Assessment of a Novel Photocatalytic TiO 2 -Zirconia Ultraﬁltration Membrane and Combination with Solar Photo-Fenton Tertiary Treatment of Urban Wastewater

: The objective of this study was to assess the combination of a photocatalytic TiO 2 -coated ZrO 2 UF membrane with solar photo-Fenton treatment at circumneutral pH for the ﬁltration and treatment of urban wastewater treatment plant (UWWTP) efﬂuents. Photocatalytic self-cleaning properties were tested with a UWWTP efﬂuent under irradiation in a solar simulator. Then, both the permeates and retentates from the membrane process were treated using the solar photo-Fenton treatment. The UWWTP efﬂuent was spiked with caffeine (CAF), imidacloprid (IMI), thiacloprid (THI), carbamazepine (CBZ) and diclofenac (DCF) at an initial concentration of 100 µ g/L each. Retention on the membrane of Pseudomonas Aeruginosa (P. Aeruginosa) , a Gram-negative bacterial strain, was tested with and without irradiation. It was demonstrated that ﬁltration of a certain volume of UWWTP efﬂuent in the dark is possible, and the original conditions can then be recovered after illumination. The photocatalytic membrane signiﬁcantly reduces the turbidity of the UWWTP efﬂuent, signiﬁcantly increasing the degradation efﬁciency of the subsequent solar photo-Fenton treatment. The results showed that the membrane allowed consistent retention of P. Aeruginosa at an order of magnitude of 1 × 10 3 –1 × 10 4 CFU/mL.


Introduction
Membranes have proven to be economic, fast and reliable solutions for the world's increasing water contamination problems. These vast and growing problems are due to the increasing use of pesticides and other chemicals known as microcontaminants (MCs), which are usually found in the already limited fresh water supply and agricultural lands, combined with salination [1][2][3]. MCs are found at trace levels up to significant concentrations, accumulating in wastewater streams and bodies, especially urban wastewater treatment plant (UWWTP) effluent and groundwater. As they are usually within the range of ng/L to µg/L, they often pass through conventional water treatment methods without being treated or recorded. These Contaminants of Emerging Concern (CECs) are found worldwide but are not yet regulated [4,5].
CECs can be removed from UWW with advanced oxidation processes (AOPs), which by generating highly reactive and nonselective hydroxyl radicals ( • OH), can be applied for the elimination of MCs in a wide variety of UWWs, such as ozonation, photo-Fenton or electrooxidation processes [6][7][8]. On the other hand, other technologies have also been used successfully to remove pharmaceuticals contained in the UWWTP effluents, especially

Reagents and Chemicals
The UWWTP effluent was taken after secondary treatment in the 'El Bobar' UWWTP located in southeast Spain (Almeria) (see Table S1 for physicochemical characterization). Sigma Aldrich supplied Fe 2 (SO 4 ) 3 ·xH 2 O for Fe(III)) and the HPLC-grade solvents for MC monitoring. MCs were obtained mainly from Sigma Aldrich (IMI, THI, CBZ and DCF), whereas CAF was supplied by Fluka. The reagents, H 2 O 2 (35% w/v) and sodium persulfate (Na 2 S 2 O 8 ), were also purchased from Sigma Aldrich. A schematic overview of the five selected MCs is shown in Table S2. UPLC column retention time, LOQ and maximum contaminant absorption data are summarized in Table S3. Type CM0067 Nutrient Broth No. 2 was supplied by OXCID LTD., England, and the Pseudomonas Chromogenic Agar by Condalab, Spain, whereas the Pseudomonas Aeruginosa was from the Spanish Culture Collection (CECT).

Analytical Determinations
Scanning electron microscopy (SEM) images were taken with a Zeiss Sigma 300 VP Gemini Technology electron microscope at variable pressure. The SEM was equipped for energy-dispersive X-ray spectroscopy (EDX) to determine the chemical composition of the ceramic membrane samples by identifying the elements present, along with their concentration and distribution. Concentrations of selected MCs were determined by ultraperformance liquid chromatography (UPLC) using an Agilent Technologies 1200 series device equipped with a UV-DAD detector and Poroshell 120 EC-C18 column (3.0 × 50 mm). Eluent conditions started from 95% water with 25 mM formic acid (mobile phase A) and 5% acetonitrile (ACN) (mobile phase B) at 1 mL/min for 5 min. This was followed by a 10 min linear gradient to 68% ACN, which was maintained for 2 min. The working temperature was 30 • C and the injection volume was 100 µL. The samples prepared at a 1:9 ACN ratio were mixed and filtered through a hydrophobic PTFE 0.2 µm (Millipore Millex-FG) syringe filter into a 2 mL HPLC vial. Molecular weight, UPLC column retention time, LOQ and maximum contaminant absorption data are given in Table S3. Dissolved organic carbon (DOC) and inorganic carbon values were determined using a Shimadzu TOC-VCN analyzer. H 2 O 2 values were determined according to DIN 38402H15 with titanium (IV) oxysulfate. Iron values were determined after filtration through a 0.45 µm nylon filter using 1,10 phenanthroline according to ISO 6332. The Liang method [33] was adapted to persulfate determination. Turbidity was measured using a Hach 2100N Turbidimeter. Total suspended solids (TSS) were measured via filtration without binder through a predried and weighed GF 52 090 Glass Microfiber Filter from ALBET LabScience, Germany. The filter paper was then placed in an oven at 105 • C for 60 min. Bacteria were cultured in a Series BD standard incubator at 36 • C.

Experimental Setup
The UWWTP effluent was filtered through a photocatalytic ceramic ZrO 2 ultrafiltration flat-sheet membrane designed and fabricated in a previous study. The membrane consists of a 100 ± 1-mm-wide, 150 ± 2-mm-long, highly porous, multi-channeled SiC support with 25 2 × 2 mm channels and 15 µm pore size supplied by LiqTech Ceramics A/S (Ballerup, Denmark). This is followed by a ZrO 2 /SiC intermediate layer with 60 nm pores, covered with a wash coating of TiO 2 particles with a mean size of 10 nm. Membrane function follows the submerged outside-in filtration principle and is able to operate in a pH range of 0-14. Further details are described elsewhere [28]. SEM images of a cut surface can be  Figure 1. Only the irradiated side of the membrane was left open; the remaining sides and bottom of the membrane were covered with a watertight polymer film to ensure that the permeation flow rate was only measured on the 150 cm 2 irradiated surface. The membrane was held in place by a fixture to provide an evenly divided crossflow over both membrane surfaces and was placed on the bottom of a 6 L container. The container was then positioned in a SunTest XLS+ solar simulator, containing a xenon lamp with daylight filter, providing a total light irradiation of 365 W/m 2 (300-800 nm), with a total UV radiation of 30 W/m 2 (300-400 nm), while the temperature was kept at 25 • C. A vacuum was created in the membrane support with a Watson Marlow 520 S peristaltic pump through a MARPRENE 902.0048.016 #25 tube at 20 rpm to obtain the permeate volume (see Figure 2 for a schematic view). The MC mixture containing CAF, IMI, THI, CBZ and DCF in methanol had previously been prepared as a stock solution. Each compound had a final concentration of 2.5 g/L to ensure solubility and low DOC. The UWWTP effluent was spiked with this stock solution to a final concentration of 100 µg/L per MC.
channeled SiC support with 25 2 × 2 mm channels and 15 µm pore size supplied by LiqTech Ceramics A/S (Ballerup, Denmark). This is followed by a ZrO2/SiC intermediate layer with 60 nm pores, covered with a wash coating of TiO2 particles with a mean size of 10 nm. Membrane function follows the submerged outside-in filtration principle and is able to operate in a pH range of 0-14. Further details are described elsewhere [28]. SEM images of a cut surface can be seen in Figure 1. Only the irradiated side of the membrane was left open; the remaining sides and bottom of the membrane were covered with a watertight polymer film to ensure that the permeation flow rate was only measured on the 150 cm 2 irradiated surface. The membrane was held in place by a fixture to provide an evenly divided crossflow over both membrane surfaces and was placed on the bottom of a 6 L container. The container was then positioned in a SunTest XLS+ solar simulator, containing a xenon lamp with daylight filter, providing a total light irradiation of 365 W/m 2 (300-800 nm), with a total UV radiation of 30 W/m 2 (300-400 nm), while the temperature was kept at 25 °C. A vacuum was created in the membrane support with a Watson Marlow 520 S peristaltic pump through a MARPRENE 902.0048.016 #25 tube at 20 rpm to obtain the permeate volume (see Figure 2 for a schematic view). The MC mixture containing CAF, IMI, THI, CBZ and DCF in methanol had previously been prepared as a stock solution. Each compound had a final concentration of 2.5 g/L to ensure solubility and low DOC. The UWWTP effluent was spiked with this stock solution to a final concentration of 100 µg/L per MC.

Solar Photo-Fenton Treatment
Solar photo-Fenton experiments were performed in the solar simulator described above, using a cylindrical 1 L container as the reactor (18.5 cm diameter and 4.0 cm depth), which was held in place and stirred magnetically in the center of the solar simulator. In selected experiments, the bicarbonates, known to be • OH radical scavenger, naturally present in UWWTP effluents, were air-stripped from the experimental volumes before starting the solar photo-Fenton treatments by adding H2SO4 to lower the concentration of HCO3 − to 75 mg/L. The Fe 3+ :EDDS complex was formed by predissolving Fe2(SO4)3.7H2O in demineralized water at pH 3, followed by the addition of EDDS. A ratio of 1:1 was maintained during all experiments and the starting reagent concentration was 1.50 mM for both H2O2 and persulfate. Table S4 summarizes the experiments.

Retention of P. Aeruginosa
The membrane was placed in the container as described above, and the system was sterilized with H2O2 to ensure that there were no living bacteria in it. In this case, the water matrix was 5 L of natural water (Tabernas, southeast Spain). The P. Aeruginosa inoculum was prepared 20 h prior to the start of experiment and added to the reactor at a concentration of 1 × 10 6 CFU/mL, which is known for its predominant bacterial growth and is commonly used as a standard inoculum concentration to be followed by bacterial inactivation [34][35][36]. Petri dishes with bovine agar were prepared for sampling by applying a 10-fold serial dilution with sample volumes of 50 µL each for the concentrate and a total sample volume of 500 µL for the permeate. Samples of the membrane surface were taken with a swab before the experiment and after 90 min and applied on the agar for cultivation. Further samples were taken from the reactor (concentrate) at 0, 30, 60 and 90 min, and afterwards from the peristaltic pump (permeate). Permeate volumes were measured separately every 30 min, showing a constant flow. The Petri dishes with the samples were stored in an incubator at 36 °C for 48 h.

Fouling and Self-Cleaning Properties
Several batches of UWWTP effluents were filtered through the TiO2-ZrO2 UF membrane. The turbidity varied from 8.7 to 21.3 NTU. During filtration to a concentration factor (CF) of 2, the pH of the concentrate slightly increased. The resulting permeate

Solar Photo-Fenton Treatment
Solar photo-Fenton experiments were performed in the solar simulator described above, using a cylindrical 1 L container as the reactor (18.5 cm diameter and 4.0 cm depth), which was held in place and stirred magnetically in the center of the solar simulator. In selected experiments, the bicarbonates, known to be • OH radical scavenger, naturally present in UWWTP effluents, were air-stripped from the experimental volumes before starting the solar photo-Fenton treatments by adding H 2 SO 4 to lower the concentration of HCO 3 − to 75 mg/L. The Fe 3+ :EDDS complex was formed by predissolving Fe 2 (SO 4 ) 3 .7H 2 O in demineralized water at pH 3, followed by the addition of EDDS. A ratio of 1:1 was maintained during all experiments and the starting reagent concentration was 1.50 mM for both H 2 O 2 and persulfate. Table S4 summarizes the experiments.

Retention of P. Aeruginosa
The membrane was placed in the container as described above, and the system was sterilized with H 2 O 2 to ensure that there were no living bacteria in it. In this case, the water matrix was 5 L of natural water (Tabernas, southeast Spain). The P. Aeruginosa inoculum was prepared 20 h prior to the start of experiment and added to the reactor at a concentration of 1 × 10 6 CFU/mL, which is known for its predominant bacterial growth and is commonly used as a standard inoculum concentration to be followed by bacterial inactivation [34][35][36]. Petri dishes with bovine agar were prepared for sampling by applying a 10-fold serial dilution with sample volumes of 50 µL each for the concentrate and a total sample volume of 500 µL for the permeate. Samples of the membrane surface were taken with a swab before the experiment and after 90 min and applied on the agar for cultivation. Further samples were taken from the reactor (concentrate) at 0, 30, 60 and 90 min, and afterwards from the peristaltic pump (permeate). Permeate volumes were measured separately every 30 min, showing a constant flow. The Petri dishes with the samples were stored in an incubator at 36 • C for 48 h.

Fouling and Self-Cleaning Properties
Several batches of UWWTP effluents were filtered through the TiO 2 -ZrO 2 UF membrane. The turbidity varied from 8.7 to 21.3 NTU. During filtration to a concentration factor (CF) of 2, the pH of the concentrate slightly increased. The resulting permeate turbidity range, measured periodically from the start of the experiment to CF = 2, was 0.4 to 0.6 NTU. The TSS values of the different permeate volumes varied from 0.6 to 1.3 mg/L.
The membrane's self-cleaning capabilities were determined by continuously filtering the UWWTP effluents for 6 h (Figure 3a). The severely fouled membrane was then placed in a beaker under simulated sunlight with demineralized water and stirred lightly. The beaker was placed in the center of the solar simulator at the above irradiation settings turbidity range, measured periodically from the start of the experiment to CF = 2, was 0.4 to 0.6 NTU. The TSS values of the different permeate volumes varied from 0.6 to 1.3 mg/L. The membrane's self-cleaning capabilities were determined by continuously filtering the UWWTP effluents for 6 h (Figure 3a). The severely fouled membrane was then placed in a beaker under simulated sunlight with demineralized water and stirred lightly. The beaker was placed in the center of the solar simulator at the above irradiation settings for 1 h to remove the fouling ( Figure   A clear difference from the originally severely fouled state can be observed after three 1 h irradiation periods. As a control test, another fouled membrane sample was first rinsed and placed in demineralized water but without illumination, then lightly stirred for three hours. Even after manual cleaning, membrane fouling could not be removed, demonstrating the self-cleaning effects of the photocatalytic membrane surface under irradiation. Both the original state of the fouled membrane ( Figure 4a) and the membrane after 3 h of self-cleaning under simulated solar irradiation were further assessed by SEM ( Figure 4b). Figure 4a is dark gray due to the presence of organic material and fouling. The presence of organic material was confirmed by EDX as having an atomic percentage of carbon of over 90%. Figure 4b shows the membrane surface with a dark spot of residual fouling in the middle. The surrounding area is the TiO2 layer. The light-white area in the upper left corner is a minor defect of the TiO2 layer due to previous manual cleaning.
After observing a clear visual difference in the state of fouling, its physical removalthat is, recovery of most of the membrane porosity and not just a change in color-was confirmed in filtration experiments. Here, 250 mL filtered samples were taken, and the time necessary to obtain this amount was recorded to determine whether the time needed to filter a certain volume could be affected by cleaning and whether the original filtration capability could be reached (see Figure 5). The normalized flow is shown in Figure 5 to the demonstrate change in flow rate based on the original flow and how it slows down for the reasons stated below. In both cases, a clear difference in flow rate can be seen between the membrane kept in the dark and under light during filtration. This again demonstrates the self-cleaning capabilities of the photocatalytic membrane when filtering UWWTP effluents. A clear difference from the originally severely fouled state can be observed after three 1 h irradiation periods. As a control test, another fouled membrane sample was first rinsed and placed in demineralized water but without illumination, then lightly stirred for three hours. Even after manual cleaning, membrane fouling could not be removed, demonstrating the self-cleaning effects of the photocatalytic membrane surface under irradiation. Both the original state of the fouled membrane ( Figure 4a) and the membrane after 3 h of self-cleaning under simulated solar irradiation were further assessed by SEM (Figure 4b). Figure 4a is dark gray due to the presence of organic material and fouling. The presence of organic material was confirmed by EDX as having an atomic percentage of carbon of over 90%. Figure 4b shows the membrane surface with a dark spot of residual fouling in the middle. The surrounding area is the TiO 2 layer. The light-white area in the upper left corner is a minor defect of the TiO 2 layer due to previous manual cleaning.  The original flow rate of 21 mL/min through the membrane in the light can be observed to be higher than the 17 mL/min rate in the dark. This can be attributed to irradiation of the membrane inhibiting attachment of suspended matter in the UWWTP effluent on the photocatalytic membrane surface. Another phenomenon is photo-induced super-hydrophilicity (PSH) of TiO2, resulting in higher flow rates and less fouling, as also previously observed [28] and reported elsewhere [37,38]. Strong decreases in flow rate can be observed up to around 60 min, both in the light and in the dark, as fouling builds up After observing a clear visual difference in the state of fouling, its physical removalthat is, recovery of most of the membrane porosity and not just a change in color-was confirmed in filtration experiments. Here, 250 mL filtered samples were taken, and the time necessary to obtain this amount was recorded to determine whether the time needed to filter a certain volume could be affected by cleaning and whether the original filtration capability could be reached (see Figure 5). The normalized flow is shown in Figure 5 to the demonstrate change in flow rate based on the original flow and how it slows down for the reasons stated below. In both cases, a clear difference in flow rate can be seen between the membrane kept in the dark and under light during filtration. This again demonstrates the self-cleaning capabilities of the photocatalytic membrane when filtering UWWTP effluents.
The original flow rate of 21 mL/min through the membrane in the light can be observed to be higher than the 17 mL/min rate in the dark. This can be attributed to irradiation of the membrane inhibiting attachment of suspended matter in the UWWTP effluent on the photocatalytic membrane surface. Another phenomenon is photo-induced super-hydrophilicity (PSH) of TiO2, resulting in higher flow rates and less fouling, as also previously observed [28] and reported elsewhere [37,38]. Strong decreases in flow rate can be observed up to around 60 min, both in the light and in the dark, as fouling builds up inside the membrane pores where irradiated light cannot reach. From 60 to 120 min, a clear difference can be observed in the curve of the normalized flow rate in the light (Figure 5b), strongly suggesting that there is no photodegradation of fouling on the outer membrane surface in the dark experiment. After 120 min, the curve of filtration under radiation follows the same trend as the curve of filtration in the dark, suggesting that both membranes follow a similar trend in internal fouling, while the outer surface of the illuminated membrane is kept clean. Here, 1 L of permeate is generated in 120 min of operation in the dark. The next step involved the observation of the effects of self-cleaning on the fouled membrane after 60 of minutes irradiation, which is considered the necessary filtration time, because this is when the curve of the normalized flow in the light deviates from the normalized flow in the dark. Furthermore, over this time, the UWWTP effluent was also spiked with 100 µg/L of each of the target MCs, namely CAF, IMI, THI, CBZ and DCF. The original flow rate of 21 mL/min through the membrane in the light can be observed to be higher than the 17 mL/min rate in the dark. This can be attributed to irradiation of the membrane inhibiting attachment of suspended matter in the UWWTP effluent on the photocatalytic membrane surface. Another phenomenon is photo-induced super-hydrophilicity (PSH) of TiO 2 , resulting in higher flow rates and less fouling, as also previously observed [28] and reported elsewhere [37,38]. Strong decreases in flow rate can be observed up to around 60 min, both in the light and in the dark, as fouling builds up inside the membrane pores where irradiated light cannot reach. From 60 to 120 min, a clear difference can be observed in the curve of the normalized flow rate in the light (Figure 5b), strongly suggesting that there is no photodegradation of fouling on the outer membrane surface in the dark experiment. After 120 min, the curve of filtration under radiation follows the same trend as the curve of filtration in the dark, suggesting that both membranes follow a similar trend in internal fouling, while the outer surface of the illuminated membrane is kept clean. Here, 1 L of permeate is generated in 120 min of operation in the dark.
The next step involved the observation of the effects of self-cleaning on the fouled membrane after 60 of minutes irradiation, which is considered the necessary filtration time, because this is when the curve of the normalized flow in the light deviates from the normalized flow in the dark. Furthermore, over this time, the UWWTP effluent was also spiked with 100 µg/L of each of the target MCs, namely CAF, IMI, THI, CBZ and DCF. Figure 6 shows the flow rate during the first filtration of 1 L of UWWTP effluent in the dark. Then, the membrane was cleaned for 60 min without filtration under illumination. The second period began after this, with further filtration of 2 L of UWWTP effluent in the dark, followed by cleaning again using the same procedure followed by a third filtration. It was demonstrated that filtration of a certain volume of UWWTP effluent in the dark is possible only up to a certain amount of fouling, although the original conditions can then be recovered after illumination. However, after filtering a larger volume of UWWTP effluent, fouling could become irreversible, emphasizing the need for the definition and optimization of operating variables when using photocatalytic membranes to maintain their self-cleaning capability. There was no retention of MCs due to the high porosity of the photocatalytic membrane, and there was no difference in the operation of membranes with UWWTP effluent, with or without MCs, as most of the fouling was eliminated by self-cleaning. operation of membranes with UWWTP effluent, with or without MCs, as most of the fouling was eliminated by self-cleaning.
The self-cleaning mechanisms of the TiO2-ZrO2 UF membrane are not only based on the previously mentioned super-hydrophilicity. Other mechanisms that prevent fouling of different compounds on the membrane surface can be assigned to the formation of different radical species ( • OH and other) when the photocatalytic surface material is irradiated with a photon energy equal or higher than the band gap energy of the semiconductor. The produced radical species react with organics and eliminate biofouling [39]. The specific band gaps of the UF membrane lie at 2.5 and 3.1 eV, due to the mixture of TiO2-ZrO2, as described in previous work [28]. Figure 6. Filtration in the dark with self-cleaning under illumination series of UWWTP effluent spiked with 100 g/L of each MC. Note: , filtration of 1 L, followed by cleaning (60 min); , after cleaning, filtration of 2 L, followed by cleaning (60 min); '' after cleaning, filtration again.

Solar Photo-Fenton Treatment of Membrane Streams
The effect of UWWTP effluent filtration with the photocatalytic membrane was also evaluated in later photo-Fenton treatment of both the concentrate and permeate volumes, which still contained the initial concentration of MCs, as they were not retained by the membrane. The advantage of the photocatalytic membrane lies in pretreatment of UWWTP effluents because it significantly lowers the turbidity and particle concentration, thereby improving the availability of photons for the solar photo-Fenton process. Experiments were carried out using 0.1 mM Fe:EDDS 1:1 and 1.5 mM H2O2 as the Figure 6. Filtration in the dark with self-cleaning under illumination series of UWWTP effluent spiked with 100 g/L of each MC. Note: , filtration of 1 L, followed by cleaning (60 min); , after cleaning, filtration of 2 L, followed by cleaning (60 min); '2' after cleaning, filtration again.
The self-cleaning mechanisms of the TiO 2 -ZrO 2 UF membrane are not only based on the previously mentioned super-hydrophilicity. Other mechanisms that prevent fouling of different compounds on the membrane surface can be assigned to the formation of different radical species ( • OH and other) when the photocatalytic surface material is irradiated with a photon energy equal or higher than the band gap energy of the semiconductor. The produced radical species react with organics and eliminate biofouling [39]. The specific band gaps of the UF membrane lie at 2.5 and 3.1 eV, due to the mixture of TiO 2 -ZrO 2 , as described in previous work [28].

Solar Photo-Fenton Treatment of Membrane Streams
The effect of UWWTP effluent filtration with the photocatalytic membrane was also evaluated in later photo-Fenton treatment of both the concentrate and permeate volumes, which still contained the initial concentration of MCs, as they were not retained by the membrane. The advantage of the photocatalytic membrane lies in pretreatment of UWWTP effluents because it significantly lowers the turbidity and particle concentration, thereby improving the availability of photons for the solar photo-Fenton process. Experiments were carried out using 0.1 mM Fe:EDDS 1:1 and 1.5 mM H 2 O 2 as the oxidizing agents ( Figure 7). These had been found to be the optimal settings for solar photo-Fenton testing in a previous study [40].
Catalysts 2022, 12, x FOR PEER REVIEW 9 of 15 oxidizing agents (Figure 7). These had been found to be the optimal settings for solar photo-Fenton testing in a previous study [40]. When the solar photo-Fenton treatment was applied to the concentrate at its natural bicarbonate concentration (720 mg/L), 16% CAF, 12% IMI, 9% THI and 27% CBZ were eliminated. DCF was already degraded by photolysis during the filtration step, or at least to below the detection limit. A total MC elimination rate of 17% was recorded, along with an H2O2 consumption rate of 12.5 mg/L. All dissolved iron had precipitated within 15 min, as it is well-known that the Fe-EDDS complex degrades more rapidly in the presence of particulate organic matter. The same process applied to the permeate-volume-produced MC elimination of 32% CAF, 22% IMI, 19% THI, 42% CBZ and 81% DCF, resulting in a total MC elimination rate of 32%. In this case, H2O2 consumption equaled 18 mg/L at the end of the experiment. The dissolved iron concentration at 15 min was higher during treatment of the permeate volume, with 22% remaining, as the Fe-EDDS complex was degraded less rapidly due to the presence of less particulate organic matter, which was removed by the self-cleaning membrane. This demonstrated that lowering the turbidity and particle concentration during filtration using self-cleaning membranes had a beneficial effect on the following photo-oxidation treatment. However, the elimination of MCs was low in both cases.
As bicarbonates are well known radical scavengers, the solar photo-Fenton process was prevented from attaining its full potential when they were present at high concentrations in the concentrate and permeate volumes, explaining the low MC elimination rates observed. This has also been observed in other studies in the literature [41,42]. Therefore, experiments were performed on the same concentrate and permeate volumes as in Figure 7, but this time after removing bicarbonates (see Figure 8). The concentrate and permeate volumes were air-stripped using H2SO4 to lower their initial bicarbonate concentration of 720 mg/L to 75 mg/L. During this process, the pH dropped from pH 8 to circumneutral. When the solar photo-Fenton treatment was applied to the concentrate at its natural bicarbonate concentration (720 mg/L), 16% CAF, 12% IMI, 9% THI and 27% CBZ were eliminated. DCF was already degraded by photolysis during the filtration step, or at least to below the detection limit. A total MC elimination rate of 17% was recorded, along with an H 2 O 2 consumption rate of 12.5 mg/L. All dissolved iron had precipitated within 15 min, as it is well-known that the Fe-EDDS complex degrades more rapidly in the presence of particulate organic matter. The same process applied to the permeate-volume-produced MC elimination of 32% CAF, 22% IMI, 19% THI, 42% CBZ and 81% DCF, resulting in a total MC elimination rate of 32%. In this case, H 2 O 2 consumption equaled 18 mg/L at the end of the experiment. The dissolved iron concentration at 15 min was higher during treatment of the permeate volume, with 22% remaining, as the Fe-EDDS complex was degraded less rapidly due to the presence of less particulate organic matter, which was removed by the self-cleaning membrane. This demonstrated that lowering the turbidity and particle concentration during filtration using self-cleaning membranes had a beneficial effect on the following photo-oxidation treatment. However, the elimination of MCs was low in both cases.
As bicarbonates are well known radical scavengers, the solar photo-Fenton process was prevented from attaining its full potential when they were present at high concentrations in the concentrate and permeate volumes, explaining the low MC elimination rates observed. This has also been observed in other studies in the literature [41,42]. Therefore, experiments were performed on the same concentrate and permeate volumes as in Figure 7, but this time after removing bicarbonates (see Figure 8). The concentrate and permeate volumes were air-stripped using H 2 SO 4 to lower their initial bicarbonate concentration of 720 mg/L to 75 mg/L. During this process, the pH dropped from pH 8 to circumneutral. This time the concentrate degradation efficiency was much higher, showing MC degradation of 66% CAF, 52% IMI, 45% THI and 83% CBZ. Meanwhile, DCF, as previously, was degraded by photolysis during the filtration step, or at least to below the detection limit. The total MC elimination rate was 64%, more than three times higher than with the natural concentration of bicarbonates. In addition, the H2O2 consumption was higher (30 mg/L), showing that dissolved iron disappeared more slowly (better stability of the Fe-EDDS complex), with 80% of iron consumed after 15 min, which was similar to the permeate volume treatment, containing its natural concentration of bicarbonates. Lower bicarbonate concentrations trigger fewer • OH radical scavenging effects as part of the solar photo-Fenton cycle at circumneutral pH (Equations (1)-(3)) [43]. This interrupts these cycles less and regenerates the Fe(III)-EDDS complex from Fe(II)-EDDS, which is formed after quick photo degradation of the Fe(III)-EDDS complex in the first few minutes of the experiment after reacting with H2O2, whereas the Fe(II) also reacts with H2O2 to form Fe(III) and • OH (Equation (4)) [44], explaining the higher H2O2 consumption rates.
The degradation efficiency was also higher in the permeate volume, eliminating 60% CAF, 46% IMI, 39% THI, 78% CBZ and 78% DCF in 15 min. Total elimination was 59%, double that in the treatment with the natural bicarbonate concentration. This time the H2O2 consumption was almost three times lower, at 12.7 mg/L, attaining an iron concentration similar to the concentrate volume in 15 min, with 20% left. The lower H2O2 consumption can be attributed to the lower concentration of (natural) organic matter present, mainly due the lower turbidity (Section 3.1) and MC concentration profiles. Although both concentrate and permeate volumes had similar total concentrations of MCs (450 mg/L), their relative concentrations were very different, e.g., the presence and This time the concentrate degradation efficiency was much higher, showing MC degradation of 66% CAF, 52% IMI, 45% THI and 83% CBZ. Meanwhile, DCF, as previously, was degraded by photolysis during the filtration step, or at least to below the detection limit. The total MC elimination rate was 64%, more than three times higher than with the natural concentration of bicarbonates. In addition, the H 2 O 2 consumption was higher (30 mg/L), showing that dissolved iron disappeared more slowly (better stability of the Fe-EDDS complex), with 80% of iron consumed after 15 min, which was similar to the permeate volume treatment, containing its natural concentration of bicarbonates. Lower bicarbonate concentrations trigger fewer • OH radical scavenging effects as part of the solar photo-Fenton cycle at circumneutral pH (Equations (1)-(3)) [43]. This interrupts these cycles less and regenerates the Fe(III)-EDDS complex from Fe(II)-EDDS, which is formed after quick photo degradation of the Fe(III)-EDDS complex in the first few minutes of the experiment after reacting with H 2 O 2 , whereas the Fe(II) also reacts with H 2 O 2 to form Fe(III) and • OH (Equation (4) The degradation efficiency was also higher in the permeate volume, eliminating 60% CAF, 46% IMI, 39% THI, 78% CBZ and 78% DCF in 15 min. Total elimination was 59%, double that in the treatment with the natural bicarbonate concentration. This time the H 2 O 2 consumption was almost three times lower, at 12.7 mg/L, attaining an iron concentration similar to the concentrate volume in 15 min, with 20% left. The lower H 2 O 2 consumption can be attributed to the lower concentration of (natural) organic matter present, mainly due the lower turbidity (Section 3.1) and MC concentration profiles. Although both concentrate and permeate volumes had similar total concentrations of MCs (450 mg/L), their relative concentrations were very different, e.g., the presence and absence of DCF. Along with differences in the type, size and morphology of the organic matter, less H 2 O 2 reacted with this particulate matter, instead of being mainly consumed by the solar photo-Fenton reaction, producing • OH without the interference of the effects of radical scavenging by bicarbonates, as the concentration was equal in both concentrate and permeate volumes. The significantly lower H 2 O 2 consumption at a similar MC degradation rate shows the economic advantages of a pretreatment with the photocatalytic UF membrane. An attempt was also made to use persulfate [45] as an oxidizing agent instead of H 2 O 2 , at the same molar concentration, although it was found to be ineffective for the degradation of the selected MCs in the UWWTP effluent. Only 19.6% of the MCs were degraded (results not shown) compared to 64% under conventional solar photo-Fenton treatment.

Retention of P. Aeruginosa
The retention of P. Aeruginosa was also tested to assess future applications of the TiO 2 -ZrO 2 UF membrane. Figure 9 shows the results of filtering natural water (from Tabernas, Spain; see Table S5 for physicochemical characterization) with a starting concentration of 1 × 10 6 CFU/mL. The membrane showed a stable flow rate of 31 mL/min over a 150 cm 2 surface. Samples were taken and prepared, as described in Section 2.3.2, and P. Aeruginosa CFU values were counted and calculated after 48 h of cultivation at 36 • C. Samples of the permeate volume at t = 0 min contained no P. Aeruginosa, or at least it was under the detection limit. Two replicates, Rep1 and Rep2, were run. absence of DCF. Along with differences in the type, size and morphology of the organic matter, less H2O2 reacted with this particulate matter, instead of being mainly consumed by the solar photo-Fenton reaction, producing • OH without the interference of the effects of radical scavenging by bicarbonates, as the concentration was equal in both concentrate and permeate volumes. The significantly lower H2O2 consumption at a similar MC degradation rate shows the economic advantages of a pretreatment with the photocatalytic UF membrane. An attempt was also made to use persulfate [45] as an oxidizing agent instead of H2O2, at the same molar concentration, although it was found to be ineffective for the degradation of the selected MCs in the UWWTP effluent. Only 19.6% of the MCs were degraded (results not shown) compared to 64% under conventional solar photo-Fenton treatment.

Retention of P. Aeruginosa
The retention of P. Aeruginosa was also tested to assess future applications of the TiO2-ZrO2 UF membrane. Figure 9 shows the results of filtering natural water (from Tabernas, Spain; see Table S5 for physicochemical characterization) with a starting concentration of 1 × 10 6 CFU/mL. The membrane showed a stable flow rate of 31 mL/min over a 150 cm 2 surface. Samples were taken and prepared, as described in Section 2.3.2, and P. Aeruginosa CFU values were counted and calculated after 48 h of cultivation at 36 °C. Samples of the permeate volume at t = 0 min contained no P. Aeruginosa, or at least it was under the detection limit. Two replicates, Rep1 and Rep2, were run. The results in Figure 9 show that the membrane allowed the consistent retention of P. Aeruginosa of an order of magnitude of 1 × 10 3 -1 × 10 4 CFU/mL within 90 min of filtration during Rep1 (permeate). Although the concentration of P. Aeruginosa in the permeate volume significantly decreased compared to the concentrate volume, lower values were expected as a function of the photocatalytic UF membrane pore size, the zirconia intermediate layer (60 nm) and the size of the rod-like P. Aeruginosa, which has nominal size range of 500-800 nm by 1500-3000 nm [28,46,47]. The lower-than-expected retention rate of P. Aeruginosa may have been due to minor damages in the photocatalytic TiO2 membrane surface due to previous manual cleaning, as can also be observed in Figure 4b, and minor defects in the intermediate zirconia layer, making it possible for P. Aeruginosa to permeate the intermediate zirconia layer. However, the decreasing trend in Rep2 (permeate) in Figure 9 shows the effects (irreversible) of fouling at these sites with minor The results in Figure 9 show that the membrane allowed the consistent retention of P. Aeruginosa of an order of magnitude of 1 × 10 3 -1 × 10 4 CFU/mL within 90 min of filtration during Rep1 (permeate). Although the concentration of P. Aeruginosa in the permeate volume significantly decreased compared to the concentrate volume, lower values were expected as a function of the photocatalytic UF membrane pore size, the zirconia intermediate layer (60 nm) and the size of the rod-like P. Aeruginosa, which has nominal size range of 500-800 nm by 1500-3000 nm [28,46,47]. The lower-than-expected retention rate of P. Aeruginosa may have been due to minor damages in the photocatalytic TiO 2 membrane surface due to previous manual cleaning, as can also be observed in Figure 4b, and minor defects in the intermediate zirconia layer, making it possible for P. Aeruginosa to permeate the intermediate zirconia layer. However, the decreasing trend in Rep2 (permeate) in Figure 9 shows the effects (irreversible) of fouling at these sites with minor damage. A continuous fouling mechanism that can also be seen in Figure 6 resulted in a satisfactory P. Aeruginosa concentration in the permeate volume when operating for more than 180 min. Certain controlled fouling mechanisms are commonly applied in ceramic membrane production and industry as a pretreatment for similar reasons.
An irradiated membrane surface would prevent the adhesion of bacteria through the abovementioned hydrophilicity, as well as degradation of any preliminary extracellular polymeric substances (EPS) produced by the bacteria to bind themselves to surfaces and for cohesion promoting protection from unfavorable conditions. Thus, the membrane would show less fouling and longer operating times. These substances can normally be observed as a slime-like biofilm layer [48].

Conclusions
The TiO 2 -ZrO 2 ultrafiltration membrane proved to have self-cleaning capabilities when irradiated by simulated solar light after UWWTP effluents were filtered. This reduced the turbidity in the permeate by a factor of 20-50 NTU. It has been demonstrated that filtration of a certain volume of effluent in the dark is possible only up to a certain amount of fouling, then the original conditions can be recovered after illumination. However, after filtering a larger volume of effluent, fouling could become irreversible. Furthermore, the irradiated membrane showed lower fouling rates during the filtration of UWWTP effluents when irradiated. This was mainly attributed to the properties of irradiated TiO 2 , including the formation of • OH, which is able to degrade microorganisms as well as microcontaminants. The TiO 2 -ZrO 2 membrane significantly reduces the turbidity of the UWWTP effluent, which in turn significantly increases the degradation efficiency of the subsequent solar photo-Fenton treatment. Lowering the bicarbonate concentrations in both the concentrate and permeate resulted in higher efficiency of the solar photo-Fenton process.
The membrane shows retention of P. Aeruginosa to an order of magnitude of 1 × 10 3 -1 × 10 4 CFU/mL. The positive effects on retention of fouling of the larger pores and minor defects can be seen when repeating short term filtration experiments. Longer filtration with the TiO 2 -Zirconia ultrafiltration membrane could have positive effects on the retention of microorganisms, which would then be reversible after illumination.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/catal12050552/s1, Table S1: Composition of Urban Wastewater Treatment Plant Effluent; Table S2: Schematic overview of molecular structures of selected MCs; Table S3: Molecular weight, UPLC column retention time, LOQ and maximum contaminant absorption; Table S4: Overview of oxidation treatments and their parameters; Table S5: Composition of natural water (Tabernas, Spain).