Strong Activity Enhancement of the Photocatalytic Degradation of an Azo Dye on Au / TiO 2 Doped with FeO x

: The doping of Au / TiO 2 with FeO x is shown to result in a strong enhancement of its photocatalytic activity in the degradation of the azo dye Orange II. In order to examine the source of this enhancement, Au-FeO x / TiO 2 nanocomposites containing di ﬀ erent molar ratios of Au:Fe were synthesized, and X-ray di ﬀ raction (XRD), X-ray photoemission spectroscopy (XPS), and high-resolution transmission electron microscope (HRTEM) analyses indicated that the TiO 2 -supported Au nanoparticles were partially covered with an amorphous layer of FeO x species, in which the iron was present as Fe 2 + and Fe 3 + . The metal-semiconductor system, i.e., Au / TiO 2 , showed only a moderate degradation rate, whereas doping with FeO x strongly enhanced the degradation activity. The bandgap energy decreased gradually from Au / TiO 2 (3.13 eV) to the catalyst with the highest FeO x loading Au-FeO x (1:2) / TiO 2 (2.23 eV), and this decrease was accompanied by a steady increase in the degradation activity of the catalysts. XPS analyses revealed that compared to Au / TiO 2 , on Au-FeO x / TiO 2 a much higher population density of chemisorbed and / or dissociated oxygen species was generated, which together with the decreased bandgap resulted in the highest photocatalytic activity observed with Au-FeO x (1:2) / TiO 2 . The processes occurring during reaction on the catalyst surface and in the bulk liquid phase were investigated using operando attenuated total reﬂection IR spectroscopy (ATR-IR) combined with modulation excitation spectroscopy (MES), which showed that the doping of Au / TiO 2 with FeO x weakens the interaction of the dye with the catalyst surface and strongly enhances the cleavage of the azo bond.


Introduction
Increasing public concern about the detrimental environmental impact of organic pollutants in wastewater effluents from industrial and domestic sources has urged the development of suitable technologies for their cleaning. Conventional large-scale water-treatment systems comprising 2θ = 27.4 • , 36.1 • , 54.3 • [27]. The XRD peak at 38.1 • is attributed to the reflection of Au(111) lattice planes of gold nanoparticles. The continuous shift of the Au peak to higher 2θ values with increasing Fe concentration (Figure 1b) indicates that the FeO x species were deposited on or near to Au particles as mono or multilayers [28]. The Au diffraction peak became sharper with increasing FeO x content, in line with the increase of the Au particle size (see Table 1). Note that no diffraction peak of crystalline FeO x was discernible, inferring that the FeO x species were well-dispersed as an amorphous phase. Using the Scherrer equation, the crystallite sizes of anatase and rutile particles were estimated to be ca. 17 and 18 nm, respectively. The amount of deposited FeO x had no significant effect on the grain size of TiO 2 . However, the specific surface area of Au-FeO x /TiO 2 determined by BET analyses increased with the amount of loaded FeO x (see Table 1).
Catalysts 2020, 10, x FOR PEER REVIEW 3 of 15 at 2θ = 27.4°, 36.1°, 54.3° [27]. The XRD peak at 38.1° is attributed to the reflection of Au(111) lattice planes of gold nanoparticles. The continuous shift of the Au peak to higher 2θ values with increasing Fe concentration (Figure 1b) indicates that the FeOx species were deposited on or near to Au particles as mono or multilayers [28]. The Au diffraction peak became sharper with increasing FeOx content, in line with the increase of the Au particle size (see Table 1). Note that no diffraction peak of crystalline FeOx was discernible, inferring that the FeOx species were well-dispersed as an amorphous phase. Using the Scherrer equation, the crystallite sizes of anatase and rutile particles were estimated to be ca. 17 and 18 nm, respectively. The amount of deposited FeOx had no significant effect on the grain size of TiO2. However, the specific surface area of Au-FeOx/TiO2 determined by BET analyses increased with the amount of loaded FeOx (see Table 1).   Figure 2 displays the HRTEM images of the supported nanogold catalysts. The Au nanoparticles were highly dispersed on the surface of the TiO2 microspheres. The mean size of Au nanoparticles was in the range of 4 to 5 nm, but particles of bigger size were also discernible. As mentioned above, a slight increase in the mean Au particle size was observed with the FeOx loading. The Au nanoparticles were in intimate contact with the TiO2 support, which favors a photo-generated electron transfer between the Au and TiO2 [24,25].   Figure 2 displays the HRTEM images of the supported nanogold catalysts. The Au nanoparticles were highly dispersed on the surface of the TiO 2 microspheres. The mean size of Au nanoparticles was in the range of 4 to 5 nm, but particles of bigger size were also discernible. As mentioned above, a slight increase in the mean Au particle size was observed with the FeO x loading. The Au nanoparticles were in intimate contact with the TiO 2 support, which favors a photo-generated electron transfer between the Au and TiO 2 [24,25]. The bandgap energies were estimated from the Kubelka-Munk function via an exploration of the corresponding plots, using the relation αhν = K(hν − Eg) n/2 , in which K is a material independent constant; α is the absorption coefficient (2.303); hν is the incident photon energy; Eg is the optical bandgap energy of the material (1240 eV); and n depends on the type of transition in the semiconductor [29]. The calculated bandgap energies are listed in Table 1 and corresponding absorbance spectra are shown in Figure 3. The bandgap energy of TiO2 decreased upon deposition of Au from 3.23 to 3.12 eV, followed by its further gradual decrease to 2.23 eV caused by the doping with FeOx. The bandgap energies were estimated from the Kubelka-Munk function via an exploration of the corresponding plots, using the relation αhν = K(hν − E g ) n/2 , in which K is a material independent constant; α is the absorption coefficient (2.303); hν is the incident photon energy; E g is the optical bandgap energy of the material (1240 eV); and n depends on the type of transition in the semiconductor [29]. The calculated bandgap energies are listed in Table 1 and corresponding absorbance spectra are shown in Figure 3. The bandgap energy of TiO 2 decreased upon deposition of Au from 3.23 to 3.12 eV, followed by its further gradual decrease to 2.23 eV caused by the doping with FeO x .

Catalytic Performance in Photo-Degradation of Orange II
The photocatalytic performance of the nanogold catalysts was evaluated for the degradation of Orange II under UV-light irradiation (365 nm). Three absorbance bands of Orange II, two bands in the visible region and one in the ultraviolet region, were used to monitor the photocatalytic process. The UV band at 310 nm is ascribed to the naphthalene rings of the dye [30], while the bands in the visible region at 484 nm and 430 nm are assigned to Orange II in hydrazone and azo forms (Scheme 1), respectively. As shown in Figure 4a, Au/TiO2 exhibited an inferior degradation of Orange II in an irradiation time of 180 min, which is mainly caused by the inability to generate enough free radicals during the photocatalysis. However, the degradation rate of Orange II was greatly enhanced when FeOx species were impregnated onto the Au/TiO2, as seen in Figure 4. Increased loading of FeOx species further accelerated the photodegradation process. The reaction rate obeyed fairly well pseudo first-order kinetics r = k Corange II, and the estimated reaction rate constants were: Au/TiO2, 0.003 min −1 ; Au-Fe(1:0.5)/TiO2, 0.008 min −1 ; Au-Fe(1:1)/TiO2, 0.013 min −1 ; and, Au-Fe(1:2)/TiO2, 0.014 min −1 , respectively. Thus, the catalytic efficiency followed the order of Au-Fe(1:2)/TiO2 ≈ Au-Fe(1:1)/TiO2 > Au-Fe(1:0.5)/TiO2 > Au/TiO2. This order is in line with the decrease of the bandgap energies (Table 1), as frequently observed in the photodegradation of azo dyes. As seen in Table 1, a slight increase in the BET surface area was confirmed. However, considering the huge difference in the catalytic activity highlighted in Figure 4, the slight increase of the surface area cannot account for the observed activity enhancement. By adding FeOx, the reaction rate slightly decreased in the initial period of

Catalytic Performance in Photo-Degradation of Orange II
The photocatalytic performance of the nanogold catalysts was evaluated for the degradation of Orange II under UV-light irradiation (365 nm). Three absorbance bands of Orange II, two bands in the visible region and one in the ultraviolet region, were used to monitor the photocatalytic process. The UV band at 310 nm is ascribed to the naphthalene rings of the dye [30], while the bands in the visible region at 484 nm and 430 nm are assigned to Orange II in hydrazone and azo forms (Scheme 1), respectively. As shown in Figure 4a, Au/TiO 2 exhibited an inferior degradation of Orange II in an irradiation time of 180 min, which is mainly caused by the inability to generate enough free radicals during the photocatalysis. However, the degradation rate of Orange II was greatly enhanced when FeO x species were impregnated onto the Au/TiO 2 , as seen in Figure 4. Increased loading of FeO x species further accelerated the photodegradation process. The reaction rate obeyed fairly well pseudo first-order kinetics r = k C orange II , and the estimated reaction rate constants were: Au/TiO 2 , 0.003 min −1 ; Au-Fe(1:0.5)/TiO 2 , 0.008 min −1 ; Au-Fe(1:1)/TiO 2 , 0.013 min −1 ; and, Au-Fe(1:2)/TiO 2 , 0.014 min −1 , respectively. Thus, the catalytic efficiency followed the order of Au-Fe(1:2)/TiO 2 ≈ Au-Fe(1:1)/TiO 2 > Au-Fe(1:0.5)/TiO 2 > Au/TiO 2 . This order is in line with the decrease of the bandgap energies (Table 1), as frequently observed in the photodegradation of azo dyes. As seen in Table 1, a slight increase in the BET surface area was confirmed. However, considering the huge difference in the catalytic activity highlighted in Figure 4, the slight increase of the surface Catalysts 2020, 10, 933 6 of 15 area cannot account for the observed activity enhancement. By adding FeO x , the reaction rate slightly decreased in the initial period of reaction (Figure 4e,f). This could originate from changes in the surface configuration and oxidation state of FeO x reaching a steady state after a certain induction period. Importantly, no colored intermediate species were generated in the solution during the photocatalytic process, as shown by the UV-Vis spectra presented in Figure 4. The advantage of the addition of Fe oxide to TiO 2 in photocatalysis was demonstrated by Mercyrani et al. [20], who reported that the best degradation performance was achieved with a catalyst containing 80% Fe 3 O 4 and 20% TiO 2 . The addition of Fe 3 O 4 contributes to the activation of O 2 molecules via e − transfer from the conduction band (CB) of TiO 2 to the one of Fe 3 O 4 as follows [20]: reaction (Figure 4e,f). This could originate from changes in the surface configuration and oxidation state of FeOx reaching a steady state after a certain induction period. Importantly, no colored intermediate species were generated in the solution during the photocatalytic process, as shown by the UV-Vis spectra presented in Figure 4. The advantage of the addition of Fe oxide to TiO2 in photocatalysis was demonstrated by Mercyrani et al. [20], who reported that the best degradation performance was achieved with a catalyst containing 80% Fe3O4 and 20% TiO2. The addition of Fe3O4 contributes to the activation of O2 molecules via e − transfer from the conduction band (CB) of TiO2 to the one of Fe3O4 as follows [20]: Fe3O4 (e − CB) + O2 → Fe3O4 + O2 − In our study with UV light, a small amount of FeOx (equal molar ratio with Au) was sufficient for the above e − transfer through the Au nanoparticles due to the combination of Au, FeOx, and TiO2.  reaction (Figure 4e,f). This could originate from changes in the surface configuration and oxidation state of FeOx reaching a steady state after a certain induction period. Importantly, no colored intermediate species were generated in the solution during the photocatalytic process, as shown by the UV-Vis spectra presented in Figure 4. The advantage of the addition of Fe oxide to TiO2 in photocatalysis was demonstrated by Mercyrani et al. [20], who reported that the best degradation performance was achieved with a catalyst containing 80% Fe3O4 and 20% TiO2. The addition of Fe3O4 contributes to the activation of O2 molecules via e − transfer from the conduction band (CB) of TiO2 to the one of Fe3O4 as follows [20]: In our study with UV light, a small amount of FeOx (equal molar ratio with Au) was sufficient for the above e − transfer through the Au nanoparticles due to the combination of Au, FeOx, and TiO2.
Catalysts 2020, 10, 933 7 of 15 In our study with UV light, a small amount of FeO x (equal molar ratio with Au) was sufficient for the above e − transfer through the Au nanoparticles due to the combination of Au, FeO x , and TiO 2 .
The oxidation state of Au nanoparticles is well known to be crucial in the photocatalytic degradation reactions [25]. Hence, we performed XPS measurements to elucidate the chemical properties of the Au, Fe, and O species. Figure 5 depicts the measured spectra of Au4f, Fe2p, and O1s of the Au/TiO 2 and Au-FeO x /TiO 2 catalysts. Only Au and Fe and no extra peak corresponding to any impurities were discernible. The fitted Au4f spectra show that gold nanoparticles were present in a single oxidation state, i.e., metallic Au 0 . The 4f 7/2 photoelectron peak was located at the binding energy (BE) in the range of 83.1 to 83.4 eV. The BE was slightly lower than that of the reference Au foil (83.4 eV) caused by an electron transfer from TiO 2 to gold nanoparticles. The Au 4f 7/2 signal showed a negative shift in binding energy for Au/TiO 2 and a continuous shift to higher BE with increasing FeO x content. The reported negative shifts of the Au 4f 7/2 BE are scattered in a wide range from 0.2 to −1.2 eV. The particle-support interaction is suggested to be the possible reason for the negative energy shift. According to literature, the shift of the Au 4f 7/2 peak toward lower BE can be better explained by an initial state effect associated with electron transfer from Ti 3+ surface defect sites to Au clusters [31]. Interestingly, the Au4f 7/2 peak showed a continuous shift to higher BE with increasing FeO x content due to the electron transfer from Au particles to the FeO x species.
Catalysts 2020, 10, x FOR PEER REVIEW 7 of 15 The oxidation state of Au nanoparticles is well known to be crucial in the photocatalytic degradation reactions [25]. Hence, we performed XPS measurements to elucidate the chemical properties of the Au, Fe, and O species. Figure 5 depicts the measured spectra of Au4f, Fe2p, and O1s of the Au/TiO2 and Au-FeOx/TiO2 catalysts. Only Au and Fe and no extra peak corresponding to any impurities were discernible. The fitted Au4f spectra show that gold nanoparticles were present in a single oxidation state, i.e., metallic Au 0 . The 4f7/2 photoelectron peak was located at the binding energy (BE) in the range of 83.1 to 83.4 eV. The BE was slightly lower than that of the reference Au foil (83.4 eV) caused by an electron transfer from TiO2 to gold nanoparticles. The Au 4f7/2 signal showed a negative shift in binding energy for Au/TiO2 and a continuous shift to higher BE with increasing FeOx content. The reported negative shifts of the Au 4f7/2 BE are scattered in a wide range from 0.2 to −1.2 eV. The particle-support interaction is suggested to be the possible reason for the negative energy shift. According to literature, the shift of the Au 4f7/2 peak toward lower BE can be better explained by an initial state effect associated with electron transfer from Ti 3+ surface defect sites to Au clusters [31]. Interestingly, the Au4f7/2 peak showed a continuous shift to higher BE with increasing FeOx content due to the electron transfer from Au particles to the FeOx species. The BE peaks at ~709.8 eV (Fe2p3/2) and 722.7 eV (Fe2p1/2) are characteristic of Fe 2+ species in the form of FeO [32]. However, Fe 3+ species (Fe2O3) also existed in all catalysts, as evidenced by the Fe2p3/2 at ~713.5 eV and Fe2p1/2 at ~723.6 eV. The concentration of Fe 2+ species increased with the FeOx content (see Table 2), indicating an electron transfer from gold nanoparticles to FeOx, which is in agreement with the results of Au XPS analyses. The asymmetric O1s spectra were deconvoluted into three peaks, i.e., the BE values at ca. 529.5 eV, 530 eV, and 531.4 eV, as shown in Figure 5c. These peaks are associated with lattice oxygen species (OL), oxygen vacancies or defects (OV), and chemisorbed or dissociated oxygen species (OC), respectively [33][34][35]. Figure 6 shows the relative concentrations of these species on the catalyst surface together with the corresponding degrees of degradation of the catalysts achieved after 30 min under UV light irradiation. Obviously, the catalytic activity of Au-FeOx/TiO2 composites depends on their OC concentration, whereas OL is unfavorable for the photodegradation of Orange II molecules, as emerges from Figure 6, which shows the dependence of the concentration of OL, OV, and OC species. The chemisorbed OC is normally existing in the form of hydroxyl groups and adsorbed H2O/O2 on the surface of oxides. The surface hydroxyl groups and adsorbed O2 can facilitate the trapping of photoinduced electrons and holes, thus enhancing the photocatalytic degradation process [6]. These hydroxyl free radicals (·OH) and possibly also superoxide radicals (O2 −• ) are generated by the conduction band electrons via the reduction of the dissolved O2, reacting with Orange II molecules. The doping of the Au/TiO2 catalyst with FeOx enhances the production of the chemisorbed oxygen species on the surface. This is supported by the fact that the Au-Fe(1:2)/TiO2 photocatalysts with abundant OC species exhibited the highest activity in the photocatalytic degradation of Orange II. The BE peaks at~709.8 eV (Fe2p 3/2 ) and 722.7 eV (Fe2p 1/2 ) are characteristic of Fe 2+ species in the form of FeO [32]. However, Fe 3+ species (Fe 2 O 3 ) also existed in all catalysts, as evidenced by the Fe2p 3/2 at~713.5 eV and Fe2p 1/2 at~723.6 eV. The concentration of Fe 2+ species increased with the FeO x content (see Table 2), indicating an electron transfer from gold nanoparticles to FeO x , which is in agreement with the results of Au XPS analyses. The asymmetric O1s spectra were deconvoluted into three peaks, i.e., the BE values at ca. 529.5 eV, 530 eV, and 531.4 eV, as shown in Figure 5c. These peaks are associated with lattice oxygen species (O L ), oxygen vacancies or defects (O V ), and chemisorbed or dissociated oxygen species (O C ), respectively [33][34][35]. Figure 6 shows the relative concentrations of these species on the catalyst surface together with the corresponding degrees of degradation of the catalysts achieved after 30 min under UV light irradiation. Obviously, the catalytic activity of Au-FeO x /TiO 2 composites depends on their O C concentration, whereas O L is unfavorable for the photodegradation of Orange II molecules, as emerges from Figure 6 Figure 4).

Concentration Modulation
The Orange II concentration modulation experiments were performed by periodically altering the feeding of two different solutions (Orange II + H2O + O2 + UV ↔ H2O + O2 + UV) at 25 °C. The surface spectra and the liquid phase spectra in the phase-domain are depicted in Figure 7. In the surface spectra, the IR band at 1510 cm −1 is assigned to the characteristic azo bond (N=N stretching vibration) of Orange II [16,18]. The band at 1454 cm −1 is due to the aromatic skeletal vibrations or the bending vibration mode (N-H) of the hydrazone form [14]. These bands became very weak with Au-Fe(1:2)/TiO2, indicating a rapid decomposition of the azo dye and aromatic ring cleavage of Orange II. This was also confirmed by the weak band at 1554 cm −1 , originating from the aromatic skeletal vibrations of Orange II [31]. Interestingly, Au-Fe(1:2)/TiO2 showed a negative band (at 1260 cm −1 ) [14] for C-O-H stretching vibration. This is because the Orange II is subjected to intramolecular hydrogen bonding tautomeric interactions between the oxygen of the naphthyl group and the β-hydrogen of the corresponding azo linkage, converting Orange II into the hydrazone form. This transformation might be key for the high photodegradation activity of the Au-Fe(1:2)/TiO2 catalyst. The weak intensity of IR bands at 1554 cm −1 (C=C) [14] and 1213 cm −1 (=C-H bending) [12] also reflects the destruction of the aromatic group of the dye on Au-Fe(1:2)/TiO2. The sharp band at 1120 cm⁻ 1 is ascribed to the symmetric vibration of the sulfonate groups (SO3) [19]. The bands at 1033 and 1006  Figure 4).

Concentration Modulation
The Orange II concentration modulation experiments were performed by periodically altering the feeding of two different solutions (Orange II + H 2 O + O 2 + UV ↔ H 2 O + O 2 + UV) at 25 • C. The surface spectra and the liquid phase spectra in the phase-domain are depicted in Figure 7. In the surface spectra, the IR band at 1510 cm −1 is assigned to the characteristic azo bond (N=N stretching vibration) of Orange II [16,18]. The band at 1454 cm −1 is due to the aromatic skeletal vibrations or the bending vibration mode (N-H) of the hydrazone form [14]. These bands became very weak with Au-Fe(1:2)/TiO 2 , indicating a rapid decomposition of the azo dye and aromatic ring cleavage of Orange II. This was also confirmed by the weak band at 1554 cm −1 , originating from the aromatic skeletal vibrations of Orange II [31]. Interestingly, Au-Fe(1:2)/TiO 2 showed a negative band (at 1260 cm −1 ) [14] for C-O-H stretching vibration. This is because the Orange II is subjected to intramolecular hydrogen bonding tautomeric interactions between the oxygen of the naphthyl group and the β-hydrogen of the corresponding azo linkage, converting Orange II into the hydrazone form. This transformation might be key for the high photodegradation activity of the Au-Fe(1:2)/TiO 2 catalyst. The weak intensity of IR bands at 1554 cm −1 (C=C) [14] and 1213 cm −1 (=C-H bending) [12] also reflects the destruction of the aromatic group of the dye on Au-Fe(1:2)/TiO 2 . The sharp band at 1120 cm −1 is ascribed to the symmetric vibration of the sulfonate groups (SO 3 ) [19]. The bands at 1033 and 1006 cm −1 are assigned to the coupling between the benzene mode and SO 3 [16]. Significantly, destruction of the aromatic components occurs on the Au-FeO x /TiO 2 surface during the photocatalytic process, as evidenced by the persistence of the very weak band intensities related to the hydrazone part of the azo dye, such as the aromatic C=C, N-H, and O-H stretching vibrations at 1554 and 1454 and 1260 cm −1 , respectively (Figure 7a-c). These IR observations support the fact that Au-Fe(1:2)/TiO 2 outperforms the other catalysts in degrading Orange II. The phase-domain spectra (Figure 7a-d) were demodulated at the phase angle ϕ PSD of each 10 • step in the range of 0-350 • . Obviously, the presence of FeO x species on the Au/TiO 2 surface significantly increased the amount of adsorbed Orange II; the absorbance of the band at 1120 cm −1 was highest (4.5 × 10 −3 ) for Au-Fe(1:2)/TiO 2 . It is worthy to note that the promotion of Au/TiO 2 with FeO x species leads to slower adsorption of Orange II, whereas its desorption was rather quick as indicated in Figure 7e-h (colored time-domain spectra). This change in the adsorption-desorption dynamics indicates that the population of adsorption sites was increased, but the interaction of Orange II with the catalyst surface became much weaker upon the promotion of Au/TiO 2 with FeO x . The stronger adsorption of Orange II on Au/TiO 2 seems to suppress its photodegradation activity.  (Figure 7a-c). These IR observations support the fact that Au-Fe(1:2)/TiO2 outperforms the other catalysts in degrading Orange II. The phase-domain spectra (Figure 7a-d) were demodulated at the phase angle φ PSD of each 10° step in the range of 0-350°. Obviously, the presence of FeOx species on the Au/TiO2 surface significantly increased the amount of adsorbed Orange II; the absorbance of the band at 1120 cm −1 was highest (4.5 × 10 −3 ) for Au-Fe(1:2)/TiO2. It is worthy to note that the promotion of Au/TiO2 with FeOx species leads to slower adsorption of Orange II, whereas its desorption was rather quick as indicated in Figure 7e-h (colored time-domain spectra). This change in the adsorption-desorption dynamics indicates that the population of adsorption sites was increased, but the interaction of Orange II with the catalyst surface became much weaker upon the promotion of Au/TiO2 with FeOx. The stronger adsorption of Orange II on Au/TiO2 seems to suppress its photodegradation activity. The liquid phase spectra of Orange II concentration modulation (Figure 7i-l) also show that the intensity of the characteristic azo bond decreases faster with increasing concentration of FeOx species on Au-Fe/TiO2. This is in good agreement with the UV-Vis analyses in Figure 4. The degradation of Orange II has been shown to generate SO4 2− and BH4 + [16]. Our surface spectra showed no sign of these species adsorbed on the surface. Therefore, they must be in the liquid phase but probably under the detection limit of FT-IR. Our operando MES-ATR-IR set-up successfully monitored both the liquid and surface spectra on the same time scale. The liquid phase spectra of Orange II concentration modulation (Figure 7i-l) also show that the intensity of the characteristic azo bond decreases faster with increasing concentration of FeO x species on Au-Fe/TiO 2 . This is in good agreement with the UV-Vis analyses in Figure 4. The degradation of Orange II has been shown to generate SO 4 2− and BH 4 + [16]. Our surface spectra showed no sign of these species adsorbed on the surface. Therefore, they must be in the liquid phase but probably under the detection limit of FT-IR. Our operando MES-ATR-IR set-up successfully monitored both the liquid and surface spectra on the same time scale.

Gas Modulation
Finally, the surface and liquid phase spectra of MES-ATR-IR during gas composition modulation (replacement of O 2 by He), Orange II + O 2 + UV ↔ Orange II + He + UV, is shown in Figure 8. IR bands detected in the gas modulation were much weaker than those observed in the Orange II concentration and UV light modulation. The bands from 1560 to 1510 cm −1 are assigned to the combination of vibrations involving N=N aromatic skeletal vibrations [16], N-H bending and -N=C stretching of the NH-N=C group, and the phenyl ring vibrations [36]. These characteristic bands of the azo dye appeared as negative bands over Au-Fe(1:2)/TiO 2 and Au-Fe(1:1)/TiO 2 , and they show up as positive weak bands with Au-Fe(1:0.5)/TiO 2 , and comparatively strong bands with Au/TiO 2 . It suggests that the dye molecules are cleaved at the azo bonds and that the Orange II was totally degraded over the Au-Fe(1:2)/TiO 2 and Au-Fe(1:1)/TiO 2 catalysts in the presence of O 2 . This finding is consistent with the observations in the liquid phase spectra (Figure 8e-h).

Gas Modulation
Finally, the surface and liquid phase spectra of MES-ATR-IR during gas composition modulation (replacement of O2 by He), Orange II + O2 + UV ↔ Orange II + He + UV, is shown in Figure 8. IR bands detected in the gas modulation were much weaker than those observed in the Orange II concentration and UV light modulation. The bands from 1560 to 1510 cm −1 are assigned to the combination of vibrations involving N=N aromatic skeletal vibrations [16], N-H bending and -N=C stretching of the NH-N=C group, and the phenyl ring vibrations [36]. These characteristic bands of the azo dye appeared as negative bands over Au-Fe(1:2)/TiO2 and Au-Fe(1:1)/TiO2, and they show up as positive weak bands with Au-Fe(1:0.5)/TiO2, and comparatively strong bands with Au/TiO2. It suggests that the dye molecules are cleaved at the azo bonds and that the Orange II was totally degraded over the Au-Fe(1:2)/TiO2 and Au-Fe(1:1)/TiO2 catalysts in the presence of O2. This finding is consistent with the observations in the liquid phase spectra (Figure 8e-h).

Preparation of Au/TiO 2 and Au-FeO x /TiO 2 Catalysts
The Au/TiO 2 catalysts were prepared using a deposition-precipitation method [37]. Sixty mg of HAuCl 4 ·H 2 O was dissolved in 100 mL of deionized water. The solution was heated to 70 • C, and the pH was adjusted to 7.0 by adding 1.0 M aqueous ammonia solution. Afterward, the TiO 2 powder was added to the solution under constant stirring. The gold loading was controlled at a nominal value of 3 wt% as Au metal (97 wt% TiO 2 ). The resulting slurry was aged for 1 h, while the pH was maintained at around 7.0. Then, the solid was filtered off and washed with deionized water. The obtained powder was dried at 60 • C overnight and calcined at 300 • C for 1 h in air to obtain 3 wt% Au/TiO 2 . For the preparation of Au-FeO x /TiO 2 , the appropriate amount of iron(III) nitrate was dissolved in 100 mL of deionized water, and the Au/TiO 2 powder was added, resulting in a suspension, which was stirred for 1 h at room temperature. Subsequently, the suspension was heated to 60 • C in a rotary evaporator and aged for 1 h, followed by solvent evaporation at 35 • C under vacuum. The powders were then dried at 60 • C overnight and calcined at 300 • C for 1 h. The Au-FeO x /TiO 2 catalysts with different Au:Fe molar ratios are designated as Au-Fe(1:0.5)/TiO 2 , Au-Fe(1:1)/TiO 2 , and Au-Fe(1:2)/TiO 2 .

Characterization
UV-Vis absorption spectra of Orange II in aqueous solutions were analyzed from 200 to 600 nm on a Hewlett-Packard Agilent 8453 diode array spectrophotometer (Agilent Technologies, Inc., Santa Clara, CA, USA). The UV-Vis diffuse reflection spectra of the samples were measured using a Shimadzu UV2600 UV-Vis Spectrophotometer (Shimadzu Corporation, Kyoto, Japan). The bandgap energies of the photocatalysts were determined employing the Kubelka-Munk theory of reflectance.
The crystalline structure of the catalysts was analyzed using XRD recorded on a PANalytical X'Pert powder X-ray diffractometer (PANalytical, Almelo, Netherlands) operated at a voltage of 60 kV and a current of 55 mA using a Cu Kα radiation source in the 2θ range (λ = 1.5406 Å) from 5 to 80 • with a step of 0.02 • . The average crystallite sizes of anatase and rutile were calculated employing the Scherrer equation: D hkl = 0.9 λ/β cos θ, where λ is the wavelength, β is the FWHM (full width at half maximum) measured in radians on the 2θ scale, and θ is the Bragg angle. The average crystal size of rutile and anatase was estimated by averaging the values of all the diffraction patterns.
HRTEM images were obtained on a FEI Tecnai G 2 F30S-Twin microscope (FEI Company, Hillsboro, OR, USA) operated at an acceleration voltage of 300 kV. The specimen was prepared by ultrasonically dispersing the sample into ethanol, and droplets of the suspensions were deposited on a carbon-enhanced copper grid and then dried in air. Two-hundred Au particles were analyzed to estimate their mean size on each sample.
The specific surface areas of all catalysts were determined using nitrogen adsorption measurements at 77 K on an ASAP 2020 apparatus (Micromeritics, Norcross, GA, USA) using the Brunauer-Emmett-Teller (BET) method. Prior to adsorption measurements, all samples were degassed at 250 • C for 4 h to remove adsorbed CO 2 and H 2 O.

Catalytic Photodegradation
Catalytic photodegradation was performed in a 500 mL three-neck round bottom flask. Two-hundred mL of 0.02 mmol/L Orange II solution was photolyzed with a catalyst dose of 30 mg.
UV light with λ = 365 nm was applied for photodegradation of Orange II at 25 • C controlled by a thermostat (Julabo GmbH, Seelbach, Germany). Oxygen gas was saturated in the catalyst-containing suspension as an oxidant at a rate of 20 mL/min. Samples of the suspension were taken at intervals of 30 min and immediately treated with centrifugation to remove the catalysts. The resulting solutions were analyzed using UV-Vis to determine the catalytic performance.

Operando ATR-IR Combined with Modulation-Excitation Spectroscopy
ATR-IR studies [38][39][40] were performed on a Vertex 70 v spectrometer (Bruker, Billerica, MA, USA) equipped with a liquid-nitrogen-cooled mercury-cadmium-telluride (MCT) detector (ID316, Ge Window) and an optical filter (F321). Catalyst layers for in situ measurements were prepared using a deposition-evaporation method. Also, 180 mg of catalyst was added to 24 mL of ethanol, then the slurry was ultrasonicated for 15 min to obtain a uniform suspension. Afterward, 1.6 mL of the slurry was brought onto a Ge internal reflection element (IRE, bevel of 45 • , 52 mm × 20 mm × 2 mm), and subsequently the ethanol was evaporated at room temperature overnight. The prepared catalyst layer adhered to the IRE so that no loss of catalyst powders was observed for several hours under flow-through conditions.
Operando IR experiments were performed in the flow-through cell, which was mounted onto an ATR-IR attachment (HARRICK Scientific Products, Inc.), Pleasantville, NY, USA). The temperature of the cell was controlled by a thermostat at 25 • C. Spectra were recorded at 4 cm −1 of spectral resolution and 60 kHz of scanning velocity. Solutions were filled in glass tanks kept at atmospheric pressure. The tanks were connected to the reactor cell by Teflon tubing and a three-way valve. The solutions were introduced by a peristaltic pump installed downstream of the reactor cell. The liquid outlet was connected to another ATR-IR cell without a catalyst layer so that concentration changes in the liquid phase could also be monitored to realize operando spectroscopy.
Modulation-excitation ATR-IR spectroscopy (MES-ATR-IR) experiments [41][42][43][44] were carried out by periodically changing the reaction conditions. Three different types of modulation experiments were performed: Eight cycles were repeated, and the last five cycles were averaged into one cycle to enhance the S/N ratio and time resolution. Phase-sensitive detection (PSD) was used to further remove the noise and to obtain kinetic information of responding surface species. The acquired time-domain spectra were mathematically treated with PSD to obtain phase-domain spectra according to the following equation: A(t, v) sin(kωt + ϕ k )dt (4) where T is the length of a cycle, ω is the demodulation frequency (where ω = 1 in this study), ϕ k is the demodulation phase angle, k is the demodulation index (where k = 1 in this study), and A(t, v) and A k ( v) are the active species responses in the time and phase domains, respectively.

Conclusions
The doping of a TiO 2 -supported Au catalyst, containing 3 wt% metallic Au nanoparticles of 4-6 nm mean size, with an amorphous layer of FeO x species, in which the iron was present as Fe 2+ and Fe 3+ , resulted in a nearly five-fold higher photocatalytic degradation rate of the azo dye Orange II compared to the parent Au/TiO 2 . The reason for this rate enhancement could be traced back to a strong decrease of the bandgap energy from 3.12 eV to 2.34 eV for the Au-FeO x /TiO 2 catalyst with a molar ratio of Au:Fe = 1:2, and a dramatic increase of the surface population of chemisorbed and/or dissociated oxygen species. The metallic gold nanoparticles acted as electron scavengers, as confirmed by XPS analyses, which indicated an electron transfer from TiO 2 to Au, resulting in a lowering of the bandgap energy from 3.23 eV to 3.12 eV. Operando MES-ATR-IR spectroscopy investigations of the catalytic solid/liquid-gas interface and the bulk liquid-gas phase during the photocatalytic degradation revealed that the FeO x doping enhanced the cleavage of the azo bond and weakened the interaction of Orange II with the catalyst surface, resulting in a change of its adsorption-desorption dynamics. To the best of our knowledge, this is the first application of operando MES-ATR-IR spectroscopy to study the surface processes occurring at the solid-liquid interface of a photocatalytic reaction.

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