Development of a Novel 3D Highly Porous Structure for TiO₂ Immobilization and Application in As(III) Oxidation

Abstract

One of the main drawbacks of the application of photocatalysis for wastewater treatment is the use of dispersed photocatalysts, which are difficult to remove from effluent after the treatment process and may pose additional toxicity to the receiving bodies. As an alternative, immobilized catalysts can be applied; however, this strategy can increase the difficulties in mass and photo transfer. This work presents the development of an inert and highly porous support for TiO₂ immobilization. The produced materials have a high surface area and contribute to diminishing the difficulties in mass and phototransfer during photocatalysis. Different types of polymeric materials were tested as support, and a Taguchi experimental design with an L9 arrangement was used to optimize the immobilization process and evaluate the effect of TiO₂ content and the use of bidding agents, ultrasound, and thermic treatment. The grey automotive polyurethane foam proved to be the best support, using 5.0% of TiO₂ (wt.%) in the immobilization suspension with Triton X as the binding agent and heat treatment during immobilization. At the optimal conditions, it was possible to achieve total As(III) oxidation (below the analytical detection limit) in 240 min, with nearly 100% As(V) present in solution at the end of the reaction (almost no As adsorption on the catalyst surface). In addition, the catalytic bed was able to promote the As(III) complete oxidation in up to five consecutive cycles without significant leaching or deactivation of the immobilized TiO₂.

1. Introduction

Photocatalysis is a widely utilized method for treating water and wastewater containing toxic organic pollutants, heavy metals, and microorganisms [1,2]. TiO₂ is the most utilized photocatalyst, and even though it appears to be a promising material, in the vast majority of studies, suspensions containing powder TiO₂ are utilized. This approach highlights the major drawbacks of photocatalytic processes, including the limitation of UV light penetration due to the turbidity caused by dispersed nanoparticles [3], the difficulty in reusing the catalyst after the water treatment [4], and the need for a final filtration step to avoid the presence of nanoparticles in the final effluent [5]. To overcome these limitations, TiO₂ films can be prepared on different types of inert substrates, eliminating the need for a post-filtration step and allowing the catalyst to be reused [6].

On the other hand, the use of TiO₂ films increases the mass and photon transfer limitations, making “pollutants molecules-catalyst” contact more difficult and limiting the number of photons that reach the catalyst surface. Therefore, the ideal support should have a high surface area [7] while allowing the radiation to pass through its outer structure [8]. Several types of materials have been tested as catalyst supports, including glass, paper, ceramics, fiberglass, cellulose acetate sheets, alginate and chitosan spheres, and stainless steel, among others [5,9,10]. Likewise, different immobilization methods can be applied, such as chemical/physical vapor deposition, sputtering, electrospinning, dip-coating and sol–gel methods [8,11]. The commercial sponge-type polyurethane (PU) is an interesting candidate since it is highly porous, chemically inert, mechanically stable, flexible and elastic, easily shaped, and readily available at a low cost [12,13]. Furthermore, due to its thermoplastic properties, it is possible to enhance the TiO₂ fixation on a PU surface by using a simple thermal treatment method [14,15,16]. Despite these advantages, only a few studies have been carried out using PU as a support. Shoaebargh et al. [15] tested a new hybrid photodegradation-enzymatic process by immobilizing glucose oxidase (GOx) and TiO₂ on a PU surface inserted inside a photobioreactor. The authors, however, did not investigate the parameters that influence TiO₂ immobilization on PU or the catalyst’s reuse potential.

An in depth study of the parameters that affect the preparation of the catalyst films is fundamental to ensuring proper immobilization as well as the material’s durability. Besides the choice of the appropriate support, the immobilization method, the use of binder additives, and thermic and ultrasound treatment should be considered. The addition of additives, such as Triton X, glutaraldehyde, citric acid, and sodium alginate, may allow the formation of stronger chemical bonds between the carrier and the catalyst or aid in the dispersion of TiO₂ nanoparticles in the medium, preventing the formation of agglomerates and producing more homogeneous films. The latter effect can also be achieved by ultrasound treatment of the starting catalyst suspension, while thermic treatment may allow better fixation of the catalyst on the support surface [17,18].

Therefore, the present work aims to evaluate the performance of different 3D highly porous structures as well as the most suitable immobilization method for the production of TiO₂ thin films and their application in photocatalysis. To the best of our knowledge, a simultaneous evaluation of the parameters that affect the preparation of the catalyst films has not yet been carried out. This approach is relevant since synergistic effects can also be tracked. The efficiency in the oxidation of As(III) to As(V), catalyst leaching, and photocatalyst reuse were also evaluated. The oxidation of As(III) to As(V) in aqueous solution was chosen as the model reaction since the inorganic As(III) is described as approximately 70 times more toxic than inorganic As(V) [19]. While the two forms of arsenic are toxic, As(V) is more easily removed by precipitation or adsorption processes than As(III). Both aspects covered in the present study, TiO₂ immobilization and As removal, are of fundamental importance to the sustainable use of water resources. The TiO₂ immobilization process helps to suppress the discharge of nanoparticles in the aquatic environment, minimizing the impacts related to the discharge of nanoparticles and allowing the catalyst to be reused. In addition, arsenic is first on the priority list of the Agency for Toxic Substances and Disease Registry (ATSDR) due to its toxicity combined with the elevated potential its human exposure [20]. The newly developed material can be used as part of an efficient strategy for the removal of toxic pollutants such as arsenic, contributing to the sustainable recovery of polluted environments.

2. Materials and Methods

2.1. Chemicals

As(III) solutions were prepared from NaAsO₂ (Vetec, Duque de Caxias, Brazil, purity 98%). Nitric acid (HNO₃—Neon, Suzano, Brazil, purity 65%) was used to acidify the As(III) standards and samples for atomic absorption spectrometry analysis and material washing. An arsenic stock solution (As₂O₅, 1.0 g L⁻¹ in 2% HNO₃, Sigma Aldrich, Duque de Caxias, Brazil) was used to prepare As(V) standard solutions through appropriate dilutions. Ammonium molybdate tetrahydrate ((NH₄)₆Mo₇O₂₄·4H₂O—Qhemis, Jundiaí, Brazil, purity 83%) and ascorbic acid (C₆H₈O₆—Vetec, Duque de Caxias, Brazil, purity 99%) were used as colorimetric reagents to determine As(V) concentration and nickel nitrate hexahydrate (Ni(NO₃)₂·6H₂O—Dinâmica, Indaiatuba, Brazil, purity 99%) was used as a modifier during the As_(total) determination. TiO₂ Degussa P25 (Evonik, Essen, Germany) powder was used as delivered without further modification or purification.

Triton X-100 ((C₂H₄O)_nC₁₄H₂₂O, Dinâmica, Indaiatuba, Brazil), glutaraldehyde (C₅H₈O₂—Vetec, Duque de Caxias, Brazil, purity 25%), citric acid (C₆H₈O₇·H₂O—Nuclear, Diadema, Brazil, purity 99%), and sodium alginate ((NaC₆H₇O₆)_n—Vetec, Duque de Caxias, Brazil purity 98%) were used as additives in the TiO₂-P25 coating suspensions. Ammonium sulfate (Dinâmica, Indaiatuba, Brazil, purity 99%) and sulfuric acid (Química Moderna, Barueri, Brazil, purity 97%, 1.84 g/cm³) were used to dissolve TiO₂ nanoparticles, and hydrogen peroxide (Lafan, Várzea Paulista, Barzil, purity 35% (w/v)) was used as a colorimetric reagent to determine the concentration of TiO₂ leached during the oxidation reaction. Iron(III) chloride hexahydrate (Vetec, Duque de Caxias, Brazil, purity 97%), 1,10-Phenanthroline 1-hydrate (Dinâmica, Indaiatuba, Brazil, purity 99%), oxalic acid dehydrate (Neon, Suzano, Brazil, purity 98%), glacial acetic acid (Synth, Diadema, Brazil), and ammonium acetate (Dinâmica, Indaiatuba, Brazil, purity 98%) were used for actinometry. The above-mentioned sulfuric acid and sodium hydroxide (Lafan, Várzea Paulista, Barzil, purity 97%) were used for pH adjustment. All samples were filtered through 0.45 μm cellulose acetate membranes (Filtrilo, Colombo, Brazil) before analysis.

2.2. Analytical Determinations

Total arsenic concentration was determined by flame atomic absorption spectrometry (FAAS, model AA 6300, Shimadzu, Kyoto, Japan), with acetylene-N₂O flame, a spectral slit width of 0.7 nm, a wavelength of 197.3 nm, a 25-mA lamp current, and continuous background correction with a deuterium lamp, presenting a limit of detection of 2.67 × 10⁻² mM. Standards and diluted samples were prepared using ultrapure water acidified with HNO₃ (0.02 mM). Nickel solution (0.1 M) was used as a modifier [21]. The pentavalent arsenic concentration was determined by molecular absorption spectrometry (UV-Vis, model Cirrus 80, Femto, São Paulo, Brazil), with a detection limit of 1.7 × 10⁻⁴ mM, through the molybdate blue method [22]. The concentration of leached TiO₂ was also determined by molecular absorption spectrometry with a detection limit of 1.0 × 10⁻² mM, through the hydrogen peroxide method [23,24]. The photon flow for each reactor configuration was determined by the potassium ferrioxalate actinometry method [25].

2.3. Supports

Cellulose acetate monoliths (CAM, Wacotech GmbH & Co. KG, Herford, Germany—Figure S1a), green synthetic fiber (GSF, Scotch Brite 3M, Ribeirão Preto, Brazil—Figure S1b), blue commercial polyurethane sponge (BCP, Bettanin, São José dos Campos, Brazil—Figure S1c) and a grey automotive polyurethane sponge (GPF, FoamPartner GmbH, Wolfhausen, Switzerland—Figure S1d) were tested as support materials for the TiO₂ coating. To determine the mass changes during heating, all materials were submitted to a thermogravimetric analysis (TGA, model Jupiter STA 449 F3, Netzsch, Selb, Germany) in a temperature ranging from 25 °C to 1000 °C. Then, to determine the glass transition temperature, a differential scanning calorimetry (DSC, model Jade, Perkin Elmer, Shelton, CT, USA) was carried out, with a heating rate of 10 °C min⁻¹.

The electron microscopy analyses (SEM) were performed at the Central Laboratory of Electronic Microscopy (LCME-UFSC, Florianópolis, Brazil) with the JEOL JSM-6390LV microscope (Thermo Cientific, Brno, Czech Republic). All samples were analyzed with an acceleration voltage of 10 kV. For the chemical characterization of the samples, an elementary analysis was performed by energy dispersive X-ray spectrometry (EDS).

Catalyst Preparation

Before use, all materials used as support were immersed in ultrapure water solution with alkaline detergent (Derquim LM 01, Panreac Química, Barcelona, Spain) and sonicated in an ultrasonic device (USC 2500—Unique) for 1 h at 50 kHz. The materials were then washed with ultrapure water in abundance and then dried in an air recirculation oven (MA035—Marconi) at 50 °C for 24 h.

The photocatalyst aqueous suspensions containing 2 wt.% of TiO₂-P25 and ultrapure water were prepared and sonicated for 10 min at 50 kHz in order to better disperse the particles. A catalyst layer was deposited by immersing the polymeric structures into the aqueous suspension at a withdrawal rate of 0.5 mm s⁻¹, assuring a thin and uniform film on each substrate surface. The samples were dried at 50 °C for 1 h in the air recirculation oven. The quantity of catalyst deposited was controlled by weighing the coated supports in a precision balance (model ABS204-S, Mettler Toledo, Barueri, Brazil).

The structures were then immersed in a 100 mL ultrapure water solution through a nylon strip under 60 rpm mechanic shaking (model NL-01-01-A, New Lab, Piracicaba, Brazil) for 5 min. The samples were newly dried at 50 °C for 1 h and weighed after cooling in order to verify the amount of leached catalyst.

2.4. Chemical Agents

To improve the catalytic immobilization process, different types of additives were tested. Triton X (TX) was used as a dispersant to increase the mass of immobilized catalyst on the substrate surface [16]. Glutaraldehyde (GL) and citric acid (CA) were used as cross-linking agents [17,18]. Different suspensions were prepared, containing 2 wt.% of TiO₂-P25, 2 wt.% of the selected additive in ultrapure water. Then, the same immobilization procedure and leaching test previously applied were also used to evaluate the additives.

2.5. Experimental Design

Due to the large number of variables that affect catalytic immobilization, an experimental design was applied to optimize the process. Four influencing parameters (control factors) were chosen for the statistical analysis: (i) catalyst load in the coating suspension; (ii) type of binder agent; (iii) amount of binder agent; and (iv) ultrasound dispersion of the catalyst.

To determine the effects of control factors on the TiO₂ immobilization, three levels were considered for each factor, as shown in

Table 1. Factors, levels, and response utilized in the design of experiments.

Influencing Parameters
Factor	Levels
	1	2	3
Initial TiO2 concentration (wt.%)	0.1	1.0	5.0
Additive	Triton X (TX)	Citric acid (CA)	Glutaraldehyde (GL)
Additive concentration (wt.%)	1	3	5
Ultrasonic horn (intensity/time)	0 *	30% (5 min)	70% (5 min)
Response
% of TiO2 immobilized (xi)		Response A
% As(V) after reaction (gi)		$A_i=\frac{\left(x_i+g_i+z_i+k_i+h_i\right)}{5}$	(1)
% de TiO2 immobilized after reaction (zi)
% As(V) after reuse (ki)
% de TiO2 immobilized after reuse reaction (hi)

* Without ultrasonic treatment.

. Therefore, to decrease the number of experiments and obtain the optimal conditions, the Taguchi method was chosen [26]. The total degree of freedom for four factors at three levels each is 8. Therefore, the L9 orthogonal array with 9 experimental positions was selected, and experiments were designed accordingly (

Table 2. Design of experiments according to Taguchi L9 orthogonal array.

Trails	Initial TiO2 Concentration (wt.%)	Additive	Additive Concentration (wt.%)	Ultrasonic Horn (Intensity %)
1	0.1	TX	1	0
2	0.1	CA	3	30
3	0.1	GL	5	70
4	1.0	TX	3	70
5	1.0	CA	5	0
6	1.0	GL	1	30
7	5.0	TX	5	30
8	5.0	CA	1	70
9	5.0	GL	3	0

). The experimental runs were determined by Statistica 12 software and performed in duplicate to minimize the noise factor.

The TiO₂ immobilization on the supports was carried out using the previously described method but applying the ultrasonic cavitation technique with a high-intensity ultrasonic horn (20 kHz, Sonic Dismembrator Model 500, Fischer Scientific, Waltham, MA, USA). The drying methods were defined based on the additives’ boiling points and on previous published studies [16,17,18].

To facilitate the data analysis, a result factor was utilized (Response A, Equation (1), Table 1). Response A was based on the combination of five indicators: (i) the TiO₂ % immobilized on the support (x_i); (ii) the As(III) % oxidized after 60 min of photocatalytic reaction (g_i, As(V) %); (iii) the TiO₂ % remaining immobilized after the reaction (z_i); (iv) the As(III) % oxidized after reuse assay (k_i, As(V) %); and (v) the TiO₂ % remaining immobilized after reuse (h_i).

The Taguchi method involves performing analysis of means (ANOM) on raw data of output responses. ANOM indicates the significant factors, with their contribution to the output response (Δ), and the configuration of factor levels that maximize factor A. The range of mean values (Δ) was calculated for each factor; a large range implies a high influence on factor A. In addition, the statistical analysis of variance (ANOVA) was also performed to reveal which factor was statistically significant for Response A. From the predicted results, the condition for the maximum Response A was obtained. An experimental run was performed at the predicted condition to validate the predicted optimum Response A.

2.6. Photochemical Reactors

2.6.1. Batch Reactor with External Lighting

The As(III) oxidation reactions, during the determination of the best coating method, were carried out in a batch reactor with external light (Figure S2). The setup consisted of a 150 mL capacity reservoir with magnetic stirring (New Lab, Piracicaba, Brazil). The radiation source was the same UVA lamp (Philips Actinic BL TL TL/10 1FM/10X25CC, 6 W, λ_max = 365 nm), fixed 13 cm above the solution level in the reservoir. The potassium ferrioxalate actinometry assays indicated a photonic flux of 7.80 × 10⁻⁸ Einstein s⁻¹.

2.6.2. Annular Reactor

The catalyst reuse tests were performed in a lab scale annular reactor (Figure S3) during consecutive As(III) oxidation cycles. The total treated and illuminated volume were 1.5 L and ~700 mL, respectively. Table S1 summarizes the main characteristics of the reactor. Previous assays of potassium ferrioxalate actinometry indicated a photonic flux of 1.67 × 10⁻⁶ Einstein s⁻¹.

2.7. As(III) Oxidation Experimental Procedure

The initial As(III) concentration used in all experiments was 0.27 mM. In the annular reactor, the jacket temperature was maintained at 20 °C, pH 8.0, pump flow of 1.6 L min⁻¹, and the reservoir was stirred at 60 rpm. A total volume of 1.5 L was recirculated in the system. Experiments in the batch reactor with external light were carried out with 100 mL of As(III) solution, pH controlled at 8.0, agitation at 60 rpm, and without temperature control. In both reactors, the samples were collected after the addition of each reagent, before and after the lamp connection, and at predetermined times.

3. Results and Discussion

3.1. Effect of Polymeric Support Type in the TiO₂ Immobilization

To determine the effect of the support material in the TiO₂ immobilization, all polymeric structures were equally cut (1 × 1 × 1 cm) and evaluated over three parameters: (i) load of TiO₂ supported; (ii) thermic stability; and (iii) glass transition temperature. Regarding the amount of catalyst immobilized on the support surface and the amount that remained fixed after immersion in ultrapure water, it is possible to observe from Figure 1 that the grey automotive polyurethane sponge (GPF) foam was the material capable of retaining a larger mass of TiO₂ adhered in the formed film. In addition, the film formed on this material surface was also the one that presented lower catalyst leaching.

To evaluate the material’s thermic stability, their degradation temperature was determined by thermogravimetric analyses (TGA). It can be noticed by Figure 2 that the material’s thermic stability increases in the following order: green synthetic fiber (GSF) < cellulose acetate monoliths (CAM) < blue commercial polyurethane sponge (BCP) < grey automotive polyurethane sponge (GPF), presenting the beginning of mass loss at around 146, 178, 185, and 253 °C, respectively. These results indicated that the GPF structure is the material that could better withstand some type of thermal treatment for the fixation of the catalyst.

Finally, DSC analysis was performed to obtain qualitative and quantitative information on physical and chemical changes involving endothermic processes (heat absorption), exothermic processes (release of heat), or changes in heat capacity. Therefore, the main objective of the analysis was to determine the material’s glass transition temperature, which is the temperature where polymer molecules turn from the solid phase to a vitreous state. The glass transition temperature is easily identified by a step in the baseline of the measurement curve [27]. In this condition, the material’s internal energy is still not sufficient for melting but can still modify some of the material’s properties, such as heat capacity, coefficient of thermal expansivity, and viscoelasticity [28]. By discovering the glass transition temperature, controlled heat treatment can be applied in the structure coated with TiO₂, leading to deeper penetration of the catalyst into the polymer structure and an increase in catalyst adhesion after cooling. From Figure 3, it is possible to observe an initial endothermic behavior (melting process) related to the dehydration process, followed by an exothermic behavior (decomposition process). The valley observed during CAM analysis (Figure 3a) indicates that the crystallization process for this material is in the range 175–200 °C [29]. Furthermore, no changes in the baseline for the CAM, GSF, and BCP structures were observed. Thus, these three materials do not present a glass transition temperature, or, at least, it was not identified. By analyzing the endothermic peaks, it can be concluded that the melting of the materials occurs at approximately 183, 275, and 306 °C, respectively. Nevertheless, for the GPF, it was possible to notice the baseline change, making evident the material glass transition range, which occurs between approximately 215 and 237 °C with an average value of 226 °C.

Due to the higher thermic stability and the identification of the glass transition temperature of the gray polymeric foam (GPF), it was chosen to be used in the next experiments. In addition, this foam has higher and more dispersed pores, which minimize photon transfer limitations. Furthermore, the GPF also presents: (i) high surface area; (ii) stability against degradation by strong oxidative radicals; (iii) no toxicity, chemical, or mechanical stability; (iv) high durability; (v) hydrophobic nature; (vi) low cost and availability [12,13,14,15].

3.2. Coating Method Optimization

For the experimental design process, the TiO₂ immobilization tests were performed using GPF structures with dimensions of 2 × 2 × 2 cm and a batch reactor with an external UVA lamp for the As(III) photooxidation reaction. As the objective was to optimize the process of catalyst coating, the Taguchi method with the maximize the means (direct observation) characteristic was chosen. With this analysis, the responses versus the levels of different factors can be observed directly from a broken line plot. The mean value of Response A for the corresponding control factors at each level was calculated according to the assignment of the experiment.

The experiment sequence was conducted in a random order to avoid the incorporation of systematic errors. The raw results are shown in the supplementary material (Table S2), along with the equations for normalizing the output variables used in the experimental design. At each experimental level, two samples were prepared to minimize the noise factor. The normalized parameters mean and the resultant Response A are shown in

Table 3. Mean values of normalized experimental results and the respective factor A.

Trials	% of TiO2 Immobilized	% As(V) after Reaction	% de TiO2 Immobilized after Reaction	% As(V) after Reuse	% de TiO2 Immobilized after Reuse Reaction	Response A
	xi	gi	zi	ki	hi	Ai
1	8.66	1.96	99.80	1.06	99.41	42.18
2	4.78	1.68	99.46	0.81	98.41	41.03
3	14.01	8.53	79.00	4.58	75.67	36.35
4	1.82	1.59	99.86	2.23	99.66	41.03
5	0.51	1.13	99.17	1.32	98.82	40.19
6	0.57	10.55	96.27	8.34	94.98	42.14
7	0.63	9.08	99.65	6.54	99.28	43.03
8	0.23	4.24	93.20	5.60	91.52	38.96
9	0.48	14.55	98.06	14.65	94.70	44.37

and were used to predict the optimum configuration to maximize Response A. The mean of the correspondent Response A at different levels of the control parameters is summarized in Table S3 (ANOM). The corresponding diagrams of direct observation analysis for the individual effects of various parameters at each level are shown in Figure 4.

From Figure 4, it is possible to observe that the optimal coating condition should be achieved using 5% of TiO₂ (wt.%) in the coating suspension, Triton X (TX) as the additive with a concentration of 3% (wt.%), and without the use of the ultrasonic horn. The difference in Response A using different TiO₂ concentrations in the coating suspension can be explained by the mass transfer resistance, which decreases when there is a larger mass of catalyst in the suspension. For the same reason, Triton X presents the best results, as it aids in the dispersion of the catalyst nanoparticles in the coating suspension and thus increases the chance of contact between the TiO₂ and the support [30]. The fact that the 3% concentration of Triton X has been more efficient can be explained by the solubility of the additive, since at higher concentrations it does not dissolve and precipitates at the bottom of the coating suspension recipient with a part of the catalyst. The ultrasonic cavitation at high intensity (70%) may negatively affect Response A since the energy supplied by the ultrasonic horn can contribute to the removal of the adhered catalyst particles from the support surface. Since no significant difference between the processes without horn (WH) and with ultrasonic horn at 30% intensity were noticed, it was cheaper and more convenient to not use it.

The difference between the maximal and minimum values for Response A at each factor is defined as the extreme difference, which directly reflects the effect of each factor. The larger the extreme difference, the more significant the controlling factor is. As can be seen from Table S3, the highest difference is related to horn intensity, followed by additive concentration, TiO₂ concentration, and additive type. Analysis of variance (ANOVA) was performed to determine the significance of each coating parameter. Statistically, the F-test provides the information, at a certain confidence level, for which the results are significantly different. As can be seen in

Table 4. ANOVA. (SQ) sum of squares; (df) degrees of freedom; (SMQ) sum of the mean squares; (F) F-test.

Effect	SQ	df	SMQ	F	Porcentage Contribution	Rank
Initial TiO2 concentration (wt.%)	15.5145	2	7.7573	97.7514	17.39	3
Additive type	12.3219	2	6.1610	77.6360	13.81	4
Additive concentration (wt.%)	15.7120	2	7.8560	98.9954	17.61	2
Ultrasonic horn (% intensity)	45.6707	2	22.83540	287.7543	51.19	1
Residual	0.7142	9	0.07936

, all the variables (factors) have a large F-value, indicating that the variation in each process parameter implies a significant change in the coating results.

Confirmation Experiment

The optimal Response A value predicted by the Taguchi analysis is shown in

Table 5. Results of the confirmation together with the mean of the optimal Factor A predicted by the Statistica software.

Confirmation Experiment	Levels
Initial TiO2 concentration (w/w)	5
Additive	TX
Additive concentration (w/w)	3
Ultrasonic horn (% intensity)	WH
Output factors	Response
	1	2
% of TiO2 immobilized	0.75	0.69
% As(V) after reaction	12.55	13.25
% de TiO2 immobilized after reaction	99.95	99.90
% As(V) after reuse	12.15	12.25
% de TiO2 immobilized after reaction	99.90	99.85
Factor A (%)	45.06	45.19
Factor A (%) mean	45.13
Results predicted by Statistica
Factor A	45.49

. As can be seen from the orthogonal matrix L9 shown in Table 3, there is no experiment arrangement that involves the optimal configuration found by the statistical analysis. Therefore, it is necessary to perform an additional experiment with this configuration to verify if the response obtained is indeed the optimal one. The confirmation experiment resulted in an experimental Response A (45.13%) similar to the predicted one by the Taguchi method (45.49%), with a difference of only 0.36%. With the validation of the developed method, the optimal levels of the factors that maximize Response A were utilized to prepare enough supported catalyst to be used in the lab-scale annular photoreactor.

3.3. As(III) Oxidation by GPF-TiO₂/UVA-Irradiated Annular Reactor

The As(III) photocatalytic oxidation by TiO₂ can be driven by three different mechanisms through successive one-electron steps: (i) direct oxidation by TiO₂ photogenerated hole ($h_{VB}^{+}$), (ii) indirect oxidation by OH^• generated by the reaction of H₂O, and $h_{VB}^{+}$, and (iii) indirect oxidation by ${\mathrm{O}}_2^{•-}$/$\mathrm{H}{\mathrm{O}}_2^{•}$ generated by the reaction of O₂ and $e_{CB}^{-}$ [31]. Figure 5 shows a schematic diagram for the photocatalytic oxidation of As(III) by TiO₂ heterogeneous photocatalysis.

After the optimization of the TiO₂ immobilization process, the use of GPF-TiO₂ in the annular reactor system was evaluated to promote As(III) oxidation. In this case, the total mass of TiO₂ immobilized on the support was 293 mg (Figure S4). From Figure 6, it is possible to observe the complete oxidation of As(III) after 240 min of photocatalytic reaction during the first use of GPF-TiO₂. Although this result seems much slower than the ones reported by other authors dealing with As(III) photooxidation, the vast majority of available studies use TiO₂ in suspension. Furthermore, the different operational conditions and lack of standardization of the reported results make the direct comparison of experiments difficult. As an example, using the TiO₂@Fe₃O₄ photocatalyst, Xiao et al. [32] achieved complete As(III) oxidation in only 4 min of photocatalytic reaction. However, since the authors used 10 mg of catalyst in suspension in 40 mL of treated solution with an As(III) initial concentration of 10 mg L⁻¹, the direct comparison with the results obtained in the present article is very imprecise.

Since this is a catalytic process, an important parameter is the catalyst’s potential to be reused. Thus, the GPF-TiO₂ structures (Figure S4) were reused for six consecutive oxidation cycles using fresh solutions of 0.27 mM As(III) with a pH of 8 at 20 °C. After each experiment, the support was immersed in ultrapure water, sonicated for 5 min at 50 kHz, dried at 50 °C for 1 h, and weighed after cooling. As expected, the kinetic constant slightly decreased (Table S4) as the support was reused, from 7.0 × 10⁻³ min⁻¹ in the 1st cycle to 6.1 × 10⁻³ min⁻¹ in the 5th cycle. Still, during the first five cycles, complete As(III) oxidation was successfully achieved in up to 300 min of reactions. Similar results were reported by Fausey et al. [33] in the oxidation of 1 mg L⁻¹ of As(III) by immobilized TiO₂ nanoparticles in electrospun nanofibers modified with reduced graphene (rGO-TiO₂@fibers).

During the 6th reuse cycle, however, the complete oxidation of As(III) was not achieved, and 0.022 mM (or 8.0%) of As(III) was remaining in solution after 300 min of reaction. It was also observed an increase in the As(V) adsorption over the consecutive reuse cycles. This behavior was probably related to the increase in the amount of leached TiO₂ over the cycles, causing small flaws in the film and eventually favoring adsorption. While in the 1st cycle, the As_(Total) adsorbed was only 0.01 mM, at the end of the 6th cycle, this value was increased to 0.13 mM. The loss of TiO₂ per leaching was practically constant for each reuse, resulting in similar concentrations. The total TiO₂ mass lost in the 6 oxidation cycles was 15 mg, which leaves 278 mg (~95%) of catalyst immobilized on the support surface. During the 6 reaction cycles, a total solution volume of 9 L was treated, thus the total TiO₂ concentration in the final effluent was 1.7 mg L⁻¹. According to Kahru and Dubourguier [34], although the toxicity of TiO₂ depends on the organism type, these nanoparticles are harmful to the environment at concentrations higher than 10 mg L⁻¹. Therefore, the amount leached from GPF-TiO₂ is smaller than the limit that poses a risk to aquatic systems, evidencing the contribution of the present work to a more sustainable process.

GPF-TiO₂ Characterization

To investigate the surface of the GPF structures, the support was evaluated through scanning electron microscopy (SEM) before TiO₂ deposition (sample Y), after TiO₂ deposition (sample A), and after six reuse cycles of As(III) oxidation in the annular reactor system (sample A6). In sample Y (Figure 7a), it is possible to observe a smooth surface of the initial GPF structure. In sample A (Figure 7b), it is possible to visualize the TiO₂ particles well adhered to the surface, forming a uniform film. Since the heat treatment reached the glass transition temperature of the polymer, the catalyst particles seem fused to the surface of the material, which is in accordance with the low leaching rate during the reuse tests. Finally, this effect can also be noted in sample A6 (Figure 7c), since it shows a very similar surface to that of unused TiO₂ film (sample A, Figure 7b). The EDS analysis corroborates this observation since the Ti peak was detected in both A and A6 samples (Figure 7e,f, respectively). In sample Y (Figure 7d), as expected, it was not possible to detect the presence of a catalyst. In addition, in sample A6 (Figure 7f), it is possible to identify a small peak related to the presence of the arsenic element, which was already expected since a small percentage of the pollutant was adsorbed on the surface of the support after 6 oxidation cycles.

4. Conclusions

The oxidation of As(III) species was successfully achieved using the annular UVA photoreactor packed with the highly porous 3D GPF structure coated with TiO₂, allowing complete oxidation of the contaminant in 240 min with only 3.8% of As_(total) adsorption. The optimization of the TiO₂ immobilization method through the Taguchi L9 orthogonal array experimental design allowed the preparation of a material with good film stability. In fact, the prepared support could be used for five reaction cycles, promoting As(III) complete oxidation in 300 min without significant leaching of the immobilized TiO₂. Therefore, the GPF support prepared through the developed methodology reached the proposed objectives and can be a viable alternative for the oxidation of As(III) in aqueous medium, mainly due to the low cost of production and the reusability.