Accepted Manuscript A comparative application of two-way and three-way analysis to three-dimensional voltammetric dataset for the pKa determination of vanillin Zehra Yazan, Sevcan Erden, Erdal Dinç PII: DOI: Reference: S1572-6657(18)30519-8 doi:10.1016/j.jelechem.2018.07.047 JEAC 12529 To appear in: Journal of Electroanalytical Chemistry Received date: Revised date: Accepted date: 11 May 2018 23 July 2018 25 July 2018 Please cite this article as: Zehra Yazan, Sevcan Erden, Erdal Dinç , A comparative application of two-way and three-way analysis to three-dimensional voltammetric dataset for the pKa determination of vanillin. Jeac (2018), doi:10.1016/j.jelechem.2018.07.047 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. ACCEPTED MANUSCRIPT A comparative application of two-way and three-way analysis to three-dimensional voltammetric dataset for the pKa determination of vanillin Zehra Yazana, Sevcan Erdenb, Erdal Dinçc* Faculty of Science, Department of Chemistry, Ankara University 06100, Ankara, Turkey c General Directorate of Mineral Research and Exploration, 06800 Ankara, Turkey T b Faculty of Pharmacy, Department of Analytical Chemistry, Ankara University, 06100 Ankara, IP a CR Turkey ABSTRACT US Two-way and three-way data analysis methods, multivariate curve resolution-alternating squares AN (MCR-ALS) and parallel factor analysis (PARAFAC) were applied to the potential-frequency dataset to estimate the pKa value of vanillin, used as a flavoring agent in foods and M pharmaceuticals. Three-dimensional (3-D) voltammetric plots of vanillin at 12 different pHs in ED the range of 1.3-12.0 and 8 different frequencies in the range of 5-120 Hz were recorded as a function of potential (mV) and frequency (Hz) using square wave voltammetric (SWV) PT technique. The data matrix and data array of the 3D-voltammetric plots were processed by MCR- CE ALS and PARAFAC algorithms to predict the pKa values of the relevant flavor agent. In both MCR-ALS and PARAFAC approaches, the relative concentration profiles with the used pH AC range were used for the determination of the pKa without using extra additional software. In the MCR-ALS and PARAFAC applications, the numerical values for the pKa value of vanillin were computed as 7.91 and 7.97, respectively. The experimental results showed that the SWV technique coupled with the MCR-ALS and PARAFAC approaches have proved to be a quite promising solution to the problem of the chemometric evaluation the pKa of the compound of interest. 1 ACCEPTED MANUSCRIPT Keywords: Multivariate curve resolution-alternating least squares; Parallel factor analysis; Three-dimensional voltammogram; pKa determination; Vanillin *Corresponding author. Tel.: +90 312 215 4886; Fax: + 90 312 213 1081; e-mail: dinc@ankara.edu.tr IP T 1. Introduction CR Vanillin, chemically known as 4-hydroxy-3-methoxybenzaldehyde, has three different functional groups consisting of aldehyde, hydroxyl and ether with the molecular formula C8H8O3 (Fig. 1). US Vanillin, which is the main chemical component of the extract of the vanilla bean, is among the AN most consumed substances in the world. Natural vanillin has been widely used as flavoring in foods (e.g. desserts, sweet foods, ice cream, chocolate, cake, chocolate and biscuit), beverages M (e.g. tea, coffee, beer and milk etc.), pharmaceuticals (e.g. for the preparation of pharmaceutical ED drugs for Parkinson’s disease and in hypertension and many other drugs), health and personal care products (e.g. in cosmetics), and other industries (e.g. in rubbers and plastics and tobacco, Insert in Figure 1 CE PT cigarette paper and its filter). Artificial vanillin, instead of natural vanilla extract, is sometimes used as a flavoring agent for AC the above similar aims. Artificial vanillin is obtained from either guaiacol or from lignin, a constituent of wood, which is a product of the paper industry. As it was pointed out above, vanillin had been used in a wide range of food and pharmaceutical sectors. Therefore, a research on the pKa determination of vanillin will provide a useful material to uncover the acid-base equilibrium in terms of body system when taken together with either foods or drugs. 2 ACCEPTED MANUSCRIPT As it is known, according to the ratio of the unionized and ionized forms in the acid-base equilibrium, a chemical substance exhibits different physicochemical characteristics e.g. absorption solubility, partition coefficient, biological activity (or chemical reactivity) and permeability of membrane, etc. Due to the mentioned reasons, the issue of determining the acid- T base dissociation constants of active compounds used in foods and medicines still attracts the IP interest of scientists working in analytical chemistry and related research fields. This interest is CR to force scientists to develop new analytical techniques, methods, methodologies and approaches, which are easier, faster, cheaper and more reliable, for the determination of the dissociation In previous works, UV-Visible spectrophotometry and US constants of acids or bases. AN potentiometric methods are two of the most commonly used methods for the studies on dissociation equilibriums of weak acids and bases. methodological approaches including spectrophotometry [1-3] and capillary ED some M A literature survey indicated that the pKa value of vanillin in different media was determined by electrophoresis [4]. On the other hand, in IUPAC chemical data series, it was observed that there PT is a study on the pKa value of vanillin in aqueous solution [5]. However, the application of CE chemometric two-way and three-way data analysis approaches with square wave voltammetric technique to the vanillin’s pKa determination was not reported. AC In practice, a conventional strategy based on a graphical relationship between different pH values and related parameter of instrumental measurement methods is a common approach to the estimation of the acid dissociation constant (or pKa). However, the above mentioned methods may not always allow to exactly determine the acidity constant (or pKa on a logarithmic scale) for analyzed active compounds in related foods or pharmaceutical preparations because of experimental complexity, unwanted interference and methodological disadvantages. 3 ACCEPTED MANUSCRIPT Furthermore, these conventional methods do not provide any second order information related to chemical species and their signals in acid-base equilibriums. In recent years, MCR-ALS and PARAFAC approaches for multiway data analysis have received more attention than other chemometric tools for providing useful and easy interpretable T information for resolving complex data systems, which cannot be solve by using conventional IP analytical instrumental methods. Previously published papers on chemical equilibriums indicated CR that there were some studies on the applications of two-way and three-way data analysis methods to the estimation of acid-base dissociation constants and reaction kinetics with pH-dependent US photo-degradation [6-14]. AN In chemometry, some of the most popular algorithms used in the analysis of three-way data structures are parallel factor analysis (PARAFAC) [15,16], Tuker3 [17], direct trilinear M decomposition (DTLD) [18], alternating trilinear decomposition (ATLD) [19], the generalized ED rank annihilation method (GRAM) [20] 20 and curve resolution-alternating least squares (MCRALS) [21]. In this research paper, new applications of three-way and two-way data analysis PT methods, PARAFAC and MCR-ALS to the potential-frequency dataset were demonstrated for CE monitoring the unionized and ionized species and their voltammograms in acid equilibrium of vanillin and estimating the pKa value of vanillin. A good agreement was reported for the pKa AC values obtained from MCR-ALS and PARAFAC approaches. 1.1. Theoretical framework If HA is a weak acid, the following general reaction for the acid dissociation in aqueous solutions can be written as HA ⇌ H + + A− (1) 4 ACCEPTED MANUSCRIPT Where HA is the weak acid and A- is its conjugate base. In the above expression, an equilibrium is established between A- and HA forms, which are expressed as ionized and unionized species. From the equilibrium reaction, the following acid dissociation constant would be obtained [H+ ] [A− ] (2) [HA] IP = T according to the law of mass action. [HA] (3) [A− ] US [H + ] = CR Equation (2) can be formulated as AN If the negative logarithm of both sides of equation (3) is taken, the following HendersonHasselbalch equation for the calculation of the pH value in chemical and biological systems is M obtained as [A− ] ED pH = pK a + log [HA] (4) PT Here [HA] and [A−] are considered as coming from the acid and its salt in relevant system, CE respectively. In Equation (4), if [HA] = [A−], the pH value is equal to pKa (-log Ka). In order to determine pKa value, an appropriate parameter of experimental methods is measured as a AC function of pH. In the implementation of instrumentation techniques in the pKa analysis, a graph is obtained from the relationship between the responses of instrument and different pH values. This graph is used for the pKa determination of weak acid. A similar way is utilized for the application of multiway data analysis technique to get better pKa analysis results than that of traditional instrumental techniques as in this study. 5 ACCEPTED MANUSCRIPT 1.2. Multivariate curve resolution-alternating squares MCR-ALS is a very powerful mathematical tool for the bilinear decomposition of data matrices to get individual pure contributions of analytes in a complex mixture. In practice, MCR-ALS approach has been used in chemistry and related branches of science and industry. The bilinear IP T decomposition of experimental responses (data matrix, X) into pure component contributions can CR be given as X = P FT + E (5) US Where P (n x k) is the potential profile (or voltammetric profile) of the k components on n AN potential points, F (m x k) is the frequency profile of the k components on m frequency points, and E (n x m) is the residual data matrix which is related to model error or random noise. M Decomposition of data matrix X is performed by iterative least squares minimization of ‖E‖ ED under suitable constraining conditions i.e. non-negativity and unimodality in related profiles. PT More explanations and all theoretical details concerning the mathematical apparatus of the MCR- CE ALS iterative procedure can be found in the literature [21-23]. AC 1.3. Parallel Factor Analysis PARAFAC is a mathematical generalization of principal component analysis (PCA) to higher order arrays. Parallel Factor Analysis (PARAFAC) is one of decomposition methods of threeway data (or N-way data) into trilinear components, which are defined by three loading matrices A, B, C with the elements aif, bjf and ckf, respectively. The PARAFAC approach is based on the 6 ACCEPTED MANUSCRIPT minimization of the sum of squares of the residues, eijk in the fitting model [24, 25], which is expressed as follows: X = ∑ + (6) T here af, bf and cf denote the f th columns of the loading matrices A, B and C, respectively. In IP practice, the alternate least squares (ALS) algorithm is used to fit the PARAFAC model of higher CR order data. In the PARAFAC decomposition of three-way dataset, the use of some constraints such as orthogonality and non-negativity, etc. may be necessary for the resolution of the US component profiles of the analyzed mixtures. In some cases, constraining the PARAFAC model AN can be beneficial in terms of the interpretability or the accuracy of the model. In order to identify the optimal number of components, the criteria of core consistency diagnostic tool is described in M the literature [26]. An optimum condition for a PARAFAC fit to obtain unique parameter PT ED assessment is proposed by Kruskal [27] and then elaborated by Sidiropoulos and Bro [28]. CE 2. Materials and Methods AC 2.1. Instrument and software Voltammetric potential-frequency data were collected on a CH Instruments Model 660C connected C4 Cell Stand with working electrode consisting of sepiolite clay, TiO2 nanoparticles and multi-walled carbon nanotubes (SC/TiO2/MWCN) based on use of a carbon paste electrode, which was firstly used in the literature [29], auxiliary electrode consisting of Pt wire (CHI 115) and reference electrode consisting of Ag/AgCl (CHI 111). Square wave voltammetry was used for the oxidation of vanillin on the related modified electrode system. 7 ACCEPTED MANUSCRIPT pH adjustments were performed using a Hanna HI 2211 pH/ORP meter calibrated with buffer solutions (Thermo Scientific). Doubly-distilled and de-ionized water supplied from Human Power I+ Ultra pure water system T was used throughout in this work. IP All data were converted to Microsoft EXCEL files and imported into Matlab by means of a CR special m-file program written in-house in Matlab. The MCR-ALS algorithm written in-house in MATLAB (Math. Works, Natick, MA) was implemented for the deconvolution of potential- US frequency dataset into pure contributions of the analyzed profiles for the evaluation of the pKa AN values of vanillin. M 2.2. Chemicals, reagents and solutions Graphite powder, mineral oil, sepiolite clay, TiO2 nanoparticules, Multiwalled carbon nanotubes ED (MWCNTs) and all solvents were purchased from Sigma. All the reagents were analytical grade PT and used without any further purification. All solutions were freshly prepared with triply distilled water. Britton-Robinson (BR) buffer solutions were prepared by mixing appropriate amounts of CE 0.04 molL-1CH3COOH, 0.04 molL-1 H3BO3 and 0.04molL-1 H3PO4 and then, the buffer solutions AC at 12 different pHs (1.3, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0 and 12.0) were adjusted with additions of 0.1 M NaOH and 0.1 M HCl to stock buffer solutions. For the solution of 1.0x10-3 M vanillin dissolved in ethanol, the voltammograms at 12 different pH values between 1.3-12.0 and at 8 different frequencies between 5.0-120.0 Hz were recorded between -200.0–1400.0 mV. In the experimental applications, each sample solution containing vanillin (1.0x10-3 M, 1 mL) and 9.0 mL of BR buffer solution for each pH in the working pH ranges was pipetted into a voltammetric cell. 8 ACCEPTED MANUSCRIPT Electrochemical experiments were performed by using SWV technique with suitable voltammetric conditions consisting of supporting electrolyte Britton-Robinson (BR) buffer, pulse amplitude of 0.05 V, scan increment of 0.008 V, frequencies of 5, 10, 20, 40, 60, 80, 100, 120 T Hz. IP 3. Results and Discussion CR As it is well known, electroactive compounds (weak acidic compound in our case) in different US pH media give different electrochemical behaviors with diverse wave forms when applying some electrochemical techniques such as cyclic voltammetry, square wave voltammetry (SWV) and AN differential plus voltammetry, etc. In a similar manner, the change of the frequency in the M implementation of the mentioned techniques for the analysis of the relevant compounds in samples exhibits different electrochemical peak forms with varying peak currents. Among the ED electrochemical parameters, pH and frequency are very important particularly to monitor PT electrochemical characteristics of a weak acid (or a weak base) on the working electrode surface. In our research paper, it was observed from preliminary experiments that pH and frequency of CE the substance was influential on the variation of the peak current and waveform (or wave AC potential) of the voltammograms obtained from the oxidation of vanillin on the SC/TiO2/MWCN electrode applying square wave voltammetry. In the following experiments, the peak current of the related weak acid i.e., vanillin at 12 different pHs in the range of 1.3-12.0 and 8 different frequencies in the range of 5-120 Hz at the mentioned working electrode was plotted as a function of potential in the potential range of -2001400 mV. 9 ACCEPTED MANUSCRIPT In order to get three dimensional (3-D) voltammograms (potential x frequency), one-dimensional peak currents obtained from 8 different frequencies (Hz) were collected in a voltammetric twoway data matrix, for each pH medium. Figure 2 shows the 3D-voltammograms of vanillin at 12 different pH values. IP T Insert in Figure 2 CR Recorded 3D-voltammograms were converted into Microsoft Excel files. Two-way matrices of the voltammetric 3D-plots were imported from Microsoft Excel files into Matlab domain and US then they arranged in a three-way array , also named as a tensor, with dimensions 200x8x12 (potential x frequency x pH) as indicated in Figure 3a. In three-way dataset, the columns denote AN peak current measurements that change with potentials and the rows represent intensity M measurements that change with frequency, and the frontal slices correspond to the pH values. Insert in Figure 2 CE PT were illustrated in Figure 3b. ED The sub-structures (frontal slices, horizontal slices and vertical slices) in three-way data array 3.1. Application of MCR-ALS model AC In this application, the augmented data matrices (or two-way data matrices) obtained unfolding voltammetric three-way array were processed using MCR-ALS algorithm providing pure voltammetric profile, frequency profile and relative concentration profile to predict the pKa value of vanillin. Before performing the MCR-ALS application to the analysis of three-way data array, the tensor , which has a 200812 dimension with potential x frequency x pH was unfolded in three different directions, along the row in first mode, along the column in second 10 ACCEPTED MANUSCRIPT mode and along a third direction of the cubic data. As it can be seen in Figure 3c, it is possible to obtain three different unfolding structures, which correspond to two-way data matrices 1 with the dimension potential x (frequency x pH), 2 with dimension frequency x (potential x pH) and 3 with dimension pH x (potential x frequency). In our study, the two-way data matrices 2 and T 3 were considered as augmented matrices for the evaluation of electrochemical behavior and IP pKa value of vanillin. For the deconvolution of data matrix, the MCR-ALS algorithm was CR separately applied to the augmented matrices 2 and 3 matrices, which correspond to 2-D data US matrices, frequency x (potential x pH) and pH x (potential x frequency), respectively. In MCRALS models, the 2 and 3 , matrices were iteratively decomposed into a set of voltammetric AN profile and relative concentration for the unionized (or acid) and ionized (or congregated base) species. In case of the application of MCR-ALS with a non-negativity constraint to the two-way ED were illustrated in Figure 4a and b. M matrix 2 , the resulting profiles for unionized and ionized species of the analyzed compound CE PT Insert in Figure 4 The normalized voltammograms of unionized and ionized species of vanillin were extracted AC from the voltammetric profile presented in Figure 4a and then they were shown in Figure 5. At 12 different pH media, the electrochemical behaviors of unionized and ionized forms obtained with the oxidation of vanillin on the SC/TiO2/MWCN electrode were observed from the normalized voltammograms displayed in Figure5. These normalized voltammograms indicated that the MCR-ALS deconvolution was suitable way to uncover the electrochemical characteristics of vanillin in diverse pH media for considering chemical and biological systems. The relative concentration profile of unionized and ionized of the analyzed substance at eight 11 ACCEPTED MANUSCRIPT different frequencies was given in Figure 4b. It can be seen from this figure that the relative concentration ratio of two species is equal to the frequency of 58 Hz. Insert in Figure 5 In similar manner, MCR-ALS method with a non-negativity constraint was iteratively applied to IP T the deconvolution of the two-way data matrix 3 with dimension pH x (potential x frequency) to CR get voltammetric profile and relative concentration profile for the unionized and ionized (also named as acid and congregated base) forms. Figures 6a and b show the voltammetric profile and US relative concentration profile of both the acid (unionized) and the base (ionized) species, obtained by the MCR-ALS deconvolution of the matrix 3 . In Figure 6a, the electrochemical AN behavior of vanillin can be observed from the normalized voltammograms of two species of the M related compound at eight different frequencies. In the relative concentration, profile of two ED species at 12 different pH values was depicted in Figure 6b. Insert in Figure 6 PT As can be seen from this figure, the fraction of the acid (unionized) and the base (ionized) forms CE was equal when the pH was 7.91± 0.14. In the MCR-ALS implementation, this pH value corresponds to the vanillin’s pKa value, which corresponds to the negative logarithm of the acid AC dissociation constant, Ka (see Table 1). Insert in Table 1 Figure 7 shows the normalized voltammograms extracted from Figure 6a obtained by the application of MCR-ALS model to 3 data matrix. As it was pointed out above, the pKa value obtained from the intersection point of concentration profiles was given in Table 1. 12 ACCEPTED MANUSCRIPT Insert in Figure 7 3.2. Application of PARAFAC model For the demonstration of the electrochemical characteristics and the pKa estimation of vanillin with the PARAFAC method, the peak currents of the vanillin’s oxidation on the IP T SC/TiO2/MWCN electrode were measured as a function of potential, mV and frequency, Hz to CR record the voltammetric 3-D plots in the potential range of -200-1400 mV applying SWV technique (see Figure 2). As explained above, the data sets of two-way matrices obtained from US voltammetric 3-D plots were arranged in three-way array with dimensions 200x8x12 (potential x frequency x pH). The structure of the three-way array (or tensor) with potential in columns AN frequency in rows, and pH in frontal slices was schematized in Figure 3a. In order to estimate the M electrochemical behavior and pKa value of vanillin, some two component PRAFAC models with ED different constraints were iteratively tested for the decomposition of three-way data array into individual contributions of the unionized and ionized forms of the relevant substance using PT alternating least squares (ALS) computation. Then, it was reported from the mentioned CE preliminary test results that un-constraining PARAFAC deconvolution of three-way array was very useful approach for the assessment of the potential profile of voltammograms, the frequency AC profile and relative concentration profile of the unionized (acidic) and ionized (conjugate basic) species of vanillin. In the PARAFAC approach, no pre-treatment processes e.g., centering or scaling, was used for the data analysis. Three loadings were obtained by the PARAFAC decomposition of the tensor into trilinear components. The related loadings correspond to the potential profile of voltammograms, the frequency profile and relative concentration profile for both chemical forms (unionized and ionized species of vanilin), respectively as shown in Figure 13 ACCEPTED MANUSCRIPT 8a-c. In this implementation, the fit of PARAFAC model with two components was done and then the explained variance was found to be 99.98% of the change in three-way data. Insert in Figure 8 In the PARAFAC deconvolution of three-way array, core consistency was reported as 100.0 %. IP T Experimental results indicated that PARAFAC model was satisfactory for the demonstration of CR electrochemical characterization of vanillin and for the prediction of the pKa value of the subjected substance. In Figure 8c, it can be seen the pH point when the ratio of two chemical US species of vanillin in the relative concentration profile was equal in acid-base equilibrium. In this pH point, the pKa value and its corresponding standard deviation for the analyzed compound was M AN listed in Table 1. This pKa value was obtained from the average of three replicate experiments. ED 4. Conclusions A comparative application of the MCR-ALS and PARAFAC algorithms to two-way data array PT and three-way data array of 3D-voltammograms brings new opportunity different approaches for the pKa prediction of vanillin used in foods and pharmaceuticals. In the MCR-ALS and CE PARAFAC deconvolutions of 2-D voltammetric dataset and 3-D voltammetric dataset, AC respectively, the voltammetric profiles in different frequencies and different pH media made it possible to monitor the unionized and ionized forms of vanillin in its dissociation equilibrium and relative concentration profiles giving the fraction of two chemical species against pH values made it possible to predict the pKa value of the analyzed compound without using additional software and chemical treatment e.g. titration. It was concluded that easy interpretable results obtained from SWV combined with MCR-ALS and PARAFAC were due to second order advantage of multiway data analysis methods over traditional pKa determination methods. 14 ACCEPTED MANUSCRIPT Although MCR-ALS and PARAFAC methodologies have different mathematical algorithms, they gave comparable pKa results. This study indicated that the proposed MCR-ALS and PARAFAC were alternative tools to demonstrate the electrochemical behavior of vanillin against frequency and pH changes in chemical and biological systems with small number of experiments IP T to leads to more accurate results. CR Acknowledgments This study was performed at the Chemometrics Laboratory of Faculty of Pharmacy, which was US supported by the scientific research fund of Ankara University (Project Number 10A3336001). AN The authors would like to thank Ankara University for their support regarding Chemometrics Laboratory. Electrochemical part of the experiments was done at Faculty of Science, Department M of Chemistry and it was supported by the scientific research fund of Ankara University (Project AC CE PT ED Number 2005-07-05-094). 15 ACCEPTED MANUSCRIPT References [1] S. Al Arni, A. F. Drake, M. Del Borghi, A. Converti, Study of Aromatic Compounds Derived from Sugarcane Bagasse. Part I: Effect of pH, Chem. Eng. Technol. 33 (2010) No. 6, 895–901. [2] R. A. Robinson and A. K. Kiang, The ionization constants of vanillin and two of its T isomers. Trans. Faraday Soc. 51 (1955) 1398–1402. IP [3] Z. Rasouli, R.Ghavami, Investigating the discrimination potential of linear and nonlinear spectral multivariate calibrations for analysis of phenolic compounds in their binary and CR ternary mixtures and calculation pKa values. Spectrochim. Acta A Mol. Biomol. Spectrosc. 165 (2016) 191–200. US [4] J. M. Cabot, E. Fuguet, and M. Rosés, Internal Standard Capillary Electrophoresis as a ACS Comb. Sci. 16 (2014) 518−525. AN High-Throughput Method for pKa Determination in Drug Discovery and Development, [5] E.P. Serjeant, B. Dempsey, Ionisation Constants of Organic Acids in Aqueous Solution. M International Union of Pure and Applied Chemistry (IUPAC). IUPAC Chemical Data Series No. 23, New York, New York: Pergamon Press, Inc., 1979. ED [6] M. M. Sena, J. C. B. Fernandes, L. Rover Jr, R. J. Poppi, L. T. Kubota, Application of twoand three-way chemometric methods in the study of acetylsalicylic acid and ascorbic acid PT mixtures using ultraviolet spectrophotometry, Anal. Chim. Acta 409 (2000) 159–170. [7] R. Gargallo, M. Vives, R. Tauler, R. Eritja, Protonation Studies and Multivariate Curve CE Resolution on Oligodeoxynucleotides Carrying the Mutagenic Base 2-Aminopurine, Biophys. J. 81 (2001) 2886–2896. AC [8] C. Rodríguez-Rodríguez, J.M. Amigo, J. Coello, and S. Maspoch, An Introduction to Multivariate Curve Resolution-Alternating Least Squares: Spectrophotometric Study of the Acid–Base Equilibria of 8-Hydroxyquinoline-5-sulfonic Acid, J. Chem. Educ. 84 (7) (2007) 1190-1192. [9] J. Amador‐Hernández, A. Rojas‐Hernández, E. M. Colunga‐Urbina, I.M. De La Garza‐ Rodríguez, M. Velázquez‐Manzanares, L. F. Medina‐Vallejo, New chemometric strategies in the spectrophotometric determination of pKa, Eur. J. Chem. 5 (1) (2014) 1‐5. [10] M. A. El-Monem Hegazy, M. S. Eissa, O. İ. Abd El-Satter, M. Abd El-Kawy, Two and three-way spectrophotometric-assisted multivariate determination of linezolid in the 16 ACCEPTED MANUSCRIPT presence of its alkaline and oxidative products and application to pharmaceutical formulation, Spectrochim. Acta A Mol. Biomol. Spectrosc. 128 (2014) 231-242. [11] A. Y. Faal, T. H. Shayeste, M. Bordbar and B. Jamalyan, Acidity constants determination of triazine dye derivative in the presence of some surfactants by multiwavelength spectrophotometric and spectrofluorimetric, Current Chemistry Letters 3 (2014) 115–124. [12] M. D. Luca, R. Tauler, G. Ioelea and G. Ragno, Study of photodegradation kinetics of T melatonin by multivariate curve resolution (MCR) with estimation of feasible band IP boundaries, Drug. Test Anal. 5 (2013) 96–10. CR [13] M. D. Luca, G. Ioele, S. Mas, R. Tauler, G. Ragno, A study of pH-dependent photodegradation of amiloride by a multivariate curve resolution approach to combined US kinetic and acid-base titration UV data, Analyst 137(22) (2012) 5428-5435. [14] J. C. G. Esteves da Silva and S. A. G. Novais, Trilinear parafac decomposition of acid, Analyst 123 (1998) 2067–2070. AN synchronous fluorescence spectra of mixtures of the major metabolites of acetylsalicylic M [15] R.A. Harshman, Foundations of the PARAFAC procedure: Model and conditions for an 'explanatory' multimodal factor analysis, UCLA Working Papers in phonetics 16 (1970) 1- ED 84. [16] J.D. Carroll, J. Chang, Analysis of individual differences in multidimensional scaling via an PT N-way generalization of and Eckart-Young decomposition, Psychometrika 35 (1970) 283319. CE [17] L. R., Tucker, The extension of factor analysis to three-dimensional matrices in Contributions to Mathematical Psychology. Holt, Rinehart and Winston, New York, 1964. AC [18] E. Sanchez, B. R., Kowalski, Tensorial Resolution: A Direct Trilinear Decomposition. J. Chemom. 4(1) (1990) 29–45. [19] H.L. Wu, M. Shibukawa, K. Oguma, An alternating trilinear decompostion algorithm with aplication to calibration of HPLC-DAD for simultaneous determination of overlapped chorinated aromatic hydrocarbons. J. Chemom. 12 (1) (1998) 1-26. [20] E. Sanchez, B.R. Kowalski, Generalized rank annihilation factor analysis, Anal. Chem. 58 (1986) 496-499. [21] R. Tauler, Multivariate curve resolution applied to second order data, Chemom. Intell. Lab. Syst. 30 (1995) 133-146. 17 ACCEPTED MANUSCRIPT [22] R. Tauler, R. A. Smilde, and B. R. Kowalski, Selectivity, local rank, three-way data analysis and ambiguity in multivariate curve resolution. J Chemom. 9 (1995) 31– 45. [23] A. de Juan, S. C. Rutan and R. Tauler, Two-Way Data Analysis: Multivariate Curve Resolution–Iterative Resolution Methods, Comprehensive Chemometrics, ed. S. Brown, R. Tauler and R. Walczak, Elsevier, Oxford, 2009. [24] R. Bro, PARAFAC. Tutorial and applications, Chemom. Intell. Lab. Syst, 38 (1997) 149– T 171. CR Thesis, University of Amsterdam, Amsterdam, 1998. IP [25] R. Bro, Multi-way analysis in the food industry: Models, algorithms, and applications. PhD. [26] R. Bro, H.A.L. Kiers, A new efficient method for determining the number of components in US PARAFAC models, J Chemom. 17 (2003) 27-286. [27] J.B. Kruskal, Rank, decomposition, and uniqueness for 3-way and N-way arrays, in AN Multiway Data, Analysis, Coppi R, Bolasco S (Eds), Elsevier, Amsterdam, 1989. [28] R. Bro, N.D, Sidiropoulos, A.K. Smilde, Maximum likelihood fitting using simple least M squares algorithms, J Chemom. 16 (2002) 387–400. [29] D. Eskiköy Bayraktepe, Z. Yazan, K. Polat, Sensitive and selective voltammetric ED determination of anti˗cancer agent shikonin on sepiolite clay/TiO2 nanoparticle/MWCNTs composite carbon paste sensor and investigation of its electro˗oxidation mechanism, J. of AC CE PT Electroanal. Chem. 780 (2016) 38–45. 18 ACCEPTED MANUSCRIPT Table 1. Experimental pKa determination results of vanillin obtained by the application of MCR-ALS and PRAFAC methods to the voltammetric dataset T IP : Standard deviation RSD : Relative standard deviation AC CE PT ED M AN CI : Confidence interval at 95 % confidence level US SD Literature methods [1] [2] [3] [4] 7.75 7.40 7.15 7.36 - CR Mean SD RSD CI Proposed methods MCR-ALS PARAFAC 7.91 7.97 0.12 0.10 1.40 1.19 0.14 0.12 19 PT ED M AN US CR IP T ACCEPTED MANUSCRIPT AC CE Figure 1. Chemical structure of vanillin 20 AC CE PT ED M AN US CR IP T ACCEPTED MANUSCRIPT Figure 2. Three-dimensional voltammograms as a function of the potential and frequency (from pH=1.3 to pH =12.0) 21 ACCEPTED MANUSCRIPT (c) Frequency x pH (8 x 12) 1 2 3 T Potential (b) CR 198 199 200 IP (a) Two-way data matrix, 1 Frontal slices US Frequency AN Frequency 1 2 3 4 5 6 7 8 Two-way data matrix, 2 M Horizontal slices CE PT ED Three-way data array, Potential x Frequency (200 x 8) pH Potential Potential x pH (200 x 12) 1 2 3 4 5 6 7 8 9 10 11 12 Two-way data matrix, 3 Vertical slices AC Figure 3. a) Three-way array, b) the substructures (frontal slices, horizontal slices and vertical slices) in three-way array and c) the representation of unfolding threeway data array into two-way data matrices 1 , 2 and 3 . 22 AC CE PT ED M AN US CR IP T ACCEPTED MANUSCRIPT Figure 4. a) Normalized voltammetric profiles and b) relative concentration profiles obtained by applying the MCR-ALS method to the 2-D matrix 2 . 23 AC CE PT ED M AN US CR IP T ACCEPTED MANUSCRIPT 24 ACCEPTED MANUSCRIPT AC CE PT ED M AN US CR IP T Figure 5. Normalized voltammograms of unionized (acid) and ionized (conjugate base) forms of vanillin in 12 different pH media, obtained by the extraction from the voltamometric profiles given in Figure 3a. Figure 6. a) Voltammetric profile and b) relative concentration profile obtained by applying the MCR-ALS method to the 2-D matrix, 3 . 25 AC CE PT ED M AN US CR IP T ACCEPTED MANUSCRIPT Figure 7. Normalized voltammograms of unionized (acid) and ionized (conjugate base) forms of vanillin in eight different frequencies (Hz), obtained by the extraction from the voltamometric profiles given in Figure 5a. 26 CE PT ED M AN US CR IP T ACCEPTED MANUSCRIPT AC Figure 8. a) Voltammetric profile, b) Frequency profile and c) Relative concentration profile obtained by PARAFAC decomposition of the potential-frequency tensor. 27 US CR IP T ACCEPTED MANUSCRIPT AC CE PT ED M AN Figure: Graphical abstract 28 ACCEPTED MANUSCRIPT Highlights T IP CR US AN M ED PT CE 3D-voltammograms were obtained as a function of different frequencies and pHs. Multiway analysis of 3D-voltammograms was used for the estimation of vanillin’s pKa. Multiway analysis methods are PARAFAC and MCR-ALS deconvolutions. This study indicates possible to monitor acid and base species of vanillin in its dissociation equilibrium. MCR-ALS and PARAFAC were alternative tools to quantify pKa. AC 29

1/--страниц