**Evaluation of TiO2 and SnO Supported on Graphene Oxide (TiO2-GO and SnO-GO) Photocatalysts for Treatment of Hospital Wastewater**

## **Laura Rosero Parra 1, Lizeth Guerrero Pantoja 1, Natali Lorena Mena 2, Fiderman Machuca-Martínez <sup>1</sup> and Julian Urresta 2,\***


Received: 7 February 2020; Accepted: 7 April 2020; Published: 19 May 2020

**Abstract:** The effectiveness of two photocatalysts, TiO2 and SnO, supported on graphene oxide (TiO2-GO and SnO-GO) on the removal of organic matter from hospital wastewater effluent was evaluated at laboratory scale. The results of the experimental design allow us to conclude that variables such as catalyst type and catalyst concentration have a significant effect on the organic matter removal efficiency of the photocatalytic process. The highest levels of removal efficiencies—for chemical oxygen demand, 85%, for phenols, 80%, and for dissolved organic carbon, 94%—were achieved using a TiO2-GO catalyst with a concentration in the wastewater of 1.5 g/L.

**Keywords:** heterogeneous photocatalysis; titanium dioxide; tin oxide; graphene oxide; chemical oxygen demand

#### **1. Introduction**

Among emerging contaminants, pharmaceuticals are one of the major concerns, due mainly to their toxicity, low biodegradability and extensive use [1]. In fact, hospital wastewater is an important source of pollution, not only due to pharmaceuticals and antibiotics not absorbed by the body, but mainly due to the metabolites excreted by patients. In Colombia, most of the hospitals do not have wastewater treatment systems and their effluents are directly discharged into the sewer system, then into the rivers. Drinking water treatment plants are not designed to completely remove substances from hospitals; therefore, direct consumption of tap water could arise as a major public health concern [2].

In Colombia, before 2015, the disposal of hospital wastewater was not regulated by environmental authorities, creating a serious problem of public health and sanitary risk [3]. Nowadays, local environmental authorities, such as DAGMA (Santiago de Cali Administrative Department of Environmental Management) and local public service provider companies such as EMCALI have the power of imposing sanctions in case of non-fulfillment of environmental regulations [4]. For example, between 2017 and 2018 in Santiago de Cali (Colombia), many rehabilitation centers, institutes of radiology, skin centers and clinics have been financially penalized for not complying with wastewater discharge parameters.

Advanced oxidation processes (AOPs) are technically feasible alternatives for the oxidation of the compounds found in hospital wastewater. These methods are based on physicochemical processes which are capable of modifying the chemical structure of the pollutants using highly reactive transient species such as the hydroxyl radicals [5]. Among the AOPs, the photocatalytic process is one of the most studied due to its high efficiency and low implementation costs at commercial level; furthermore, the catalysts can be designed and tuned to absorb energy in certain regions of the visible spectrum, enhancing the process according to the local environmental conditions.

Hospital wastewater contains a complex mixture of active pharmaceutical ingredients and microorganisms. Often, this wastewater is discharged to municipal wastewater treatment plants (WWTPs) without any pre-treatment. The municipal WWTPs are not designed to remove persistent pharmaceuticals. In addition, hazardous wastewater may spread during flooding and combined sewer overflow events. Internationally, there is increasing focus on the potential environmental effects of pharmaceuticals in water environments. Hospitals have been identified as a key source of pharmaceuticals that can act as potent micropollutants. Painkillers such as diclofenac and hormones, for example, can have fatal effects on fish, crustaceans and algae at very low doses. Nowadays, there are now new technologies for treatment of hospital wastewater; conventional methods have been coupled with advanced oxidation processes to obtain high reductions in organic matter content.

For instance, oil and grease traps, sedimentation, homogenization tanks, filtration beds, Fenton processes and the implementation of new photocatalysts are the technologies used for treatment of hospital wastewater with a high content of drug compounds. The study of tetracycline content in hospital wastewater has been important: the use of boron-doped titanium catalysts which absorb visible and UV light enhances the reduction of organic matter content. On the other hand, processes which include upflow anaerobic sludge blankets, anaerobic filters, aerobic processes, and activated sludge extended aeration with final chlorination are only accepted as suitable and efficient technologies for cities with less than 10,000 inhabitants.

In 2019, Monte and collaborators evaluated a hospital wastewater treatment system consisting of a membrane-coupled bioreactor with advanced oxidation, where they showed that the biological phase of the treatment required approximately 20 h of retention, while the advanced oxidation phase took only 5 h, which ensures that processes involving radical hydroxyls as oxidizing agents decrease the production of sludge and involve short residence times. In advanced oxidation processes, groups such as double-bond C-C, activated aromatic groups and non-protonated amines and those antibiotics with higher electronic density have greater affinity with the ozonation process; that is, they will be oxidized more efficiently and quickly [6].

In the same year, Hoang and collaborators, considering the need to implement efficient processes of the reduction of organic and inorganic load, coupled a traditional process of filtration with gravel and sand followed by a biological process of enzymes and plants like *Scirpus validus*, in order to reduce the organic load attributed mainly by acetaminophen. However, it took 15 days to achieve decreases of more than 60%, which shows that traditional processes are not recommended for hospital-type wastewater treatment. At the end of the study, the authors state that, depending on the antibiotic, the biological environment can be used and, if it can be implemented, the minimum retention time should be 40 days. These conclusions lead to the consideration that advanced oxidation processes should prevail over biological processes for the hospital waters, because the flows managed are greater than 5 L/s [7].

In Colombia, there is little research associated with hospital wastewater treatment. In this field, we can mention the preliminary treatment carried out at the Quindío University Departmental Hospital called a fat and oil trap, whose function is to retain suspended solids by sedimentation and by flotation. The fatty material was subjected to a flow homogenization process, ending with a coagulation—flocculation process, which takes care of the removal and reduction of BOD, DOC and suspended solids. The coupling of these three technologies has an efficiency of 59% in the reduction of percentages of organic load in times of residence longer than 24 h, which evidences that it is not feasible for flows greater than 0.2 L/s [8].

In general, most of the recalcitrant compounds found in hospital sewage are difficult to mineralize, and conventional processes do not guarantee the reduction of organic and phenolic-type loads that are present and are continuously monitored by Colombian standards.

Previous research [9] showed that a combination of different semiconductor materials supported on graphene moves the absorption spectra to the visible region, enhancing photocatalytic activity. Characteristics and applications of titanium dioxide (TiO2) on carbon-supported materials such as activated carbon, fullerene, graphene, carbon nanotubes, etc., have been studied extensively [10]. However, there is no published research on the use of romarchite phase tin oxide (SnO) supported on graphene for photocatalytic processes. The use of SnO as a catalyst seems to be interesting due its physical properties: high surface area, structure, colour and size; SnO could also exhibit high photocatalytic activity due its properties as a semiconductor [11]. In this work, we evaluated the efficiency of two photocatalysts (TiO2 and SnO) supported on graphene for the treatment of wastewater from a hospital centre in Cali-Colombia. The effect of the catalyst type and catalyst concentration on the organic matter removal efficiency was studied.

The results of this study are a proof of concept based on punctual sampling of hospital wastewater from the Valle de Lili Foundation, which aims to demonstrate the possible interactions of variables associated with the concentrations of recalcitrant contaminants and synthesized catalysts. These results will allow us to analyse and evaluate the oxidative process of this type of wastewater using advanced oxidation processes.

#### **2. Materials and Methods**

#### *2.1. Materials and Reagents*

Graphite (99 wt%), sodium nitrate (>99 wt%), sulfuric acid (95–97 wt%), glacial acetic acid (>99 wt%), hydrochloric acid (37 wt%), potassium permanganate (99 wt%), hydrogen peroxide (30 wt%), ammonium hydroxide (28 wt%) and ethanol (>99.9 wt%) were purchased from Merck (Darmstadt, Germany) and were used without further purification. Stannous chloride (>99 wt%) was obtained from Panreac (Barcelona, Spain) and titanium tetrachloride (>99 wt%) from Fisher (Pittsburgh, PA). Industrial grade nitrogen was obtained from Cryogas (Cali, Colombia). Milli-Q type water was obtained after two successive steps of filtration, followed by deionization and distillation.

#### *2.2. Synthesis of Graphene Oxide (GO)*

In a typical procedure, 0.5 g of graphite powder and 0.5 g of sodium nitrate were mixed in a beaker, then 23 mL of sulphuric acid (H2SO4) was added to the beaker, which was previously immersed in a water bath. A quantity of 3 g of potassium permanganate (KMnO4) was added slowly into the solution and dissolved using 20 min of low-intensity (70 W, 60 Hz) indirect sonication. Next, 100 mL of water and 3 mL of hydrogen peroxide were added into the solution, and finally 40 mL of water was added [12].

The solution was centrifuged using an Ortoalresa centrifuge Digitor 20 C at 1900 rpm for 5 min; subsequently, the supernatant was decanted away, and the residual was washed with a 10% hydrochloric acid solution. This mixture was centrifuged, decanted and washed again. The washed solution was dried using a forced convection oven at 80 ◦C for 48 h to obtain the graphene oxide powder.

#### *2.3. Synthesis of Romarchite Phase SnO and GO Impregnation*

Firstly, an aqueous solution of GO was prepared by dissolving 2.1 g of GO in 210 mL of water. This solution was then mixed with the tin precursor solution, which was prepared by mixing 6.67 g of SnCl2·2H2O in 100 mL of 20 M acetic acid solution at 80 ◦C. Then, five drops of HCl fuming were added into the solution, and it was left to stand for 24 h.

The solution was quantified by titration with ammonium hydroxide (28%) up to pH 8.0 and was allowed to settle for 24 h at ambient temperature. Then, the solution was dried using a forced convection oven at 100 ◦C for 3 h. Finally, the dry solid was calcined using a continuous flow of N2 in an oven with a temperature program as follows: initially, 40 ◦C for 1 min; then, a first ramp up to 100 ◦C at a rate of 4 ◦C/min; 100 ◦C was then maintained over 25 min; then, a second ramp up to 150 ◦C at a rate of 1 ◦C/min; and, finally, 150 ◦C was maintained over 2.5 h [13].

## *2.4. Synthesis of TiO2 and GO Impregnation*

A solution was prepared by dissolving 0.0905 g of titanium tetrachloride TiCl4 into 20 mL of anhydrous methanol, then 3 mL of HCl fuming and 20 mL of water were added to the solution. A GO solution prepared by adding 0.1 g of GO in 10 mL of water was added dropwise to the titanium oxide solution by the wet impregnation method [13]. The resulting solution was introduced into a forced convection oven at 100 ◦C for three hours to obtain a dry powder. Finally, the dry solid was calcined using a continuous flow of N2 in an oven with a temperature program as follows: initially, 40 ◦C for 1 min; then, a first ramp up to 100 ◦C at a rate of 4 ◦C/min; 100 ◦C was maintained over 1 h; then, the same rate was used to ramp up to 200 ◦C; this temperature was maintained for 20 min; then, a third ramp up to 400 ◦C at a rate of 4 ◦C/min; and, finally, 400 ◦C was maintained over1h[13].

#### *2.5. Characterization*

#### 2.5.1. Photocatalysts

A FT/IR spectrometer (JASCO FT/IR-4100, USA) equipped with an ATR cell was used to characterize the functional groups of the support and the photocatalysts. Diffuse reflectance measurements were carried out with a spectrometer (Ocean Optics, USB 4000, Orlando, FL, USA) using a broadband halogen fiber optic illuminator as a light source (Nikon Inc., NI-30, Orlando, FL, USA). A scanning electron microscope (SEM JEOL, JSM 6490 LV, North Billerica, MA, USA) was used to determine the shape and size distribution of crystalline particles. For the analysis, each sample was coated with gold (Denton Vacum, Desk, North Billerica, MA, USA) and then analyzed at the microscope in a back-scattered mode with an acceleration voltage of 20 kV. Micrographs were taken at 500×.

#### 2.5.2. Samples Analysis

Dissolved organic carbon measurements were carried out in a TOC analyser (Shimadzu TOC, VCPH, Japan) according to Standard Method 5310B. Chemical oxygen demand determinations were made following the Standard Method 5220B, samples were digested for 2 h under closed-reflux at 150 ◦C, and then a spectrophotometer (Shimadzu, UV 1800, Switzerland) at 610 nm was used to measure the light absorbed by the sample. Phenolic compound measurements were carried out using the Hanna Instruments HI3864 kit.

#### *2.6. Experimental Procedure*

Hospital wastewater for the photocatalytic experiments was obtained from a hospital of Cali, Colombia and was prepared by mixing the samples collected from different sewer pipes: oncology department, clinical laboratory, hospitalization rooms, and laundry. Samples were stored in glass recipients and refrigerated at 4 ◦C. Additionally, for adequate preservation and for COD and phenol characterization, samples were mixed with sulphuric acid. Prior to the experiments, each sample was well stirred to ensure homogeneity of the aliquots.

For this research, hospital wastewater samples were collected specifically to assess the proof of concept. In subsequent stages of this preliminary research, compound sampling periods will be established. The inspection box detected for the sampling of the hospital wastewater involves all the discharges associated with the clinical laboratory, hospitalization, deliveries, oncology and surgery. The wastewater at this sampling point includes fats and oils from coffee shops and laundries. Wastewater samples were taken once monthly during three months in the afternoon.

Quantities of 0.25, 0.75 and 1.5 g of each catalyst were added to 1000 mL volumetric flasks and diluted to the mark with the hospital wastewater. Then, aliquots of 20 mL were exposed to a simulated solar radiation level of 250 W/m2 for 1 h and under constant stirring (180 rpm). Photocatalytic experiments were carried out using solar simulator equipment (Atlas, Suntest CPS+, Linsengericht, Germany). After each experiment, the sample was analysed according to Section 2.5.1.

#### **3. Results and Discussion**

## *3.1. Characterization of GO, SnO-GO and TiO2-GO by Means of FT*/*IR and SEM*

Figure 1 shows the FT/IR spectrum of the GO. The tension peak of the hydroxyl group is seen at 3225 cm−<sup>1</sup> (peak 1), which represents the water absorbed by the graphene oxide. Peaks 2 and 3 at 1713 cm−<sup>1</sup> and 1621 cm−1, can be assigned to carbonyl bonds (C=O) and double bonds C=C, respectively. The absorption peak 4 at 1041 cm−<sup>1</sup> is assigned to the C–O stretching vibrations [14]. The presence of C=O and C–O bonds are evidence of the oxidation of graphite [12].

**Figure 1.** FT/IR spectrum of the graphene oxide (GO).

The characteristic FT/IR spectrum of a SnO-GO photocatalyst is depicted in Figure 2. The absorption peak 1 at 3124 cm−<sup>1</sup> can be assigned to the water absorbed, which cannot be completely removed during the synthesis at 150 ◦C. Another characteristic peak (2) appears at 1567 cm−<sup>1</sup> and can be assigned to C=O stretching of carboxylic and/or carbonyl moiety functional groups. The C–H deformation peak (3) can be noticed at 1398 cm−1. The absorption peak (4) at about 1206 cm−<sup>1</sup> is assigned to the C–O stretching vibrations. Finally, near to the inorganic region of the spectrum, the characteristic band of the Sn–O–H bond is seen, at 560.2 cm<sup>−</sup>1. On the other hand, it is important to point out that peaks 2 and 3 can be also assigned to an overtone band of a second vibration of the O–Sn–O bond [15].

**Figure 2.** FT/IR spectrum of SnO-GO photocatalyst.

Figure 3 shows the FT/IR spectrum of the TiO2-GO photocatalyst. Peaks 1 and 2 at 3740 cm−<sup>1</sup> and 3363 cm−<sup>1</sup> represent hydroxyl groups. Meanwhile, peaks 3 and 4 at 1707.6 cm−<sup>1</sup> and 1565.9 cm−<sup>1</sup> can be assigned to the carbonyl groups and the C=C bonds, respectively. Peaks at 1217.8 cm−<sup>1</sup> and 1132.9 cm−<sup>1</sup> appear due the presence of methyl groups in the graphene oxide. In the inorganic region of the spectrum at 541 cm−<sup>1</sup> the characteristic peak of the Ti-O-Ti bond is seen [14].

**Figure 3.** FT/IR spectrum of the TiO2-GO photocatalyst.

As is mentioned above, peak 1 in the spectrum can be assigned to the hydroxyl groups. It can be noticed that if the drying or calcination temperature increases, the band is decreased, due mainly to the fact that water absorbed decreases as the temperature increases. Meanwhile, the increase of peak 3 intensity in Figure 3 can be explained due the TiO-H interaction, which is more marked due to the high calcination temperature, greater than 300 ◦C [15].

Comparing the spectrum of GO with the spectrums of SnO-GO and TiO2-GO, it can be observedthat the wavelengths are shifted to shorter wavelengths, towards the inorganic spectrum. This can be assumed to be due to titanium and tin interactions with oxygen and carbon, and the other compounds' interactions.

Figure 4 shows the SEM micrograph of the GO. The micrograph shows an irregular and wrinkled appearance and a close surface morphology, which can be attributed to the exfoliation of the GO layers during the ultrasonic process [16]. Figure 5 shows the SEM micrograph of the SnO-GO; it can be seen that the romarchite impregnated in the GO presents a thin micrometric cluster, approximating a spherical shape. The surface irregularity can be attributed to the acetic acid [17]. Finally, Figure 6 shows the SEM micrograph of the TiO2-GO. Clusters of crystalline particles with irregular surface can be noticed [18].

**Figure 4.** SEM micrograph of the GO.

**Figure 5.** SEM micrograph of the SnO-GO.

**Figure 6.** SEM micrograph of the TiO2-GO.

SEM characterization shows the presence of chlorine in the SnO-GO catalyst (around 3.5% w/w) mainly due to the use of SnCl2·2H2O as the precursor.

Diffuse reflectance measurements allow us to conclude that all photocatalysts supported on graphene absorb around 80% of the electromagnetic radiation in the visible region (see Figures S1 and S2 Supplementary Information).

#### *3.2. Evaluation of the Photocatalytic Process*

Table 1 shows the raw hospital wastewater characterization parameters and the limit values applicable to discharge of wastewater into the public sewer in Colombia [7]. Characterization parameters were evaluated according to Section 2.5.1.


**Table 1.** Raw wastewater characterization parameters and limit values for discharge.

In Colombia, unfortunately, environmental resolution 0631 of 2015 does not include dissolved organic carbon (DOC) measurement as a physicochemical control parameter; therefore, only the detection limit of the method used in the measurement is reported.

A GC-MS analysis of organic compounds in the hospital wastewater allows us to determine the functional groups of pollutants present in the wastewater, in order to know which of the photocatalysts has greater efficiency in the treatment of hospital waters. These were found: sulphonamides, phenols (3,5-dioctoxyphenol) and aromatic compounds. Moreover, we found several compounds with different functional groups, such as esters, ketones, aromatics and amines.

Figure 7 shows the removal efficiencies obtained after the photocatalytic process with Sn-GO and SnO catalysts. The highest organic matter removal efficiency with the Sn-GO photocatalyst was obtained using a concentration of 1.5 g/L; conversely, the lowest efficiency was obtained using a concentration of 0.25 g/L. Similar trends were obtained with the SnO catalyst. It is important to point out that phenol concentrations in the experiments in which we used 0.25 and 0.75 g/L of catalyst remained unaltered, which indicates that there was no phenol conversion. Moreover, a comparison between the catalysts shows that the highest efficiencies were achieved when the TiO2-GO catalysts were used.

**Figure 7.** Removal efficiencies of chemical oxygen demand, phenols and dissolved organic matter after the photocatalytic process with SnO-GO and SnO.

Figure 8 shows the removal efficiencies obtained after the photocatalytic process with TiO2-GO and TiO2 catalyst. Similar trends are obtained comparing these results with the ones obtained with the SnO-GO catalyst; lower efficiencies were achieved at low catalyst concentrations. As can be noticed, at low concentrations of catalysts (0.25 g/L and 0.75 g/L), concentrations of phenols remain almost unaltered. At the highest catalyst concentrations (0.75 g/L) removal efficiencies of 80%, 85% and 94% for phenols, chemical oxygen demand and dissolved organic matter were obtained, respectively. The use of TiO2-GO significantly increases the removal efficiencies compared with the TiO2 catalyst.

The trends obtained suggest that the generation of hydroxyl radicals increases with increasing photocatalyst concentrations, which enhances the organic matter removal efficiency. However, if the concentration of catalyst is too high, it can produce a "shielding" effect or dimming effect, due to the turbidity created by the suspended catalyst.

As was mentioned above, low concentrations of chemical oxygen demand were obtained at the end of the photocatalytic process due to the degradation of organic matter. However, specific compounds such as phenols were not completely mineralized, achieving conversions lower than 91%.

**Figure 8.** Removal efficiencies of chemical oxygen demand, phenols and dissolved organic matter after the photocatalytic process with TiO2-GO and TiO2.

pH of wastewater does not seem to affect the photocatalytic treatment; high removal efficiencies were obtained in a wide range of pH values. Nevertheless, some publications report that pH affects some properties of the catalysts, such as the particle size, the surface charge, and the maximum and minimum of TiO2 band values due to its amphoteric nature [19].

The concentration of phenol in hospital wastewater is high. In this work, we report a phenol removal efficiency up to 80%, a higher value compared to the values obtained with other methods such as wet air oxidation [20] and adsorption [21], in which conversions of 14% and 10% were respectively reported. The high removal efficiencies obtained with the photocatalytic process can be assigned to the presence of chlorine, to the high amount of absorbed energy and to the properties of the GO-supported catalysts, which enhance the photocatalytic activity.

In advanced oxidation processes that involve hydroxyl radicals, phenol firstly oxidizes to hydroquinone and then to p-benzoquinone. Then, ring opening reactions take place to form maleic acid, which is the main product of the process. Maleic acid can be oxidized in malonic acid, oxalic acid and formic acid [22]. Malonic acid oxidizes to acetic acid; oxalic acid can be converted into formic acid and vice versa. Finally, most of the carboxylic acids can be directly oxidized to CO2 and H2O [22].

Figure 9 shows the removal efficiencies obtained in the photocatalytic process with SnO-GO and TiO2-GO. As can be noticed, the highest organic matter removal efficiencies were obtained when the SnO-GO photocatalyst was used.

**Figure 9.** Removal efficiencies obtained in the photocatalytic process with SnO-GO and TiO2-GO.

#### *3.3. Statistical Analysis*

Figure 10 shows the concentration of phenols in the effluent after the photocatalytic treatment with each one of the catalysts synthesized. As it can be seen, the pattern of concentration is identical for all catalysts; low removal efficiencies were achieved in all cases, except when TiO2-GO with a concentration of 1.5 g/L was used.

**Figure 10.** Concentration of phenols in the effluent after the photocatalytic treatment with each one of the catalysts synthesized.

Figure 11 shows the concentrations of COD in the effluent after the photocatalytic treatment with each one of the catalysts synthesized. As can be noticed, the concentrations of catalysts used in the treatment have different effects on the COD removal efficiency; for instance, the lowest COD concentration was achieved using the SnO-GO catalyst with a concentration of 0.75 g/L, while at the same concentration but using the SnO catalyst, the highest COD concentration was obtained.

**Figure 11.** Concentration of COD in the effluent after photocatalytic treatment with each one of the catalysts synthesized.

Figure 12 shows the concentrations of DOC in the effluent after photocatalytic treatment with each one of the catalysts synthesized. In this case, the effect of catalyst concentration on the DOC removal efficiency is proportional; efficiency increases when catalyst concentration increases. The highest DOC removal efficiency was obtained using the SnO-GO catalyst.

A statistical analysis was carried out using the free distribution software R version 3.5.1 (R Development Core Team, 2018). Differences between the photocatalytic treatments at a significance level of 10% were determined. From the F-test, it can be concluded that the interaction (concentration level and catalyst type) has a significant effect on the COD and DOC removal efficiencies. Taking into account the descriptive analysis summarized in Figures 11 and 12 and the results from the F-test, we performed post-ANOVA Duncan tests. In these tests, we compared the interactions at which the highest COD removal efficiencies were obtained: SnO-GO with a concentration of 0.75 g/L and TiO2-GO with concentrations of 0.75 and 1.5 g/L. We found statistically significant differences between the comparing groups; from this, and considering the phenol removal efficiencies, we can suggest that the TiO2-GO photocatalyst with a concentration of 1.5 g/L can be used to obtain the highest COD, DOC and phenol removal efficiencies.

**Figure 12.** Concentration of DOC in the effluent after the photocatalytic treatment with each one of the catalysts synthesized.

#### **4. Conclusions**

In this work, we investigated the use of a heterogeneous photocatalytic process to treat hospital wastewater using two catalysts supported on graphene—SnO-GO and TiO2-GO—and tested their performance by varying the amount of catalyst loaded in the wastewater. The highest removal efficiencies for chemical oxygen demand (85%), dissolved organic carbon (94%) and phenols (80%) were achieved using the TiO2-GO catalyst with a concentration of 1.5 g/L. Moreover, we obtained an effluent which met the local environmental regulations in terms of chemical oxygen demand (COD < 200 ppm). However, further work would be needed (e.g., increased residence time, use of radical initiators, increased catalyst load, etc.) to achieve a concentration of phenols lower than 0.2 ppm (the discharge limit value). Findings obtained in this research indicate that the heterogeneous photocatalytic process is a viable alternative for degradation of the pollutants in hospital wastewater. The results obtained in this investigation are a proof of concept, based on punctual indicative sampling, and care must be taken before generalizing the results obtained. However, once this first phase is completed, it could be established that the heterogeneous photocatalytic process is a possible alternative for the degradation of pollutants present in hospital wastewater.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4441/12/5/1438/s1, Figure S1. Diffuse reflectance spectra of TiO2 and TiO2-GO catalysts; Figure S2. Diffuse reflectance spectra of SnO and SnO-GO catalysts.

**Author Contributions:** L.R.P., and L.G.P. These authors performed the data acquisition and analysis and prepared all figures. N.L.M., F.M.-M. and J.U., These authors jointly supervised this work. L.R.P. and N.L.M. wrote the main manuscript text. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Universidad del Valle (Colombia), grant number 71030.

**Acknowledgments:** The authors wish to thank to Laboratory of Applied Catalysis and Catalytic Processes (LICAP), to Chemistry Department, and to Universidad del Valle for financial support (Grant 71030). Also, we wish to acknowledge the help provided by Fundación Valle de Lili (Cali, Colombia) for the wastewater samples and the logistic support.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## **Insights into the Photocatalytic Bacterial Inactivation by Flower-Like Bi2WO6 under Solar or Visible Light, Through in Situ Monitoring and Determination of Reactive Oxygen Species (ROS)**

**Minoo Karbasi 1,2, Fathallah Karimzadeh 1,\*, Keyvan Raeissi 1, Sami Rtimi 2, John Kiwi 2, Stefanos Giannakis 3,\* and Cesar Pulgarin <sup>2</sup>**


Received: 15 March 2020; Accepted: 10 April 2020; Published: 12 April 2020

**Abstract:** This study addresses the visible light-induced bacterial inactivation kinetics over a Bi2WO6 synthesized catalyst. The systematic investigation was undertaken with Bi2WO6 prepared by the complexation of Bi with acetic acid (carboxylate) leading to a flower-like morphology. The characterization of the as-prepared Bi2WO6 was carried out by X-ray diffraction (XRD), scanning electron microscopy (SEM), X-ray photoelectron spectroscopy (XPS), specific surface area (SSA), and photoluminescence (PL). Under low intensity solar light (<48 mW/cm2), complete bacterial inactivation was achieved within two hours in the presence of the flower-like Bi2WO6, while under visible light, the synthesized catalyst performed better than commercial TiO2. The in situ interfacial charge transfer and local pH changes between Bi2WO6 and bacteria were monitored during the bacterial inactivation. Furthermore, the reactive oxygen species (ROS) were identified during *Escherichia coli* inactivation mediated by appropriate scavengers. The ROS tests alongside the morphological characteristics allowed the proposition of the mechanism for bacterial inactivation. Finally, recycling of the catalyst confirmed the stable nature of the catalyst presented in this study.

**Keywords:** flower-like Bi2WO6; *E.coli* inactivation; reactive oxygen species (ROS); photocatalysis; solar disinfection; water treatment; pollution

## **1. Introduction**

Over the last few decades, environmental contamination has shifted from the exclusive focus of organic and inorganic pollutants [1], towards the inclusion of bacteria and other organisms [2–4]. Therefore, well-organized methods are urgently required to control the spread [5] or eradicate microorganism-related issues [6]. In recent times, beside the traditional bacterial inactivation methods such as UV disinfection and chlorination, a green, efficient, and cost-effective semiconductor photocatalysis has appeared to be a more promising technique [7,8]. TiO2 has been extensively reported as an effective bactericidal semiconductor photocatalyst due to its high stability, strong redox potential, low cost, and non-toxic nature, but its band-gap of 3.2 eV allows light absorption up to 387 nm which makes up just over 4% of the total solar spectrum [9–11].

Since solar radiation contains more visible light (∼47%) than UV, the appropriate use of this fraction becomes necessary through the employment of efficient visible-light photocatalysts [12]. As a promising visible-light-driven photocatalyst with good chemical and thermal stability, Bi2WO6, beside its non-toxic and environmentally friendly nature, is a typical n-type semiconductor composed of accumulated layers of alternating (Bi2O2) <sup>2</sup><sup>+</sup> layers and (WO4) <sup>2</sup><sup>−</sup> octahedral sheets [13,14]. The valence band of Bi2WO6 consists of O 2p and Bi 6s hybrid orbitals, its narrowed band gap increases visible light absorption capacity, and photoactivity [15,16], while its photocatalytic activity greatly depends on morphology, particle size, surface area, and interface structure [17,18]. Constructing a unique micro/nano hierarchical structure usually shortens the pathways of water pollutants, absorb incidental light more efficiently, because of multiple-scattering increase, and easily separated from wastewater by filtration or sedimentation methods [13,19,20].

However, despite the long presence of this catalyst as a possible solution, most studies on Bi2WO6 have focused on the photocatalytic degradation of organic pollutants, with only a few studies investigating the photocatalytic inactivation of microorganisms. Ren et al. [21] reported *Escherichia coli* degradation in a few hours on Bi2WO6 nest-like structures in a pseudo-first order process. Helali et al. [22] prepared a 20 m2/g SSA Bi2WO6 leading to *E. coli* inactivation within four to five hours under solar light on a hydrothermally grown mixture of Bi-nitrate and Na-tungstate in a 65–35% ratio while a similar study has been reported by Amano et al. [23]. However, there is a relatively wide gap in literature on effective preparation of robust structures with high specific surface areas in order to promote efficient disinfection, and a gap in interpreting the pathways to bacterial inactivation by this catalyst.

This study aims to assess a facile preparation method for flower-like Bi2WO6 photocatalysts destined for disinfection applications. As such, we assess the preparation parameters (aging, temperature, pH) in order to modify the structural (crystalline) and morphological characteristics (flower-like, nanoparticles). These modifications are envisioned to create a series of catalysts, and their activity under low-intensity solar or visible light will be assessed. Furthermore, the robustness of the catalyst in serial reuse cycles will be evaluated for its stability. Last but not least, special focus will be given to the identification of the pathways that lead to bacterial inactivation in an effort to decrypt the mechanistic action mode of the flower-like Bi2WO6.

### **2. Materials and Methods**

## *2.1. Synthesis of Flower-Like Bi2WO6 Samples*

All chemicals were of analytical grade. They were used as received without any further purification and were purchased from Merck, Germany. All solutions were prepared with Milli-Q water (18.2 MΩ cm). In a typical hydrothermal procedure for the synthesis of flower-like Bi2WO6, 0.5 mmol of Na2WO4·2H2O was dissolved in an 80 mL solvent containing 16 mL acetic acid and 64 mL Milli-Q water until attaining a clear solution. Then, 1 mmol of Bi(NO3)3·5H2O solid was added to the solution, and a white precipitate immediately emerged. Next, the reaction mixture was stirred for 1 h, transferred into a 120 mL Teflon-lined stainless-steel reactor, and heated at 160 ◦C for 12 h. The as-formed yellow precipitates were collected, washed with distilled water, and dried in vacuum at 70 ◦C for 10 h. A schematic representation of the synthesis is illustrated in Scheme 1. The influence of the hydrothermal reaction time and temperature has been explored as shown in Table 1. In order to investigate the effect of morphology on photocatalysis, Bi2WO6 nanoparticles (BWO6) were prepared applying the same hydrothermal method at 200 ◦C for 24 h by the regulation of pH to 10.

**Scheme 1.** Schematic illustration of the preparation of flower-like Bi2WO6 by hydrothermal method.


**Table 1.** Bi2WO6 obtained at the different synthetic conditions.

## *2.2. Physical Characterization of the Bi2WO6 Flakes*

The crystallinity and phase identification of the as-prepared samples were determined by powder X-ray diffraction (XRD) using an X'Pert MPD PRO (Panalytical) analyzer, equipped with a ceramic tube (Cu anode, λ = 1.54060 Å), and with a continuous scanning rate in the range of 5◦ < 2θ < 80◦. The results were studied with Rietvield refinement by the FullProf program. The morphology developments of the samples were characterized using scanning electron microscopy (SEM, FEI Quanta 200). Before SEM imaging, the samples were coated with a thin layer of gold. The specific surface area and porosity size were obtained using Brunauer–Emmett–Teller (BET) analysis, performed with a BELSORP-mini II analyzer, Japan. The photoluminescence (PL) measurement was carried out using a fluorescence spectrophotometer (Perkin Elmer LS55) equipped with a xenon lamp at an excitation wavelength of λ = 340 nm. The surface atomic percentage of the element in the as-synthesized sample was analyzed using an AXIS NOVA photoelectron spectrometer with a mono-chromatic Al Ka X-ray (hν = 1486.6 eV) source (Kratos Analytical, Manchester, UK). The interfacial in situ voltage and pH variation during the bacterial inactivation was monitored in a pH/mV/Temp meter (Jenco 6230N) equipped with a microprocessor and a RS-232-C IBM interface for data recording.

## *2.3. Photocatalytic Antibacterial Activity on Bi2WO6 and Light Sources*

The bacterial strain used was a wild type *E. coli* K12, supplied by the German Collection of Microorganisms and Cell Cultures, DSMZ (No. 498). The master plate and stock solution were prepared according to previous research reported by our laboratory [24,25]. The bacterial concentration

of the samples was measured in Colony Forming Units (CFU/mL) and was determined by plating on a non-selective cultivation media, namely, Plate Count Agar (PCA). A total of 1 mL of the sample was withdrawn after each interval and then serial dilutions were made in a sterile 0.8% NaCl/KCl solution. A 100 μL aliquot was pipetted onto a nutrient agar plate and processed using the standard plate count method. The plates were incubated at 37 ◦C followed by the bacterial evaluation. Experimental results were carried in triplicate runs applying statistical analysis for the calculation of mean and standard deviation (reported in the graphs). Samples were irradiated in the cavity of a SUNTEST solar simulator CPS (Atlas GmbH, Hanau, Germany) with an overall light irradiance of 48 mW/cm2 (~0.8 <sup>×</sup> 1016 photons/s, Supplementary Figure S1). A cut-off filter was used in the SUNTEST cavity to filter the light <310 nm. A second cut-off filter was also used during bacterial inactivation under visible light with a cut-off blocking the wavelength < 405 nm rendering (Supplementary Figure S2). Finally, after the two filters, the visible light irradiance reaching the sample was 38 mW/cm2.

## **3. Results**

#### *3.1. Synthesis and Characterization of Bi2WO6: X-Ray Di*ff*raction (XRD), Scanning Electron Microscopy (SEM), X-Ray Photoelectron Spectroscopy (XPS), and SSA Determination*

Figure 1 depicts the XRD patterns of the as-synthesized Bi2WO6 via the hydrothermal method at different reaction times and temperatures. All of the XRD patterns illustrated that characteristic peaks were in good agreement with the orthorhombic phased Bi2WO6 in the standard JCPDS card (39-0256) [26]. No other diffraction peaks arising from possible impurities were detected. With the holding time increasing to 24 h, the characteristic peaks became much sharper due to an increase in crystallinity. Understandably, the increment of the temperature with the constant reaction time for 24 h resulted in the same trend because of grain growth. Table 2 illustrates the crystallite size of the samples (using the Scherrer formula based on the half-width of their (113) peak) calculated by the Rietveld method using the FullProf program.

**Figure 1.** X-ray diffraction patterns of the Bi2WO6 samples: (**a**) BWO1, (**b**) BWO2, (**c**) BWO3, (**d**) BWO4, (**e**) BWO5, and (**f**) BWO6. Profiles are shifted in y-scale for clarity.


**Table 2.** Rietveld structural parameters of the samples.

The scanning electron microscopy (SEM) images of the Bi2WO6 samples prepared under different experimental conditions are shown in Figure 2. Heating at 160 ◦C for 12 h and 18 h (Figure 2a,b) led to aggregated irregular small Bi2WO6 nanoparticles and flower-like microspheres. However, when the heating time was prolonged to 24 h (see Figure 2c), organized hierarchical flower-like Bi2WO6 microspheres composed of nanoplates were obtained and the aggregated nanoparticles totally disappeared (Figure 2e,f). The SEM images of as-prepared Bi2WO6 nanoparticles are also shown in Figure 2f. The joint effect of nanoparticles assembly followed by the localized ripening mechanism as well as the hierarchical assembly of nanoplates have been also previously reported for the formation mechanism of flower-like microspheres [27,28]. Owing to the absence of discrete nanoplates according to the SEM images at different reaction times (Figure 2), the former mechanism seems to predominate.

Scheme 2 illustrates the proposed formation mechanism of flower-like Bi2WO6 microspheres. Nanoparticles initially aggregated, then the self-assembled nanoparticles preferentially grew along <010>. Longer reaction times and higher temperatures result in dissolution of some nanoplates leading concomitantly to re-deposition by Ostwald ripening [27,28].

The relevant reactions leading to the Bi2WO6 synthesis in aqueous solutions when working in acetic acid media can be suggested as follows:

$$\rm{Bi(NO\_3)\_3} \rightarrow \rm{Bi^{3+}} + \rm{3NO\_3}^-\tag{1}$$

$$2\text{ Na}\_2\text{NO}\_4 \rightarrow 2\text{Na}^+ + \text{ } \text{NO}\_4^{2-} \text{ } \text{ } \tag{2}$$

$$\text{CH}\_3\text{-}-\text{COOH} + \text{H}\_2\text{O} \rightarrow \text{CH}\_3\text{-}-\text{COO}^- + \text{H}\_3\text{O}^+ \quad pK \text{4.75}, \tag{3}$$

$$\left|Bi\_3^{\text{+}} + WO\_4^{\text{2-}} + CH\_3-COO^- \rightarrow CH\_3COO^- + \begin{array}{c} CH, \neg COO^- \\\\ CH, \neg COO^- \end{array} \right| \tag{4}$$

$$\begin{array}{rcl} \stackrel{CH\_3\text{-}COO^-}{\longrightarrow} & & \stackrel{}{\longrightarrow} Bi\_2WO\_6 & \quad & \text{follower-like microsphere} \end{array} \text{(see Figure 2)} \\ \text{?} \begin{array}{rcl} \text{(\ $ \text{-}COO^-)}{\longrightarrow} & & \text{(\$  \text{-}COO^-)} \\ \text{(\ $ \text{-}COO^-)} & & \text{(\$  \text{-}COO^-)} \end{array} \text{(\ $ \text{-}COO^-)} \\ \text{(\$  \text{-}COO^-)} & & \text{(\ $ \text{-}COO^-)} \end{array} \text{(\$  \text{-}COO^-)} \\ \text{(\ $ \text{-}COO^-)} & & \begin{array}{rcl} \text{(\$  \text{-}COO^-)} & \text{(\ $ \text{-}COO^-)} \\ \text{(\$  \text{-}COO^-)} & & \text{(\ $ \text{-}COO^-)} \end{array} \text{(\$  \text{-}COO^-)}$$

**Figure 2.** Scanning electron microscopy (SEM) images of Bi2WO6 samples prepared under different conditions: (**a**) BWO1, (**b**) BWO2, (**c**) BWO3, (**d**) BWO4, (**e**) BWO5, and (**f**) BWO6.

The initial complex between Bi and acetic acid presents a stability constant of 102.6–2.7 [29], which is not in the range found for insoluble complexes/precipitates >1011–12 [30–32]. This coordination complex is suggested in Equation (4) (Bi = M). The complex formation which is the precursor of Bi2WO6 does not lead to precipitate formation and gradually decomposes releasing Bi3<sup>+</sup> which reacts with WO4 2−. Therefore, the nanoplate formation leads to aggregates which present inner pores/voids and provide the required contact area for the photocatalytic bacterial inactivation.

In addition to the crystal structure and morphology, the surface chemical composition of the as-synthesized flower-like sample at 200 ◦C for 24 h was examined by XPS. As shown in the survey XPS spectrum in Figure 3, the Bi, O, W, and C elements were present in the pure Bi2WO6. The C element peak can be attributed to adventitious carbon from the sample preparation and/or the XPS instrument itself [33]. The surface atomic concentration ratio of Bi:W:O estimating from XPS peak areas is around 2.0:0.8:5.4, which further confirms its composition of Bi2WO6. Furthermore, the peaks centering at 164.7 and 159.4 eV are attributed to the binding energies of Bi 4f5/<sup>2</sup> and Bi 4f7/2, respectively (inset of Figure 3), confirming Bi3<sup>+</sup> ions in the crystalline structure [34–36]. The W4f energy region can be designated to be the +6 oxidation state of tungsten in accordance with previous reports [33,36].

**Scheme 2.** Schematic illustration of the growth process of the flower-like Bi2WO6 microspheres.

**Figure 3.** XPS survey spectra of the hydrothermally prepared Bi2WO6 sample at 200 ◦C for 24 h. Inset is the zoom of XPS scans over the Bi4f7/<sup>2</sup> peak in the 154–170 eV region.

The N2 adsorption–desorption isotherms of the well-organized flower-like (BWO5) and nanoparticles (BWO6) Bi2WO6 are presented in Figure 4. According to IUPAC classification, it can be seen that the isotherm shape for both samples exhibited a typical type IV isotherm with a clear hysteresis loop H3, suggesting the presence of mesopores in the size range of 2–50 nm [37].The insets show the Barrett–Joyner–Halenda (BJH) pore-size distributions and present the evidence for the existence of mesopores (2–50 nm). Table 3 summarizes the BET specific surface areas (SSA) and the pore volumes of BWO5 and BWO6.

**Figure 4.** N2 adsorption–desorption isotherm of the samples: (**a**) flower-like Bi2WO6, (**b**) nanoparticle Bi2WO6. The insert shows the pore size distribution.



*3.2. E. Coli Inactivation Kinetics: E*ff*ect of the Bacterial Concentration, Amount of Catalyst, Light Dose, and Applied Light Wavelength*

Figure 5 shows the complete bacterial inactivation mediated by the BWO5 being faster under low-intensity simulated solar light, compared to the other samples. The *E. coli* inactivation was 95% after 2 h. The effectiveness of a disinfection process resides in the time necessary to inactivate a determined percentage of bacteria. In the Chick–Watson model [38,39], the simplest inactivation model, the inactivation rate shown in Figure 5 is seen to be dependent on the residual bacteria after each specific time during the inactivation process and this allows comparing the effect of the different Bi2WO6 samples. Neither irradiation in the absence of Bi2WO6 (photolysis) nor runs in the presence of

this catalyst in the dark lead to bacterial inactivation of up to 4 h. The latter provides the proof that Bi2WO6 is not toxic to *E. coli* and a photocatalytic process is required for their inactivation. As the treatment time increased, the photocatalytic process became more effective, owing to the formation of hierarchical flower-like Bi2WO6 microspheres and loss of aggregates and the higher crystallite size. Nevertheless, the nanoparticles (BWO6), which presented lower specific surface area than BWO5, led to lower inactivation rates. The pseudo first-order rates of the Bi2WO6 samples during flower-like development (BWO1 and BWO5) compared with Bi2WO6 nanoparticles (BWO6) are given in the supplementary material, Figure S3. The pseudo first-order rate constants (*kapp*) of the BWO1, BWO5, and BWO6 were estimated to be 0.0331 min<sup>−</sup>1, 0.0488 min−1, and 0.0195 min−1, respectively. As can be seen, the photocatalytic inactivation of bacteria mediated by as-developed flower-like Bi2WO6 (BWO5) is around 2.5 times faster compared with nanoparticles.

**Figure 5.** Photocatalytic inactivation of *Escherichia coli* in aqueous dispersions on different Bi2WO6 samples in the dark and under simulated solar light (SSL). Experimental conditions: [Catalyst]0 = 0.2 g/L, [bacteria]0 <sup>=</sup> <sup>2</sup> <sup>×</sup> <sup>10</sup><sup>6</sup> Colony Forming Units (CFU)/mL and light intensity: 48 mW/cm2.

The photoluminescence spectrum of the prepared catalysts was used as a practical method to verify the separation efficiency of photo-generated electron–hole pairs in the semiconductors. Generally, a lower photoluminescence (PL) intensity represents a lower recombination rate of photo-generated charge carriers. The photoluminescence (PL) spectra of the Bi2WO6 samples during the flower-like development (BWO1 and BWO5) in comparison with Bi2WO6 nanoparticles (BWO6) is shown in Figure 6. The wide absorption-band was observed between 350 nm and 600 nm which is due to the Bi2WO6 electron-hole recombination giving rise to the free and bound-exciton luminescence [40]. The PL spectra of the as-synthesized samples through flower-like development (BWO1 and BWO5) exhibited significantly decreased PL intensity related to that of the Bi2WO6 nanoparticles. It could be ascribed that the recombination of photo-generated charge carriers is greatly inhibited in the hierarchically flower-like composed of nanosheets. Hence, the efficient separation of photo-generated electron–hole pairs and rapid transfer of electrons to the surface of crystal would be obtained. Moreover, the lower PL-intensity bands shown in BWO5 reflected a higher crystallinity in comparison with BWO1, allowing a lower amount of crystal defects, leading to a higher electron-hole separation and an

increased photocatalytic activity [41], a fact that corroborates with the faster inactivation of bacteria (Figure S3).

**Figure 6.** Photoluminescence (PL) spectroscopy of the synthesized samples at different conditions. BWO1: 12 h, 160 ◦C. BWO5, 24 h, 200 ◦C. BWO6: 24 h, 200 ◦C. pH = 10.

Following, the effects of initial catalyst or bacterial concentration were studied, and the results are summarized in Figure 7. The effect of the Bi2WO6 concentration on *E. coli* inactivation is shown in Figure 7a. Although increasing Bi2WO6 concentration of up to 0.2 mg/mL resulted in higher inactivation rates, increasing the catalyst concentration to 0.4 mg/mL resulted in a slower bacterial inactivation kinetics, most possibly due to a loss in surface area by catalyst agglomeration (particle–particle interactions), as well as a decrease in the penetration of the photon flux by the solution opacity, thereby decreasing the photocatalytic inactivation rate [42]. The effect of the initial concentration on the *E. coli* kinetics mediated by Bi2WO6 catalysts is presented in Figure 7b, showing a delay in the time necessary for bacterial inactivation at higher bacterial concentrations. Although this effect can be ascribed to the exhaustion of surface active sites due to opacity in solution [43], we note here that in absolute numbers, the higher the amount of bacteria in solution, the higher the number of available bacteria (for inactivation). Hence, by calculating the amount of cells inactivated in 4 h per mg of catalyst and per minute, we get 2075, 208, and 21 cells min−<sup>1</sup> mg−<sup>1</sup> for 108, 107, and 106, respectively. As a result, we report that this catalyst can effectively disinfect higher amounts of microorganisms, albeit in a higher residence time.

**Figure 7.** Effect of catalyst and bacterial concentration on inactivation kinetics. (**a**) *E. coli* survival on Bi2WO6 samples in the dark and under low intensity solar simulated light. Experimental conditions: (bacteria)0 = 2 <sup>×</sup> 10<sup>6</sup> CFU/mL and light intensity: 48 mW/cm2. (**b**) Initial concentration of *E. coli* (CFU/mL) effects on the bacterial inactivation kinetics mediated by Bi2WO6 (200 ◦C for 4 h) under low intensity solar simulated light. Experimental conditions: (Catalyst)0 = 0.2 g/L and light intensity: 48 mW/cm2.

Next up in the operational parameters investigation, we assessed the possibility of photonic limitation or saturation of the system. As such, Figure 8a,b shows the effects of the light intensity and composition (UVA–vis or Vis only) on the bacterial degradation kinetics. A higher light dose accelerated the bacterial inactivation because of a higher amount of charges generated in the semiconductor during bacterial disinfection under band-gap irradiation (Figure 8a), since the direct inactivation by light was previously excluded. Figure 8 b illustrates that under visible light, a solution containing 0.2 g/L of Bi2WO6 was still efficiently inactivating bacteria and was more effective compared to commercial TiO2 P25 Degussa (used as reference). These results come from the optical absorption of up to ~450 nm in the visible region by Bi2WO6, which is significantly wider than that of TiO2 P25 Degussa with an absorption of up to 387 nm for the 20 nm particles, making up the bulk of this mixed TiO2 P25 Degussa rutile–anatase [44].

**Figure 8.** Effect of light irradiance and composition on inactivation kinetics. (**a**) *E. coli* inactivation on Bi2WO6 (200 ◦C for 24 h) under different solar light irradiation intensities. (**b**) *E. coli* inactivation mediated by Bi2WO6 (200 oC for 24 h) and TiO2 under low intensity solar simulated (48 mW/cm2) and visible light (38 mW/cm2). Experimental conditions: (Catalyst)0 = 0.2 g/L and (bacteria)0 = 2 <sup>×</sup> 106 CFU/mL.

### *3.3. Mechanistic Interpretation: ROS-Species Involvement, Interfacial Charge Transfer, and Catalyst Reuse During Bacterial Inactivation*

༦ ༦ ༦ ༦ The reactive oxygen species (ROS) such as **·** OH, O2 **·**−, and vb (h+) play a pivotal role in the photo-degradation of organic pollutants and bacterial inactivation [22,45–47]. To determine the main ROS followed by the photodegradation mechanism, appropriate radical-scavengers such as isopropanol ( **·** OH scavenger), sodium oxalate (a vbh<sup>+</sup> hole scavenger), and superoxide dismutase (O2 **·**<sup>−</sup> scavenger) were used in the present study. Figure 9 depicts the results of scavenging experiments mediated by the

༦

optimized flower-like Bi2WO6 (BWO5). The photocatalytic bacterial inactivation could be remarkably suppressed by the addition of isopropanol and sodium oxalate. It is very likely that **·** OH and h<sup>+</sup> intervene jointly in the bacterial inactivation. Meanwhile, the addition of SOD (O2 **·**<sup>−</sup> scavenger) inhibits the bacterial inactivation to a smaller degree compared to vb(h+) and the **·** OH-radical as shown in Figure 9, traces (a) and (b). ༦ ༦

**Figure 9.** Effect of the scavengers during *E. coli* inactivation on Bi2WO6 under solar simulated light for (a) isopropanol as OH-radical scavenger, (b) sodium oxalate a hole vb(h+) scavenger, (c) superoxide dismutase (SOD) as an O2 .- scavenger, (d) no scavenger. Runs under low intensity solar simulated light (48 mW/cm2). The solutions contained Bi2WO6 (0.2 g/L) and scavenger concentration of 0.1 mM.

The possible reaction mechanism for the inactivation of *E. coli* mediated by Bi2WO6 can be proposed as the following, which is shown in Scheme 3. Under visible-light irradiation, the photo-excitation of Bi2WO6 implies the transfer of an electron from the valence band (Equation (6)).

$$
\mathcal{B}i\_2\mathcal{V}\mathcal{O}\_6 + hv \to e^-(\mathcal{C}B) + h^+(VB). \tag{6}
$$

As mentioned before, the valence band of Bi2WO6 is a hybrid band made up by the O2p and Bi6s orbitals. Under light irradiation, the O2p and Bi6s hybrid orbitals increase the charge transfer in the W5d orbitals of Bi2WO6. This moves the valence band (VB) potential to a more positive potential energy narrowing the band-gap and inducing a higher photocatalytic activity [48].

Based on the references, CB and VB potentials of Bi2WO6 are 3.08 and 0.36 eV, respectively [49,50]. The redox potential for the dissolved oxygen/superoxide couple (E0 (O2/ O2 **·**<sup>−</sup>)), O2/HO2 **·** , and OH- / **·** OH are −0.33 eV, −0.046 eV, and 1.98 eV vs NHE [49], respectively. Comparing the band edge energy level of Bi2WO6 with the redox potentials of ROS, it is obvious that the excited holes in the valence band of Bi2WO6 were sufficiently more positive than that of OH- / **·** OH, suggesting that the photogenerated holes on the surface of Bi2WO6 could react with OH- /H2O to form "non-selective" **·** OH radicals (Equation (7)). However, the conduction band edge potential of Bi2WO6, which is more positive than the standard redox potential of O2/ O2 **·**<sup>−</sup> and O2/HO2 **·** , cannot directly reduce O2 to O2 **·**<sup>−</sup> or HO2 **·** . As shown in Figure 9, the bacterial inactivation is reduced in the presence of SOD-scavengers, which confirms the presence of the HO2 **·** radicals. Considering the redox potential of O2/H2O2 = +0.682 eV vs NHE [51], H2O2 seems to be generated initially (Equation (8)) which is followed by the formation of different species according to the relations 9–10 in the photocatalytic reaction. It is worth noting that the powerful hole can directly attack bacteria cells in the photocatalytic oxidation process, which was also confirmed by the hole scavenger [45,52].

**Scheme 3.** Schematic diagram showing the photocatalytic inactivation of bacteria on the Bi2WO6.

$$\rm{H}^+(VB) + \rm{H}\_2\rm{O} \to \rm{HO} \bullet + \rm{H}^+, \tag{7}$$

$$2\text{ O}\_2 + 2H^+ + 2e^- \rightarrow H\_2\text{O}\_2,\tag{8}$$

༦ ༦

$$
\theta\_2 O\_2 + \varepsilon^- \rightarrow OH\bullet + OH^-,\tag{9}
$$

$$H\_2O\_2 + h^+ \rightarrow O\_2^- \\
\bullet + 2H^+ \quad or \quad HO\_2^- \\
\bullet + H^+. \tag{10}$$

Figure 10 shows the variation of the interfacial potential and the local pH shift under simulated solar light. At pH ~6, the bacterial inactivation preferentially proceeds via the O2 **·**<sup>−</sup> species over HO2 −**·** as shown in Equation (11) and Figure 9, trace (c).

$$\text{PbO}\_2^{\text{--}} \Leftrightarrow \text{O}\_2^{\text{--}} + \text{H}^+ \qquad \qquad \text{pK}\_4 \quad \text{4.8.} \tag{11}$$

The initial pH at time zero in Figure 10 was observed to decrease slightly from 6.0 to 5.9 within four hours of irradiation. The initial pH of 6.0 in this figure is seen to decrease drastically to 5.4 after 8000 s due to the concomitant production of long-lived intermediates carboxylic acids, owing to the degradation of the bacterial membrane. The interface potential is shown to drastically drop within 8000 s (2.2 h) when the bacterial reduction is reduced by 99.90%, which is equivalent to 3 logs as shown in Figure 5. The interface potential recovers to its initial value as shown in Figure 10 after the inactivation of bacteria [52]. The recovery to the initial pH-level occurs when the intermediate acids are mineralized to CO2 by the photo-Kolbe reaction according to Equation (12) [53,54].

$$\rm{RCOO^{-}} + \rm{solar light} \rightarrow \rm{R} + \rm{CO^{2}}.\tag{12}$$

**Figure 10.** Evolution of the interfacial potential and local pH of an *E. coli* suspension in contact with Bi2WO6 under low intensity light irradiation (48 mW/cm2). Catalyst concentration 0.2 g/L.

Finally, we provide the evidence for synthesizing a stable Bi2WO6 flower-like photocatalyst by a repetitive inactivation of a *E. coli* test, which results are shown in Figure 11. In order to evaluate the bacterial inactivation after each cycle, the pseudo first-order rate constants (*kapp*) were calculated and are reported in Table 4. The recycled sample used in Figure 11 was thoroughly washed after each cycle. Practically, no loss of bacterial inactivation was observed. These results show the stable repetitive bacterial inactivation mediated by flower-like Bi2WO6 up to five cycles and confirm the potential for the practical application of this photocatalyst in *E. coli* inactivation.

**Figure 11.** Reusability of flower-like Bi2WO6 under low intensity solar simulated light (48 mW/cm2). Solution parameters: (Catalyst)0 <sup>=</sup> 0.2 g/L and (bacteria)0 <sup>=</sup> <sup>2</sup> <sup>×</sup> <sup>10</sup><sup>6</sup> CFU/mL.


**Table 4.** Pseudo first-order rate constants (*kapp*) for *E. coli* inactivation under different conditions consistent with Figure 11.

## **4. Conclusions**

In the present study, Bi2WO6 flower-like samples were prepared at 200 ◦C attaining a high crystallinity and led a low amount of crystal by hydrothermal growth in acetic acid media. By SEM, XRD, XPS, and PL analysis, the properties of the flower-like Bi2WO6 samples and nanoparticles were investigated. These catalysts resulted in effective bacterial inactivation even under visible light and were faster than TiO2. In addition to higher SSA of flower-like Bi2WO6, its lower PL intensity leads to lower recombination of photo-generated electron–hole pairs as a consequence of more efficient photocatalytic activity. The photocatalytic inactivation of bacteria mediated by as-developed flower-like Bi2WO6 (BWO5) is around 2.5 times faster when compared with nanoparticles. The samples under light lead to effective Bi2WO6 charge separation and the generation of ROS inducing bacterial inactivation. The intermediate ROS species produced by Bi2WO6 were identified by the use of the appropriate scavengers, and the **·** OH-radical was identified to be the dominant inactivation mechanism. Finally, the stable performance of the synthesized catalyst during recycling indicates its robustness and may suggest practical application potential.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4441/12/4/1099/s1, **Figure S1.** SUNTEST solar simulator light wavelength emission spectrum (Manufacturer: Atlas, CPS+/CPS Instruments Brochure), **Figure S2.** Transmittance of the polymethylmethacrylate filter used to block UV light, **Figure S3.** Pseudo first-order rates of the Bi2WO6 samples during flower-like development (BWO1 and BWO5) compared with Bi2WO6 nanoparticles (BWO6).

**Author Contributions:** Conceptualization, M.K., S.G., C.P., S.R., and J.K.; methodology, M.K., S.G., S.R., J.K., and C.P.; software, M.K.; validation, M.K., S.G., S.R., and J.K.; investigation, M.K.; resources, C.P., F.K., and K.R.; writing—original draft preparation, M.K., F.K., K.R., S.R., J.K., S.G., and C.P.; writing—review & editing, M.K. and S.G.; visualization, M.K.; supervision, F.K., K.R., S.R., J.K., and C.P.; project administration, C.P., F.K., and K.R.; funding acquisition, S.G., C.P., F.K., and K.R. All authors have read and agreed to the published version of the manuscript.

**Funding:** Minoo Karbasi obtained exchange scholarship from the Ministry of Science, Research and Technology of Iran and the Isfahan University of Technology, and its contribution is hereby acknowledged. Stefanos Giannakis would like to acknowledge the Spanish Ministry of Science, Innovation and Universities (MICIU) for the Ramón y Cajal Fellowship (RYC2018-024033-I).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## **Degradation of Hexacyanoferrate (III) from Gold Mining Wastewaters via UV-A**/**LED Photocatalysis Using Modified TiO2 P25**

## **Augusto Arce-Sarria 1,2, Kevin Mauricio Aldana-Villegas 1, Luis Andres Betancourt-Buitrago 1, Jose Ángel Colina-Márquez 3, Fiderman Machuca-Martínez <sup>1</sup> and Miguel Angel Mueses 3,\***


Received: 6 July 2020; Accepted: 6 September 2020; Published: 10 September 2020

**Abstract:** The photocatalytic degradation of potassium hexacyanoferrate (III) was assessed in a bench-scale compound parabolic collectors (CPC) reactor assisted with a light-emitting diode (LED) UV-A source emitting at 365 nm, and using a modified TiO2 as a catalyst via the hydrothermal treatment of commercial Aeroxide P25. The experiments were performed under oxic and anoxic conditions in order to observe a possible reduction of the iron. The modified TiO2 showed a specific surface area 2.5 times greater than the original Aeroxide P25 and its isotherm and hysteresis indicated that the modified catalyst is mesoporous. The bandgap energy (Eg) of the modified TiO2 increased (3.34 eV) compared to the P25 TiO2 band gap (3.20 eV). A specific reaction rate constant of 0.1977 min−<sup>1</sup> and an electrical oxidation efficiency of 7.77 kWh/m<sup>3</sup> were obtained in the photocatalytic degradation. Although the TiO2 P25 yields a photocatalytic degradation 9.5% higher than that obtained one with the modified catalyst (hydrothermal), this catalyst showed better performance in terms of free cyanide release. This last aspect is a significant benefit since this can help to avoid the pollution of fresh water by reusing the treated wastewater for gold extraction. A photocatalytic degradation of the cyanocomplex of 93% was achieved when the process occurred under oxic conditions, which favored the removal. Summarizing, the hydrothermal method could be a promising treatment to obtain TiO2-based catalysts with larger specific areas.

**Keywords:** photocatalysis; UV-LED; TiO2; hexacyanoferrate; mining; hydrothermal method

## **1. Introduction**

Small and medium industries of gold extraction use the leaching process with sodium cyanide for mining the gold contained in the extracted ore, before precipitation of the metallic gold in the presence of zinc. During the process, the cyanide extracts undesired metals and thus forms several types of cyano complexes. The produced wastewater is rich in metallic complexes that are formed when the free cyanide interacts with the different metals present in the ores such as Ni, Fe, Co, Au, Ag, etc. These cyano complexes are very stable and recalcitrant compounds, which are hard to remove by natural remediation, resulting in the pollution of rivers, lakes and groundwater sources. Besides, solar photolysis releases free cyanide, which is highly harmful to ecosystems [1]. Advanced oxidation processes (AOPs), such as ozone-based treatments, alkaline chlorination, hydrogen peroxide-based

processes, biological and photocatalytic processes, can be used as alternative treatment technologies for these mining wastewaters [2].

The heterogeneous photocatalysis is an AOP where a solid semiconductor, assisted by UV radiation, promotes the generation of free hydroxyl radicals (•OH) and the degradation of diverse pollutants. The most commonly used semiconductor is the titanium dioxide (TiO2), and it can be used as a base oxide for the synthesis of other photoactive catalysts as well. The TiO2 is preferred because of its low cost, easy handling, and low toxicity. In general, when the photocatalyst is irradiated with photons with energy greater than the bandgap (Eg) of the semiconductor, the excited electrons are promoted from the valence band to the conduction band of the semiconductor, leading to the formation of electron–hole pairs. The strong oxidative potential of the holes (*h*+) oxidizes the hydroxyl anions of water for generating •OH, whereas the electrons of the conduction band can react with oxygen for generating superoxide ions (O2 •—) or promote other reduction reactions. Those radicals are the main species responsible for the oxidation reactions in the photocatalytic process [3,4].

To improve the semiconductors' •OH-generating performance, several studies have been focused on the preparation of semiconductors with enhanced radiation absorption. Different methods of preparation have been reported, namely hydrothermal [5], sol-gel [6], anodic oxidation, template method, and chemical vapor deposition (CVD) [7,8]. TiO2 catalysts doped with rare earth and transition metals have been modified to improve their •OH electron transfer properties. Some modifications on their morphology have also been made to produce structures such as nanorods, nanotubes, nanospheres, nanoflowers, among others [9–12].

The hydrothermal method has been widely used for the nanomaterial synthesis of TiO2 with diverse morphologies. This methodology is controlled by different variables, namely the precursors used, pH, temperature and reaction time [13]. Nowadays, TiO2-based nanowires, TiO2 nanotubes [14], carbon nanotubes [15], nanofibers, nanoflowers, and others have been successfully modified by hydrothermal treatment [16]. This method has become a very important tool for obtaining advanced materials due to its advantages, such as low cost, low operating temperatures, energy saving and lower impact to the environment (according to the principles of green chemistry) [10,12,17], in comparison to anodic oxidation and CVD methods. The hydrothermal treatment has been applied to the synthesis of nitrogen and carbon co-doped TiO2 [18], Sn-doped TiO2 nanoparticles composites [19], silica-titania combination of sol-gel-hydrothermal TiO2 nanoparticles [20], and both anatase and rutile TiO2 [21]. Moreover, several applications of TiO2 nanoparticles synthesized by the hydrothermal method have been reported such as hydrogen production via CO2 reduction, degradation of emergent pollutants and selective oxidation [22].

Huang and Chien [23] showed that the degradation of methylene blue increases from 65% to 95% with titania nanotubes compared to the powder. Camposeco et al. [24] compared the degradation between nanotubes and Evonik P25, showing that the catalytic activity was improved from 54 to 93% for methylene blue degradation and from 37% to 60% for the elimination of methylene orange. However, there is a lack of specific information about the use of titania modified via hydrothermal process for treating gold mining wastewater under UV/LED radiation.

In this work, the degradation of potassium hexacyanoferrate, which is a complex occurring as a by-product in gold mining wastewaters, via photocatalysis with hydrothermally treated TiO2, was studied. The mechanism proposed by Grieken et al. 2005 [25] or the hexacyanoferrate (III) reduction to hexacyanoferrate (II) and the subsequent degradation by heterogeneous photocatalysis is depicted in Figure 1. After the progressive abatement of the CN− groups in the molecule, the free cyanide can remain stable in solution due to the high pH of treatment or to produce cyanate by photocatalytic degradation, which is less toxic than the free cyanide. Nonetheless, the free cyanide is an advantage if the treated wastewater can be reused for the gold extraction. This would reduce the fresh water and cyanide consumptions and a consequent diminution of cyanide presence in water bodies.

The mechanism of free cyanide release is congruent with the reported literature [26–29]. The oxic conditions were analyzed in order to compare these results with the obtained ones in our previous work [27]. A further contribution respect to the reported literature is the use of the modified P25 via hydrothermal treatment and its potential improvement for the potassium hexacyanoferrate removal.

**Figure 1.** Mechanism of the heterogeneous photocatalytic degradation of Hexacyanoferrate [25]. Reprinted from Applied Catalysis B: Environmental, 55, Rafael van Grieken \*, José Aguado, María-José López-Muñoz, Javier Marugán, Photocatalytic degradation of iron–cyanocomplexes by TiO2 based catalysts, 201–211, Copyright (2005), with permission from Elsevier.

The photocatalytic performance of the obtained titania was evaluated by analyzing the effect of the catalyst load on the overall efficiency of the photodegradation under both oxic and anoxic conditions. In addition, the impact of the variation of the power supplied by the UV source and of the initial concentration of the cyanocomplex, was assessed. All the experiments were carried out in a bench-scale compound parabolic collector (CPC) photoreactor with artificial UV/LED radiation.

#### **2. Materials and Methods**

#### *2.1. Catalyst Treatment*

The catalyst was modified by using the hydrothermal treatment [30–34]. Six grames of Aeroxide P25 (Evonik®, Essen, Germany) were mixed with 100 mL of a 10-M solution of NaOH (Merck, Darmstadt, Germany). The solution was stirred to avoid the formation of agglomerates and then it was decanted into a 120-mL beaker. Subsequently, it was transferred to a stainless-steel sealed reactor. The reactor temperature increased up to 120 or 180 ◦C during 24 or 72 h, according to the 23 experimental design described in Table 1. The white precipitate was washed with a 0.1-M HCl (Merck, Darmstadt, Germany) solution under stirring. The solid was recovered by centrifugation followed by a series of washing cycles with deionized water until the pH of the supernatant was 7.4. After drying the solid at 100 ◦C for 24 h, it was calcinated at 400 or 500 ◦C during four hours, with a heating gradient of 10 ◦C/min. Figure 2 shows the detailed procedure for the synthesis of photocatalysts.

Table 1 shows the different conditions of reaction time, reaction temperature and calcination temperature used to prepare each of the eight catalysts. For the statistical analysis, an analysis of variance (ANOVA) was carried out, considering a significance level of 0.05.

**Figure 2.** Schematic diagram of the hydrothermal synthesis method used.



#### *2.2. Catalyst Evaluation*

The evaluation of the performance of the modified catalysts was carried out in a bench-scale CPC reactor assisted by a UV/LED radiation source [27,35]. The reactor consisted of four Pyrex tubes with an outside diameter of 2 cm and a length of 11 cm, which were connected to a 750-mL container through a centrifugal pump. The input power of the centrifugal pump was 50 W. The container was sealed at the top with a stopper, which had openings for sampling and oxygen/nitrogen inlet to the gas diffuser [27].

Four 30 W LEDs (TaoYuan Electron Ltd. TY-365 nm, Hong Kong, China) connected in parallel, were used as the artificial light source. The light output was set up with a tilt angle of 115–125◦ and of 900–1200 mW of radiation intensity per LED [27,35]. Each LED (model GW GPS-3030D, GWINSTEK, Veldhoven, Netherlands) was equipped with a cooling system consisting of heat sinks and a 12-V fan. The UVA radiation intensity was measured with a UV radiometer (DELTA OHM model HD2102.2, Deltha Ohm S.r.l., Padova, Italy) and it was varied by adjusting the current intensity supplied to the LEDs at a constant voltage of 30 V. The reactor had a reactive volume and a total irradiated area of 138.23 cm<sup>3</sup> and 276.4 cm2, respectively. The ratio of the illuminated volume to the total volume was 0.23. This ratio is useful to characterize the reactive system volume used with respect to those used by other authors and thus be able to compare its performance.

Once the system was loaded with the matrix to be degraded, the LEDs were placed above the tubes at approximately 3 cm of height, whereas the parabolic collectors were placed below the reactor. The use of these reflective surfaces provides a more homogeneous distribution of the radiation reflected to reactor walls since the bottom of the tubes could be illuminated evenly [36].

The hexacyanoferrate III (K3[Fe(CN)6], CAS 13746-66-2, (Panreac AppliChem, Darmstadt, Germany) was selected as the model cyanocomplex of the gold mining wastewaters. The control experiments (physical adsorption, i.e., without light; or photolysis, i.e., without catalyst) were carried out with 60 mL of solutions of 100 ppm of the pollutant. For the physical adsorption experiment, the solution was kept under continuous stirring in a 500-mL beaker, under darkness conditions. For the photolysis experiment, the power of the UV-LEDs was set at 30 W that supplies the maximum intensity of UV radiation. For both experiments, an aliquot of 5 mL was taken every 10 min during two hours (time set for the reaction).

The results obtained for the removal were estimated with the Equation (1):

$$\%Dergradiation = \left(1 - \frac{C}{C\_0}\right) \times 100\tag{1}$$

where *C* is the final concentration and *C*<sup>0</sup> the initial concentration

For each optimization step, 500 mL of a solution of 100 ppm of hexacyanoferrate was prepared. For keeping the solution pH above 12, 1 mL of a 10 M solution of NaOH was previously added to 500 mL of hexacyanoferrate solution. After an adsorption stage carried out under darkness conditions for 20 min, the LEDs were turned on to perform the photocatalytic runs. The experiments were carried out at room temperature (20 ◦C) and 10 mL aliquots (less than 10% of the total volume) were taken at different time intervals. For oxic and anoxic experiments, air or nitrogen was sparged, according to the case, into the solution at a constant flow rate of 0.5 L/min. The optimization study was executed in four stages:


The hexacyanoferrate (III) concentration was followed by UV-VIS (JASCO V-730 spectrophotometer, Easton, MD, USA) at 303 nm, corresponding to its maximum absorbance wavelength in the UV spectrum. The measurement of total dissolved iron was performed using atomic absorption spectrometry (Thermo Scientific iCE 3000, Waltham, MA, USA) and the measurement of CN<sup>−</sup> by titration with AgNO3 according to the Standard Methods 4500 [37].

A kinetic law with a two-step reaction was used to describe the degradation of hexacyanoferrate (III). The first step (faster) corresponds to the adsorption of Fe(CN)6 <sup>3</sup><sup>−</sup> onto the surface of TiO2 and degradation of the iron modified, whereas the second step (slower) corresponds to the reduction of the iron present in the cyano-metallic complex (that corresponds to the removal of dissolved iron) [38].

For the kinetic analysis of the photo reductive process of the iron cyanocomplex, a pseudo first-order reaction rate equation was proposed (Equations (2) and (3)), as suggested by previous studies [39–41]:

$$-\frac{d\mathcal{C}}{dt} = k'\mathcal{C} \tag{2}$$

$$\ln(\mathbb{C}\_0/\mathbb{C}) = k't \tag{3}$$

where *k* is the pseudo first-order rate constant (min<sup>−</sup>1), *C*<sup>0</sup> and *C* are the initial and final concentrations of the iron complex in solution, respectively. The *ln* (*C*0/*C*) was plotted versus time for obtaining the *k*value, which is the slope of the equation of the line.

### *2.3. Characterization*

The crystalline phases of the resulting solid from the hydrothermal synthesis were characterized using X-ray diffraction (XRD) on a X'per PRO-PANalytical diffractometer with CuKα radiation (0.1542 nm) with a 2θ sweep between 0◦ and 90◦. The surface area was determined by the Brunauer–Emmett–Teller method (BET) by adsorption–desorption of nitrogen (N2) at 77 K and the volume and size of the pore were determined by the Barrett–Joyner–Halenda method (BJH) in a Micromeritics equipment ASAP 2020 V4.01 (Micromeritics, Norcross, GA, USA).

The morphology was analyzed by scanning electron microscopy (SEM) and X-ray energy dispersion spectrometry (EDS) was used for the analysis of elemental composition of the catalyst in a JEOL JSM 6490 LV brand equipment. The semiconductor bandgap (Eg) was estimated by measuring the material transmittance with UV-vis diffuse reflectance spectroscopy (UV DRS) in a Thermo Scientific Evolution 300 PC series EVOP068001 spectrophotometer. Finally, the Fourier-transform infrared spectroscopy (FT-IR) was used to identify the functional groups of the inorganic and organic substances (FT/IR-4100 type-A).

## *2.4. Estimation of the Electric Oxidation E*ffi*ciency (EEo)*

The IUPAC has proposed methods to calculate the electrical consumption of an AOP, depending on the type of reactor and the amount of contaminant to be treated. For low concentrations, it is proposed to use the electric energy per order (*EEo*). This parameter consists of the electrical energy (kWh) required to remove the pollutant up to 90% of its initial concentration per volume unit. The *EEo* can be calculated using the Equation (4), following the methodology proposed by Shirzad-Siboni et al. [41] and Daneshvar et al. [40]:

$$E\_{E\_\oplus} = \frac{1000 \text{ P } t}{60 \text{ V } \log(\text{C}\_0/\text{C}\_f)} \tag{4}$$

where *P* is the power supplied to the system (kW) and it is defined as the product of electric potential and the current intensity (A); *V* is the total reactive volume (L), and *t* is time (h). From Equations (3) and (4), the *EEo* can be calculated as follows:

$$E\_{E\_3} = \frac{38.4P}{Vk'} \tag{5}$$

#### **3. Results and Discussion**

#### *3.1. Photolysis and Adsorption*

The control tests in 3-h experiments showed that the photolysis contributes moderately to the removal of contaminants and the release of free cyanide. A 17% of photolytic removal of hexacyanoferrate and a 12% of cyanide release were achieved, which is in agreement with the results reported in this literature review [26]. On the contrary, the adsorption had a minor effect both in the elimination of contaminants and in the release of cyanide, respectively, 10% and less than 5% after three hours of experimentation. It was observed that 8% of the initial hexacyanoferrate concentration was adsorbed during the first 20 min of the experiment and therefore the dark period for the photocatalytic runs was set at 20 min.

#### *3.2. Evaluation of Synthesized Materials*

#### 3.2.1. Catalyst Load

This behavior observed in the Figure 3 is explained by the lower flow of photons into the reactive system resulting from the higher turbidity (catalyst loads higher than 0.5 g/L) of the slurry to treat [42]. This screening effect limits the effectiveness of the treatment by decreasing the local volumetric rate of photon absorption for tubular photoreactors, which has been analyzed by Colina-Marquez et al. in 2010 [43] and Mueses et al. in 2013 [44]. Those studies reported an optimal catalyst load of 0.3 g/L for CPC reactors, approximately. In turn, Osathaphan et al. [45] used catalyst loads between 0.1 and 4 g/L without affecting the reductive treatment considerably. Given the best results when using 0.5 g/L of both SL400 and SL500, both catalysts were promising to degrade the cyanocomplex. To select the

best performing catalysts modified, photocatalytic experiments were performed using 0.5 g/L of each catalyst to degrade the pollutant during 2 h of reaction. For the further experiments, 0.5 g/L of SL400 was selected, due to the better performance and also in order to save energy in the calcination process.

**Figure 3.** Evaluation of the best catalyst load to remove the cyanocomplex (hexacyanoferrate III) after 1 h of reaction (**a**) Catalysts calcined at 400 ◦C; (**b**) Catalysts calcined at 500 ◦C. Operating conditions: Initial pollutant concentration of 100 ppm, 20 min of adsorption, LED power supply of 20 W, air flow of 0.5 L/min.

Table 2 shows the results of the degradation obtained at different synthesis temperatures, calcination temperature and synthesis times.


**Table 2.** Degradation percentage of K3[Fe(CN)6] after 2 h of reaction, using a catalyst load of 0.5 g/L.

Operating conditions: initial pollutant concentration of 100 ppm, 20 min of adsorption, LED nominal power of 20 W, air flow of 0.5 L/min. Each experiment was done in duplicate.

A statistical analysis (see Table 3) of the information reported in Table 2 was carried out by using Statgraphics® Centurion XVI (version 16.2.04, Statpoint Technologies Inc., The Plains, VA, USA) and it was found that the calcination temperature was the most significant effect on the response variable within the evaluated intervals (see Figure 4), obtaining better results with 400 ◦C. The second most significant effect was the synthesis temperature and the best results were obtained at 120 ◦C; however, it is not statistically significant. Comparing the information of the table with the Pareto chart (Figure 4), it can be observed that the calcination temperature has a negative effect; that means that an increase of this variable represents a degradation decrease. This behavior can be attributed to the reduction of the surface area of the catalyst or material sintering at higher temperatures [6].


**Table 3.** ANOVA for degradation percentage of K3[Fe(CN)6].

**Figure 4.** Standardized Pareto Chart for degradation percentage of K3[Fe(CN)6].

On the other hand, although the synthesis time was not significant, its interactions with the other variables were meaningful and synergistic. This behavior is interesting because it means that a simultaneous increase of the calcination and synthesis temperatures with the synthesis time represents an improvement on the pollutant removal. In fact, the interaction between the synthesis time and the synthesis temperature (BC) is as significant as the effect of the calcination temperature. In addition, it was found that the best results for the degradation of the cyanocomplex were obtained for the catalyst modified at 24 h—120 ◦C to 400 ◦C (SL400). Considering all these facts, the following stages were carried out using SL400.

#### 3.2.2. Tests Under Oxic and Anoxic Conditions

The degradation of the cyanocomplex by photocatalysis using SL400 was evaluated under oxic and anoxic conditions, to evaluate the importance of the presence of oxygen (Figure 5).

**Figure 5.** (**a**) Cyanocomplex (hexacyanoferrate III) degradation; and (**b**) release of free cyanide. Operating conditions: Initial concentration of K3[Fe(CN)6] of 100 ppm, catalyst load of 0.5 g/L, 20 min of adsorption, Power supply: 20 W, air or nitrogen flow of 0.5 L/min, reaction time of 2 h.

After two hours of reaction, 56% of the cyanocomplex was degraded in the presence of oxygen, whereas it was only 29% when air was replaced by nitrogen. In turn, the cyanide release was two times higher when air containing oxygen was used (18 ppm in the presence of oxygen and 9 ppm using nitrogen). Finally, for the total removal of iron, a removal of 40% was achieved in the presence of oxygen and only 15% under an inert atmosphere. The higher degradation of the cyanocomplex and release of free cyanide in the presence of oxygen can be ascribed to the, electrons directly reducing iron and the oxidation of the complex by holes, hydroxyl radicals and superoxide anions. In contrast to our results that showed that the presence of oxygen during the reaction increases the degradation of the complex, Yang et al. [46] and Ku and Jung [47] reported a better performance of the P25 TiO2 for the removal of the studied contaminants under anoxic conditions. In these reports, the authors observed that the presence of oxygen did not have a significant effect on the contaminant removal, whereas a higher reduction was showed with nitrogen.

#### 3.2.3. Effect of the Radiation Intensity

The availability of UV photons directly affects the generation of electron–hole pairs. By comparing the results obtained at 10, 20 and 30 W (Figure 6a), it can be observed that the radiation intensity higher effect when increasing from 10 to 20 W than after a further increase to 30 W. Regarding to the degradation of the cyanocomplex, removals of 55, 73 and 79% were obtained with 10, 20 and 30 W, respectively. Additionally, iron removals of 30, 48 and 60% were achieved for 10, 20, and 30 W, respectively. The dissolved iron concentration was analyzed to corroborate its removal from the solution and its deposition onto the catalyst surface (Figure 6b).

**Figure 6.** (**a**) Degradation of the cyanocomplex (hexacyanoferrate III) at 10 W, 20 W and 30 W; (**b**) Total removal of dissolved iron. Operating conditions: Initial concentration of K3[Fe(CN)6] of 100 ppm, catalyst load of 0.5 g/L, 20 min of adsorption, air flow of 0.5 L/min, reaction time of 3.5 h.

The degradation values obtained with 20 and 30 W exhibited similar behaviors. An energy increase of 33% (20 to 30 W) yielded just an increase of 8.12% for the cyanocomplex degradation. This means that this energy increase is not enough to significantly affect the degradation performance. Therefore, the radiation intensity of 20 W was selected as the best condition due to the less energy consumption. Similar results were obtained by Rodriguez and Ossa [27], reporting a better but not significant performance when working at 30 than 20 W, and thus the selection of an inferior power supply to avoid an additional electrical consumption.

## 3.2.4. Comparison between Modified TiO2 and the Raw P25

By comparing the raw and treated TiO2, the degradation efficiency obtained with SL400 was 70%, whereas TiO2 P25 led to a photocatalytic removal of 80% (Figure 7a). In turn, 20 ppm of cyanide are released by SL400 and less 10% is observed for TiO2 P25, with 18 ppm of cyanide released (Figure 7b). Although for the complex degradation, the TiO2 P25 showed better results; regarding to the free cyanide release, the SL400 showed a performance 10% higher. As the initial concentration of contaminant

increases, the degradation decreases, as it was documented in the studies of Yang et al. [46] and Samarghandi et al. [39]. The cyanide release can be beneficial since it can be reused in the mining processes where such cyanide can be returned for the mineral (gold) re-extraction process. This feature would make the use of the synthesized material economically and environmentally attractive and also attenuate its weakness against P25 in terms of degradation of the hexacyanoferrate complex.

**Figure 7.** (**a**) Cyanocomplex degradation; (**b**) Free cyanide released. Operating conditions: 0.5 L/min of air bubbled, 0.5 g cat/L solution, 3.5 h of reaction with 30 min of adsorption and power supply of 20 W.

Van Grieken et al. [25] reported that the oxidative degradation of hexacyanoferrate (100 ppm of initial concentration) releases around 20 ppm of CN− in 240 min of irradiation by using mercury lamps. In this study, the same amount of cyanide ion was released in 210 min by using a UVA/LED photon source.

Table 4 shows the values found for the pseudo first-order speed constant (min<sup>−</sup>1) for a reaction time of 210 min. As it can be seen, the P25 TiO2 rate constants are higher than the SL400 ones for both initial concentrations of the pollutant. This can be explained because of the differences in superficial area, particle size distribution, semiconductor purity and other features in electronic properties.


**Table 4.** Pseudo first order rate constants.

3.2.5. Electric Oxidation Efficiency

Table 5 shows the *EEo* values obtained for the P25 and the SL400 sample with two different concentrations of hexacyanoferrate.

**Table 5.** Electrical oxidation efficiency for the catalysts used.


The P25 still exhibits better performance regarding to the energy consumption. This behavior is related to the higher activity of the commercial standard, which was discussed previously. The obtained results are similar to the reported ones by Daneshvar et al. [40], which did not exceed 10 kWh/m3. On the other hand, when the value obtained is compared with the study of Rodriguez and Ossa [27], it was found that the *EEo* is 40 and 20 times lower, respectively, than the presented ones in Table 5. In these works, it was reported the same concentration of Fe(CN)6 but with the use of different catalysts.

### *3.3. Characterization of the Photocatalyst*

#### 3.3.1. Fourier-Transform Infrared Spectroscopy (FT-IR)

The Figure 8 shows the IR spectra of the SL400 before and after usage in the photocatalytic experiments. Four bands are highlighted that are common in both spectra. As described by Thennarasu et al. [48], the peaks observed around 3300–3400 cm−<sup>1</sup> correspond to the stretching vibrations (stress) of the •OH and around 1600 cm−<sup>1</sup> arises from the water bending mode that can be associated with water absorbed by the catalyst due to the presence of moisture in the materials by contact with air. The main bands below 1000 cm–1 were attributed to the Ti-O and Ti-O-Ti bending vibrations. The band around 1300 cm−<sup>1</sup> is attributed to the C-H bending vibrations.

**Figure 8.** FTIR spectra of SL400 before and after the photocatalytic reaction.

#### 3.3.2. XRD Results

According to Mozia et al. [33], the peaks found at 2θ of 24◦, 28◦ and 48◦ as those observed for SL400 (Figure 9) correspond to titanates of the form A2Ti2O5·H2O and A2Ti3O7. The sodium titanates (Ti12O36Na4 or Ti3O9Na) exhibit peaks at 10◦, 24◦, 28◦, 48◦ and 62◦, which evidence the presence of the anatase phase of TiO2 at 25◦, 62◦, and 82◦. The analysis showed no significant amount of rutile since may be found at calcination temperatures over 600 ◦C.

**Figure 9.** XRD pattern for SL400 without use. The blue dotted lines represent the peaks associated with titanates and the black dotted line represents the peaks associated with the anatase phase of TiO2.

The most significant difference between the SL400 diffractogram (Figure 9) and that of P25 (Figure 10) without modifications [49], is the sharper peaks obtained by XRD for the commercial P25. This means a more crystalline structure for the unmodified P25 and some amorphous characteristics for the modified material (SL400). This modification affected the overall performance of the modified material regarding to the activity and, therefore, the pollutant removal. In addition, the XRD of SL400 does not have characteristic peaks of rutile phase as P25, which are known to improve the photocatalytic activity thanks to its synergistic effect with the anatase.

**Figure 10.** XRD pattern for TiO2 P25 [49].

#### 3.3.3. EDS Results

The Figure 11 shows a micrograph obtained from SL400. Additionally, an energy-dispersive X-ray spectroscopy analysis (EDS) was performed for elemental detection of the modified catalyst (see Figure 11). This analysis shows the type of elements present in different analyzed areas of the catalyst, where the presence of Carbon (C), Oxygen (O), Sodium (Na) and Titanium (Ti) were exhibited, with their respective composition, as shown in Table 6.

**Figure 11.** EDS analysis of the modified catalyst SL400.


**Table 6.** EDS results in % weight in the modified catalyst.

According to the EDS results, the presence of carbon in the material (3–6%), probably from impurities in the precursors used for the synthesis, can affect negatively the photocatalytic performance because of the number of active sites on the semiconductor surface decreases as the carbon occupies them.

#### 3.3.4. Surface Area Results

The surface area was 127.84 m2/g, which is greater than the surface area of the precursor material (50 m2/g). The pore volume of the total amount absorbed was 0.197 cm3/g and the pore size distribution analyzed by the BJH method was approximately 58 Å (5.8 nm) for an average particle size of 469 Å (46.9 nm). An isotherm of type IV was observed (Figure 12) with a hysteresis type III, which suggests that this catalyst is a mesoporous solid (2–50 nm).

Although the sample SL400 has a surface area higher than the P25's one, the number of active sites could not exceed the amount of sites of the TiO2 P25, since the modified catalyst did not exceed the photocatalytic activity of the precursor. In addition, the absence of rutile phase affects the overall activity of the TiO2, since this phase in the P25 acts synergistically with the anatase to improve the activity of the catalyst. The surface area is similar to those obtained by Turki et al. [50], Sikhwivhilu et al. [51] and Fen et al. [52]. On the other hand, some studies have obtained values higher than 200 m2/g as is the case of Thennarasu et al. [48] and Camposeco et al. [53] with important photocatalytic activity.

**Figure 12.** Isotherm absorption-desorption of the modified catalyst SL400.

#### 3.3.5. Bandgap Energy Estimation by DRS

The bandgap energy (Eg) is one of the most important parameters in the photocatalytic activity of TiO2 since it determines the effective wavelength interval for photon absorption. This parameter was estimated with the Kubelka-Munk theory according to the methodology reported by López and Gómez [54] (see Figure 13). It has to be considered that the crystal size, the particle size, the aggregation state of the particles, and the impurities present in the solid and the method of synthesis, can significantly affect the Eg.

**Figure 13.** Estimation of bandgap energy (Eg) of the catalyst SL400 using the Kubelka-Munk function, the red points represent those used to obtain the slope of the line and obtain the intercept on the x axis.

The energy of the bandgap obtained was 3.34 eV and the wavelength (estimated with Equation (6)) at which the catalyst is activated is 370 nm.

$$
\lambda = \frac{h \times c}{h\_v} \tag{6}
$$

If these values are compared with those reported for TiO2 P25 (Eg = 3.20 eV; λ = 385 nm), it is expected that the modified catalyst underperform respect to the commercial standard, regarding the UV photons absorption. This can be a significant drawback when it is intended to use a wide spectrum photons source.

#### **4. Conclusions**

The modified TiO2 P25, via the hydrothermal method, did not improve the Fe(CN)6 removal with respect to the obtained one with the original P25. This could be attributed to the loss of both the rutile phase and the material crystallinity. In addition, the increase of the bandgap energy for the modified P25 is another drawback since it affects the photon absorption by the semiconductor. Although the higher free cyanide release achieved with the modified material can be considered as a shortcoming regarding to the environmental potential of this material, in this particular case, this can be beneficial since this free cyanide could be reused for the gold extraction process and so, obtain a closed cycle for the water use. Furthermore, the increase of the specific surface area can be a promising result, in terms of physical adsorption of the studied pollutant or metallic cations.

While at a first sight the hydrothermal method did not improve the activity of the P25, further studies should be carried out to obtain more information about the structural modifications of the catalyst and potential advantages for photocatalytic applications.

**Author Contributions:** Conceptualization, A.A.-S. and L.A.B.-B.; methodology, A.A.-S. and L.A.B.-B.; software, L.A.B.-B. and A.A.-S.; validation, A.A.-S., L.A.B.-B., F.M.-M. and J.Á.C.-M.; formal analysis, K.M.A.-V. and A.A.-S.; investigation, K.M.A.-V., A.A.-S. and L.A.B.-B.; resources, F.M.-M.; data curation, K.M.A.-V. and L.A.B.-B.; writing—original draft preparation, K.M.A.-V.; writing—review and editing, M.A.M. and J.Á.C.-M.; visualization, M.A.M.; supervision, A.A.-S.; project administration, A.A.-S.; funding acquisition, F.M.-M. and M.A.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received external funding from "Departamento Administrativo de Ciencia, Tecnología e Innovación (COLCIENCIAS) through GRANT 1106-669-45250 "Recuperación de oro y tratamiento de aguas residuales cianuradas en la industria aurífera de la región pacífico colombiana" and the Universidad del Valle.

**Acknowledgments:** The authors are grateful to Universidad del Valle and the COLCIENCIAS for the Ph.D. scholarship 567-2012. Also, the authors thank the Biotechnology and Nanotechnology Laboratory of Tecnoparque (Nodo Cali) for the support with the analytical techniques. Colina-Márquez and Mueses thank the University of Cartagena. All authors send thanks to Ana Rita Lado Ribeiro from the University of Porto for the much-appreciated writing revisions of this manuscript.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## **A Pilot Study Combining Ultrafiltration with Ozonation for the Treatment of Secondary Urban Wastewater: Organic Micropollutants, Microbial Load and Biological E**ff**ects**

**Cátia A. L. Graça 1, Sara Ribeirinho-Soares 2, Joana Abreu-Silva 3, Inês I. Ramos 4, Ana R. Ribeiro 1, Sérgio M. Castro-Silva 5, Marcela A. Segundo 4, Célia M. Manaia 3,\*, Olga C. Nunes 2,\* and Adrián M. T. Silva 1,\***


Received: 8 October 2020; Accepted: 6 December 2020; Published: 9 December 2020

**Abstract:** Ozonation followed by ultrafiltration (O3 + UF) was employed at pilot scale for the treatment of secondary urban wastewater, envisaging its safe reuse for crop irrigation. Chemical contaminants of emerging concern (CECs) and priority substances (PSs), microbial load, estrogenic activity, cell viability and cellular metabolic activity were measured before and immediately after O3 + UF treatment. The microbial load was also evaluated after one-week storage of the treated water to assess potential bacteria regrowth. Among the organic micropollutants detected, only citalopram and isoproturon were not removed below the limit of quantification. The treatment was also effective in the reduction in the bacterial loads considering current legislation in water quality for irrigation (i.e., in terms of enterobacteria and nematode eggs). However, after seven days of storage, total heterotrophs regrew to levels close to the initial, with the concomitant increase in the genes 16S rRNA and *intI*1. The assessment of biological effects revealed similar water quality before and after treatment, meaning that O3 + UF did not produce detectable toxic by-products. Thus, the findings of this study indicate that the wastewater treated with this technology comply with the water quality standards for irrigation, even when stored up to one week, although improvements must be made to minimise microbial overgrowth.

**Keywords:** advanced oxidation; membrane technology; micropollutants; biological contaminants; cytotoxicity; wastewater reuse

## **1. Introduction**

Urban wastewater reuse is considered an important strategy when addressing water scarcity issues [1]. This is a common practice in some countries, where the treated wastewater is mostly directed

for agricultural irrigation [2]; however, urban wastewater often contains a variety of contaminants, such as salts, metals, metalloids, pathogens, and organic micropollutants, such as residual drugs, endocrine-disrupting chemicals, and residues from personal care products, among others [3,4]. Moreover, there is growing evidence that conventional urban wastewater treatment plants (UWWTPs) are not completely effective in eliminating bacteria and chemical micropollutants [5,6], rendering the effluent unsuitable for crops irrigation. Failure to properly treat and manage wastewater can generate adverse health effects, accumulation of heavy metals in crops, and the production of low-quality agricultural goods [3]. A new regulation on minimum preconditions for water reuse for agricultural irrigation has entered into force in the EU, which encompasses coordinated water-quality monitoring requisites for the safe reuse of treated urban wastewater [7]. These new rules will be put into practice in 2023 and are expected to promote water reuse. This regulation also demands an established water reuse risk management plan that should consider the environmental quality standards for priority substances and certain other pollutants, as well as additional requirements, such as heavy metals, pesticides, disinfection by-products, pharmaceuticals, and other substances of emerging concern, including micropollutants and microplastics. It also addressed the identification of some preventive measures that can be taken to limit risks, namely additional disinfection or pollutant removal measures.

Advanced oxidation processes (AOPs) and technologies (AOTs), such as ozonation, have emerged as effective tertiary treatments for the removal of both chemical and biological contaminants in UWWTPs [8,9]. Ozonation is among the few AOTs that have been applied to large-scale water treatment, due to its strong oxidation ability and broad-spectrum disinfection [10]. Ozone can react either by direct oxidation of organic pollutants (mostly at acidic conditions), or via hydroxyl radical formation (mainly produced under alkaline conditions) [10]. Studies employing ozone-based AOTs in UWWTP effluents have yielded remarkable results regarding the simultaneous removal of CECs and the reduction in the microbial load at different ozone doses and contact times [11–16]; however, bacterial regrowth in stored treated wastewater has been observed [14–16], which might be the result of the bacteria's ability to repair injuries, promoting fast regrowth, when stress levels are lowered. This may jeopardize water quality in the long term, thus prompting its immediate reuse rather than storing this water. Additionally, the use of chlorine as the traditional disinfection agent in stored water may not ensure its safety, because injured bacteria can also survive and regrow at low chlorine doses [17]. A suitable approach would be a physical separation step, using membrane-like technology. Although ozone may damage cell components, such as lipids, proteins and DNA, membrane filtration acts via size exclusion and adsorption, retaining microorganisms [18]. Among the available options in the market for full-scale applications, ultrafiltration (UF) membranes are favourable alternatives for bacteria removal due to their small pore size (0.01 to 0.1 μm). Moreover, studies have shown that UF is preferred to other filtration alternatives to avoid the regrowth of antibiotic-resistant bacteria (ARB) [19,20]. For example, Hembach et al., 2019 [18] reported the efficiency of UF in the disinfection of a secondary effluent of a UWWTP, and the results were compared with those obtained with single ozonation. The authors reported that UF (using a membrane pore size of 20 nm) was not able to remove the entire bacterial community, whereas ozonation presented limited effectiveness on the reduction in the same contaminants when using an ozone concentration optimised for micropollutant removal. Thus, these authors suggested further investigations coupling both technologies to achieve both micropollutant removal and bacteria mitigation, which was the target of the present study.

Thus, the present study investigated the potential of using UF in combination with ozonation, operating in continuous mode at a pilot scale, for the treatment of the secondary effluent of a UWWTP. Parameters commonly legislated in different countries were considered when assessing the suitability of treated wastewater for reuse in irrigation (Portuguese laws, US EPA, FAO guidelines and WHO). Moreover, envisaging higher quality criteria, the following parameters were also included in this work: (i) priority substances and CECs identified in Directive 2013/39/EU and Decision 495/2015/EU [21,22], respectively; (ii) load of selected microbial groups; and (iii) potential estrogenic activity, cytotoxicity, and cell viability (biological effects). All these parameters were analysed in both freshly collected and

O3 + UF treated wastewater to assess treatment efficiency. Biological effects are particularly important to evaluate, due to the possibility of formation of toxic by-products after ozonation. Moreover, microbiological indicators were re-examined after a 7-day storage period to assess potential bacteria regrowth. Regarding other studies coupling O3 to UF, only a few evaluate the feasibility of this system for urban wastewater reclamation [23–26] and, as far as it is known, none of those comprise the simultaneous evaluation of physico-chemical parameters, removal of priority substances and CECs, microbial inactivation and regrowth, and investigation of biological effects, which are important parameters for safe wastewater reuse, this work bringing a valuable contribution to the knowledge on this field.

#### **2. Materials and Methods**

#### *2.1. Chemicals and Materials*

All reference and isotopically labelled internal standards for liquid chromatography (>98% purity) were acquired from Sigma-Aldrich (Steinhein, Germany). Ethanol 99.5% (HPLC grade) was obtained from Fisher Scientific U.K. Ltd. (Loughborough, UK). Acetonitrile (MS grade) was purchased from VWR International (Fontenay-sous-Bois, France), whereas formic and sulphuric acid were obtained from Merck (Darmstadt, Germany). Multichannel tubular ceramic membranes with a selective layer of α-Al2O3 (nominal pore size of 10 nm) were provided by Rauschert Distribution GmbH, Inopor® (Schesslitz, Germany). Membrane dimensions were 305 mm in length with 15 mm glazed ends. The external diameter was 25 mm, and it contained 19 internal channels of 3.5 mm diameter each.

For microbial culture analyses, water samples were filtered through cellulose nitrate membranes (0.22 μm pore size, 47 mm diameter), provided by Sartorius (Gottingen, Germany). For DNA-based analyses, water samples were filtered through track-etched polycarbonate membranes (0.22 μm pore size, 47 mm diameter) from Whatman® NucleporeTM, provided by VWR (Alfragide, Portugal).

For cell culture experiments, dimethyl sulfoxide (DMSO; ≥99.9%), Triton™ X-100, and thiazolyl blue tetrazolium (MTT) were purchased from Sigma-Aldrich (Steinhein, Germany). Dulbecco's modified Eagle medium (DMEM; ref: 31966-021), heat-inactivated foetal bovine serum (FBS), penicillin-streptomycin (PenStrep), and trypsin-EDTA (1X) were purchased from Gibco® through Life Technologies™ (Warrington, UK). Murine fibroblasts L929 were obtained from the American Type Culture Collection (ATCC, Wesel, Germany). Caco-2 cell line was also purchased from ATCC and used between passage number 35 and 42. LDH Cytotoxicity Detection Kit was acquired from Takara Bio Inc. (Shiga, Japan). The XenoScreen YES/YAS assay kit for estrogenic activity assessment was acquired from Xenometrix® (Allschwil, Switzerland).

The ultrapure water used in the experiments and analytical methods was supplied by a Milli-Q water system (18.2 MΩ cm).

#### *2.2. Secondary E*ffl*uent and Treated Samples*

The secondary effluent used in the advanced treatment assays was collected at three different dates (between September and October 2019) from a full-scale UWWTP located in northern Portugal. In this UWWTP, the water line treatment includes a preliminary step (trash racking and dredging) followed by decantation, biological treatment with activated sludge, and a final decantation stage before discharging the effluent to the river. In this study, freshly collected samples of this UWWTP secondary effluent were divided into two aliquots, one of which was immediately analysed (WW) and another was directed to the O3 + UF treatment unit. Details of the analytical methods employed to characterise the UWWTP secondary effluent (WW) are given in Section 2.4, and its chemical and biological characterisation can be found in Tables 1 and 2. Samples collected after O3 + UF treatment (TWW0) were also immediately processed for microbiological analyses and DNA extraction. In addition, aliquots of TWW0 were stored for seven days in sterile glass bottles under dark conditions and at room temperature (herein named as TWW7) to assess possible bacterial regrowth in a hypothetical storage scenario for wastewater reuse.

**Table 1.** Characterisation of the urban wastewater treatment plant (UWWTP) secondary effluent, before (WW) and immediately after treatment (TWW0), and standards of water for irrigation (Decree-Law 236/98) and wastewater reuse in irrigation without restriction, for urban wastewaters which treatment includes a disinfection step (Decree-Law 119/2019) and for wastewater reuse in the Eastern Mediterranean Region—WHO, 2016.


DL stands for detection limit; MVR stands for maximum value recommended; n.a stands for not applicable/available; SAR stands for sodium adsorption ratio; PV stands for parametric value; TDS stands for total dissolved solids; TSS stands for total suspended solids. <sup>a</sup> Analysed by an external laboratory—the maximum value allowed (MVA) for this parameter in the Decree-Law 236/98 is 1. <sup>b</sup> Value up to which there is no restriction to use in irrigation. <sup>c</sup> Permitted limit for greywater reuse in irrigation of vegetables likely to be eaten uncooked.


**Table 2.** Additional analyses made to the UWWTP secondary effluent, before (WW) and immediately after treatment (TWW0).

DL stands for detection limit; MVA stands for maximum value allowed; MVR stands for maximum recommended value; PV stands for parametric value; n.a stands for not applicable/available.

#### *2.3. Experimental Setup and Procedure*

A scheme of the experimental apparatus is depicted in Figure 1. Ozonation was performed in a packed-bed column (2.2 I.D × 70 cm height) with a useful volume of approximately 0.35 L and containing glass Raschig rings (6 mm I.D × 6 mm height), because the water–ozone mass transfer achieved in the column packed with these Raschig rings was up to 3 times higher than that in a bubble column [31]. Firstly, the reactor was filled with ultrapure water (through a peristaltic pump) to regulate the desired concentration of ozone in the liquid phase. Ozone was produced from pure oxygen in a BMT 802X ozone generator and bubbled at the bottom of the column. The ozone concentration in the gas inlet was regulated by adjusting the oxygen gas flow rate with a mass flow controller and the electric intensity of the ozone generator (BMT 802X). The concentration of ozone in the liquid phase (dissolved ozone) was measured with an ATI model Q45H dissolved ozone analyser placed at the exit of the column. High ozone doses and contact time increase the capital and operating costs, therefore a low ozone dose (0.9 ± 0.1 gO3/gDOC) and a short hydraulic retention time (HRT: 8 min obtained with a liquid flow rate of 46 mL min<sup>−</sup>1) were investigated. These experimental conditions were selected in preliminary tests and fixed for all the subsequent experiments.

After a period, ultrapure water in the inlet liquid stream was replaced by the UWWTP effluent to start the ozonation experiments. Samples of ozonised wastewater were only collected after a period of two residence times (~16 min), in order to ensure that the steady state was achieved (i.e., when the outlet wastewater achieved a constant concentration of pollutants in two subsequent measurements). Then, the ozonised effluent was directed to the feed tank of the UF pilot reactor, aiming for the physical removal of microbial cells. Fifteen litres of ozonised effluent was pumped to the UF pilot through a peristaltic pump (Varmec®) and filtered through the 10 nm α-Al2O3 membrane operating in cross-flow mode (1 bar of transmembrane pressure). The UF pilot was designed in a way that the liquid flow of ozonised wastewater was automatically regulated to maintain the pressure constant inside the membrane housing compartment. The concentrate was recirculated to the feed tank [32], while a composite sample of the permeate was collected and split for microbiological and chemical analysis (TWW0 immediately after O3 + UF treatment and TWW7 after being stored for seven days). UF was performed after O3 and not the other way around, because by doing so, the membrane fouling is minimised [33,34]. At the end of the treatment, the membrane was left with H2O2 (30% *w*/*v*) overnight, followed by abundant washing with boiling water and autoclaved before starting another experiment. This cleaning procedure was defined to restore the membrane permeance and sterility.

**Figure 1.** Scheme of the experimental apparatus. (**a**) feed tank containing deionised water or UWWTP effluent; (**b**) peristaltic pump; (**c**) ozone generator (c.1—O2 entrance; c.2—O3 exit); (**d**) mass flow controller; (**e**) ozone diffuser; (**f**) packed-bed column; (**g**) Raschig rings; (**h**) ozone analyser; (**i**) ozone destroyer; (**j**) feed to the ultrafiltration (UF) pilot; (**k**) UF pilot system; (**l**) membrane housing; (**m**) 19 channel ceramic membrane (top view); (**n**) permeate stream; (**o**) concentrate stream.

#### *2.4. Chemical Analyses*

The anionic and cationic contents (Cl<sup>−</sup>, NO3 <sup>−</sup>, SO4 <sup>2</sup>−, Na+, K+) in water samples were determined by ion chromatography, as, using a Metrohm 881 Compact IC Pro apparatus equipped with a Metrosep C4 Cationic Exchange Column (250 mm × 4.0 mm) for the quantification of cations and a Metrosep A Supp 7 Anionic Exchange Column (250 mm × 4.0 mm) for quantification of anions. The content of metals was determined by using an inductively coupled plasma-optical emission spectrometer (ICP-OES, thermo scientific, model iCAP 7000 Series). The pH and conductivity of water were measured with pHenomenal® pH 1100L apparatus (VWR, Germany) and a conductivity meter (Crison GLP 31), respectively. Other relevant parameters (referred to as "additional analyses" in Table 2) were considered to assess the quality of water for irrigation: dissolved organic carbon (DOC) determined in a TOC-L analyser (Shimadzu TOC-5000A); turbidity measured with a turbidimeter (Hanna instruments, model HI88703); chemical oxygen demand (COD) determined by the closed reflux method (EPA standard method 5220D); and biochemical oxygen demand measured according to the EPA standard method 5210B (respirometric method) for a 5 day period (BOD5). These analyses were performed as recommended in the standard methods for the examination of water and wastewater [35].

Moreover, the concentration of target organic micropollutants was determined using ultra-high performance liquid chromatography with tandem mass spectrometry (UHPLC-MS/MS) with Shimadzu Corporation apparatus (Tokyo, Japan) consisting of a triple quadrupole mass spectrometer detector (Ultra-Fast Mass Spectrometry series LCMS-8040) with an ESI (Electrospray Ionisation) source operating in both positive and negative ionisation modes. The mobile phase and operating conditions of the UHPLC-MS/MS system for the detection and quantification of the target pollutants are described elsewhere [16,36]. Prior to UHPLC-MS/MS analysis, WW and TWW0 samples were pre-concentrated and cleaned up by solid-phase extraction (SPE) using Oasis® HLB (Hydrophilic-Lipophilic-Balanced sorbent, 150 mg, 6 mL) cartridges (Waters, Milford, Massachusetts, USA), according to the methodology described elsewhere [37]. For internal calibration, isotopically labelled internal standards were added to the samples before SPE. The preconcentration procedure was performed in duplicate for all the samples. This methodology allows to determine a total of 14 organic micropollutants.

#### *2.5. Microbial Culture Analyses*

Volumes ranging from 100 mL to 1 mL of WW, TWW0 or TWW7 samples or of serial 10-fold dilutions thereof were filtered in triplicate and placed onto the appropriate culture media of the target microbial group: Plate Count Agar (PCA, VWR International (Pennsylvania, USA)) (30 ◦C, 48 h) for culturable heterothrops, m-Faecal Coliform Agar (mFC, Thermo Fisher Scientific, Massachusetts, USA) (37 ◦C, 24 h) for enterobacteria, Slanetz Bartley Agar (Thermo Fisher Scientific, Massachusetts, USA) (37 ◦C, 48 h) for enterococci, and Rose Bengal Chloramphenicol Agar (Thermo Fisher Scientific, Massachusetts, U.S.A.) (25 ◦C, 5 days) for fungi. Results were expressed as colony forming units per 100 mL of sample (CFU/ 100 mL).

## *2.6. DNA Extraction, 16S rRNA and Inti1 Genes Quantification*

Volumes of 100 mL of WW, 2 L of TWW0, and 800 mL to 1 L of TWW7 were vacuum-filtrated and processed in three independent samplings as biological replicates. DNA extraction was performed using the DNeasy® PowerWater® Kit (QIAGEN, Hilden, Germany) according to Rocha et al., 2020 [38] and with two additional steps suggested in the manufacturer's troubleshooting guide: after adding the lysis solution, a heating step at 65 ◦C for 10 min was included in the protocol; and to ensure the removal of residual ethanol before DNA elution, the centrifugation step was conducted in a clean collection tube for an additional minute. DNA samples were stored at −20 ◦C until quantitative PCR (qPCR) analysis.

The 16S rRNA gene (a marker for total bacteria) and the *intI*1 gene encoding a class 1 integronintegrase (a marker of anthropogenic impact) were quantified based on qPCR to assess the removal efficiency of bacteria after treatment [39,40]. Gene-specific primer sequences are listed in previous studies [41,42] and provided as supplementary information in Table S1. Gene quantification was based on SYBR Green qPCR assays in a StepOnePlus™ Real-Time PCR System (Life Technologies, USA) and interpolation to the standard curve run in each assay, as described elsewhere [39,43].

The data that met the quality criteria described in Rocha et al., 2018 [44] were expressed as the ratio of gene copy number per 100 mL of water sample (WW, TWW0, and TWW7). The secondary wastewater effluent (WW) was used as reference to assess the removal efficiency of both 16S rRNA and *intI*1 genes in treated samples, immediately after treatment (TWW0) and after storage for 7 days (TWW7). The duration of 7 days was selected to allow enough time for eventual injured cells surviving the treatment to fully recover, as we have verified in previous works with other treatment solutions [14,15].

## *2.7. Biological E*ff*ect Assays*

#### 2.7.1. Cell Culture and Incubation with Water Samples

Murine fibroblasts L929 and Caco-2 cells were cultured in Dulbecco's modified Eagle medium (DMEM) with d-glucose (4.5 g L−1), sodium pyruvate (0.11 g L−1), l-alanyl-l-glutamine (0.86 g L−1) and further supplemented with 10% (*v*/*v*) heat inactivated foetal bovine serum (FBS), and 5% (*v*/*v*) of PenStrep (37 ◦C, 5% CO2 and 95% of humidity). For cell viability and cytotoxicity assessment, the cells were detached from the culture flask as described elsewhere [45]. After cell counting in Neubauer chamber (Boeco, Germany), the suspension was centrifuged at 300 *g* for 5 min, and the cell pellet was suspended in culture medium to a final concentration of 5 <sup>×</sup> <sup>10</sup><sup>4</sup> cells per well. Cells were then seeded in a 96-well microplate (100 μL per well) and cultured for 24 h at 37 ◦C (5% CO2 and 95% humidity).

#### 2.7.2. Thiazolyl Blue Tetrazolium Reduction (MTT) and Lactate Dehydrogenase (LDH) Assays

Cellular metabolic activity was evaluated as indicator of cytotoxicity by the thiazolyl blue tetrazolium reduction (MTT) assay, whereas cell membrane integrity was evaluated through the lactate dehydrogenase (LDH) assay, providing information about cell viability. Briefly, test water samples were filtered using Corning® syringe filters (Sigma-Aldrich®, St. Louis, MO, USA) with 0.20 μm pore diameter and diluted 1:10 and 1:5 in DMEM. After discarding culture supernatant, 100 μL of diluted samples were added to cell layers and incubated at 37 ◦C (5% CO2 and 95% humidity). After 24 h, the supernatant was removed for LDH assay, while the remaining content of the wells was used for MTT assay. For MTT assay, absence of cytotoxicity (100%) was estimated by replacing water test sample by culture medium. For LDH assay, the absence of cell viability (100%) was estimated by replacing water test sample by 1% (*v*/*v*) Triton X-100 solution prepared in culture medium.

#### 2.7.3. Yeast Estrogen Screen (YES) Assay for Estrogenic Activity Assessment

WW and TWW0 samples were filtered through 0.21 μm hydrophilic membranes and analysed directly, without any preconcentration. The YES assay and data analysis were performed according to the kit manufacturer's instructions. Calibration was established using standard solutions of the natural estrogen 17-β-estradiol (E2), at concentrations between 10−6–10−<sup>9</sup> mol L−1. E2 also worked as positive control while ultrapure water was used as negative control. E2 standard solutions were prepared in DMSO (<1% in the assay medium), therefore a solvent blank was also assayed. Samples, standards, and control solutions were transferred to a 96-well microplate, mixed with assay medium, and inoculated with the transformed yeast cells. The mixture was then incubated for 48 h at 31 ◦C under orbital shaking. Spectrophotometric measurements at 570 nm (β-galactosidase expression) and 690 nm (yeast growth) were carried out using a Cytation3® microplate reader (Bio-Tek Instruments, Winooski, USA). The potential estrogen agonistic activity was estimated through the calculation of the parameters growth factor (G) and induction ratio (IR). The G parameter was calculated as the ratio of absorbance values measured at 690 nm for the sample and for the solvent (A690)sample/ (A690)solvent. The IR parameter was calculated as (1/G) × ((A570 − A690)sample/(A570 − A690)solvent).

#### **3. Results and Discussion**

#### *3.1. Micropollutant Removal, Mineralisation, and Other Physico-Chemical Parameters*

Under the regulation on minimum requirements for water reuse in agricultural irrigation, the environmental quality standards for priority substances and certain other pollutants should be targeted [7,21]. Moreover, the same regulation refers to additional requirements for risk assessment, including micropollutants. From the chemical organic micropollutants analysed in fresh (WW) and O3+UF treated water samples (TWW0), only 9 out of 14 were detected. The antiplatelet clopidogrel, the herbicide isoproturon, the anti-inflammatory diclofenac, the industrial compound PFOS (perfluorooctanesulfonic acid), and the lipid regulator bezafibrate were detected with a frequency of 100% in WW samples during the sampling campaign (Figure 2). Alachlor was also detected in all WW samples but below the limit of quantification (LOQalachlor < 25 ng L<sup>−</sup>1), whereas warfarin, citalopram, and clofibric acid were detected only in some samples. According to the Directive 2013/39/EU and Decision 495/2015/EU [21,22], alachlor, isoproturon and PFOS are considered PSs, whereas the others are considered CECs. After treatment, most micropollutants presented values below LOD—Limit Of Detection. Only alachlor, clopidogrel, citalopram and isoproturon were detected: the first two were below the LOQ—Limit Of Quantification (25 and 5 ng L<sup>−</sup>1, respectively), whereas the latter two were found at concentrations up to 529 and 10.6 ng L−1, respectively. In fact, isoproturon was the micropollutant with the lowest removal percentage (i.e., 80% of maximum removal). All priority substances (alachlor, isoproturon and PFOS) were below their environmental quality standards defined in the EU Directive 2013/39 [21], complying with the requirements of the EU Regulation 2020/741 [7].

**Figure 2.** Logarithmic range of concentrations (ng L<sup>−</sup>1) of the detected micropollutants in WW (black bar) and TWW0 (striped bar) for samples with concentrations above LOQ. The frequency of occurrence was 100% (3/3) for all compounds, except when indicated in brackets after the compound name. \* <LOQ and \*\* <LOD (compounds with concentrations < LOD before treatment are not shown in this figure for the sake of simplicity).

DOC and pH values did not remarkably vary after treatment (Tables 1 and 2, respectively). Values for DOC are not regulated and both pHs (before and after treatment) comply with the maximum value allowed (MVA). Thus, considering that regulations of water quality for irrigation often do not inform about adequate levels of organic matter, it can be assumed that the achieved values of DOC and micropollutants in treated water do not invalidate its use for irrigation. Moreover, the available literature mentioning the monitoring of DOC in water for irrigation recommends the evaluation of DOC when COD and BOD5 are at the so-called alarming levels (>60 mgO2 L−<sup>1</sup> and >10 mgO2 L−1, respectively) [28,30,46], which is not the case of TWW0 (Table 2).

In the combined process, ozonation was expected to be mainly responsible for the removal of micropollutants and dissolved organic matter rather than UF [18]. These results are coherent with other studies performing solely ozonation, in which the authors attributed the low yield of mineralisation to the formation of recalcitrant organic intermediates deriving from the organic micropollutants or, more likely, from the oxidation of dissolved organic matter naturally present in the wastewater [14,16]. For instance, using a similar experimental apparatus for the continuous ozonation of a secondary-UWWTP effluent (without UF), Moreira et al., 2016 [14] reported a DOC removal of ~30% (retention time of 26 min), whereas Iakovides et al., 2019 [16] obtained a DOC removal of ~10% (with similar ozone dosage and retention time).

Regarding other physico-chemical parameters, TWW0 presents values below the maximum recommended in the Portuguese Laws of (i) water for irrigation [31] and (ii) treated wastewater for reuse [28,30]. The only exception is for the concentration of chloride in Table 1 (ca. 80 mg L−<sup>1</sup> before and after treatment) which is slightly higher than the maximum value recommended (MVR) of 70 mg L−<sup>1</sup> in the oldest law [27], which is not included in the newest one [28]. It is worth mentioning that this maximum value recommended for chloride was stipulated considering the sensitivity of tobacco crops; therefore, TWW0 might not be appropriate for irrigation of this specific crop, but not necessarily inappropriate in the case of crops tolerant to these concentrations of chloride. For instance, some crops of fruits and vegetables are highly tolerant to chloride, such as Rangpur lime and cauliflower, for which

the water for irrigation can contain up to 600 and 710 mg L−<sup>1</sup> of chloride, respectively [47]. In fact, TWW0 can be applied for irrigation according to the WHO (World Health Organization) and FAO (Food and Agriculture Organization of the United Nations) guidelines of water quality for surface irrigation, where the allowed chloride concentration is up to 142 mg L−<sup>1</sup> (Table 1) [29], i.e., well above the value determined for the wastewater in this study (ca. 80 mg L<sup>−</sup>1). The value of salinity (782 μS/cm) is slightly higher than that recommended by FAO and WHO [28] for the use of water for irrigation with no restriction (<700 μS/cm), but this value is not defined in Portuguese guidelines. Another interesting observation is the increase in the nitrate concentration after treatment, although still below the maximum value recommended [27], which can be attributed to the oxidation of nitrogen-containing substances that are likely to be present in the secondary effluent of UWWTPs [48]. Sulphate and copper contents also suffer a slight increase after treatment, which can be due to their release from sediments/soil particles after ozonation [49,50].

Future work must consider the energy demand of these processes [51] and life cycle assessment (LCA) [52–54] for the elimination of micropollutants from urban wastewater—these studies being particularly scarce with data at full scale. For instance, it has been concluded that ozonation has a lower energy demand compared to the use of membranes or UV/H2O2 [9]. Conversely, the electrical energy demand of ozonation is higher than those determined for powdered activated carbon (PAC) addition or granular activated carbon (GAC) filtration, but always being a plant-specific issue [51]. Performing LCA, it was suggested that ozonation has a better overall environmental performance than the photo-Fenton process [53], whereas reverse osmosis causes higher environmental burdens than ozonation due to the high energy and material consumption [52]. In these processes, generated impacts result mainly from the production of energy needed (and the respective energy mix) and from the use of some specific reagents [54].

#### *3.2. Microbial Inactivation and Regrowth*

As expected, a reduction in the load of the microbiological groups analysed was observed immediately after treatment (Figure 3). Reductions of nearly 3.5 log-units of 16S rRNA gene (indicative of the abundance of total bacteria) and 3.7 log-units of culturable heterotrophs occurred. The abundance of *intI*1 followed a similar trend, with a reduction of ~4.6 log-units immediately after treatment, whereas enterobacteria, enterococci, and fungi, with reductions higher than five log-units, reached values below the detection limit (0.33 CFU per 100 mL). Microbial inactivation can be transient [14,55,56], therefore further assays testing the regrowth capacity after seven days of storage of the treated wastewater were performed (TWW7 samples). It is known that bacterial reactivation is influenced by factors such as storage conditions, in particular temperature, availability of nutrients, ultraviolet light, and assimilable organic carbon content, among others [57,58]. Therefore, the conditions to perform this assessment were selected to mimic the most common real storage conditions, i.e., room temperature (25 ± 2 ◦C) and absence of light to minimise DNA repair mechanisms [59]. The abundance of the 16S rRNA and *intI*1 genes, as well as the heterotrophic counts, recovered to values close to those observed in WW samples. The same pattern was observed for fungi, although with a lower regrowth extent (~1.6 log-units). The transient effect of single ozone-based processes for the treatment of UWWTP effluents was reported before [14–16]. In fact, even when operating with close ozone doses (0.75 gO3/gDOC) and higher HRT (10–60 min) to those used here (0.9 gO3/gDOC, HRT 8 min), reactivation of all the microbial groups analysed in the current study has been described in the literature [14,15]. In contrast, in the present study, regrowth of faecal indicators (enterobacteria and enterococci) was not observed in TWW7 samples. Notwithstanding, from a microbiological quality point of view, both TWW0 and TWW7 comply with the biological parameters included in the quality standards of water for crops irrigation, both in Portugal [27] and United States [60], or the Portuguese/European Union quality standards of wastewater reuse in irrigation without restriction [28,61]. In fact, faecal coliforms or *Escherichia coli* (enterobacteria) and nematode eggs are the only biological parameters included in these quality standards, for which values were found below the stipulated thresholds (Table 1).

**Figure 3.** Microbiological water quality. (**a**) Culturable heterotrophs, enterobacteria, enterococci, and fungi, expressed as log (CFU/100 mL of sample); and (**b**) qPCR-based quantification of 16S rRNA and intI1 genes, expressed as log (gene copy number/ 100 mL of sample). \* below the detection limit (0.33 CFU/100 mL).

Based on the abundance of enterobacteria in wastewater immediately after ozonation (102–103 CFU 100 mL <sup>−</sup>1) or after 3 day storage (103–104 CFU 100 mL<sup>−</sup>1) reported by Moreira et al. (2016) [14] and Iakovides et al., 2019 [16], the utilisation of ozonation alone would not produce wastewater compatible with its further use in irrigation. In contrast, the combination of UF with O3 utilised here improved the efficiency of the treatment. The membrane fouling observed during the filtration process, which was evidenced by the permeate flow decrease from ~60 mL min−<sup>1</sup> to ~16 mL min−1, was most likely derived from bacteria that survived ozonation, cell debris and undissolved (in)organic matter. Nevertheless, the total suspended solid (TSS) value after O3 was unquantifiable. In spite of the considerable improvements demonstrated in this study, the post-storage increase in total heterotrophs and genes shows that there is still room for additional tuning of the process to prevent the possible contamination of the permeate tank with spores of heterotrophic bacteria or fungi.

#### *3.3. Evaluation of Biological E*ff*ects*

Cytotoxic and cell viability effects of wastewater collected before (WW) and after treatment with O3 + UF (TWW0) were evaluated for skin (L929) and digestive epithelium (Caco-2) cell models by performing complementary MTT and LDH assays (Table 3). Considering that cell viability upon exposure to water samples depends on the final composition of the growth medium [62], test samples were diluted 5 and 10 times in culture medium before incubation with cell layers. Similar cytotoxicity (MTT) and cell viability (LDH) values were obtained for both dilution levels (Table 3). Moreover, cell viability was equivalent to that obtained for cell incubation with a plain culture medium. For both cell

lines, no difference in cytotoxicity was observed for water samples collected before and after treatment (Table 3, MTT assay). Cell viability was also maintained after treatment (Table 3, LDH assay), providing similar or even higher values than those obtained for plain culture media or tap water. Additionally, samples analysed right after ozonation (i.e., before UF) rendered percentages of 91 ± 6% and 23 ± 8% in the MTT and LDH assays for L929 cells, respectively, indicating that no cytotoxic compounds were produced during this step.

**Table 3.** Results (percentage) from MTT <sup>a</sup> and LDH <sup>b</sup> assays obtained for urban wastewater before (WW) and after treatment (TWW0).


<sup>a</sup> Values for culture media were 100% (Relative Standard Deviation—RSD < 20%) and between 1 and 9% for Triton X-100 (total disruption of cells). Samples were diluted 5 times in culture media before incubation with cells. <sup>b</sup> Values for Triton X-100 (total disruption of cells) were 100% (RSD < 10%). Values for tap water were 111 ± 5 for L929 cells and 110 ± 12 for Caco-2 cells. <sup>c</sup> Values between brackets correspond to blank values obtained in culture media only (intact cells). Values for tap water were 26.3 ± 5.6 for L929 cells and 55.9 ± 4.7 for Caco-2 cells.

The presence of estrogenic activity was also evaluated using the YES assay for WW and TWW0 samples. Yeast growth inhibition was not observed for any of the tested samples. Induction ratios (IR) were 1.02 ± 0.09 for WW, and 0.74 ± 0.02 for TWW0. These values were below the kit threshold value IR10 (corresponding to 10% of the maximum IR, value of 2.82, obtained for E2 standards), which indicated no estrogenic activity.

Work on toxicity assessment of effluents treated by ozonation has provided contradictory evidence. The biological toxicity of the influent of sewage treatment plants was significantly decreased after applying different advanced treatment processes, including ozone combined with UV, using *Daphnia magna*, zebrafish (*Danio rerio*), and *Vibrio fischeri* [63] as target organisms. However, when ozone and hydrogen peroxide were used together, a slight acute toxicity was perceived for *V. fischeri* while acute toxicity was observed for *D. magna* [64]. Other work, also applying the algae *Desmodesmus quadricauda*, indicated that the toxicity class of treated wastewater may change from completely non-toxic to very high hazard category, with a clear relationship between the time of ozonation and the increase in ecotoxicity [65]. This compound-dependent behaviour was also observed in a study with zebrafish embryos where different pharmaceutical compounds were tested [66]. Therefore, our results with cell lines are in agreement with previous works, where no toxic effect was observed after treatment, particularly when low doses of ozone are applied.

#### **4. Conclusions**

The results of this study indicate that UF performed after ozonation can be a suitable approach to allow the safe reuse of urban wastewater for irrigation. The combined process resulted in an effective treatment, especially against micropollutants detected in the UWWTP secondary effluent, and in the reduction in the microbial load. Treated wastewater stored for seven days maintained the quality required for irrigation, with the physico-chemical parameters, and enterobacteria and nematode egg counts below the maximum values recommended in water quality standards. In addition, no harmful biological effects were detected concerning the viability and estrogenicity tests. However, the fact that total bacterial cells, total cultivable heterotrophs as well as the *intI*1 gene reactivated to values close to those observed for untreated wastewater, shows that there is still room for additional improvement of this process.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4441/12/12/3458/s1, Table S1: Quantitative PCR conditions used in the present study for absolute gene quantification in all WW, TWW0, and TWW7 samples.

**Author Contributions:** Conceptualisation, A.M.T.S., O.C.N., C.M.M., M.A.S., S.M.C.-S.; methodology, A.M.T.S., O.C.N., C.M.M., M.A.S., S.M.C.-S.; investigation, C.A.L.G., S.R.-S., J.A.-S., I.I.R., A.R.R.; writing—original draft preparation, C.A.L.G., S.R.-S., J.A.-S., I.I.R.; writing—review and editing, A.M.T.S., O.C.N., C.M.M., M.A.S., A.R.R.; supervision, A.M.T.S., O.C.N., C.M.M., M.A.S.; project administration, A.M.T.S., O.C.N., C.M.M., M.A.S., S.M.C.-S.; funding acquisition, A.M.T.S., O.C.N., C.M.M., M.A.S., S.M.C.-S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work is a result of the project "DEPCAT—Demonstration of new EquiPment involving integrated CATalytic processes for treatment of organic pollutants and disinfection of water", with the reference NORTE-01-0247-FEDER-033330, co-funded by European Regional Development Fund (ERDF), through the North Portugal Regional Operational Programme (NORTE2020), under the PORTUGAL 2020 Partnership Agreement.

**Acknowledgments:** We would like to thank the scientific collaboration under Base Funding-UIDB/50020/2020 of the Associate Laboratory LSRE-LCM and Base Funding-UIDB/00511/2020 of the Laboratory for Process Engineering, Environment, Biotechnology and Energy—LEPABE, both funded by national funds through the FCT/MCTES (PIDDAC), and FCT project UID/Multi/50016/2013 (Associate Laboratory CBQF) and UIDB/50006/2020 (LAQV, REQUIMTE).

**Conflicts of Interest:** The authors declare no conflict of interest.

## **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## **A MATLAB-Based Application for Modeling and Simulation of Solar Slurry Photocatalytic Reactors for Environmental Applications**

## **Raúl Acosta-Herazo 1,\*, Briyith Cañaveral-Velásquez 2, Katrin Pérez-Giraldo 2, Miguel A. Mueses 2, María H. Pinzón-Cárdenas <sup>1</sup> and Fiderman Machuca-Martínez <sup>1</sup>**


Received: 2 July 2020; Accepted: 27 July 2020; Published: 4 August 2020

**Abstract:** Because of the complexity caused by photochemical reactions and radiation transport, accomplishing photoreactor modeling usually poses a barrier for young researchers or research works that focus on experimental developments, although it may be a crucial tool for reducing experimental efforts and carrying out a more comprehensive analysis of the results. This work presents PHOTOREAC, an open-access application developed in the graphical user interface of Matlab, which allows a user-friendly evaluation of the solar photoreactors operation. The app includes several solar photoreactor configurations and kinetics models as well as two variants of a radiation absorption-scattering model. Moreover, PHOTOREAC incorporates a database of 26 of experimental solar photodegradation datasets with a variety of operational conditions (model pollutants, photocatalyst concentrations, initial pollutant concentrations); additionally, users can introduce their new experimental data. The implementation of PHOTOREAC is presented using three example cases of solar photoreactor operation in which the impact of the operational parameters is explored, kinetic constants are estimated according to experimental data, and comparisons are made between the available models. Finally, the impact of the application on young researchers' projects in photocatalysis at the University of Cartagena was investigated. PHOTOREAC is available upon request from Professor Miguel Mueses.

**Keywords:** computer-based learning; solar photocatalysis; water contaminants; kinetic modeling; photoreactor design

## **1. Introduction**

Heterogeneous photocatalysis is an example of an emerging environmental technology with a variety of promising applications, such as air and water disinfection and decontamination, clean fuel production and green product manufacturing [1–3].

Modeling and computer simulation of photoreactors are crucial for their design, scale-up and technology transfer; since they allow engineers and researchers to understand the role of the design parameters and operational conditions without performing an excessive number of experiments. However, modeling a solar photoreactor is a very complicated task, because it requires a combination of knowledge in applied solar energy, geometric optics, radiative transfer, materials science and photochemical reaction engineering.

The implementation of commercial packages for photoreactor simulations is limited. Simulation packages for chemical plants, such as Aspen HYSYS® (Aspen Technology, Inc., Bedford, MA, USA) or Aspen plus® (Aspen Technology, Inc., Bedford, MA, USA), do not incorporate photocatalytic reactors. On the other hand, modeling and simulation of photoreactors can be carried out in Computational Fluid Dynamics (CFD) packages, such as COMSOL Multiphysics® (COMSOL, Inc., Burlington, MA, USA) and ANSYS® Fluent (ANSYS, Inc., Southpointe, PA, USA). However, they do not have modules dedicated to photoreactor engineering. Therefore, the simulations are performed by adapting the existing simulation modules for the simulation of photocatalytic reactors. This configuration of the CFD modules must be carried out manually by the user, which may result in an approach not intuitive enough for non-experts in photoreactor engineering. Another alternative is to perform the direct coding of the photoreactor model in a programming language. Still, this may result in a challenge for researchers that have not taken advanced courses in programming and numerical methods.

For the above reasons, the direct coding or the use of CFD simulators to implement a photoreactor model could be found inconvenient by non-expert researchers in photoreactors engineering, such as young researchers or those focused on experimental developments. However, implementing a photoreactor model may be a crucial tool for reducing the experimental efforts and carrying out a more comprehensive analysis of the results.

In this work, we present PHOTOREAC, an open-access computational application developed in the graphical user interface of Matlab wholly dedicated to the modeling and simulation of large-scale slurry solar photocatalytic reactors for environmental applications. It is based on the experience gathered by our research groups at Cartagena University (Cartagena de Indias, Colombia) and the Universidad del Valle (Cali, Colombia) during the last twenty years of research in heterogeneous solar photocatalysis, and also on extensive literature research in photoreactor engineering.

The application aims to provide non-expert researchers in photoreactors engineering a user-friendly, dedicated and efficient tool for the modeling and simulation of solar photoreactors, providing them with valuable information without implementing very sophisticated methods.

By employing PHOTOREAC, the users will be able to explore the role of critical parameters of the system on the radiation absorption performance of the photoreactor and the overall kinetic behavior of the photocatalytic process; parameters include the photoreactor geometry, the photoreactor dimensions, the model pollutant, the kinetic expression, the photocatalyst concentration, the photocatalysts optical properties, the initial pollutant concentration, the volume of treated water and the incident radiation. Additionally, PHOTOREAC incorporates a database of experimental information collected in our laboratory regarding the solar photodegradation of a variety of model pollutants under different operational conditions. Therefore, users will have empirical data available to carry out analyses and comparisons with their data.

#### **2. Solar Photoreactors Modeling by PHOTOREAC**

PHOTOREAC performs the modeling and simulation of the photoreactors following the general algorithm described in Figure 1. The algorithm considers mathematical simplifications to maintain the approach as rigorously and computationally efficient as possible, and thus it provides the users with valuable information without implementing sophisticated numerical methods that, although they can improve the quantitative results, may not affect the qualitative analysis. These assumptions and simplifications will be described and discussed in the upcoming sections.

The basis of the PHOTOREAC approach is that the radiation field modeling can be carried out independently of the photocatalytic kinetics modeling since the radiation balance in the photoreactor is not a function of the concentration of the chemical species. Therefore, the radiation balance is decoupled from the mass and momentum balances of the system. Besides, the radiation field described by the local volumetric rate of photon absorption (LVRPA) profile inside the photoreactor is considered to be in a steady-state, i.e., it does not vary along the reaction rime its reaction time does not change [4,5]. On the other hand, to carry out a kinetic analysis independent of the radiation absorption effects,

i.e., the optimized kinetic parameters are not a function of the irradiation conditions, it is mandatory to know the radiation field in the photoreactor beforehand [6,7].

Thus, PHOTOREAC considers two modules: (i) the photon absorption-scattering module, in which the user will be able to determine the radiation field of the available photoreactor configurations by following the procedure described by the red box in Figure 1; and (ii) the kinetic modeling module, in which the user will be able to estimate the radiation-independent kinetic parameters for the four available kinetics expressions following the procedure described in the blue box in Figure 1.

**Figure 1.** General algorithm for the modeling and simulation of solar slurry photoreactors in PHOTOREAC. ODE: Ordinary Differential Equation; SFM: Six Flux Model.

#### *2.1. The Photoreactors Set-Up in PHOTOREAC*

PHOTOREAC includes three configurations of pilot-scale solar photoreactors: a flat plate photoreactor (FPP), a compound parabolic collector photoreactor (CPCP) and a tubular-type photoreactor (TTP). These are the most common configurations for solar-pilot applications of heterogeneous photocatalysis; a detailed description of them can be found in the literature [3,8,9]. For the TTP, a novel prototype is also included, the offset multi-tubular photoreactor (OMTP) [10]. All of the photoreactors operate in recirculation, a flow-through mode with the water passing through an external tank, as shown in Figure 2. The photoreactor is exposed to the sunlight, facing the sun, while the reservoir tank is in the dark. The flow consists of an aqueous suspension of photocatalyst powder and the dissolved contaminant. The Evonik TiO2 P25 was selected as the model photocatalyst in PHOTOREAC because it is considered the most promising alternative for commercial applications due to its low cost, photochemical stability, and high oxidation power [3]. Therefore, it is widely studied, and its physicochemical and optical properties are well known in the literature [11].

**Figure 2.** Scheme of a solar pilot photoreactor set-up.

#### *2.2. The Input Data for the Use of PHOTOREAC*

The availability and reliability of the input data provided to PHOTOREAC are crucial for good results. Table 1 shows a summary of the input information that is required to simulate the photoreactors in PHOTOREAC. Additionally, it is indicated in which module is the information used.

**Table 1.** Summary of the input information for the PHOTOREAC modules.


<sup>a</sup> PASM: photon absorption-scattering module; KMM: kinetic modeling module. <sup>b</sup> If it is a multicomponent mixture, *Ci* is replaced by *TOC*.

At the same time, the experimental photodegradation data for kinetic analysis in PHOTOREAC deserves special attention. The effects of the adsorption must be carefully considered in the solar photocatalytic experimental test. The photodegradation data used to feed PHOTOREAC must be reported at the zero-point of photodegradation, where adsorption has already been allowed to homogenize, which is usually achieved by allowing the system to recirculate under darkness for 30 min to establish adsorption–desorption equilibrium conditions before being exposed to solar light. Thus, although the kinetic models in PHOTOREAC do not contemplate the competitive effects of molecular adsorption, the data used will already be corrected with that effect. Therefore, there is no problem with the application of the models [10,12].

During the exposure time to sunlight, the data should be reported as the pollutant concentration *Ci* against the accumulated ultraviolet (UV) energy ξ*AE*. The experiments finish when the desired accumulated UV total energy in J/m2 is reached. Additionally, it is required to record the corresponding standard time for each sample.

### *2.3. The PHOTOREAC Photon Absorption-Scattering Module*

The PHOTOREAC photon absorption-scattering module performs the radiation field modeling of the three available configurations of solar photoreactors: FPP, CPCP and OMTP. It provides the LVRPA spatial distribution inside the photoreactor and the overall volumetric rate of photon absorption (OVRPA), which corresponds to the LVRPA averaged over the entire volume of the reactor. The latter is a critical magnitude for the kinetic assessment [12].

The PHOTOREAC modeling approach is focused on the six-flux absorption-scattering model (SFM). SFM is an analytical equation in which the leading hypothesis is that scattering only occurs in the six Cartesian directions [5]. Despite being a simplified model, it retains the key aspects of the radiation field modeling in photoreactors and has been implemented successfully at the solar pilot scale [13,14]. Other modeling approaches for solar photoreactors, such as the discrete ordinate method (DOM) or the Monte Carlo simulation, offer a more accurate description of the radiation transport phenomena. However, they are more time-consuming in the computations and their mathematical formulation is of high complexity. The SFM short computation times are ideal for exploring the impact of operational parameters, including the photocatalyst concentration, photoreactor dimensions and incident radiation, in particular for users that are dabbling in photoreactor engineering, to which PHOTOREAC is oriented. Independently of the photoreactor configuration, the central equation of SFM is given by [14]:

$$LVRPA = \frac{I\_0}{\lambda\_{\omega\_{\rm corr}} \omega\_{\rm corr} (1 - \gamma)} \left[ \left( \omega\_{\rm corr} - 1 + \sqrt{1 - \omega\_{\rm corr}^2} \right) e^{-\frac{\gamma p}{\lambda\_{\rm corr}}} + \gamma \left( \omega\_{\rm corr} - 1 - \sqrt{1 - \omega\_{\rm corr}^2} \right) e^{\frac{\gamma p}{\lambda\_{\rm corr}}} \right] \tag{1}$$

where *I*<sup>0</sup> is the incident solar radiation in W/m2 and *rp* is a spatial coordinate in the reactor domain whose definition depends on the reactor geometry. Finally, the corrected photon path length λω*corr* in m, the dimensionless corrected scattering albedo ω*corr* and the dimensionless parameter γ are all parameters derived from the SFM formulation. PHOTOREAC also includes a more recent variant of the SFM, the Six Flux Model coupled to the Henyey-Greenstein scattering phase function (SFM-HG). In it, the Henyey–Greenstein (HG) scattering phase function is used to describe the optical properties of the TiO2 P25 photocatalyst. By contrast, the SFM describes TiO2 based on a diffuse reflectance scattering phase function [15]. By incorporating both variants of SFM, the users will be able to observe the role of the scattering phase function. The parameters and implementation of Equation (1) are detailed in the literature, and the modeling details for the FFP are given in previous work [16].

On the other hand, for the CPCP and the OMTP, a ray-tracing technique together with Equation (1) must be implemented, since, besides the incident radiation, the direction with which solar rays impact the photoreactor is crucial. A complete description of the SFM implementation for CPCP and OMTP is reported elsewhere [10,13,14].

#### *2.4. The PHOTOREAC Kinetic Modeling Module*

The PHOTOREAC kinetic modeling module estimates the kinetic parameters from the photodegradation experimental data provided. Table 2 shows the photocatalytic kinetic models in PHOTOREAC. These models explicitly consider the effect of the radiation absorption on the average reaction rate in −*riVR* by including the *Eg*, and the overall rate of photon absorption (OVRPA) in W/m3, which corresponds to the LVRPA averaged over the entire volume of the reactor. Additionally, *Ci* is the concentration of the water contaminant in mol/m3, κ*<sup>P</sup>* = *2*/*Sg Ccat* is the particle constant in m3/m2, *Sg* is the catalyst specific surface area m2/kg, *CCat* is the photocatalyst concentration kg/m3, *CO*<sup>2</sup> is the oxygen concentration in mol/m<sup>3</sup> and <sup>φ</sup>*eff <sup>g</sup>* is the effective quantum yield in mol/(s watts). Finally, *kL*−*H*, *Kkin*, α<sup>1</sup> and α<sup>2</sup> are the kinetic constants of the models, which are independent of the irradiation conditions.


**Table 2.** Photocatalytic kinetic models in PHOTOREAC.

Each of these previous expressions has its features and limitations, from either a phenomenological or a numerical point of view. For instance, the Langmuir–Hinshelwood expression is a semi-empirical model. By contrast, the other models were deduced from a detailed reaction mechanism. Zalazar et al. and Mueses et al.'s kinetic expressions consider the effect of the effective quantum yield φ*eff <sup>g</sup>* explicitly, a critical parameter to evaluate photocatalytic reactions. However, the expression proposed by Mueses et al. is the only one with three fitting parameters, unless the effective quantum yield of the system is previously known [12].

To determine the kinetics parameters, it is necessary to follow a rigorous approach to account for the effects of the diffusion and convection in the material balance of the photoreactor. Although the inclusion of these effects will provide more accurate results for the kinetic parameters (such parameters will be independent of the diffusion and convection), it also implies the implementation of more advanced numerical techniques, e.g., finite differences and orthogonal collocation [7,19]. PHOTOREAC considers the photoreactor-tank system as a batch mode reactor; therefore, the effects of the diffusion and convection are lumped in the kinetic parameters, which simplifies the numerical approach.

The following assumptions are established for the mass balance of the system (represented by Figure 2): (i) the system is perfectly mixed; (ii) there are no mass transport limitations; (iii) the conversion per pass in the reactor is differential; and (iv) parallel dark reactions can be neglected. The mass balance in the reservoir tank can then be expressed as follows [7,18]:

$$\frac{dC\_i}{dt} = \frac{V\_R}{V\_T} \langle -r\_i \rangle\_{V\_R} \tag{2}$$

where *Ci* is the concentration of the water contaminant in mol/m3 at time *t*, *t* is time in s, −*riVR* is the average reaction rate in (mol m<sup>3</sup> s<sup>−</sup>1), and *VR* and *VT* are the volumes of the photoreactor and the total reaction volume in m3, respectively. However, for solar photoreactors, the standard time may not be the more appropriate magnitude for following the concentration of the water pollutant due to the fluctuation of the incident solar irradiance because of the atmospheric phenomena and the time of day. Therefore, a change of variable is proposed as follows [10]:

$$\frac{d\mathbf{C}\_i}{dt} = \left(\frac{d\mathbf{C}\_i}{d\xi\_{AE}}\right) \left(\frac{d\xi\_{AE}}{dt}\right) \tag{3}$$

$$\frac{d\mathbb{C}\_i}{d\xi\_{AE}} = \frac{\beta}{\xi\_t} (-r\_i)\_{V\_R} \tag{4}$$

With the initial condition, *Ci* (ξ*AE* = 0) = *Ci,*0, where *Ci* is the water contaminant concentration for a given ξ*AE* is the accumulated energy in J/m2, ξ*<sup>t</sup>* = *d*ξ*AE dt* in J/m2s is the slope of the straight line resulting from the experimental data relationship of the accumulative incident solar radiation vs. time for each experimental test, and the dimensionless factor β = *VR*/*VT*.

The search for the best values for the kinetic parameters of the model is carried out using a non-linear regression procedure, as is shown in Figure 1. It starts with an initial guess and follows an optimization criterion until the required convergence is reached. The error function is given by the sum of the squared errors of the experimental water contaminant concentration *Ci,*exp and the value determined from the numerical solution of Equation (4) *Ci,calc*:

$$F\_{obj} = \sum\_{i=1}^{N} \left(\mathbb{C}\_{i, \text{exp}} - \mathbb{C}\_{i, \text{calc}}\right)^2 \tag{5}$$

where *N* is the number of experimental data. The Matlab function fminsearch, which uses the Nelder–Mead algorithm, is implemented as the optimization solver together with the Matlab function ode 45 for solving the ordinary differential equation (ODE) given by Equation (4).

For the photodegradation of multicomponent mixtures, the concentration *Ci* may be replaced by a global concentration parameter such as total organic carbon (*TOC*) [12]. Therefore, Equation (4) is written as:

$$\frac{dTOC}{d\xi\_{AE}} = \frac{\beta}{\xi\_t} (-r\_{TOC})\_{V\_R} \tag{6}$$

with the initial condition *TOC* (ξ*AE* = 0) = *TOC*0, where *TOC* is the total organic carbon of the mixture mol/m3 for a given ξ*AE*, ξ*AE* is the accumulated energy in J/m2, *TOC*<sup>0</sup> is *TOC* of the mixture measured at the starting point of the experiment, and *VR* and *VT* are the volumes of the photoreactor and the total reaction volume in m3, respectively. −*rTOCVR* is the average reaction rate of the *TOC* of the mixture in (mol m3 <sup>s</sup><sup>−</sup>1). The mathematical expressions for −*rTOCVR* are the same given in Table 2, replacing −*riVR* by −*rTOCVR* and *Ci* by *TOC*.

Similarly, Equation (5) is rewritten as:

$$F\_{obj} = \sum\_{i=1}^{N} \left( TOC\_{i, \text{exp}} - TOC\_{i, \text{calc}} \right)^2 \tag{7}$$

Then, for multicomponent mixtures, the *TOC* of the mixture must be provided to PHOTOREAC as a function of the accumulated energy instead of the concentration of a pure component water contaminant. This approach is particularly useful in real environmental applications because in such cases the most usual situation is that the content of the wastewater is unknown, and it would be tough and resource-consuming to determine it. Therefore, it is easier to establish a global parameter such as the *TOC*, which shows the mineralization of both intermediates and the precursor compounds in the wastewater. By contrast, the monitoring of each initial pure component in the mixture does not consider the formation of intermediates.

#### The Kinetic Modeling Module Database

In the kinetic modeling module, PHOTOREAC incorporates a database that consists of 26 datasets of the solar photocatalytic degradation of water contaminants using TiO2 P25 Evonik as a photocatalyst. The information was collected by the Modeling and Applications of Advanced Oxidation Technologies Research Group at Cartagena University (Cartagena de Indias, Colombia) and the Research Group on Advanced Processes for Biological and Chemical Treatments (GAOX) at the Universidad del Valle (Cali, Colombia). Table 3 details the information available in the database: two solar photoreactor configurations (CPCP and OMTP) and five model pollutants at different initial concentrations and photocatalyst concentrations. By selecting the dataset to perform the kinetic analysis, PHOTOREAC loads the information about the experimental test: the pollutant concentration vs. accumulated energy data, the OVRPA and the β = *VR*/*VT* factor.


**Table 3.** PHOTOREAC database of the solar photodegradation of water contaminants.

#### **3. Implementation of PHOTOREAC in Solar Photoreactors**

In this section, three example cases to demonstrate the use of the PHOTOREAC application are presented. All of the cases are based on an experimental test already performed in the solar photoreactor platforms of our research groups in Cartagena, Colombia (10◦25- 25" N, 75◦31- 31" W) and Cali, Colombia (3◦27- 00" N, 76◦32- 00" W). Further information about the set-up and operation of the experimental solar tests can be found in previous works [10,12].

#### *3.1. Example Case I: Solar Photodegradation of Dichloroacetic Acid (DCA) in a CPCP*

This example shows the implementation of PHOTOREAC for an analysis of the solar photocatalytic degradation of DCA in a CPCP. The photoreactor consists of ten borosilicate tubes with radius *R* = 0.016 m and length *L* = 1.2 m providing a reaction volume of *VR* = 9.7 L. The DCA initial concentration was *Ci* = 30 ppm using a TiO2 P25 Evonik concentration of *Ccat* = 0.5 g/L. The main objective of the example case was to determine the radiation-independent kinetic parameters of the system from the experimental data provided to the application using the SFM as the radiative model.

First, the radiation field is determined by the photon absorption-scattering module. Figure 3 shows the main screen of the PHOTOREAC GUI: (1) the photoreactor panel, where the photoreactor configuration was selected; (2) the system properties panel, where the input data were introduced for the simulation; (3) the SFM model panel, where the SFM variant for the simulation is selected; (4) the SFM scattering phase function probabilities are displayed according to the SFM variant that was selected, in this case, the SFM; (5) the resulting LVRPA spatial distribution in the cross-section of the CPCP tube is plotted; (6) the resulting OVRPA of the system is displayed; (7) the options menu. Together with the main screen shown in Figure 3, PHOTOREAC generates a secondary screen with the results of the ray-tracing simulation (Figure 4).

**Figure 3.** Radiation field simulation for a CPCP in the photon absorption-scattering module. (1) photoreactor panel; (2) system properties panel; (3) the SFM panel; (4) display the corresponding SFM scattering probabilities used in the simulation; (5) LVRPA spatial distribution plot; (6) OVRPA; (7) options menu.

**Figure 4.** Ray-tracing simulation for the CPCP.

From the results presented by the photon absorption-scattering module, the user will be able to extract essential findings regarding the impact of variables on the photocatalyst concentration. For instance, for this example case, in the LVRPA distribution plot shown in Figure 3, it is observed that the highest values of the LVRPA are around *y* = −0.015 m and *y* = −0.005 m. This result is due to the fact that at these coordinates there is a high concentration of rays that come from the CPCP reflectors, as can be observed in Figure 4. Additionally, it is observed that the LVRPA is concentrated near to the CPCP wall, and the center of the tube shows very low LVRPA values, as a result of the relatively high photocatalyst concentration used in the simulation (*Ccat* = 0.5 g/L). This behavior is well-known in the literature: at high concentrations of the photocatalyst, the photons cannot penetrate deeply into the tube and the absorbed energy is concentrated around the boundary wall [14].

Once the radiation field for the CPCP is determined, the application proceeds to the kinetic modeling module. Figure 5 shows the input panel displayed by PHOTOREAC. The application loads the system parameters determined previously, such as the TiO2 concentration and the OVRPA. The remaining system parameters must be provided manually by the user. Similarly, the photodegradation vs. accumulated energy data should be introduced in the experimental data panel. Finally, the user may proceed to the kinetic modeling module's main screen.

**Figure 5.** Input panel for the kinetic modeling module. (1) System parameters panel; (2) experimental data panel.

Figure 6 shows the main screen of the PHOTOREAC GUI at the kinetic modeling module: (1) kinetic models panel, where the user can choose the kinetic models to be fitted; (2) experimental data and models simulations plot, where the experimental data and the fitting curves of the models that were previously selected are displayed; (3) fitted kinetic parameters panel, where the values of the fitting parameters of each model chosen are displayed; (4) the *x*-axis magnitude panel, where the user can determine if the displayed data are presented in accumulated energy or standard time as the *x*-axis magnitude; (5) correlation coefficient panel, which displays the higher *R*<sup>2</sup> among the selected kinetic models; (6) correlation coefficient panel, which shows the kinetic model with the highest *R*<sup>2</sup> value among the chosen ones; (7) export data button, which exports the results of the fitting curve to a Microsoft Excel file; (8) options menu panel.

From the PHOTOREAC kinetic modeling module screen in Figure 6, it is observed that the best fitting is achieved for the Ballari et al. model with *R*<sup>2</sup> = 0.97392. The other models reported *R*<sup>2</sup> = 0.97365 for Mueses et al., *R*<sup>2</sup> = 0.63843 for Langmuir–Hinshelwood and *R*<sup>2</sup> = 0. 63585 for Zalasar et al. Due to PHOTOREAC only displaying the model with the higher value for the correlation coefficient *R*2, it selected the Ballari et al. model. However, Mueses et al.'s expression showed an almost identical *R*2, and it should not be discarded without further analysis. From Table 2, it is observed that the mathematical structure of the Ballari et al. and Mueses et al. expressions are very similar; indeed, the Ballari et al. expression is considered a particular case of the Mueses et al. model for systems with high molecular adsorption [12]. Therefore, it is expected that both models performed similarly, as is the case for the DCA photodegradation. The Langmuir–Hinshelwood and Zalasar et al. expressions may not lead to successful results due to the fact that they do not describe the effects of the absorbed radiation (OVRPA) accurately. On the other hand, Ballari et al. and Mueses et al. may perform better since they include an OVRPA squared root correction factor. the same can be said for Ballari et al. and Mueses et al. regarding the OVRPA squared root correction factor.

**Figure 6.** DCA fitting for the available kinetics models in the kinetics modeling module. (1) Kinetic models panel; (2) experimental data and models simulations plot; (3) fitted kinetic parameters panel; (4) *x*-axis magnitude panel; (5) correlation coefficient panel; (6) display of the kinetic model with the highest *R*<sup>2</sup> value among the ones selected; (7) export data button; (8) options menu panel.

### *3.2. Example Case II: Solar Photodegradation of Methylene Blue in an OMTP*

In the previous example case, the user must provide all the required information to perform the computations. In this example, the use of the database incorporated in the PHOTOREAC kinetic modeling module is shown. Figure 7 shows the PHOTOREAC screen of the kinetic modeling module: (1) the photoreactor configuration panel, for selecting the photoreactor to be studied; (2) the model pollutant panel, for choosing the water contaminants from the five available options in the database; (3) the photocatalyst-pollutant panel, for choosing the photocatalyst concentration-initial pollutant concentration combination from the available options in the database; (4) the experimental data panel, for loading the pollutant concentration vs. accumulated energy (or standard time); (5) the system parameters panel, which displays the OVRPA and the β = *VR*/*VT* factor charged.

**Figure 7.** Database in the kinetic modeling module. (1) photoreactor configuration panel; (2) model pollutants panel; (3) photocatalyst-pollutant panel; (4) experimental data panel; (5) system parameters panel.

In this case, an OMTP with methylene blue (MB) was selected as a model pollutant with an initial concentration of *Ci* = 10 ppm and a photocatalyst concentration of *Ccat* = 0.2 g/L. Figure 8 shows the results obtained by PHOTOREAC. It is observed that the best fitting is achieved for the Mueses et al. and Ballari et al. models with *R*<sup>2</sup> = 0.99737 for both. The other models reported *R*<sup>2</sup> = 0.001. As the Ballari et al. model is a particular case of the Mueses et al. model, the first is considered the more appropriate option since it is more specific for this case. These results agree with the discussion presented in the previous section.

**Figure 8.** Methylene blue solar photodegradation fitting in an offset multi-tubular photoreactor (OMTP) for the four available kinetics models in the kinetic modeling module.

## *3.3. Example Case III: Radiation Field Modeling in a Flat Plate Photoreactor (FPP)*

In this case, the objective was to compare the radiation field simulation for an FPP using SFM and SFM-HG. The photoreactor consists of a titled squared flat plate of length *L* = 1 m, which is placed facing the sun and uniformly irradiated. A water film of 1 cm thickness flows over its surface, providing a reaction volume of *VR* = 10 L. The TiO2 P25 Evonik concentration is *Ccat* = 0.2 g/L. Figures 9 and S1 show the LVRPA profile in the FPP calculated with the SFM-HG and the SFM, respectively. In both cases, the highest LVRPA values are found near to the surface of the water film (thickness = 0–0.2 cm) because this is the boundary that the solar light irradiates. After 0.2 cm, exponential decay in the LVRPA occurs as a result of the absorption and scattering of photons by the suspended photocatalyst. Due to the photoreactor being considered as uniformly irradiated, the changes in the LVRPA profile are only significant along with the water film thickness.

In Figure S1, which uses the SFM, a shaper exponential LVRPA profile is observed, with higher values near the irradiated boundary (at thickness = 0–0.2 cm) when comparing to values in Figure 9, which uses the SFM-HG. These results are due to the difference in the scattering phase function; the SFM-HG uses a predominantly forward scattering phase function, which causes photons to penetrate deeper into the water film. By contrast, the SFM uses a predominantly backward phase function, which causes that photons to be redirected toward the irradiated boundary and be mostly absorbed in the beginning of the film or escape from the system [15].

**Figure 9.** Radiation field simulation of a flat plate photoreactor (FPP) with SFM-HG.

## **4. PHOTOREAC Implementation in Research Projects in Heterogeneous Photocatalysis and Photoreactor Engineering by Chemical Engineering Undergraduates**

Since the year 2015, different versions of PHOTOREAC have supported the final degree projects of chemical engineering students belonging to the Modeling and Applications of Advanced Oxidation Technologies Research Group at Cartagena University. The students developed research on heterogeneous photocatalysis and photoreactor engineering. A survey was done amongst them to determine the perceived impact of PHOTOREAC on their final degree projects. Table 4 shows the results of the survey. Between 2015–2018, ten final degree projects were developed in the research group, with an average impact of 37.5% perceived by the students. The use of PHOTOREAC can be summarized as follows: in four of the degree projects, both modules of PHOTOREAC were implemented since they performed the radiation field simulation and kinetic modeling of model pollutants; in two other projects, the photon-scattering module was used to determine the radiation field in photoreactors; finally, in four projects, the application was used in the learning process for modeling solar photoreactors. As a relevant outcome, two of the degree projects supported publications in high-impact journals. In all of the projects, the authors highlighted the use of PHOTOREAC as a user-friendly tool that allows them to reach the main objective of the projects or to achieve a fast advance in the learning curve, therefore allowing them to focus on more complex research.




**Table 4.** *Cont.*

#### **5. Analysis of the Overall Performance of PHOTOREAC**

PHOTOREAC was shown to be a useful tool for modeling and simulation of solar photoreactors, and in particular for a non-expert public. Its user-friendly interface developed in the graphical user interface of Matlab proved to be intuitive enough to be used successfully by chemical engineering undergraduates, which develop research in heterogeneous photocatalysis.

In Section 3, the application was evaluated for disparate operational conditions, showing that it can fit and simulate the photodegradation experimental data provided for the two cases evaluated: CPCP-DCA (example case I) and OMTP-MB (example case II). These example cases were very different from each other, mainly because of the different photoreactor geometries: CPCP can capture more

solar radiation per length than the OMTP as a result of being equipped with reflectors. However, OMTP has more volume than the CPCP [10]. Additionally, the employed model pollutant, its initial concentration, and the photocatalyst concentration were different. In both examples, PHOTOREAC performed successfully, allowing the user to evaluate different kinetic expressions and extract relevant findings from it. Finally, example case III focused on the photon absorption-scattering module, in which the impact of the radiation model was evaluated and discussed for an FFP. In this case, PHOTOREAC shows its versatility for researchers with an interest in studying the energy absorption behaviors of photoreactors.

The dedicated interface of PHOTOREAC for photoreaction engineering, together with its numerical algorithm, allowed the evaluation of the performance of large scale solar photoreactors without time-consuming computations and a complex mathematical formulation. The time invested in preparing and launching a simulation in PHOTOREAC is between 5–10 min, and the calculation time does not exceed 45 s. In contrast to commercial CFD simulators in which preparing and starting a first-time simulation may take a couple of hours needed for generating the photoreactor geometry in the system (or it importing it from CAD software), preparing the simulation modules and their models and selecting the proper meshing and numerical algorithms; besides, the computational time for each simulation is, generally, measured in hours [21–23]. Nevertheless, the results obtained by CFD simulators are much more complete and accurate than the results that PHOTOREAC may offer; for instance, CFD simulators provide detailed flow patterns for studying the hydrodynamics in the photoreactor. However, its high computational time may result in a barrier when exploring the impact of numerous parameters on a wide range of values.

Moreover, the most common CFD commercial simulators used in modeling photoreactors are very expensive licensed software. At the same time, PHOTOREAC is an open-access application that is available on-demand, by email to one of the authors of the paper, professor Miguel A. Mueses (mmueses@unicartagena.edu.co).

In conclusion, PHOTOREAC is recommended for the following cases: (i) for an introduction to photoreactor engineering; (ii) when a quantitative margin of error is still acceptable in the calculations; (iii) when qualitative results are the main objective of the work; and (iv) when the parametric space in the study is extensive, i.e., it is required to study the impact of numerous variables in broad ranges. In this case, PHOTOREAC may be employed to reduce the parametric space and then to implement a CFD simulator.

### **6. Limitations and Future Work**

As with every modeling software, PHOTOREAC is limited by the availability and reliability of the input data provided by the users. Additionally, the computational application is limited to Titanium Dioxide P25 Evonik as photocatalyst. Although TiO2 P25 is the most common photocatalyst, the capability of performing simulations for any photocatalyst will be crucial for the software, since an area of intensive research in heterogeneous photocatalysis is the development and testing of new photoactive materials. On the other hand, expanding the available kinetic models would also be a considerable improvement, because it will allow users to make a more comprehensive analysis by comparing the results of the kinetic models' fitting. Moreover, it is necessary to implement the option that users introduces their own kinetic expression, since some pollutants will require concrete mathematical expression because their kinetic mechanism may not follow the most common postulates. These drawbacks are expected to be overcome in the upcoming version of PHOTOREAC.

In the authors' opinion, some important challenges for PHOTOREAC and, in general, for photoreaction engineering at the pilot-solar scale are that the models account for the variability of the incident radiation on the solar photoreactor caused by fluctuations in atmospheric conditions. This improvement will allow more accurate quantification of the energy absorbed by the suspended photocatalyst, and therefore better quantification of the chemical species produced by the photoactivation of the photocatalyst.

*Water* **2020**, *12*, 2196

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4441/12/8/2196/s1, Figure S1: Radiation field simulation of a flat plate photoreactor (FPP) with SFM.

**Author Contributions:** Conceptualization, B.C.-V., K.P.-G. and M.A.M.; data curation, B.C.-V., and K.P.-G.; formal analysis, R.A.-H.; funding acquisition, M.H.P.-C. and F.M.-M.; investigation, B.C.-V. and K.P.-G.; methodology, B.C.-V. and K.P.-G.; project administration, M.H.P.-C. and F.M.-M.; supervision, R.A.-H. and M.A.M.; validation, B.C.-V. and K.P.-G.; writing—original draft, R.A.-H.; writing—review and editing, R.A.-H., M.H.P.-C., M.A.M. and F.M.-M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Universidad del Valle, grant number C.I. 21022.

**Acknowledgments:** The authors gratefully thank Universidad de Cartagena (Cartagena, Colombia) with the project 017-2018: "Plan de Fortalecimiento del Grupo de Investigación Modelado y Aplicación de Procesos Avanzados de Oxidación" and Universidad del Valle (Cali, Colombia) with the Project CI. 21022 "Estudio del efecto hidrodinámico y de transporte de energía radiante en el diseño y optimización de reactores fotocatalíticos heterogéneos solares. CI 21022." for financial support. Acosta-Herazo thanks the CEIBA foundation with the program "Bolivar Gana con Ciencia" for financing his doctoral studies.

**Conflicts of Interest:** The authors declare no conflict of interest.

## **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Water* Editorial Office E-mail: water@mdpi.com www.mdpi.com/journal/water

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18