Simultaneous determination of and in soft drinks by using lanthanide-sensitized luminescence

M. P. Aguilar-Caballos, A. Gómez-Hens and D. Pérez-Bendito

Analytical Chemistry Division, Faculty of Sciences, University of Córdoba, 14004-Córdoba, Spain

Received 26th March 1999, Accepted 24th May 1999

A simple and fast approach is used for the first time to develop a time resolved lanthanide-sensitized luminescence method for the simultaneous determination of a and a sweetener, namely benzoic acid (BZ) and saccharin (SC), respectively, in samples. The method involves the formation of the corresponding ternary chelates with terbium(iii) and trioctylphosphine oxide (TOPO) in the presence of Triton X-100, and the measurement of the initial rate and equilibrium signal of this system, which were obtained in 0.1 and 5 s, respectively. The dynamic ranges of the calibration graphs, obtained by using kinetic and equilibrium measurements, were 0.2Ð36 mg ml21 and 0.15Ð30 mg ml21, respectively, for BZ, and 3.3Ð24 mg ml21 and 4Ð36 mg ml21 for SC and the detection limits were 0.07 and 0.04 mg ml21, respectively, for BZ, and 1.1 and 1.2 mg ml21, respectively, for SC. The relative standard deviation ranged between 2.3 and 3.0%. Both compounds were determined simultaneously by using a system of two equations which were resolved by using the calibration data obtained individually for each analyte and by multiple linear regression. Mixtures of BZ and SC in ratios between 3:1 and 1:9 were satisfactorily resolved by using both approaches. The method was applied to the direct analysis of several soft drinks. Analytical recoveries ranged between 89.3 and 108.5%.

Introduction cence mode so eliminating scatter, Raman and any fluorescence background signal. Benzoic acid (BZ) and saccharin (SC) are two common Benzoate was previously reported as a counter ion in studies additives widely used in numerous as a preservative and on lanthanideÐcrown ether complexes extracted with ethyl an artificial sweetener, respectively. They are considered acetate.12 Later, the possibility of using the benzoate ion for innocuous for humans at the allowed concentration levels, but it energy transfer to the lanthanides in an aqueous medium was was reported1 that benzoates can promote an allergic response studied and applied to the determination of these ions.13 in sensitive people and the potential toxicity of SC, which is Detection limits of 1029 M for terbium and 1028 M for prohibited in some countries considering possible carcinogenic dysprosium and europium were reported. Several methods effects, still continues to be a topic of investigation. The based on terbium-sensitized luminescence have been reported simultaneous determination of these compounds in food for the determination of benzoic acid derivatives such as p- analysis is mainly carried out by using liquid chromatography aminobenzoic, salicylic and p-aminosalicylic acids and di- and UV photometric detection.227 Although these methods flunisal and applied to clinical14,15 and pharmaceutical16 allow other additives together with BZ and SC to be determined, analysis. However, the application of this approach to the some of them involve a solid-phase extraction by using individual and simultaneous determination of BZ and SC and its 224 quaternary ammonium or C18 cartridges prior to the liquid use in routine food analysis have not been described up to date. chromatographic analysis, which causes difficulties in the The proposed method combines the selectivity of lanthanide- application of these methods to routine food analysis. Thus, the sensitized luminescence with kinetic and equilibrium measure- main aim of this paper has been the development of a simple, ments of both BZ and SC systems. The high initial rate of these automatic and fast method for the simultaneous determination reactions requires the use of stopped-flow mixing technique, of BZ and SC, which does not require any previous separation which also enables ready automation of the method since step and can be easily applied to the analysis of soft drink reactant manipulation is minimal and measurements are samples. performed shortly after mixing. In spite of the high sensitivity of photoluminescence techniques, they have been rarely applied to the individual determination of BZ8 and SC9 and their simultaneous determi- nation has not been described up to date. Thus, the lack of Experimental luminescent methods for the resolution of mixtures of these compounds led us to study their reactions with lanthanide ions. Instrumentation As known, many lanthanide chelates feature a special lumines- cent behaviour as a result of the intramolecular energy transfer An SLM-Aminco (Urbana, IL, USA) AB2 luminescence process from the ligand to the central ion, which avoids or spectrometer, equipped with a 150 W continuous xenon lamp minimizes selectivity limitations of luminescence methods.10,11 and a 7 W pulsed xenon lamp, was used. The instrument was They show large Stokes shifts and narrow emission bands, furnished with a stopped-flow module17 supplied by Quimi-Sur which allow the spectral discrimination of the analytical signal. Instrumentation (Seville, Spain), which was fitted with an Also, their relatively long lifetimes allow the temporal discrim- observation cell of 1 cm path length. The excitation and ination by making analytical measurements in the phosphores- emission slits were adjusted to provide an 8 nm band-pass. The

Analyst, 1999, 124, 1079Ð1084 1079 temperature of the solutions in the stopped-flow module and cell compartment was kept constant at 25 ¡C by circulating water from a thermostated tank.

Reagents

All chemicals used were of analytical reagent grade. Stock solutions of BZ (100 mg ml21) (Sigma, Poole, Dorset, UK) and SC (400 mg ml21) (Sigma) were prepared in ethanolÐdistilled water (1 + 99). A 2 3 1022 M terbium(iii) solution was made by dissolving an appropriate amount of terbium(iii) nitrate penta- hydrate (Aldrich, Gillingham, Dorset, UK) in distilled water. A 6 3 1024 M solution of TOPO (Sigma) was prepared in 50% . An 1% Triton X-100 (Fluka, Buchs, Switzerland) aqueous solution and an imidazol buffer solution (0.2 M, pH 6.0) were also used. Fig. 1 Kinetic curves obtained for SC (1), BZ (2) and a mixture of both (3) in the presence of terbium(iii) (2 3 1023 M), Triton X-100 (8 3 1022 M) 25 21 21 and TOPO (5.6 3 10 M). [SC] = 10 mg ml , [BZ] = 1 mg ml . lex 284 nm, lem 545 nm (IL = luminescence intensity). Procedures

measuring the initial rate (v) and equilibrium signal (IL) of the Individual determination of BZ and SC. A solution kinetic curve and applying the following equations: containing terbium(iii) (2 3 1023 M), Triton X-100 (8 3 22 25 22 10 %), TOPO (5.6 3 10 M), imidazol buffer (2.4 3 10 v = a0 + a1[BZ] + a2[SC] M) and ethanol (4%) was used to fill one of the two 2 ml drive IL = aA0 + aA1[BZ] + aA2[SC] syringes of the stopped-flow module. The other syringe was filled with a solution containing standard or sample BZ or SC, where [BZ] and [SC] are the concentrations of the analytes imidazol buffer (2.4 3 1022 M) and ethanol (4%). The final expressed in mg ml21. Two methods were used to resolve this 21 concentration ranges of BZ and SC were 0.2Ð30 mg ml and equation system. In the first method (conventional method), a0 21 4Ð36 mg ml , respectively, in the equilibrium method, and and aA0 are the sum of the intercepts of the corresponding 0.3Ð36 mg ml21 and 4Ð34 mg ml21, respectively, in the kinetic calibration graphs obtained individually for each compound by method. In each run, 0.15 ml of each solution was mixed in the using the initial rate or the equilibrium signal as the analytical mixing chamber. The variation of the luminescence intensity parameter, and a1, a2, aA1 and aA2 are the corresponding slope of with time throughout the reaction was monitored at lex 284 nm each calibration graph; the equations were resolved by using a and lem 545 nm by using the phosphorescence mode, with delay simple BASIC program. In the second method, the coefficients and gate times of 0.1 and 5 ms, respectively. Kinetic data were of these equations were estimated by multiple linear regression processed by linear regression using the microcomputer, (MLR) from 20 mixtures of both analytes in the above furnished with a program for application of the initial-rate established ranges, and the system was resolved by using a method. The dynamic and equilibrium signals were measured in laboratory-developed FORTRAN 77 program.18 ca. 0.1 s and 5 s, respectively. All measurements were made at 25 ¡C and each standard or sample was assayed in triplicate. Determination of BZ and SC in soft drink samples. Each Although the contribution of the blank signal was very low, it sample (2.5 ml) was degassed, neutralised with 2 M sodium was subtracted from the kinetic and equilibrium measure- hydroxide and diluted to 10 ml with distilled water. A volume ments. (1 ml) of this solution was treated as described above.

Simultaneous determination of BZ and SC. The procedure used was the same as that described above, but the syringe Results and discussion containing the analyte was filled with appropriate amounts of both compounds to obtain a final concentration level of each In order to choose the adequate systems for the determination of analyte within the corresponding calibration graph. The concen- BZ and SC, based on lanthanide-sensitized luminescence, a tration of each analyte in the mixture was calculated by series of previous assays were carried out, taking into account

Fig. 2 Effect of terbium(iii) (A), Triton X-100 (B) and TOPO (C) concentrations on the initial rate (1, 2, 3) and equilibrium signal (1A, 2A, 3A) values obtained 21 21 for SC (1, 1A), BZ (2, 2A) and a mixture of both (3, 3A).[SC] = 5 mg ml , [BZ] = 1 mg ml . lex 284 nm, lem 545 nm (v = initial rate, IL = luminescence intensity).

1080 Analyst, 1999, 124, 1079Ð1084 Table 1 Analytical features of the methods

BZ SC

Methoda K E K E

Linear range/mg ml21 0.2Ð36 0.15Ð30 3.3Ð24 4.0Ð36 (Slope ± s)b (3.32 ± 0.01) 3 1021 (2.01 ± 0.05) 3 1021 (2.12 ± 0.02) 3 1022 (7.5 ± 0.3) 3 1023 (Intercept ± s) (5 ± 2) 3 1022 (1 ± 0.7) 3 1022 (7 ± 3) 3 1023 (9 ± 6) 3 1023 Correlation coefficient (r)c 0.998 0.993 0.992 0.999 SEEd 0.31 0.15 0.01 0.006 Detection limit/mg ml21 0.07 0.04 1.1 1.2 Precision (RSD%)e 0.8f 2.6 2.4 10 2.3 2.3 5 3.0 2.7 15 2.6 2.5 a b 21 21 21 c d e K, kinetic; E, equilibrium. units: K, s mg ml; E, IL mg ml (IL, luminescence intensity). n = 10. SEE, Standard error of estimate. RSD, Relative standard deviation (n = 11). f mg ml21. that the luminescence of lanthanide chelates can be greatly The kinetic behaviour of both systems was studied with the enhanced in the presence of a surfactant and a second aim of using the initial rate together with the equilibrium signal ligand.10,11 Thus, the surfactant protects these chelates from as analytical parameters, which allow the establishment of a non-radiative processes, while the second ligand only has a system of two equations for the simultaneous determination of synergistic effect as it removes water molecules from the BZ and SC. Fig. 1 shows the kinetic curves obtained for each coordination sphere of the lanthanide ion, avoiding the non- analyte and a mixture of both. Although the initial rate of these radiative deactivation caused by these molecules. systems is very high, it can be easily measured by using Terbium(iii) was the lanthanide ion chosen to carry out this stopped-flow mixing technique. The additivity of the initial rate study because, although BZ also reacts with europium(iii) and and equilibrium signal allows the application of the afore- dysprosium(iii),11 the luminescence signal obtained with these mentioned equation system. two ions was about three-times lower than that obtained with terbium. Also, SC only gave a luminescence signal with this ion. Three surfactants, namely cetyltrimethylammonium bro- Optimization of variables mide (CTAB), sodium dodecyl sulfate (SDS) and Triton X-100, were assayed. While the cationic and anionic surfactants did not Variables affecting the BZ and SC systems were optimized cause any effect on these systems, the presence of Triton X-100 separately and jointly by the univariate method using initial rate noticeably increased the luminescence signal. The synergistic ligands assayed were EDTA, 1,10-phenanthroline and trioctyl- Table 2 Effect of foreign substances on the determination of 4 mg ml21 of phosphine oxide (TOPO), obtaining the best positive effect with BZ and SC TOPO. The excitation spectra of BZ and SC in the presence of Maximum tolerated concentration/mg ml21 terbium(iii), Triton X-100 and TOPO showed a relatively wide band with maximum signal at 284 nm, while the emission BZ SC spectra showed the three characteristic bands of terbium(iii) at Compound Ka E K E 490, 545 and 580 nm, the second band being the most intense. Although this emission band partially overlaps with the second > 400b > 400b > 400b > 400b order scatter from the excitation wavelength, this is eliminated Cyclamic acid 350 350 280 280 by the time-resolved mode, which can be used in these systems 110 110 110 110 since they have a relatively long luminescence lifetime. Also, 200 80 200 80 Ascorbic acid 80 28 80 30 this mode increases the relative luminescence signal as this is a b integrated during a longer time than in the fluorescence K, kinetic method; E, equilibrium method. Maximum tested concentra- tion was 100-fold the analyte concentration. mode.

Table 3 Resolution of BZ and SC mixtures

BZ/mg ml21 SC/mg ml21

Founda Founda BZÐSC mixtures (ratio) Prepared CONVb MLRb Prepared CONVb MLRb

3:1 12 12.1 ± 0.4 11.8 ± 0.3 4 3.6 ± 0.2 3.7 ± 0.2 1.5:1 6 6.4 ± 0.3 6.1 ± 0.2 4 3.96 ± 0.08 4.1 ± 0.1 1:2 4 4.2 ± 0.3 4.1 ± 0.2 8 7.9 ± 0.3 7.8 ± 0.2 1:3 4 3.7 ± 0.2 3.99 ± 0.1 12 13.4 ± 0.2 13.1 ± 0.3 1:4 4 3.99 ± 0.09 3.9 ± 0.1 16 15.4 ± 0.5 14.9 ± 0.3 1:8 1 0.97 ± 0.03 0.97 ± 0.02 8 7.9 ± 0.2 7.9 ± 0.2 1:9 2 2.1 ± 0.1 2.02 ± 0.08 18 19 ± 0.5 18.3 ± 0.3 Comparison of results rc 0.998 0.9997 0.993 0.993 Slope 1.015 0.984 1.056 0.999 Intercept 25 3 1023 0.06 20.39 20.018 a Mean of three determinations ± s. b CONV, conventional method; MLR, multilinear regression method. cCorrelation coefficient.

Analyst, 1999, 124, 1079Ð1084 1081 and equilibrium signal measurements from the kinetic curves in terbium(iii). However, the imidazole buffer, in a concentration order to develop methods for the individual and simultaneous range 1.6 3 1022-4.0 3 1022 M, did not cause any effect on the determinations of these compounds. All concentrations given systems. are initial concentrations in the syringes (twice the actual Fig. 2A shows the variation of the initial rate and equilibrium concentrations in the reaction mixture at time zero after signal of both systems, individually and in mixture, with the mixing). Those values yielding the minimum possible standard terbium(iii) concentration. A 2 3 1023 M concentration was deviation for the initial rate under conditions where the reaction chosen as it allows maximum values of both parameters and the order was zero or near zero, and those where the initial rate and additivity of both systems to be obtained. As commented above, equilibrium signal obtained were additive for BZ and SC the presence of a surfactant, Triton X-100, and a synergistic mixtures, were taken as optimal. Each result was the average of ligand, TOPO, is required in order to obtain both luminescent three measurements. systems. The study of the effect of the concentration of these The initial rate and equilibrium signal of both BZ and SC variables (Fig. 2B and C) showed that both measurement systems were independent of the pH in the range 5.5Ð6.4. The parameters are maximum, additive and independent of each study of this variable was carried out by adjusting the pH value variable in the concentration range 0.06Ð0.1% for Triton X-100 of each solution in the syringes, which allowed the same value and 5 3 1025-7 3 1025 M for TOPO. The optimum to be obtained in the mixing chamber as checked in the wastes. concentration range of Triton X-100 is higher than the reported Three buffer solutions (ammonium acetate, hexamine and critical micelle concentration (0.018%), which suggests that the imidazole) were assayed to adjust the pH. The ammonium micelles shield the chelates and favour the intramolecular acetate buffer slightly decreased both measurements parame- energy transfer process from the analytes to the lanthanide ion. ters, while the additivity of both systems was poor in the Because the analytes and TOPO solutions were prepared in a presence of the hexamine buffer, which could be attributed to hydroalcoholic medium, the effect of the ethanol content was the fact that hexamine can act as an additional ligand of also studied and it was found that both the initial rate and signal amplitude of both systems are not affected by the presence of ethanol up to a concentration of 6%. Increasing temperatures in Table 4 Determination of BZ and SC in soft drinks the range 20-45 ¡C only caused a very slight increase in the Analyte content/mg ml21 b values of both parameters, so that a temperature of 25 ¡C was selected. BZ SC The initial slopes of the kinetic curves obtained for solutions containing different concentrations of BZ and SC revealed a Samplea CONVc MLRc CONVc MLRc first-order dependence on each analyte. Under optimum conditions, the other reactants showed a pseudo-zero order 1 312 ± 8 305 ± 6 178 ± 5 185 ± 4 2 362 ± 5 358 ± 7 172 ± 5 163 ± 7 dependence. Thus, the following kinetic equations are pro- 3 326 ± 9 315 ± 6 183 ± 9 186 ± 6 posed: v = k [BZ] and vA = kA [SC], where v and vA are the rates 4 373 ± 5 364 ± 4 188 ± 4 194 ± 5 of formation of the complexes, k and kA the corresponding 5 298 ± 4 305 ± 3 — — conditional constants and [BZ] and [SC] the concentration of 6 312 ± 7 309 ± 4 — — each analyte. 7 450 ± 3 471 ± 2 — — 8 — — 172 ± 6 185 ± 3 9 — — 192 ± 1 186 ± 2 Analytical features a Trade mark, manufacturer and composition of samples: 1, La Casera, La Casera, S.A., carbonated water, , orange juice, citric acid, sweeteners: sodium saccharin and ; sodium citrate; stabilizers: arabic The kinetic curves obtained under the optimum conditions for gum, wood colofonia esters and pectins; preservative sodium benzoate; BZ and SC, with 284 and 545 nm as the excitation and emission natural aroma; artificial colours: , Sunset Yellow and ; wavelengths, respectively, were processed using two quantita- 2, Rioco, Refrescos del Sur Europa, S.A, carbonated water, glucose syrup, tion methods involving initial-rate and equilibrium signal orange juice, citric acid, sweeteners: sodium saccharin and sodium measurements, which were obtained in ca. 0.1 and 5 s, cyclamate; sodium citrate; stabilizer: arabic gum; preservative: sodium benzoate; artificial aroma; artificial colours: carotenoids; 3, La Casera, La respectively. Table 1 summarizes the features of the corre- Casera, S.A., carbonated water, sugars, lemon juice, citric acid, sweeteners: sponding methods developed for the two analytes. As can be sodium saccharin and sodium cyclamate; sodium citrate; stabilizers: arabic seen, the determination is about ten times more sensitive for BZ gum, wood colofonia esters and pectins; preservative sodium benzoate; than for SC. The correlation coefficients and standard errors are natural aroma; artificial colours: Quinoline Yellow; 4, Coca Cola, the Coca- indicative of good calibration linearity for both systems. The Cola company, carbonated water, sweeteners: acesulphame K, detection limits were obtained according to IUPAC recom- and sodium saccharin; acidifiers: phosphoric and citric acids; artificial 19 aroma; artificial colours: ammonium sulfate of caramel; preservative: mendations. These results were obtained by using the time- benzoic acid and caffeine; 5, Fanta, the Coca-Cola company, carbonated resolved mode, with a delay time of 0.1 ms and a gate time of water, sugars, orange juice (8%), acidifier: citric acid; sodium citrate; 5 ms, which allowed the scatter to be eliminated. Table 1 also stabilizers: guar gum; preservative sodium benzoate; antioxidant: ascorbic includes the precision (RSD) of the proposed methods, which acid; aroma; artificial colours: carotenoids; 6, Fanta,the Coca-Cola was studied at two concentrations of each analyte and ranged company, carbonated water, sugars, orange juice (8%), acidifier: citric acid; from 2.3 to 3.0%. sodium citrate; stabilizers: guar gum; preservative: sodium benzoate; antioxidant: ascorbic acid; aroma; artificial colours: Quinoline Yellow; 7, Several potential interferent compounds were assayed in Rioco, Refrescos del Sur Europa, S.A, carbonated water, sugars, glucose order to determine the selectivity of these determinations. Table syrup; acidifiers: benzoic acid; vegetal extracts; aroma; preservative: 2 summarizes the tolerated limits obtained by using kinetic and sodium benzoate; artificial colours: caramel, azorrubin and Ponceau 4R; 8, equilibrium measurements. A substance was considered not to Rioco, Refrescos del Sur Europa, S.A, carbonated water, sugars, glucose interfere at a given concentration if the corresponding analytical syrup; acidifiers: benzoic acid; vegetal extracts; aroma; preservative: parameter obtained in the presence of this substance was within sodium benzoate; artificial colours: caramel, azorrubin and Ponceau 4R; Rioco, Refrescos del Sur Europa, S.A, carbonated water, sweeteners: one standard deviation of the value obtained with the analyte sodium saccharin and sodium cyclamate; acidifiers: citric acid; artificial alone. As can be seen, the compounds assayed cause a similar aroma; artificial colors: caramel and caffeine; 9, La Casera, La Casera, S.A., effect on the determination of both analytes, but the tolerated carbonated water, acidifier: citric acid, sweeteners: sodium saccharin and levels for citric and ascorbic acids were higher for the kinetic sodium cyclamate; sodium citrate; lemon aroma. b Mean of three method than for the equilibrium method. As this selectivity c determinations ± s. CONV, conventional method, MLR, multilinear study was carried out in the presence of 4 mg ml21 of each regression method. analyte, it is foreseeable that the assayed compounds do not

1082 Analyst, 1999, 124, 1079Ð1084 Table 5 Recovery of BZ and SC added to soft drink samples

BZ SC

CONVb MLRb CONVb MLRb Added/ Added/ Samplea mg ml21 Foundc Recovery (%) Found Recovery (%) mg ml21 Found Recovery (%) Found Recovery (%)

1 200 202 ± 3 101.0 191 ± 8 95.5 200 189 ± 5 94.5 192 ± 6 96.0 600 610 ± 5 101.7 596 ± 6 99.3 600 570 ± 12 95.0 579 ± 6 96.5 800 780 ± 10 97.5 776 ± 8 97.0 800 770 ± 8 96.3 786 ± 3 98.3 2 200 195 ± 6 97.5 198 ± 8 99.0 200 203 ± 8 101.5 210 ± 5 105.0 600 580 ± 9 96.7 575 ± 6 95.8 600 596 ± 6 99.3 606 ± 5 98.5 800 816 ± 6 102.0 802 ± 7 100.3 800 714 ± 8 89.3 735 ± 3 91.8 3 200 190 ± 11 95.0 192 ± 6 96.0 200 208 ± 5 104.0 203 ± 4 101.5 600 605 ± 5 100.8 606 ± 3 101.0 600 615 ± 7 102.5 606 ± 3 101.0 800 815 ± 8 101.9 809 ± 4 101.1 800 804 ± 6 100.5 799 ± 4 99.9 4 200 187 ± 9 93.5 186 ± 4 93.0 200 217 ± 7 108.5 205 ± 6 102.5 600 578 ± 3 96.3 569 ± 6 94.8 600 628 ± 6 104.7 615 ± 7 102.5 800 769 ± 8 96.1 763 ± 6 95.4 800 781 ± 12 97.6 775 ± 6 96.9 5 200 216 ± 7 108.0 211 ± 2 105.5 200 186 ± 8 93.0 180 ± 6 90.0 600 581 ± 3 96.8 578 ± 8 96.3 600 610 ± 2 101.7 605 ± 4 100.8 800 810 ± 6 101.3 803 ± 6 100.3 800 810 ± 4 101.3 803 ± 3 100.4 6 200 212 ± 9 106.0 215 ± 5 107.5 200 202 ± 9 101.0 198 ± 9 99.0 600 598 ± 4 99.7 601 ± 8 100.2 600 577 ± 5 96.2 573 ± 4 95.5 800 775 ± 2 96.9 778 ± 3 97.3 800 835 ± 5 104.4 828 ± 5 103.5 7 200 198 ± 5 99.0 195 ± 6 97.5 200 188 ± 7 94.0 179 ± 4 89.5 600 615 ± 4 102.5 607 ± 3 101.2 600 586 ± 13 97.7 580 ± 8 96.7 800 782 ± 1 97.8 775 ± 4 96.9 800 818 ± 9 102.3 808 ± 6 101.0 8 200 208 ± 2 104.0 206 ± 5 103.0 200 213 ± 3 106.5 206 ± 1 103.0 600 587 ± 6 97.8 584 ± 8 97.3 600 597 ± 7 99.5 590 ± 4 98.3 800 780 ± 11 97.5 778 ± 9 97.3 800 820 ± 11 102.5 813 ± 8 101.6 9 200 210 ± 7 105.0 206 ± 6 103.0 200 205 ± 8 102.5 210 ± .3 105.0 600 586 ± 9 97.7 590 ± 8 98.3 600 624 ± 9 104.0 630 ± 6 105.0 800 790 ± 18 98.8 790 ± 7 98.8 800 785 ± 12 98.1 793 ± 7 99.1 a See Table 4. b CONV, conventional method; MLR, multilinear regression method. c mg ml21; Mean of three determinations ± s. interfere in the determination of BZ and SC in soft drink according to the procedure described above. Both BZ and SC samples, taking into account that the maximum tolerated were present in four of the nine samples assayed, while the other concentration showed in the official positive list of additives in five samples only contained one of them. Table 4 summarizes these samples20 is 600 ppm for BZ, 200 ppm for SC, 4000 ppm the concentrations obtained by using the conventional and MLR for cyclamic acid, 150 ppm for caffeine, 10000 ppm for citric methods and the qualitative composition of each sample. The acid and 300 ppm for ascorbic acid. application of the paired t-test21 showed that both methods do The additivity of the initial rate and equilibrium signal not give significantly different values. Table 5 lists the obtained for BZ and SC in mixtures allowed the simultaneous analytical recoveries obtained by both methods, which were determination of both analytes by using the system of the two calculated by adding three different amounts of each additive to equations described under Procedure. This system was resolved each sample and subtracting the results obtained for similarly by using separate standard solutions of each analyte and prepared unspiked samples. The recoveries ranged from 93.0 to standard mixtures of both analytes. In the first instance, the data 108.0% for BZ and from 89.3 to 108.5% for SC. Although the were processed by a BASIC program (conventional method) potential interference of many of the ingredients of the samples and, in the second, by multiple linear regression (MLR method) analysed was not previously studied, the recovery values using a laboratory-developed FORTRAN 77 program.18 Both obtained show the adequate selectivity of the proposed methods were applied to solutions containing known BZ and SC methods. concentrations in different ratios. Table 3 summarizes the results obtained and, also, representative least-squares statistics for comparison of the results obtained for each method with the Conclusions standards. In all instances, the regression coefficients suggest good linearity and the slopes are close to unity. Also, there is not The results obtained in this work show the usefulness of the a significant difference between the results obtained by both simultaneous method described to control the presence of BZ conventional and MLR methods. Thus, both methods can be and SC in soft drinks. In spite of the complex composition of the used for the simultaneous determination of BZÐSC mixtures in + + samples analysed, the method can be directly applied without ratios from 3 1 and 1 9. Mixtures with BZ:SC ratios higher and any previous separation step. This can be ascribed to the spectral lower than this range were also tested but the relative errors and temporal discrimination of the analytical signal achieved by obtained were relatively high, so that the proposed simultaneous the combined use of lanthanide-sensitized luminescence and the methods would not be suitable. One major feature of this time-resolved mode. Also, the use of stopped-flow mixing simultaneous determination is expeditiousness: analytical technique allows the automation of the method and its easy measurements for each mixture can be obtained in only ca. application to the simple and fast determination of these 5 s. analytes in soft drinks. As it has been shown, the two methods applied to the Applications simultaneous determination of both analytes give similar results so that they can be indistinctly used, although the MLR method The simultaneous proposed method was applied to the analysis allows analytical results to be obtained more rapidly and the of several soft drink samples. Each sample was analysed errors obtained are slightly lower.

Analyst, 1999, 124, 1079Ð1084 1083 The authors gratefully acknowledge financial support from the 10 E. P. Diamandis and T. K. Christopoulos Anal. Chem., 1990, 62, Spanish DGICyT (Grant No. PB96-0984). 1149A. 11 J. Georges, Analyst, 1993, 118, 1481. 12 C. D. Tran and W. Zhang, Anal. Chem., 1990, 62, 835. 13 S. Peter, B. S. Panigrahi, K. S. Viswanathan and C. K. Mathews, Anal. Chim. Acta, 1992, 260, 135. References 14 E. S. Lianidou and P. C. Ioannou, Clin. Chem., 1996, 42, 1659. 15 S. Panadero, A. Gómez-Hens and D. Pérez-Bendito, Anal. Chim. 1 H. M. G. Doeglass, Br. J. Dermatol., 1975, 93, 135. Acta, 1996, 329, 135. 2 U. Ostermeyer, Dtsch. Lebensm.-Rundsch., 1995, 91, 307. 16 S. Panadero, A. Gómez-Hens and D. Pérez-Bendito, Talanta, 1998, 3 M. Moors, C. R. R. R. Teixeira, M. Jimidar and D. L. Massart, Anal. 45, 829. Chim. Acta, 1991, 255, 177. 17 A. Loriguillo, M. Silva and D. Pérez-Bendito, Anal. Chim. Acta, 4 U. Hagenauer-Hener, C. Frank, U. Hener and A. Mosandl, Dtsch. 1987, 199, 29. Lebensm.-Rundsch., 1990, 86, 348. 18 V. González, B. Moreno, D. Sicilia, S. Rubio and D. Pérez-Bendito, 5 C. Borchert and E. Krueger, Monatsschr. Brauwiss, 1989, 42, 438. Anal. Chem., 1993, 65, 1897. 6 M. Veerabhadrarao, M. S. Narayan, O. Kapur and C. S. Sastry, 19 G. L. Long and J. D. Winefordner, Anal. Chem., 1983, 55, 712A. J. Assoc. Off. Anal. Chem., 1987, 70, 578. 20 A. Madrid, Manual de utilización de los aditivos en alimentos y 7 L. P. Valenti J. Assoc. Off. Anal. Chem., 1985, 68, 782. bebidas, Madrid Ed., Madrid 1987, p. 294. 8 Y. X. Zhu, Y. Zhang, S. X. Li and X. Z. Huang, Fenxi Huaxue, 1995, 21 J. C. Miller and J. N. Miller, Statistics for Analytical Chemistry, Ellis 23, 1313. Horwood, Chichester, 1984, p. 56. 9 T. S. Tibbels, R. A. Smith and S. M. Cohen, J. Chromatogr., 1988, 441, 448. Paper 9/02402F

1084 Analyst, 1999, 124, 1079Ð1084