Choice of the working electrode
Vitamins B1 and V3 are reduced at highly negative potentials , therefore, for their determination, a working electrode with a high over-potential of hydrogen evolution reaction (HER) is required. In case of the Glassy Carbon Electrode (GCE) and Silver Annular Band Electrode (AgABE), a rapid increase in current was observed as the potential shifted towards more negative values (Fig. 1a). These high background current values will impair the determination of compounds reduced in this potential range, including VB1 and VB3. For those reasons, no peak of VB3 reduction was observed for the three mentioned electrodes. A completely opposite situation was detected in case of the CGMDE, which provided a well-defined peak of VB3 reduction, low background current, and high over-potential of HER. Those unique features, combined with the possibility to precisely control the electrode surface and to renew it before the recording of each subsequent voltammogram, provided high repeatability and reproducibility of the measurements. This was essential to compare the results obtained in separate experiments. Among tested supporting electrolytes, 0.1 mol dm−3 acetate buffer of pH 6.0 provided the best signal-to-noise ratio and the highest sensitivity. Furthermore, due to the high over-potential of the HER of CGMDE in this buffer, it facilitated the interpretation of the VB1 reduction peak.
Under previously optimized conditions, quantitative measurements for VB1 and VB3 separately were performed (Fig. 1b, c), based on which the figures of merit were calculated. The obtained values, summarized in Table 1, indicate the superiority of the VB1 determination protocol in terms of the sensitivity, linear range, as well as LOD and LOQ values. In the case of VB3, the deviation from linearity was observed for the concentrations above 1.0 mg dm−3, as depicted in the inset in Fig. 1b. Moreover, the VB3 reduction peak is wider (half-width 73 mV) than the peak of VB1 reduction (40 mV), and the sensitivity ratio of VB1 and VB3 determination protocol is equal to 1.56. These features are going to be crucial at the stage of the separation of overlapped peaks.
Figure 2 presents the DP voltammograms for simultaneous reduction of VB1 and VB3 for the three cases described in the Measurement procedure section. It points out how the presence of a compound with a peak potential close to that of the studied compound can hinder the quantitative evaluation of the recorded voltammograms. Moreover, it demonstrates the issues related to various sensitivities towards the two compounds and differences in their concentrations, that is, the two factors that influence the shape of the resulting voltammograms.
Continuous wavelet transform
The proposed mother wavelet ΨV, ΨVB1, and ΨVB3 were first normalized and standardized. The obtained curves, depicted in Fig. 3, were used in the signal separation by means of CWT.
In the next step, the scale of the CWT was selected for each proposed mother wavelet based on the result of the transformation of the DP voltammogram for simultaneous reduction of VB1 and VB3 in the concentration of 1.0 mg dm−3. Among tested scale values, ranging from 10 to 70, the scale 40 for ΨV and the scale 20 in the case of ΨVB1 and ΨVB3 provided the best separation of the signals (Fig. 4). For those scales, there was at least one zero-crossing point in the potential range between the peak potential of VB3 and VB1, thus they were used in the subsequent calculations.
The transformed curves exhibited the increase in the wavelet coefficient CWTx following the rise in the concentration of VB1 and VB3 in the first variant of the simultaneous reduction experiment, as depicted in Fig. 5. This particularly applies to the CWTx values at the reduction potential of VB1 and VB3, which plotted against the concentration of the respective vitamin resulted in the linear calibration functions. The figures of merit, summarized in Table 2, fulfilled the defined criteria for the acceptance of devised protocols for quantitative analyses. Noteworthy, despite omitting the background correction step, the intercept did not differ from 0 in every case. The obtained LOD, LOQ, and linear range are comparable with those presented in Table 1. As anticipated, CWT has not improved the range of linear response.
In case no. 2, where the concentration of VB3 was kept constant, hardly any change has been observed in the transformed signals presented in Fig. 6a in the potential range corresponding to the reduction of VB3. Unfortunately, CWT with the ΨV mother wavelet has not provided the correct intercept value, thus failing to comply with the set criteria. The application of ΨVB1 to extract the voltammetric signal of VB1 provided an ideally linear response with the LOD being the lowest among all established. Similar remarks are accurate for case no. 3, characterized by the constant concentration of VB1 (Fig. 6b). The best results have been obtained for the ΨVB3 mother wavelet, as in this specific case all of the established criteria have been fulfilled. These observations proved the legitimacy of using the experimentally based mother wavelets for specific applications.
Mother wavelets ΨVB1 and ΨVB3, based on the experimentally recorded voltammograms, ensure similar slopes for a given vitamin between the investigated cases. Moreover, those mother wavelets offer lower detection and quantification limits. In case of ΨV, which resembles the shape of the theoretical DP peak, the highest sensitivities (slopes) have been observed, however, it failed to provide acceptable results in some of the examined conditions.
For the potential more positive than -1.3 V, hardly any change in current values was observed in the DP voltammogram of the VB1 reduction (Fig. 7a). Therefore, the first and second derivatives of this voltammogram did not significantly differ from 0 in this potential range. This also caused that the DP voltammogram of simultaneous reduction of VB1 and VB3 overlapped with the DP voltammogram of VB3 reduction (in the absence of VB1). The same applies to the first and second derivatives of the recorded voltammograms. Therefore, for potential more positive than -1.3 V, the DP voltammogram and its derivatives depended only on the concentration of VB3. This allowed applying the peak-to-peak approach to establish the value of the analytical signal for VB3 (denoted as hVB3 in Fig. 7a). Similarly, for the potentials more negative than the potential of VB1 reduction, the derivative voltammograms of simultaneous VB1 and VB3 reduction resembled the derivative voltammograms of the VB1 reduction. For this reason, choosing the value for the analytical signal of VB1 (hVB1) in the manner presented in Fig. 7a allowed accounting only for the contribution of VB1. In the potential range between the maximum of the VB1 and VB3 reduction peaks, the derivative voltammograms were influenced by both tested vitamins, therefore, applying this region for calibration purposes would lead to erroneous results. The disparity in the height of VB1 and VB3 analytical signals was caused by the difference in the peak width, since differentiation enhanced the narrower signal of VB1 reduction.
As illustrated in Fig. 7b, the value of the second derivative increased as the concentrations of VB1 (CVB1) and VB3 (CVB3) were raised (case 1). For VB1, the linear relationship was observed in the whole range of tested concentrations; however, the calibration curve for VB3 was linear only for CVB3 lower than 1.0 mg dm−3. These tendencies, as well as the obtained figure of merit summarized in Table 3, resemble those obtained for calibrations performed separately for VB1 and VB3, which is easily understandable considering the linear nature of differentiation.
Cognately, in case no. 2. the hVB1 values were proportional to the concentration of VB1, whereas the hVB3 remained hardly unchanged (Fig. 8a) and vice versa in case 3 (Fig. 8b). The calibration curves and figures of merit obtained for the three cases are comparable to those established in the measurement performed separately for each vitamin. Moreover, the slopes of the calibration curves for a given vitamin are similar between experiment variants, which proves that the chosen analytical signals hVB1 and hVB3 did not depend on the concentration of VB3 and VB1, respectively. For this reason, the necessity to build complex models with multiple input variables can be avoided. The sensitivity ratio of VB1 and VB3 for the calibrations based on the derivative voltammograms is 4.5 times higher than for the calibration performed for VB1 and VB3 individually (Table 1). Therefore, it can be assumed that quantitative analysis of VB1 should be relatively straightforward and far less problematic than in the case of VB3, for which the range of linear response is narrower. The direct comparison of sensitivities obtained based on the DP voltammograms (Table 1) and their transforms (Table 2 and 3) is not possible, since they are expressed in different units.
Real sample analysis
Finally, the proposed methods of signal separation were applied for the determination of VB1 and VB3 in dietary supplements Plusssz Mg and Plusssz Ginseng. In terms of the content of vitamins, they do not differ from other products of this type available on the market. This is justified in the recommended daily dose of vitamins at the level of 1.1 mg VB1 and 16 mg VB3, which can be administered with one supplement tablet. Considering that and the obtained linear ranges of the reported procedures, it is clear that voltammetric determination of VB1 and VB3 has to be performed separately to comply with the accepted standards of quantitative analysis.
Figure 9a presents the DP voltammograms for the determination of VB3 in the Plusssz Mg supplement. The presence of the large peak corresponding to VB3 reduction and a small inflection resulting from the VB1 reduction is an early sign that the above-mentioned contents of VB1 and VB3 may be conserved. The obtained shape of curves in the wavelet domain is similar to that depicted in Fig. 5a because the same mother wavelet (ΨV) has been utilized during the transformation. Based on the calculated wavelet coefficients, the standard additions plots have been drafted, from which vitamin contents have been determined as the intercept with the horizontal axis. The values reported in Table 4 for CWT are the average of the vitamin contents estimated by means of CWT with the use of ΨV, ΨVB1, and ΨVB3 mother wavelets, whereas their standard deviation has been adopted as the determination error. Relating this to the declared content, the recoveries have been calculated.
The recorded voltammograms for every analysis were also subjected to differentiation, and an example of that is presented in Fig. 9b for the determination of VB1 in the Plusssz Ginseng. Based on the obtained second derivatives, the calibration plots have been drafted (inset in Fig. 9b), and the vitamin contents have been calculated (Tab. 4). Moreover, Fig. 9b perfectly illustrates how differentiation facilitates the interpretation of the overlapped peak appearing at the falling edge of another signal even in the case of a very unfavorable ratio of the peak currents.
All of the obtained vitamin contents, summarized in Table 4, are consistent with the declaration of the manufacturer, which proves the accuracy of the developed protocols. Due to the lack of necessity to perform any manual interpretation step, these procedures are more objective and possible to computerize. Thus, it is an important step towards complete automation of the quantitative analysis.