Fenton processes for AOX removal from a kraft pulp bleaching industrial wastewater: optimisation of operating conditions and cost assessment

AOX emissions to water are a key environmental aspect in the pulp and paper industry being limited by EU Ecolabel criteria. In this work, response surface methodology was used to optimise Fenton and photo-Fenton processes for AOX removal from a kraft pulp mill bleaching wastewater. Focus on the specific stream where AOX is higher reduces the wastewater volume to treat and the associated costs. Moreover, there is a need to assess the effect of treatment in parameters other than AOX, such as BOD 5 , COD and colour content, which were also quantified in this work. Operational costs were determined for both processes, including chemical consumption and energy input (in the case of photo-Fenton process). The photo-Fenton process exhibited better performance than the classic Fenton process, achieving 90 % AOX removal against 80 % for the Fenton process, higher colour content removal and enhanced wastewater’s biodegradability (BOD 5 /COD ratio). Moreover, photo-Fenton process showed lower operational costs for maximum AOX removal: 46.5 €·m -3 wastewater treated or 0.45 €·g -1 AOX removed, against 70.0 €·m -3 wastewater treated or 0.78 €·g -1 AOX removed by the Fenton process.


INTRODUCTION
Pulp and paper industry (PP) is one of the main producers of wastewater worldwide. Kraft pulping is the prevalent chemical pulping process accounting for 90 % of the worldwide paper production [1]. From the different stages of the pulping process, bleaching process produces probably the most toxic wastewater stream, exhibiting high chemical oxygen demand -COD (1 -7 g·L -1 ), high content of suspended solids (0.5 -2 g·L -1 ), low biodegradability ratio of BOD5/COD (0.02 to 0.07, with BOD5 meaning 5-day biochemical oxygen demand), dissolved lignin, colour, complexing agents like EDTA, an array of inorganic elements (e.g. Ca, K, Mg, P and Al) and hundreds of organochlorine compounds, such as dioxins, furans, chlorophenols, chloroform, chloromethane, phenols, chlorinated hydrocarbons, etc. [2][3][4][5][6][7]. However, the most concerning pollutant found in PP wastewater is perhaps the adsorbable organic halogens -AOXwhich show extremely concerning effects on fish and zooplankton, including respiratory stress, mixed function oxygenase activity, carcinogenicity and mutagenicity, liver damage, effect on sexual maturation and ability to reproduce, and even lethal effects [3,4,8,9]. Therefore, AOX emissions from the pulp and paper industry are regulated with the EU Ecolabel imposing a maximum of 0.17 kg·ton -1 of air-dried pulp at the point of discharge [10].
The toxic and recalcitrant nature of the PP bleaching wastewater makes conventional biological processes insufficient to ensure efficient treatment [11][12][13][14][15][16]. Regarding AOX removal, Ranganathan et al. [8] cited more than one study reporting biological processes' efficiencies below 50 % in removing these compounds from wastewaters, even with hydraulic retention time of 15 h.
The Fenton process is one of the most widely studied advanced oxidation process (AOP) for recalcitrant compounds removal. It is based on the catalytic effect of Fe 2+ on the dissociation of the oxidant, H2O2, into hydroxyl radicals, HO • , which is a strong and non-selective oxidant capable of oxidizing highly recalcitrant organic pollutants [11,[17][18][19][20]. The main drawbacks of the Fenton process are the high consumption of chemicals and the formation of a dark iron sludge that requires additional costs for separation from the treated wastewater and subsequent disposal [21][22][23][24]. Adding UV light to the treatment (the so called photo-Fenton process) has been proved effective for regenerating the catalytic Fe 2+ ion by photo-reduction of Fe 3+ (reducing the consumption of Fe 2+ ), simultaneously improving the production of HO • and the degradation of organic pollutants [7,11,19,22,23,25,26].
The efficiency of Fenton processes is mainly ruled by the conditions of operation, namely oxidant and catalyst concentrations, pH, temperature, and time of treatment [18,23]. Moreover, the choice of the most suitable experimental conditions to treat any particular wastewater depends mainly on the characteristics of that wastewater [19]. Fenton processes are strongly influenced by pH mainly due to the speciation of iron [11,25,26], with the optimum pH being around 2.5 -3 [11,22,27]. At lower pH values, complex iron species like [Fe(H2O)6] 2+ react with H2O2 at slower rates, reducing HO • formation and changing quantum yield of light absorption (in the case of photo-Fenton) [11,28,29]. On the other hand, higher pH values can favour: (i) H2O2 decomposition to O2 and H2O [26,28], and (ii) the formation and precipitation of iron oxides/hydroxides, meaning less iron ions are available to react with H2O2 and form HO • [2,11,26]. Temperature has been reported to have a negligible effect on Fenton/photo-Fenton efficiency [11,28,30] given the low activation energy of the first steps of the reaction mechanism [31].
In this work, response surface methodology (RSM) was used to find the optimal operating conditions to remove AOX from a real kraft pulp bleaching wastewater by Fenton and photo-Fenton processes. Since AOX arises essentially in pulp bleaching operations, this work focused on AOX removal from the bleaching stage wastewater before mixing with other wastewater streams and being sent to the wastewater treatment plant. This strategy reduces the volumes to be treated by the Fenton processes and thus reduces the consumption of chemicals and the capital and operational costs. Despite focusing on AOX removal, and contrasting to what is sometimes found in the literature, in this work an integrated analysis has been performed to assess also the "side-effects" of the Fenton/photo-Fenton treatment, regarding COD, BOD5 and colour reduction, plus the operational cost associated with the Fenton and photo-Fenton processes.

WASTEWATER
The bleaching wastewater was provided by a Portuguese PP industry that produces kraft pulp, mainly from Eucalyptus globulus, using a 4-stage elemental chlorine free bleaching process comprising three chlorine dioxide (D) stages and an intermediate alkaline extraction with H2O2 addition (Ep) -D0EpD1D2. The wastewater samples were collected (grab sample) from the industrial plant, immediately after the D0 bleaching stage, and stored in the dark at 0 -4 ºC. The wastewater samples were used within one month and characterized by standard methods (see Section 2.4). The parameters listed in Table 1 represent mean and standard deviation values of replicate analysis performed monthly throughout two years.

EXPERIMENTAL DESIGN
The experimental work was planned and executed according to central composite experimental design (CCED). This method allows to assess the effect of a priori established factors (independent variables) on a target response (dependent variable). Factors are studied within a predefined range (minimum "-1", median "0" and maximum "1"), originating a matrix of experiments to perform, to which at least 3 replicates of the central point (performed at the "0" values of each factor) and 2 axial points (± "1.42") are added, to allow for statistical inference and quadratic effects assessment, respectively. In this work, the response was AOX removal from the D0 bleaching wastewater and the independent variables tested were concentration of the oxidant [H2O2] and concentration of the catalyst [Fe 2+ ], for both Fenton and photo-Fenton processes. Considering that the pH of the wastewater used in this work (2.2 ± 0.2, see Table 1) was within the optimal pH range (2.5 -3.0) for Fenton processes [11,22,27,32], the experiments were performed without pH correction. Regarding temperature, since the wastewater is generated at 60 ± 5 ºC the experiments were performed at 60 ± 2 ºC envisioning a future industrial implementation (i.e. to avoid additional costs related to temperature adjustment). The treatment time (t) was set at 10 min based on preliminary experiments that revealed treatment times above 10 minutes have no statistically significant effect on AOX removal for this wastewater [33]. repetitions at the central point. Irradiance* (W·m -2 ) 142 *Photo-Fenton process (medium pressure UVA+UVB lamp)

EXPERIMENTAL SETUP
Experiments were performed in a 0.7 L quartz photoreactor, equipped with a 150W UV medium pressure UVA+UVB (297 -436 nm) TQ150 lamp (for the photo-Fenton experiments) and magnetic stirring. For each experiment, 0.5 L of bleaching wastewater was placed inside the reactor and, after reaching the pre-set temperature for the essay, H2O2 (30 % w/v, Panreac) and Fe 2+ (0.8 M aqueous solution of FeSO4·7H2O, Panreac, 97% purity) were added under constant stirring (200 rpm) to initiate the Fenton reaction. In the photo-Fenton experiments, the UV lamp was turned on immediately after adding the chemicals. After the pre-set treatment time, samples were taken from the reactor and immediately quenched by adding an aqueous solution of sodium sulphite (1.5 M aq. solution of Na2SO3, Fisher, 98% purity). Each sample was split into three subsamples, which were preserved as follows: (i) acidification (pH< 2) for AOX and COD analysis; (ii) neutralization (pH ≈ 7) for BOD5 analysis; (iii) no pH adjustment for colour measurement. The three subsamples were stored below 4 ºC until analysis.

ANALYTIC METHODS
AOX was quantified by coulometric titration, in accordance with EN 16166:2012, ISO 9562:2004 and EPA Method 1650C, using a Thermo TOC 1200 AOX/Total Carbon Analyser. COD and BOD5 measurements were performed using an Aqualytic® COD photometer and a WTW Oxitop® Control equipment, respectively, following the commonly accepted methodologies 5220D and 5210D [34]. pH was measured with a Denver Instrument® model 25 pH/ION METER. Colour was determined by absorbance measurements at 410 nm, according to ISO 7887:2015, using a T80+ UV/VIS spectrometer from PG instruments.

STATISCAL DATA TREATMENT
RSM was used to analyse the experimental results, allowing to explore the relationships between the variables and the studied response from a reduced number of experiments. It is used for mapping a response surface over a special region of concern, or for selecting operating conditions to obtain purpose requirements [35].
In this work, a regression model (second order polynomial function, equation (1)  The adequacy of the suggested regression model was evaluated by carrying out an analysis of variance (ANOVA) of the regression; p-value and Lack of Fit parameters were used, as well as the determination coefficients (R 2 and R 2 adj). A 95 % confidence level was adopted in this study.
The validation of the model assumptions was done through a residuals plots analysis. First, the residuals were normalized with respect to their standard deviations (studentized) and its distribution was analysed to determine if they followed a normal distribution (i.e. a normal distribution function was fitted to the residuals) and were random (the observed residuals were plotted against predicted AOX removal). An outlier plot was also assessed to determine if any of the experiments had particularly large residuals; usually a cut-off of  3.5 standard deviations is used to define an outlier. All statistical analyses were performed with StatSoft Statistica ® v. 8.0 software.

RESULTS AND DISCUSSION
The bleaching wastewater sample used in this work had 113.9 mg·L -1 AOX, 3703 mg·L -1 COD, 174.3 mg·L -1 BOD5 and 1804 mg Pt·L -1 colour. These values are within the average values of the bleaching wastewater from the industrial plant as shown in Table 1.

Fenton process
The second-order polynomial model of equation (1) and three simplifications of this model were fitted to the experimental results for the Fenton processsee Table S1 in the Supplementary Material. Since the interaction between the independent variables was not significant (p = 0. 37) and decreased the quality of the fitting (decreasing R 2 adj), the polynomial fitting was performed disregarding the interaction term from equation (1).
The ANOVA results and the regression coefficients of the fitted model (without interactions) are shown in Table 3 and Table 4, respectively. The determination coefficient (R 2 ≈ 0.815) indicated the model explains 81.5 % of the response variability (see Figure 1d). This value is very close to the minimum value (R 2 = 0.80) usually pointed out as acceptable [36,37]. Most important, the Lack of Fit was not significant relatively to pure error (p-value = 0.08), which proves that the quadratic model without the interaction term fits well to the data for the Fenton process. Therefore, the model fitted to the experimental results was obtained according to Eq. (2).
The linear effect of the oxidant concentration, [H2O2], was the only statistically significant variable, meaning that AOX removal increases linearly with increasing [H2O2] (p<0.01, see Table 3). Concerning the validation of the model assumptions, the residual plot shown in Figure 1a shows small deviation from the straight line, indicating that the studentized residuals may be considered to follow a normal distribution. Figure 1b   and predict the AOX removal for any given operating condition within the experimental range of this work, and within that range only (see Table 2). Results showed increasing AOX removal for increasing [H2O2], which is in accordance with several published works (e.g. [11,17,29]) dealing with Fenton processes. Enough H2O2 will maximize the production of HO • , which is the driving force of the main reactions of degradation of organics (R) involved in Fenton processes: hydroxylation, dehydrogenation or hydrogen abstraction, and redox reactionssee equations The general degradation pathways of the organics show that hydroxylation of organic aromatic pollutants leads to the cleavage of the aromatic moieties yielding short-linear aliphatic carboxylic and dicarboxylic acids, which chain is shortened, giving rise to the accumulation of oxalic and formic acid that are finally mineralized to CO2 [38,40]. Care must be taken to prevent excess use of H2O2 given its counter-productive effect of HO • scavenging; moreover, unused H2O2 will increase COD and toxicity of the treated wastewater, which may pose environmental threats and/or hinder Fenton application prior to biological treatment [2,11,19]. In fact, Bautista et al. [39] recorded an EC50 value of 12 mg·L -1 of H2O2 for marine bacterium Vibrio fischeri.
Moreover, for short treatments times, the intermediates in the solution may still be structurally close to the initial recalcitrant compounds [41], contributing to toxicity, although Xie et al registered that 28 out of 33 kinds of organic compounds, among which 11 out of 13 kinds of AOX were completely removed by the Fenton process applied to pharmaceutical wastewater [42]. min. The higher AOX removal yield for lower concentrations of H2O2 and Fe 2+ might be due to the biological pre-treatment undergone by the wastewater before the Fenton treatment [43].
Hence, easily biodegradable compounds had already been removed from the wastewater matrix and did not compete for HO • . These findings prove the influence of the wastewater composition on the treatment efficiency, and the need for optimization procedures such as those performed in this work.  achieved (a confirmation experiment was performed yielding 8% COD removal). For the highest [H2O2] tested in this work, maximum COD (around 4500 mg O2·L -1 ) in the treated wastewater was registered, which would correspond to an increase in COD (around 20%). This is in agreement with Babuponnusami and Muthukumar [11], who also observed that excess chemicals would contribute to increase COD in the substrate. The explanation for this fact is probably related to the auto-scavenging of HO • radicals, which forms water and perhydroxyl radicals (OH2 • ) reducing the efficiency of the process [20,38,44,45]. Indeed, a decrease in AOX removal in the region of higher oxidant concentrations was also observed in this work (see   [29,43]. At optimal oxidant and catalyst concentrations for AOX removal, model prediction (confirmed experimentally) showed that colour became around two-fold the colour of the untreated wastewater, which is a disadvantage of the Fenton process.

Photo-Fenton Process
The second-order polynomial model of equation (1) and three simplifications of this model were fitted to the experimental results for the photo-Fenton process; Table S2 The p-value of the Lack of Fit (see Table 5) indicates that the fit of the regression model is significant. Moreover, the R 2 of this model was also quite good (≈0.99) which highlights the quality of the fit (see also Figure 4d). Table 5 and Table 6 present the ANOVA results and the regression coefficients, respectively, for the fitted model. The model adequacy was checked by residual plots analysis. The normal probability plot presented in Figure 4a shows that residuals mainly follow a normal distribution. The random scatter of the residuals in Figure 4b indicates that the regression model is an appropriate description of the process. In Figure 4c all observations lied in the interval  3.5, thus no outliers were identified in this set of experiments.

. AOX removal by photo-Fenton process: (a) Studentized residuals and normal (%) probability; (b) Studentized residuals and predicted response; (c) Outlier t plot; (d) predicted and experimental values.
The response surface obtained from the photo-Fenton experiments, concerning the AOX removal efficiency, is shown in Figure 5, and can be used to analyse and predict the AOX removal for any given operating condition within the experimental range of this work, and within that range only (see Table 2). AOX removal by photo-Fenton process increased with increasing [H2O2] (p < 0.0001).
Significance of both linear and quadratic effects shows that this dependence is well described by a second-order relationship. Maximum AOX removal attained was around 90 %, for [H2O2] ≥ 178 mM, and providing a minimum 1 mM Fe 2+ , which is substantially lower than for Fenton process. Under these conditions, the predicted AOX removal for the Fenton process was only 68 % (see also Figure 2).   (6)). In addition, a different (but slower) pathway of HO • production takes place under photo-Fenton conditions, which is the direct photolysis of H2O2, according to equation (7) [

Operational cost analysis
The experimental results presented in this work show that it is possible to attain a certain value for AOX removal by more than one combination of for the photo-Fenton process operational cost. Figure 7 shows the minimum attainable cost for the different AOX removal targets from the bleaching wastewater by both Fenton and photo-Fenton processes.  increases several times the removal efficiency. In another work, Cañizares et al. [53] reported 5 €·m -3 as the cost to achieve 70 % COD removal from olive oil mill wastewater, considering 18 g·L -1 H2O2 and 0.7 g·L -1 Fe 2+ as chemical input. One should note, however, that the costs presented in this work were based in lab-scale prices for the chemical reagents. At full-scale, market prices of the chemicals used would be several times lower, which would mean much lower costs would be attained. For instance, Belalcázar-Saldarriaga et al. [52] considered market price of H2O2 as half of the one considered in this work and the price of FeSO4 was almost 10-fold lower. Cañizares et al. [53] also considered prices for H2O2 and FeSO4 around ten-fold lower than the prices considered herein.
Besides allowing for higher AOX removal efficiencies, the photo-Fenton process provided higher colour removal, increased biodegradability of the wastewater, lower operational costs, and has the potential to reduce the formation of iron sludge. This is still the main drawback of the Fenton processes and still requires further investigation. For maximum AOX removal, BOD5/COD ratio (i.e., the biodegradability) slightly increased from 0.05 to 0.09 for photo-Fenton treated wastewater, against 0.07 using Fenton process; colour decreased almost 80 % for photo-Fenton, while for Fenton it became almost two-fold higher than in the untreated wastewater. The operational cost was also an advantage of the photo-Fenton process: under optimal operating conditions, the costs were estimated to be around 70 €·m -3 or 0.78 €·g -1 AOX removed for photo-Fenton, against around 46.5 €·m -3 or 0.45 €·g -1 AOX removed for Fenton.
Finally, this work highlighted the advantages of (photo-)Fenton processes over other technologies regarding AOX removal, such as the higher removal efficiencies compared with biological treatment (usually between 50 -60 % [54,55]), and the lower cost and shorter treatment time when compared to other AOP such as ozonation or photocatalysis [21,56]. The coupling between Fenton processes and biological treatment is probably the most suitable to treat PP wastewater, due to the coupled efficiency in removing recalcitrant pollutants in a first AOP stage, followed by strong mineralization of organic matter by biological processes. The results from this work prove the suitability of Fenton processes to be applied prior to biological treatment, which could help solving the issue of wastewater treatment for industries that generate very high amounts of wastewaters.