**Fast Microwave Synthesis of Gold-Doped TiO2 Assisted by Modified Cyclodextrins for Photocatalytic Degradation of Dye and Hydrogen Production**

**Cécile Machut 1,\*, Nicolas Kania 1, Bastien Léger 1, Frédéric Wyrwalski 1, Sébastien Noël 1, Ahmed Addad 2, Eric Monflier <sup>1</sup> and Anne Ponchel <sup>1</sup>**


Received: 29 June 2020; Accepted: 16 July 2020; Published: 18 July 2020

**Abstract:** A convenient and fast microwave synthesis of gold-doped titanium dioxide materials was developed with the aid of commercially available and common cyclodextrin derivatives, acting both as reducing and stabilizing agents. Anatase titanium oxide was synthesized from titanium chloride by microwave heating without calcination. Then, the resulting titanium oxide was decorated by gold nanoparticles thanks to a microwave-assisted reduction of HAuCl4 by cyclodextrin in alkaline conditions. The materials were fully characterized by UV-Vis spectroscopy, X-Ray Diffraction (XRD), Transmission Electron Microscopy (TEM), and N2 adsorption-desorption measurements, while the metal content was determined by Inductively Coupled Plasma Optical Emission Spectroscopy (ICP-OES). The efficiency of the TiO2@Au materials was evaluated with respect to two different photocatalytic reactions, such as dye degradation and hydrogen evolution from water.

**Keywords:** photocatalysis; photodegradation; nanoparticles; gold; TiO2; cyclodextrins

#### **1. Introduction**

During the past decades, photocatalysis received extensive research interest for both limiting toxic wastes and developing clean and renewable sources of energy. Indeed, the association of a semiconductor with the sunlight in order to remove pollutants [1,2] or to produce hydrogen fuel by water splitting [3] could provide a sustainable solution to the crucial problems of environmental pollution and energy shortages [4].

Among a large number of photocatalysts, TiO2 has been extensively investigated due to its good properties such as low cost, non-toxicity, and good stability [5,6]. Anatase phase is particularly recognized for its high photocatalytic efficiency. However, its large band gap (3.2 eV) combined with a high recombination rate of the photogenerated electron/hole pairs (e−/h+) reduce the photon-to-charge carriers conversion efficiency, but also limit the use in photochemical applications under visible or solar light.

In order to overcome these drawbacks and improve the photocatalytic performance of semiconductors, one of the promising strategies consists in introducing noble metals at the surface of TiO2, such as gold nanoparticles [7]. Indeed, the combination with gold nanoparticles aims at inhibiting the electron-hole pair recombination by trapping electrons and facilitating the transfer of holes on the TiO2 surface [8]. Gold nanoparticles (Au NPs) are also known to enhance the activity

of TiO2 under visible-light irradiation due to the localized surface plasmon resonance of Au NPs in the visible light spectrum [9]. However, it is well accepted that the photocatalytic activity of such TiO2@Au composites can strongly depend on the particle size of Au NPs [10] and optimal synthetic conditions must be found, especially to prevent the aggregation of gold nanoparticles [11,12].

Over the last decade, cyclodextrins and derivatives have received great interest in the field of synthesis and stabilization of metallic nanoparticles in aqueous medium [13]. These macrocyclic oligosaccharides, which are well-known to form inclusion complexes with numerous guest molecules via supramolecular interactions [14], can also be used as capping agents to stabilize zerovalent metal nanoparticles, such as Au NPs. Owing to the numerous hydroxyl groups attached to the CD rims, they can also act as efficient reducing agents for the synthesis of Au NPs [15–19]. However, to the best of our knowledge, the use of cyclodextrins to prepare TiO2@Au composites through simple methods of synthesis has been scarcely investigated. Most synthetic routes involve the use of chemically modified cyclodextrins bearing thiol pendant groups as metal binding sites. Their preparations require multistep and complex synthetic procedure as well as the use of time-consuming purification methods. For instance, Zhu et al. developed a method to synthesize TiO2 decorated by the assembly of per-6-thio-β-cyclodextrin and gold nanoparticles. The resulting composite showed very good efficiency for the degradation of methyl orange (MO) under UV light [20]. More recently, TiO2 nanosheets consisting of the combination of Au nanoparticles and mono-6-thio-β-cyclodextrin were prepared for the electrochemical detection of trace of methyl parathion pesticide [21].

Recently, we have reported a sol-gel method using cyclodextrins as both structure-directing agents and metal-complexing agents to self-assemble titania and gold colloids in composite materials with controlled porosity and uniform metal dispersion [22]. Among the various cyclodextrins examined, the TiO2@Au material prepared using the commonly used randomly methylated β-CD (RAME-β-CD) have shown, after calcination, the best catalytic performance for the photodegradation of organic pollutants in water under visible light, due to a good compromise between its textural properties, crystallinity, and Au particle size. However, the preparation of such plasmonic photocatalysts involved a multistep process that occurred over several days (including acid hydrolysis, peptization, maturation, drying, and finally calcination at a high temperature of 500 ◦C to form Au NPs).

In recent years, microwave (MW) irradiation techniques have received considerable attention in the field of nanomaterial synthesis by inducing or enhancing chemical reactions [23–25]. The use of microwave heating may offer several advantages over conventional heating, such as shorter reaction times, higher heating rates as well as higher uniformities of the products. In the literature, a few articles were already devoted to microwave-assisted synthesis of gold nanoparticles protected by cyclodextrin derivatives [16,26,27]. As a matter of fact, Aswathy et al. synthesized β-cyclodextrin capped Au NPs with a mean diameter of 20 nm within a few minutes [16]. More recently, Stiufiuc et al. used native cyclodextrins as reducing and capping agents during the microwave reduction of the gold precursor and obtained stable monodispersed gold nanospheres covered with either α-, β- or γ-CD [26]. However, to the best of our knowledge, the stabilization and anchorage of Au NPs on titania support thanks to CDs under microwave irradiation have never been explored.

In this context, we reported hereby a novel method for elaborating TiO2@Au materials from a two-step microwave-assisted synthetic route without the need for high temperature calcination. Herein, the TiO2 support is synthesized using a microwave-method by hydrolysis of titanium tetrachloride while the cyclodextrins are employed afterwards to produce size-controlled gold metallic nanoparticles anchored on the support, once again under microwave irradiation. We have focused our efforts on randomly methylated-β-CD (RAME-β-CD) and 2-hydroxypropyl-β-CD (HP-β-CD), which are both highly water-soluble and readily available commercially at relatively low cost. The impact of the nature of the carbohydrate precursor is investigated and discussed on the basis of different physicochemical characterizations, including X-ray diffraction (XRD), N2 adsorption-desorption analysis, transmission electron microscopy (TEM), thermogravimetry analyses (TGA), and diffuse reflectance UV-Vis spectroscopy (DRUV-Vis). Finally, the efficiency of these photocatalysts is examined

with respect to two photocatalytic reactions carried out under near-UV-light irradiation (λ > 365 nm), i.e., the oxidative photodegradation of methyl orange and the hydrogen evolution reaction (HER).

#### **2. Results and Discussion**

As described in the Experimental Section, gold-doped TiO2 materials have been synthesized at 150 ◦C with a fast microwave heating using cyclodextrins as reducing agent of the metal precursor and stabilizer of Au NPs. The synthetic procedure is schematically depicted in the Figure 1.

**Figure 1.** Schematic illustration of the two-step microwave (MW) procedure used for the TiO2@Au materials synthesis.

Note that a bare TiO2 control was also prepared in the same conditions as those described for the first step. These conditions were selected based on preliminary experiments, by varying the duration and power of the microwave irradiation, in order to optimize the crystallinity of our titania support. Indeed, the crystallinity is known to be a key factor in the photoactivity of TiO2 particles. The XRD patterns of titania materials prepared from different heating programs (10, 30, and 45 min) and powers (320 and 600 W) are given for comparison in the Figure 2. With increasing the duration of heating at 320 W, we observe that the intensity of the XRD lines progressively increases and narrows, suggesting a growth in crystallite size. The planes (101), (004), (200), (105), and (211) associated to 2θ = 25.3◦, 37.7◦, 48◦, and 55.2–55.9◦ respectively correspond to the anatase phase (Ti-A, JCPDS 21-1272). No XRD signals related to the presence of other crystalline phases such as rutile and brookite are detected. However, the most interesting effects are produced with the power of 600 W, which offers a very good compromise between crystallinity state and rate of anatase formation since this crystalline phase was obtained after only 10 min, this duration being considerably shorter than that applied for conventional sol-gel synthesis [28]. In line with this first optimization, the heating power of microwave irradiation was set to maximum (600 W) for all the further investigations, with a duration of temperature rise of 2 min from room temperature to 150 ◦C (isothermal step-time of 10 min).

The impact of the addition of gold by microwave-assisted reduction of the TiO2 support was further investigated using mixtures of HAuCl4 and modified cyclodextrins in alkaline conditions (see Figure 1, second step). We decided to use the randomly methylated β-cyclodextrin (RAME-β-CD) and the hydroxypropylated β-cyclodextrin (HP-β-CD) to stabilize Au NPs. Indeed, we particularly focused on these two CDs because of their high solubility in water and their beneficial effect on previously described gold-doped TiO2 [22]. RAME-β-CD and HP-β-CD have also the advantages to

offer a number of available hydroxyl groups (8.4 per RAME-β-CD and 21 per HP-β-CD), which are known to play an important role in the reduction processes of metal cations [16,29].

The XRD patterns of these microwave-prepared titania@Au materials are reported in the Figure 3. For comparison, the XRD pattern of a control gold-doped TiO2 prepared in ethanol (selected as model reducing agent), but without cyclodextrin, was also included (TiO2@Au).

It can be noticed, that in addition to the reflections of anatase (Ti-A, JCPDS 21-1272), TiO2@Au, TiO2@Au-RB, and TiO2@Au-HP present broad and low intense peaks at 2θ = 38.2◦, 44.2◦, 64.3◦, and 78.1◦, which could be respectively indexed to the (111), (200), (220), and (311) planes of gold with face-centered cubic crystalline structure phase (JCPDS 04-0784). The Au crystallite sizes could have been estimated from the line broadening of the (200) diffraction peak at 2θ = 44.2◦ by the Debye-Sherrer equation. Interestingly, for the control-doped TiO2 material prepared using ethanol, the size of gold crystallites is ca. 15 nm, while it significantly decreases to ca. 8–10 nm for the materials prepared with HP-β-CD and RAME-β-CD.

The textural characteristics of the titania-based materials were then evaluated by N2 adsorption-desorption analysis. All the samples exhibit type IV adsorption isotherms with distinct hysteresis loops appearing at P/P◦ ≈ 0.5–0.8, thus supporting the mesoporous character of the samples with a monomodal pore size of 4 nm (Figure S1 ESI and Table 1). The specific surface areas of the bare TiO2 (TiO2-control) is close to those prepared by gold-doped TiO2 (240–260 m2.g−1). In the same way, the pore volumes and pore size values are substantially the same whether there is gold or not. The detailed surface properties and the gold loading determined by Inductively Coupled Plasma Optical Emission Spectroscopy (ICP-OES) measurements are summarized in the Table 1. ICP-OES analysis was used to quantitatively determine the gold content in our composites. Interestingly, the gold loading (≈2 wt %) corresponds to a gold incorporation efficiency around 80–90% of the initial amount of metal used during the synthesis.

**Figure 2.** X-Ray Diffraction (XRD) patterns of titania-based materials prepared by microwave heating with different programs of heating (**a**) 320 W 3 min ramp then 10 min at 150 ◦C, (**b**) 320 W 3 min ramp then 30 min at 150 ◦C, (**c**) 320 W 3 min ramp then 45 min at 150 ◦C, and (**d**) 600 W 2 min ramp then 10 min at 150 ◦C.

**Figure 3.** XRD patterns of titania materials prepared by microwave heating: (**a**) bare TiO2, (**b**) TiO2@Au prepared with ethanol, (**c**) TiO2@Au-RB, and (**d**) TiO2@Au-HP.


**Table 1.** Surface properties and gold loading of the different TiO2 materials.

<sup>a</sup> Specific surface area determined by the BET (Brunauer, Emmett et Teller) method in the relative pressure range of 0.1−0.25. <sup>b</sup> Pore volume computed by BJH. <sup>c</sup> Pore size determined by BJH. <sup>d</sup> Gold loading determined by ICP-OES analysis.

The morphology and structure of the TiO2@Au materials were further characterized by TEM analyses (Figure 4). Whatever the materials, the presence of Au NPs deposited onto the surface of TiO2 is observed. When the synthesis is performed under cyclodextrin-free conditions, with ethanol as reducing agent, TEM images (Figure 4a,b) show the presence of gold nanoparticles with a mean diameter of 13.5 nm but with a relatively broad size distribution ranging from 5 to 30 nm and a standard deviation of 5.3 nm (see histogram in Figure 4c). Note that larger gold nanoparticles with diameter ranging from 44 to 78 nm can be also observed (See Figure S2 in ESI).

Although a modest decrease in the mean particle size is noticed when modified β-cyclodextrins (12.5 nm for HP-β-CD and 12.9 nm for RAME-β-CD) are introduced during the microwave-assisted synthesis, it can be seen that, for these two catalysts, gold nanoparticles are more uniformly dispersed over the TiO2 support. Narrower size distributions with standard deviations as low as 2.5–2.8 nm (see histograms in Figure 4f,j) can be clearly found, evidencing the stabilization of small and well-dispersed spherical Au NPs, as can be seen at high magnification (See Figure S3 in ESI). It provides an intimate contact between Au NPs and the TiO2 mesoporous support. Conversely to what was observed with the ethanol procedure, no aggregation or formation of larger particles were observed over TiO2@Au-HP and TiO2@Au-RB.

**Figure 4.** Transmission electron microscopy (TEM) images at magnification of ×25,000 (Scale bar = 100 nm) and ×62,000 (Scale bar = 50 nm) and size distribution of (**a**–**c**) TiO2@Au, (**d**–**f**) TiO2@Au-RB, (**h**–**j**) TiO2@Au-HP.

As previously observed by several teams, cyclodextrins can stabilize metallic nanoparticles in aqueous solution [13]. Because of different types of interactions between the metal and the CDs (hydrophobic-hydrophobic interactions [30], non-covalent interactions between metal ions and hydroxyl groups of the CD [15]) the aggregation of gold nanoparticles can be avoided and it will result in a smaller particle size. As already observed with native CDs, RAME-β-CD and HP-β-CD are able to reduce Au3<sup>+</sup> thanks to their hydroxyl groups and then interact with the gold nanoparticles in order to prevent their agglomeration [31]. To the best of our knowledge, it is the first time that modified cyclodextrins are employed as both reducing agent of gold precursor and also stabilizing agent of gold nanoparticles.

Our materials were then characterized by UV-visible diffuse reflectance spectroscopy experiment. Figure 5 shows UV–Visible absorbance spectra and Tauc plots of TiO2-control, TiO2@Au, TiO2@Au-RB and TiO2@Au-HP materials. All the titania samples exhibit a broad absorption band around 330 nm corresponding to the charge transfer from O 2p valence band to Ti 3d conduction band [32]. Thus, the large band gap energy (Eg) of 3.20 eV estimated for the unmodified TiO2 is in agreement with typical values reported in the literature for anatase structures. However, it is worth noting that, for the gold-doped TiO2 samples prepared from cyclodextrins, a slight red-shift of the absorption edge of the TiO2 semiconductor toward higher wavelengths was observed compared to pure TiO2. The following sequence can be established in terms of Eg: TiO2@Au-RB (2.70 eV) < TiO2@Au-HP (2.95 eV) < TiO2@Au (3.20 eV) = TiO2 (3.20 eV).

As previously reported, the electrons can be transferred from the excited TiO2 to the metallic nanoparticles and the electron accumulation increases the Fermi level of the nanoparticle to more negative potentials. Therefore, the involved edge energy in the electron transfer from TiO2 to the metallic nanoparticles is lower than bare TiO2 [33]. The lowest band gap values for the TiO2@Au-RB and the TiO2@Au-HP materials suggest that the contact between the two inorganic phases (gold and TiO2) is enhanced when cyclodextrin is used during the Au NPs synthesis and this result is in good agreement with the TEM observations. However, the smallest value was found for the TiO2@Au-RB so that we can suppose that the use of the RAME-β-CD promotes the most intimate contact between the semiconductor and the metal. Further, another band is revealed at approximately 550 nm, confirming the presence of gold particles embedded in the TiO2 matrix [34]. When neither cyclodextrin nor ethanol is added to the gold salt in the second step, no reduction of Au3<sup>+</sup> was noticed, the resulting powder remained white and its UV-Vis spectra was similar to that obtained for the bare TiO2 (see Figure S4 in ESI).

**Figure 5.** Diffuse Reflectance UV-Vis (DRUV-Vis) spectra of titania-based materials prepared by microwave heating: (**a**) bare TiO2, (**b**) TiO2@Au, (**c**) TiO2@Au-HP, (**c**) and (**d**) TiO2@Au-RB. In the inset, Tauc plots for the determination of the band gap values Tauc (indirect bad gap energy).

UV-Vis experiment and TEM images proved that modified CDs can act as both reducing agent of the metal precursor and capping agent of well-dispersed homogeneously dispersed Au NPs even in the presence of titanium dioxide. But to further characterize our materials and specially to know if cyclodextrins still remained in the TiO2@Au-RB and the TiO2@Au-HP samples, thermogravimetric analyses (TGA) were performed. The thermal profiles of TiO2@Au, TiO2@Au-RB, and TiO2@Au-HP are shown in Figure 6.

The thermal patterns of the bare TiO2 and the TiO2@Au exhibit a one-step decomposition process with a weight loss in the 50–400 ◦C temperature range corresponding to the desorption of physically adsorbed water. The total weight loss for these samples are estimated to be 6.0 and 6.7 wt.%, respectively. The thermal profile of TiO2@Au-RB exhibits a two-step decomposition process with a total weight loss of ca. 10.4 % at 1000 ◦C. The first weight loss (≈4%) in the 50–250 ◦C temperature range corresponds to the removal of physically adsorbed water, whereas the second weight loss (≈6%) in the 250–450 ◦C temperature range with a major weight loss at ca. 380 ◦C attributed to the thermal decomposition of the modified β-CD (Figure S5 in ESI). A similar profile was obtained with the TiO2@Au-HP (Figure 6c) since this sample exhibited also a two-step decomposition process attributed to the removal of physically adsorbed water (≈6%) and to the thermal decomposition of residual HP-β-CD or its residues (≈11%) (see Figure S5 in ESI for the thermal profile of HP-β-CD alone). These thermal analyses proved that a small amount of saccharidic compounds (≈6 wt.% for the TiO2@Au-RB and 11 wt.% for the TiO2@Au-HP) remains adsorbed on our composite materials prepared with modified CDs even after the washing cycles. This result could be explained by the ability of CD derivatives to interact both with the gold nanoparticles and with the titania support. As previously described, we can suppose that after the

microwave reduction, cyclodextrin derivatives could be linked to the gold nanoparticles through weak interactions and covered the outer surface of the Au NPs [16,26]. In addition, cyclodextrins are known to be able to interact with the titanium dioxide through hydrogen bounds [35,36]. In fact, the hydroxyl groups located at the exterior of the torus favored the interactions of the cyclic oligosaccharides with the surface OH groups of titania. This latter hypothesis could also explain why the amount of organic compounds is higher in the TiO2@Au-HP than in the TiO2@Au-RB composite: the quantity of saccharidic compounds adsorbed on titania increases with the number of hydroxyl groups of the CD [37]. Because of a higher number of hydroxyl groups (21 vs. 8.4), the HP-β-CD is more adsorbed on the titania support than the RAME-β-CD. This hypothesis could also explain the larger Eg observed for the TiO2@Au-HP compared to the TiO2@Au-RB composite: the residual organic compounds may reduce the contact between the Au NPs and the titanium dioxide [38].

**Figure 6.** Thermogravimetric profiles for (**a**) bare TiO2, (**b**) TiO2@Au, (**c**) TiO2@Au-RB, and (**d**) TiO2@Au-HP.

According to the textural and structural studies, our titania-based materials exhibited interesting characteristics for photocatalytic applications. Indeed, the catalytic efficiency is known to be linked to two major physical properties: crystallinity and surface area of the photocatalysts [39].

With this microwave synthesis, only the anatase crystalline phase was obtained at low temperature (150 ◦C) without any additional calcination (or another thermal treatment) and this phase is known for its good activity in photocatalysis. On the other hand, good textural properties in terms of specific surface area, pore size, and pore volume could facilitate adsorption and diffusion of the target molecules onto the surface of the catalyst [40].

To confirm these hypotheses, the photocatalytic performances of the microwave gold-doped TiO2 materials have been investigated through two different experiments. The redox properties of these materials have been firstly evaluated in the photodegradation of methyl orange (MO) in water. Briefly, an aqueous solution of MO (50 ppm) in the presence of the semi-conductors was irradiated at 365 nm and the concentration of the residual dye was regularly quantified by HPLC measurements. Prior to the photocatalytic study, the photostability of the organic dye was checked in a preliminary test without photocatalyst (Figure S6), and it was found that the concentration of MO remained unchanged during

the 1h test period. The performances of TiO2@Au-RB and TiO2@Au-HP are reported in the Figure 7. For comparison, TiO2@Au prepared with ethanol was also tested (Figure 7a).

**Figure 7.** Photocatalytic performances of the gold-doped titania materials prepared by microwave heating for the degradation of methyl orange in near UV (λ = 365 nm): (**a**) Conversion of methyl orange after 60 min of irradiation (**b**) Evolution of the methyl orange concentration during one hour of irradiation for TiO2@Au-RB (yellow) and TiO2@Au-HP (green).

After 60 min under near UV irradiation, the dye was hardly degraded in the presence of the TiO2@Au prepared without CD by microwave heating and this result is similar to thus obtained with bare TiO2 (Figure S6). In contrast, after one hour of irradiation, the MO concentration was close to zero for the tests realized with the gold-doped TiO2 prepared with modified cyclodextrins (TiO2@Au-RB and TiO2@Au-HP). The addition of modified CD during the synthesis of Au NPs in the presence of TiO2 improved drastically the performances of the photocatalyst and this result is probably linked to the good dispersion of nanosized gold nanoparticles obtained from CDs over the support. In fact, small and well-dispersed metal islands deposited on the TiO2 core are known to provide a favorable geometry for facilitating the interfacial charge transfer under UV irradiation [41]: the electrons of the titanium oxide are excited from the valence band to the conduction band and then migrate to Au clusters, which prevent the direct recombination of electrons and holes. For the TiO2@Au-RB and TiO2@Au-HP samples, we can suppose that the small and spherical Au NPs observed on the surface of the semi-conductor by TEM experiments act as electron sink to favor the oxidation and the reduction reactions. Conversely, large particles of metal are often harmful to the photocatalytic activity so that the TiO2@Au prepared with ethanol as reducing agent was less efficient in our conditions [42]. Logically, large nanoparticles mobilize more gold atoms than small ones. With an equal metal loading, materials doped with large Au NPs offer fewer electronic reservoirs than those with small particles.

Additionally, the Figure 7b showed that the decrease of the MO amount was significantly faster in the presence of the microwave-assisted gold-doped TiO2 prepared with RAME-β-CD compared to that prepared with HP-β-CD (Figure 7b). The lowest efficiency of the TiO2@Au-HP compared to the TiO2@Au-RB might be correlated to the highest band gap (as evidenced by DRUV-Vis experiment) and also to the amount of CDs residues in the final material (as evidenced by thermogravimetric analysis). In fact, we can suppose that the residual organic compounds decrease the contact between the semi-conductor and the gold and so reduce the electron transfer. Furthermore, the CDs residues could maybe mask some of the active sites of the semi-conductor or reduce the potential adsorption of the MO [43].

The recyclability and reuse of the most efficient photocatalyst (TiO2@Au-RB) was also evaluated in the degradation of the MO. From Figure S7, it can be seen that the photocatalytic activity is stable during at least 3 runs. This study clearly showed the robustness of the catalyst and the strong embedment of the Au NPs onto the TiO2 support.

Finally, we studied the behavior of the gold-doped TiO2 in the production of hydrogen by photoreduction of water. Aqueous suspensions of the TiO2@Au, TiO2@Au-HP, and TiO2@Au-RB were

irradiated at 365 nm in the presence of ethanol as the sacrificial agent. The result of the amount of hydrogen produced by photoreduction of water is reported in the Figure 8a.

When the TiO2@Au-HP and the TiO2@Au-RB were irradiated in water, hydrogen was quickly detected and the amount of H2 was quantified as about 160 and 300 μmol.h−1.g−<sup>1</sup> of catalyst, respectively. Compared to other TiO2 catalysts in the literature [2,44], these amounts of produced hydrogen are promising since the power of our lamp is very low in comparison to Xe lamp usually used in such photocatalytic experiments. Moreover, the yield of hydrogen produced with our gold-doped catalyst was very high in comparison with that obtained with commercial anatase TiO2 (<2 μmol.h<sup>−</sup>1.g−1) or with TiO2@Au prepared with ethanol as reducing agent in the same conditions (about 3 μmol.h<sup>−</sup>1.g−1). As observed with the first photocatalytic test, the TiO2@Au-RB was also more efficient than theTiO2@Au-HP to produce hydrogen from water, probably due to the same reasons discussed above (i.e., twice as many organic compounds on the surface of the photocatalyst for the TiO2@Au-HP than for the TiO2@Au-RB). Finally, the amount of hydrogen produced is reproducible after several cycles of illumination (see for example Figure 8b with TiO2@Au-RB) and stable during more than 10 h (ESI, Figure S8). This catalytic result proved that the introduction of small and uniform gold nanoparticles thanks to CDs reduction leads to a real boost of the photocatalytic performances of titanium dioxide even under UV irradiation and clearly confirmed the need of intimate contact with TiO2 and Au to enhance the electron transfer between them.

**Figure 8.** (**a**) Amount of hydrogen produced by photoreduction of water in the presence of gold-doped TiO2 prepared by microwave heating process (100 mg of photocatalyst, 80 mL water, 20 mL ethanol, λ = 365 nm) (**b**) Evolution of the hydrogen production by photoreduction of water in the presence of TiO2@Au-RB during 3 cycles of illumination.

#### **3. Materials and Methods**

#### *3.1. Chemicals*

Randomly methylated β-cyclodextrin (RAME-β-CD) with an average degree of substitution of 1.8 methyl groups per glucopyranose unit (MW 1310 g.mol−1) was a gift from Wacker Chemie GmbH (Lyon, France). Hydroxypropyl-β-cyclodextrin (HP-β-CD) with an average substitution of 0.6 CH2CH(OH)-CH3 groups per glucopyranose unit (MW 1380 g.mol<sup>−</sup>1) was purchased from Roquette (Lestrem, France). Ethanol, methyl orange (MO) and TiCl4 were purchased from Sigma-Aldrich (Quentin-Fallavier, France) while HAuCl4 (49 wt.%) was provided by Strem Chemicals (Bischheim, France). All these reagents were used without purification.

#### *3.2. Preparation of the Au*/*TiO2 Materials with Cyclodextrins*

In a typical preparation, TiO2 was prepared from TiCl4 by microwave heating (CEM Mars instrument, Power 600 W) inspired by a method previously described by Wang et al. [45]. TiCl4 (0.9 mL, 8.21 mmol) was quickly added to ethanol (25 mL) and stirred at room temperature during 10 min. Then, the yellow solution was introduced in a Teflon microwave reactor equipped with temperature and pressure probes and heated to 150 ◦C during 10 min. The white suspension was centrifuged at 3000 rpm during 5 min. The supernatant was evacuated and the resulting white powder of TiO2 was added to 20 mL of an aqueous solution of HAuCl4 (3.73 <sup>×</sup> 10−<sup>5</sup> mmol) and cyclodextrin (4.04 <sup>×</sup> <sup>10</sup>−<sup>4</sup> mmol). Note that this CD/Au molar ratio of about 10 has been chosen to promote the synthesis of spherical gold nanoparticles, in line with a previous work reported in the literature [46]. NaOH (0.5 M) was slowly added to the solid suspension in order to adjust the pH value at about 9. Then the mixture was transferred in a Teflon microwave reactor and was finally heated under microwave irradiation with the same program used to prepare TiO2 from TiCl4 (600 W, 150 ◦C, 10 min). To promote the synthesis of gold nanoparticles. At the end of the heating microwave program, the suspension was centrifuged and the purple powder was thoroughly washed with water before overnight drying at 100 ◦C. The gold-doped TiO2 materials synthesized by microwave heating from RAME-β-CD and HP-β-CD were named as TiO2@Au-RB and TiO2@Au-HP, respectively. Additionally, note that a control gold-doped TiO2 (denoted as TiO2@Au) was also prepared in a very similar manner as the above described procedure, by substituting cyclodextrin for ethanol during the reduction process. The syntheses and characterizations have been reproduced several times.

#### *3.3. Characterization Methods*

#### 3.3.1. Powder X-ray Diffraction

Powder X-ray diffraction data were collected on a Siemens D5000 X-ray diffractometer (Bruker, Palaiseau, France) in a Bragg-Brentano configuration with a Cu Kα radiation source. Scans were run over the angular domains 10◦ < 2θ < 80◦ with a step size of 0.02◦ and a counting time of 2 s/step. Crystalline phases were identified by comparing the experimental diffraction patterns to Joint Committee on Powder Diffraction Standards (JCPDS) files for anatase. The treatment of the diffractograms was performed using the FullProf software [47] and its graphical interface WinPlotr [48]. The average crystallite size D was calculated from the Scherrer formula, D = Kλ/(β cos θ), where K is the shape factor (a value of 0.9 was used in this study, considering that the particles are spherical), λ is the X-ray radiation wavelength (1.54056 Å for Cu K), β is the full width at half-maximum (fwhm), and θ is the Bragg angle.

#### 3.3.2. Nitrogen Adsorption-Desorption Isotherms

Nitrogen adsorption-desorption isotherms were collected at −196 ◦C using an adsorption analyzer Micromeritics Tristar 3020 (Merignac, France). Prior to analysis, 200–400 mg samples were outgassed at 100 ◦C overnight to remove the species adsorbed on the surface. From N2 sorption isotherms, specific surface areas were calculated by the BET method while pore size distributions were determined using the BJH model assuming a cylindrical pore structure. The relative errors were estimated to be the following: SBET, 5%; pore volume (pv) (BJH), 5%; pore size (ps) (BJH), 20%.

#### 3.3.3. Diffuse Reflectance UV-Visible

Diffuse reflectance UV-visible spectra were collected using a Shimadzu UV-Vis NIR spectrometer (Marne-la-Vallée, France). BaSO4 was used as the reference. Tauc plot analysis was performed for the calculation of the band gap energy (Eg). In fact, the Eg can be estimated by plotting (F(R) hν) <sup>n</sup> vs. hν and extrapolated from linear part of the curve to the hν x-axis intercept. To determine values of these forbidden energies, the absorption data were fitted to the Tauc relation for indirect band-gap transitions (*n* = <sup>1</sup> <sup>2</sup> ) [49].

#### 3.3.4. Thermogravimetric Analysis (TGA) Coupled with Differential Scanning Calorimetry (DSC)

Thermogravimetric Analysis (TGA) coupled with Differential Scanning Calorimetry (DSC) analyses were performed using a Mettler Toledo TGA/DSC3+ STARe system unit (Viroflay, France). The samples were placed in aluminum oxide crucibles of 70 μL and heated from 40 to 1000 ◦C at 10 ◦C.min−<sup>1</sup> under a 50 mL.min−<sup>1</sup> air flow.

#### 3.3.5. ICP Optical Emission Spectrometry

ICP optical emission spectrometry was performed on an iCAP 7000 Thermo Scientific spectrometer (Les Ulis, France). For the quantification of gold loading, 10 mg of the Au/TiO2 materials were introduced in 20 mL of aqua regia and then heated to 130 ◦C during one hour. Then the remaining TiO2 was removed using a 0.2 μm pore filter. The resulting solution is finally diluted with pure water up to a final volume of 100 mL. The amount of gold incorporated in the material was determined using an external calibration with a gold ICP standard solution.

#### 3.3.6. Transmission Electron Microscopy (TEM)

Transmission Electron Microscopy (TEM) bright field observations were performed on a Tecnai G2 microscope (FEI, Hillsboro, Oregon, USA) operating at an accelerating voltage of 200 kV. The Au/TiO2 powder was deposited directly on a carbon coated copper grid. Metal particle size distributions have been determined from the measurement of about 200 Au NPs. The nanoparticles were found in arbitrarily chosen area of the images using the program ImageJ software.

#### *3.4. Photocatalytic Experiments*

#### 3.4.1. Photodegradation of Methyl Orange

The photocatalytic efficiency of the titania-based materials was first evaluated in the photodegradation of methyl orange (MO) carried out using quartz reactors of 5 mL. In a typical experiment, 10 mg of photocatalyst was added to 4 mL of a solution of methyl orange (50 ppm). After 30 min in the dark, UV irradiation was performed using a led UV light lamp (Opsytec λ = 365 nm, beam size = 0.785 cm2, power of 0.2 W.cm−2). Aliquots were centrifuged at regular intervals and the MO concentration in the supernatant was determined by high-performance chromatography (HPLC, PerkinElmer, Villebon-sur-Yvette, France) analyses using a PerkinElmer Pecosphere C18 (83 mm length × 4.6 mm diameter) column. An aqueous mixture of acetonitrile (20% (*v*/*v*)) was used as the mobile phase at a flow rate of 1 mL.min−1. Aliquots of 50 μL of the sample was injected and analyzed using a photodiode array detector. The MO conversion given in percentage refers to the difference in the MO concentration before irradiation (C0) and after 1 h of irradiation (C) divided by the MO concentration before irradiation (i.e., 100 × (C0 − C)/C0).

#### 3.4.2. Production of Hydrogen by Photoreduction of Water

Photocatalytic measurements for H2 generation were carried out in a cylindrical pyrex reactor equipped with a quartz window by irradiating the titania-based materials in a 20 vol% ethanol-water solution (ethanol was used as hole-scavenger). As light source, we used the same LED UV light as that employed for the photocatalytic degradation of MO experiments described above in Section 3.4.1. The reactor operated at room temperature and atmospheric pressure and was kept under stirring at a constant speed of 1250 rpm. In a typical experiment, 100 mg of photocatalyst was added to a 100 mL of ethanol-water solution in the reactor. The catalytic solid suspension was then flushed with argon gas (420 mL.h<sup>−</sup>1) for 60 min prior to photocatalysis. The amount of H2 produced was measured on-line using a micro gas chromatograph (Micro-GC Agilent 490, Les Ulis, France) equipped with a thermal conductivity detector and two separating columns (Microsieve 10 m (5 Å) and 8 m-Paraplot U) operating with backflush injection (Ar as carrier gas).

#### **4. Conclusions**

In this work, an easy and fast preparation of Au loaded TiO2 without calcination step is described. The addition of common modified cyclodextrin (methylated or hydroxypropylated) during the

microwave reduction of a gold precursor in the presence of TiO2 led to an efficient photocatalyst both for pollutant photodegradation and photoreduction of water under near UV irradiation. The saccharidic macrocycle was responsible for a good stabilization of gold nanoparticles in aqueous solution so that these latter could not aggregate during the microwave synthesis and were deposited uniformly on the TiO2 surface. Because of its lowest number of hydroxyl groups, the RAME-β-CD seems to be less adsorbed onto the surface of the final composite after the gold reduction and represents the most promising photocatalyst. It could be now interesting to study the photocatalytic performances of our materials under solar simulated lamp. However, this new and fast synthetic approach offers promising perspectives for photocatalytic depollution process and green energy production.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4344/10/7/801/s1, Figure S1. N2 adsorption desorption isotherms of TiO2-control (a) gold decorated titania materials prepared without CD (TiO2@Au) (b) gold decorated titania materials prepared with HP-β-CD (TiO2@Au-HP) (c) gold decorated titania materials prepared with RAME-β-CD (TiO2@Au-RB) (d), Figure S2. TEM images of TiO2@Au catalyst at magnification of ×62,000, Figure S3. TEM images of (a) TiO2@Au-RBand (b) TiO2@Au-HB at magnification of ×490,000, Figure S4. UV-vis spectra of titania materials prepared by a two-step microwave heating procedure with HAuCl4 in a second step but without CD and without ethanol, Figure S5. TGA profiles for the RAME-β-CD and the HP-β-CD, Figure S6. Evolution of methyl orange concentration under irradiation (λ = 365 nm) as a function of time in the absence (open circle) or presence of the bare TiO2 prepared by microwave process (filled circle). Reaction conditions: TiO2, m = 10 mg; methyl orange solution, V = 4 mL (50 ppm) Figure S7. Performance of TiO2@Au-RB in three consecutive tests with reuse of the catalyst. Reaction conditions: 4 mL of a solution of methylorange (50 ppm), 10 mg of TiO2@Au-RB (λ = 365 nm, t = 10 min), Figure S8. Production of hydrogen by photoreduction of water (80 mL) in the presence of TiO2@Au-RB (100 mg) and ethanol (20 mL) as sacrificial agent (λ = 365 nm).

**Author Contributions:** Synthesis, catalytic tests, ICP, and UV experiments, C.M.; N2 adsorption-desorption measurements and IR spectroscopy, N.K.; TEM analysis, B.L. and A.A.; XRD, A.P. and F.W.; catalysis, S.N.; supervision and reviewing results, reviewing the manuscript, English writing, C.M., E.M., and A.P. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** The TEM facility in Lille (France) is supported by the Conseil Régional du Nord Pas de Calais and the European Regional Development Fund (ERF). Chevreul Institute (FR 2638), Ministère de l'Enseignement Supérieur, de la Recherche et de l'Innovation, Région Hauts-de-France and FEDER are acknowledged for supporting and funding partially this work. The authors are grateful to the University of Artois for supporting this research through the Quality Research Bonus (Micro-GC Agilent 490 in 2018).

**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/).

### *Article* **Effect of Potential and Chlorides on Photoelectrochemical Removal of Diethyl Phthalate from Water**

**Laura Mais, Simonetta Palmas, Michele Mascia and Annalisa Vacca \***

Dipartimento di Ingegneria Meccanica, Chimica e dei Materiali, Università degli Studi di Cagliari, Via Marengo 2, 09123 Cagliari, Italy; laura.mais@unica.it (L.M.); simonetta.palmas@dimcm.unica.it (S.P.); michele.mascia@unica.it (M.M.)

**\*** Correspondence: annalisa.vacca@dimcm.unica.it

**Abstract:** Removal of persistent pollutants from water by photoelectrocatalysis has emerged as a promising powerful process. Applied potential plays a key role in the photocatalytic activity of the semi-conductor as well as the possible presence of chloride ions in the solution. This work aims to investigate these effects on the photoelectrocatalytic oxidation of diethyl phthalate (DEP) by using TiO2 nanotubular anodes under solar light irradiation. PEC tests were performed at constant potentials under different concentration of NaCl. The process is able to remove DEP following a pseudo-first order kinetics: values of kapp of 1.25 <sup>×</sup> <sup>10</sup>−<sup>3</sup> min−<sup>1</sup> and 1.56 <sup>×</sup> <sup>10</sup>−<sup>4</sup> min−<sup>1</sup> have been obtained at applied potentials of 1.8 and 0.2 V, respectively. Results showed that, depending on the applied potential, the presence of chloride ions in the solution affects the degradation rate resulting in a negative effect: the presence of 500 mM of Cl− reduces the value of kapp by 50 and 80% at 0.2 and 1.8 V respectively.

**Keywords:** diethyl phthalate; photoelectrochemical degradation; persistent organic pollutants; chloride ions; TiO2 nanotubes

#### **1. Introduction**

The application of photoelectrochemical process for polluted waters and wastewaters has been gaining more and more attention thanks to the possibility to obtain electrical energy from renewable energy sources, rather than from fossil fuels [1]. The technique exploits the synergy between photochemistry and electrochemistry: from one side, the photochemical process increases its efficiency as the bias potential lowers recombination of the photogenerated charges, from the other side the photo-potential generated on the semiconductor depolarizes the cell improving the yield of the electrochemical process [2].

Considering the application to real matrices, the effect of the composition of the water to be treated plays a crucial role, with particular regard to the presence of chlorides, which are ubiquitous ions in water and wastewater. Several studies on the photochemical process using TiO2 highlighted a negative effect of the presence of chloride: the inhibiting effect has been ascribed both to the competitive adsorption between the pollutant molecules and Cl− towards the surface-active sites of TiO2, or to the scavenging function of chloride ions towards holes and hydroxyl radicals [3,4]. Piscopo et al. [5] showed different effects on the degradation rate of two pollutants depending on the chloride concentration, the nature of the organics and the pH: in the case of poorly adsorbed molecules, if the pH favored the adsorption of Cl−, even low concentration of chloride strongly affected the degradation.

Several papers evidenced the key role of pH in the photocatalytic degradation using TiO2: point of zero charge (pHpzc) plays a crucial role in determining the surface charge of photocatalyst and, in turn, its interaction with charged molecules or ions. When the pH is higher than the pHpzc, the polarity of TiO2 surface is negative and the electrostatic repulsion toward anionic compounds dominates [6–8]. Moreover, since hydroxyl radicals can be formed by the reaction between hydroxide ions and positive holes, the hydroxyl radicals are

**Citation:** Mais, L.; Palmas, S.; Mascia, M.; Vacca, A. Effect of Potential and Chlorides on Photoelectrochemical Removal of Diethyl Phthalate from Water. *Catalysts* **2021**, *11*, 882. https://doi.org/10.3390/catal11080882

Academic Editor: Bruno Fabre

Received: 17 June 2021 Accepted: 19 July 2021 Published: 22 July 2021

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

**Copyright:** © 2021 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 (https:// creativecommons.org/licenses/by/ 4.0/).

considered as the predominant species at neutral or high pH, while at low pH the holes are considered the major oxidizing species [9]. Regarding the scavenging effect, chloride can react with HO• radicals and holes, allowing the formation of less reactive chloride radical (Cl•) and dichloride radicals (Cl2•−) [10–12]: the oxidized chloride may also recombine with photogenerated electrons quenching the photogenerated charge carriers [13].

Different considerations may be made when photoelectrocatalysis is considered: in this case, heterogenous photocatalysis can be improved by the application of a bias potential to obtain a more effective separation of photogenerated charges, thereby increasing the lifetime of electron–hole pairs. In the photoelectrocatalytic process, the increases of the applied potential can accelerate the photogenerated electrons toward the external circuit, generating the bending of the conduction and valence bands, with the consequent formation of a space charge layer. Thus, the recombination of the e−/h<sup>+</sup> pairs may be decreased or totally prevented, improving the photocatalytic performance [14,15]. Moreover, increase in the potential can empty the defects where the photogenerated charges are trapped, enhancing the photoactivity [16].

The presence of chloride in a photoelectrochemical process exerts a different effect with respect to the photochemical one: in fact, unlike the inhibitory effects found in photocatalysis, in photoelectrochemical removal of pollutants, enhancing effect in the degradation process has been often highlighted. Zanoni et al. [17] reported the highest discoloration rate and TOC removal for solution containing Remazol Brilliant Orange 3R at pH 6.0 in presence of 0.5 M of NaCl applying +1.0 V (SCE) to the TiO2 photoanode. Also, in the case of other dyes or organics, the presence of Cl− has been found beneficial to accelerate the degradation rate [18,19]. The improvement in the degradation has been explained by the synergistic action of the strong oxidizing species HO•, chlorine-based radicals Cl• and Cl2•−, and active chlorine species like HClO and Cl2 that can give a bulk contribution [20,21]. Moreover, at the anode the adsorption of negative charged ions, such as chloride, can be enhanced both by the polarization and the promotion of reactions that can generate local acidic pH variation near the anodic surface.

In this framework, our work is devoted to study the photoelectrochemical degradation of a persistent organic pollutant at two levels of applied potentials and in the presence of different concentrations of chloride under simulated solar light conditions, using TiO2 nanotubular electrodes. The pollutant selected for the study is the diethylphthalate (DEP). Phthalate esters (PAEs) are a group of widely used plasticizers that can lead to endocrine system disorders, affecting reproductive function, and inducing some tumors [22–24]. Due to their wide utilization and the difficulty to completely remove them with conventional treatment processes, PAEs are ubiquitous persistent organic pollutants in the environment, being the short chain phthalate as DEP, the most detected in surface marine waters, freshwaters, and sediments [25–27]. To the best of our knowledge, only few papers reported on the photoelectrochemical degradation of the diethyl phthalate [28,29]. Moreover, the influence of the presence of chloride during their treatment and the effect of the applied potential are not yet presented by the literature.

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

#### *2.1. Preparation of TiO2 Nanotubes*

TiO2 nanotube electrode (TiO2-NT) used for the photoelectrochemical degradation of DEP was prepared by electrochemical anodization as reported in our previous work [30]. Briefly, Ti foils (0.25 mm thickness, 99.7% metal basis, Aldrich, St. Louis, MO, USA) were cut in circular disks of 5 cm diameter. After ultrasonic treatments in acetone, isopropanol and methanol (10 min each), Ti was rinsed with deionized water, and dried with a nitrogen stream. The anodization was performed in a two-electrode cylindrical cell made by Teflon (inner dimension: diameter = 4.4 cm and height = 5 cm). The working electrode was located at the bottom of the cell where the electrical contact was an aluminum disc. The exposed geometrical area of the Ti electrodes was 15 cm2. A platinum titanium grid placed in front of the anode at 1 cm distance constituted the counter electrode.

The anodization was performed in (10%) deionized water/(90%) glycerol solution with 0.14 M of NH4F at room temperature. A potential ramp was imposed from open circuit voltage (OCV) to 20 V with a scan rate of 100 mVs−1; then the applied potential was maintained at this fixed value for 4 h. TiO2-NT was annealed in air atmosphere at 400 ◦C for 1 h to transform the amorphous structure into crystalline one. The phase transformation depends on both the structure morphology and annealing temperature: it has been shown that the anatase-to-rutile transformation starts near 430 ◦C for the 500 nm long nanotubes [31], while the same transformation has been reported to occur at 550 ◦C for nanotubes up to 200 nm [32]. In our case, after 1 h at 400 ◦C, a unique anatase phase was present [33]. The morphological characterization of TiO2-NT was presented in [30]: the average diameter of tubes ranged between 40–50 nm, while the tube length of around 700 nm was measured.

#### *2.2. Photoelectrochemical Tests*

Photoelectrochemical tests were performed in a three-electrode beaker cell using TiO2 nanotubes as photoanode, a platinized titanium grid as cathode, and a saturated calomel electrode (SCE) as reference. The cell was filled with 100 mL of solution and connected with a potentiostat-galvanostat (Metrhom Autolab 302N, Metrohm, Herisau, Switzerland) controlled by Nova software. The photoanode was irradiated by UV-vis light using a 300 W xenon lamp equipped with air mass (AM) 0 and 1.5 D filters to simulate the solar irradiation.

Photocurrent measurements were carried out by linear sweep voltammetric (LSV) runs, starting from the OCV to 2.5 V at a scan rate of 10 mVs−1, with hand-chopped light. The photocurrent-time measurements were recorded applying a constant potential in the dark for 10 min; afterward, the electrode was exposed to light for 200 s, followed by dark condition.

Photoelectrochemical oxidation of diethyl phthalate was performed under potentiostatic conditions at 0.2 and 1.8 V vs. SCE. The initial concentration of the organic compound was 40 mg dm−<sup>3</sup> and 0.1 M NaClO4 was used as supporting electrolyte. Moreover, different amount of NaCl (1, 100, 500 mM) were added to the solution, to investigate on the effect of chloride concentration during the photoelectrochemical oxidation of DEP. The pH of the solution was neutral. During degradation experiments, samples of electrolyte were withdrawn for qualitative and quantitative analyses of the model organic compound.

#### *2.3. Analytical Methods*

Analyses of the model organic compound were carried out by HPLC (Waters), equipped with a column Varian C18 and a dual band UV detector set to 283 and 229 nm. The mobile phase was Acetonitrile and aqueous solution 0.1% H3PO4 = 40:60 with a flow rate of 1 mL min<sup>−</sup>1.

The oxidant concentration, expressed as μM of active chlorine, was measured using the N,N-diethyl-p-phenylenediamine (DPD) colorimetric method. DPD oxidizes to form a red-violet product, the concentration of which is determined measuring the absorbance at 515 nm.

The trend of mineralization was monitored by measuring the total organic carbon (TOC) by a Shimatzu TOC 500L instrument.

For each sample a repeatability within ±5% has been evaluated.

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

Figure 1 shows the trend of polarization curve performed at the TiO2-NT electrode during LSV in aqueous solution of DEP under irradiation and in the dark.

**Figure 1.** LSV of TiO2-NT performed at 10 mV s−<sup>1</sup> of scan rate, under dark and irradiation condition. Blue symbols indicate the potentials selected for the degradation runs.

A typical trend is observed, with an onset potential of −0.25 V, followed by an ohmic behavior of the system, in which the positive influence of the potential is strictly connected to the increase in the space charge depletion region of the semiconductor; in the central range of potential (0.7–1.8 V) the saturation of the current is reached, in which increase of the potential is no more effective in terms of a corresponding increasing of the current. In the final range, at potentials higher than the value of band gap of the semiconductor, the barrier breakdown effect could be responsible for the sharp rising in the photocurrent along with the dark current contribution [30].

The degradation tests have been performed selecting two applied potentials: the first one in the ohmic region and the second one in the saturation region. The two blue diamonds in Figure 1 indicate the values of potential selected.

Figure 2a shows the trend with time of the DEP concentration, normalized with respect to the initial concentration, during electrolysis at the two different potentials. For comparison, the trend with time of the DEP concentration at the open circuit potential in the dark was also reported in the same figure: no significant adsorption of DEP on the electrode surface was detected that can be explained considering the neutral pH of the solution, the iso-electrical point of TiO2 located around pH = 6, and the non-ionic nature of the molecule of DEP. When the runs were performed in potentiostatic conditions and under illumination, the concentration of DEP decreased, being the highest reaction rate achieved at 1.8 V.

**Figure 2.** (**a**) Trends with time of the concentration of DEP, normalized to the initial concentration C0, during runs performed with solutions containing 40 mg dm−<sup>3</sup> DEP in 0.1 M NaClO4 as supporting electrolyte at different applied potentials. (**b**) Fraction of reactant removed as a function of the specific charge supplied during the related runs.

However, since the mean current intensity measured during the potentiostatic runs was 0.1 mA at 0.2 V and 1.2 mA at 1.8 V, it could be useful to compare the trend of fraction of the removed reactant as a function of the specific supplied charge (Figure 2b): in this case, the highest yield of the removal process is measured at the lowest potential, indicating that most of the charge passed at 1.8 V has been used for the side reaction of water oxidation.

An analogous behavior was observed in our previous work, where the photo-electrocatalytic degradation of 2,4-dichlorophenoxyacetic acid was investigated: higher efficiency and slower kinetics of degradation were detected in the ohmic region of the polarization curve with respect to those in the saturation region [30].

Degradation curves of DEP at various chloride concentration at the two applied potentials are shown in Figure 3a,b as semilogarithmic plots. A linear trend of ln(C/C0) vs. time is observed under all the experimental conditions, indicating that a pseudo-first order kinetics could be used to interpret the data, as follows:

$$\text{dC/dt} = -\mathbf{k\_{app}} \,\mathrm{C} \tag{1}$$

**Figure 3.** Trends with time of lnC/C<sup>0</sup> during photoelectrochemical degradations using solutions containing 40 mg dm−<sup>3</sup> DEP, 0.1 M NaClO4, and different chloride concentration. (**a**) Applied potential: 0.2 V; (**b**) applied potential: 1.8 V.

The values of the apparent kinetic constant kapp, evaluated from the slope of each straight line at the relevant operative conditions are reported in Figure 4, as a function of the chloride concentration. As already observed in absence of chloride, the fastest kinetics of the reactant removal are obtained at 1.8 V for each level of chloride concentration. Moreover, at 0.2 V, the increase of chloride concentration scarcely affects the reaction rate, except for 500 mM of Cl−, which halves the kapp. At 1.8 V, the effect of chloride is more evident: at 1 mM of Cl− the kapp is reduced by 40% while at 500 mM of Cl− by 80%, in respect to the kapp evaluated without chloride.

Figure 5 shows the trend of the ratio between kapp evaluated at 1.8 V and that at 0.2 V measured at different chloride concentrations. In absence of chloride, an increment of one order of magnitude is obtained, while in presence of the highest concentration of chloride kapp increases of two-fold when the potential values change from 0.2 to 1.8 V. This behavior indicates that the higher the potential, the higher is the negative effect of the concentration of chloride.

**Figure 4.** Pseudo-first order kinetic constants of the reactant removal process performed in solutions of 40 mg dm−<sup>3</sup> of DEP, 0.1 M NaClO4, and different chloride concentration. (**a**) Applied potential: 0.2 V; (**b**) applied potential: 1.8 V.

**Figure 5.** Ratio between the apparent kinetic constant evaluated at 1.8 and 0.2 V for different chloride concentrations.

The inhibiting effect observed in presence of Cl− agrees with observations reported for photocatalytic processes at TiO2-based materials. Several mechanisms have been proposed to explain the inhibiting effect on the photocatalytic degradation [13]:


Moreover, due to the complexity of the processes, a combination of mechanisms is often claimed to explain the inhibiting effect [5,13,39–41].

In the case of a photo-electrochemical process, also the effect of the applied potential should be considered, as well as the pH modification due to the side reactions that occur to a greater or lesser extent depending on the applied potential.

In order to verify the effect of the concentration of chloride and the applied potential on the behavior of the semiconductor, photocurrent transients have been recorded applying different potential and varying the chloride concentration during chopped light chronoamperometries.

Figure 6 shows the results obtained without chloride. For TiO2 nanotubes, the thickness of the wall can be determinant for the extension of the space charge depletion layer; this in turn, can be relevant for the recombination phenomena, which are strictly connected to the applied potential. As can be seen, at the lowest potential, a typical spike of the anodic current is observed, followed by an exponential decrease of the photocurrent with time until a stationary value is reached. The positive spike is no more visible at the highest potential. According to the literature [42], the positive current transient when the light is turned on represents the accumulation of holes at the electrode/electrolyte interface without injection to the electrolyte. Since any fast faradic reaction is occurring, the charge recombination is responsible for the subsequent decrease of the measured current.

**Figure 6.** Potentiostatic tests performed with solution containing 0.1 M NaClO4 at 0.2 (**pink**) and 1.8 V (**black**).

At low potentials, when we operate in the ohmic region of the polarization curve, where the charge depletion layer thickness is not fully developed inside the nanotubes wall, the photogenerated holes may rapidly recombine in the regions of the material that do not experience beneficial space charge effects, i.e., that are non-depleted of the majority carriers (electrons). When the experiment is performed at the highest potential (in the saturation region of the polarization curve) the depletion layer extends in the whole wall of nanotubes and the recombination is suppressed.

Photocurrent transients in presence of chloride are reported in Figure 7.

**Figure 7.** Potentiostatic tests performed with solution containing 0.1 M NaClO4 and different concentration of chloride ions at 0.2 and 1.8 V.

At low applied potential, the rate of the photocurrent decreasing (i.e., the rate of the charge recombination process) is scarcely influenced by the presence of chloride, when they are present at low concentration levels: overlapped curves are obtained related to the runs performed at 0, 1, and 100 mM of Cl− ions. Only at 500 mM of Cl−, slower decay of the photocurrent can be observed, indicating an inhibition of the recombination processes. Moreover, when the light was turned off, negative current transients were observed for high chloride concentration. Negative spikes were often detected during photocurrent transient of semiconductors and can be related to slower electron/hole pairs recombination due to the presence of holes trapped in the surface [43].

These transients in photocurrent can be explained as follows: at lower chloride concentration, the charge recombination prevails since chloride is poorly adsorbed onto the semiconductor electrode, so it is not able to react with the photogenerated holes faster than the electrons. However, at the highest concentration of chloride, it is likely that the adsorption effect would predominate, so that chloride can act as hole scavenger, according to the following adsorption phenomena:

$$\rm TiO\_2\rm \rm \rm h^+ + Cl^- = TiO\_2\rm \rm Cl\_{ads} \tag{2}$$

This process promotes the separation of electron-hole pair limiting the charge recombination as suggested by other authors [20,44,45].

At the highest potential, the recombination is suppressed, and positive transient and negative spikes disappear also in presence of high chloride ions. Moreover, very small increment in the steady state photocurrent was observed, increasing the concentration of chloride. So, at 0.2 V, the highest variation in the value of kapp obtained at 500 mM of chloride, can be connected to the blocking effect of adsorbed Cl− and the competitive adsorption, with respect to water molecules, which reduces the formation of HO• radicals.

Similar considerations should be done also to explain the result at 1.8 V, but, as we noticed, the inhibiting effect at this potential is evident also at low concentration of chloride. This can be explained by considering two aspects connected to the applied potential: the electrode works in a region of potential where the oxygen evolution reaction occurs to a large extent, so that a local acidic pH near the surface can generate a positive charge (pH < isoelectric point). Moreover, the application of high anodic potentials can generate a build-up of a positive surface charge. In this condition, the competitive adsorption or

blocking of active surface sites by chloride anions will be favored due to the electrostatic attraction of Cl−, also at low concentration of chloride.

The adsorbed chloride can react to form chlorine by the following reaction [17,44]:

$$\text{TiO}\_2\text{-Cl}\_{\text{ads}} + \text{Cl}^- \rightarrow \text{Cl}\_2 + \text{TiO}\_2 + \text{e}^- \tag{3}$$

Dissolved chlorine reacts with water to give hypochlorous acid and hypochlorite ions (Equations (3) and (4)), being the distribution of the three forms of active chlorine dependent on pH:

$$\text{Cl}\_2 + \text{H}\_2\text{O} \rightarrow \text{Cl}^- + \text{HClO} + \text{H}^+ \tag{4}$$

$$\text{HClO} \leftrightarrow = \text{H}^+ + \text{ClO}^-\tag{5}$$

Chlorine-based oxidants (active chlorine) have been detected during the photoelectrochemical degradation of DEP in different operating conditions.

At 0.2 V after 130 C dm−<sup>3</sup> of supplied charge, 2.0 and 4.2 μM of active chlorine concentrations were detected at 100 and 500 mM of Cl−, respectively. These small amounts agree with the poor adsorption of chloride at this value of applied potential. At 1.8 V, higher concentration of active chlorine was detected. As an example, the trend with time of the concentration of active chlorine obtained during DEP degradation in presence of 100 mM of Cl− is reported in Figure 8. The higher amount of active chlorine confirms a better reactivity of chloride with the positively charged surface of TiO2 at 1.8 V.

**Figure 8.** Trends with time of the concentration of active chlorine produced during a degradation run at 1.8 V with solution containing 40 mg dm−<sup>3</sup> of DEP, 0.1 M NaClO4, and 100 mM of Cl−.

The formation of chlorine-based oxidants during photoelectrochemical treatment of water containing chloride has been studied by several authors: some of them found that the presence of Cl− suppressed the degradation rate of organic pollutants, while others found opposed result [44].

The positive effect was generally observed when the active chlorine was able to give a bulk contribution to the reaction, i.e., in the cases where the organic pollutants can be oxidized also by active chlorine. For example, during photo-electrochemical discoloration of solutions containing Methylene Blue, low pH, and high concentration of Cl− were beneficial [18]. Also, Zanoni et al. [17] found the highest TOC removal for solution containing Remazol Brilliant Orange 3R, working at pH 6.0, 1.0 M NaCl, when the photoelectrode was biased at +1 V (versus SCE).

In our case, the formation of active chlorine seems not sufficient to contribute to the overall reaction rate at such an extent to make up for the negative effect.

Some specific tests were performed to evaluate the effectiveness of the photo-electrogenerated active chlorine on the DEP degradation. To this aim, during photoelectrocatalytic degradation runs, the light was turned off and the application of bias potential was stopped. In

this condition, in the solution, 25 μM of active chlorine accumulated, and residual 26 mg dm−<sup>3</sup> of DEP were present: the solution was monitored by following the concentration of the residual DEP with time. Negligible variation in the concentration of DEP was found after two hours indicating that the HO• radicals may be considered as the main factor responsible for the degradation, while active chlorine seems to give a not significant contribution to the overall oxidation rate. Similar behavior was found during the electrochemical degradation of the dimethyl phthalate ester on a fluoride-doped Ti/β-PbO2 anode: the lower removal of the pollutant in the presence of chloride ions was explained considering the lower reactivity of dimethyl phthalate with chlorine radical species in respect to hydroxyl radicals. Also, the active chlorine can react with HO• radicals thus reducing their availability for organic oxidation [46].

The low reactivity of active chlorine towards DEP obtained in our experimental conditions may indicate that the formation of harmful chlorinated intermediates is unlikely, even if the possible reaction of DEP intermediates with active chlorine during the runs cannot be excluded. Table 1 reports the ratio (*ϕ*) between the removal percentages of TOC and DEP evaluated at the end of each run, which indicates the level of total mineralization as defined by the following equation [47]:

$$\varphi = \frac{\% \text{[TOC]}\_{\text{removal}}}{\% \text{[}DEP\text{]}\_{\text{removal}}} \tag{6}$$


**Table 1.** TOC removal and ϕ evaluated at the end of each run.

At 1.8 V, a higher degree of mineralization was evaluated at the end of the runs, in which *ϕ* approached the unity. However, at 0.2 V, the high values of *ϕ* also indicate that the possible intermediates are almost completely removed.

#### **4. Conclusions**

In this work, the photoelectrochemical degradation of diethyl phthalate has been studied at two levels of applied potentials and in the presence of different concentrations of chloride under simulated solar light conditions, using TiO2 nanotubular electrodes. The process is able to remove DEP following a pseudo-first order kinetics: values of kapp of 1.25 × <sup>10</sup>−<sup>3</sup> min−<sup>1</sup> and 1.56 × <sup>10</sup>−<sup>4</sup> min−<sup>1</sup> were obtained at applied potentials of 1.8 and 0.2 V, respectively. Higher current efficiency and slower kinetics of degradation were detected in the ohmic region of the polarization curve at 0.2 V. The presence of chloride ions in the solution affects the degradation rate to different extents depending of the applied potential: the higher the potential, the higher the negative effect of the increase of chloride concentration. The presence of 500 mM of Cl− halves the kapp at 0.2 V, while at 1.8 V its value decreases to 1.56 × <sup>10</sup>−<sup>4</sup> min<sup>−</sup>1. This behavior can be connected to the blocking effect of adsorbed Cl− and the competitive adsorption, with respect to water molecules, which reduces the formation of HO• radicals: at 0.2 V, the adsorption of chloride predominates only at the highest concentration of chloride, while at 1.8 V, the positive surface charge due to the applied potential and the possible acidification of the anodic layer allow the adsorption also at low chloride concentrations.

**Author Contributions:** Conceptualization, A.V. and S.P.; methodology, M.M.; validation, L.M., A.V. and S.P.; formal analysis, L.M.; investigation, L.M.; writing—original draft preparation, A.V. and S.P.; writing—review and editing, S.P., L.M. and M.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This paper is part of the research project funded by P.O.R. SARDEGNA F.S.E. 2014–2020— Axis III Education and Training, Thematic Goal 10, Specific goal 10.5, Action partnership agreement 10.5.12—"Call for funding of research projects—Year 2017".

**Data Availability Statement:** Data is contained within the article.

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

#### **References**


### *Article* **Analysis of Photocatalytic Degradation of Phenol with Exfoliated Graphitic Carbon Nitride and Light-Emitting Diodes Using Response Surface Methodology**

**Adeem Ghaffar Rana 1,2 and Mirjana Minceva 1,\***


**Abstract:** Response surface methodology (RSM) involving a Box–Benkhen design (BBD) was employed to analyze the photocatalytic degradation of phenol using exfoliated graphitic carbon nitride (g-C3N4) and light-emitting diodes (wavelength = 430 nm). The interaction between three parameters, namely, catalyst concentration (0.25–0.75 g/L), pollutant concentration (20–100 ppm), and pH of the solution (3–10), was examined and modeled. An empirical regression quadratic model was developed to relate the phenol degradation efficiency with these three parameters. Analysis of variance (ANOVA) was then applied to examine the significance of the model; this showed that the model is significant with an insignificant lack of fit and an R<sup>2</sup> of 0.96. The statistical analysis demonstrated that, in the studied range, phenol concentration considerably affected phenol degradation. The RSM model shows a significant correlation between predicted and experimental values of photocatalytic degradation of phenol. The model's accuracy was tested for 50 ppm of phenol under optimal conditions involving a catalyst concentration of 0.4 g/L catalysts and a solution pH of 6.5. The model predicted a degradation efficiency of 88.62%, whereas the experimentally achieved efficiency was 83.75%.

**Keywords:** g-C3N4; photocatalysis; response surface methodology; wastewater treatment; phenol

#### **1. Introduction**

For all living beings, water is considered to be the most important resource. Easy access to clean water is one of the biggest challenges for mankind. In the last few decades, advancements in science, technology, and industrialization have led to considerable benefits to mankind but at the cost of a more polluted environment, particularly water [1]. There are multiple categories of pollutants in water, such as heavy metals, dyes, pesticides, pharmaceuticals, and other organic pollutants. Amongst organic pollutants, phenolic compounds, with ~3 million tons of global production, are an emerging contaminant detected in water [1–4].

Phenols or phenolics are essential because of their wide range of applications in the processing and manufacturing industry. However, the ecosystem's contamination by phenolics is concerning because of the adverse implications on human health such as their endocrine-disrupting abilities and carcinogenic behavior [1,5,6]. Moreover, these chemicals cause environmental issues such as water hardness, pH change, and a decrease in dissolved oxygen level. Furthermore, the Environmental Protection Agency (EPA) and the European Union (EU) have included a few phenols in their priority pollutants list. It is necessary to make this polluted water containing phenols and other pollutants suitable for human use and aquatic life using certain techniques to minimize the usage of these chemicals [5].

The removal of phenolic compounds from wastewater has attracted considerable attention from researchers [5]. Many biological, chemical, and physical techniques such

**Citation:** Rana, A.G.; Minceva, M. Analysis of Photocatalytic Degradation of Phenol with Exfoliated Graphitic Carbon Nitride and Light-Emitting Diodes Using Response Surface Methodology. *Catalysts* **2021**, *11*, 898. https:// doi.org/10.3390/catal11080898

Academic Editor: Annalisa Vacca

Received: 2 July 2021 Accepted: 23 July 2021 Published: 25 July 2021

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

**Copyright:** © 2021 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 (https:// creativecommons.org/licenses/by/ 4.0/).

as membrane filtration, coagulation–flocculation, adsorption [7,8], ion exchange, bacterial and fungal biosorption [9], aerobic and anaerobic processes [10] are used for phenol removal. In these processes, there are many constraints such as high cost, and low efficiency; furthermore, these methods do not completely remove phenol from wastewater [11,12]. Moreover, using these techniques, phenol is transferred from wastewater to a solid phase that requires treatment for safe disposal, which leads to additional cost for the whole process. Thus, it is necessary to develop an alternative effective and cost-efficient method for phenol removal from wastewater.

Advanced oxidative processes (AOP) are successful for achieving the complete removal of pollutants [13]. The degradation process using AOP can be performed in several ways, such as using only oxidizing agents, light irradiance in addition with oxidizing agents, and photocatalysis [14]. For all these processes, the degradation process is conducted using OH− radicals that are generated during the oxidation reaction. Among these processes, photocatalysis has attracted considerable interest because it can harvest solar light with the help of semiconductor materials (catalysts). The catalysts can help solve environmental issues related to water contaminations; these semiconductor materials are nontoxic and efficient. Note that different semiconductor materials such as ZnO [15], TiO2 [16], SiO2, Al2O3 [8], and g-C3N4 [17,18], are used for environmental applications in photocatalysis; these have considerable advantages because of the large surface areas, adsorption capacities, and better absorption of light. Among these materials, g-C3N4 offers improved visible light absorption [17,19–21].

g-C3N4, a polymeric semiconductor, composed of C, N, and H, has gained considerable interest from researchers for novel generation of photocatalysts because of its widespread catalytic uses in oxidation and reduction processes, such as pollutant degradation, water splitting, and CO2 reduction. These materials have been extensively used for environmental remediation because they are easy to synthesize, metal-free, inexpensive, and easily available [22–24]. Furthermore, g-C3N4 possesses higher thermal and chemical stability because of π-conjugated frameworks connecting the 2D layered structure of tri-striazine building blocks. g-C3N4 can be activated by visible light of 420–460 nm because of its low bandgap energy (2.7 eV) [25,26]. There are, however, certain challenges associated with the application of g-C3N4 in phenol removal such as low surface area, fast recombination rate, and low conductivity, thus resulting in lower efficiency. To overcome these limitations, multiple strategies have been used to improve the surface electronic structures and activity of the bulk g-C3N4 in visible light. To improve the activity of pristine g-C3N4, strategies such as metal and non-metal doping, exfoliation, hard and soft templating, and metal oxide heterojunctions have been used [27–31].

Factors affecting the removal efficiency can be tuned by the morphology and/or chemistry of the catalyst and by optimizing the operating parameters. Multiple operating parameters play an important role in the photocatalytic degradation process, thus making their optimization important for achieving good photocatalytic degradation of the target pollutant. Response surface methodology (RSM) is one of the most commonly applied optimization techniques; it is a powerful optimization tool for an experimental design that efficiently helps in systemic analysis [5,11,14]. RSM uses mathematics and statistics to analyze the relative significance of influencing factors on the response of the studied system. RSM is suitable for predicting the effect of individual experimental operating parameters, in addition to locating interactions between parameters and their impact on a response variable. RSM uses a systematic technique to simultaneously vary all parameters and evaluate the influence of these parameters on photocatalytic degradation [32,33]. The greatest advantage of RSM lies in the systematic approach for the experimental design, which mostly requires fewer experiments, thus reducing the time required and thereby being more economical. For designing these experiments, a central composite design (CCD) [3] and Box–Benkhen design (BBD) [11,12] are most commonly used. For the same number of parameters, BBD requires fewer experiments than CCD [3]; therefore, in this study, BBD is selected as a preferred design approach.

92

The objective of this study was to analyze the photocatalytic degradation of phenol with metal-free g-C3N4 and visible LED light and to model the process using RSM. In this study, the operating parameters considered were catalyst concentration, phenol concentration, and pH of the solution. BBD was used for the experimental design and RSM was applied to determine the mathematical relationship between operating parameters and phenol degradation. Finally, the correlation determined by RSM was experimentally validated.

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

#### *2.1. Chemicals and Materials*

Melamine (C3H6N6, 99%) was purchased from Alfa Aesar. Phenol (C6H5OH, 99%) was purchased from Merck. Acetonitrile (C2H3N, 99.99%) and ultra-pure water for highperformance liquid chromatography (HPLC) were purchased from Sigma Aldrich. NaOH and HCl were purchased from VWR chemicals. All chemicals used were of analytical grade and used as-received without any further purification.

#### *2.2. Photocatalyst Synthesis*

Photocatalyst was prepared as per the procedure used in our previous study [18]; the synthesis process is briefly reported here. Melamine was placed in a muffle furnace (Carbolite Gero, GPC 1200, Derbyshire, UK) in a closed crucible to prepare bulk g-C3N4 using thermal decomposition. The synthesis process comprised two steps: A heating ramp rate of 2 ◦C min−<sup>1</sup> was programmed up to 450 ◦C; this temperature was maintained for 2 h. Then, the temperature was increased to 550 ◦C using a heating ramp rate of 2 ◦C min−<sup>1</sup> and then maintained for 4 h. The material synthesized was crushed in mortar after cooling, then rinsed with ultrapure water, and dried overnight at 80 ◦C. The exfoliation process was conducted in an open crucible at 500 ◦C for 2 h at a heating ramp rate of 2 ◦C min−<sup>1</sup> in a muffle furnace.

#### *2.3. Characterization of the Photocatalyst*

Fourier transform infrared (FTIR) measurements (4000–400 cm−1) were performed on a Spectrum Two FT-IR Spectrometer (PerkinElmer, Switzerland) with a universal ATR (UATR Two) cell equipped with a ZnSe single crystal. The acquisition performed using 60 scans and the resolution was set to 4 cm−1. Zetasizer Nano ZEN5600 (Malvern, UK) was used to measure the zeta potential of the synthesized material. SU8030 (Hitachi, Japan) SEM-type microscope operated at an acceleration voltage of 10 kV and a probe current of 15 pA was used to examine the morphology of the material with scanning electron microscopy (SEM).

#### *2.4. RSM with Box–Behnken Experimental Design*

The influence of three independent operating parameters, i.e., catalyst concentration (A), phenol initial concentration (B), and pH of the solution (C), was considered in RSM. The remaining reaction conditions, namely, the airflow rate (50 mL/min) and reaction time (3 h), was kept constant in the experiment based on previous study [18]. The degradation efficiency of phenol (Equation (1)) was set as a response variable. Note that a previous study [18] was conducted to obtain the upper and lower limits of the parameters. Table 1 shows the ranges and levels of independent parameters A, B, and C. BBD was used to examine the combined effect of these three variables. Section 3.3 lists the set of experiments in table; it includes a replication of experiments at the central point. Regression analysis was the performed using OriginPro 2021 9.8.0.200 (OriginLab Corporation, Northampton, MA, USA) software. The suggested model's data were analyzed for significance and suitability using analysis for variance (ANOVA).


**Table 1.** Independent parameters and their ranges and levels.

#### *2.5. Photocatalytic Experiments*

Figure 1 shows the photocatalytic experiments that were conducted in a jacketed glass reactor (working volume 225 mL) (Peschl Ultraviolet GmbH, Mainz, Germany) with a safety cabinet. The reactor was irradiated from inside using a custom-made LED immersion lamp; the LED has maximum emission at 430 nm. Glass reactor was then sonicated with a reaction mixture for uniform dispersion, followed by stirring with continuous airflow to maintain adsorption–desorption equilibrium for 30 min. Subsequently, lights were turned on, which is considered as zero time (to). Nine to ten samples (1 mL) were periodically collected from the reaction mixture. After centrifugation and filtration, the samples were analyzed using HPLC. For acidic and basic reaction conditions, the pH of the mixture was adjusted using 0.1 M HCl and NaOH. The phenol degradation efficiency was determined using the following Equation:

$$\text{Depradiation efficiency } (\%) = \frac{\text{C}\_o - \text{C}}{\text{C}\_o} \times 100 \tag{1}$$

where *Co* is the initial phenol concentration and *C* is the residual phenol concentration in the solution at an irradiation time t.

**Figure 1.** Photocatalytic reactor setup.

The reduction of the reaction mixture volume due to the sampling was less than 5% at the end of the experiments and was therefore not considered in the calculation of the phenol degradation efficiency.

#### *2.6. Analytical Techniques*

A prominence HPLC system from Shimadzu (Kyoto, Japan) was used for analyzing the samples obtained from the reactor. The system is equipped with a binary pump (Model LC-20AB), an autosampler (Model SIL-20A), a degasser (Model DGU-20A3,) and a diodearray detector (Model SPD-M20A). Phenomenex (C18, 150 × 4.6 mm, 3 μm) column was used with a fixed flow rate of 0.8 mL/min, with the mobile phase gradient of water (A) and acetonitrile (B): starts with 15% B, followed by 60% B in 7 min and back to 15% B in 8 min;

injection of 5 μL; UV light of 254 nm. Phenol was analyzed at a maximum absorption wavelength (λmax) of 270 nm.

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

#### *3.1. Photocatalyst Characterization*

The metal-free g-C3N4 used in this study was synthesized and characterized in our previous study [18] using transmission electron microscopy (TEM), Brunauer–Emmett–Teller isotherms (BET), X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), photoluminescence (PL), and UV-Vis spectroscopy. In this study, scanning electron microscopy (SEM), Fourier transform infrared spectroscopy (FTIR), and zeta potential analyses were performed. Table 2 lists the physical properties of metal-free g-C3N4 before and after its exfoliation.


**Table 2.** Summary of characterization results [18].

The exfoliated material has a significantly higher surface area than the bulk material, while the average pore size of both materials is almost the same (Table 2 and Figure S1). Using XRD, the material shows two characteristic peaks of g-C3N4 (Figure S4) [34,35]. The strong and weak peaks of N1s and C1s observed in XPS confirm the chemical state of g-C3N4 (Figure S3) [17,36–41]. Table 2 lists the maximum absorption wavelength and bandgap of the material, which are presented in Figure S2 [42,43].

In Figure 2, the selected SEM images of bulk and exfoliated g-C3N4 are presented. The thermal exfoliation transformed the stacked and aggregated structure of bulk g-C3N4 in a porous nanosheet structure. The reduction in layer thickness (Figure 2b) leads to an increase in the specific surface area of g-C3N4 [17,44–46].

**Figure 2.** SEM images of the bulk (**a**) and exfoliated (**b**) g-C3N4.

Figure 3 shows the catalysts' FTIR spectra. A broad peak is observed between 3200 and 3000 cm−1, which can be attributed to the stretching vibrations of N–H bonds from

residual amino groups and adsorbed H2O. The sharp peak that appears at 806 cm−<sup>1</sup> can be attributed to the breathing mode of triazine units [47,48], whereas the strong bands between 1636 and 1242 cm−<sup>1</sup> belong to the C=N and C–N bonds of heterocyclic rings. Because the spectra of both materials show the same absorption bands, the chemical structure remained unaltered after treatment.

**Figure 3.** Fourier transform infrared spectra of bulk and exfoliated g-C3N4.

Figure 4 shows the effect of pH on the zeta potential of the exfoliated g-C3N4. The catalyst surface is positively charged at acidic pH (3) and negatively charged at natural (6) and basic pH (10).

**Figure 4.** Zeta potential at different pH of the synthesized exfoliated g-C3N4. Reproduced with permission from [18].

The optical properties (PL/UV-Vis) and surface area (BET) of the material have changed with exfoliation; however, the chemical state (XPS), phase (XRD), and the chemical structure (FTIR) remained the same after exfoliation.

#### *3.2. Photodegradation Studies*

The photodegradation efficiency of exfoliated g-C3N4 photocatalyst was evaluated under visible light irradiation using 430 nm wavelength LEDs. The influence of individual operation parameters, catalyst concentration, phenol concentration, and pH of the solution, in their preselected ranges (Table 1), was examined. For all experiments, an adsorption time of 30 min was used before the light irradiation was started. Moreover, the photolysis experiment was performed to verify the removal of phenol in the absence of the catalyst. Phenol removal with adsorption in the dark and photolysis is insignificant compared to the removal of phenol obtained in the presence of light (Figure 5a). Figure 5a shows the effect of g-C3N4 photocatalyst concentration in the range of 0.1–0.75 g/L on phenol degradation, which increased with the increase in catalyst concentration up to 0.75 g/L because of an increased number of active sites available for the reaction to occur. However, there is no significant increase at >0.5 g/L because an additional increase of the catalyst concentration might cause light scattering and hindrance in light absorption. The effect of phenol concentration on the performance of the catalyst on phenol degradation was examined for three concentrations between 20 and 100 ppm and is shown in Figure 5b. The phenol degradation efficiency decreased as the concentration increased because of the higher number of molecules for adsorption on the available active sites, which hinders the absorption of light. Figure 5c shows the effect of different pH on phenol degradation. Increasing the pH decreases the degradation efficiency of exfoliated g-C3N4. Note that acidic pH is most favorable for phenol degradation because as per the zeta potential (Figure 3) and the surface charge of the catalyst is positive at an acidic pH, which helps attract OH– ions produced in the solution due to dissociation of H2O2 to the surface and improves the degradation efficiency.

**Figure 5.** Phenol degradation at preselected (**a**) catalyst concentration (at 20 ppm and natural pH) (**b**) pollutant concentration (at 0.5 g/L and natural pH), and (**c**) pH of the solution (at 0.5 g/L and 20 ppm); airflow = 50 mL/min. Reproduced with permission from [18].

#### *3.3. Response Surface Methodology*

#### 3.3.1. Model Equation

To analyze the combined effect of three variables: catalyst concentration (A), phenol concentration (B), and pH of the solution (C) on the degradation efficiency of phenol (Equation (1)), a three-variable BBD was used in the experimental design for RSM. Table 3 lists the set of performed experiments and the obtained phenol degradation (in 3 h and under an airflow of 50 mL/min).

**Table 3.** Box–Behnken design with experimental and predicted phenol degradation efficiency values with Equation (2).


Experimental data were fitted with four different models: two-factor interaction (2FI), linear, quadratic, and cubic model to obtain regression equations. Three different tests, namely, the sequential model sum of squares, lack of fit, and model summary statistics, were conducted to determine the adequacy of various models; the results are presented in Table 4. The response surface model is then used to select the best model based on the following criterion: the highest-order polynomial with additional significant terms and the model is not aliased (Table 4). The cubic model has the highest polynomial model because there are no sufficient unique design points to independently estimate all terms for that model. The aliased model results in unstable and inaccurate coefficients and graphs. Thus, the aliased model cannot be selected [49,50]. The criteria used in the lack of fit test is the non-significant lack of fit (*p*-value > 0.05) based on which a quadratic model is selected. Moreover, multiple summary statistics are calculated to compare models or to confirm the adequacy of the model. These statistics include adjusted R2, predicted R2, and prediction error sum of squares (PRESS). A good model will have a largely predicted r2, and a low PRESS. According to the aforementioned criteria, adjusted R2 (0.967) and predicted R<sup>2</sup> (0.805) are in reasonable agreement with each other and have a low PRESS. Thus, the quadratic model is finally selected to build the response surface.


**Table 4.** Adequacy of the models tested.

Based on regression coefficients from Table 5, the following empirical second-order polynomial equation was obtained:

#### Degradation Efficiency (%)

<sup>=</sup> 85.72 <sup>+</sup> 6.36 A <sup>−</sup> 24.86 B <sup>−</sup> 15.22 C <sup>+</sup> 3.71 AB <sup>−</sup> 2.83 AC <sup>−</sup> 2.38 BC <sup>−</sup> 5.05 A2 <sup>−</sup> 5.22 B2 <sup>−</sup>16.07 C2 (2)

> where, A, B, and C are the catalyst concentration, phenol concentration, and pH of the solution, respectively.


**Table 5.** Coefficients of the second-order polynomial (quadratic) equation.

The influence of model terms on the degradation of phenol as per p-values (Table 5) is in the following order B < C < C2 <A<B<sup>2</sup> < A<sup>2</sup> < AB < AC < BC. The mixed interaction terms AB, AC, and BC are not significant because their *p*–value is > 0.05 and may be removed from Equation (2).

An ANOVA of the second-order polynomial (Equation (2)) for phenol degradation was conducted; the results are shown in Table 6. In statistics, the significance of the model can be confirmed by a large F-value (53.31) and a small *p*-value (<0.0001). Furthermore, the significance of the model can be confirmed by the lack of fit test. In this study, the lack of fit is not significant because its *p*-value is >0.05. The accuracy of the model is confirmed by the low coefficient of variation (CV) value of 5.79%. The results showed that the signal-to-noise ratio of 24.89 is adequate.


**Table 6.** Analysis of variance ANOVA of the second-order polynomial (Equation (2)).

Furthermore, the coefficient of determination R2 confirmed the fit of the model. For the used model, the value of the predicted R<sup>2</sup> = 0.810 (Table 6) is in agreement with adjusted R<sup>2</sup> = 0.967, which indicates that the obtained model is significant.

Equation (2) provides a suitable relationship (R<sup>2</sup> = 0.810) between the response (degradation efficiency) and the parameters, which can be seen in Figure 6. In this figure, the experimental values of phenol degradation are plotted against the predicted values obtained from the RSM model; these values of the percentage phenol degradation fit well.

**Figure 6.** The experimental phenol degradation efficiency (%) plotted against the predicted values from the RSM model.

3.3.2. Interaction Effects of Independent Operating Parameters

Three dimensional (3D) response surface and contour plots were generated using the regression model (Equation (2)) to visualize the influence of the independent operating parameters on phenol degradation; they are presented in Figures 7–9. In surface and contour plots, one parameter is maintained constant at its zero levels, whereas the other two are varied in the studied range reported in Table 1.

**Figure 7.** Effect of catalyst concentration and pH on the degradation of phenol: pollutant concentration was kept constant at 60 ppm.

**Figure 8.** Effect of pollutant concentration and pH on the degradation of phenol: catalyst concentration was kept constant at 0.5 g/L.

**Figure 9.** Effect of catalyst concentration and pollutant concentration on the degradation of phenol: pH was kept constant at 6.5.

Figure 7 shows the influence of pH and catalyst concentration on the degradation efficiency of phenol at a constant phenol concentration of 60 ppm. The contour lines show a decrease in the degradation efficiency with an increase in pH; there is no considerable increase in efficiency, even at higher catalyst concentrations. However, an increase in degradation efficiency with a decrease in pH is observed. These results demonstrate that pH has a significant effect on phenol degradation and a low pH favors the degradation process. This phenomenon is linked with the zeta potential of the catalyst surface [18]. There is a positive charge at the surface of the catalyst at an acidic pH (Figure 2), which attracts the OH− ions produced in the solution due to dissociation of H2O2 and significantly increases the degradation process. However, at a basic pH, the surface charge is negative and there could be electrostatic repulsion that reduces the efficiency of the degradation process.

Figure 8 shows the influence of pH and pollutant concentration on phenol degradation at a constant catalyst concentration of 0.5 g/L. For selecting the catalyst concentration, the effect of initial pollutant concentration is important. The contour lines demonstrate that simultaneously increasing both parameters (pH and phenol concentration) considerably decreases the degradation efficiency of phenol (33%), which is 62% at a low pH. As shown in Figure 5b, at low pH and low pollutant concentration, 100% degradation is achieved in a considered reaction time of 3 h. An increase in degradation efficiency from high to low pH can then be associated with catalyst surface charge. However, a decrease in efficiency at low pH from low to high phenol concentration is attributed to the increased number of pollutant molecules compared with the available active sites.

Figure 9 shows the effect of catalyst concentration and pollutant concentration at a constant pH of 6.5. The contour lines demonstrate that both parameters independently affect the degradation efficiency. By increasing the catalyst concentration at a lower pollutant concentration, phenol degradation increases; however, at a higher pollutant concentration, the degradation efficiency decreases. This can be attributed to the availability of active sites on the catalyst surface for OH− radicals, as well as phenol molecules. The electron–hole pair generated from the catalyst surface improves the degradation rate.

#### 3.3.3. Experimental Validation of RSM Model

To demonstrate the applicability of the model, a hypothetical case study for water with a phenol concentration of 50 ppm was considered. The model equation was used to identify the optimum catalyst concentration and pH, leading to maximal phenol degradation in 3 h under an airflow rate of 50 mL/min. According to the model prediction, maximal phenol degradation of 88.62% is achievable using 0.4 g/L of catalyst concentration and operating at a pH of 6.5. To examine the accuracy of the model prediction, an experiment was conducted under these conditions. The experimentally obtained phenol degradation was 83.75%, which is less than a 5% deviation from the predicted value. Thus, the optimum operating point obtained by RSM was successfully confirmed; this suggests that RSM can be a useful tool for optimizing photocatalytic processes. Similarly, the model developed can be used for minimizing the catalyst amount or for maximizing the degradation efficiency of phenols for any set of parameters in range.

#### **4. Conclusions**

Metal-free g-C3N4 was used for the photocatalytic degradation of phenol from an aqueous solution. The morphology of the catalyst was confirmed by SEM, and the surface charge was confirmed using zeta potential. Based on zeta potential, the catalyst surface was confirmed to have a positive surface charge under acidic conditions and a negative surface charge under basic conditions; therefore, acidic pH favors the degradation process. A RSM based on the BBD was used to analyze the degradation efficiency of phenol. The influence of experimental parameters, namely, catalyst concentration, pollutant concentration, and pH of the solution, and their interaction at a different level was examined for phenol degradation. An empirical regression quadratic model was developed for the response variable. Analysis of variance (ANOVA) demonstrated that the model is significant with an insignificant lack of fit and a high coefficient of determination (R2) of 0.96, which can be helpful to navigate the design space. Furthermore, an optimized degradation efficiency of 83.75% was achieved for phenol concentration of 50 ppm, catalyst concentration of 0.4 g/L, and a solution pH of 6.5 pH (in 3 h and under an airflow of 50 mL/min). Thus, the results suggest that the RSM can be used for the optimization of parameters for maximizing the photocatalytic degradation of phenol using g-C3N4 and LEDs.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/10 .3390/catal11080898/s1, Figure S1 N2 adsorption-desorption isotherms of bulk and exfoliated g-C3N4. The inset shows the corresponding BJH pore size distribution curves of the sample, Figure S2 (**a**) UV-Vis absorption spectra and (**b**) PL spectra of bulk and exfoliated g-C3N4; insets of (**a**) showing the Tauc plots, Figure S3 XPS spectra of bulk and exfoliated g-C3N4 C1s, N1s, Figure S4 X-ray diffraction patterns of bulk and exfoliated g-C3N4.

**Author Contributions:** Conceptualization, A.G.R.; Formal analysis, A.G.R.; Investigation, A.G.R.; Methodology, A.G.R.; Resources, M.M.; Supervision, M.M.; Writing—original draft, A.G.R.; Writing review and editing, M.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** A.G.R. acknowledges the financial support from the Higher Education Commission, Pakistan, and Deutscher Akademischer Austauschdienst (DAAD), Germany.

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

#### **References**

