High-performance liquid chromatography as a tool to evaluate the performance of the catalytic wet peroxide oxidation of 4-nitrophenol: pre-validation of analytical methods

A high-performance liquid chromatography (HPLC) method capable to detect 4nitrophenol (4-NP) in aqueous solutions with concentration in the range 0.0495-118.8 mg L-1 was developed and validated under the typical criteria for in-house prevalidations. Accordingly, linearity was demonstrated through the Fisher’s exact probability test. Accuracy and precision were then assessed in three concentration levels over the linear range. The reproducibility of the catalytic wet peroxide oxidation (CWPO) of 4-NP was evaluated together with the analytical error of 4-NP determination in independent experiments. In this case, the sum of both contributions to error never reached 3%. Two multi-component HPLC methods were also developed for the determination of possible aromatic intermediates (hydroquinone, 1,4-benzoquinone, catechol, 4nitrocatechol and phenol) and carboxylic acids (oxalic, formic, malic, malonic, acetic and maleic acids) resulting from the 4-NP degradation. The combination of these analytical methods already led to the proposal of an oxidation/mineralization mechanism for the CWPO of 4-NP. Subject Headings. Chemical analysis, chromatography, water pollution. Author


Introduction
Like other advanced oxidation processes, catalytic wet peroxide oxidation (CWPO) is a water treatment technology based on the in-situ generation of hydroxyl radicals (HO • ) -very reactive species with high oxidizing potential (between +2.8 V and +2.0 V), known to be effective in the destruction of a huge range of organic pollutants (Gogate andPandit 2004, Navalon, Alvaro andGarcia 2010). For that purpose, hydrogen peroxide (H2O2) is used as oxidation source, provided that a suitable catalyst is employed to decompose H2O2 into HO •according to a previously reported reaction mechanism (Ribeiro et al 2013a).
In recent years, CWPO has been well established as a treatment option that is suitable for the degradation of a broad range of toxic and bio-recalcitrant pollutants in aqueous phase (Bautista et al 2010, Domínguez et al 2014b, Lücking et al 1998, Melero et al 2009, Neamţu et al 2004, Pinho et al 2015, Ramirez et al 2007a, Rey et al 2008, Ribeiro et al 2012, Ribeiro et al 2013b, Santos et al 2009, Taran et al 2010. Although several studies have been focused on the degradation of real industrial wastewaters (Bautista et al 2010, Domínguez et al 2014b, Melero et al 2009, most CWPO applications still deal with model pollutants such as dyes (Neamţu et al 2004, Ramirez et al 2007b, Ribeiro et al 2012, Santos et al 2009, phenol (Pinho et al 2015, Rey et al 2008, Taran et al 2010 or other phenolic compounds (Lücking et al 1998, Ribeiro et al 2013b, Ribeiro et al 2015a. Typically, real effluents are complex mixtures of several compounds; therefore, generic (lumped) parameters such as chemical oxygen demand (COD) and/or total organic carbon (TOC) are usually determined in order to assess the efficiency of the CWPO treatment process (Bautista et al 2010, Domínguez et al 2014b, Melero et al 2009. On the other hand, model pollutants are usually prepared by dissolution of the target compound in water; in this case, to monitor the concentration of the model pollutant during the CWPO experiments, very specific analytical methods are required (Lücking et al 1998, Neamţu et al 2004, Pinho et al 2015, Ramirez et al 2007b, Rey et al 2008, Ribeiro et al 2012, Ribeiro et al 2013b, Santos et al 2009, Taran et al 2010. For that purpose, dyes have been usually determined by UV-Vis spectrophotometry (Neamţu et al 2004, Ramirez et al 2007b, Ribeiro et al 2012, Santos et al 2009, whereas high-performance liquid chromatography (HPLC) has been found as the most suitable analytical technique for the determination of phenol or phenolic compounds (Pinho et al 2015, Rey et al 2008, Ribeiro et al 2013b, Taran et al 2010.
In addition to the determination of the model compounds, HPLC techniques have been also used successfully in several works for the identification and quantification of reaction intermediates and by-products resulting from the degradation of the model pollutants by CWPO, allowing the identification of the oxidation/mineralization mechanisms that are involved in these processes (Domínguez et al 2013, 2014a, Inchaurrondo et al 2012, Rey et al 2008, Ribeiro et al 2015a. The main goal of the present work is the development of a single-component HPLC analytical method suitable for the determination of 4-nitrophenol (4-NP), a compound chosen as model pollutant for the screening of new catalysts for CWPO. Once validated, this single-component method will allow to follow the 4-NP abatement during the CWPO experiments.
Furthermore, to perform the identification and quantification of reaction intermediates and by-products resulting from the degradation of 4-NP by CWPO, two other multi-component HPLC analytical methods were developed. Specifically, one of these multi-component methods is devoted to the determination of possible aromatic intermediates (i.e., hydroquinone, 1,4-benzoquinone, catechol, 4-nitrocatechol and phenol), whereas the other is focused on the determination of carboxylic acids (i.e., oxalic, formic, malic, malonic, acetic and maleic acids). Once validated, these multi-component methods allow the identification of the oxidation/mineralization mechanisms involved in the CWPO of 4-NP. Finally, in order to assess reproducibility and error of the 4-NP determinations in real applications, a suitable catalyst was synthesized and used in three consecutive CWPO runs.

Analytical methods
UV-Vis absorption spectra were obtained using a T70 Spectrophotometer (PG Instruments, Ltd.) in the wavelength range of 200-660 nm, with a scan interval of 1 nm. The amount of 4-NP was determined by HPLC, adapting the procedure described elsewhere (Apolinário et al 2008). For that purpose, a Jasco HPLC system equipped with an UV-Vis detector (UV-2075 Plus), a quaternary gradient pump (PU-2089 Plus) for solvent delivery (1 mL min -1 ) and a Kromasil 100-5-C18 column (15 cm x 4.6 mm; 5 µm particle size; reversedphase) was employed. Mobile phase consisted in an isocratic method of A:B (40:60) mixture of 3% acetic acid and 1% acetonitrile in methanol (A) and 3% acetic acid in ultrapure water (B). The 4-NP absorbance peaked at 318 nm, as determined from the corresponding UV-Vis absorption spectrum. Possible intermediates of 4-NP oxidation (e.g., hydroquinone, 1,4benzoquinone, 4-nitrocatechol, catechol and phenol) were determined using the same system, the absorbance wavelength being adjusted to 277 nm. The amount of the carboxylic acids (e.g., formic, acetic, oxalic, malonic, maleic and malic acids) was also determined using the same Jasco HPLC system, but in this case using an YMC -Triart C18 column (25 cm x 4.6 mm; 5 µm particle size; reversed-phase), adapting procedures reported elsewhere (Rocha et al 2011, Yang et al 2000. Mobile phase consisted in an isocratic method of A:B (95:5) mixture of 1% sulfuric acid in ultrapure water (A) and acetonitrile (B), delivered to the system at 0.6 mL min -1 . The UV/Vis detector was set to 210 nm.

Synthesis of the catalyst
The catalyst used in the CWPO experiments was prepared by polycondensation of resorcinol with formaldehyde (with a molar ratio of 1:2), adapting a procedure previously reported (Gomes et al 2008). Namely, 9.91 g of resorcinol were added to 18.8 mL of deionized water in a glass flask, to which a calculated mass of iron (III) chloride hexahydrate was added (Fe/resorcinol molar ratio of 0.05). After complete dissolution, 13.5 mL of formaldehyde solution were also added. In order to achieve the desired initial pH of the precursor solution (6.1), sodium hydroxide solution was added dropwise under continuous stirring and pH monitoring. The gelation step was allowed to proceed at 85 o C during 3 days. After this period the gel was dried in oven during several days from 60 to 150 o C, defining a heating ramp of 20 o C day -1 . After drying, the gel was calcined under a nitrogen flow (100 cm 3 min -1 ) at 120, 400 and 600 o C during 60 min at each temperature and then at 800 o C for 240 min, defining a heating ramp of 2 o C min -1 . Finally, the calcined materials were washed with 1 L of deionized water at 50 o C under vacuum filtration, and then with 1 L of HCl solution (pH = 3), also at 50 o C under vacuum filtration, being afterwards dried overnight in an oven at 60 o C, resulting in the material named CX/Fe0.05.

Analyte
Std

CWPO experiments
Batch CWPO experiments were performed as previously described (Ribeiro et al 2015a), in a 250 mL well-stirred (600 rpm) glass reactor equipped with a condenser, a temperature measurement thermocouple, a pH measurement electrode and a sample collection port. The reactor was loaded with 50 mL of a 4-NP aqueous solution (5.0 g L -1 ) and heated by immersion in an oil bath at controlled temperature. Upon stabilization at the desired temperature, the solution pH was adjusted to 3 by means of H2SO4 and NaOH solutions, and the experiments were allowed to proceed freely, without further pH adjustment. A calculated volume of H2O2 (30% w/v) was injected into the system, in order to reach the stoichiometric amount of H2O2 needed to completely mineralize 4-NP (17.8 g L -1 ). The catalyst (CX/Fe0.05, cf. Section 2.4) was added after complete homogenization of the resulting solution, that moment being considered as t0 = 0 min. The experiments were conducted during 24 h, at T = 50 o C, pH = 3 and catalyst load = 2.5 g L -1 .

Acceptance criteria for methods validation
Based on a literature survey, acceptance criteria were established for each one of the parameters that were considered in this in-house validation (cf. Table 2). Individual expressions and requirements for each parameter are detailed in the next subsections.
It should be noted that the ruggedness of the analytical methods was not assessed since the experimental work was fully performed by the same operator and using the same equipment.

Accuracy
The difference between the true value (Xtrue) and the measured value (Xmeasured) was assessed through the analysis of bias values obtained as described in Eq. 1 (Taverniers, De Loose and Van Bockstaele 2004). The values of Xmeasured were given by the average result obtained by three replicate measurements of three samples over the working range, representing high, medium and low concentration levels (cf . Table 1), as typically required (Snyder, Kirkland andGlajch 1997, Taverniers, De Loose andVan Bockstaele 2004).
In the particular case of bias, the results obtained should be compared with a second validated reference method (Taverniers, De Loose and Van Bockstaele 2004). Therefore, since no validated reference methods were found, the acceptance criterion was established as 20%, i.e., the accuracy of the analytical method is accepted if the absolute value of bias is ≤ 20% (Peters and Maurer 2005, UNODC 2009, USEPA 1997. bias, % = (X measured -X true ) X true × 100 (1)

Precision
During the pre-validation of the methods, only repeatability precision and intermediate precision were assessed, since these are the levels of precision that are usually related to measurements performed within the same laboratory (Snyder, Kirkland andGlajch 1997, Taverniers, De Loose andVan Bockstaele 2004). Both levels of precision were expressed in terms of relative standard deviation (RSD, also known as coefficient of variation) of a data set, as usually performed (Snyder, Kirkland andGlajch 1997, Taverniers, De Loose andVan Bockstaele 2004). In repeatability precision (i.e., intra-run precision) conditions, three samples over the working range, representing high, medium and low concentration levels (cf. Nowadays, the so-called Horwitz ratio (HorRat) is widely recognized as a suitable acceptance criterion for the evaluation of the precision of an analytical method, namely by several research groups (Brown and Yu 2013, Indyk et al 2014, Sasaki et al 2014 and organizations worldwide (Horwitz andAlbert 2006, Latimer 2012). Specifically, the HorRat is a normalized performance parameter indicating the acceptability of the precision of an analytical method, which is obtained by the ratio between the observed RSD and the predicted relative standard deviation (PRSD) calculated from the Horwitz equation, as described in Eq. 4 (Horwitz and Albert 2006). PRSD is obtained as described in Eq. 5, where C is the concentration of a given analyte, expressed as a dimensionless mass fraction (Horwitz andAlbert 2006, Latimer 2012).
Based on recommendations made by AOAC International -Association of Official Analytical Chemists, regarding standard method performance requirements, the optimum target for HorRat values should be set to 0.5 in validations performed within the laboratory (Latimer 2012). Nevertheless, a broader acceptance criterion has been defined by the same organization considering HorRat values of 0.3 and 1.3 as the minimum and maximum acceptance values, respectively (Latimer 2012). Therefore, these values were used to set the HorRat acceptance criterion used in this work for repeatability precision (HorRatr) and intermediate precision (HorRatint).

Range
Lower and upper analyte concentrations for which the analytical method has adequate accuracy, precision and linearity LOD Determined based on the average peak area of 10 independent blank samples, plus 3 times the corresponding standard deviation LOQ Determined based on the average peak area of 10 independent blank samples, plus 10 times the corresponding standard deviation

Linearity and range
Statistical tests, such as the Lack-of-fit test, the Mandel's fitting test or the Fisher's exact test, are the more suitable options for the validation of a linear calibration model (Fisher 1934, Loco et al 2002. In this work, Fisher's exact probability test was used to assess the linearity of the calibration curves (Fisher 1934). Specifically, the data was adjusted to a second-degree polynomial model (y = β0+ β1x + β2x 2 ), and then the statistical significance of the null value of the second-degree coefficient was tested through its p-value [P(H0,2nd)], with a significance level of P ≤ 0.05. Therefore, the acceptance criterion for the linearity of the calibration curve was set to P(H0,2nd) > 0.05. The question of whether to force the regression line through the origin (i.e., β0 = 0) has been the subject of several debates on linear regression (Meier and Zünd 2000). Although theory can justify the β0 = 0 assumption, the reality may be somewhat more complex (Meier and Zünd 2000). Taking this into account, the statistical significance of the null value of the yintercept was tested through its p-value [P(H0,int)], with a significance level of P ≤ 0.05. Therefore, a linear model with y-intercept (y = β0+ β1x) was used when P(H0,int) ≤ 0.05, whereas a linear model without y-intercept (y = β1x) was used when P(H0,int) > 0.05. All the linear and polynomial regressions were performed using R Software (version 3.1.2), Copyright © 2014, The R Foundation for Statistical Computing. The range of each method was defined as the lower and upper analyte concentrations for which the analytical method has adequate accuracy, precision and linearity (Snyder, Kirkland and Glajch 1997).

Limit of detection
The determination of the limit of detection (LOD) of each analyte was based on a signal-tonoise ratio of 3, as suggested by Eurachem and IUPAC (Snyder, Kirkland andGlajch 1997, Taverniers, De Loose andVan Bockstaele 2004). The practical assessment of these values was performed as previously described (Taverniers, De Loose and Van Bockstaele 2004). Namely, 10 independent blank samples were analyzed, in order to determine the average peak area of that samples (xbl) and the corresponding standard deviation (sbl). Afterwards, the LOD value was obtained as described in Eq. 6, where β1 is the slope of the linear regression performed for the quantification of each component.

Limit of quantification
The limit of quantification (LOQ) of each analyte was determined exactly as in the case of the LOD, except that it was based on a signal-to-noise ratio of 10, as described in Eq. 7 (Snyder, Kirkland andGlajch 1997, Taverniers, De Loose andVan Bockstaele 2004).

Development and validation of the analytical methods
In this Section, the development of three distinct and independent analytical methods will be described and discussed. Furthermore, the resulting methods will undergo in-house validation studies.

Selection of the HPLC method
The first task of the development process was the choice of the most accurate HPLC method. Accordingly, taking into account that 4-NP, the aromatic intermediates and carboxylic acids considered in this work are non-ionic compounds soluble in water, reversed-phase HPLC was found the most appropriate solution (Lindsay 1992). For each particular analytical method, the choice of the most suitable reversed-phase HPLC column was made based on procedures previously reported elsewhere. Namely, a Kromasil 100-5-C18 column was used for 4-NP and for aromatic intermediates determinations (Apolinário et al 2008), whereas a YMC -Triart C18 column was used for carboxylic acids determination (Rocha et al 2011).

Selection of the detector
The next step in the development process was the selection of the most appropriate detector. It is known that UV-Vis detectors are by far the most used in HPLC (Christian 1994, Lindsay 1992, mainly due to their high sensitivity, relatively low cost and robustness towards slight temperature changes (Christian 1994). Furthermore, UV-Vis detectors are selective, since they only detect compounds that absorb a specific UV or visible radiation (Lindsay 1992). Keeping these assumptions in mind, an UV-Vis detector was chosen (Jasco UV-2075 Plus). Afterwards, UV-Vis absorbance spectra were obtained for each component, using a spectrophotometer (cf. Section 2.3). UV-Vis detector settings (i.e., appropriate absorbance wavelengths) were then selected based on the resulting spectra, which are shown in Figure 1. It should be mentioned that the three analytical methods will be employed to the analysis of complex samples containing 4-NP, and possibly several aromatic intermediates and carboxylic acids. Therefore, the wavelength of each method should be set to a value in which potential interferences have minimal absorbance (Snyder, Kirkland and Glajch 1997). Keeping this in mind, it is observed in Figure 1a that the absorbance spectrum of 4-NP has a very well pronounced maximum at 318 nm. On the opposite, the aromatic intermediates and the carboxylic acids all have negligible absorbance at that wavelength, except 4-nitrocatechol, as it can be seen in Figures 1b and c. Therefore, the UV-Vis detector wavelength was set to 318 nm in the analytical method developed for the determination of 4-NP. The analysis of the UV-Vis absorbance spectra of the aromatic intermediates is more complex, since several compounds must be simultaneously determined. As observed in Figure 1b, several of these aromatic intermediates have their maximum absorbance peak around 220 nm and/or around 277 nm. On the other hand, as seen in Figure 1c, most of the carboxylic acids also show maximum absorbance around 220 nm, but negligible absorbance at 277 nm. Therefore, in order to minimize possible interferences from carboxylic acids, the UV-Vis detector wavelength was set to 277 nm in the analytical method developed for the determination of aromatic intermediates. Finally, the UV-Vis detector wavelength was set to 210 nm in the analytical method developed for the determination of carboxylic acids, in this way minimizing possible interferences from aromatic intermediates. Nevertheless, it should also be mentioned at this point that possible interferences only occur with overlapped compounds (i.e. compounds leaving the HPLC column with the same retention time). Therefore, the selection of the mobile phase for the determination of carboxylic acids also takes this into consideration.

Selection of the mobile phase
Once the detector settings were selected, the mobile phase and the way the mobile phase is supplied to the system were also selected. A combination of mobile phase and mobile phase rate, which was initially reported for the determination of dinitrophenol and trinitrophenol (Apolinário et al 2008) was successfully adapted for the determination of 2-nitrophenol by our group (Ribeiro et al 2013b, Ribeiro et al 2015b. That adaptation was used with success in a preliminary run for the determination of 4-NP, thus being selected for the analytical method used for the determination of 4-NP. Specifically, the mobile phase consists in an isocratic method of A:B (40:60) mixture of 3% acetic acid and 1% acetonitrile in methanol (A) and 3% acetic acid in ultrapure water (B), delivered at 1 mL min -1 . Furthermore, this mobile phase also showed good performance when applied to the determination of aromatic intermediates. In this case, complete separation of the individual components in the mixture was achieved, as observed by the high resolution of the resulting chromatographic peaks (as confirmed by the retention times given in Table 3). On the other hand, the development of a combination of mobile phase and mobile phase rate, suitable for carboxylic acids determination, was slightly more complex. Both the mobile phase and the mobile phase rate had to be optimized through a systematic approach in which each parameter was individually assessed. In a first preliminary run, the mobile phase consisted in a sulfuric acid solution (4 mmol L -1 ) delivered at flow rate of 0.6 mL min −1 (Rocha et al 2011). This combination did not allowed proper separation of the mixture components. Therefore, different sulfuric acid concentrations (in the range 4-200 mmol L -1 ) and delivery rates (in the range 0.4-1 mL min -1 ) were tested. None of the tested combinations resulted in a suitable separation of the mixture components. In the second approach, a mobile phase consisting in an isocratic method of A:B (95:5) mixture of 1% sulfuric acid in ultrapure water (A) and acetonitrile (B), delivered to the system at 0.8 mL min -1 (Yang et al 2000), was tested. This time, the resulting separation of the mixture components was better. Thus, distinct sulfuric acid concentrations (in the range 20-200 mmol L -1 ) and delivery rates were tested (in the range 0.4-1 mL min -1 ). The best separation was obtained when using a delivery rate of 0.6 mL min -1 and the composition which was previously reported by Yang et al. (Yang et al 2000). As observed by the retention times given in Table 3, complete separation of the individual components in the mixture was achieved in this way.

Quantitative determination
Assuming that the separation conditions were fully optimized for each analytical method, the next step was the quantitative analysis of each component. Within this scope, the peak area   High-performance liquid chromatography as a tool to evaluate the performance of the catalytic wet peroxide oxidation of 4-nitrophenol Rui S. Ribeiro, Adrián M.T. Silva, Helder T. Gomes, Joaquim L. Faria Analyte LOD (mg L -1 ) LOQ (mg L -1 ) bias (%)  Table 4: Limit of detection (LOD), limit of quantification (LOQ) and bias obtained for each analyte, in low, medium and high concentration levels over the linear range High-performance liquid chromatography as a tool to evaluate the performance of the catalytic wet peroxide oxidation of 4-nitrophenol Rui S. Ribeiro, Adrián M.T. Silva, Helder T. Gomes, Joaquim L. Faria

Application of the analytical methods
So far, a single-component HPLC analytical method suitable for the determination of 4-NP was developed and validated throughout its linear range. Although the proficiency of this particular method has already been shown in the assessment of 4-NP abatement during CWPO experiments (Ribeiro et al 2015a), further studies were now performed. Specifically, three CWPO runs were carried out under the same operating conditions, in order to assess reproducibility and error of the 4-NP determinations. Through this approach, both the analytical error and CWPO process reproducibility were evaluated. The results obtained in this study are shown in Figure 4, where it is observed that the RSD of the individual measurements never reaches 3%. Likewise, the suitability of the two multi-component methods -one of which is devoted to the determination of possible aromatic intermediates, whereas the other is devoted to the determination of carboxylic acids -was already demonstrated in a previous work (Ribeiro et al 2015a). Figure 5 shows the oxidation/mineralization mechanism resulting from the CWPO of 4-NP, as previously proposed (Ribeiro et al 2015a), based on the application of the two analytical methods that were developed in this work.

Target value
High-performance liquid chromatography as a tool to evaluate the performance of the catalytic wet peroxide oxidation of 4-nitrophenol Rui S. Ribeiro, Adrián M.T. Silva, Helder T. Gomes, Joaquim L. Faria

Conclusions
The main conclusion withdrawn from this work is that an analytical HPLC method for the determination of the 4-NP concentration in aqueous solutions was successfully developed and validated under the acceptance criteria that were previously established for this in-house pre-

RSD (%)
High-performance liquid chromatography as a tool to evaluate the performance of the catalytic wet peroxide oxidation of 4-nitrophenol Rui S. Ribeiro, Adrián M.T. Silva, Helder T. Gomes, Joaquim L. Faria validation. The linearity of the calibration curve used for the determination of 4-NP was demonstrated through the Fisher's exact probability test. Accuracy studies have shown that the bias of this method is in the range 0.4-10%, depending on the concentration level of the analyte. Repeatability precision and intermediate precision were also assessed in three concentration levels over the linear range, through the so-called Horwitz ratio (HorRat). In this case, HorRat values never exceeded 0.26, the maximum acceptable value being 1.3. Taking all this into account, the range of the analytical method developed for the determination of 4-NP was 0.0495 -118.8 mg L -1 . The CWPO process reproducibility was evaluated together with the analytical error of 4-NP determination. In this case, the sum of both contributions to error never reached 3%. Furthermore, two multi-component HPLC analytical methods were successfully developed for the determination of possible aromatic intermediates (i.e., hydroquinone, 1,4-benzoquinone, catechol, 4-nitrocatechol and phenol) and carboxylic acids (i.e., oxalic, formic, malic, malonic, acetic and maleic acids). These methods were validated in the linear range of each analyte, under the acceptance criteria that were previously established for this in-house prevalidation. In addition, the combination of these two analytical methods already led to the proposal of an oxidation/mineralization mechanism for the CWPO of 4-NP.