Regarding the Nature of Charge Carriers Formed by UV or Visible Light Excitation of Carbon-Modiﬁed Titanium Dioxide

: Although titanium dioxide gathers many of the required properties for its application in photocatalytic processes, its lack of activity in the visible range is a major hurdle yet to be overcome. Among di ﬀ erent strategies, the post-synthesis modiﬁcation of TiO 2 powders with organic compounds has already led to commercially available materials, such as KRONOClean 7000. In this work, we apply di ﬀ use reﬂectance transient absorption spectroscopy on this visible-light active photocatalyst and study the dynamics of the charge carriers alternatively induced by UV or visible light laser irradiation, under inert or reactive atmospheres. Our results can be interpreted by considering the material as TiO 2 sensitized by an organic-based layer, in agreement with previous studies on it, and show that the oxidative power of the material is considerably diminished under visible light irradiation. By complementarily performing continuous visible light irradiation photocatalysis experiments in aerated aqueous suspensions, we show that, although the oxidation of methanol proceeds at a very slow rate, the oxidation of chlorpromazine occurs much faster thanks to its better suited redox potential.


Introduction
In 1972, Fujishima and Honda reported the photocatalytic splitting of water over titanium dioxide (TiO 2 ) photoelectrodes irradiated with UV light [1]. Since this discovery, much work has been done to understand the underlying processes in photocatalytic reactions [2] and develop new materials with activity in the visible range [3]. Despite these efforts, TiO 2 continues to be the most employed photocatalyst, in particular in its anatase polymorph, due to its higher activity in comparison with the thermodynamically stable form rutile [4,5]. However, anatase has a bandgap of 3.23 eV [6] and, thus, it can make use of not more than 3.5% of the energy emitted by the sun [7]. Different strategies have been considered to increase this fraction. One possibility is to replace the material altogether. For example, iron based photocatalysts generally show a strong absorption in the visible range; however, they are far less active under visible light irradiation than TiO 2 under UV irradiation [8,9]. Besides its replacement can be used to enhance the light absorption by TiO2, such as doping or the adsorption of sensitizers on the surface. While doping of TiO2 (either with metals or non-metals) typically leads to bandgap narrowing [10,11], dye-sensitization extends the visible-light response by coupling the semiconductor with a strongly absorbing molecule, finding applications not only in the field of solar cells [12], but also for photocatalysis [13,14]. These attempts can be considered successful after the commercial development of the KRONOClean 7000 photocatalyst from Kronos International, synthesized by the modification of TiO2 with pentaerythritol [15]. Although initially it was thought to be carbon-doped TiO2, it was later proven to consist of anatase particles with a carbon-based sensitizer layer around them [15]. By performing a set of simple but elegant experiments, Ząbek et al. were able to remove the layer around the particles and to reassemble it afterwards, while getting exactly the same properties as before the removal. Importantly, the material was shown to be superior to Evonik P25, both for the photocatalytic abatement of NOx under white light excitation and for the degradation of acetone under visible light [16,17]. Nevertheless, to the best of our knowledge, the tradeoffs incurred by extending the photocatalytic activity of TiO2 to the visible range, and the similarities or differences of this process compared to those initiated with UV light, remain so far unclarified.
With the goal of understanding the nature and reactivity of its photogenerated charge carriers, here we investigated KRONOClean 7000 by means of transient absorption spectroscopy using both UV (355 nm) and visible light (455 nm) laser excitation. Furthermore, we analyzed the influences of inert (argon), oxygen, and methanol atmospheres on the charge carrier dynamics. Complementarily, we studied the photocatalytic oxidation of two organic compounds, chlorpromazine and methanol, using aerated KRONOClean 7000 suspensions under continuous visible light irradiation, from which we were able to establish boundaries on its photocatalytic activity under such conditions.

Characterization of the Photocatalyst
We start the material characterization by analyzing the basic structural properties of KRONOClean 7000 (from here on, K-7000). As can be observed in Figure 1, the XRD pattern matches well with the standard data of anatase TiO2 (JCPDS card No. , indicating that the carbon modification does not significantly modify the crystalline structure of the parent material. The pattern precludes the presence of a significant fraction of the rutile or brookite TiO2 polymorphs. To determine the light absorbing properties of the photocatalyst, we measured its diffuse reflectance spectra in the spectral range from 300 to 800 nm (Figure 2(a)). As a comparison, the results for the pure-anatase photocatalyst Hombikat UV100 are also shown. From these spectra, we determined the Kubelka-Munk functions, shown in Figure 2(b). A long-tailed absorption is clearly observed for K-7000, spanning the visible range from 400 nm to 550 nm. On the contrary, UV100 only To determine the light absorbing properties of the photocatalyst, we measured its diffuse reflectance spectra in the spectral range from 300 to 800 nm ( Figure 2a). As a comparison, the results for the pure-anatase photocatalyst Hombikat UV100 are also shown. From these spectra, we determined the Kubelka-Munk functions, shown in Figure 2b. A long-tailed absorption is clearly observed for K-7000, spanning the visible range from 400 nm to 550 nm. On the contrary, UV100 only absorbs in the ultraviolet range. By processing the spectra of K-7000 and Hombikat UV100 with the Tauc plot ( Figure  S1) we obtained bandgap values of 3.31 eV (corresponding to a wavelength of 374 nm) for K-7000 and 3.32 eV (~373 nm) for UV100.
Catalysts 2019, 9, x FOR PEER REVIEW 3 of 16 absorbs in the ultraviolet range. By processing the spectra of K-7000 and Hombikat UV100 with the Tauc plot ( Figure S1) we obtained bandgap values of 3.31 eV (corresponding to a wavelength of 374 nm) for K-7000 and 3.32 eV (~373 nm) for UV100.

Mott-Schottky Measurements
In order to determine the flatband potential of K-7000 and Hombikat UV100, we performed electrochemical measurements. By plotting the inverse of the square of the measured capacitance versus the applied potential, the flatband potential can be obtained by a linear fit. As shown in Figure  3, this value was determined to be equal to −0.46 V vs. NHE and −0.51 V vs. NHE, for K-7000 and Hombikat UV100, respectively. The small difference is within experimental error.

Transient Absorption Spectroscopy
To study the dynamics of the photogenerated charge carriers, we investigated K-7000 via diffuse reflectance transient absorption spectroscopy. Figure 4 presents the transient absorption spectra of the K-7000 powder obtained after excitation with 355 nm (Figure 4(a)) and 455 nm (Figure 4(b)) at 100 ns after the laser excitation. The spectra were obtained at three different conditions, as follows: in the presence of oxygen (electron scavenger), of methanol (hole scavenger), and in the absence of

Mott-Schottky Measurements
In order to determine the flatband potential of K-7000 and Hombikat UV100, we performed electrochemical measurements. By plotting the inverse of the square of the measured capacitance versus the applied potential, the flatband potential can be obtained by a linear fit. As shown in Figure 3, this value was determined to be equal to −0.46 V vs. NHE and −0.51 V vs. NHE, for K-7000 and Hombikat UV100, respectively. The small difference is within experimental error.
Catalysts 2019, 9, x FOR PEER REVIEW 3 of 16 absorbs in the ultraviolet range. By processing the spectra of K-7000 and Hombikat UV100 with the Tauc plot ( Figure S1) we obtained bandgap values of 3.31 eV (corresponding to a wavelength of 374 nm) for K-7000 and 3.32 eV (~373 nm) for UV100.

Mott-Schottky Measurements
In order to determine the flatband potential of K-7000 and Hombikat UV100, we performed electrochemical measurements. By plotting the inverse of the square of the measured capacitance versus the applied potential, the flatband potential can be obtained by a linear fit. As shown in Figure  3, this value was determined to be equal to −0.46 V vs. NHE and −0.51 V vs. NHE, for K-7000 and Hombikat UV100, respectively. The small difference is within experimental error.

Transient Absorption Spectroscopy
To study the dynamics of the photogenerated charge carriers, we investigated K-7000 via diffuse reflectance transient absorption spectroscopy. Figure 4 presents the transient absorption spectra of the K-7000 powder obtained after excitation with 355 nm (Figure 4(a)) and 455 nm (Figure 4(b)) at 100 ns after the laser excitation. The spectra were obtained at three different conditions, as follows: in the presence of oxygen (electron scavenger), of methanol (hole scavenger), and in the absence of

Transient Absorption Spectroscopy
To study the dynamics of the photogenerated charge carriers, we investigated K-7000 via diffuse reflectance transient absorption spectroscopy. Figure  of oxygen (electron scavenger), of methanol (hole scavenger), and in the absence of electron scavengers or donors (Ar atmosphere). In the latter case, the photogenerated charge carriers can only react with each other (i.e., recombine). electron scavengers or donors (Ar atmosphere). In the latter case, the photogenerated charge carriers can only react with each other (i.e., recombine).
As can be seen in Figure 4(a), after excitation with 355 nm, there are no significant differences in the transient absorption spectra under Ar or O2 atmospheres, while the experiments in Ar atmosphere and methanol atmosphere show a considerable difference. Moreover, this difference is larger at shorter wavelengths (~400 nm) than at the longer wavelength range (~650 nm). On the other hand, as revealed in Figure 4(b), the transient absorption spectra obtained after excitation with 455 nm in all the atmospheres are, within experimental error, identical to each other.   (Figure 5(a)), both under Ar and O2 atmospheres the transient absorption strongly increases right after the laser pulse and is followed by a rapid decrease down to ~15% of the initial value at 8 μs. In the case of the methanol atmosphere, however, the transient absorption is initially lower than the other two cases and increases with time in the investigated window. By comparing the transient absorption decays after excitation with 355 nm and 455 nm, Figure 5(a) and Figure 5(b), it can be seen that the initial absorptions after excitation with 455 nm are much lower than those values after excitation with 355 nm. After excitation with 455 nm the decrease of the transient absorption is significantly slower and decreases only to ~60% of the initial value after 8 μs. Furthermore, at this excitation wavelength no significant difference was observed regarding the type of atmosphere. As can be seen in Figure 4a, after excitation with 355 nm, there are no significant differences in the transient absorption spectra under Ar or O 2 atmospheres, while the experiments in Ar atmosphere and methanol atmosphere show a considerable difference. Moreover, this difference is larger at shorter wavelengths (~400 nm) than at the longer wavelength range (~650 nm). On the other hand, as revealed in Figure 4b, the transient absorption spectra obtained after excitation with 455 nm in all the atmospheres are, within experimental error, identical to each other. Figure 5 displays the transient absorption decays detected at 650 nm for the K-7000 powder in O 2 , methanol and Ar atmospheres after excitation with 355 nm (left) and 455 nm (right). In the case of excitation with 355 nm (Figure 5a), both under Ar and O 2 atmospheres the transient absorption strongly increases right after the laser pulse and is followed by a rapid decrease down to~15% of the initial value at 8 µs. In the case of the methanol atmosphere, however, the transient absorption is initially lower than the other two cases and increases with time in the investigated window. By comparing the transient absorption decays after excitation with 355 nm and 455 nm, Figure 5a,b, it can be seen that the initial absorptions after excitation with 455 nm are much lower than those values after excitation with 355 nm. After excitation with 455 nm the decrease of the transient absorption is significantly slower and decreases only to~60% of the initial value after 8 µs. Furthermore, at this excitation wavelength no significant difference was observed regarding the type of atmosphere.
Catalysts 2019, 9, x FOR PEER REVIEW 4 of 16 electron scavengers or donors (Ar atmosphere). In the latter case, the photogenerated charge carriers can only react with each other (i.e., recombine). As can be seen in Figure 4(a), after excitation with 355 nm, there are no significant differences in the transient absorption spectra under Ar or O2 atmospheres, while the experiments in Ar atmosphere and methanol atmosphere show a considerable difference. Moreover, this difference is larger at shorter wavelengths (~400 nm) than at the longer wavelength range (~650 nm). On the other hand, as revealed in Figure 4(b), the transient absorption spectra obtained after excitation with 455 nm in all the atmospheres are, within experimental error, identical to each other.   (Figure 5(a)), both under Ar and O2 atmospheres the transient absorption strongly increases right after the laser pulse and is followed by a rapid decrease down to ~15% of the initial value at 8 μs. In the case of the methanol atmosphere, however, the transient absorption is initially lower than the other two cases and increases with time in the investigated window. By comparing the transient absorption decays after excitation with 355 nm and 455 nm, Figure 5(a) and Figure 5(b), it can be seen that the initial absorptions after excitation with 455 nm are much lower than those values after excitation with 355 nm. After excitation with 455 nm the decrease of the transient absorption is significantly slower and decreases only to ~60% of the initial value after 8 μs. Furthermore, at this excitation wavelength no significant difference was observed regarding the type of atmosphere.  The behavior of K-7000 under visible-light excitation was compared with that of Hombikat UV100 in Ar atmosphere. Figure 6a shows the transient absorption decays in K-7000 and UV100 powders detected at 650 nm, after excitation with 455 nm. While UV100 shows virtually no transient absorption, a significant signal with a long-lived component can be observed for K-7000. Similarly, the transient absorption spectrum of UV100 ( Figure 6b) at 100 ns after the pulse is negligible with respect to that of K-7000. The behavior of K-7000 under visible-light excitation was compared with that of Hombikat UV100 in Ar atmosphere. Figure 6(a) shows the transient absorption decays in K-7000 and UV100 powders detected at 650 nm, after excitation with 455 nm. While UV100 shows virtually no transient absorption, a significant signal with a long-lived component can be observed for K-7000. Similarly, the transient absorption spectrum of UV100 ( Figure 6(b)) at 100 ns after the pulse is negligible with respect to that of K-7000.

Photocatalytic Experiments
In order to evaluate the visible-light activity of K-7000, we performed photocatalytic experiments using methanol and chlorpromazine as organic model compounds. For that matter, we analyzed their conversion in the presence of a K-7000 suspension upon irradiation at 455 nm. The results of these experiments are shown in Figure 7.

Photocatalytic Experiments
In order to evaluate the visible-light activity of K-7000, we performed photocatalytic experiments using methanol and chlorpromazine as organic model compounds. For that matter, we analyzed their conversion in the presence of a K-7000 suspension upon irradiation at 455 nm. The results of these experiments are shown in Figure 7. The behavior of K-7000 under visible-light excitation was compared with that of Hombikat UV100 in Ar atmosphere. Figure 6(a) shows the transient absorption decays in K-7000 and UV100 powders detected at 650 nm, after excitation with 455 nm. While UV100 shows virtually no transient absorption, a significant signal with a long-lived component can be observed for K-7000. Similarly, the transient absorption spectrum of UV100 ( Figure 6(b)) at 100 ns after the pulse is negligible with respect to that of K-7000.

Photocatalytic Experiments
In order to evaluate the visible-light activity of K-7000, we performed photocatalytic experiments using methanol and chlorpromazine as organic model compounds. For that matter, we analyzed their conversion in the presence of a K-7000 suspension upon irradiation at 455 nm. The results of these experiments are shown in Figure 7.  The oxidation of methanol was monitored by assessing the production of formaldehyde throughout the reaction period ( Figure S2). After 60 min of visible light irradiation in the presence of K-7000, only a slight degradation of methanol was observed.
In the case of chlorpromazine, it was found to be stable in the absence of K-7000 throughout the photolytic reaction time ( Figure S3) under visible light irradiation (455 nm), as expected from the very low overlap between its absorption spectrum and the emission spectrum of the light source ( Figure S4). In the presence of K-7000, however, chlorpromazine is completely converted within the first 30 min of irradiation. The main product of this reaction is chlorpromazine sulfoxide, identified by comparing its retention time with a standard sample. The sulfoxide seems to be stable with respect to its photocatalytic conversion, since its concentration increases monotonically before reaching a constant value at~30 min ( Figure S5). As a comparison, we have also investigated the photocatalytic conversion of chlorpromazine in the presence of Hombikat UV100 under visible light irradiation under identical conditions. However, in this case, the conversion of chlorpromazine was negligible ( Figure S6).

Discussion
The XRD analysis of the K-7000 powder ( Figure 1) confirms that anatase is the only crystalline phase significantly present, as also reported by Ząbek et al. [15]. Thus, no effect of the carbon-based layer on the crystalline structure of the titanium dioxide is observed. The anatase phase of the photocatalyst was further evinced by the UV-vis reflectance measurements (Figure 2a). From these, a bandgap value of 3.31 eV was found ( Figure S1), which was in good agreement with previous reports [15,17]. Moreover, the comparison of the absorption properties of K-7000 with those of UV100 (pure anatase) reveals a broad absorption band in the visible region for K-7000, not observed in anatase and, thus, ascribed to the carbon-based layer. The flatband potential of K-7000 was determined via Mott-Schottky measurements to be −0.46 V vs. NHE at pH 7 (Figure 3a), similar to a previously reported value for its quasi-Fermi level (−0.50 V vs. NHE at pH 7) [15]. Both values match very well with those measured for anatase TiO 2 (Hombikat UV100) and, therefore, it can be concluded that the modification performed on K-7000 does not significantly affect the position of its conduction band.
Transient absorption spectroscopy measurements were performed in different atmospheres using excitation with UV light (355 nm) and visible light (455 nm). In the Ar atmosphere the only possible fate of the photogenerated charge carriers is recombination with each other, since no other species are present to react with them. Since oxygen can act as an electron scavenger, it would be expected for it to affect the transient absorption spectra and decays. However, no difference can be observed between inert and O 2 atmospheres in both transient absorption spectra and transient decays, irrespective of the excitation wavelength. The standard reduction potential of oxygen is −0.33 V vs. NHE [18], i.e., more positive than the flatband potential of K-7000 (−0.46 V vs. NHE in neutral solution). Therefore, from a thermodynamic point of view, the transfer of electrons from the conduction band of TiO 2 to oxygen molecules is expected to happen. The scavenging of photogenerated electrons in TiO 2 by O 2 molecules in the gas atmosphere has been studied by Yamakata et al., who found a characteristic reaction time of tens of microseconds. Consequently, no significant reaction can be observed before 10 µs [19]. Accordingly, O 2 shall not affect the transient absorption spectra in the considered time scale. For longer time scales it would be possible to observe an effect of the O 2 atmosphere; however, our focus is on the oxidative half-reaction, and thus, for consistency, we limited ourselves to an 8 µs window.
Methanol is a well-known hole scavenger and reacts in a first step to form an α-hydroxyl radical (Equation (1)) [20,21]. For this oxidation reaction, a potential of 1.03 V vs. NHE is required at pH 7 [22]. With the flatband potential of −0.46 V vs. NHE and the determined bandgap of 3.31 eV, the valence band edge can be calculated to be at 2.85 V vs. NHE. For the abstraction of an electron from a methanol molecule, the valence band edge and thus the potential of the photogenerated holes must be more positive compared to the redox potential of the methanol oxidation. Accordingly, this reaction is likely to take place. In general, a transient absorption spectrum contains contributions from both, photogenerated electrons and holes. Herein, methanol acts as a hole scavenger; therefore, only the spectrum of the remaining electrons can be detected. Furthermore, since there is no electron scavenger present, photogenerated electrons accumulate in the particles, a well-known phenomenon for TiO 2 [23]. It can be concluded from the spectrum (Figure 4a) that the photogenerated electrons mainly absorb at higher wavelengths, as reported before [24]. This explains the observed differences in the spectra under Ar and methanol atmospheres.
For completeness, we note that in the absence of oxygen, the formed α-hydroxyl radicals react in a further step to produce formaldehyde (Equation (2)). This reaction possesses a redox potential of −1.41 V vs. NHE at pH 7 [22]. According to this potential, which is more negative than the flatband potential of K-7000 (−0.46 V vs. NHE), the formation of formaldehyde can occur by the injection of an electron into the conduction band of TiO 2 , known as the current doubling effect [21].
The transient absorption spectroscopy experiments using 455 nm light excitation offer a different picture. The photon energy (2.73 eV) is not sufficient for bandgap excitation in anatase TiO 2 , as illustrated by the bandgap of K-7000 (3.31 eV). Accordingly, the pure anatase material UV100 shows no transient response after visible light excitation ( Figure 6), with the exception of some experimental noise. On the contrary, a notable and long-lived signal can be detected in K-7000 under the same conditions, originated from its light absorption in the visible range, as described above. Regarding the physical origin of such absorption, we recall that investigations performed in the group of Prof. Kisch have determined that, in K-7000, carbon is not incorporated into the lattice, as originally thought, but is rather deposited on the surface in the form of a molecular sensitizer [15]. This important distinction may have a profound impact on the photocatalytic mechanism under visible light, since doping usually causes bandgap narrowing, as demonstrated, for instance, in N-and S-doped TiO 2 [25,26]. In agreement with the molecular sensitizer description, we found no bandgap narrowing for K-7000, but rather an absorption band in the visible range related to the sensitizer. Incidentally, the sensitizer-TiO 2 assembly can be treated akin to dye-sensitized solar cells, where the electronic structure of the former is characterized by defined HOMO (highest occupied molecular orbital) and LUMO (lowest unoccupied molecular orbital) levels [27]. Therefore, under excitation with visible light, the transition does not involve the valence band and the conduction band of TiO 2 , but rather the HOMO and LUMO levels of the sensitizer.
In coincidence with the experiments using excitation with UV light, no difference in the spectra and decays in Ar and oxygen atmosphere was observed after visible light excitation of K-7000 (Figures 4  and 5). This indicates that no reaction takes place with oxygen in the considered time scale. A plausible explanation is that, after excitation of the carbon-based sensitizer, electrons are injected from the LUMO to the conduction band of TiO 2 , as known for dye-sensitized solar cells [27]. Therefore, under visible light irradiation, oxygen reduction is ultimately limited in the same way as for bandgap excitation [19].
At odds with the transient absorption results using UV excitation, we observed no differences in the decays and spectra in methanol atmosphere with respect to inert atmosphere, indicating that no degradation of methanol takes place within the considered time scale under visible light excitation. Furthermore, the decays observed in all atmospheres after excitation with visible light were found to be rather long-lasting. This behavior is comparable to the decay obtained in methanol atmosphere after excitation with UV light (Figure 5a), where electrons accumulate in the TiO 2 particles. In the present case, electron transfer from the LUMO of the sensitizer to the conduction band of TiO 2 gives rise to their accumulation, since either no electron scavenger is present (Ar or methanol) or the rate of electron consumption is very low (O 2 ). Concomitantly, the radical cations of the sensitizer are produced after electron injection, which may lead to a bleaching of the sensitizer [28].
To obtain further insights into the oxidation of methanol over K-7000 under visible light irradiation (or the lack thereof), we performed photocatalytic experiments under continuous irradiation in aerated aqueous suspensions. After 60 min of irradiation in the presence of K-7000, less than 2% of a 100 µM methanol solution was transformed to formaldehyde (Figure 7 and Figure S2). These observations are in good agreement with the transient absorption spectroscopy results, revealing that, under visible light irradiation, the degradation of methanol is negligible or, at best, slow. To understand this behavior, the possible degradation mechanisms of methanol under irradiation with UV light as well as visible light in the presence of the photocatalyst must be discussed. Generally, after excitation with UV light, electron-hole pairs are generated in the TiO 2 particles (Equation (3)). In this case, the carbon-based sensitizer is possibly excited as well. However, considering the smaller amount of the sensitizer layer compared to TiO 2 , the excitation of the sensitizer can be neglected in the mechanism. The photogenerated electrons can react with oxygen molecules forming a superoxide radical (O − 2 ) (Equation (4)), while the photogenerated holes can be trapped either at terminal protonated bridging oxygen ions (OH − br ) to form a hydroxyl radical (-OH br ) (Equation (5)) or at adsorbed methanol molecules (CH 3 OH ads ) forming an α-hydroxyl radical ( CH 2 OH) (Equation (6)). Methanol molecules in aqueous solution (CH 3 OH aq ) can be oxidized via hydroxyl radicals yielding the same radical (Equation (7)). In an aerated suspension, the produced α-hydroxyl radicals are further oxidized to formaldehyde through reaction with oxygen molecules (Equation (8)) [29].
On the other hand, after excitation with visible light the mechanism will be different, since the excitation energy is not enough to generate electron-hole pairs in TiO 2 . Thus, under visible light irradiation, only the carbon-based sensitizer layer (CS) can be excited (CS * ), after which an electron is transferred from the LUMO to the conduction band of the TiO 2 (Equation (9)), simultaneously forming the radical cation of the sensitizer (CS + ). Afterwards, the electron in the conduction band can react with molecular oxygen (Equation (4)), while the sensitizer radical cation could react with methanol to form an α-hydroxyl radical ( CH 2 OH) (Equation (10)). Exactly as after irradiation with UV light, this radical can react further with molecular oxygen, yielding formaldehyde (Equation (8)).
However, the reaction shown in Equation (10) does not occur at a significant rate, since neither the transient absorption spectroscopy measurements nor the photocatalytic experiments suggest a degradation of methanol. The reason for the lack of reaction is discussed below. In passing, we represent the different reactions and processes occurring after excitation with UV and visible light in Scheme 1.
In contrast to methanol, chlorpromazine, which belongs to the phenothiazines group, could be completely degraded within the first 30 min of continuous visible light irradiation in the presence of K-7000 (Figure 7). In order to explain this difference, the mechanism of the visible light induced conversion of chlorpromazine must be taken into account. Upon visible light irradiation, excitation of the carbon-based sensitizer leads to electron transfer to the conduction band of TiO 2 (Equation (9)). In the presence of oxygen, these electrons (slowly) reduce it, forming superoxide radicals (Equation (4)). The sensitizer radical cation (CS + ) is able to oxidize chlorpromazine to form a chlorpromazine radical cation (Equation (11)), simultaneously regenerating the sensitizer. As reported for similar compounds [30], the next step is most likely the reaction between the produced chlorpromazine radical cation and a superoxide radical to form a persulfoxide molecule (Equation (12)). The persulfoxide molecule forms a thiadioxirane (Equation (13)), which reacts with a further chlorpromazine molecule yielding two chlorpromazine sulfoxide molecules (Equation (14)), as expected from the mechanism proposed by Somasundaram and Srinivasan for the oxidation of aryl methyl sulfides in the presence of irradiated TiO 2 [30]. In contrast to methanol, chlorpromazine, which belongs to the phenothiazines group, could be completely degraded within the first 30 minutes of continuous visible light irradiation in the presence of K-7000 (Figure 7). In order to explain this difference, the mechanism of the visible light induced conversion of chlorpromazine must be taken into account. Upon visible light irradiation, excitation of the carbon-based sensitizer leads to electron transfer to the conduction band of TiO2 (Equation 9). In the presence of oxygen, these electrons (slowly) reduce it, forming superoxide radicals (Equation 4). The sensitizer radical cation (CS •+ ) is able to oxidize chlorpromazine to form a chlorpromazine radical cation (Equation 11), simultaneously regenerating the sensitizer. As reported for similar compounds [30], the next step is most likely the reaction between the produced chlorpromazine radical cation and a superoxide radical to form a persulfoxide molecule (Equation 12). The persulfoxide molecule forms a thiadioxirane (Equation 13), which reacts with a further chlorpromazine molecule yielding two chlorpromazine sulfoxide molecules (Equation 14), as expected from the mechanism proposed by Somasundaram and Srinivasan for the oxidation of aryl methyl sulfides in the presence of irradiated TiO2 [30]. In contrast to methanol, chlorpromazine, which belongs to the phenothiazines group, could be completely degraded within the first 30 minutes of continuous visible light irradiation in the presence of K-7000 (Figure 7). In order to explain this difference, the mechanism of the visible light induced conversion of chlorpromazine must be taken into account. Upon visible light irradiation, excitation of the carbon-based sensitizer leads to electron transfer to the conduction band of TiO2 (Equation 9). In the presence of oxygen, these electrons (slowly) reduce it, forming superoxide radicals (Equation 4). The sensitizer radical cation (CS •+ ) is able to oxidize chlorpromazine to form a chlorpromazine radical cation (Equation 11), simultaneously regenerating the sensitizer. As reported for similar compounds [30], the next step is most likely the reaction between the produced chlorpromazine radical cation and a superoxide radical to form a persulfoxide molecule (Equation 12). The persulfoxide molecule forms a thiadioxirane (Equation 13), which reacts with a further chlorpromazine molecule yielding two chlorpromazine sulfoxide molecules (Equation 14), as expected from the mechanism proposed by Somasundaram and Srinivasan for the oxidation of aryl methyl sulfides in the presence of irradiated TiO2 [30].
Contrary to methanol, which was not significantly oxidized upon visible light irradiation of K-7000, chlorpromazine is able to react with the sensitizer radical cation, yielding its sulfoxide. The reaction of the organic compound with the sensitizer radical cation is essential for the further photocatalytic reaction steps. If this initial reaction does not take place, the further steps will not occur as well. To explain the reason why this initial reaction only happens with chlorpromazine and not with methanol, the redox potentials for the corresponding reactions have to be considered. As mentioned before, the required potential for the methanol oxidation (Equation 1) is 1.03 V vs. NHE at pH 7 [22]. The redox potential for the one-electron oxidation of chlorpromazine is 0.53 V vs. NHE (13) (13) (14) Contrary to methanol, which was not significantly oxidized upon visible light irradiation of K-7000, chlorpromazine is able to react with the sensitizer radical cation, yielding its sulfoxide. The reaction of the organic compound with the sensitizer radical cation is essential for the further photocatalytic reaction steps. If this initial reaction does not take place, the further steps will not occur as well. To explain the reason why this initial reaction only happens with chlorpromazine and not with methanol, the redox potentials for the corresponding reactions have to be considered. As mentioned before, the required potential for the methanol oxidation (Equation 1) is 1.03 V vs. NHE at pH 7 [22]. The redox potential for the one-electron oxidation of chlorpromazine is 0.53 V vs. NHE (14) Contrary to methanol, which was not significantly oxidized upon visible light irradiation of K-7000, chlorpromazine is able to react with the sensitizer radical cation, yielding its sulfoxide. The reaction of the organic compound with the sensitizer radical cation is essential for the further photocatalytic reaction steps. If this initial reaction does not take place, the further steps will not occur as well. To explain the reason why this initial reaction only happens with chlorpromazine and not with methanol, the redox potentials for the corresponding reactions have to be considered. As mentioned before, the required potential for the methanol oxidation (Equation (1)) is 1.03 V vs. NHE at pH 7 [22]. The redox potential for the one-electron oxidation of chlorpromazine is 0.53 V vs. NHE at pH 7 and, thus, more negative [31,32]. From this information, it can be concluded that the redox potential of the sensitizer radical cation is more positive than that of the chlorpromazine oxidation and either more negative or close to that of the methanol oxidation, which hinders its reaction with the latter. We note that, being electron transfer reactions, their kinetics are governed by Marcus theory [33,34]. As such, even if the reaction is moderately exergonic, it still may occur at a reduced rate due to the influence of the solvent reorganization energy. Additionally, we can estimate that the position of the LUMO must be located at a more negative potential than the conduction band edge of TiO 2 ; otherwise, no electron transfer from the excited sensitizer to the conduction band of TiO 2 would be possible. However, it has to be mentioned that the exact position of the LUMO cannot be determined from these experiments. All relevant redox potentials, as well as the conduction band and valence band position of the photocatalyst, including the possible positions of the LUMO and HOMO, are gathered in Scheme 2. [33,34]. As such, even if the reaction is moderately exergonic, it still may occur at a reduced rate due to the influence of the solvent reorganization energy. Additionally, we can estimate that the position of the LUMO must be located at a more negative potential than the conduction band edge of TiO2; otherwise, no electron transfer from the excited sensitizer to the conduction band of TiO2 would be possible. However, it has to be mentioned that the exact position of the LUMO cannot be determined from these experiments. All relevant redox potentials, as well as the conduction band and valence band position of the photocatalyst, including the possible positions of the LUMO and HOMO, are gathered in Scheme 2.

Scheme 2.
Band positions of the anatase photocatalyst K-7000 with the corresponding redox potentials of the oxygen reduction, methanol oxidation, and chlorpromazine (CPZ) oxidation at pH 7 and the possible positions of the LUMO and the HOMO of the sensitizer [18,22].
To confirm the results of the photocatalytic experiments, we performed transient absorption spectroscopy measurements with the K-7000 powder in the presence of pre-adsorbed chlorpromazine. Unlike the case of methanol ( Figure 5(b)), the transient absorption of bare K-7000 changes in the presence of chlorpromazine, giving rise to a signal attributable to the chlorpromazine radical cation [35] ( Figure S7). Thus, both the continuous irradiation experiments and the transient absorption spectroscopy measurements evince a prompt degradation of chlorpromazine and a negligible one for methanol. We note that chlorpromazine oxidation under UV irradiation of K-7000 has been the subject of a recent study [36] and, thus, we focused here on its visible light degradation.
Interestingly, the oxidation product of the chlorpromazine degradation, namely chlorpromazine sulfoxide, was found to be highly persistent throughout the visible light irradiation of K-7000 ( Figure  S5). This product was found to be a long-lasting compound upon visible light driven photocatalysis as well, as previously reported [36]. This result can be understood on the same grounds as for methanol and chlorpromazine. Although the redox potential for the one-electron oxidation of chlorpromazine sulfoxide has not, to the best of our knowledge, been determined, potentials in the range of 2.5 V to 2.8 V have been reported for similar sulfoxides [30], i.e., very close to the valence band edge of K-7000. By assuming a potential for chlorpromazine sulfoxide in this range, no oxidation can be expected, since this value is located at a position even more positive than that for the methanol oxidation. To confirm the results of the photocatalytic experiments, we performed transient absorption spectroscopy measurements with the K-7000 powder in the presence of pre-adsorbed chlorpromazine. Unlike the case of methanol (Figure 5b), the transient absorption of bare K-7000 changes in the presence of chlorpromazine, giving rise to a signal attributable to the chlorpromazine radical cation [35] ( Figure S7). Thus, both the continuous irradiation experiments and the transient absorption spectroscopy measurements evince a prompt degradation of chlorpromazine and a negligible one for methanol. We note that chlorpromazine oxidation under UV irradiation of K-7000 has been the subject of a recent study [36] and, thus, we focused here on its visible light degradation.
Interestingly, the oxidation product of the chlorpromazine degradation, namely chlorpromazine sulfoxide, was found to be highly persistent throughout the visible light irradiation of K-7000 ( Figure S5). This product was found to be a long-lasting compound upon visible light driven photocatalysis as well, as previously reported [36]. This result can be understood on the same grounds as for methanol and chlorpromazine. Although the redox potential for the one-electron oxidation of chlorpromazine sulfoxide has not, to the best of our knowledge, been determined, potentials in the range of 2.5 V to 2.8 V have been reported for similar sulfoxides [30], i.e., very close to the valence band edge of K-7000. By assuming a potential for chlorpromazine sulfoxide in this range, no oxidation can be expected, since this value is located at a position even more positive than that for the methanol oxidation.
In summary, our results with the commercial photocatalyst K-7000 illustrate that photogenerated charge carriers in visible light active materials may have a limited ability to drive oxidation reactions, which is at odds with the general wisdom stating that photogenerated holes in TiO 2 possess a very high oxidizing power. Moreover, while this fact may be detrimental for the application of such materials in environmental remediation, it may however be exploited to increase the selectivity of the oxidation half-reaction in photocatalytic processes. Instead of complete mineralization, under the right conditions it might be possible to selectively oxidize an unwanted compound into a valuable one [37].

Characterization of K-7000
X-ray diffraction (XRD) analysis was performed using Cu K α radiation (λ = 0.154178 nm) with a Bruker D8 Advance diffractometer. The patterns were recorded at room temperature in the 2θ range between 20 • and 80 • , with steps of 0.011 • in Bragg-Brentano geometry (θ-θ). The diffuse reflectance spectra of the photocatalysts powders were determined in the spectral range from 300 to 800 nm by means of a Varian Cary-100 UV-vis spectrophotometer equipped with an integrating sphere, with barium sulfate as a reflectance reference. The optical bandgap was determined using a Tauc plot analysis assuming indirect transitions ( Figure S1).
The flatband potential was determined through the Mott-Schottky method, performed using a ZENNIUM Electrochemical Workstation (ZAHNER-elektrik) with a 0.1 M KNO 3 solution as the electrolyte and a frequency of 100 Hz. For this measurement, a three-electrode electrochemical cell made of Teflon, with a Pt counter electrode and an Ag/AgCl/NaCl (3 M) reference electrode, was used. The screen-printing technique was applied to prepare the K-7000 and Hombikat UV100 films on a 3 cm × 3 cm fluorine-doped tin oxide (FTO) glass (Sigma-Aldrich) calcined at 450 • C, as the working electrode. The paste used to prepare the films was obtained following the procedure described by Ito et al. [38].

Transient Absorption Spectroscopy
The transient absorption spectroscopy measurements were performed with the catalyst powder inside a flat quartz cell. All samples were purged prior to the experiments with either argon, a mixture of nitrogen and methanol, or oxygen for 30 min. The experiments were performed in diffuse reflectance mode by means of an Applied Photophysics LKS 80 Laser Flash Photolysis Spectrometer with a pulsed Nd:YAG laser (Quantel, Brilliant B). For excitation of the samples, the third harmonic of the laser (355 nm, 6 ns pulses) or the third harmonic equipped with an optical parametric oscillator (OPO, 455 nm) was used with an average energy of 3 mJ·cm −2 . The laser light and the analyzing light (Xenon lamp, pulsed, Osram XBO, 150 W) were focused on the surface of the powder sample. The reflected light was guided to a detector (Hamamatsu PMT R928), which was connected with a 100 Ω resistance to an oscilloscope. The photomultiplier converted the incoming photons directly into a current. For every detection wavelength, a different voltage was applied to the photomultiplier to adjust the light level to a constant value. Absorbance values, Abs, were calculated from the reflected light before (J 0 ) and after the laser excitation (J). The values of the change of reflectance, ∆J, were obtained by applying the following equation: For each detection wavelength, a time scale of 10 µs was considered, 25 shots were averaged and the data points were reduced to 200. The spectra were recorded from 400 nm to 650 nm in 25 nm steps for excitation with 355 nm and from 550 nm to 650 nm in 25 nm steps for the excitation with 455 nm.

Photocatalytic Procedure
For the photocatalytic experiments, chlorpromazine and methanol were chosen as probe organic molecules. The irradiation source was a monochromatic light emitting diode (LED) with an emission maximum at 455 nm (Thorlabs, M455L3, 339 W·m −2 ; emission spectrum shown in Figure S5). The photocatalytic setup included a barrel-shaped borosilicate photoreactor (diameter: 6 cm, height: 3 cm) covered with a black polymer case, which had a connection on the top for the attachment of the LED light source. For these experiments, initially a 60 mL aqueous suspension of 100 µM chlorpromazine/methanol and 1 g·L −1 photocatalyst was prepared at its natural pH. The suspension was then stirred for 60 min under dark conditions in the photoreactor to reach the adsorption-desorption equilibrium. Afterwards, the LED light source began to irradiate the reaction medium for 60 min; throughout the reaction period, the suspension was purged with oxygen. At regular time intervals, samples were taken, centrifuged, and filtered (PVDF, 0.2 µm pore size). For the experiments with chlorpromazine, the samples were analyzed with a HPLC-UV-MS (ESI) device using an Alliance 2795-HT HPLC (Waters, UK) coupled with a 1050 UV-detector (type 79853C, detection wavelength: 254 nm, HP, USA) and an LCT Premier ESI mass spectrometer (Waters, UK). The resulting data of the conversion of chlorpromazine into its main product chlorpromazine sulfoxide were presented in terms of the change of their concentration (µM) versus irradiation time (min). For the methanol degradation, the production of formaldehyde was monitored by derivatizing it with Nash's reagent [39] and then following the colored product with a Varian Cary-100 UV-vis spectrophotometer. The conversion of methanol was compared with that of chlorpromazine in terms of the concentration at the sampling time divided by the initial concentration after the equilibrium (C/C 0,e ). Photocatalytic experiments were performed at least twice to ensure reproducibility.

Conclusions
We studied the reactivity upon UV and visible light irradiation of the commercially available photocatalyst KRONOClean 7000. Its characterization revealed that its crystalline structure, flatband potential, and light absorption in the UV range closely match those of pure anatase TiO 2 . Its visible light photocatalytic activity originates from a long-tailed absorption in this range, which is ascribed to the carbon modification employed for its preparation. By means of transient absorption spectroscopy, we investigated charge carrier generation and recombination in the material using Ar, O 2 , and methanol atmospheres. Excitation with UV light did not show a significant reaction of photogenerated charge carriers with oxygen in the investigated time window, explained by a slow electron transfer. In the presence of methanol, we observed a gradual accumulation of electrons due to its fast reaction with photogenerated holes. Upon excitation with visible light, all spectra and decays show similar features, indicating that neither with methanol nor with O 2 a reaction takes place to a significant degree. Complementarily, we performed photocatalytic reactions under visible light irradiation, after which we observed no methanol oxidation, but a fast degradation of chlorpromazine. We rationalized these results by considering K-7000 akin to a dye-sensitized system, in which the carbon-based layer acts as a molecular sensitizer. As such, excitation with UV light leads to the generation of electron-hole pairs in TiO 2 , while visible light causes the excitation of the sensitizer followed by electron injection into TiO 2 and subsequent formation of the radical cation. Ultimately, the oxidation ability of the material depends on the difference between the redox potential of the radical cation and that of the molecule to be oxidized: while the oxidation potential of methanol is too negative for it to react, that of chlorpromazine allows the reaction to swiftly occur.
All in all, our results show that excitation of the photocatalyst with visible light leads to a lower reactivity than for UV light, limited by the redox potential of the radical cation of the sensitizer. Moreover, although photogenerated charge carriers in this visible light active material may have a limited ability to drive oxidation reactions, under the right circumstances this could be exploited to increase the selectivity of photocatalytic transformations.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4344/9/8/697/s1, Figure S1: Determination of the bandgap of K-7000 and UV100 via the Tauc plot, Figure S2: Kinetic profile for the formation of formaldehyde upon the photocatalytic oxidation of methanol (C 0 = 100 µM) under visible light irradiation in the presence of K-7000, Figure S3: Photolysis and photocatalytic conversion (in the presence of K-7000) of chlorpromazine under visible light irradiation, Figure S4: Absorption spectrum of a 100 µM aqueous solution of chlorpromazine (black line), and emission spectrum of the employed light source (blue line), Figure S5: Photocatalytic conversion of chlorpromazine to the main product chlorpromazine sulfoxide in the presence of K-7000 under visible light irradiation, Figure S6: Photocatalytic conversion of chlorpromazine under visible light irradiation in the presence of K-7000 and Hombikat UV100, Figure S7: (a) Transient absorption decays of K-7000 observed at 550 nm and (b) transient absorption spectra of K-7000 measured at 100 ns after the laser excitation in an Ar atmosphere and in the presence of chlorpromazine after excitation with 455 nm.
Funding: This research was supported by Saint-Petersburg State University via a research Grant ID 32706707. The publication of this article was funded by the Open Access fund of Leibniz Universität Hannover.