 80 water:ethanol solution; (13) sample or standard solution; (14) 0.5 M NH3 in 20 : 80 water:ethanol solution
2 Experimental 2.1 Apparatus An 8453 Hewlett-Packard UV-V diode array spectrophotometer was used for the spectrophotometric measurements. The FI system (Fig. 1 a) used was made up of the following components: a Gilson minipuls-3 peristaltic pump; a six-channel Rheodyne valve; a 10 mm QS 1000 (Hellma, Müllheim/Baden, Germany) flow-through cell (volume 18 µL); Teflon tubing (0.5 mm i.d.). The SI system depicted in Fig. 1 b was constructed with the following components: a Crison 2031 autoburette (Alella, Barcelona, Spain) equipped with a 10 mL syringe and an eight-channel Crison 2030 automatic valve (Alella, Barcelona, Spain) connected to a personal computer via an RS 232 C interface and controlled by home made software. A Hitachi HPLC chromatograph equipped with an L-7100 high pressure pump and an L-7420 UV-V detector was used to carry out the determinations by the reference procedure. A Lichrospher RP-18 (12.5 cm length, 4 mm i.d., 5 µm particle size) (Merck) column was used.
There are no standard methods for the determination of oxybenzone. A method based on a chromatographic procedure [7] with some modifications was therefore used as a reference. 0.5–3 g of sunscreen sample was solved and diluted to 50 mL with THF. This solution was filtered through a Whatman 42 filter. 2 mL of a 0.6 mg/mL internal standard solution and 10 mL of distilled water were added to 2 mL of the filtrate and diluted to 25 mL with THF. Oxybenzone solutions (from 10 to 50 µg/mL) in THF containing the same concentration of internal standard and water as the sample solutions were used as standards. A 20 µL aliquot of sample or standard solution was injected into the HPLC system using a mobile phase 55 : 0.1 : 44.9 (v/v/v) THF:HAc:H2O with a 0.5 mL/min flow rate. The determination was carried out at 313 nm. 2.3.2 FI spectrophotometric determination 0.02–0.12 g of sunscreen sample was dissolved and diluted to 25 mL with ethanol. A 200 µL aliquot of sample solution was injected into the double channel FI manifold shown in Fig. 1 a through a 0.45 µm filter syringe. 0.02 M Ni(II) in 20 : 80 water : ethanol solution and 1.0 M NH3 in 20 : 80 water:ethanol were, respectively, aspirated by each channel. The determination was carried out at 376 nm using standards of oxybenzone in ethanol (working range up to 50 µg/mL). 2.3.3 SI spectrophotometric determination 0.02–0.12 g of sunscreen sample was dissolved and diluted to 25 mL with ethanol. Sample solution was aspirated in the SI system shown in Fig. 1 b; oxybenzone in ethanol solutions (up to 40 µg/mL) were used as standard. 0.01 M Ni(II) in 20 : 80 water:ethanol solution and 0.5 M NH3 in 20 : 80 water:ethanol were aspirated to carry out the colorimetric reaction. Absorbance was measured at 376 nm and corrected at 700 nm to minimize the effect of refractive index variations. Three measurements were carried out for all sample and standard solutions. The following analytical cycle was used: 1) 2) 3) 4)
Aspiration of 100 µL 0.01 M Ni(II) solution Aspiration of 200 µL of sample or standard solutions Aspiration of 100 µL 0.5 M NH3 solution Propulsion of 1.5 mL to the detector through the reaction coil.
2.2 Reagents Oxybenzone (2-hydroxy-4-methoxybenzophenone) (Aldrich, Barcelona, Spain) was used to prepare the standards. Sunscreens containing different components were analyzed following the flow procedures and the HPLC reference methods. The analyzed commercial samples were: Nievina solar milk SPF 8 (Valencia, Spain), Pryca solar milk SPF 8 (Lab. Expanscience, Madrid, Spain), Pryca solar milk SPF 20 (Lab. Expanscience, Madrid, Spain), Shiseido solar oil SPF 4 (Tokyo, Japan) and Clinique solar lotion SPF 6 (London, U.K.), purchased from the local market.
3 Results and discussion 3.1 Reference method The data obtained by the HPLC reference method for five commercial sunscreen formulations studied and their standard deviation are shown in Table 1. The curve obtained by the internal standard (IS) calibration method was linear in the working range (up to
686 Table 1 Results obtained for oxybenzone determination by the reference and the proposed methods Sunscreen sample
A B C D E
Oxybenzone ± s (mg/g) Reference procedure
FI-spectrophotometry
SI-spectrophotometry
15.7 ± 0.2 9.5 ± 0.3 29.7 ± 0.4 5.0 ± 0.2 25.4 ± 0.2
15.1 ± 0.5 9.6 ± 0.3 30.3 ± 0.5 5.0 ± 0.1 26.5 ± 0.9
15.2 ± 0.8 9.5 ± 0.3 31 ± 2 5.9 ± 0.2 26 ± 2
50 µg/mL). The equation obtained was: Aoxybenzone / AIS = (–2 ± 1) 10–2 + (0.0248 ± 0.0004) C (N = 5, C is expressed in µg/mL). Retention times were 4.1 min for the internal standard and 5.2 min for oxybenzone, but the complex matrix composition of some sunscreen samples requires higher run times to assure the elution of other components.
3.2 Colorimetric reaction Irima et al. [29] proved that Ni(II) can be extracted from a solution by complexation with biquelate ligands such as 2-hydroxy-5-methylbenzophenone or 2-hydroxy-4-methoxybenzophenone. The reaction between Ni(II) and oxybenzone was studied in the present work in order to identify the optimal analytical conditions for oxybenzone determination. The OH phenolic groups are easy to dissociate, especially when they have a carbonyl in the ortho position which can delocalize the negative charge density generated; then, the stability of the formal complexes increases with the pH. Because of this, an ammoniacal medium, which allows this enhanced without crystallization of nickel hydroxyde, is proposed here. A 0.5 M ammoniacal medium (pH 11.5) allows the reaction between oxybenzone and a Ni(II) solution (0.01 M). The electronic states of the oxybenzone are affected and a bathochromic shift of the absorption maximum is produced; the band allows the determination of oxybenzone without matrix interferences. Figure 2 shows the effect of the reagents on the oxybenzone spectra. The Job method was applied and showed the reaction stoichiometry to be Ni(II):oxybenzone is 1:1. 2-Ethylhexyl-4-dimethylaminobenzoate and 2-ethylhexyl-4-methoxycinnamate are the UV-filters usually mixed with oxybenzone in sunscreen formulations because they have absorbance higher than oxybenzone in the UVB range but do not present absorbance in the UVA range; therefore a broad spectrum UVB-UVA photoprotection products are achieved with these combinations. These UV-filters do not react with Ni(II) in ammoniacal medium. 4-t-Butyl-4′-methoxydibenzoylmethane is a UV-filter whose absorbance band overlaps with that of the Ni(II)-
Fig. 2 Effect of the reagents on the spectrum of a 20 : 80 water:ethanol solution containing 20 µg/mL of oxybenzone. Oxybenzone (); oxybenzone and 0.01 M Ni(II) (); oxybenzone and 0.5 M NH3 (); oxybenzone, 0.01 M Ni(II) and 0.5 M NH3 ()
oxybenzone complex but it is not usually combined with oxybenzone in sunscreen formulations because it only absorbs in the UVA range. Sulisobenzone, homosalate and 2-ethylhexyl salicylate react with Ni(II), but they are not usually associated with oxybenzone because the first has similar absorbance properties than oxybenzone and the other do not absorb in the UVA range and have lower molar extinction coefficients than oxybenzone in the UVB range. Other components usually present in the sunscreen formulations such as base cream, mineral or vegetable oils, emulsifying agents and preservatives do not undergo the reaction. 3.3 Continuous flow injection analysis 3.3.1 Experimental variables affecting the analytical signal Since previous experiments with large sample volumes provided wide peaks, which increase the time of analysis, a loop volume of 200 µL was selected, which provides a suitable dilution to the oxybenzone content expected in the samples. Figure 3 a shows the influence of flow rate (from 2 to 6 mL/min) and coil length (from 10 to 50 cm) on the absorbance signal and its relative standard deviation. Figure 3b shows as an example the fiagram obtained for a 25 cm reaction coil at different flow rates. A 3 mL/min total flow rate (1.5 mL for each reagent) and 25 cm reaction coil were chosen in order to obtain good reproducibility, analysis time and sensitivity for oxybenzone determination in sunscreen samples. The relative standard deviations (RSD) were lower than 2%.
687
cept and the slope were 0.78 and 1.40, respectively, which shows the comparability of the results obtained by the two methods. The RSD of the concentration values obtained by FI-spectrophotometry was in the 1.5–3.5% range. In order to study the ruggedness of the method, the concentration of NH3 was varied ±20% for a fixed Ni(II) concentration and the results obtained for the oxybenzone concentration in a sample were compared by an ANOVA test. The Fexp was 1.46, which is lower than the theoretical F = 3.42 (P = 95%, freedom degrees = 2 and 24). The concentration of Ni(II) was varied for a fixed NH3 concentration and the results for the oxybenzone concentration in a sample were also compared by the ANOVA test. The Fexp was 1.64 which is lower than theoretical F = 3.42 (P = 95%, freedom degrees = 2 and 24). These results show the ruggedness of the FI procedure. The measurement time of a solution is 15 s whereas a complete cycle where each solution is measured three times consumes 50 s. The sample frequency is therefore 72 h–1. 3.4 Sequential injection analysis 3.4.1 Study of the experimental variables
Fig. 3 a Influence of the flow rate on the absorbance measurements using different reaction coil lengths. (Mean absorbance ± standard deviation obtained for three absorbance measurements of each point are drawed ); b fiagram obtained for different flow rates. (Reaction coil used in data show in the fiagram: 25 cm)
3.3.2 Analytical figures of merit The calibration curve was linear in the working range (up to 50 µg/mL), with a regression coefficient r2 = 0.9990. The calibration equation was: A = (0.099 ± 0.003) + (0.0115 ± 0.0001) C (N = 18), where C is expressed in µg/mL. The sensitivity estimated by the slope of the calibration curve was 0.0115 mL/µg. The limit of detection estimated by 3 sy/x / b (where sy/x is the standard deviation of the calibration curve and b the slope) was 1.9 µg/mL. The recovery of the method, obtained by spiking the five commercial sunscreens with oxybenzone, was 100 ± 5%. In order to evaluate the accuracy of the method, the five commercial sunscreen samples were analyzed by the proposed FI procedure (Y) and by the HPLC reference method (X). The results are given in Table 1. The correlation line found for the mean values obtained by FI-spectrophotometry vs reference procedure was Y = (–0.04 ± 0.05) + (1.04 ± 0.03) X (N = 5, r2 = 0.998), where X and Y are expressed in % oxybenzone (m/m). The theoretical t value for a 95% confidence level and N-2 degrees of freedom is 3.18, and the experimental values for the inter-
The parameters used in this study (except the one corresponding to the modified parameter in each experiment) were: 200 µL aspirated sample volume, 100 µL aspirated volume for each segment of reagent solutions, 7.5 mL/ min propulsion flow rate, 75 mL/min aspiration flow rate, 25 cm reaction coil length, 150 cm holding coil length. Different sandwich arrangements were assayed to ensure adequate sensitivity and precision. Figure 4 shows the siagrams obtained. The Ni(II)-oxybenzone-NH3 sequence provided the best results, with and RSD lower than 2%. Moreover, this sequence provided sample frequency higher than the other. Volumes from 50 to 250 µL of sample were assayed, as shown in Fig. 5 a. The standard deviations were not af-
Fig. 4 Influence of the sandwich arrangement on the absorbance measurements. Type of sandwich: A, Ni(II)-oxybenzone-NH3; B, NH3-oxybenzone-Ni(II); C, NH3-Ni(II)-oxybenzone-Ni(II)-NH3; D, Ni(II)-NH3-oxybenzone-NH3-Ni(II); E, NH3-Ni(II)-oxybenzone-NH3-Ni(II); F, Ni(II)-NH3-oxybenzone-Ni(II)-NH3
688 Fig. 5 Influence of the experimental conditions of the SI system on the absorbance measurements. (Mean absorbance ± standard deviation obtained for three absorbance measurements of each point are drawed) a Aspirated sample volume, b volumes of Ni(II) and NH 3 solutions, c propulsion flow rate, d reaction coil length
fected by this variable (according to Cochran and Fisher tests). A 200 µL sample volume was chosen because it provided adequate sensitivity and good precision (RSD of measurements was lower than 2%). Different volumes of the Ni(II) and NH3 solutions were assayed, as shown in Fig. 5 b. The standard deviations were not affected by this variable (according to Cochran and Fisher tests). A total volume of 200 µL (100 µL of each reagent solution) provided adequate sensitivity and precision (RSD of measurements was lower than 2%). The aspiration flow rate was kept constant (75 mL/ min) for technical reasons. The reaction does not depend on the propulsion flow rate or reaction time because it can be considered instantaneous, but mixing could be affected by the propulsion flow rate. Propulsion flow rate between 7.5–25 mL/min were assayed; minor propulsion flow rates are not permitted by the instrument and higher ones were not considered since sensitivity decreased with the flow rate and the standard deviation increased. Figure 5 c shows that 7.5 mL/min provided adequate sensitivity and precision (RSD of measurements was lower than 2%). Reaction coil lengths between 25–100 cm were assayed (Fig. 5 d); minor coil lengths were not adequate to the manifold geometry. Short coils provided high sensitivity and high sample throughput, whereas long coils strongly decreased the sensitivity and increased the analysis time. The standard deviations were not affected by this variable (according to Cochran and Fisher tests). A reaction coil length of 25 cm, which provides adequate sensitivity and precision (RSD lower than 2%), was chosen.
3.4.2 Analytical figures of merit The calibration curve was linear in the working range (up to 40 µg/mL) with a regression coefficient of r2 = 0.995. The calibration equation was: A = (–0.011 ± 0.009) + (0.0184 ± 0.0004) C (N=15), where C is expressed in µg/mL. The sensitivity of the instrumental measurements estimated by the slope of the calibration curve was 0.0184 mL/µg. The limit of detection estimated by 3 sy/x / b (where sy/x is the standard deviation of the calibration curve and b the slope) was 2.4 µg/mL. The recovery of the method, obtained by spiking of the five commercial samples with oxybenzone, was 101 ± 4%. The RSD of the concentration values obtained by SI-spectrophotometry ranged from 3% to 8%. In order to evaluate the accuracy of the method, five commercial sunscreen samples were analyzed by the proposed SI procedure (Y) and by the HPLC reference method (X). The results are given in Table 1. The correlation line for the mean values obtained by SI-spectrophotometry vs the reference procedure was Y = (0.00 ± 0.07) + (1.03 ± 0.04) X (N = 5, r2 = 0.996), where X and Y are expressed in % oxybenzone (m/m). The theoretical t value for a 95% confidence level and N-2 degrees of freedom is 3.18, and the experimental values for the intercept and the slope were 0.02 and 0.71, respectively, which shows the comparability of the results obtained by the two methods. The concentration of NH3 was varied ± 20% for a fixed Ni(II) concentration, and the results obtained for oxybenzone concentration in a sample were compared by an ANOVA test. The Fexp was 1.72 which is lower than the theoretical F = 3.42 (P = 95%, freedom degrees = 2 and 24). The concentration of Ni(II) was varied for a fixed NH3 concentration and the results for oxybenzone con-
689
centration in a sample were also compared by the ANOVA test. The Fexp was 2.95, which is lower than theoretical F = 3.42 (P = 95%, freedom degrees = 2 and 24). These results show the ruggedness of the method. The measurement time for a same solution was 34 s, whereas a complete cycle including three measurements per sample took 150 s. The sample frequency is therefore 24 h–1.
4 Conclusions The complexation reaction between oxybenzone and Ni(II) in ammoniacal medium is proposed to determine oxybenzone in sunscreen formulations. Flow injection and sequential injection procedures were studied in order to carry out the automatization of the method. The FI and the SI systems were optimized with respect to precision and sensitivity. Under the selected experimental conditions the detection limits, reproducibilities, accuracy, ruggedness and sample frequency were established. The proposed methods permit analysis of oxybenzone in commercial sunscreens. No prior separation is necessary and simple, rapid determinations can be performed by complexation on-line. Good precision and sensitivity were obtained by both methods. Cosmetic ethanol, wich offers the advantages of low cost and low toxicity can be used as the carrier. SI provided a sensitivity higher than FI and a similar LOD, but standard deviation was lightly higher for SI. The sample frequency of FI was three fold higher than that of SI. This finding has also been reported by other authors studying other applications [30]. The major advantages of SI were the more cost effective use of reagents and a high level of automation due to the control of the system by the software. Acknowledgements The authors should like to express their heartfelt gratitude to the Division of Analytical Chemistry of the FECS and also to Springer-Verlag for the Second Poster Prize awarded in Euroanalysis XI to a part of this work (poster number 317). The authors acknowledge the financial support of the Spanish Ministry of Education and Culture for our research project on the development of analytical methods for sunscreen formulations (PM-98–0210). A.Chisvert expresses his gratitude to Consellería d’Educació i Cultura (Generalitat Valenciana, Valencia, Spain) and to Ministerio de Educación y Cultura (Spain) for a predoctoral grant.
References 1. Schmidt T, Ring J, Abeck D (1998) Dermatology 196:354–357 2. EEC Directive 83/574, 1983, No. L332, pp 38–42 3. FDA, Department of Health and Human Services, 21CFR Parts 310, 352, 700 and 740, RIN 0910-AA01, Sunscreen Drug Products for Over-The-Counter Human Use; Final Monograph, Federal Register, Vol 64, No.98/May 21, 1999/Rules and Regulations, pp 27666–27693 4. Ohshima T, Saito E, Yakuji G (1987) Gekkan-Yakuii 29: 2477–2482 5. Tan HSI, Sih R, Moseley SE, Lichtin JL (1984) J Chromatogr 291:275–282 6. Meijer J, Loden M (1995) J Liq Chromatogr 18:1821–1832 7. DiNunzio JE, Gadde RR (1990) J Chromatogr 519:117–124 8. Ikeda K, Suzuki S, Watanabe Y (1989) J Chromatogr 487:240–245 9. Gagliardi L, Cavazzutti G, Montarella L, Tonelli D (1989) J Chromatogr 464:428–433 10. Gagliardi L, Amato A, Basili A, Cavazzutti G, Tonelli D (1987) J Chromatogr 408:409–415 11. Yao XY, Zheng XQ, Qin XY, Qi QP (1998) Sepu 16:223–225 12. Tomasella FP, Pan Z, Love LJC (1991) J Chromatogr 587: 325–328 13. Ro KW, Choi JB, Lee MH, Kim JW (1994) J Chromatogr A 688:375–383 14. Hany J, Nagel R (1995) Deutsch Lebensm Rundsch 91:341– 345 15. Ikeda K, Suzuki S, Watanabe Y (1990) J Chromatogr 513:321–326 16. Mori K, Itoh K, Suzuki S, Nakamura H (1996) Japan J Toxicol Environ Health 42:60–66 17. Ruzicka J, Hansen HE (1980) Anal Chim Acta 114:19–44 18. Valcárcel M, Luque de Castro MD (1987) Flow Injection Analysis: Principles and Applications. Ellis Horwood, Chichester 19. Ruzicka J, Hansen EH (1998) Trends Anal Chem 17:69–73 20. Ruzicka J, Marshall GD (1990) Anal Chim Acta 237:329–343 21. Cerdá A, Oms MT, Forteza R, Cerdá V (1998) Anal Chim Acta 371:63–71 22. Zagatto EAG, Vicente S, Oliveira CC, Sartini RP, Costa-Lima JLF (1998) J Flow Injection Anal 15:226–233 23. March JG, Simonet BM, Grases F (2000) Anal Chim Acta 409:9–16 24. Sultan SM, Desai NI (1998) Talanta 45:1061–1071 25. Barnett NW, Lenehan CE, Lewis (1999) Trends Anal Chem 18:346–353 26. Taljaard RE, Van Staden JF (1998) Lab Rob Autom 10:325–337 27. Liu XZ, Fang ZL (1999) Fenxi Kexue Xuebao 15:70–78 28. Reis BF, Morales-Rubio A, de la Guardia M (1999) Anal Chim Acta 392:265–272 29. Irimia G, Mocanu R, Olaru R, Särghie I (1997) Revista de Chimie 48 :395–398 30. Van Staden JF, du Plessis H, Linsky SM, Taljaard RE, Kremer B (1997) Anal Chim Acta 354:59–64
