**Magnetic Sorbent for the Removal of Selenium(IV) from Simulated Industrial Wastewaters: Determination of Column Kinetic Parameters**

#### **Andrew Ying 1,**†**, Samuel F. Evans 1,2,**†**, Costas Tsouris <sup>3</sup> and M. Parans Paranthaman 1,2,\***


Received: 1 April 2020; Accepted: 22 April 2020; Published: 26 April 2020

**Abstract:** A novel meso- and microporous tire-derived-carbon support with magnetic iron oxide nanoparticle adsorbents that selectively adsorbs Se(IV) ions from simulated contaminated water has been developed. In this work, the physicochemical characteristics of the composite adsorbent are characterized with respect to porosity and surface area, chemical composition, and microstructure morphology. The kinetics of this composite adsorbent in a fixed-bed setting has been determined. Several column runs were conducted and analyzed by inductively coupled plasma-optical emission spectroscopy (ICP-OES) to determine the concentration gradient vs time. These results were then fit to a pseudo-second order rate law to obtain equilibrium values. Combining calculated equilibrium values with effluent concentration data, enabled the application of the Adams–Bohart model to determine reaction constants and column coefficients. Column parameters obtained from different flow rates and fittings of the Adams–Bohart model were remarkably consistent. These findings enable the application of this sorbent to fixed-bed column systems and opens up further research into mixed pollutants tests with real wastewater and scaling of selenium pollutant removal.

**Keywords:** selenium removal; wastewater purification; nanoadsorbents; carbon magnetic iron oxide particles; bench scale column extraction; column kinetics

#### **1. Introduction**

Mining, fossil, and petrochemical operations often produce significant quantities of wastewater with high concentrations of toxic metals such as selenium, arsenic, and lead [1–4]. The millions of gallons of wastewater produced from industrial activities present an economic, legal, and technological challenge as the water must be treated before being stored or reintroduced back into the environment [5–8]. Previous research has explored the use of composite materials derived from sources of industrial waste, such as tires and iron chloride [9], as adsorbents that can affordably and sustainably reduce selenium concentrations to EPA standards of ≤50 ppb. These materials can be easily integrated into an industrial purification process by flowing wastewater through a fixed bed of composite adsorbents in a continuous-flow system.

Ion uptake rates of adsorbents vary when scaling from batch adsorption to continuous column tests due to transport effects. Fixed beds can experience uneven flow distribution, z-axis concentration gradients, mass transport limitations due to the presence of inert media, and other mass transport limitations. However, fixed-bed columns are of simple design and perform effectively in the removal of low concentration pollutants [10]. Kinetics models, such as the Adams–Bohart model, have been developed to model the sorbent behavior in continuous settings [11]. Employing such models to understand the behavior at a small scale can ease transition to full-scale industrial processes. By quantifying the maximum adsorption capacity, rate of adsorption, and breakthrough curve (a plot of concentration of the adsorbate in the effluent vs. time) [12], the operation of adsorption columns can be optimized. Rate law and kinetic constants can also be used to normalize columns to account for varying volumes, masses, and flow rates.

A variety of materials are known to adsorb selenium species effectively including mesoporous materials [13], graphene based composites [14], and a variety of metal oxides [15], layered-double hydroxides [16], metal-organic frameworks (MOFs) [17], MgO nanosheets [18], carbon nanotubes [19], and bioremediation-based methods [20–22]. Other methods, such as mineralization [23], capacitive deionization, and catalyzed reduction [24] are also known but not the focus of the current work. As separation materials are developed, they need to be aligned with a more sustainable future in mind. There are twelve principles of circular chemistry of particular importance [25]. In this work, the adsorbent material addressed two concerns in this area, namely the collection and use of waste and the optimization of resource efficiency. The sorbent was synthesized from waste tire material and iron chloride, FeCl3, both of which are waste products from different sectors. Additionally, the sorbent was able to be regenerated and reused, decreasing the need for large amounts of adsorbent material.

A tire-derived carbon has recently been demonstrated in our laboratory as an effective support for iron oxide magnetic adsorbents for selectively removing Se(IV) ions from contaminated water [9]. While it is most effective in a fluidized bed, due to the ease of removal of the magnetized particles, determining its viability in fixed-bed systems is still of importance. This is due to the lower cost, simplicity, and predictable behavior in the removal of low concentration pollutants. Herein, the physicochemical characteristics of this composite adsorbent were characterized, and its kinetic behavior with respect to Se(IV) uptake in a fixed-bed setting was determined and analyzed to inform its effective use.

#### **2. Experimental Methods**

#### *2.1. Carbon Supported Magnetic Nanoparticle Adsorbents (C-MNA) Synthesis*

Tire derived carbon was synthesized as described in previous work [9]. To enhance its activity, it was ground with potassium hydroxide pellets (Sigma Aldrich, >85%) in a 1:4 weight ratio. The mixture was then placed in a furnace and heated at a 10 ◦C/min ramp rate to 800 ◦C, where it was kept under nitrogen atmosphere for 1.5 h. The mixture was then removed, cooled to ambient temperature, and neutralized with 3M hydrochloric acid (EMD Millipore).

C-MNA was synthesized by suspending 0.5 g of activated carbon in 120 mL of deionized water containing 13 mmol of iron sulfate, FeSO4·7H2O (Sigma Aldrich, ≥99.0%) and 15.6 mmol FeCl3·6H2O (Sigma Aldrich, ≥97.0%). The mixture was then sonicated for 5 min and stirred at 70 ◦C for 1 h. 3M sodium hydroxide solution was added in excess to maximize iron oxide nanoparticle (FeNP) precipitation. C-MNA was then vacuum filtered and separated from the excess iron chloride mother liquor via magnetic filtration and washing with DI water.

#### *2.2. Column Set Up*

A solution containing 5 ppm Se(IV) at pH 5 was prepared by dissolving sodium selenite (Sigma Aldrich, 99%) into deionized water. The pH was adjusted with dilute hydrochloric acid. As demonstrated in Figure 1, the column was prepared by using a standard jacketed borosilicate glass column with a radius of 0.5 cm and a bed height of 13 cm, lining the bottom with approximately 2.5-cm height of cotton, adding sand containing 5 weight % of C-MNA, and finally adding 2.5-cm height of pure sand on the top. A peristaltic pump was used to deliver the Se(IV) solution at varying flow rates. The effluent was collected manually in increments of 9.5 mL and sampled for analysis via inductively

coupled plasma optical emission spectroscopy (ICP-OES) [26]. The column was prepared for trials by flowing 20 mL of dilute hydrochloric acid of pH 5 through the column before adding the Se(IV) solution. After the adsorbent was fully saturated with Se(IV), the column was flushed with 1 M sodium hydroxide solution of pH 11 to desorb Se(IV). This demonstrates that the adsorbents can be recycled as shown in Figure 1.

**Figure 1.** Schematic of the bench scale column set up. After Se (IV) removal from the contaminated stream, 1M NaOH solution can be flushed into the column to desorb Se (IV) in the form of sodium selenite by exiting through the top of the column.

#### *2.3. Characterization*

X-ray powder diffraction (XRD) patterns were collected using a PANalytical Empyrean instrument (Malvern, UK) with a Cu Kα radiation. All data were processed with HighScore Plus (Malvern, United Kingdom). Brunauer-Emmett-Teller (BET) surface areas and pore-size distributions were determined from nitrogen adsorption isotherms at 77 K using Autosorb-1 from Quantachrome (Anton Paar GmbH, Austria). The pore-size distributions and pore volumes were calculated from the DFT/Monte Carlo method using the QSDFT adsorption branch model. ICP-OES compositional analysis was performed to determine selenite removal from the solution using a Thermo Fisher iCAP Model 7400 ICP-OES Duo. The ICP-OES has a minimum selenium detection limit of 0.77 μg/L at wavelength 196 nm. During measurements, the linear standard curve had an R<sup>2</sup> value of 1 and a limit of detection of 0.049 ppm. A Zeiss Merlin VP scanning electron microscopy (SEM) (White Plains, NY, USA) operated at 3 kV and a Hitachi HD-2300A scanning transmission electron microscope (STEM) with a field emission source operated at 200 kV in bright-field imaging mode at a 2.1 Å resolution, were used to characterize the surface morphologies of the samples.

#### *2.4. Kinetic Model*

The kinetic equations were derived from the kinetic rate law equation for pseudo-second order reactions, as a variety of carbon based materials display this general adsorption behavior [27,28]. Equation (1) assumes that the adsorption capacity is correlated to the number of active sites on the surface.

$$\frac{dq\_t}{dt} = k(q\_t - q\_t)^2\tag{1}$$

*Water* **2020**, *12*, 1234

In this equation, *qe* and *qt* refer to the equilibrium concentration and the concentration at time *t*, respectively. The *k* parameter is the kinetic constant. With the boundary conditions of *t*<sup>0</sup> = 0, *t* = *t*, *q*<sup>0</sup> = *q*(*t* = 0) = 0, and *q*(*t*) = *qt*, a solution was derived and rearranged to a linear form as seen in Equations (2)–(5). Equation (5) is then used to fit experimental data.

$$\frac{1}{q\_c - q\_t} - \frac{1}{q\_c} = kt \tag{2}$$

$$\frac{q\_t}{q\_c^2 - q\_c q\_t} = kt\tag{3}$$

$$q\_t = \frac{ktq\_c^2}{1 + ktq\_c} \tag{4}$$

$$\frac{1}{q\_t} = \left(\frac{1}{kq\_\varepsilon^2}\right)\frac{1}{t} + \frac{1}{q\_\varepsilon} \tag{5}$$

The derivation of the Adams–Bohart model starts with a modified chemical rate law expression as shown in Equation (6) [29].

$$\frac{\partial q}{\partial t} = k\_{AB}\mathcal{C}(q\_0 - q) \tag{6}$$

The analytical solution provided by Adams and Bohart is shown in Equations (7) and (8) [30].

$$\frac{\mathcal{C}}{\mathcal{C}\_0} = \frac{\exp(a)}{\exp(a) + \exp(\beta) - 1} \tag{7}$$

$$
\alpha = k\_{AB}C\_0 \left( t - \frac{Z}{\upsilon} \right); \beta = \frac{k\_{AB}\rho\_p q\_0 Z}{\upsilon} \frac{1 - \varepsilon}{\varepsilon} \tag{8}
$$

Here, the following simplifications are made [28]:

1. *exp*(α),*exp*(β) 1

*v*

The quantity 1 at the denominator of Equation (7) is thus regarded as insignificant.

2. *<sup>t</sup> <sup>Z</sup>*

Since the time of the experiment far outweighs the residence time, the residence-time term can be ignored.

The following definitions are also used:


Equation (7) is then simplified into the recognizable form in Equation (9) [12]. Although this equation is used to calculate the breakthrough time, or the time before the effluent concentration exceeds acceptable levels [12], the Adams–Bohart equation can be used to predict changes in the effluent concentration over time.

$$\ln\left(\frac{\mathbf{C}\_o}{\mathbf{C}\_B} - 1\right) = \frac{k\_{AB}N\_oZ}{\mu} - k\_{AB}\mathbf{C}\_o\mathbf{t}\_B\tag{9}$$

Equation (9) can be rewritten with the following adjustments as seen in Equation (10).

$$\frac{N\_0 Z}{u} = \frac{q\_0 M}{Q}$$

We can rewrite *u* as *<sup>Q</sup> <sup>A</sup>* and then rewrite *N*0*ZA* as *qoM* since both reduce to the amount of solute adsorbed in the system.

$$\ln\left(\frac{C\_o}{C\_B} - 1\right) = \frac{k\_{AB}q\_oM}{Q} - k\_{AB}C\_ot\_B\tag{10}$$

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

The C-MNA adsorbent utilized in this study was fully characterized, including SEM, EDS, and magnetic property measurements in previous work [9]. XRD patterns of the column mixture of adsorbent and inert packing material shown in Figure 2 indicates the presence of both C-MNA and sand in the mixed material. SEM images of sand, C-MNA and the composite mixture are shown in Figure 3. SEM images of C-MNA and the porous carbon surfaces are clearly seen. Additionally, the large sand particles are mixed with smaller C-MNA particles homogenously in the mixture. Finally, a BET analysis was performed to determine the surface area of the different materials as another assurance of C-MNA mixing with the inert packing material. In Figure 4, surface areas of 6.8 m2/g for sand, 638 m2/g for C-MNA, and 45.5 m2/g for the mixture were determined, confirming that the high surface area C-MNA was mixed with inert sand packing material.

**Figure 2.** X-ray powder diffraction (XRD) patterns of column materials sand, C-MNA, and a mixture of the two.

**Figure 3.** SEM images of (**a**) sand; (**b**) C-MNA adsorbent material; and (**c**) mixture of sand and C-MNA composite material.

**Figure 4.** BET analysis of (**a**) sand; (**b**) C-MNA adsorbent material; and (**c**) mixture of sand and C-MNA used in the column.

In continuous flow systems, kinetics data are normalized for comparison to the residence time. The residence time, or the total time that the solution is in contact with the sorbent, is calculated by dividing the total bed volume by the volumetric flow rate.

$$\text{Residence Time} = \frac{\text{Bcd Volume}}{\text{Volumetric Flow Rate}} \tag{11}$$

By expressing kinetics relative to residence time, we can make sure that the effects of bed volume size or flow rate are not omitted. In our column runs, we assumed that there was good radial mixing and no axial dispersion. As seen in Figure 5, the composite adsorbent steadily adsorbed Se(IV) until reaching a plateau between 0.4 and 0.5 mg Se per g adsorbent, suggesting monolayer formation on the carbon surface of the adsorbent [29]. Subsequent adsorption is thought to be a result of mass transfer of Se(IV) into the mesopores of the carbon support structures and formation of a multilayer on the surface of the iron nanoparticles, since the adsorption profile fits the two stage adsorption characteristic of a Type IV adsorption isotherm [19]. At a higher flow rate of Se(IV) solution, the initial rate of Se(IV) adsorption is higher than Se(IV) adsorption at a lower flow rate because there is a greater concentration gradient between the bulk concentration and the concentration adsorbed [27], which means a greater driving force for mass transfer.

**Figure 5.** Se adsorption of column at differing flow rates to determine breakthrough times.

Sand was mixed with adsorbent because the small grain size of the adsorbent caused packing issues. Previous studies used sand as an inert medium to house sorbents [26]. However, sand is known to physisorb selenium as well as other contaminants [31,32]. To quantify the adsorption capacity of the sand, a column was filled with pure sand (still including the cotton layer at the bottom), and water containing 5 ppm Se(IV) at a pH of 5 was flowed through the column. The concentration of the effluent was compared to the initial concentration to calculate how much Se(IV) was adsorbed per gram of sand. At a flow rate of 2.92 mL/min and a pH value of 5, the adsorption capacity was low (6.8 μg/g), as shown in Figure 6, compared to the total of ~500 μg/g in the presence of the composite adsorbent. Based on these tests, the adsorption of sand was determined to be negligible. Instead, the adsorption is attributed to the embedded FeNPs and surface adsorption on the carbon support. Although correction for Se(IV) adsorption by sand did not change fundamental conclusions about the reaction rate or adsorption behavior, the equilibrium concentrations and kinetics constants were affected. As a result, all data subsequently shown and discussed contain corrections made for sand adsorption.

**Figure 6.** Adsorption of Se in sand as a control for inert column materials present.

#### *3.1. Rate Law*

To further elucidate the mechanism and kinetics of the composite adsorption of previously studied carbon and iron-based adsorbents, the adsorption rate was modeled using a second-order kinetic rate equation, as shown in Equation (7). From Figure 7, the adsorption kinetics is strongly consistent with pseudo-second order reactions and corroborates trends observed in kinetic data taken from other studies [33,34]. The equilibrium concentration and kinetic constant were also obtained from this model and subsequently used in Adams–Bohart kinetic model calculations (see Table 1).

**Table 1.** Equilibrium concentration and kinetic constants from 2nd order kinetic model fitting.


Previous batch studies report the maximum adsorption concentration achieved was 1.14 mg/g [9]. When the sorbent is mixed with inert sand in a fixed-bed column, several factors could lead to a decrease in the observed uptake of Se. Channeling could restrict water flow to certain sections of the column, preventing the sorbent in that region from interacting with selenite. The inert medium could cover the mesopores in the composite sorbent, again restricting adsorbate flow to the iron nanoparticles housed within the mesopores. In addition, batch tests are performed with the adsorbent free flowing in solution, thus increasing the wettability of the sorbent in comparison to the fixed column. Finally, due to lower concentrations of the adsorbate near the bottom of the column, preventing the full utilization of the adsorbent for selenite removal, the adsorbent in that area could require extended time to reach equilibrium.

**Figure 7.** Linear regression fit of data to second order kinetics equation in agreement with expected pseudo-second order behavior.

#### *3.2. Adams–Bohart Model*

The Adams–Bohart model utilizes the following assumptions: (1) flow rate is constant, (2) absence of axial dispersion, (3) behavior matches the rectangle (irreversible) isotherm (i.e., highly favorable adsorption), and (4) adsorption rate follows second-order reaction kinetics [30]. The data were fit with a linear regression line as shown in Figure 8, and the rate constant (*kBA*) was calculated from the m and b terms of the linear regression equation (y = mx + b).

The literature shows two forms (Equations (9) and (10)) of the Adams–Bohart model [11,30]. Although both models yield kinetic constants while maintaining dimensional homogeneity, they differ slightly in the method by which the kinetic constant is derived. One equation can be used to calculate the constants by equilibrium adsorption per unit volume, while the other can be used to determine the constants on a per unit mass basis. The two forms of the same model would be considered identical if the contents of the column were homogenized and uniform [30]. However, the presence of both an inert medium and the composite sorbent in our column means that local densities and adsorption capacities could vary. In our case, we chose to include the mass and volume of sand in our calculations in order to obtain more precise values. As discussed above, further tests indicated that the sand did play a minor role in the adsorption of Se(IV).

**Figure 8.** Linear regression fit of data to Adams–Bohart model. Values from these regressions were used in the calculation of kinetic constants in Tables 2 and 3.


**Table 2.** Kinetic constants calculated from the slope.



Linear regression fits were performed on the data in Figure 8, providing two equations to determine kinetic parameters from. For a resonance time of 3.5 min and 6 min, the regressions were y = −0.0502x + 2.1309 and y = −0.0617x + 2.4508, respectively. Since the rate constant appears in both the m and b terms of the linear regression equation, each fitting could yield two rate constants. Between these two terms the slope (m) produced more consistent parameter values, making them more useful for the predictions of the fixed-bed behavior [30]. A better agreement with the observed column adsorption behavior was seen by only including the initial adsorption of the column. This is reflective of the constraints of the model, as the Adams–Bohart model is typically used to depict breakthrough curves rather than overall adsorption curves [12]. As seen in Tables 2 and 3, many of the kinetic constants derived from the m and b components across all forms of the Adams–Bohart model and all residence times were similar. Only the values calculated on a per unit mass basis (Equation (10)) deviated significantly, potentially due to variance in equilibrium concentration calculations or sand adsorption behavior. The differing rate constants from the m and b terms potentially originate from

multiple kinds of adsorption taking place or multiple reactions occurring. Another possibility is the existence of a pseudo-solution which can be eliminated by comparing the kinetic rate constants from both equations and over both runs. When compared to values obtained from experimental data, the theoretical values calculated from concentration measurements resulted in rate constants that displayed a small initial deviation from the experimentally obtained values that subsequently decreased as the column reached full saturation.

#### **4. Conclusions**

Column kinetics were determined by first analyzing the rate law to obtain equilibrium concentrations. Subsequently, this data was fitted to the Adams–Bohart model, enabling the calculation of critical data on column kinetics of C-MNA adsorbents in a fixed-bed system. Since the data displayed strong correlation to a pseudo-second order linear regression fit, the equilibrium concentrations could be applied to an Adams–Bohart providing a better understanding of column behavior. Within a continuous-flow setting, the Adams–Bohart model predicts column behavior with reasonable accuracy, given certain limitations. Given the mixture of semi-inert sand and composite adsorbent, calculations were found to be more accurate when the mass and volume of the sand were included in the model. Given the lack of literature on the Se(IV) adsorption behavior in a fixed-bed column, this analysis provides valuable insight as corporations and governments continue to require new technologies for waste and contaminated water processing. This work provides insights to enable further work on mixed pollutant removal in addition to industrial scaling of this column.

Finally, although magnetism has not been taken advantage of in the current work, it can be used to create macroporosity in a sorbent bed and thereby increase mass transfer, while reducing pressure drop. It is well known that magnetically stabilized beds can have the benefits of fixed beds, in terms of simplicity and high separation efficiency, and fluidized beds with respect to high mass-transfer rates and low pressure drop. Future work will be focused on removing Se(IV) from real wastewater using magnetically stabilized sorbent beds with magnetic sorbent based on tire-derived carbon.

**Author Contributions:** Conceptualization, A.Y., S.F.E., C.T. and M.P.P.; methodology, A.Y., S.F.E. and C.T.; formal analysis, A.Y., S.F.E. and C.T.; resources, M.P.P.; data curation, A.Y. and S.F.E.; writing—original draft preparation, A.Y., S.F.E.; writing—review and editing, C.T., M.P.P.; supervision, M.P.P.; funding acquisition, M.P.P. All authors have read and agreed to the published version of the manuscript.

**Funding:** The synthesis of carbon composite materials was sponsored by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, Materials Sciences and Engineering Division. This work was supported in part by the U.S. Department of Energy, Office of Science, Office of Workforce Development for Teachers and Scientists (WDTS) under the Science Undergraduate Laboratory Internship (SULI) program. S.F.E. is grateful for a fellowship from the Bredesen Center. We would like to thank Dr. Rich Lee, RJ Lee Group for providing tire derived carbon.

**Acknowledgments:** This manuscript has been authored by UT-Battelle LLC under Contract No. DE-AC05-00OR 22725 with the U.S. Department of Energy. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes. The Department of Energy will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan (http://energy.gov/downloads/doe-public-access-plan).

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **Abbreviations**

#### **Relevant Variables**


#### **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* **Au**/**ZnO Hybrid Nanostructures on Electrospun Polymeric Mats for Improved Photocatalytic Degradation of Organic Pollutants**

#### **Laura Campagnolo 1,2, Simone Lauciello 3, Athanassia Athanassiou <sup>1</sup> and Despina Fragouli 1,\***


Received: 31 July 2019; Accepted: 22 August 2019; Published: 28 August 2019

**Abstract:** An innovative approach for the fabrication of hybrid photocatalysts on a solid porous polymeric system for the heterogeneous photocatalytic degradation of organic pollutants is herein presented. Specifically, gold/zinc oxide (Au/ZnO)-based porous nanocomposites are formed in situ by a two-step process. In the first step, branched ZnO nanostructures fixed on poly(methyl methacrylate) (PMMA) fibers are obtained upon the thermal conversion of zinc acetate-loaded PMMA electrospun mats. Subsequently, Au nanoparticles (NPs) are directly formed on the surface of the ZnO through an adsorption dipping process and thermal treatment. The effect of different concentrations of the Au ion solutions to the formation of Au/ZnO hybrids is investigated, proving that for 1 wt % of Au NPs with respect to the composite there is an effective metal–semiconductor interfacial interaction. As a result, a significant improvement of the photocatalytic performance of the ZnO/PMMA electrospun nanocomposite for the degradation of methylene blue (MB) and bisphenol A (BPA) under UV light is observed. Therefore, the proposed method can be used to prepare flexible fibrous composites characterized by a high surface area, flexibility, and light weight. These can be used for heterogeneous photocatalytic applications in water treatment, without the need of post treatment steps for their removal from the treated water which may restrict their wide applicability and cause secondary pollution.

**Keywords:** nanocomposite fibers; photocatalysis; mineralization; water remediation; organic pollutants

#### **1. Introduction**

The growing demand for clean water urges an improvement in commonly used methods for the wastewater management [1]. Depending on the type of water contamination, these conventional wastewater treatments can be physical, chemical, and biological processes [2]. However, these treatments present several limitations from a technological point of view, including the necessity for wastewater pre-treatment steps, secondary pollution, limited practicability, and often, low performance when emerging organic pollutants are involved [2–4]. For the removal of emerging and recalcitrant pollutants, new strategies have been proposed in combination with the available treatments. Among them, advanced oxidation processes (AOPs) can be used to ideally destroy organic pollutants from the contaminated water, thanks to the in situ production of strong oxidant species as the result of the interaction of the developed materials with solar, chemical, or other forms of energy [5–8].

Heterogeneous photocatalytic oxidation is one of the most promising AOPs for oxidizing organic pollutants into more biodegradable and less harmful compounds [5]. Typically, wide band gap, nano-, or micro-scale sized semiconductors are used as photocatalysts able to absorb incident phonons causing the subsequent formation of a separation charge with a conduction-band electron and a valence-band hole. The photo-excited carriers can generate hydroxyl radicals and other reactive oxygen species (ROS), which react with the organic molecules present in the contaminated water causing their degradation, assuring an eco-compatible process since the oxidation occurs in mild conditions [5].

The semiconductor particles are usually dispersed in the contaminated water to ensure a high exposed surface area in contact with the pollutants, and therefore, a high volumetric generation rate of ROS. However, the main limitation in their use for water remediation applications is the demand of energy and time consuming additional steps for their recovery from the treated water. To avoid these post-treatment steps, the photo-catalyst powders can be immobilized on suitable solid supports [9,10]. To this aim, polymeric supports have been recently introduced due to their low weight, flexibility, and easy conversion in different morphologies [11–16]. Among them, fibrous polymeric mats have high surface area, which allows an enhanced interaction with the polluted water, and their preparation process is relatively simple and easily scaled up [9,17].

For the fabrication of the functional semiconductor based polymeric materials, an innovative approach has been recently proposed, dealing with the direct synthesis of NPs inside different polymeric matrices through a solid-state thermally, chemically, or photo-induced reaction [13,17–19]. Compared to traditional mixing processes, this approach offers the advantageous formation of NPs homogeneously distributed not only in the whole volume of the polymer matrix, but also exposed on its surface. Furthermore, limitations such as the complex rheology of polymer/NP solutions, which can limit their electrospin-ability and therefore the formation of fibrous nanocomposites, can be avoided [13,17]. Among the diverse types of nanocomposites developed so far following this approach [13,17,20], the ZnO-based mats is a promising system for the water remediation through the photocatalytic approach. In fact, ZnO is of great interest not only because it is environmental friendly, non-toxic, of low-cost, and abundant, but also because of its optoelectronic properties [9,21–24]. In particular, the n-type ZnO is a direct wide band gap semiconductor (band gap energy, Eg = 3.37 eV), characterized by a high exciton binding energy (60 meV), which allows efficient excitonic emission even at room temperature [25]. However, due to the high photo-generated electron–hole pair recombination, the photocatalytic performance and, therefore, the utilization of the ZnO in wastewater purification, can be limited. For this reason, the enhancement of the separated electron–hole lifetime is one of the factors on which the efforts are focused in order to improve the ZnO photocatalytic performance [26–28]. This aspect can be achieved through various methods, such as the modification of the morphology of the ZnO nanostructures [22,29], the incorporation of metallic [30,31] and non-metallic dopants [32] in their crystalline structure, and the preparation of hybridized structures with other materials [21,22,33,34].

Concerning the last method, the combination of ZnO NPs with noble metal NPs has been recently considered an effective route to enhance the local density of states at the metal–semiconductor interface, prolonging thus the electron–hole separation lifetime [21,33]. Specifically, the use of Au NPs to modify the ZnO surface has been proven to improve the photocatalytic performance of the semiconductor due to the surface plasmonic resonance effect (which allows the establishment of the Schottky barrier simplifying the charge carrier separation), the absorption in the visible range of the electromagnetic spectrum, and the ROS generation [35,36].

Considering the above, the development of Au/ZnO-based PMMA fibrous mats is here presented and their application in the remediation of organic aqueous pollutants through the heterogeneous photocatalytic approach is thoroughly explored. Specifically, ZnO NPs are synthetized by a thermally-induced solid state reaction directly in the electrospun polymeric fibers, obtaining an easily handleable flexible solid porous material. The subsequent dipping of the PMMA/ZnO composite mats in Au precursor solutions of different concentrations and their thermal treatment result in the direct formation of Au NPs on the surface of the fibers and, therefore, on the exposed ZnO NPs, forming a high metal–semiconductor interfacial area. Different concentrations of the Au precursor solutions are used to explore the effect of the amount, size, and localization of the formed Au NPs on the

ZnO/PMMA fiber surface. Moreover, the photocatalytic performance of the different nanocomposite mats under UV light irradiation is investigated, focusing on the degradation of the model molecule MB and of the emerging pollutant BPA. In particular, it is demonstrated that the presence of Au NPs affects significantly the photocatalytic performance of the PMMA/ZnO mat for both the organic pollutants. Therefore, with the presented straightforward fabrication method, it is possible to form light-weight and flexible polymeric fiber mats of high surface area that contain functional NP combinations for advanced photocatalytic applications. This valuable alternative to the existing systems paves the way to an innovative route for the fabrication of multifunctional membranes for water purification without the need of costly and time consuming post treatment separation processes as in the case of powders.

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

#### *2.1. Materials*

PMMA (average Mw ~350 kDa), zinc acetate dihydrate (Zn(CH3CO2)2, 99,999%), *N*,*N*-dimethylformamide (DMF, ≥99.8%), chloroauric acid trihydrate (HAuCl4·3H2O, 99.999%), MB, BPA (≥99%), acetone (≥99.5%), ethanol (EtOH, ≥99.8%), sodium hydroxide (NaOH, ≥98%, anhydrous), hydrochloridric acid (HCl, 37%), and nitric acid (HNO3, 70%), were purchased from Sigma Aldrich, Italy. All chemicals were used without any further purification.

#### *2.2. Preparation of PMMA*/*ZnO Composite Mats*

The nanocomposite fibers were prepared as described elsewhere [17]. In brief, 1.5 g of PMMA were added in 10 mL of DMF and left under stirring at 50 ◦C until the complete dissolution; then 1.1 g (c.a. 40 wt %) of Zn(CH3CO2)2 were added to the polymer solution and stirred at 40 ◦C, until a clear solution is obtained. The electrospinning technique was then used in a vertical setup to obtain the PMMA/Zn(CH3CO2)2 composite mats. The prepared solution was placed in a 10-mL syringe with a 22 G needle and the flow rate was set to 600 <sup>μ</sup>L·h−1. The voltage applied to the syringe needle was 23 kV and the grounded collector was kept at a distance of 16 cm from the needle. The electrospun fibers were then dried to remove the eventual residual solvent for 8 h under dynamic vacuum. The in situ growth of the ZnO NPs on the electrospun mats was achieved by placing the electrospun composite mats in a convection oven for 48 h at 110 ◦C, through the thermal decomposition of Zn(CH3CO2)2.

The obtained PMMA/ZnO composite mats were then washed in 10 mL of a H2O/EtOH mixture (8:2 *v*/*v*) for 24 h and dried at 60 ◦C for 4 h. Following the methodology described to Section 2.4, it is defined that the total zinc loss during this washing step was around 1.170 ± 0.160 wt % of the total zinc. This can be attributed either to the loss of ZnO NPs not perfectly fixed on the surface of the fibers, or, also, to the water-soluble zinc acetate, which was not fully converted during the synthesis.

#### *2.3. Preparation of PMMA*/*ZnO-Au Composite Mats*

Three PMMA/ZnO composite mats (40 mg each) were dipped in 20 mL of a H2O/EtOH mixture (8:2 *v*/*v*) of HAuCl4·3H2O, which was previously neutralized to pH 6.5 by NaOHaq 0.1 M for 24 h in dark under gentle stirring. The aqueous-based HAuCl4 3H2O solution was prepared at different initial concentrations of gold precursor—1.5 mM, 3 mM, and 4.6 mM—to explore the effect of the different amounts of Au on the photocatalytic performance of the PMMA/ZnO-Au mats. The composite mats were then collected and washed three times with water. The reduction of the gold ions was obtained by the subsequent thermal treatment of the mats at 60 ◦C for 4 h, obtaining three PMMA/ZnO-Au samples named PMMA/ZnO-Au1, PMMA/ZnO-Au3, and PMMA/ZnO-Au6, respectively. Each sample was then divided in four similar pieces of 10 mg each for further analysis.

#### *2.4. Characterization*

The fiber morphology was investigated by a scanning electron microscope (SEM, JEOL JSM 6490LA, Tokyo, Japan) and a high-resolution scanning electron microscope (HR-SEM, JEOL JSM 7500FA, Tokyo, Japan), equipped with a cold field emission gun applying an accelerating voltage of 15 kV. The specimens were previously coated with a 10-nm-thick carbon layer by a carbon coater (Emitech K950X, Quorum Technologies Ltd., East Susse, UK). The morphology of the ZnO and Au NPs in the composite mats was investigated by a transmission electron microscope (TEM, JEOL JEM 1400 Plus, Tokyo, Japan) operating at an acceleration voltage of 120 kV. To do so, the polymeric fibers were dissolved in acetone and the suspension was sonicated and centrifuged in order to separate the NPs from the polymer. Then, 30 μL of the suspension were deposited on a copper grid (CF300-Cu-UL, Electron Microscopy Science, USA) for further investigation. High-angle annular dark field scanning TEM (HAADF-STEM) images and energy dispersive X-ray spectroscopy (EDS) maps were acquired with a high-resolution FEI Tecnai G2 F20 (Oregon, USA) equipped with a Schottky field emission gun (FEG). The average size and the size distribution of the fibers and NPs were explored using the open source Fiji/ImageJ software.

The X-ray diffraction (XRD) analysis was performed on a PANalytical Empyrean X-ray diffractometer with a 1.8 kW CuKα ceramic X-ray tube (λ = 1.5418 Å) and a PIXcel3D area detector (2 <sup>×</sup> 2 mm2), operating at 45 kV and 40 mA. The diffractograms were recorded in parallel-beam geometry and symmetric reflection mode, for a 2θ range from 25◦ to 80◦, with a step time of 383.67 s, and a step size of 0.026◦. The average crystallite size was calculated by using the Debye–Scherrer equation (details in Supplementary Materials). The zinc and the gold content in the composite mats was obtained by ICP-OES analysis (iCAP 6300, Thermo, USA) of the digested solid samples. Specifically, 2.5 mL of aqua regia (HCl/HNO3, 3:1) were added to 2.5 mg of the composite fibers and the solid degradation reaction was performed by using a microwave digestion system (MARS Xpress, CEM) at 180 ◦C for 15 min. The samples were then diluted with milliQ water up to 50 mL and filtered through polytetrafluoroethylene (PTFE) syringe filters (diameter 15 mm, pore size 0.45 μm, Sartorius). In addition, to evaluate the loss of ZnO and Au NPs from the composite mats, the ICP-OES analysis was performed on the liquids after the photocatalytic degradation process. The samples (125 μL) were previously treated with 2.5 mL of aqua regia, and then they were diluted up to 25 mL with milliQ water and filtered through PTFE syringe filters (diameter 15 mm, pore size 0.45 μm, Sartorius, Germany).

The optical properties of the nanocomposite fibers were evaluated by diffuse-reflectance measurements in the range between 200 and 800 nm, using a Varian Cary 6000i (Agilent, USA) UV-visible-NIR spectrophotometer equipped with integrating sphere. The energy bandgap (Eg) of ZnO and ZnO/Au NPs was then extrapolated by applying the Kubelka-Munk method to the reflectance results (see Supplementary Materials). The Raman spectra of the composite mats were acquired by a Raman LabRam HR800 (Horiba Jobin-Yvon Inc., France) spectrometer equipped with a built-in microscope with objectives 10 × (NA 0.25) and 50 × (NA 0.75), using the 632.8 nm He-Ne laser excitation. The experimental setup consists of a grating 600 lines·mm−<sup>1</sup> with a spectral resolution of approximately 1 cm<sup>−</sup>1.

#### *2.5. Photocatalytic Performance*

The photocatalytic performance of the developed materials was evaluated by immersing 10 mg of the mats in quartz cuvettes filled with 3 mL of MB or BPA aqueous solutions (0.0125 mM and 0.044 mM, respectively). To evaluate the adsorption of the organic molecules on the developed mats, experiments in dark conditions were performed overnight. For the photocatalytic experiments, the solutions containing the samples were placed at a distance of 10 cm under a UVA lamp emitting at a wavelength range from 315 nm to 400 nm (937 μW/cm<sup>2</sup> at 365 nm, 10 cm from source) connected to a climatic chamber (ICH 110 L, Memmert, Germany). The UVA irradiance (μW/cm2) of the UV light source was measured by using a combined photo-radiometric probe (LP 471 P-A, Delta Ohm, Italy) at the fixed experimental distances between the source and the samples. Before starting the UV irradiation, the mats were maintained in the dark for 60 min. The decrease of the organic molecule concentrations was monitored by recording the UV-vis absorption spectra at specific time intervals. The test was also performed for a reference sample, namely a quartz cuvette filled with the MB or BPA solution, in order to estimate the self-degradation of the pollutants under the UV light irradiation.

The photocatalytic degradation efficiency (*DE*%) was obtained from the Equation (1), while the first-order rate constant *(k*1, min−1) was calculated by fitting the experimental data of the first 5 h of irradiation with the pseudo-first order model expressed in Equation (2)

$$
\Delta DE \% = \left(1 - \frac{C}{C\_0}\right) \cdot 100 \tag{1}
$$

$$\ln\left(\frac{C\_0}{C}\right) = k\_1 t \tag{2}$$

in which *C*<sup>0</sup> is the initial pollutant concentration and *C* is the concentration at t irradiation time. The organic molecule concentrations were determined by using two calibration curves with linear regression (*R*<sup>2</sup> values of 0.9983 and 0.9998 for MB and BPA, respectively). For the MB, the calibration curve was obtained by monitoring the absorbance of the MB at the wavelength of maximum absorbance (λmax = 664 nm) with concentrations ranging from 0.00087 to 0.0125 mM. For the BPA calibration curve, the aqueous solution concentrations ranged from 0.0031 to 0.05 mM, and the absorbance variation was monitored at 276 nm.

Before and after the photocatalytic degradation of the organic molecules, the total organic carbon (TOC) of the solutions was analyzed by a membraPure uniTOC lab analyzer, Gemany. The total amount of organic bound carbon in the aqueous samples was measured by mean of a UV/reagent promoted oxidation into carbon dioxide (CO2) which was then quantified by the non-dispersive infrared detector. After the photocatalytic degradation experiment, the MB and BPA aqueous solutions were collected from the quartz cuvettes and diluted with MilliQ water, in order to reach 15 mL in each case. To remove any possible solid residue, the aqueous solutions were centrifuged and then filtered using a polytetrafluoroethylene (PTFE) syringe filters (diameter 15 mm, pore size 0.20 μm, Sartorius, Germany). The reagent used was an acidic aqueous solution of sodium persulfate (10%*W*/*V*) and phosphoric acid (3%*V*/*V*). The inorganic carbon content was firstly removed and measured by the internal acidification step. The TOC value was used to evaluate the mineralization of the MB and of the BPA, and therefore, the efficiency of the photocatalytic degradation.

The stability of the composite mats was evaluated by repeating the photocatalytic experiments on the same samples over three cycles. The composite mats were washed before starting the subsequent cycle by dipping in clean water for 1 h under UV light irradiation (same conditions as previously described) in order to remove the possible adsorbed pollutants.

#### **3. Results**

The electrospun composite mats (photographs shown in Figure S1 in the Supplementary Materials) consisted of continuous defect-free polymeric fibers characterized by smooth surface, whose random distribution resulted in the formation of a porous material. Compared to the PMMA fibers (Figure S2), the PMMA/Zn(CH3CO2)2 fibers exhibited a higher mean diameter with a broader size distribution, while their morphology remained rather similar, in accordance with our previous work [17]. In particular, even if the electrospinning operating parameters were preserved, the mean diameter of the polymeric fibers increased from 1.5 ± 0.3 to 6.2 ± 3.7 μm (for PMMA and PMMA/Zn(CH3CO2)2, respectively). This difference was mainly due to the presence of the high content of Zn(CH3CO2)2 which causes the enhancement of the viscosity of the solution (from 1.57 ± 0.05 Pa·s to 4.10 ± 0.15 Pa·s) [17].

After the thermal treatment of the PMMA/Zn(CH3CO2)2 fibrous mats, ZnO NPs were formed, well distributed on the surface of the fibers but also in their bulk, as shown in Figure 1a and Figure S3. The content of ZnO NPs obtained after the zinc salt conversion was about 11.7 ± 0.9 wt.% with respect to the composite, as calculated by the ICP-OES analysis. Apart the ZnO NPs formation, it should be mentioned that the thermal treatment did not affect the overall size distribution of the polymeric fibers,

which remained unchanged (Figure S3b). In order to obtain the ZnO/Au NPs hybrid structures on the polymeric fibers, the PMMA/ZnO composite mats were dipped in the gold precursor aqueous solutions of different initial concentrations. The neutralization of the acidic gold precursor solution, performed to avoid the solubilization of the amphoter ZnO NPs, determined the presence of [AuCl2(OH)2] − species which were favorably adsorbed on the positively charged surface of the ZnO NPs.

**Figure 1.** High-resolution scanning electron microscope (HRSEM) images of the (**a**) PMMA/ZnO, (**b**) PMMA/ZnO-Au1, (**c**) PMMA/ZnO-Au3, and (**d**) PMMA/ZnO-Au6 composite mats. Insets report the details on the surface of the different composite mats. PMMA: poly(methyl methacrylate).

The subsequent thermal treatment led to the formation of Au NPs on the surface of the fibers, which appeared not to be further affected by these steps. In fact, as shown in Figure 1b–d, the formed Au NPs appeared well distributed on the surface of the composite mats. The Au NPs were mainly attached on the surface of the ZnO NPs, as shown in the corresponding insets of the Figure 1b–d, while their size was not significantly affected by the initial concentration of the gold precursor solution. However, a higher initial concentration of the gold precursor solution led to an increase in the coverage of the composite fibers by the Au NPs, which, for the highest concentration (PMMA/ZnO-Au6), tended to form nanowires and aggregate on the surface of the composite mats. The amount of Au NPs formed on the composite mats was approximately 1%, 3%, and 6% wt. of Au with respect to the composite, for PMMA/ZnO-Au1, PMMA/ZnO-Au3, and PMMA/ZnO-Au6 respectively, as defined from the ICP-OES analysis.

As shown in Figure 2a,b, the formed ZnO NPs had a spherical branched structure with an average diameter of 135 ± 38 nm, which, in accordance with the SEM analysis, corresponded to the NPs grown on the surface of the fibers (Figure S3a). The selected area electron diffraction (SAED) pattern, inset of Figure 2a, confirmed the attribution of the NPs to ZnO of hexagonal wurtzite phase. In Figure S4a it is possible to observe also the presence of smaller irregular ZnO NPs. This second type of particles was formed in the internal part of the polymeric fibers (inset of Figure S3a), which restrained their growth [17].

**Figure 2.** (**a**) Transmission electron microscope (TEM) image of the ZnO nanoparticles (NPs). The two insets show the selected area electron diffraction (SAED) pattern and the diameter size distribution of the NPs. Dark-field TEM images of (**b**) ZnO and of (**c**) ZnO decorated with Au NPs.

After the dipping in the gold precursor solutions of the PMMA/ZnO fibers and the subsequent heating process, small Au NPs were grown, mainly localized on the surface of the ZnO NPs, as shown in Figure 2c and in the EDS maps of Figure S5. In all the composite mats, the formed Au NPs had an average diameter of c.a. 8 ± 3 nm, 13 times smaller than the ZnO branched NPs. The lower mean diameter of the Au NPs compared to the ZnO allowed a high superficial contact between the two types of NPs and, therefore, an efficient charge separation.

The structures and the crystallites size of the ZnO NPs in the different composite mats were explored by XRD analysis (Figure 3). In all the composite mats, the ZnO NPs exhibited the characteristic diffraction peaks of the hexagonal wurtzite phase. In presence of Au NPs, a new diffraction peak appeared at 38◦, which was assigned to the (111) crystal plane of the cubic Au phase. For the PMMA/ZnO-Au6 composite mat, it was also possible to observe an additional diffraction peak at 44◦, related to the Au (002) lattice plane due to the higher amount of Au NPs on the surface compared to the other composite mats. The presence of Au did not cause shifts of the position of the ZnO diffraction peaks, and therefore the crystal structure of the ZnO NPs was not affected by the post-synthetic growth of the Au NPs. Furthermore, no extra phase or other peaks were observed, confirming the absence of crystalline impurities inside the samples. The average crystallite size of the ZnO NPs was approximated to 9 nm, as estimated by using the Debye–Scherrer equation.

The metal–semiconductor interface in the Au/ZnO NPs hybrid structure was investigated through diffuse reflectance and Raman spectroscopy. The diffuse reflectance spectra of the different composite mats are shown in Figure 4a. In all cases, it is possible to observe the characteristic absorption of the ZnO NPs in the UV region originated by the direct band gap transitions [26]. The presence of the Au NPs led to the addition of another absorption band with a maximum at 550 nm, attributed to the surface plasmon resonance of the Au NPs. The Eg of the ZnO was estimated by applying the Kubelka–Munk (KM) method to the data obtained by the reflectance spectra (details of the KM equation and graphical representation can be found in Supplementary Materials and Figure S6). The calculated Eg values demonstrate that the presence of the Au NPs did not cause any modification to the Eg of the ZnO present in the composite mats. In particular, the Eg was 3.25 ± 0.01 eV for the PMMA/ZnO, and 3.23 ± 0.02 eV, 3.23 ± 0.04 eV, and 3.25 ± 0.01 eV for the PMMA/ZnO-Au1, PMMA/ZnO-Au3, and PMMA/ZnO-Au6, respectively.

**Figure 3.** X-ray diffraction (XRD) patterns of the different composite mats. Stick reference pattern of ZnO (black) and of Au (red) are shown along the x-axis. The arrows indicate the detected Au diffraction peaks.

**Figure 4.** (**a**) Diffusive reflectance spectra of the developed composite mats. (**b**) Raman spectra of the polymeric fibers and of the PMMA/ZnO before and after the superficial modification with Au NPs.

The formation of the Au/ZnO NP hybrid structures was confirmed by Raman spectroscopy. Typically, the optical phonon modes predicted by the group theory for the hexagonal wurtzite ZnO structures are the polar Raman and infrared active modes A1, E1, the non-polar Raman active 2E2, and the silent 2B1 modes. The A1 and E1 modes are split in the transverse optical (TO) and longitudinal optical (LO) modes with different frequencies due to the manifestation of the macroscopic electric field of the LO phonons. The two E2 are split in the high- and low-frequency modes (E2 high and E2 low) which are associated with the oxygen atoms and the zinc sublattice, respectively [37,38]. In the Raman spectrum of the PMMA/ZnO composite mat (Figure 4b), all the peaks observed were attributed to the PMMA [39], whose assignments are listed in Supplementary Materials (Table S1), while the Raman active modes of the wurtzite ZnO were not evident, possibly due to the presence of the polymer matrix

and to the used experimental conditions [40]. It was possible to observe the characteristic modes of the ZnO only after the plasmonic enhancement provided by the contact with the Au NPs in the ZnO/Au hybrid systems. In fact, in the Raman spectrum of PMMA/ZnO-Au1 (Figure 4b), additional peaks appeared at 450 cm−<sup>1</sup> and 560 cm<sup>−</sup>1, which were respectively assigned to the E2 high and A1(LO) modes of the ZnO, proving the non-resonant surface enhanced Raman scattering of the ZnO optical modes in the proximity of Au NPs [37,41,42]. The Raman spectra of the PMMA/ZnO-Au3 and PMMA/ZnO-Au6 (Figure S7) show analogue enhancement of the Raman scattering of the ZnO optical modes.

The effect of different amounts of Au NPs on the photocatalytic degradation performance of the composite mats was then investigated using MB and BPA aqueous solutions. The MB is a cationic dye which can be found in wastewater deriving mainly from the textile, paper, and plastic industries. It is commonly adopted as a model pollutant for testing the photocatalytic performance of various materials under UV light irradiation. BPA is a recalcitrant aqueous pollutant classified as an endocrine disruptor, and is widely present in water due to its broad applications in the plastic industry [43]. In order to evaluate the adsorption of the organic pollutants on the fibrous composites, the mats were dipped in the organic pollutant solutions in dark conditions. As shown in Figure S8, there was a negligible MB adsorption on the fibers. The concentration of BPA slightly decreased over time in the presence of the PMMA/ZnO and of the PMMA/ZnO-Au with 1 wt % and 3 wt % Au contents. On the contrary, the PMMA/ZnO-Au6 showed a higher adsorption capacity compared to the other composites, adsorbing c.a. 39% of the initial BPA, due to the instauration of dispersive bonding interactions between the aromatic rings of the BPA and the gold surface, which was in a relatively high amount on the composite fibers [44,45].

Under UV irradiation (Figure 5), the concentration of both organic pollutants decreased in time only in the presence of the nanocomposite fibrous mats. In fact, their self-degradation was negligible, while for the pure PMMA fibrous mat there was no significant change in the concentration of the pollutants, confirming the absence of photocatalytic activity. When the composite mats were dipped in the solutions, it was possible to observe a more efficient photocatalytic degradation of the organic pollutants in the presence of the Au/ZnO hybrid structures.

**Figure 5.** Photocatalytic degradation curves of (**a**) methylene blue (MB) and (**b**) bisphenol A (BPA) in presence of the composite mats under UV light. Before the UV irradiation, the solutions were kept in dark for 60 min in the presence of the composite mats. The self-degradation of MB and BPA under UV is also presented.

As shown in Figure 5a, the PMMA/ZnO mat was able to decolorize 77% of the initial MB solution while in the presence of the PMMA/ZnO-Au1, the efficiency reached 88% after 20 h of irradiation. By increasing the content of Au, the MB photocatalytic degradation tended to slightly decrease, reaching 80% for the PMMA/ZnO-Au3, while for the PMMA/ZnO-Au6, it was lower compared to that

observed for the PMMA/ZnO (Figure 5a). The photocatalytic degradation curves of the pollutants followed a pseudo-first order kinetic model for all the developed composite mats (Figure S9), and the enhanced performance of the PMMA/ZnO-Au1 composite was also reflected on its *k*<sup>1</sup> for the first 5 h of irradiation. In fact, the photocatalytic degradation of MB with the PMMA/ZnO-Au1 reached the maximum value of *<sup>k</sup>*<sup>1</sup> 2.33 <sup>×</sup> <sup>10</sup>−<sup>3</sup> min<sup>−</sup>1, significantly higher than that obtained using the PMMA/ZnO, 1.44 <sup>×</sup> <sup>10</sup>−<sup>3</sup> min−<sup>1</sup> and the other composite mats, as shown in Table S3 in the Supplementary Materials.

A similar behavior is also observed for the photocatalytic degradation of BPA. As shown in Figure 5b, the PMMA/ZnO-Au1 induced a 63.5% decrease of the initial BPA concentration after 20 h of UV irradiation, remarkably higher compared to the BPA reduction achieved with the PMMA/ZnO mat (*DE*% of 34%). For higher amounts of Au NPs, the photocatalytic degradation efficiency was around 61% and 41%, for the PMMA/ZnO-Au3 and PMMA/ZnO-Au6, respectively. The *k*<sup>1</sup> values calculated by applying the first order kinetics model on the experimental data were in accordance with the photodegradation efficiency results. The PMMA/ZnO-Au1 displayed a faster reaction rate (1.12 <sup>×</sup> <sup>10</sup>−<sup>3</sup> min<sup>−</sup>1) compared to the PMMA/ZnO-Au3 and PMMA/ZnO-Au6 (0.96 <sup>×</sup> 10−<sup>3</sup> and 0.82 <sup>×</sup> <sup>10</sup>−<sup>3</sup> min−<sup>1</sup> respectively) and to the PMMA/ZnO (*k*<sup>1</sup> 0.69 <sup>×</sup> 10−<sup>3</sup> min−1). Despite the results obtained for the MB, the PMMA/ZnO-Au6 composite mat displayed a more effective reduction of the initial BPA concentration compared to the PMMA/ZnO mat. However, this is not attributed to the photocatalytic degradation but to the high adsorption of the BPA on the surface of the Au NPs, as demonstrated with the adsorption experiment (Figure S8b).

For both the pollutants studied, the enhanced photocatalytic activity using the PMMA/ZnO-Au1 hybrid structures indicates that the specific ZnO-Au combination is favorable for the formation of the Schottky barrier at the metal–semiconductor interface, which enhances the charge carrier separation and therefore the ROS formation [21,46]. In fact, as observed in the SEM analysis, when the content of Au was low, the NPs were better distributed on the fibers without forming large aggregates, as in the case of higher amount of Au (PMMA/ZnO-Au6). The presence of Au aggregates on the ZnO NPs catalyst causes a decrease of their active surface exposed towards the organic pollutants and also a less effective light penetration due to a screening effect [22,28,47]. This result is also confirmed by previous studies; as a matter of fact, there is an optimal concentration of metallic NPs for the surface modification of the metal oxide semiconductors, beyond which the photocatalytic performance declines [46,48].

In order to evaluate the conversion of the organic pollutants into CO2 and, therefore, the efficiency of the developed materials to mineralize the organic pollutants upon photocatalytic degradation, the mineralization of MB and BPA has been investigated by TOC studies of the solutions before and after the irradiation experiments. Based on the obtained results (Table S2 in Supplementary Materials), it can be assured that 60% of MB is mineralized in the presence of the PMMA/ZnO-Au1 mat upon UV irradiation, higher compared to the 45.5% obtained using the PMMA/ZnO. In the case of BPA, the PMMA/ZnO-Au1 was able to mineralize the 15% of the 63.5% photodegraded molecules, while no changes in the TOC were detected after using PMMA/ZnO. The lower mineralization of BPA compared to that obtained for MB can be ascribed to the recalcitrant nature of this organic pollutant. The effective removal of the BPA molecules is greatly challenging due to their complex aromatic structure, which, together with their low biodegradability, makes BPA one of the principal pollutants detected in the effluents of the wastewater treatment plants [49].

Since the PMMA/ZnO-Au1 mats show the best performance in terms of photocatalytic degradation of both MB and BPA organic pollutants, their stability is investigated by performing three consecutive irradiation cycles. The loss of Zn and Au from the composite mats was negligible after the first irradiation cycle, since about 0.022 ± 0.001 wt % of Zn with respect to the total Zn amount in the composites was detected in the liquids, while the content of Au was well below the instrument's detection limit. The overall photocatalytic performance of the PMMA/ZnO-Au1 was maintained unvaried for two consecutive cycles in terms of degradation efficiency of MB (Figure 6a), while, in the third cycle, the performance slightly decreased to 72.5%. This was also the case of the PMMA/ZnO sample (Figure S10a of Supplementary Materials) indicating the fact that this reduction in

the performance can be possibly attributed to a low photo-corrosion effect of the ZnO [50]. In the case of BPA, the photocatalytic performance of the PMMA/ZnO-Au1 showed a more evident decrease after the second cycle compared to that observed for the MB, lowering further in the third cycle (Figure 6b). Despite this reduction, the degradation efficiency of the third cycle was higher than that reached for the PMMA/ZnO mat, in which only 10% of the BPA was photodegraded (Figure S10b). This aspect confirms the improvement of the catalytic activity of the mat due to the presence of 1% of Au NPs also in the removal of a recalcitrant aqueous pollutant.

**Figure 6.** Photocatalytic degradation activity under UV light irradiation of (**a**) MB and (**b**) BPA for three consecutive cycles using the PMMA/ZnO-Au1.

#### **4. Conclusions**

The enhancement of the photocatalytic activity of the PMMA/ZnO composite mats was here investigated through a surface modification of the semiconductor nanostructures with Au NPs by means of a straightforward fabrication approach. The hybrid Au/ZnO nanostructures were fixed on a polymeric support through a thermally induced solid-state synthesis and a subsequent adsorption dipping process. Thanks to this accessible and scalable fabrication procedure, the Au/ZnO nanostructures were formed directly on the surface of electrospun polymeric fibers, obtaining a flexible, easily handleable, and light-weight porous system. The photoactive material was stable on the polymeric matrix, since its loss in water after 24 h was negligible. The presence of 1% of Au NPs on the PMMA/ZnO composite mats enhanced the photocatalytic degradation and the mineralization of both the organic pollutants studied, MB and BPA. In fact, the Au NPs were homogeneously distributed on the surface of the ZnO NPs without the formation of aggregates, thus allowing an effective metal–semiconductor interface, which is fundamental for the improvement of the photocatalytic performance of the ZnO under UV light irradiation. Although the BPA degradation was less effective than that obtained for MB, due to the intrinsic nature of the BPA molecules, the presence of Au significantly improved the photodegradation performance of ZnO for such a persistent pollutant. Therefore, the proposed fabrication approach can be considered a valid alternative to the conventional routes for the development of efficient supported photocatalysts for water remediation applications.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4441/11/9/1787/s1, Details on the Debye–Scherrer equation and Kubelka–Munk method, Figure S1: Photos of (**a**) PMMA/ZnO, (**b**) PMMA/ZnO-Au1, (**c**) PMMA/ZnO-Au3, and (**d**) PMMA/ZnO-Au6. Figure S2: SEM images and size distribution analysis of the diameter of the fibers of (**a**,**b**) PMMA and (**c**,**d**) PMMA/Zn(CH3CO2)2 mats. Figure S3: (**a**) HRSEM image of PMMA/ZnO with cross-sectional detail. (**b**) Diameter size distribution of the PMMA/ZnO composite mat. Figure S4: TEM images of (**a**) the smaller ZnO NPs formed in the bulk of the polymeric fibers and of (**b**) the ZnO/Au hybrid structure in the PMMA/ZnO-Au1 composite mat. Figure S5: Dark field TEM image of (**a**) the ZnO-Au hybrid structure and EDS mapping of (**b**) Au, (**c**) Zn, and (**d**) O. Figure S6: Kubelka–Munk plots of the composite mats. The energy band gap is extrapolated from a linear regression. Table S1: Assignment of the Raman modes of the PMMA. Figure S7: Raman spectra of PMMA/ZnO-Au3 and PMMA/ZnO-Au6. Figure S8: Evolution

*Water* **2019**, *11*, 1787

of the normalized concentration of (**a**) MB and (**b**) BPA solutions in presence of the developed mats in dark. Table S2: Degradation and mineralization values obtained from the photocatalytic degradation of the MB and BPA aqueous solution in presence of the mats after 20 h under UV light irradiation. Figure S9: The pseudo-first-order reaction kinetics for (**a**) MB and (**b**) BPA, applied on the experimental data obtained in the first 5 h of reaction. Table S3: Photo-degradation rate constants and linear regression coefficients obtained from the linear fitting of the experimental data by using the pseudo-first order model. Figure S10: Photocatalytic degradation activity of (**a**) MB and (**b**) BPA for three consecutive UV irradiation cycles using PMMA/ZnO composite mats.

**Author Contributions:** L.C. and D.F. conceived and designed the experiments. L.C. realized the materials, performed the characterizations and the photocatalytic experiments, analyzed the data, and accomplished the original draft. S.L. acquired HRSEM images and contributed to the draft editing. A.A. provided the funding and the resources. D.F. supervised the research, validated the results, and edited and reviewed the draft.

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

**Acknowledgments:** The authors kindly acknowledge Alice Scarpellini for the acquisition of the dark-field TEM images and Filippo Drago for the ICP-OES analysis. The useful discussion with Davide Morselli concerning the preparation and the characterization of the materials is also gratefully acknowledged.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


© 2019 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*
