A Single Calibration Graph for the Direct Determination of Ascorbic

Massimo Zorzi, Paolo Pastore,* and Franco Magno. Department of Inorganic, Metallorganic and Analytical Chemistry, University of Padova, via Marzolo 1,...
0 downloads 0 Views 77KB Size
Download PDF
Anal. Chem. 2000, 72, 4934-4939

A Single Calibration Graph for the Direct Determination of Ascorbic and Dehydroascorbic Acids by Electrogenerated Luminescence Based on Ru(bpy)32+ in Aqueous Solution Massimo Zorzi, Paolo Pastore,* and Franco Magno

Department of Inorganic, Metallorganic and Analytical Chemistry, University of Padova, via Marzolo 1, 35131, Padova, Italy

Ascorbic (H2A) and dehydroascorbic (DA) acids were for the first time directly determined in a single chromatographic run by means of the tris(2,2′-bipyridine)ruthenium(II) (Ru(bpy)32+) based electrogenerated chemiluminescence (ECL) detection. For the first time, it was demonstrated that DA, a nonelectroactive compound, is ECL active and is responsible for the ECL behavior of H2A. This fact, together with the lack of a DA standard, suggested the use of a calibration graph obtained for H2A, for determining both analytes. The proven ECL activity of DA, together with literature data relative to the standard redox potentials of the different species coming from H2A, led to a reconsideration of the proposed ECL reaction mechanism for H2A. The role of the OH- ion in the reaction mechanism of the two analytes appeared to be crucial. H2A and DA could be separated by a suitable C18reversed-phase HPLC column using an aqueous 30 mM H3PO4 solution as the mobile phase. The optimal ECL response was achieved by polarizing the working electrode at 1.150 V vs SCE (standard calomel electrode) (oxidation diffusion limiting potential for both H2A and Ru(bpy)32+). The Ru(bpy)32+ solution, at pH 10 for carbonate buffer, was mixed to the eluent solution in a postcolumn system, obtaining, still at pH 10, the final 0.25 mM Ru(bpy)32+ concentration. The detection limit found for the two analytes was 1 × 10-7 M. The method was successfully applied to the determination of the analytes in a commercially available orange fruit juice. L-ascorbic (H2A) and dehydroascorbic (DA) acids (1) are the two forms of vitamin C, the first discovered vitamin.1

The importance of H2A as a remedy in many diseases2 and as an antioxidant compound in the food industry is well-recognized. * Corresponding author. E-mail: [email protected]. Phone: ++3949-827-5182. Fax: ++39-49-827-5161.

4934 Analytical Chemistry, Vol. 72, No. 20, October 15, 2000

DA, the oxidized form of H2A, can be present, even at high concentration levels, in many tissues and physiological fluids, especially in the presence of several diseases and of aging processes,3,4 and in food products.5,6 Its quantification, therefore, is very important both to evaluate some pathologic conditions and to evaluate the state of conservation of food. DA exists in different forms,7 being quite unstable both in solid crystalline and in solution forms. Moreover, it hydrolyzes irreversibly to 2,3 diketogulonic acid (DKA) and then, in cascade, to many other products depending on the pH of the solution.8,9 Analytical techniques for the determination of H2A and DA have been thoroughly reviewed,10,11 but no one-run direct analytical method has been yet proposed. Colorimetric titrations with different reactants such as the Tillmans’ reagent or Fe(III) are effective to determine H2A but suffer from many interferences11 and do not allow the determination of DA. A colorimetric determination of DA and DKA with 2,4 dinitrophenyl hydrazine has been proposed11 but the determination of H2A can be done only by difference after a second run on a solution in which H2A has been oxidized (DKA is actually an interfering species). The two-run procedure has to be adopted also with the fluorimetric method, on the basis of the selective reaction of o-phenylenediamine with DA.12 Enzymatic methods are very sensitive to H2A but do not allow the determination of DA.13,14 Chromatographic methods are potentially able to separate the two analytes but they (1) Vernin, G.; Chakib, S.; Rogacheva, S. M.; Obretenov, T. D.; Parkanyi, C. Carbohydr. Res. 1998, 305, 1-15. (2) Jaffe, G. In Handbook of Vitamins; Machlin, L., Ed.; Dekker: New York, 1990; pp 199-244. (3) Chatterjee, I. B.; Banerjee, A. Anal. Biochem 1979, 98, 368-374. (4) Farber, C. M.; Kanegiser, S.; Stahl, R.; Liebes, L.; Silber, R. Anal. Biochem. 1983, 134, 355-360. (5) Rose, R. C.; Narhrwold, D. L. Anal. Biochem. 1981, 114, 140-145. (6) Ziegler, S. J.; Meyer, B.; Sticher, O. J. Chromatogr. 1987, 391, 419-426. (7) Seib, P. A., Tolbert, B. M., Eds. Ascorbic Acid: Chemistry, Metabolism and Uses; American Chemical Society: Washington, DC, 1982; Chapter 5, pp 101-123. (8) Kimoto, E.; Tanaka, H.; Ohmoto, T.; Choami, M. Anal. Biochem. 1993, 214, 38-44. (9) Kagawa, Y. J. Biochem. (Tokyo) 1962, 51, 134. (10) Barthe`lemy, J. P. Analusis 1996, 24, 95-103. (11) Washko, P. W.; Welch, R. W.; Dhariwal, K. R.; Wang, Y.; Levine, M. Anal. Biochem. 1992, 204, 1-14. (12) Moeslinger, T.; Brunner, M.; Sieckermann, P. G. Anal. Biochem. 1994, 221, 290-296. (13) Uchiyama, S.; Kobayashi, Y.; Suzuki, S. Anal. Chem. 1991, 63, 2259-2262. (14) Stevanato, R.; Avigliano, L.; Finazzi-Agro, A.; Rigo, A. Anal. Biochem. 1985, 149, 537-542. 10.1021/ac991222m CCC: $19.00

© 2000 American Chemical Society Published on Web 09/14/2000

still suffer from many drawbacks: gas chromatography requires long and tedious lyophilization and derivatization steps15 while liquid chromatography has no suitable detector16 for the direct determination of DA. The only possibility is the reduction of DA to H2A with sulfur-containing compounds such as D,L-homocysteine17 or dithiotreitol (2,3 dihydroxybutane 1,4 dithiol, DTT)18 and others11 and then the detection of H2A either by UV or by electrochemical detection.19 Recently, the detection of H2A via Ru(bpy)32+-based ECL has been proposed20 but no attention has been paid by the authors to DA. ECL is the production of light as the result of highly energetic electron transfer reactions between reactants electrochemically generated. In particular, Ru(bpy)32+ (used throughout our study), a good ECL reagent in aqueous solutions, can be used to determine organic and inorganic substrates able to form very reactive intermediates, such as oxalic acid,21 peroxodisulfate,22 and others.23 The lowest Ru(bpy)32+ excited state (2.12 eV) has a lifetime sufficiently short (about 0.5 µs) to be analytically useful24 and can be produced by reacting the electrogenerated Ru(bpy)33+ or Ru(bpy)3+ with suitable substrates which act as electron donors or acceptors. In this case, a single oxidation/reduction step is required to produce either Ru(bpy)33+ or Ru(bpy)3+ species that act as electrocatalysts.21,22 In particular, when the Ru(bpy)33+ species is produced in aqueous media (E° (Ru(bpy)33+/Ru(bpy)32+) ) 1.05 V vs SCE), the production of luminescence due to the reaction with OH- has to be taken into account.25,26 Using the UV detection as the reference tool, this paper suggests the use of the Ru(bpy)32+-based ECL detection system to determine directly both the forms of the vitamin C, H2A and DA, previously separated with a suitable chromatographic procedure. EXPERIMENTAL SECTION Reagents. All the chemicals were of analytical grade. Solutions were prepared with high-purity water produced with a Millipore Milli-Q apparatus (Bedford, MA). In particular, the following compounds were used without further purification: Ru(bpy)3Cl2‚ 6H2O supplied by Fluka (Fluka, Buchs, Switzerland); H3PO4, CH3COOH, NaHCO3, NaOH, (COOH)2‚2H2O, all supplied by Carlo Erba (Milan, Italy); L-H2A, DA, and DTT supplied by Aldrich (Milwaukee, WI). The eluent solution for the reversed-phase separation and the postcolumn solution were solutions of 30 mM H3PO4 and 0.5 mM Ru(bpy)32+ in 0.2 M carbonate buffer (pH 10), respectively. The L-H2A and DA solutions were prepared in a degassed and cooled 30 mM H3PO4 solution, to increase their stability. In the ion-exclusion chromatographic separation, a 0.4 mM H3PO4 solution was used as the eluent. (15) Deutsch, J. C. Anal. Biochem. 1998, 260, 223-229. (16) Koshiishi, I.; Imanari, T. Anal. Chem. 1997, 69, 216-220. (17) Hughes, R. E. Biochem. J. 1956, 64, 203-208. (18) Okamura, M. Clin. Chim. Acta 1981, 115, 393-405. (19) Sofic, E.; Riederer, P.; Burger, R.; Gsell, W.; Heuschneider, G. Fresenius’ J. Anal. Chem. 1991, 339, 258-260. (20) Chen, X.; Sato, M. Anal. Sci. 1995, 2, 749-754. (21) Rubinstein, I.; Bard, A. J. J. Am. Chem. Soc. 1981, 103, 512-516. (22) White, H. S.; Bard, A. J. J. Am. Chem. Soc. 1982, 104, 6891-6895. (23) Gerardi, R. D.; Barnett, N. W.; Lewis, S. W. Anal. Chim. Acta 1999, 378, 1-41. (24) Juris, A.; Balzani, V.; Barigelletti, F.; Campagna, S.; Belser, P.; Von Zelewsky, A. Coord. Chem. Rev. 1988, 84, 85-277. (25) Hercules, D. M.; Lytle, F. E. J. Am. Chem. Soc. 1966, 88, 4745-4746. (26) Brune, S. N.; Bobbitt, D. R. Talanta 1991, 38, 419-424.

The ECL emission of Ru(bpy)32+ as a function of pH was studied using a suitably buffered 0.25 mM complex solution. The buffers chosen were 0.01 M H2SO4 (pH ≈ 1.5), H3PO4/NaOH (pH 2, 3, 7, 8), CH3COOH/NaOH (pH 4, 5, 6), NaHCO3/NaOH (pH 9, 10, 11), all 0.2 M. Electrochemical Apparatus and Procedures. An EG&G Princeton Applied Research model 175 function generator (Princeton, NJ) connected to a homemade three-electrode potentionstat was used for the amperometric measurements. The ECL thinlayer flow cell, built in our laboratory, is composed of a threeelectrode amperometric cell (2-mm diameter Pt working, stainless steel counter, and SCE reference electrodes) faced to a Hamamatsu model 5600-04 photomultiplier (Shimokanzo, Japan) equipped with a 900 V dc power supply. The working and counter electrodes, together with the inlet and outlet pipes, were sealed in a Struers Epofix HQ two-component epoxy resin (Copenhagen, Denmark) polymerized in a suitable mould. The SCE reference electrode was externally connected to the cell via the outlet tube. The photomultiplier tube, placed above the working electrode, was separated from it by a 2-mm-thick glass window. A hole in a 200-µm spacer created the void volume of the thin-layer flow. The ECL system was placed in an iron dark box which acted also as a Faraday cage. The constant-potential macroelectrolysis experiments were performed with a EG&G Princeton Applied Research (Princeton, NJ) model 273 potentiostat using Pt gauzes as working and counter electrodes and a SCE as the reference electrode. All the pH measurements were made with a Orion model 720-A pHmeter (Boston, MA) equipped with an Ingold glass electrode. Chromatographic Apparatus and Procedures. The chromatographic/postcolumn system was comprised of a Waters ActIon Analyzer (Milford, MA) chromatograph for the eluent supply and by a Dionex (Sunnyvale, CA) gradient pump for the postcolumn solution supply. A “tee” junction allowed the mixing of the two solutions. A single bead string reactor and a delay tube (1mL volume) were inserted between the “tee” junction and the detector to allow a correct homogenization of the solutions. The two tubes were placed into a thermostatic device. An injection loop of 20 µL was always used. The analytical column was a 3-µm packing 150 × 40 i.d. mm Violet Nucleosil C18. The UV detector was a Waters UV-vis photodiode array model 996 (Milford, MA). The Waters system ran under the Waters Millenium software. ECL signals were recorded with a Spectra-Physics Integrator (San Jose, CA) model 4290. A 150 × 4 i.d. mm Dionex ICE-AS9 column was used to confirm some of the results obtained. RESULTS AND DISCUSSION ECL Behavior of the Ru(bpy)32+/H2A System in Aqueous Solution. Before acquiring the ECL behavior of the analytes, preliminary tests dealing with the ECL emission of the OH- ion in flowing solutions were carried out to monitor the background. Figure 1a reports, as a function of pH, the emission due to the overall process27

Ru(bpy)33+ + OH- ) *Ru(bpy)32+ + 1/2 H2O + 1/4 O2 (2) The emissions, relative to suitably buffered 0.25 mM Ru(bpy)32+ (27) Creutz, C.; Sutin, N. Proc. Natl. Acad. Sci. U. S. A. 1975, 72, 2858-2862.

Analytical Chemistry, Vol. 72, No. 20, October 15, 2000

4935

Figure 1. (a) ECL of Ru(bpy)32+ as a function of pH. Experimental conditions: Ru(bpy)32+ concentration, 0.25 mM buffered at different pH values. Solution flow rate, 0.6 mL/min; electrode potential 1.15 V vs SCE. (b) ECL of 20-µL injections of H2A, 0.05 mM, as a function of pH. Other experimental conditions are the same as for the curve a. Inset: relative stability of H2A as a function of pH at 25 °C. Experimental conditions: H2A concentration, 0.1 mM; injection volume, 20 µL; solution flow rate, 0.6 mL/min (permanence time in the mixing coil of the analyte, 1.5 min); UV detection at 245 nm for pH 2, 4.5, 5, 6, and at 265 nm for higher pHs. The ordinates are the ratios between actual readings and that of a nondecomposed sample.

solutions, were monitored polarizing the Pt electrode at 1.150 V vs SCE, a diffusion-limiting potential value for the production of Ru(bpy)33+. On the basis of the standard potential values involved, E° (Ru(bpy)33+/Ru(bpy)32+) ) 1.05 V vs SCE and E°′(O2/H2O)pH 10 ) 0.398 V vs SCE, the total amount of energy available is

-∆G° ) F(E° (Ru(bpy)33+/Ru(bpy)32+) E°′(O2/H2O)pH 10) ) 0.652 eV This energy is much lower than that required to populate the excited state of Ru(bpy)32+ (2.12 eV), so that a conceivable reaction mechanism is cumbersome to propose. A tentative explanation could be, as suggested by Martin and co-workers,28 a mechanism involving the presence of very reactive intermediates coming from OH-, such as OH‚ or O2-‚, and having a lifetime long enough to react with Ru(bpy)33+. The intermediate so produced reacting with a second Ru(bpy)33+ also accounts for the 2:1 stoichiometry. These considerations will be used in a next section to explain the ECL of DA. Figure 1b shows the ECL emission (each point represents the average of three independent measurements) relative to injections of a 5 × 10-5 M H2A solution into Ru(bpy)32+ flowing solutions. All the experimental conditions were the same as in Figure 1a. It must be remarked that the chosen working potential realized oxidation diffusion limiting conditions also for H2A at all the pH values, as demonstrated by cyclic voltammetry measurements. The reported values are the differences between the recorded signal and the corresponding background. As is pointed (28) Martin, J. E.; Hart, E. J.; Adamson, A. W.; Halpern, J. J. Am. Chem. Soc. 1972, 94, 9238-9240.

4936

Analytical Chemistry, Vol. 72, No. 20, October 15, 2000

out in the figure, the maximum emission is observed at pH 10 and the ECL reaction between Ru(bpy)33+ and the analyte does not occur significantly at pHs lower than 8, which is different from that reported by Chen and Sato20 who observed the maximum emission at pH 7. Usually, Ru(bpy)32+ ECL can be used in the anodic potential range for the detection of vicinal dicarbonyl or R-hydroxyl carbonyl compounds. In analogy with the behavior of these substrates, a reaction mechanism between H2A and Ru(bpy)33+ has been proposed20 on the basis of the hypothesis that the real partners of the ECL reaction were Ru(bpy)33+ and the protonated radical anion of ascorbic acid produced at the electrode surface, HA‚.20 Actually, literature data relative to the standard potentials of the redox couples made by H2A and its oxidation intermediates29 suggest that the proposed mechanism is not true, because those potentials are too high (see Table 1 in ref 29) to give a ∆G° for the redox reaction sufficient to promote the Ru(bpy)32+ to the excited state. For this reason, the mechanism must follow a different pathway. In particular, as it will be argued in the next sections, we believe that the ECL mechanism starts from the electroinactive species DA, electrochemically produced at the electrode surface, then involves the OH- ion which has to play a crucial role. Therefore, at present, the rising part of the peakshaped emission reported in Figure 1b cannot be explained by the increased reducing power of H2A due to the pH increase (the pKa values of H2A are 4.2 and 11.6, respectively). We note, on the contrary, that the emission trend decreases at pH values greater than 10 in concomitance with the reduced stability of H2A above pH 10, for a residence time of the analyte in the mixing coil of 1.5 min (see the inset of Figure 1). Another reason of the decrease of the emitted light above pH 10 can be ascribed to the reduced availability of Ru(bpy)33+ due to the competitive reaction with the OH- ion which assumes quite high concentration levels at these high pHs. Aside from the knowledge of the real reaction pathway, experimental findings show that pH 10 has to be chosen for the detection of H2A because this value allows an optimal balance between the availability of OH- required for the reaction between Ru(bpy)32+ and H2A (see above), the parallel decomposition of H2A, the competitive reactions with OH- at higher pH values. The choice of the optimal applied potential was done by means of the ECL-hydrodynamic voltammetric responses. The ECL signal develops in parallel to the anodic voltammetric curve of Ru(bpy)32+, reaches a maximum at 1.000-1.150 V, decreases first between 1.150 and 1.250 V, and then decreases more markedly above 1.250 V, where the oxygen evolution may decrease the effective electrode area. The electrochemical oxidation of OH- at these high potentials could also be responsible for the overall ECL signal decrease, suggesting again its importance as a reactant in the ECL reaction. The potential of 1.15 V, limiting diffusion oxidation potential of both Ru(bpy)32+ and H2A, was therefore chosen. The optimized conditions just found were used for the calibration graph of H2A, which was linear in the range 3 × 10-7 M to 1 × 10-4 M (a ) 14.2 ( 16.9 mV; b ) 1.6 × 107 ( 0.2 × 107 mV/M; r2 ) 0.9997 for Y ) a + bX). Since the standard deviation of the a parameter is found to be greater than the estimated parameter, the slope of the calibration plot is the response factor of H2A. The repeatability of the single sample analysis is quite good (better than 3%) provided that the pH and the Ru(bpy)32+ (29) Tur’ian, Y. I.; Kohen, R. J. Electroanal. Chem. 1995, 380, 273-277.

Figure 3. Chromatogram relative to a 20-µL injection of DA nominally 0.05 mM with ECL detection. Chromatographic conditions: column, Violet Nucleosil C18; eluent, H3PO4, 30 mM; flow rate, 0.3 mL/min; postcolumn Ru(bpy)32+ concentration, 0.5 mM buffered at pH 10; flow rate, 0.3 mL/min; electrode potential, 1.15 V vs SCE.

Figure 2. Chromatogram relative to a 20-µL injection of DA nominally 0.05 mM with UV detection. Chromatographic conditions: column, Violet Nucleosil C18; eluent, H3PO4, 30 mM; flow rate, 0.3 mL/min; postcolumn flow rate, 0.3 mL/min. (a) direct detection at 228 nm (the postcolumn solution was the same as the eluent); (b) detection at 267 nm after postcolumn reaction with DTT buffered at pH 8.

concentration are always the same. The detection limit computed with the Hubaux-Vos method30 is 1.1 × 10-7 M. This value places the described procedure among the most sensitive ones reported in the literature for the considered analyte. Considerations on the DA Standard Solutions. The analytical approach to the DA determination was quite different because, as a matter of fact, DA, also in solid crystalline form, is not a standard. Indeed, a nominally pure solution of DA, even freshly prepared from the solid, shows many chromatographic peaks when an UV detector at 228 nm is used,31 see Figure 2a. We tried to recognize which peaks were the real DA species analyzing the DA sample by means of a procedure reported in the literature.6 In the described method, DA, chemically produced by oxidation of H2A with Br2, is selectively reduced with DTT at pH 8 back to H2A, which is spectrophotometrically detected at 267 nm. Taking advantage of the reducing properties of DTT, it was demonstrated that among the species present in the solution, prepared dissolving the solid DA, only two were reduced to H2A, as shown in Figure 2b. The attribution was done by the very close matching of their UV spectra, obtained with the photodiode array detector, with that of an authentic H2A sample. This result suggests that a solution of DA made from the solid is not a standard solution since DA is partially decomposed and two forms of it regenerate H2A. In particular, the UV spectra recorded during the chromatogram reported in Figure 2a show that the shorter (30) Vanatta, L. E.; Coleman, D. E. J. Chromatogr., A 1997, 770, 105. (31) Lloyd, L. L.; Warner, F. P.; White, C. A.; Kennedy, J. F. Chromatographia 1987, 24, 371-376. (32) Kalianasundaram, K. Coord. Chem. Rev. 1982, 46, 159-244.

retained form of DA (DA1) has a weak UV absorption with a maximum wavelength at 228 nm, while the longer retained form (DA2), eluted only 15 s before H2A, is characterized by a weak absorption at wavelengths shorter than 220 nm. The use of the ion chromatographic approach with the postcolumn DTT reaction system confirmed these findings, showing the two mentioned peaks in an inverted elution sequence with the H2A peak well separated from them. However, the use of the IC system for the analysis of these solutions could not be proposed because an oncolumn H2A decomposition was markedly evident. When an aged or an oxidized solution of H2A is analyzed, only the DA1 peak is found, proving once more that the solid “standard” is inadequate to quantify DA. The proof of the formation of only the DA1 form in the aging process was done with an experiment consisting of the two following chomatographic runs. Run 1. A H2A standard solution was analyzed with the postcolumn DTT system after an aging period of 24 h. Two peaks appeared corresponding to DA1 reconverted to H2A and to the residual H2A plus the possibly overlapped DA2 form reduced to DA. The peak area of this last peak was 4.03 × 107 au s (average of five injections). Run 2. The same partially oxidized solution was analyzed in the same conditions but in the absence of DTT. Only one peak corresponding to the H2A species was observed at λ ) 267 nm, and its area was 4.01 × 107 au s (average of five injections). A t-test demonstrated that the two peak areas were equal and that, therefore, the only DA form present in the aged H2A solution was the DA1. These results were again confirmed by the ion chromatographic analysis performed with a Dionex column in which the DA2 form (eluted practically with the dead volume) was absent. ECL Behavior of the Ru(bpy)32+/DA System in Aqueous Solution. On the basis that DA is the oxidation product of H2A and of its intrinsic chemical nature (vicinal polycarbonyl compound or polyhydroxyl compound if hydrated), we tested the possible ECL signal due to DA using the ECL conditions optimized for H2A. The resulting chromatogram, reported in Figure 3, shows two ECL peaks with retention times coincident with those relative to H2A generated by the DA-DTT postcolumn system. This result points out that DA also is ECL active even though not electroactive. Since DA standards are not available, its Analytical Chemistry, Vol. 72, No. 20, October 15, 2000

4937

Figure 4. Conversion of H2A into DA1 by controlled potential macroelectrolysis. Chromatographic conditions as in Figure 3. H2A concentration, 0.1 mM. (___) 0% electrolysis, 10% electrolysis, (-‚-‚), 30% electrolysis (- - -).

quantification was tried with a two-step experiment starting from a standard solution of H2A: (i) a freshly prepared 10-4 M H2A solution, made in cooled and degassed eluent, was injected, and the ECL peak area was then recorded; (ii) 20 mL of the same solution were electrolyzed at 0.6 V vs SCE up to 10% and then up to 30% of the molar quantity. Deeper electrolyses involved longer times with decomposition of both H2A and DA. The two resulting solutions were then chromatographed by using the ECL detection system. As the electrolysis of H2A produces DA, the chromatograms showed only two well-separated peaks, one relative to the residual H2A and the second referring to DA1. The decrease of the H2A peak area was very close to the expected 10 and 30% levels, respectively, and was exactly equal to the area of the new DA1 peak (see Figure 4). By repeating this experiment with five aliquots of the same H2A solution, each electrolyzed up to 10% of consumption of H2A, the constancy of the sum of the peak areas of DA and H2A was better than 3%. By changing the H2A concentration to 10-5 and 10-6 M, the same results were obtained. All these facts point out four important things: (i) also in the case of the H2A ECL detection, the real ECL active species is DA or a related species; (ii) DA can be directly quantified, via the ECL procedure, together with H2A, in a single chromatographic run without any postcolumn reduction reaction; (iii) the response factors being the same, a single calibration plot relative to H2A, expressed in terms of peak area vs H2A concentration, can be used to quantify both the analytes, thus avoiding the necessity of a standard for DA. The only requirement is the knowledge of the retention time of DA, which can be easily obtained from a chromatographic run of either an electrolyzed or an aged H2A standard solution; (iv) it is reasonable to think that the kinetics of the H2A-DA ECL reactions are the same and that the electrochemical (or chemical by Ru(bpy)33+) oxidation of H2A to DA is not the rate-determining step even if necessary for the detection of H2A. Role of the OH- Ion in the ECL of DA. The following considerations explain the reported results and point out the role of the OH- ion in the reaction mechanism. (i) In total agreement with Figure 1b, it was verified that solutions of H2A in acidicneutral media were rapidly oxidized by a stoichiometric amount 4938 Analytical Chemistry, Vol. 72, No. 20, October 15, 2000

Figure 5. Spectrophotometric titration curve of 15 mL of coulometrically generated 0.9 mM Ru(bpy)33+ with (a) 1 mM H2A and with (b) nominally 1 mM DA in 30 mM H3PO4. Detection wavelength, 670 nm.

Figure 6. Chromatogram relative to a 20-µL injection of a commercially available orange fruit juice with ECL detection. Chromatographic conditions as in Figure 3.

of Ru(bpy)33+ (separately prepared by massive electrolysis of Ru(bpy)32+) without production of light, while the introduction of Ru(bpy)33+ into the alkaline solution produced chemiluminescence which was increased by the addition of H2A. (ii) The analogous oxidation reaction of DA with Ru(bpy)33+ in acidic and neutral media did not occur, at least at short times: only a slow change from the green color of Ru(bpy)33+ to the orange color of Ru(bpy)32+ was observed, most likely owing to the oxalic acid slowly produced by the decomposition of DA.9 The presence of oxalic acid in the solution was suggested by the feeble and continuous chemiluminescence detectable around pH 5-6.21 In alkaline medium the contribution of DA to the overall chemiluminescence was again found. All these results are summarized in Figure 5 which shows the spectrophotometric titration curve, recorded at 670 nm,32 of Ru(bpy)33+ at pH 2 with the two analytes. As is shown in the figure, DA did not reduce Ru(bpy)33+, while the titration end point obtained with H2A corresponded to a 1:2 stoichiometry, underlying once more the production of DA also in acidic media. These findings strongly suggested that the OH- ion plays a fundamental role in the ECL reaction mechanism of the two considered analytes. Its ability to increase the reducing power of H2A appears meaningless since the decrease of the redox potential of H2A with the increase of pH is thermodynamically insufficient to produce ECL. Therefore, in our opinion, the OH- ion must be

regarded as the precursor of a very reactive intermediate, even if, at present, it is hard to say whether the claimed intermediate is connected to the OH- itself, to the ruthenium complex, or to DA. Work is in progress to study this kinetic aspect and to verify the different species of DA. In particular, a HPLC-MS procedure fully confirmed the results reported in this paper demonstrating the presence of different forms of DA in the commercial product and only one form in an aged solution of H2A. According to this study, which will be published in the next months, we have the possibility to suggest a reaction scheme, which is a tentative explanation of the ECL reaction considered

Ru(bpy)32+ S Ru(bpy)33+ + eRu(bpy)33+ + OH- f Ru(bpy)32+ + OH‚ 1 Ru(bpy)33+ + OH‚ f *Ru(bpy)32+ + O2 + H+ 2 H2A f DA + 2e- + 2H+ DA + OH‚ f [DAOH‚] Ru(bpy)33+ + [DAOH‚] f *Ru(bpy)32+ + P *Ru(bpy)32+ f Ru(bpy)32+ + hν In this scheme, DA is in its hydrated hemiketal form (as found by HPLC-MS) which is supposed to react with the very reactive OH‚ species produced in the first step of the ECL reaction of Ru(bpy)33+ with OH-. The species written as [DAOH‚] is undefined so far and should be the real ECL reaction partner. In some way, this mechanism is similar to that proposed by Rubinstein and Bard21 for the Ru(bpy)33+/oxalate ECL system, but in this case the OH- species is necessary to start the reaction. A kinetic study to confirm this mechanism will begin soon. Detection of H2A and DA in Real Matrixes. The determination of H2A and DA in a fruit juice was carried out with the proposed method. In particular, a commercially available orange juice has been injected using an on-line filter after a 1:100 dilution,

and the resulting ECL-chromatogram is reported in Figure 6. H 2A and DA were quantified in 2.16 mM (2.4 mM declared) and 0.19 mM (amount not declared), respectively. The chromatogram shows also other unknown peaks, thus demonstrating that the ECL detection can be valid also for the determination of other compounds present in these kinds of real matrixes. Anyway, these species should belong to the R-hydroxyl acids or to R-hydroxyl ketone classes, as this detection technique is specific for these kinds of compounds. CONCLUSIONS This work has demonstrated the possibility to determine directly both H2A and DA acids, suitably separated by HPLC, by means of an ECL technique based on Ru(bpy)32+ added in a postcolumn system. It has to be remarked that a single calibration plot obtained with H2A standards is suitable to determine the two analytes since they show the same ECL behavior. The ECL activity of DA, a nonelectroactive compound, suggests that it is the real ECL active species and further indicates a new point of view for the formulation of the ECL reaction mechanism for H2A that seems to produce ECL only because of its previous oxidation to DA. In any case, the role of the OH- ion in the ECL emission seems to be crucial. The detection limit obtained for the two analytes is 1 × 10-7 M and places this detection technique among the most sensitive for these kinds of analytes. The method can be successfully applied to the determination of H2A and DA in a commercially available fruit juice and most likely in other kinds of real matrixes. The ECL detection is valid also for the determination of other compounds present in real matrixes, provided that they are known together with their ECL sensitivity. ACKNOWLEDGMENT We gratefully acknowledge the financial support of the Italian National Council of Research (CNR) and the Ministry of the University and Scientific and Technological Research (MURST).

Received for review October 25, 1999. Accepted July 18, 2000. AC991222M

Analytical Chemistry, Vol. 72, No. 20, October 15, 2000

4939