Abstract
In this work, a PVC tubular ion-selective electrode without internal reference [composed of 6.35% (w/w) of β-cyclodextrin as ionophore, 1.59% (w/w) of potassium tetrakis (p-chlorophenyl) borate as lipophilic additive, and 63.5% (w/w) of 2-fluorophenyl-2-nitrophenyl ether as mediator solvent] was used to accomplish the simultaneous determination of thiamine and pyridoxine in multivitamin products. The electrode responds selectively to thiamine but presents high initial diffusion potential contribution from pyridoxine. This fact was exploited under flow conditions for the quantitative assessment of both species. A multitask flow system was optimized to dilute appropriately real samples and prepare in-line the set of calibrating solutions. The potentiometric signal was modeled by partial least squares regression, enabling the prediction of the content of each vitamin in the range of 1 × 10−3 to 1 × 10−4 mol L−1. Absolute errors of prediction inferior to 0.06 × 10−4 mol L−1 and 0.16 × 10−4 mol L−1 for pyridoxine and thiamine respectively were obtained when compared with the chromatographic method.
Ana R. Pires thanks Fundação para a Ciência e Technologia (FCT) and Fundo Social Europeu (FSE) for the grant (SFRH/BD/25995/2005).
Notes
a These parameters were used for initial potentiometric measurements of externally prepared solutions (steps 1 and 2) and after in-line mixing of solutions at variable ratios (steps 3 and 4). Steps 3 and 5 were used for preparation and measurements of synthetic mixtures and for samples analysis.
a No peaks were achieved under these conditions.
In-line solutions (IS) were prepared throughout dilution of a 0.1 mol L−1 PYD (MS) solution in LiCl carrier solution. Tx: on/off cycle period (s); x: SV switch on time for each Tx (in %); RSD: relative standard deviation (in %, n = 4); RE: relative error toward the externally prepared solutions (in %); ES: 5 × 10−2, and 5 × 10−3 mol L−1 externally prepared solutions.
Notes. SG, Savitzky–Golay filter; MNCN, mean centering; LOG, decimal logarithm; LV, latent variables; RMSECV, root mean square error of cross validation.
STD, standard deviation of model predictions; CL95%, 95% confidence limits for model predictions.