**Enhancement of Hydrogen Productions by Accelerating Electron-Transfers of Sulfur Defects in the CuS@CuGaS2 Heterojunction Photocatalysts**

#### **Namgyu Son, Jun Neoung Heo, Young-Sang Youn, Youngsoo Kim, Jeong Yeon Do \* and Misook Kang \***

Department of Chemistry, College of Science, Yeungnam University, Gyeongsan, Gyeongbuk 38541, Korea; sng1107@naver.com (N.S.); hjn2521@naver.com (J.N.H.); ysyoun@yu.ac.kr (Y.-S.Y.); kimys6553@yu.ac.kr (Y.K.)

**\*** Correspondence: daengi77@ynu.ac.kr (J.Y.D.); mskang@ynu.ac.kr (M.K.); Tel.: +82-53-810-3798 (J.Y.D.);

+82-53-810-2363 (M.K.); Fax: +82-53-815-5412 (M.K.)

Received: 29 November 2018; Accepted: 28 December 2018; Published: 4 January 2019

**Abstract:** CuS and CuGaS2 heterojunction catalysts were used to improve hydrogen production performance by photo splitting of methanol aqueous solution in the visible region in this study. CuGaS2, which is a chalcogenide structure, can form structural defects to promote separation of electrons and holes and improve visible light absorbing ability. The optimum catalytic activity of CuGaS2 was investigated by varying the heterojunction ratio of CuGaS2 with CuS. Physicochemical properties of CuS, CuGaS2 and CuS@CuGaS2 nanoparticles were confirmed by X-ray diffraction, ultraviolet visible spectroscopy, high-resolution transmission electron microscopy, scanning electron microscopy and energy dispersive X-ray spectroscopy. Compared with pure CuS, the hydrogen production performance of CuGaS2 doped with Ga dopant was improved by methanol photolysis, and the photoactivity of the heterogeneous CuS@CuGaS2 catalyst was increased remarkably. Moreover, the 0.5CuS@1.5CuGaS2 catalyst produced 3250 μmol of hydrogen through photolysis of aqueous methanol solution under 10 h UV light irradiation. According to the intensity modulated photovoltage spectroscopy (IMVS) results, the high photoactivity of the CuS@CuGaS2 catalyst is attributed to the inhibition of recombination between electron-hole pairs, accelerating electron-transfer by acting as a trap site at the interface between CuGaS2 structural defects and the heterojunction.

**Keywords:** hydrogen production; methanol photo-splitting; heterojunction; CuS@CuGaS2; electron-hole recombination

#### **1. Introduction**

Copper sulfide (Cu2−xS, 0< x < 1), a non-toxic and conductive chalcogen compound, has been continuously noted for its excellent photoelectric behavior, potential thermal/electrical properties, and unique biomedical properties for decades, and much extensive research on Cu2−xS micro/nano structures is still being actively conducted. In particular, micro/nanostructured Cu2−xS with well-controlled shapes, sizes, structures and compositions have already been applied as photocatalytic materials [1], energy conversion materials [2], biosensing materials [3], and bioimaging materials [4] and have shown reasonable results. However, comprehensive reviews of the Cu2−xS structure in-depth in applications are still lacking. Therefore, it is necessary to categorize new functions or orientations of Cu2−xS-based nanocomposites and to develop and improve their essential elements for specific applications. Many researchers have already published a number of strategies for synthesizing 0D (dimension) , 1D, 2D, and 3D micro/nanostructures (including polyhedra) [5–7], and their efforts have made important progress in identifying Cu2−xS micro/nanostructures. Furthermore, improved Cu2−xS

composites with hollow structures or super-lattices could be extended to a variety of applications in terms of performance [8]. Cu2−xS belongs to the covellite mineral group with a hexagonal crystal structure, and belongs to the crystal group of P63/mmc. However, bond lengths and angles can be varied in various ways depending on the oxidation state of the copper or other anion exchanges instead of S2<sup>−</sup> present in the surroundings. For example, Cu2−xS is very different from Cu2−xO but shows a similar structure to Cu2−xSe (klockmannite) [9]. CuS compound is paramagnetic due to the 3d9 electron arrangement, and some studies have reported that all Cu atoms in CuS have an oxidation state of Cu<sup>+</sup> based on XPS results [10]. However, it can be attributed to (Cu+)2Cu2+S2 with both Cu(I) and Cu(II) in the XRD crystal structure results [11]. There are many applications of Cu2−xS as a photocatalyst, among which Saranya et al. suggested that the morphology of CuS was influenced by the reaction time and surfactant, and its photocatalytic activity for decolorization of methylene blue (MB) dye under visible-light irradiation was 87% [12]. However, CuS is mainly used in combination with other types of photocatalysts rather than alone [13,14]. On the other hand, substitution of Cu2−xS with other metal ions instead of Cu(II) can produce a complex crystal structure with varying performance. For example, CuInS2 and CuGaS2, which are called CIS or CIGS, are used as light absorbers in thin film solar cells; they absorb visible light in a wide area and are stable to light, unlike Cu2−xS. In particular, Salak et al. reported that CuM3+S2, a chalcogen compound, formed Cu defects on tetrahedrons and facilitated the separation of electrons and holes, which could maintain photoactivity for a long time [15]. Yue et al. concluded that CuInS2 was most sensitive at 500 nm with an optimal apparent quantum yield of 23.85% [16]. Han et al. found that with an increase in the reaction temperature, the excitonic absorption peaks and band gap emission peaks were systematically red-shifted, thus exhibiting a quantum confinement effect, and the CuGaS2 quantum dots showed promising visible-light-driven photocatalytic activity during degradation of rhodamine 6G [17]. We have already confirmed in previous studies that the CuS@CuInS2:In2S3 catalyst has the ability to decompose water to produce hydrogen with high efficiency. Especially, it was found that the CuInS2 layer inserted between CuS and In2S3 acts as an electron-rich interface to accelerate the reduction of water in this layer [18]. Furthermore, we have also found in previous studies that Ga has excellent performance in decomposing methanol to produce hydrogen, and that the hydrogen reverse-spillover phenomenon of Ga has a great influence on the removal of hydrogen from methanol [19]. Despite many previous studies, the excellent photosensitivity of chalcogen compounds, including Cu2−xS, in multifunctional complexes is broadly applicable to a wide range of photochemical reactions, so research on chalcogen compounds remains of interest. In particular, if chalcogen compounds are expected to perform well in the photoreaction for hydrogen production from water decomposition and their photostability is guaranteed for a long time, development of the chalcogen photocatalyst will be quite a desirable area of study for the next generation of environmentally friendly energy sources.

Therefore, in this study, we applied two particles of CuS and CuGaS2 as base catalysts and applied them to hydrogen production from water degradation by hetero-connecting them between two particles. The effect of the pure structure or heterojunction structure of the two particles on photoactivity was investigated. Five types of catalysts were prepared: CuS, CuGaS2, 1.0CuS@1.0CuGaS2, 0.5CuS@1.5CuGaS2, and 1.5CuS@0.5CuGaS2. The molar ratios of the two particles at the heterojunction were CuS:CuGaS2 = 1:1, 0.5:1.5, and 1.5:0.5, respectively, to determine which particles most influence catalytic activity.

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

#### *Characteristics of CuS, CuGaS2 and CuS@CuGaS2 Nanoparticles*

The XRD patterns (A) and high-resolution TEM (HRTEM) images (B) are shown in Figure 1 to confirm the crystallinity of synthesized CuS, CuGaS2 and heterojunction CuS@CuGaS2 nanoparticles. The main XRD peaks of CuS were observed at 2θ = 27.65◦ (101), 29.25◦ (102), 31.75◦ (103), 32.77◦ (006), 38.77◦ (105), 47.89◦ (110), 52.61◦ (108), 59.24 (116), 73.87◦ (208) and 78.96◦ (213) and were

assigned to the covellite CuS of the hexagonal crystal structure (P63 / mmc space group, JCPDS card No. 01-078-0876) [20]. On the other hand, the peaks at 31.74◦ and 46.11◦ correspond to Cu2S (JCPDS card no. 00-053-0522) of cubic crystal structure, meaning that the two structures are finely mixed [21]. The XRD patterns of the CuGaS2 nanoparticle showed a CuGaS2 peak with a tetragonal crystal structure (I-42d space group, JCPDS card no. 01-085-1574) [22], although CuS was mixed. The main XRD peaks of CuGaS2 were assigned to 2θ = 29.11◦ (112), 33.49◦ (200), 48.63◦ (204) and 57.20◦ (312). Meanwhile, the XRD pattern of the heterojunction CuS@CuGaS2 nanoparticles was very similar to the XRD pattern of the CuGaS2 corresponding to the shell, but there was a slight difference in the intensity and position of the peaks as the ratio of CuS:CuGaS2 varied. In particular, in the 0.5CuS@1.5CuGaS2 sample, the peak corresponding to the (112) crystal plane migrated at a higher angle than CuGaS2. This is probably due to the small ionic radius of Ga3+ ions compared to Cu2+ [23], and it is expected that lattice parameters will be reduced according to Bragg's law [24], nλ = 2d sin (θ) (where n is the order of reflection; the wavelength of the X-rays, d = the distance between two layers of the crystals, and θ = the angle of the incident light). It can be expected that lattice defects are formed as compared with pure CuS or CuGaS2, since Ga2+ ions can be incorporated into the CuS lattice or the lattice gap in the process of heterojunction. This can act as an active site of the catalyst and enhance catalyst performance. Figure 1B) shows the high-resolution TEM (HRTEM) (a), the selected area electron diffraction (SAED) (b), and the elemental mapping image (c) of the heterojunction 0.5CuS@1.5CuGaS2 particles. This result not only shows the overall shape of the particles, but also explains the intrinsic crystal structure of the particles based on the lattice parameter values in relation to the XRD results. The 0.5CuS@1.5CuGaS2 particles are shown in polycrystalline form as aggregates of single crystals with different orientations, which are evidenced by the lattice images and the SAED patterns. In general, when a certain point in a SAED pattern is clearly marked, it signifies a single crystal, and when a continuous ring is drawn, it signifies a polycrystalline. Therefore, heterojunction 0.5CuS@1.5CuGaS2 particles appeared to be polycrystalline, and lattice patterns of CuS and CuGaS2 were observed. We can expect CuS and CuGaS2 to be bonded as shown by the difference in shading in the TEM image. A lattice corresponding to 0.288 nm (103 diffraction plane) and 0.305 nm (102 diffraction plane) of CuS was observed in the bright portion, and a lattice pattern of 0.266 nm (200 diffraction plane) and 0.188 nm (204 diffraction plane) of CuGaS2 was observed in dark areas. In particular, the 200 diffraction plane of CuGaS2 has a very distinct lattice pattern, and a clear diffraction spot was also observed in the SAED pattern. This is consistent with the XRD results of the heterojunction 0.5CuS@1.5CuGaS2 particles and demonstrates that the particle is a continuous polycrystalline structure with a ring pattern with partially defined diffraction spots. On the other hand, the element mapping (c) results show that the Cu, Ga and S elements are uniformly distributed in the 0.5CuS@1.5CuGaS2 particles, thus demonstrating the microstructure and composition of the heterogeneous bonded particles.

**Figure 1.** X-ray diffraction (XRD) patterns (**A**) and high-resolution transmission electron microscopy (HRTEM) images (**B**) of prepared samples.

Figure 2 shows the scanning electron microscope (SEM) image and energy-dispersive X-ray spectroscopy (EDS) analysis of synthesized CuS, CuGaS2, and heterojunction CuS@CuGaS2 nanoparticles, showing the composition of the components present on the surface of the particles. Figure 2A shows a SEM image of each sample, showing significant aggregation between the particles in the CuGaS2 sample compared to CuS. On the other hand, the heterojunction CuS@CuGaS2 sample inhibited the agglomeration of particles and the particle size became smaller. Figure 2B shows the EDS spectra of each sample and the atomic composition ratios are shown in the Table 1. Determination of the atomic ratio of the main metal species in the catalyst is very important because it relates to the density of the crystal lattice defects [25]. CuS and CuGaS2, heterojunction CuS@CuGaS2 particles show that Cu, Ga and S atomic components are precisely contained, and no other components are included. The atomic composition of pure CuS is close to the ideal stoichiometric mole fraction with a Cu:S ratio of 46.74:53.26. The composition of the CuGaS2 sample was 31.46:27.26:41.27 with a Cu:Ga:S ratio slightly different from the stoichiometric ratio, but this can be predicted as a limitation of the EDS surface methodology. In addition, the heterojunction CuS@CuGaS2 samples had a relatively low proportion of Ga. It is also expected that the difference in the sizes of Cu and Ga ions causes Ga to enter into the lattice of the crystal structure to reduce the amount of Ga exposed on the surface. Taking these factors into account, the overall molar ratio of Cu to Ga, S was almost quantitatively and reliably obtained.

**Figure 2.** The SEM images (**A**) and energy-dispersive X-ray spectroscopy curves (**B**) of the CuS, CuGaS2 and CuS@CuGaS2 catalysts.



Figure 3 shows the UV-visible reflectance spectra of the synthesized CuS, CuGaS2 and heterojunction CuS@CuGaS2 nanoparticles. UV-visible absorption spectroscopy is widely used to investigate microscopic changes caused by the chemical characteristics of the surface of the particles [26], and the optical band gap can be calculated through absorption spectra. Figure 3A clearly shows that the CuS, CuGaS2 and heterojunction CuS@CuGaS2 samples show absorption spectra in the region of 300~800 nm, and CuS in particular exhibited a pronounced absorption shoulder at about 620 nm. In the CuGaS2 and CuS@CuGaS2 samples, the apparent absorption shoulder peak

disappeared, but the overall absorption range was similar to CuS. These results suggest that CuS, CuGaS2 and CuS@CuGaS2 samples can be used as promising photocatalytic materials to absorb visible light. Based on the absorption spectrum, the band gap was calculated by the Tauc equation [27], αhν = A (hν = Eg) n. Where hυ is the photon energy, α is the absorption coefficient, A is the constant relative material, and n is the value that depends on the transitional nature (2 is direct allowed transition, 2/3 is direct suppressed transition, 2/3 is indirectly permissible transition). In addition, the energy band gap can be predicted from the wavelength extrapolated from the exciton peak called λ1/2, or the point where the end of the absorption curve meets the *x* axis. The band gaps of pure CuS samples and CuGaS2 were 1.63 and 2.34 eV, respectively. This value is similar to the band gap reported in other papers [28]. The bandgap of the heterojunction 0.5CuS@1.5CuGaS2 particles was 2.32 eV. As CuGaS2, which has a longer band gap, is bonded to CuS, CuGaS2 first acts as a photosensitizer, and excited electrons can be transferred to CuS. If the bandgap is long, the recombination between the photogenerated electrons and the hole pair is delayed, and the photoactivity can be increased.

**Figure 3.** UV–visible spectroscopy curves (**A**) and Tauc plots (**B**) of CuS, CuGaS2 and CuS@CuGaS2 catalysts.

In general, the behavior of photogenerated carriers is closely related to photocatalytic activity. The generation, separation, transport and recombination of photogenerated electron-hole pairs have a great influence on photocatalytic activity [29]. Therefore, the photocurrent measurement and the photoluminescence measurement results are shown in Figure 4 in order to understand the separation efficiency and recombination characteristics of the photogenerated electrons and hole pairs. Figure 4A shows the results of measuring the photocurrent value when the light was irradiated by controlling the switch of the light source at intervals of 30 s. Electrons were excited by the irradiated light to generate excited electron and hole pairs. Their separation efficiency and mobility are closely related to the photocurrent value [30]. The photocurrent density values were increased in the order of CuS < 1.5CuS@0.5CuGaS2 < CuGaS2 < 1.0CuS@1.0CuGaS2 < 0.5CuS@1.5CuGaS2. The pure CuS catalyst showed only a slight tendency for photocurrent density to drop momentarily when the light was turned off. In general photocurrent results, the photogenerated holes migrate to the catalyst surface and are captured or trapped by the reduced species in the electrolyte, and the electrons undergo backside contact through the catalyst, leading to an increase in the initial anodic photocurrent. Then, after the competitive separation of the electron and hole pairs and the equilibrium of the recombination, the photocurrent is kept constant, while the CuS temporarily decreases the photocurrent. This is presumably due to the fact that the traced holes on the catalyst surface are not captured or trapped by the reduced species in the electrolyte, but instead are competitively recombined with the electrons in

the conduction band of the catalyst [31]. On the other hand, the photocurrent density of CuGaS2 and heterojunction CuS@CuGaS2 catalysts was not only increased, but also showed excellent stability and reliability. It is believed that the addition of the Ga dopant induces more exciton formation and that structural or surface defects caused by heterojunction act as capture sites, or accelerate electron-transfer, resulting in more efficient photogenerated charge separation.

**Figure 4.** Photocurrent responses (**A**) and photoluminescence spectra (**B**) of CuS, CuGaS2 and CuS@CuGaS2 catalysts.

The transfer process of the photogenerated charge carrier is closely related to photoluminescence, and the photoluminescence measurement results are shown in Figure 4B. The position and intensity of the emission peak of photoluminescence (PL) differs depending on the kind of the catalyst and the vacancy [32]. Generally, as the intensity is increased, the recombination of the photogenerated electron and the hole pair is promoted and the photocatalytic activity is decreased [33]. When an excitation wavelength of 300 nm was irradiated, an emission peak was observed at about 420 nm in all of the CuS, CuGaS2 and heterojunction CuS@CuGaS2 nanoparticles. Especially, when Ga was added, the photoluminescence peak intensity of CuGaS2 and the heterogeneous CuS@CuGaS2 catalyst was slightly shifted to the long wavelength side. As the number of defect lattices generated in the Ga doping and heterojunction process increases, the electron trap site increases, which results in suppression of the recombination of photogenerated electrons and hole pairs. On the other hand, CuGaS2 and heterojunction CuS@CuGaS2 catalysts to which Ga was added, in comparison with pure CuS, had new emission peaks near 400 nm and 480 nm. According to Mehmood et al. [34], the blue emission peak at around 400 nm is due to a high energy defect due to the dopant. In addition, the electrons trapped by the dopant cannot generate excitons because they are bound by surface oxygen defects and other defects, thereby reducing the overall PL intensity. The blue-green emission peak at about 480 nm corresponds to the radiative transition of an electron to the deep donor level of the metal interstitials to an acceptor level of neutral Vmetal [35]. In contrast to photocurrent measurement, the photoluminescence intensity decreased in the order of CuS > 1.5CuS@0.5CuGaS2 > CuGaS2 > 1.0CuS@1.0CuGaS2 > 0.5CuS@1.5CuGaS2. As a result, the heterojunction 0.5CuS@1.5CuGaS2 catalyst showed the lowest photoluminescence intensity, and the Ga doping and heterojunction structure can play an important role in increasing the light efficiency by slowing the recombination between exciton and hole pairs the most.

Based on the results of photoluminescence measurements, further intensity modulated photovoltage spectroscopy (IMVS) measurements were performed to compare the excited electron recombination lifetime for CuS, CuGaS2 and heterojunction CuS@CuGaS2 catalysts, and the results are shown in Figure 5A. IMVS is a useful method to study lifetime of electrons, which relates to the electron recombination process [36]. The IMVS plot (A) of all samples showed a semicircular shape, and the larger the size of the semicircle in the IMVS results, the longer the recombination lifetime [37]. The electron recombination lifetime (B) was calculated from the IMVS plot using the equation [38] sr = 1/2 πfmin , where fmin is the frequency of the minimum imaginary component of the plot. As with the results of the photoluminescence measurement, the electron recombination lifetime increased in the order of: CuS < CuGaS2 < CuS@CuGaS2. The increased IMVS electron lifetime indicates that the residence time at the electron trap site increases. Figure 5B shows that CuGaS2 increases the recombination lifetime due to the effect of Ga dopant compared to CuS, but shows the longest recombination life when the two catalysts are hetero-bonded. These results indicate that defects or catalyst surfaces formed during the heterojunction process generate more trap sites, and that the electron recombination lifetime is accelerated while accelerating electron-transfer, which may exert an excellent photocatalytic activity.

**Figure 5.** Intensity modulated photovoltage spectroscopy (IMVS) curves (**A**) and electron recombination lifetime (**B**) of CuS, CuGaS2 and CuS@CuGaS2 catalysts.

XPS was measured to investigate the chemical states of Cu, Ga and S ions in the surface state of the heterojunction 0.5CuS@1.5CuGaS2 sample. The results are shown in Figure 6. In Cu atomic spectra, peaks corresponding to Cu 2p3/2 and 2p1/2 were observed at 934.27 and 953.71 eV. A typical satellite peak of the Cu2+ oxidation state was observed at 942.78 and 963.06 eV, except for these two distinct peaks, indicating a defect in Cu2+. This implies a vacancy in the surface state that occurs in vacancies or 0.5CuS@1.5CuGaS2 heterojunction processes in skeletal distortions due to ion size differences due to Ga addition [39]. In the Ga atomic spectrum, peaks corresponding to 2p3/2 and 2p1/2 were observed at 1118.87 and 1145.80 eV, which corresponds to the Ga-S bond. On the other hand, three fitting curves were separated in the S atomic spectrum. The peak at low binding energy (162.00 eV) corresponds to

the Cu-S bond, and the peak at the high binding energy (165.28 eV) corresponds to the Ga-S bond [40]. Furthermore, the intermediate peak observed near 163.74 eV corresponds to sulfur vacancy, which is attributed to sulfur defects formed at the CuGaS2 or heterojunction interface [41]. These XPS results demonstrate the presence of structural defects on the heterojunction 0.5CuS@1.5CuGaS2 catalyst surface and predict that the site can act as a reactive site.

**Figure 6.** XPS spectra of heterojunction 0.5CuS@1.5CuGaS2 catalysts.

Figure 7 summarizes the evolution of hydrogen from the photo splitting of aqueous methanol solution to CuS, CuGaS2 and CuS@CuGaS2 catalysts. Figure 7A shows the amount of hydrogen generated under 365 nm UV light source conditions of the catalyst. The pure CuS catalyst showed little hydrogen evolution even after 10 h of reaction. On the other hand, the amount of hydrogen generation in the CuGaS2 catalyst was remarkably increased, and the amount of hydrogen produced after the reaction for 10 h reached 3000 μmol. In particular, the CuS@CuGaS2 and 0.5CuS@1.5CuGaS2 catalysts increased the amount of hydrogen generation further, reaching 3100 and 3250 μmol, respectively. Meanwhile, the heterojunction 1.5CuS@0.5CuGaS2 catalyst with a high CuS ratio produced less hydrogen than CuGaS2. This is considered to be due to the fact that the portion exposed on the surface of the catalyst has a large amount of CuS and is less influenced by CuGaS2. According to the previous studies [42], the photo splitting process of methanol aqueous solution follows the following reaction:

$$\text{CH}\_3\text{OH} + \text{H}\_2\text{O} \rightarrow \text{CO}\_2 + 6\text{H}^+ + 6\text{e}^- \tag{1}$$

$$6\text{H}^+ + 6\text{e}^- \rightarrow 3\text{H}\_2\tag{2}$$

$$\text{CH}\_3\text{OH} + \text{H}\_2\text{O} \rightarrow \text{CO}\_2 + 3\text{H}\_2\tag{3}$$

According to the above photolytic decomposition method of methanol, hydrogen and carbon dioxide are produced, but in this study, carbon dioxide was not observed because it exists as a CO2 ion in an aqueous solution. In addition, photolysis of methanol aqueous solution was further performed using a 150 W Xe lamp to confirm the catalytic activity in the visible region. The amount of hydrogen produced was reduced by about 1/20 compared to the UV light source, and the 0.5CuS@1.5CuGaS2 catalyst, which was heterogeneously bonded to the CuGaS2 catalyst, produced about 200 μmol of hydrogen. From these results, we have confirmed that CuGaS2 and the heterogeneous CuS@CuGaS2 catalyst exhibit optical activity even in the visible region, albeit in a small amount compared to UV light sources.

From these results, we proposed an improved photocatalytic decomposition of aqueous solution of CuS@CuGaS2 catalyst as shown in Scheme 1. The valence band, conduction band, and band gap values of CuS [18] and CuGaS2 [43] have already been reported in other studies and based on this, energy potential diagrams are shown together. In the heterojunction CuS@CuGaS2, the band gap of CuGaS2 contains CuS, and CuGaS2 first acts as a light-absorbing agent. Electrons are excited from the valence band to the conduction band by light irradiation, and electrons generated from CuGaS2 can move to the conduction band of the adjacent CuS. This transfer is thermodynamically favorable

by band gap alignment, and the photogenerated electrons react with H+ to produce H2. At this time, S2−/Sx <sup>2</sup><sup>−</sup> , which is a sacrifice material of the metal sulfide, captures holes by the following equation [44], and suppresses recombination between electron-hole pairs.

(1) 2 S2<sup>−</sup> +2h+ → S2 2− (2) S2<sup>−</sup> + 2h<sup>+</sup> → <sup>S</sup>

**Figure 7.** Evolution of H2 for methanol aqueous solution photo-splitting under UV light source (**A**) and visible light source (**B**) for CuS, CuGaS2 and CuS@CuGaS2.

**Scheme 1.** The expected mechanism for methanol aqueous solution photo-splitting in photosystem with CuS@CuGaS2 catalyst.

In addition, structural defects of the CuGaS2 catalyst caused by the interface between the heterojunction CuS@CuGaS2 catalyst and the GaGaS2 catalyst may act as a trap site to accelerate hole transfer and prolong the electron lifetime. In CuGaS2 structure, free electrons are gathered around Ga3+ for charge balance based on CuS structure, and sulfur vacancies due to structural defects are formed. These sites provide a place where the reactants can be adsorbed better, resulting in more products, and ultimately an improvement in photocatalytic activity. Moreover, structural defects at the CuS@CuGaS2 catalyst interface form quasi-continuous energy levels and reduce ohmic contact to induce ohmic contact [28]. This contact recombines the holes formed in VB of CuS and the excited electrons in CB of CuGaS2, which ultimately promotes the efficiency of photocatalytic activity by promoting the separation efficiency of the photogenerated charge pair in the heterojunction CuS@CuGaS2 complex catalyst.

#### **3. Experimental**

#### *3.1. Preparation of CuS, CuGaS2 and CuS@CuGaS2 Nanoparticles*

CuS and CuGaS2 nanoparticles were prepared using a typical sol-gel synthesis method [45]. Copper (II) nitrate trihydrate (Cu(NO3)2·3H2O, 99.0%, Junsei Chemical, Tokyo, Japan), Gallium (III) nitrate hydrate (Ga(NO3)3·xH2O, 99.9%, Alfa Aesar, Tewksbury, MA, USA) and Thiourea (CH4N2S, 98.0%, Junsei Chemical, Tokyo, Japan) were used as starting materials for Cu, Ga and S, respectively. First, in order to synthesize CuS, Cu(NO3)2·3H2O and CH4N2S were dissolved in ethylene glycol at a molar ratio of 1:2, mixed well and aged at 180 ◦C for 8 h. The resulting powder was treated at 400 ◦C for 4 h under an argon atmosphere to obtain black CuS nanoparticles. In the synthesis of CuGaS2 particles, only the molar ratio of Cu(NO3)2·3H2O, Ga(NO3)3·xH2O and CH4N2S was changed to 1:1.25:4 during the synthesis of CuS, respectively.

On the other hand, heterogeneous CuS@CuGaS2 nanoparticles were obtained by the impregnation method [46] using prepared CuS and CuGaS2. The synthesis procedure was as follows. The amounts of CuS and CuGaS2 added were different. CuS was added to ethanol, and the mixture was stirred for 2 h, then CuGaS2 was added and stirred sufficiently. The homogeneously stirred solution was separated into powdery samples by centrifugation and dried at 80 ◦C for 24 h. Thereafter, the resultant was again annealed at 200 ◦C for 2 h in order to remove impurities and increase the bonding strength, finally obtaining a heterogeneous CuS@CuGaS2 catalyst.

#### *3.2. Characterization of CuS, CuGaS2 and CuS@CuGaS2 Nanoparticles*

X-ray diffraction (XRD, MPD, PANalytical, Almelo, The Netherlands) was used to analyze the crystal structure of the prepared CuS, CuGaS2 and heterojuntion CuS@CuGaS2 nanoparticles. XRD was measured at a 2θ angle of 20–100◦ using nickel-filtered CuKα (λ = 1.5056 Å) radiation (40 kV, 30 mA). The shape and size of the particles were confirmed using high-resolution transmission electron microscopy (TEM, H-7600, Hitachi, Tokyo, Japan) and scanning electron microscopy (SEM, S-4100, Hitachi, Tokyo, Japan). In addition, energy-dispersive X-ray spectroscope (EDS, EX-250, Horiba, Kyoto, Japan) analysis was used to identify the atomic composition of CuS, CuGaS2, and heterogeneous CuS@CuGaS2 nanoparticles.

The diffuse reflection spectra of the particles were obtained using a UV-Vis spectrophotometer (Neosys-2000, SCINCO, Daejeon, Korea). The recombination tendency between the photogenerated electron-hole pair (e−/h+) of the catalyst was determined using a photoluminescence spectroscopy (PL, FS-2, SCINCO, Daejeon, Korea) equipped with a 150 W continuous Xenon lamp light source.

In addition, photocurrent and intensity modulated photovoltage spectroscopy (IMVS) measurements were taken with a two-electrode system to confirm the behavior of the photogenerated charge carrier. The catalyst was coated on fluorine doped tin oxide (FTO) glass to form a cell, and a platinum wire was used as a counter electrode. The catalyst coated on the FTO glass with a certain unit area was used as the working electrode and photocurrent was measured by irradiating light at intervals

of 30 s. The IMVS measurement was also performed using a visible light source in a two-electrode system, and the recombination lifetime of electrons was confirmed through the measurement.

#### *3.3. Hydrogen Production by Photo Splitting of Methanol Aqueous Solution Using CuS, CuGaS2 and Heterojunction CuS@CuGaS2 Catalyst*

The photocatalytic decomposition of methanol aqueous solution was carried out using a liquid photoreactor prepared in our laboratory, which was reported in previous work [47]. First, the photocatalytic decomposition of methanol aqueous solution using a UV light source was performed using a pyrex reactor. 1.0 L of a mixed solution of 500 mL of methanol and 500 mL of distilled water was put into the reactor, and 0.5 g of the synthesized CuS, CuGaS2 and CuS@CuGaS2 powder was added. Light was irradiated using a UV-lamp (3 × 6 cm−<sup>2</sup> = 18 W cm−2, length 30 cm, diameter 2.0 cm, Shinan, Pochon, Korea) at a wavelength of 365 nm and the reaction was performed for a total of 10 h. The photocatalytic decomposition of methanol aqueous solution using a visible light source was performed in a quartz reactor using a 150 W Xe lamp. The resulting gas was analyzed by gas chromatography (GC, DS7200, Donam Company, Gwangju, Korea) equipped with a thermal conductivity detector (TCD). The following GC conditions were used: TCD detector, Carboxen-1000 column (Bruker, Billerica, MA, USA), and the injection, oven and detector temperatures of 423, 393 and 473 K, respectively.

#### **4. Conclusions**

We have synthesized CuS, CuGaS2, and heterojunction CuS@CuGaS2 catalysts for hydrogen production through methanol aqueous photo splitting. The interface between the heterojunction CuS@CuGaS2 catalyst and the structural defect of CuGaS2 formed by the addition of Ga3+ to CuS acted as a trap site. This trap site accelerates the electron-transfer, indicating a high photocurrent density value in the photocurrent results and excellent charge separation efficiency. In addition, compared to pure CuS and CuGaS2, as shown by photoluminescence and IMVS measurements, recombination between the excited electron-hole pairs in the heterojunction catalyst was suppressed, resulting in higher electron lifetime. As a result, the heterojunction CuS@CuGaS2 catalyst produced a significant amount of hydrogen gas, up to 3250 and 200 μmol, through photo splitting of aqueous methanol solution under UV and visible light irradiation, showing a significant increase in photocatalytic activity.

**Author Contributions:** Conceptualization, M.K.; Data curation, N.S. and J.Y.D.; Formal analysis, N.S. and J.Y.D.; Investigation, J.N.H. and Y.K.; Methodology, J.N.H. and Y.-S.Y.; Supervision, M.K.; Writing—original draft, J.Y.D.; Writing—review & editing, M.K.

**Funding:** This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIT) (No. 2018R1A2B6004746).

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

#### **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/).

## **In-Situ Synthesis of Nb2O5/g-C3N4 Heterostructures as Highly Efficient Photocatalysts for Molecular H2 Evolution under Solar Illumination**

**Faryal Idrees 1,2,3,\*, Ralf Dillert 1,2, Detlef Bahnemann 1,2,\*, Faheem K. Butt <sup>4</sup> and Muhammad Tahir <sup>3</sup>**


Received: 7 January 2019; Accepted: 7 February 2019; Published: 11 February 2019

**Abstract:** This work focuses on the synthesis of heterostructures with compatible band positions and a favourable surface area for the efficient photocatalytic production of molecular hydrogen (H2). In particular, 3-dimensional Nb2O5/g-C3N4 heterostructures with suitable band positions and high surface area have been synthesized employing a hydrothermal method. The combination of a Nb2O5 with a low charge carrier recombination rate and a g-C3N4 exhibiting high visible light absorption resulted in remarkable photocatalytic activity under simulated solar irradiation in the presence of various hole scavengers (triethanolamine (TEOA) and methanol). The following aspects of the novel material have been studied systematically: the influence of different molar ratios of Nb2O5 to g-C3N4 on the heterostructure properties, the role of the employed hole scavengers, and the impact of the co-catalyst and the charge carrier densities affecting the band alignment. The separation/transfer efficiency of the photogenerated electron-hole pairs is found to increase significantly as compared to that of pure Nb2O5 and g-C3N4, respectively, with the highest molecular H2 production of 110 mmol/g·h being obtained for 10 wt % of g-C3N4 over Nb2O5 as compared with that of g-C3N4 (33.46 mmol/g·h) and Nb2O5 (41.20 mmol/g·h). This enhanced photocatalytic activity is attributed to a sufficient interfacial interaction thus favouring the fast photogeneration of electron-hole pairs at the Nb2O5/g-C3N4 interface through a direct Z-scheme.

**Keywords:** Niobium(V) oxide; graphitic carbon nitride; hydrothermal synthesis; H2 evolution; photocatalysis; heterostructures; Z-Scheme

#### **1. Introduction**

Renewable energy sources are currently needed by our society to address the foreseable future energy crisis and growing environmental issues. The production of molecular H2 through photoelectrochemical or photocatalytic water splitting is a viable replacement of fossil fuels [1]. In the past, due to certain limitations of the most frequently employed photocatalyst, TiO2, many other photocatalysts have been developed and explored for photocatalytic molecular H2 evolution. Their significant limitations are the high recombination rate of photogenerated electron-hole pairs and unfavorable band edges, hence their low photocatalytic activity. For favorable band edges, CuO and Cu2O are thought of as good alternatives but their stability and effective light absorption present additional issues. Somehow, these issues have been solved by adopting the atomic layer deposition (ALD) technique and doping, etc. [2–4]. The non-toxicity, facile synthesis, high visible light absorption and good physicochemical stability of g-C3N4 have made it an amiable photocatalyst [5–9]. However, the fast recombination rate of photogenerated charge carriers has reduced its charismatic effect. Some researchers proposed to solve this problem by forming heterojunctions with TiO2 [10], WS2 [11], BiOCl [12] and WO3 [13,14]. Still, there is considerable scope to develop other photocatalysts, with the particular focus being on facile synthesis, low toxicity, easy accessibility, and, most importantly, high photocatalytic activity [15].

Niobium pentoxide (Nb2O5) is an n-type and wide bandgap (3.4 eV) semiconductor, which has been extensively investigated in recent years for electrochemical, photocatalytic, and energy storage applications [16,17]. Its light absorption can be effectively shifted into the visible region by synthesizing composites with small bandgap materials. Noticeably, heterojunctions prepared with small amounts of Nb2O5 have shown a significant improvement in the photocatalytic activity of TiO2 [8], ZnO [18], and g-C3N4 [7,9,19]. To date, there are only a few reports regarding the synthesis of g-C3N4/Nb2O5 heterojunctions and their photocatalytic properties. Y. Z. Hong et al. [19] and Q. Z. Huang et al. [20] have prepared g-C3N4/Nb2O5 heterojunctions facing, however, considerable limitations concerning the control of the size and shape of Nb2O5, and resulting in a rather small specific surface area. Moreover, to the best of our knowledge, Nb2O5/g-C3N4 heterostructures have so far not been explored for molecular H2 evolution.

To increase photocatalytic activities, semiconductor-semiconductor heterostructures exhibiting lower recombination rates of photogenerated electron-hole pairs have been studied. Depending on the energetic situation, these heterostructures have been classified as type-II heterostructures and Z-scheme heterostructures, respectively (Figure 1; for a detailed discussion see the Supporting Information) [20–22]. The direct Z-scheme system seems promising for overcoming the limitations associated with enhanced photocatalytic activity, due to the strong oxidation and reduction potential developed at different active sites [21].

As depicted in Figure 1, a conventional type-II heterostructure can easily be converted into a Z-scheme structure by controlling the Fermi level or the band potentials. Moreover, the Fermi level and band potentials can also be modulated to attain the Z-scheme by the addition of suitable hole scavengers such as triethanolamine (TEOA), which has been reported to exhibit a larger H2 evolution rate than methanol [19,22].

**Figure 1.** Schematic illustration of Type-II and Z-Scheme systems and their interfacial band bending under bandgap irradiation.

In the present work, Nb2O5/g-C3N4 Z-scheme heterojunctions were prepared considering their suitable band edges for photocatalytic H2 production, i.e., Nb2O5 (ECB = −0.69 V, EVB = 2.32 V) and g-C3N4 (ECB = −1.68 V, EVB = 0.88 V) vs. NHE (pH = 7). In comparison to previous reports, we obtained a controlled shape of the heterojunction with a specific surface area as high as 227 m<sup>2</sup> g−1. The prepared heterojunctions were tested for H2 evolution without and with deposited platinum (Pt) acting as a cocatalyst, thus faciliating the interfacial electron transfer. The role of hole scavengers in the overall mechanism of the hydrogen evolution reaction was also investigated. A significant increase in the H2 evolution rate was observed in the presence of TEOA compared with methanol. These

findings have suggested a possible change in the photocatalytic mechanism to increase the evolution rate to such an extent. An excellent H2 production rate of 110 mmol/h·g was found by employing platinized Nb2O5/g-C3N4 as the photocatalyst and TEOA as the hole scavenger. Noticeably, the high molecular H2 rate could also be associated with the tuning of shape, size and structures. The compact interfacial development between Nb2O5/g-C3N4 also suggested that a Z-scheme system was formed, thus facilitating fast charge carrier separation and excellent photocatalytic performance as compared with simple physical mixing. The effect of different ratios of g-C3N4 to Nb2O5 has also been studied in detail.

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

#### *2.1. Synthesis Procedure*

Nb2O5 (NBO) and Nb2O5/g-C3N4 (NBCN) heterostructures were synthesized via a hydrothermal synthesis considering its possible principal advantages such as: (a) attaining porous structures with high surface areas; (b) reagents mixing at the atomic level, and (c) high reaction rates at a low reaction temperature due to the atomic mixing level [23–25]. Thus, a suitable strategy to tune the shape, size, and structure of Nb2O5 (NBO) and Nb2O5/g-C3N4 (NBCN) heterostructures with a high specific surface area and a sufficient contact interface was employed here.

NBO was synthesized via the oxidant-peroxo method (OPM). A niobium salt (NbCl5) was dissolved in diluted nitric acid (HNO3) to avoid salt residuals. In a second step H2O2 was added to the prepared solution to remove chloride ions by an oxidation–reduction process. The resulting yellow solution (pH = 0.5) indicated the presence of the water-soluble niobium peroxo-complex [Nb(O2)4] 3− (named as NPC). A possible reaction is provided in Figure 2. The decomposition of H2O2 into molecular oxygen possibly accelerated the condensation reaction. Amorphous hierarchical spheres of Nb2O5 and Nb2O5·nH2O were obtained due to an excess amount of H2O2. Finally, annealing in the 200–500 ◦C range resulted in the formation of Nb2O5 with adequate composition and controlled morphology.

$$\text{2NbCl}\_5 + \text{ 5n H}\_2\text{O} \xrightarrow{\text{H}\_2\text{O}} \text{Nb}\_2\text{O}\_5 \cdot \text{nH}\_2\text{O} + \text{10HCl}$$

$$\text{3H}\_2\text{O}\_2 + 2\text{HCl} \xrightarrow{\quad} \xrightarrow{\quad} + \text{H}\_2\text{O} + \text{O}\_2 + \text{Cl}\_2$$

**Figure 2.** The proposed reaction mechanism for Nb2O5 formation.

For the preparation of the heterostructures, a g-C3N4 (GCN) suspension was prepared in de-ionised water by 1 h of continuous stirring. The prepared suspension was added to the solution of NPC (as described above) which changed the pH from 0.5 to 0.7. The pH change resulted in a positive surface charge on GCN. The developed electrostatic attraction between NPC and GCN favors the effective in-situ formation of heterojunctions with a controlled morphology [26]. This procedure was employed to prepare other heterostructures with a different wt % of GCN (as described in the experimental sections). Heterostructures have also been prepared by physical mixing of NBO and g-C3N4 (the method is described in the supporting information). A schematic presentation of the synthetic procedure is given in Figure 3.

**Figure 3.** Schematic illustration of the hydrothermal routes followed for Nb2O5 (NBO) and Nb2O5/g-C3N4 (NBCN) synthesis.

#### *2.2. Structural and Morphological Characterization*

The phases and the crystal structures of as-prepared NBO, GCN, and NBCN-X were analyzed by XRD (Figure 4). The XRD data obtained for a NBO sample before annealing corresponded to a mixture of niobic acid and amorphous NBO. By annealing at 200 ◦C the mixture converted to the amorphous pseudohexagonal phase of Nb2O5 (TT-NBO, JCPDS#30-0873). No significant change in the phase was observed for samples annealed at temperatures between 200 ◦C and 400 ◦C. However, annealing at 500 ◦C results in the transformation of the amorphous pseudohexagonal phase of NBO in to the pseudohexagonal phase (JCPDS#30-0873) of NBO. The corresponding XRD results are provided in the supporting information in Figure S1. In the XRD of GCN, two peaks are observed at 13.4◦ and 27.0◦, which were associated with the (100) and (002) diffraction planes, respectively. The (100) distinct peak was due to the characteristic inter-planar staking of aromatic systems, while the (002) peak corresponded to the inter-layer structural packing [26]. The XRDs of NBCN-X (X = 1–4) exhibited combinatory characteristic peaks of GCN and NBO. The characteristic peak of GCN (002) become stronger with an increasing amount of GCN, indicating a significant coupling between the GCN and NBO.

**Figure 4.** XRD patterns of NBO, g-C3N4 (GCN), NBCN-1, NBCN-2, NBCN-3, and NBCN-4.

The FTIR patterns further confirmed the development of compact heterostructure interfaces, as shown in Figure 5a. The amorphous pseudohexagonal phase of NBO possessed slightly distorted NbO6, NbO7, and NbO8 groups and some highly distorted NbO6 sites (a scheme is provided in Figure 5b). The peak at 665 cm−<sup>1</sup> has been assigned to the symmetric stretching mode of Nb–O polyhedra (NbO6, NbO7, and NbO8) [27]. The broad shoulder between 850 and 970 cm−<sup>1</sup> has been assigned to the stretching of Nb=O groups. The removal of coordinated water further distorted the highly distorted NbO6 octahedra, which led to the formation of Nb=O bonds [28]. The peaks at 414 cm−<sup>1</sup> and 459 cm−<sup>1</sup> have been attributed to the symmetric stretching *vs*[Nb(O)2] and the asymmetric stretching *va*[Nb(O)2], respectively. Their occurrence indicated the presence of a small amount of coordinated peroxide on the Nb(V) ions [29]. With increasing calcination temperatures, the peak at 459 cm−<sup>1</sup> first decreased from 200–400 ◦C and then disappeared entirely at 500 ◦C. The peaks at 3753 cm−<sup>1</sup> and 1559 cm−<sup>1</sup> have been attributed to the vibration of OH groups *v*(O–H) of adsorbed water molecules [30], which disappeared after increasing the calcination temperature (as shown in Figure S2).

The FTIR spectrum of GCN shows numerous bands in the 1100–1630 cm−<sup>1</sup> region corresponding to the typical stretching modes of GCN heterocycles. The FTIR spectra of the heterostructures NBCN-X(X = 1–4) exhibited characteristic peaks of NBO and GCN. However, for NBCN-1 and NBCN-2 the GCN band's peaks were found to be weaker than NBCN-3 and NBCN-4, due to the lower amount of GCN incorporated in the heterostructure. The results were consistent with the XRD results. Moreover, the sharp band at 810 cm−<sup>1</sup> was associated with the tri-s-triazine forming the GCN structure. The broad band around 3163 cm−<sup>1</sup> could be associated with the N–H or O–H bonds of the residual amino groups or absorbed H2O molecules [31,32]. Detailed FTIR spectra are also provided in Table S1.

**Figure 5.** (**a**) FTIR spectra of NBO, GCN and NBCN-2 and (**b**) Highly and slightly distorted NbO6 octahedra coexist in the structure.

SEM and TEM images of as-synthesized photocatalysts are shown in Figure 6. The bulk of the GCN and the hierarchical nanospheres of NBO can be seen in Figure 6a,b, respectively, while the NBCN heterostructures are given in Figure 6c–f. A significant change in the morphology can be observed for the NBO and NBCN heterostructures. BET analysis helped to further analyze the change in the surface area of bare and heterostructure photocatalysts. The high surface area of NBO-BA gradually decreased with increasing the annealing temperature from 200 to 500 ◦C (see Table S2). A high annealing temperature favored pore coalescence due to the crystallisation of walls separating mesopores in their structures. The BET surface areas of NBCN-X (X = 1–4) annealed at 200 ◦C and are provided in Table S3. The specific surface area decreased gradually with the increase in the amount of GCN, due to the low surface area of GCN.

**Figure 6.** SEM images of (**a**) GCN, (**b**) NBO, (**c**) NBCN-1, (**d**) NBCN-2, (**e**) NBCN-3, and (**f**) NBCN-4.

A TEM analysis for NBO and NBCN-2 was conducted, as shown in Figure 7. For NBO (Figure 7a), the lattice fringes had 0.39 nm d-spacing's corresponding to the (001) lattice plane of Nb2O5. A small number of lattice fringes were observed due to the amorphous nature of NBO. The SAED pattern

(Figure 7a inset) indicated that the NBO was in the transforming state from the amorphous to the crystalline phase (consistent with XRD results). More TEM images at different resolutions are provided in Figure S3. For NBCN-2 (Figure 7b), lattice fringes for NBO and GCN were observed. However, due to a smaller amount of GCN, only a few lattice fringes for GCN were observed (0.31 nm d-spacing of (002) plane). The (001) plane of NBO was found to be compatible with the (002) plane of GCN, thus favoring the in-situ growth of NBO on the surface of GCN.

**Figure 7.** (**a**) HRTEM (high electron transmission electron microscopy) image and SAED (selected area electron diffraction) pattern (inset) of NBO, (**b**) HRTEM image and TEM (inset) image of NBCN-2.

#### *2.3. Photocatalytic Evolution of Molecular H2*

A systematic study was conducted to signify the role of all possible parameters that could change the rate of generated molecular hydrogen (H2). Initially, the photocatalytic formations of molecular H2 for P25, GCN, NBO and NBCN-X (X = 1–4) were studied with and without Pt deposition in the presence of TEOA (a hole scavenger). The evolved H2 is shown in Figure 8a,b, respectively. The total liberated amount in 7 h is also shown in Figure 8c. All heterostructures prepared generated more molecular H2 than P25, NBO or GCN, both with and without Pt deposition. This increase in the H2 formation rate with Pt deposition can be taken as evidence for the successful interfacial charge separation. Among all heterostructures, NBCN-2 showed the highest H2 generation rate of 110 mmol/g·h, which could be associated with both its high surface area and suitable band positions. The evolution rates for other prepared composites: NBCN-1 = 103.27 mmol/g·h, NBCN-3 = 77.88 mmol/g·h, and NBCN-4 = 56.04 mmol/g·h, are also higher than those for P25 = 45.46 mmol/g·h, GCN=33.46 mmol/g·h or NBO= 41.20 mmol/g·h. Thus, the enhanced photocatalytic activity may have been caused by GCN loading acting as a visible light active material and as an efficient electron-hole separator for NBCN prepared heterostructures. However, this effect was only obvious provided that the optimised ratio of GCN and NBO was used.

For P25, the formation rate of molecular H2 was increased in the first 4 h then a gradual decrease was observed. In comparison to bare materials the H2 formation rate of P25 was larger than GCN and NBO, i.e., 45.46 mmol/g·h, but the rate was not stable after Pt deposition, as shown in Figure 8a,b, respectively. The H2 formation rate was also smaller than all heterostructures prepared and the ~66% increase in molecular H2 evolution rate was observed after Pt deposition, which was less than all heterostructures prepared. Moreover, the prepared composites showed more of a stable increase in the molecular H2 formation rate than P25. The effect of methanol over platinized P25 could be studied through the Kandiel papers [33,34].

**Figure 8.** The photocatalytic formation of molecular H2 of P25, NBO, GCN, NBCN-1, NBCN-2, NBCN-3 and NBCN-4: (**a**) without Pt, (**b**) with Pt deposition, and (**c**) total evolution rate; (**d**) photocatalytic formation of molecular H2 of NBCN-2, in the presence of methanol and TEOA.

Secondly, to understand the role of the scavenger, the NBCN-2 activity employing methanol as hole scavenger was also investigated with and without Pt deposition. The change in H2 evolution with time and in comparison with TEOA is shown in Figure 8d. A total of 5.53 mmol/g·h and 9.3 mmol/g·h of H2 were liberated without and with Pt deposition, respectively. This increase was just 25%, and was 60% with TEOA. Moreover, in the presence of methanol, the H2 amount was 67% (without Pt) and 85% (with Pt) less than TEOA. In light of the above results, TEOA not only gave higher activity, but also favored interfacial charge separation after Pt deposition. The findings indicated a possible change in the photocatalytic scheme. NBCN-2 results have also been compared with the physically mixed compound (with the same composition) in the presence of methanol. The NBCN-2 photocatalyst and physically mixed photocatalyst with the same ratio of NBO and GCN were used. A drastic change in the evolution rate was observed, as shown in Figure 8d, and more clearly in Figure S4. The NBCN-2 molecular H2 generation rate showed a 35% increase over the physically mixed photocatalyst. Thus, in-situ heterostructures have higher activity than physically mixed heterostructures, and this can be attributed to good intimate contact. The total generated amount of H2 for all studied photocatalysts has been provided in Table S4.

The present results have also been compared with previous reports on GCN, NBO and NBCN in the presence of TEOA/Methanol. The comparison is provided in Table 1. A detailed analysis of the table has proven the significance of our work in the following ways: (a) higher molecular H2 rate in less time, (b) less concentration of photocatalysts and (c) the synthesized heterostructures have a higher molecular H2 rate than many of them, even without the Pt deposition.



To estimate the electronic structure and possible interfacial band bending of materials, we conducted Mott–Schottky measurements (C−<sup>2</sup> versus applied potential). The flat band/conduction band potentials (EFB/ECB), intercepted for NBO and GCN, are −0.61 V (−0.69) and −1.6 (−1.68) V vs. NHE (pH = 7), as shown in Figure 8a. The values have been converted to pH = 7 by using equations Equation (S1) and Equation (S2), respectively. By using the bandgap energy calculated through UV-vis-Absorption, the estimated valence band potentials (EVB) are 2.32 V and 0.88 V for NBO and GCN at pH = 7, respectively (provided in Figure S5). Judging by the bandgap positions, typically a type-II mechanism should be followed, e.g., the photogenerated electrons in GCN could easily move to the NBO conduction band following the reduction there, and the photogenerated holes could easily move from the NBO valence band to the GCN valence band following the oxidation process. However, based on the recent report by Huang, Z.-F et al. [22], when TEOA is adsorbed on the surface the electron transfer is inverted as shown in Figure 1, and hence the Z-Scheme mechanism is followed. Following their results, the NBO electrons recombined with the holes of the GCN and consequently, reduction and oxidation reactions at the GCN and NBO occurred, respectively. The high production rate also favoured the direct Z-Scheme.

We estimated the charge carrier densities (ND) to understand the interfacial band bending by using the Mott–Schottky slope (Figure 9). For a theoretical overview, the inclination angle of the M–S plot fo GCN was smaller than NBO. Since the relative dielectric constant (ε) was directly proportional to the inclination angle (θ) and inversely proportional to donor density (ND) of the material. Hence, GCN should have a smaller dielectric constant and a higher donor density than NBO. In the literature, ε ~7–8 for GCN [32] and ε ~38 for pseudohexagonal NBO [45] has been reported. GCN ε is six times larger than NBO, so the ND for GCN should be around six times larger than NBO. Mathematically, ND has been calculated by using Equation (1) [46]:

$$\mathrm{N\_D} = \frac{2}{\varepsilon\_0 \varepsilon\_0 \varepsilon} \left[ \frac{d \left( 1/C^2 \right)}{dV} \right]^{-1} \tag{1}$$

where, *<sup>e</sup>*0, <sup>ε</sup>0, <sup>ε</sup> and *<sup>d</sup>*(1/*C*<sup>2</sup>) *dV* are electron charge, vacuum permittivity, material dielectric constant and Mott–Schottky slope, respectively. The approximated values for GCN ND ≈ 6.1 × 1021 cm−<sup>3</sup> and for NBO ND ≈ 7.9 × 1020 cm−3, have been calculated, i.e., NBO ND is seven times smaller than GCN, as expected.

**Figure 9.** (**a**) Mott–Schottky plot for NBO and GCN at 1000 Hz vs. Ag/AgCl, (**b**) EIS analysis for NBO, GCN and NBCN-2.

For the n-type, due to high donor density, the Fermi level lay close to the bottom of the conduction band. Eventually, upward band bending occurred due to the high electron density of the conduction band and low acceptor density of the valence band. On the other hand, for low donor density of n-type, the Fermi level lay close to the middle of the bandgap. Moreover, a downward band bending of the conduction band and the valence band occurred. Thus, following the concept, after irradiation, the equilibrium Fermi level (EF) of NBO and GCN was lying close to the redox potential (−0.41 V) vs. NHE (pH = 7). However, to justify the Z-Scheme and high production rate, fast recombination at the interface and fast charge transfer at the proposed conduction and valence bands should be followed.

Electrochemical Impedance Spectroscopy (EIS) analysis helped to further estimate the recombination and charge transfer process at the working electrode and electrolyte interfaces. Nyquist plots for NBO, GCN and NBCN-2 in the frequency range of 1000 kHz–0.01 Hz under UV-Vis light irradiation have been recorded (Figure 8b). The Randles circuit model [47] has been employed to describe the behavior of the electrode, as shown in the inset of Figure 8b. The arc radius has been associated with charge transfer resistance (RCT) across the interface of working electrode/electrolyte, i.e., the small radius means efficient interfacial charge transfer/slow recombination rate of photogenerated electron/hole pairs. All three electrodes showed explicit arcs and the fitted values of RCT are: NBO = 342 kΩ, GCN = 348 kΩ and NBCN-2 = 159 kΩ. The lowest RCT for NBCN-2 has been recorded, which indicated that the composites have better efficiency of charge transfer than NBO and GCN. The NBO showed slightly better photoactivity than GCN. The large RCT for GCN indicated its poor charge transfer characteristics, which may be associated with a low charge transfer rate leading to the fast recombination of photogenerated electron/hole pairs and poor photoactivity.

Thus, there was two times reduced resistance for the NBCN-2 heterostructures than NBO and GCN, which favored the reduced charge-transfer barrier with an increase in charge carrier density, and hence the Z-scheme system is followed. The proposed energy diagram and mechanism is shown in Figure 10. The proposed mechanism has been associated with the compatible band positions of NBO and GCN, which favored the direct Z-scheme transfer of charges and thus a higher molecular H2 production rate. That is, the 1 wt % platinizied NBCN heterostructure exhibits a high activity for H2 generation of 110 mmol/g·h because the direct transfer and recombination of photogenerated electrons in NBO and holes in GCN could greatly extend the lifetime time of carriers. The negative shift of the flat band of NBCN heterojunction (provided in Figure S6) further confirms the proposed Z-scheme mechanism.

**Figure 10.** Schematic illustration of possible photocatalytic mechanism.

#### **3. Methods**

Niobium pentachloride NbCl5 (99.9%, Sigma Aldrich, Munich, Germany), melamine C3H6N6 (99.0%, Sigma Aldrich, Munich, Germany), hydrogen peroxide solution H2O2 (30 wt %), nitric acid HNO3 (60 wt %), Evonil Aeroxide TiO2 P25, chloroplatinic acid hexahydrate H2PtCl6.6H2O (≥37.05%, Sigma Aldrich, Munich, Germany) and triethanolamine (HOCH2CH2)3N (99.5%, Sigma Aldrich, Munich, Germany) were used without further purification.

#### *3.1. Material Synthesis*

#### 3.1.1. Preparation of Nb2O5 (NBO)

0.5 g of hygroscopic yellow powder of NbCl5 was added into a mixture of 20 mL de-ionised water and 0.5 mL HNO3. The solid dissolved immediately yielding a transparent solution. 10 mL of aqueous H2O2 solution was added drop-wise into a clear solution under vigorous stirring. The solution became light yellow, confirming the formation of the niobium-peroxo complex (NPC). This solution was diluted to 60 mL by further addition of de-ionised water. The prepared light-yellow solution was subesquently transferred into a 250 mL autoclave and maintained at 110 ◦C for 24 h in an oven. The synthesised white precipitate was centrifuged and washed several times with de-ionised water. Subsequently, the prepared powder was dried in an oven at 90 ◦C for 12 h (named as NBO-BA). The resultant NBO-BA samples were further annealed in a muffle furnace at different temperatures ranging from 200 ◦C to 500 ◦C for 2 h with a heating rate of 10 ◦C min−1. In the present work, the photocatalyst annealed at 200 ◦C has been named as NBO, throughout.

#### 3.1.2. Preparation of g-C3N4 (GCN)

10 g of melamine were placed in a crucible with a cover lid on top and then annealed at 550 ◦C for 3 h in a muffle furnace employing a heating rate of 10 ◦C min−1. After natural cooling, a yellow powder of bulk g-C3N4 was obtained (named as GCN).

#### 3.1.3. Preparation of Nb2O5/g-C3N4 (NBCN)

The Nb2O5/g-C3N4 heterostructures named as NBCN-X (where X = 1, 2, 3, 4) were prepared by using different amounts of GCN. Typically, 2.5 mg of NBO-BA were obtained by hydrothermal synthesis. Therefore, for the heterostructure preparations, the selected amount of GCN was added to the prepared solution of NBO-BA, i.e., X: 1 = 5 wt %, 2 = 10 wt %, 3 = 30 wt % and 4 = 60 wt %. The synthesis process was as follows: initially, GCN suspension was prepared in 20 mL of de-ionised water following the ultra-sonication for 30 min. Then, the prepared GCN suspension was added to a light-yellow solution of NPC under continuous stirring. The amount was adjusted to 60 mL by further addition of de-ionised water and stirring for 30 min. Choosing pH = 0.7 should assure the positive and negative surface charges have been developed over GCN and NPC, respectively, for favorable interface development. The prepared concentrated solution was subsequently transferred to a 250 mL autoclave and maintained at the 110 ◦C for 24 h in the oven. After washing and drying at 90 ◦C for 12 h, a light-yellow powder is obtained, which was further annealed at 200 ◦C for 2 h in a muffle furnace with a 10 ◦C min−<sup>1</sup> ramp rate.

#### *3.2. Photodeposition of Platinum (Pt)*

The photo-deposition technique was applied to deposit 1 wt % platinum (Pt) on the samples. Hexachloroplatinic acid (H2PtCl6·6H2O) was used as the Pt precursor and methanol as a reducing agent. The calculated amounts of H2PtCl6·6H2O and the photocatalyst were added to 10 vol/vol. % of aqueous methanol solution. The resulting suspension was transferred into a closed reactor and placed under UV-light for 12 h with continuous stirring. The suspension was washed several times

with de-ionised water to remove non-deposited Pt and subsequently with ethanol. The obtained Pt loaded photocatalysts were dried in an oven for 12 h at 90 ◦C.

#### *3.3. Material Characterization*

A Bruker (D8 Advance) instrument using Cu Kα (α = 0.15406 nm) radiation was used to record the X-ray diffraction (XRD) data of the as-synthesized samples. Scanning electron microscopy (SEM, JEOL JSM-6700F, JEOL, Tokyo, Japan) with a LEI detector (Lower Secondary Electron Image) was employed to analyze the morphologies. Transmission electron microscopy (TEM), using an FEI Tecnai G2 F20 microscope operating at 200 kV was used to characterise the samples further. Micrographs were taken in bright field (BF) and in a selected area electron diffraction mode. A Varian Cary 100 Bio was used to measure the UV–vis absorption spectra. A Bruker Vertex 80v spectrophotometer (Bruker, Billerica, MA, USA) was used to measure the FTIR spectra from 4000 to 400 cm−<sup>1</sup> in vacuum.

#### *3.4. Photocatalytic Molecular Hydrogen (H2) Formation*

The prepared photocatalysts without and with Pt deposits were used to conduct photocatalytic molecular hydrogen (H2) evolution reactions. In a typical experimental run, 0.01 g of the photocatalyst were added into 45 mL of de-ionised water (maintaining pH 5.6). Methanol or TEOA were added resulting in a suspension which contains 10 vol % of a hole scavenger. The suspension inside the photoreactor was thoroughly degassed for 30 min with Ar to remove air and then stirred for 30 min in the dark to establish the adsorption equilibrium. Afterwards, the photoreactor was placed under a Xenon light source (1000 W, 1.5G) suitable to simulate solar light for 7 h. The temperature was maintained constant by using a homemade cooling system. The evolved amounts of H2 were measured every 1 h by using a gas chromatograph (Shimadzu 8A (Shimadzu, Kyoto, Japan) equipped with a TCD detector and a 5 Å molecular sieve packed column; Ar was used as the carrier gas).

#### *3.5. Photoelectrochemical Measurements*

The photoelectrochemical (PEC) measurements were conducted by using an electrochemical workstation (CHI-660B, CH Instruments, Inc., Austin, TX, USA) accompanying a ZAHNER PECC-2 reactor and 450W xenon lamp as a light source. Moreover, a standard three-compartment cell (consisting of a photo-/working electrode (WE), a Pt wire counter electrode (CE) and an Ag/AgCl reference electrode (RE)) with 0.2 M Na2SO4 electrolyte solution (pH = 5.6) were used. The working electrode was prepared using a screen-printing method and then annealed at 400 ◦C for 2 h to remove organic chemicals. Mott–Schottky measurements were performed at a frequency range of 10–1000 Hz with 10 mV amplitude. Electrochemical impedance spectra were obtained under irradiation at open circuit voltage over a frequency range from 1000 Hz to 0.01 Hz, with an AC voltage at 250 mV vs. Ag/AgCl reference electrode.

#### **4. Conclusions**

Z-scheme Nb2O5/g-C3N4 heterostructures with excellent molecular H2 production activity were prepared via in-situ hydrothermal syntheses. The prepared heterostructures exhibited excellent photocatalytic activity compared to individual g-C3N4 and Nb2O5 under simulated solar light illumination. The highest reported H2 evolution rate was 110 mmol/g·h (7.7 mmol). We found that by increasing the amount of g-C3N4, the molecular H2 production rate decreased, indicating more intimidating interface development does not favor photocatalytic reactions. However, the molecular H2 evolution rate for all prepared heterostructures was higher than many other semiconductors reported in the literature. Moreover, we justified our results by a reduced recombination rate, high charge carrier density and complemented band positions. For future work, we suggest that the various heterojunction materials possessing diverse structural morphology exhibiting a higher photocatalytic activity be prepared by the simple methodology described here. These studies will pave the way for a new dimension in photocatalytic studies of Nb2O5 and g-C3N4 nanocomposites for enhanced molecular H2 production.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4344/9/2/169/ s1, Figure S1: XRD patterns BA and AA at different calcination temperatures, Figure S2: FTIR spectra of all photocatalysts, Figure S3: TEM images of NBO, Figure S4: Molecular H2 generation of NBCN-2, Chemical and Physical Mixing in the presence of methanol, Figure S5: Bandgap vs. photon energy by UV-vis-Diffuse Absorption Spectra of a) NBO, GCN, NBCN-1, NBCN-2, NBCN-3, and NBCN-4 and b) with the change in the annealing temperature, Figure S6: Mott-Schottky plot of NBO and NBCN-2, Table S1: Associated Bands in FTIR Spectra, Table S2: Specific surface area (m2/g) of samples with different calcination temperatures, Table S3: Specific surface area of NBO and NBCN-X (X = 1–4) without calcination, Table S4: Liberated Amount of H2 after 7h with and without Pt, Equation (S1) and Equation (S2).

**Author Contributions:** F.I. conceived, designed and performed the experiments; F.K.B. and M.T. analyzed the data; R.D. contributed reagents/materials/analysis tools; F.I., R.D. and D.B. wrote the paper. Substantial paper changes have been made by Ralf and Bahnemann.

**Funding:** The work was supported by the Alexendar Von Humboldt Foundation (Project No. 60421802) and PSF/NSFC/Eng-P-UoL(02).

**Conflicts of Interest:** We declare no conflict of interest.

#### **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*

## **Synthesis of Spherical TiO2 Particles with Disordered Rutile Surface for Photocatalytic Hydrogen Production**

#### **Na Yeon Kim, Hyeon Kyeong Lee, Jong Tae Moon and Ji Bong Joo \***

School of Chemical Engineering, Konkuk University, Gwangjin-gu, Seoul 05029, Korea; kny960403@konkuk.ac.kr (N.Y.K.); hyeonk@konkuk.ac.kr (H.K.L.); aidf91@gmail.com (J.T.M.) **\*** Correspondence: jbjoo@konkuk.ac.kr; Tel.: +82-245-03-545

Received: 28 April 2019; Accepted: 24 May 2019; Published: 28 May 2019

**Abstract:** One of the most important issues in photocatalysis research has been the development of TiO2-based photocatalysts that work efficiently under visible light conditions. Here, we report the monodispersed, spherical TiO2 particles with disordered rutile surface for use as visible-light photocatalysts. The spherical TiO2 particles with disordered surface were synthesized by sol-gel synthesis, followed by sequential calcination, and chemical reduction process using Li/Ethylenediamine (Li/EDA) solution. Variation of the calcination temperature allowed the crystalline properties of the calcined TiO2 samples, such as the ratio of anatase and rutile, to be finely controlled. The content ratios of anatase phase to rutile phase leads to different degrees of disorder of the rutile surface, which is closely related to the photocatalysis activity. Chemical reduction using the Li/EDA solution enables selective reduction of the rutile surface of the calcined TiO2, resulting in enhanced light absorption. As a result, we were able to synthesize spherical TiO2 photocatalysts having a disordered rutile surface in a mixed crystalline phase, which is beneficial during photocatalysis in terms of light absorption and charge separation. When used as photocatalysts for hydrogen production under solar light conditions, the chemically-reduced TiO2 particles with both the disordered rutile surface and mixed crystalline phase showed significantly enhanced catalytic activity.

**Keywords:** TiO2; spherical particle; disordered surface; photocatalysts; hydrogen production

#### **1. Introduction**

In modern society the energy crisis is becoming one of the biggest issues to directly impact our lives. Hydrogen as a green energy carrier has attracted much attention, due to its high energy capacity, environment-friendly characteristics, and sustainability [1]. As people are recognizing that high concentration of carbon dioxide is closely related to global warming and climate change, hydrogen can be considered as a one of the representative alternative-energy resources to either reduce or replace the use of depletable fossil fuels. There are several strategies to produce hydrogen, such as the reforming of either fossil fuel or renewable biomass, water electrolysis, ammonia decomposition, photo-electrochemical water splitting, and photocatalytic water splitting [2–8]. Among them, photocatalytic hydrogen production from water is considered as an ideal and economically-feasible method, since infinite solar energy can be used with any other type of energy resource [9].

Titanium dioxide (Titania, TiO2) is one of the most well-known semiconductor photocatalysts. TiO2 materials have a few advantages, which include low cost, considerable photocatalytic activity, low toxicity, high chemical stability, and abundance on Earth [10–12]. Since Honda and Fujishima first discovered hydrogen production by the photoelectrochemical splitting of water under UV light conditions [13], TiO2 has not only been intensively studied with fundamental researches, but also widely used for practical systems for solar energy conversion [14–17]. Although TiO2 has been intensively

studied over the past decades and has become known as a superb photocatalyst, the practical use of TiO2 has still been limited, due to its intrinsic optical property. Since TiO2 has a wide band-gap energy of 3.0–3.2 eV, its photocatalytic performance is limited to the ultra-violet (UV) region [12,17,18]. Even though the greater portion of solar light is in the visible and IR region, TiO2 can absorb mainly UV light, resulting in it generally showing low solar-to-chemical efficiency under solar light conditions. In addition, rapid recombination of photogenerated electron-hole pairs also leads to low quantum efficiency. Thus, this results in low overall photocatalytic activity [19,20].

In order to overcome the above drawbacks, various novel approaches have been taken to improve its optical, electronic, and chemical properties. To narrow the band-gap, either nonmetal or metal ions are doped into TiO2 crystal lattice, regulating either the level of conduction band or valence bands [21–27]. Surface sensitization using organic dyes can allow TiO2 to utilize the exited electron from dye molecules under visible light irradiation [28,29]. Recently, decoration of plasmonic metal nanoparticles on the TiO2 surface have also been suggested to improve hot electron transfer from metal nanoparticles to the conduction band of TiO2, resulting in unexpected photocatalytic performance under visible light conditions [30,31]. Since Mao and co-worker made the pioneering discovery of black TiO2 nanocrystal through surface-disordering using hydrogen [32], there have been many further reports about the interesting strategy of the synthesis of surface-disordered TiO2 by chemical reduction [33–35]. Unlike other TiO2-based photocatalysts by conventional disordering processes, they can achieve significantly disordered surface of TiO2 nanocrystal with well-maintained crystalline property, resulting in high photocatalytic activity on both organic dye decomposition and hydrogen production under solar light [32]. It is also well known that active metal, such as Al and Mg, can reduce TiO2, resulting in the formation of colored TiO2. Huang et al. developed a new approach to prepare colored TiO2 based on Al-reduction [36]. Sinhamahapatra and Yu synthesized black TiO2 by mixing the commercial TiO2 nanocrystals with Mg powder, followed by annealing in the hydrogen environment [37,38]. Park et al. also developed chemical reduction using a Lithium/Ethylenediamine mixture, and prepared blue-colored TiO2 that had the selectively disordered rutile surface. Lithium/Ethylenediamine (Li/EDA) solution, a metal chelate compound, is a very strong reducing agent. Ethylenediamine (EDA) is a superbase chemical that provides high pH conditions. The chelated metallic Li derived from the Li/EDA solution then selectively breaks the bonds between Ti and O of the rutile phase TiO2, resulting in various defects and a disordered surface. Defects in the disordered surface can narrow the band gap, forming interstates between the conduction band and valence band. Therefore, white TiO2 turns into a blueish color [39].

As shown in the previous literature, several chemical reduction approaches can allow the band-gap of pure TiO2 to be narrow, and the colored TiO2 to have high performance under solar light photocatalysis. Although intensive investigation of the synthesis of colored TiO2 was conducted using commercial TiO2 nanocrystals, such as P25 [39–41], there has been only limited study reported on the fabrication of colloidal TiO2 nanostructure. We recently found that uniform colloidal TiO2 particles with tunable crystalline properties, such as the ratio of anatase–rutile phase and crystallinity, can be produced by the sol-gel synthesis of titanium-alkoxide precursors, followed by calcination at different temperatures [42]. As-calcined TiO2 samples show monodispersed colloid particles and well-developed TiO2 crystallinity, with finely-tunable anatase–rutile mixed phase. As previously mentioned, Li/EDA solution can selectively reduce the rutile phase of TiO2, and disorder its surface [39]. Since the crystalline phase of spherical TiO2 particles between anatase and rutile can be finely tuned by varying the calcination temperature while maintaining the morphological dimension, the degree of surface disordering can be systemically controlled, resulting in precise control of the band-gap energy.

In this work, we report the synthesis of the spherical TiO2 particle with the disordered rutile surface for photocatalytic hydrogen production. Specifically, monodispersed TiO2 particles with tunable crystalline property are synthesized by sol-gel synthesis, followed by calcination at different temperatures. Since TiO2 samples have different ratios of anatase to rutile phases, the optical properties are conveniently controlled by Li/EDA treatment. The resulting Li/EDA-treated TiO2 samples showed advantageous characteristics, such as uniform particle dimension, favorable dispersity, facile absorption of visible light, and controllable degree of disorder of the surface. Corresponding with such reduced TiO2 spherical photocatalysts, it was possible to achieve enhanced performance in photocatalytic hydrogen production under solar light irradiation. We systemically study and discuss the optical properties, physicochemical characteristics, and photocatalytic performance of the spherical TiO2 with disordered rutile surface.

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

Colored TiO2 spherical particles with the disordered rutile surface were synthesized by a modified sol-gel synthesis, followed by sequential calcination at the desired temperature and chemical reduction, respectively (Figure 1a). More specifically, the synthesis consisted of the following steps: (i) Preparation of a colloidal amorphous TiO2 sphere (AT, amorphous TiO2) (ii) calcination of a TiO2 sphere to convert to the crystalline counterpart (CT-x); and (iii) chemical reduction of crystalline TiO2, to the colored one (RT-X), by using Li/EDA (Li in ethylenediamine) as reducing agent. During preparation of the colloidal TiO2 particle, the monodispersed amorphous TiO2 spheres were synthesized by the sol-gel reaction of titanium n-butoxide (TBOT) in mixed solvent of ethanol and acetonitrile, in the presence of base ammonia and surfactant. The hydrolysis and condensation of the TBOT were highly influenced by several synthetic parameters, such as the solvent environment, the amount of precursor, and the concentration of surfactant. Recently, we systemically studied the effect of the synthetic parameters on the physical–chemical properties of colloidal TiO2 particles, and successfully synthesized uniform TiO2 spheres with controllable crystalline properties [42]. In this work, we also synthesized the uniform spherical TiO2 particles with particle diameter of ca. 290 nm by adapting the previous synthetic method [42]. The SEM image (Figure 1b) clearly shows that uniform spherical particles with a white color powder were well synthesized. After the calcination step, amorphous TiO2 could be crystallized to the crystalline counterparts, which consist of either anatase or rutile phases. The spherical morphology was well maintained, even after high temperature calcination, and the calcined sample showed white color, which is an intrinsic property of crystalline TiO2 (Figure 1c). Chemical reduction using Li/EDA solution could selectively reduce the rutile phase of crystalline TiO2. As previously reported by Park et al., Li-EDA as a strong reducing agent in a superbase can selectively reduce rutile TiO2 to disorder the surface, resulting in black rutile TiO2 having a small band-gap [39]. Since the calcined TiO2 spherical particles had different crystalline ratios of anatase and rutile depending on the calcination temperature, the degree of surface disordering was highly influenced, resulting in the reduced spherical TiO2 particle with different colors. In practical terms, the reduced TiO2 sample showed a blue color with a uniform particle dimension (Figure 1d). After photo-deposition of Pt nanoparticle on the surface of the reduced TiO2 sample, the sample could be used as a photocatalyst for photochemical hydrogen production under solar-light irradiation (Figure 1a).

**Figure 1.** (**a**) Schematic illustration for synthesis of reduced TiO2 samples (RT-x) and Pt-reduced TiO2 catalyst. Corresponding SEM and digital images of (**b**) amorphous TiO2 (AT), (**c**) calcined TiO2 (CT-x), and (**d**) reduced TiO2 (RT-x), respectively.

The morphologies of the calcined TiO2 and the reduced TiO2 samples are investigated by SEM. Figure 2a shows the CT-600 (Calcined TiO2 at 600 oC) sample calcined at 600 ◦C, which reveals uniform spherical morphology with diameter of ca. (281 ± 31) nm. As the calcination temperature increases to 800 ◦C, the spherical morphology is well maintained, indicating the high thermal stability of the synthesized TiO2 particles (Figure 2b). The average diameter of CT-800 is ca. (280 ± 37) nm, which is almost similar to that of CT-600. Although chemical reduction is carried out to produce the colored TiO2 particles, the overall morphology with diameter is unchanged. All RT-x (Reduced TiO2) samples show the monodispersed, spherical morphology with similar diameter compared to the mother CT sample, indicating high chemical stability (Figure 2c,d).

**Figure 2.** SEM images of (**a**) CT-600, (**b**) CT-800, (**c**) RT-600, and (**d**) RT-800. The letters CT and RT indicate calcined and reduced TiO2, respectively. The number after CT and RT indicates its calcination temperature.

The crystalline characteristics of the calcined TiO2 and the reduced TiO2 samples are investigated by X-ray diffraction (XRD). Figure 3a shows the CT-500 sample revealed representative diffraction peaks of TiO2 anatase phase at 2θ = (25.4◦, 37.84◦, 48.12◦, 54.02◦, 55.08◦, and 62.68◦), which are attributed to the (101), (004), (200), (105), (211), and (204) planes, respectively. As the calcination temperature increases, other peaks related to the TiO2 rutile phase dramatically appear. CT-600 exhibits not only the sharp anatase peak, but also new rutile peaks at 2θ = (27.44◦, 36.12◦, 41.2◦, 44◦, 54.32◦, and 56.6◦), corresponding to the (110), (101), (111), (210), (211), and (220) planes, respectively. When the sample is calcined at ever higher temperatures of 700 ◦C (CT-700), the dominant rutile peaks become even sharper. As the calcination temperature increases further to 800 ◦C, major rutile peaks with a small trace of the anatase peak are observed. Based on the above results, it should be noted that the metastable anatase phase was continuously converted to the rutile phase by the thermal transformation of the TiO2 crystalline phase.

The average composition of anatase to rutile phase was calculated from the relative peak area of anatase (101) and rutile (110) peaks, using the following equation [43]:

$$\mathrm{I[A]}\% = 100 \times \mathrm{I\_A/(I\_A + 1.265 \times I\_R)}\tag{1}$$

where IA and IR correspond to the relative areas of the anatase (101) and rutile (110) peaks, respectively. Hence, the rutile content is [R] = 100 − [A]. The rutile contents of the CT-X samples were estimated to be approximately (0%, 67%, 88%, and 100%) for the CT-500, CT-600, CT-700, and CT-800, respectively.

Figure 3b also shows the XRD patterns of the reduced TiO2 sample. RT-500 sample shows the anatase diffraction peaks that are identical XRD patterns to the CT-500 sample. This indicates that the crystalline properties of the RT-500 sample are well maintained, even after the chemical reduction process using Li/EDA solution. RT-600 sample shows similar mixed crystalline patterns of both the anatase phase and rutile phase, but it shows a slightly higher relative peak intensity of anatase, compared to that of CT-600. As the calcination temperature increases, the anatase peak intensity of RT-x samples is interestingly enhanced, compared to that of CT-x. RT-700 and RT-800 exhibit the more obvious anatase (101) peaks, compared to CT-700 and CT-800 samples. In practical terms, the rutile contents were calculated to be approximately (0%, 66%, 84%, and 93%) for the RT-500, RT-600, RT-700, and RT-800 samples, respectively. It should be noted that after Li/EDA reduction, the anatase ratio is obviously increased, while that of the rutile is slightly decreased. It was recently reported that the rutile phase could be selectively disordered by Li in superbase EDA solution [39]. Interestingly, Li/EDA solution as the reducing agent could reduce the ordered white rutile to the disordered black one, resulting in the colored TiO2 with diminution of the rutile phase. In our study, we also observed similar phenomena by using our calcined TiO2 samples, which can have either anatase or rutile phase. Since Li/EDA reducing solution can selectively disorder the surface of rutile phase in the calcined TiO2 spherical particles, significant chemical reduction can induce a decrease of the rutile surface orderliness, resulting in both a decrease of the rutile peak intensity and an increase of the anatase peak, respectively. Thus, the relative content of the anatase phase of RT-x sample increases, compared to that of the CT-x one. However, because the Li/EDA solution cannot completely reduce all the rutile content, the majority of rutile phases on RT-600, RT-700, and RT-800 samples still remain, with obvious appearance of the anatase peaks.

**Figure 3.** X-ray diffraction (XRD) patterns of (**a**) calcined TiO2 sample (CT-x) and (**b**) reduced TiO2 sample (RT-x).

The Raman spectra are also obtained to confirm the surface-structural changes of TiO2 samples (Figure 4). As expected, the as-synthesized amorphous TiO2 particle (AT) does not display any obvious peaks related to the crystalline structure. It showed small intensity changes at the Raman shift of ca. 1000 cm<sup>−</sup>1, indicating the existence of a disordered surface [44]. CT-600 showed the obvious Raman peaks at Raman shift of ca. 400, 550, and 650 cm−1, respectively, indicating representative ordered anatase characteristics. After the disordering process using the Li/EDA solution, RT-600 not only showed the similar Raman peaks at ca. 400, 550, and 650 cm<sup>−</sup>1, but also obvious signal changes at ca. 1000 cm<sup>−</sup>1. It should be noted that the disordered surface was formed by the Li/EDA reduction process. The RT-800 sample, which consists of mainly the rutile phase, displays a significant peak related to the disordered surface of rutile. The above features are also consistent with the trend of the XRD results. Based on the above XRD and Raman spectra, we conclude that the surface disordering of rutile can be easily achieved by a simple Li/EDA reduction process. Based on our observation, it can be considered that the original Ti4<sup>+</sup> state in the boundary of the rutile crystalline grain is preferentially reduced to Ti3<sup>+</sup> in superbase conditions in the presence of Ethylenediamine, which is a similar phenomenon to that previously reported [39].

We also confirmed if the disordered rutile surface could be recovered to the ordered surface by recalcining the RT-800 sample at 800 ◦C. As shown in Figure S1, CT-800 showed the obvious peaks at Raman shift of ca. 451 and 615 cm<sup>−</sup>1, respectively, indicating the ordered rutile surfaces. After Li/EDA treatment, Raman peaks related with the ordered rutile surface of RT-800 were completely disappeared indicating the selective disordering of the rutile surface. When the sample is recalcined again at 800 ◦C, RT-800 recalcination exhibited the peaks of the ordered rutile surface again, indicating recovery of the disordered surface to the ordered counterpart. It should be noted that the ordered rutile surface can first be disordered through the Li/EDA reduction process, then the disordered surface can be recovered to the original ordered rutile surface by heat treatment. Based on our observation, it can be concluded that the original Ti4<sup>+</sup> state in the boundary of the rutile crystalline grain is reduced to Ti3+, then re-oxidized to Ti4<sup>+</sup> states by sequential Li/EDA reduction and recalcination, respectively. It is consistent with the results of a previous study reported by Park et al. [39]. Since the disordered surface indicates different optical property, such as narrowed band-gap and enhanced absorption in visible light, it could be believed that our reduced TiO2 sample showed different band-gap properties and enhanced photochemical performance under visible light conditions.

**Figure 4.** Raman spectra of amorphous TiO2 (AT), CT-600, RT-600, and RT-800.

To investigate the light absorption ability and optical property of TiO2 samples, we carried out UV-Vis diffuse reflectance spectroscopy (UV-Vis DRS). We obtained the absorption spectra of the TiO2 samples employed in this work by UV-Vis DRS techniques (Figure 5). The as-synthesized amorphous TiO2 sample (AT) can only absorb the UV region, indicating poor light absorbance towards solar light. While the calcined TiO2 sample at 600 ◦C (CT-600) displays its main absorption of UV light until ca. 400 nm, the reduced TiO2 sample (RT-600) shows the absorption edge red-shifted and significant increase of light absorption in the range ca. 400–650 nm, indicating large adsorption in the range of visible light. The RT-800 sample has even broader absorption in the range of visible light. We also estimated the band-gap energy of the above samples by using Tauc plot [45]. The band gap energy values of CT, CT-600, RT-600, and RT-800 are calculated as ca. 3.18, 2.93, 2.44, and 0.82 eV, respectively.

It is well known that pure TiO2 shows a band-gap of ca. 3.0–3.2 eV [18]. In this work, the CT-600 sample, which has the mixed phase of anatase and rutile, displays a quite wide band-gap value of ca. 2.93 eV, even though there is small difference compared to previous results [10,11,18]. After chemical reduction using the Li/EDA solution, the RT-600 sample shows a narrowed band-gap (ca. 2.44 eV) with large absorption toward visible light. As previously reported, the Li/EDA solution can selectively make the disordered rutile surface of TiO2 with a mixed phase of anatase and rutile [39]. Although undergoing the same Li/EDA reduction process, the anatase surface of the RT-500 sample can be well maintained, but some rutile surface of RT-600 can be selectively disordered. It is well known that the disordered surface can increase visible light absorption and utilize low-energy light on photocatalysis, which originates from either the regulated conduction band (CB) or valence band (VB) position, and indirect electron recombination [46]. In this study, we also observed that the reduced TiO2 samples (RT-600 and RT-800) had enhanced light absorption ability and optical property, which should originate from the disordered surface of the rutile phase. Thus, during photocatalysis, our reduced TiO2 samples should show enhanced performance.

**Figure 5.** UV-Vis diffuse reflectance spectroscopy (UV-Vis DRS) spectra of amorphous TiO2 (AT), CT-600, RT-600, and RT-800.

The photocatalytic hydrogen production activity of the TiO2 sample was investigated under solar light (1.5 air mass, 1.5 AM) irradiation using methanol/H2O solution (Figure 6a,b). Before employing the synthesized TiO2 samples as photocatalysts, 1 wt.% of Pt was deposited on the surface of each TiO2 sample. Figure 6a shows that when the Pt/CT-600 sample was used as the photocatalyst, only a negligible amount of hydrogen was produced. This indicates the low photocatalytic activity of Pt/CT-600, which is attributed to small photon absorption of the CT-600 sample toward solar light, which consists of mainly visible and infrared light. However, Pt/RT-600 shows remarkable improvement in photocatalytic hydrogen production (Figure 6a). It should be noted that both the improved visible light absorption and the narrowed band-gap of RT-600 enhance the light absorption, resulting in the dramatically improved photocatalysis activity of Pt/RT-600. Figure 6b shows the effect of calcination temperature of RT-x samples on photocatalytic hydrogen production with light irradiation time. Among the catalysts tested, the Pt/RT-600 catalyst shows the highest catalytic activity. The relative photocatalytic activity of the catalysts toward hydrogen production follows the order: Pt/RT-600 > Pt/RT-500 > Pt/RT-700 ≈ PT/RT-800.

**Figure 6.** Hydrogen production amount of 1 wt % Pt-deposited TiO2 photocatalyst: (**a**) CT-600 and RT-600; and (**b**) RT-500, RT-600, RT-700, and RT-800.

The photocatalytic activity of RT-x samples can be explained from the above characterization results. Since all RT-x samples was prepared from the same colloidal amorphous TiO2 by calcination at different temperatures, followed by the same Li/EDA reduction process, it should be noted that the performance differences originate from the different degrees of rutile surface disordering, which is ascribed to different crystalline properties. The RT-500 sample, which is mainly anatase phase, has a small amount of disordered surface. In addition, since there is a small portion of UV-light in solar light, Pt/RT-500 can generate a considerable amount of hydrogen. The RT-600 sample, which consists of a mixed crystalline phase of anatase and rutile, can not only favorably absorb visible light over the reduced rutile, but it can also separate charge carriers to anatase, resulting in a lot of elongated charge carriers to accelerate photocatalysis. It is well known that the outstanding activity of P25-TiO2 is mainly contributed by the mixed phase composition of anatase and rutile, which has beneficial effects on the absorption and separation of excited charges. Even though there is still controversy over the exact functions of each crystalline phase in the mixed phase, it is certain that the existence of the anatase–rutile mixed phase can have an unexpected and beneficial performance in photocatalysis [47,48]. Our RT-600 sample also had similar beneficial effects on light absorption and charge separation, due to the existence of the mixed phase. In practical terms, Pt/RT-600 shows the best hydrogen production performance. However, as the calcination temperature increases, the crystalline phase becomes mainly a rutile phase. RT-700 and RT-800 have a major rutile crystalline phase and a large portion of the disordered surface, resulting in the large absorption of visible light in the UV-DRS data. Although they can absorb visible light, the recombination of excited electron-hole pairs is severely constrained, due to the highly disordered surface and negligible portion of anatase. Although RT-800 sample can absorb the most visible light and the color of the catalyst is dark blue, Pt/RT-800 shows negligible hydrogen production.

To further confirm the photochemical properties of the TiO2 samples, we measured the photocurrent by conducting chronoamperometry (CA) under an inducing potential of 0.6 V (vs. Ag/AgCl) and a periodic irradiation of solar light with a 400 nm cut-off filter. Figure 7 shows that the photocurrent is closely related to the degree of charge separation under light irradiation. Without light exposure, samples display electrochemical currents. When the catalysts are irradiated by light conditions, the current density can be increased due to the contribution from photo-generated electrons. The TiO2 sample (CT-600) calcined at 600 ◦C exhibits a considerable photocurrent, indicating the existence of charge separation by light irradiation. The RT-600 sample shows the largest photocurrent among the TiO2 samples tested. However, the RT-800 sample displays smaller photo-generated current than RT-600, indicating the restricted recombination of electron-holes, even though it can absorb a large portion of solar light. Based on both the characterization results and photoelectrochemical data, it can be concluded that the RT-600 photocatalyst, which consists of mixed phase with a disordered rutile surface, can have advantageous effects, such as visible light absorption, and favorable charge separation, resulting in improved photocatalysis activity on photocatalytic hydrogen production.

**Figure 7.** Photoelectrochemical chronoamperometry curves of various TiO2 samples obtained at 0.6 V vs. Ag/AgCl under solar light illumination with 400 nm cut-off filter.

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

#### *3.1. Materials*

Ethyl alcohol (C2H5OH, 99.9%, anhydrous), Methyl alcohol (CH3OH, 99.9%, anhydrous), acetonitrile (ACN, CH3CN 99.9%, special guaranteed grade), and ammonium hydroxide (NH4OH, 28%) were obtained from Daejung Chemical Company. Titanium (IV) n-butoxide (TBOT, 97%, reagent grade) and Hydroxypropyl cellulose (HPC, MW ≈ 80,000) were obtained from Aldrich. Metallic Li foil was purchased from Alfa Aesar chemical company. Ethylenediamine (EDA, 99%, guaranteed grade) was obtained from Sigma-Aldrich. All chemicals were used as received.

#### *3.2. Synthesis*

Spherical TiO2 particles with average diameter of 290 nm were prepared by a sol-gel synthesis in mixed solvent, followed by calcination, as recently reported [42]. HPC (50 mg), as a surfactant, was completely dissolved in the mixed solvent solution (100 mL) of ethanol and acetonitrile with a volume ratio of 3:1. After completely dissolving the HPC, ammonia solution (0.8 mL) was added to the above solution. After stirring for 20 min, a mixture of TBOT (4 mL) in the mixed solvent of ethanol (12 mL) and acetonitrile (4 mL) was quickly injected into the above solution. The mixture was vigorously stirred for 2 h at room temperature. The white precipitate was isolated by centrifugation, and washed with ethanol and with de-ionized (D.I.) water several times. Then, the amorphous AT sample was obtained by drying under vacuum.

The dried AT sample was charged in an alumina boat in a furnace, and calcined at the desired temperature for 3 h under air conditions. The amorphous AT samples were crystallized to either anatase or rutile phase, which is highly dependent on the calcination temperature. Calcined CT samples are denoted as CT-X (where X is the calcination temperature). To synthesize colored TiO2 spherical particle, we used the Li-EDA reduction method, which was recently developed by Park et al. [39]. A piece of metallic Li foil (45 mg) was dissolved in ethylenediamine (40 mL) to form a solvated electron solution. The calcined CT-X (400 mg) samples were added into the above solution, and vigorously stirred for 6 days under inert (N2) conditions. After 6 days, a diluted HCl solution (1 M, 6.5 mL) was slowly added dropwise into the mixture, in order to quench the excess electron. The reduced TiO2 particles were isolated by centrifugation, washed with D.I. water 3 times, and dried in vacuum chamber at room temperature to give RT-X (where X is the calcination temperature).

#### *3.3. Characterization*

Digital photo images of TiO2 samples were obtained by the digital camera function of iPhone (Apple Inc.). The particle morphology and uniformity were investigated by scanning electron microcopy (SEM, JSM-6060, JEOL). The crystalline properties of the TiO2 samples were investigated by X-ray diffraction (XRD, D/MAX 2200, Rigaku). Optical absorbance spectra were studied by UV-vis spectroscopy using a UV-vis spectrophotometer with diffuse reflectance accessary (UV-DRS, V670, Jasco). Photoelectrochemical analyses were carried out using a conventional three-electrode system, with Ag/AgCl as a reference electrode, and Pt gauze as a counter electrode. The working electrode was prepared by deposition of a sample slurry on indium-tin oxide (ITO) glass (1 × 1 cm) [12]. An aqueous Na2SO4 (0.1 mol/L) solution containing methanol (10 vol.%) was used as the electrolyte. Chronoamperometry tests were conducted using a potentiostat (SP-150, BioLogic).

#### *3.4. Photocatalytic Hydrogen Production*

Photocatalytic hydrogen production was conducted in a Pyrex glass reactor. TiO2 photocatalyst samples (20 mg) were well dispersed in an aqueous methanol solution (50%, 50 mL). An ABET 150 W Xe lamp (ABET technologies inc. USA) with an AM 1.5 G air mass filter was used as a light source for solar light irradiation. The amount of hydrogen produced was determined by conventional gas chromatography with a thermal conductivity detector (GC-TCD, HP-5890 equipped with a Molecular Sieve-5A packed column).

#### **4. Conclusions**

We synthesized the uniform spherical TiO2 particles with disordered rutile surface, characterized both the physicochemical and optical properties, and demonstrated photocatalytic performances on photochemical hydrogen production under solar light conditions. The synthesis involves several sequential processes: (i) Synthesis of uniform-sized amorphous TiO2 particles by sol-gel reaction of the TiO2 precursor, (ii) calcination of amorphous TiO2 sample to convert to its crystalline counterpart, and (iii) chemical reduction of the calcined TiO2 sample to make the disordered rutile surface. The as-synthesized amorphous TiO2 sample showed uniform and monodispersed spherical morphologies. Calcination at varied temperatures induced different crystalline characteristics, such as different ratios of anatase and rutile phases, and chemical reduction using Li/EDA enables selective disordering of the rutile surface, resulting in different optical characteristics, such as the degree of visible light absorption ability and band-gap energy. The chemically-reduced TiO2 sample (RT-600) prepared by calcination at 600 ◦C, followed by Li/EDA reduction, displays beneficial characteristics in terms of light absorption and charge separation, such as a disordered rutile surface, and a mixed crystalline phase of anatase and rutile. Among the TiO2 samples employed in this work, the RT-600 sample showed significantly enhanced catalytic activity in photocatalytic hydrogen production. We believe that the proposed technique reported in this study can provide an effective method for developing visible light-responsive TiO2-based photocatalysts.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4344/9/6/491/s1, Figure S1: Raman spectra of CT-800, RT-800, and RT-800 recalcination.

**Author Contributions:** Investigation, writing—original draft, N.Y.K.; investigation, H.K.L.; resources, J.T.M.; conceptualization, writing—review and editing, J.B.J.

**Funding:** This research was funded by Konkuk University in 2016.

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

#### **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*

## **Photoelectrochemical Hydrogen Evolution and CO2 Reduction over MoS2**/**Si and MoSe2**/**Si Nanostructures by Combined Photoelectrochemical Deposition and Rapid-Thermal Annealing Process**

#### **Sungmin Hong, Choong Kyun Rhee and Youngku Sohn \***

Department of Chemistry, Chungnam National University, Daejeon 34134, Korea; qwqe212@naver.com (S.H.); ckrhee@cnu.ac.kr (C.K.R.)

**\*** Correspondence: youngkusohn@cnu.ac.kr; Tel.: +82-42-821-6548

Received: 17 April 2019; Accepted: 25 May 2019; Published: 28 May 2019

**Abstract:** Diverse methods have been employed to synthesize MoS2 and MoSe2 catalyst systems. Herein, a combined photoelectrochemical (PEC) deposition and rapid-thermal annealing process has first been employed to fabricate MoS2 and MoSe2 thin films on Si substrates. The newly developed transition-metal dichalcogenides were characterized by scanning electron microscopy, Raman spectroscopy and X-ray photoelectron spectroscopy. PEC hydrogen evolution reaction (HER) was demonstrated in an acidic condition to show a PEC catalytic performance order of MoOx/Si < MoS2/Si << MoSe2/Si under the visible light-on condition. The HER activity (4.5 mA/cm<sup>2</sup> at <sup>−</sup>1.0 V vs Ag/AgCl) of MoSe2/Si was increased by 4.8× compared with that under the dark condition. For CO2 reduction, the PEC activity was observed to be in the order of MoS2/Si < MoOx/Si << MoSe2/Si under the visible light-on condition. The reduction activity (0.127 mA/cm2) of MoSe2/Si was increased by 9.3× compared with that under the dark condition. The combined electrochemical deposition and rapid-thermal annealing method could be a very useful method for fabricating a thin film state catalytic system perusing hydrogen production and CO2 energy conversion.

**Keywords:** MoS2; MoSe2; photoelectrochemical deposition; rapid-thermal annealing; hydrogen evolution; CO2 reduction

#### **1. Introduction**

Transition-metal dichalcogenides (TMDCs) have widely been studied for applications to energy and environment such as hydrogen evolution and CO2 reduction [1–10]. Especially, molybdenum disulfide and diselenide (MoS2 and MoSe2) materials with two-dimensional character have been synthesized using diverse synthesis methods for their applications [11]. Ye et al. employed a chemical vapor deposition (CVD) method to synthesize monolayer MoS2 followed by oxygen plasma treatment or hydrogen annealing. They showed that hydrogen evolution reaction (HER) activity (e.g., onset potential and current density) was increased substantially by engineering the defects [12]. To increase HER activity of MoS2 or MoSe2, various defect engineering techniques have been employed, which include laser irradiation [13], ion irradiation [14,15] and NaClO chemical etching [16]. Li et al. examined various defect sites of MoS2 such as edge sites, S vacancies and grain boundaries and showed that edge sites and S vacancies (with optimal vacancy density of 7–10%) were main HER active sites [17]. Chang et al. employed lithium molten salts to synthesize 2H- and 1T-MoS2 monolayers at calcination temperatures of 400~600 ◦C and above 1000 ◦C, respectively [18]. They observed that metallic 1T-MoS2 showed a higher HER activity than that of semiconducting 2H-MoS2. Two step hydrothermal method was employed to synthesize 1T@2H-MoSe2 nanosheets, which showed a higher HER activity [19]. Wang et al. prepared MoS2 nanosheets by mechanical exfoliation, transferred onto a SiO2 surface and made a HER device [20]. Afterwards, they showed that HER activity was increased by applying an extra positive electric field. Guo et al. prepared oxygen-incorporated MoS2 sheets on graphene by a hydrothermal method [21] and showed that the active edge sites and conductivity were increased by oxygen-incorporation and the electrical transfer was increased by hybridization. Consequently, the HER activity was found to be substantially increased. Zhu et al. fabricated h-MoO3/1T-MoS2 heterostructures and tested photoelectrocatalytic HER activity to show better activity compared with those of 1T@2H-MoS2 and α-MoO3/MoS2 [22]. A two-step (MoO3 + H2 → MoO2 and MoO2 + Se vapor → MoSe2) chemical vapor deposition (CVD) process was employed to fabricate vertically aligned core–shell MoO2/MoSe2 nanosheet arrays which showed better HER activity than those of MoO2 and MoSe2 [23]. Electrochemical CO2 reduction is another potential application area for MoS2 and MoSe2 [4,24–29]. Francis et al. tested a single crystal MoS2 electrode for CO2 reduction and showed a Faradaic efficiency of ~3.5% at −0.59 V (vs Reversible Hydrogen Electrode) for a reduction product of 1-propanol [4]. Asadi et al. reported that electrochemical CO2 reduction product for vertically aligned MoS2 in an ionic liquid was found to be CO with a CO2 reduction current density of 130 mA cm−<sup>2</sup> at −0.764 V [28,29].

Electrodeposition has popularly been employed for the cheap fabrication of thin films on a substrate, where major factors determining the nature of thin film include electrolyte, pH, deposition time and an applied potential [30]. For electrodeposition (under the dark condition) of Mo oxides on a substrate, some studies have been reported [31–34]. However, no studies have been reported for photoelectrochemical deposition of Mo oxides. Petrova et al. used Al substrates for electrodeposition of Mo oxides in Mo ion electrolyte (Mo(NH4)6Mo7O24·4H2O, 20 g/L) at pH of 8–10 adjusted by a NH3-CH3COONH4 buffer [31]. Pd–MoO*<sup>x</sup>* catalyst on glassy carbon electrode was reported to be fabricated by electrodeposition at potential ranges between −0.73 and +0.2 V in a mixed solution of 2 mM PdCl2, 15 mM Na2MoO4 and 0.2 M HCl [32]. The dominant oxidation state of Mo was found to be +6. Uniform Mo oxide (+6, +5 and +4 oxidation states) nanostructure arrays (nanotubes at pH = 2.7 and nanowires at pH = 5.5) were prepared by a template electrodeposition in Mo ion electrolyte ((NH4)6Mo7O24·4H2O, 50 g/L) [33]. Electroless-photochemical deposition (PCD) has also been demonstrated for the preparation of metal sulfide thin films. Soundeswaran et al. prepared CdS films on indium tin oxide (ITO) glass using 1–10 mM Cd(CH3COO)2 and 100 mM Na2S2O3 solution at pH = 3.0–4.5 under irradiation of ultraviolet (UV) light (100 mW/cm2) [35]. For this reaction, S and electrons were initially formed by UV-excitation of S2O3 2- ions and reacted with Cd metal ions to form CdS. Podder et al. prepared CuxS thin films on ITO glass using a similar method [36].

Herein, a new methodology of combined photoelectrochemical deposition (Mo6<sup>+</sup> <sup>+</sup> 6OH<sup>−</sup> <sup>→</sup> MoO3 + 3H2O accelerated by an enhanced photocurrent) and rapid-thermal annealing (RTA) sulfurization (or selenization) process (2MoO3 + 4S or Se → 2MoS2 or MoSe2 + 3O2) was introduced to fabricate thin MoS2 and MoSe2 films on Si substrates. HER and CO2 reduction tests were demonstrated to show a potential applicability to energy and environment. A major advantage of the combined method is time-saving and cost effective. Another advantage is morphology and thickness-controlled by tuning applied voltage, deposition time and the electrolyte condition and so forth. Overall, the present developed method could be further improved and widely used for developing better thin film systems for diverse application areas.

#### **2. Results**

Surface morphology, crystal phase formation and surface chemical states were examined using scanning electron microscopy (SEM), Raman and X-ray photoelectron spectroscopy (XPS), respectively. Hydrogen evolution reaction (HER) and CO2 reduction were tested using the three electrode system. The experimental results are described below.

#### *2.1. SEM Morphology*

Figure 1 shows the SEM images of MoOx/Si, MoS2/Si and MoSe2/Si samples. For the as-photoelectrodeposited MoOx/Si sample, larger (200~400 nm diameter) and smaller (< 50 nm) nanoparticles were formed on the Si surface. Two different particle size distributions may be due to mixed Volmer-Weber island growth mode and Ostwald ripening process. The oxidation states were confirmed by XPS, discussed below. Under the dark condition, no electrodeposition was observed. For the SEM image of MoS2 formed by RTA process of MoOx/Si sample, two different size distributions were also observed for the sample as expected. For the SEM image of MoSe2 formed by RTA process of MoOx/Si sample, the surface morphology showed more uniform nanostructure.

**Figure 1.** Scanning electron microscope (SEM) images of MoOx/Si (**A** and **A1**), MoS2/Si (**B** and **B1**) and MoSe2/Si (**C** and **C1**) samples.

#### *2.2. Raman Spectroscopy*

Raman spectra (Figure 2) were obtained to examine the detailed crystal phase formation for the catalyst systems. For all the samples, the strongest peak was commonly observed at 524 cm−<sup>1</sup> (not shown in the Figure), attributed to Si used as a support [37]. A weak peak at ~300 cm−<sup>1</sup> for MoOx/Si was due to the phonon mode of Si [37]. No Raman peaks of Mo oxides were observed, indicating that the as-photoelectrodeposited sample was ultrathin and/or amorphous. For MoS2 sample (Figure 2B), two Raman peaks were observed at 385 and 411 cm<sup>−</sup>1, attributed to in-plane Mo-S (E2g) and out-of-plane Mo-S (A1g) vibration modes of hexagonal MoS2 [38,39]. For MoSe2 sample (Figure 2C), a strong Raman peak was observed at 238 cm<sup>−</sup>1, attributed to Mo-Se A1g mode. Other two peaks at 169 and 286 cm−<sup>1</sup> were assigned to E1g and E2g modes, respectively [40,41]. The A1g mode was found to be stronger than the E2g mode. Overall, based on the Raman data, the RTA process was found to be efficient for the fabrication of MoS2 and MoSe2.

**Figure 2.** Raman spectra of (A) MoOx/Si, (B) MoS2/Si and (C) MoSe2/Si samples.

#### *2.3. X-ray Photoelectron Spectroscopy*

X-ray photoelectron spectra (XPS) of the three catalyst systems are displayed in Figure 3. For the as-photoelectrodeposited MoOx/Si, Mo and O elements were dominantly observed. Mo 3d3/<sup>2</sup> and 3d5/<sup>2</sup> XPS peaks were observed at 235.5 and 232.5 eV, respectively with a spin-orbit splitting energy of 3.0 eV. This is attributed to Mo6<sup>+</sup> oxidation state of MoO3 [42]. Furthermore, the other Mo 3d3/<sup>2</sup> and 3d5/<sup>2</sup> XPS peaks were observed at 233.8 and 230.8 eV, respectively with a spin-orbit splitting energy of 3.0 eV. This could be due to Mo5<sup>+</sup> oxidation state. Based on the Mo 3d XPS fitting, 42% and 52% of Mo were 6+ and 5+ oxidation states, respectively. The corresponding broad O 1s peak was observed at 530.6 eV, attributed to the lattice oxygen of Mo oxides (MoOx). For the Mo 3d XPS of the MoS2/Si sample, Mo 3d3/<sup>2</sup> and 3d5/<sup>2</sup> XPS peaks were observed at 232.2 and 229.0 eV, respectively with a spin-orbit splitting energy of 3.2 eV. This is attributed to Mo4<sup>+</sup> oxidation state of MoS2 [43,44]. A small peak at 226.3 eV was due to S 2s [45]. The corresponding S 2p1/<sup>2</sup> and S 2p3/<sup>2</sup> XPS peaks were observed at 163.1 eV and 161.9 eV, respectively with a spin-orbit splitting of 1.2 eV. This is in good agreement with the literature reported by Jian et al. for 1T phase MoS2 [44]. For the Mo 3d XPS of the MoSe2/Si sample, Mo 3d3/<sup>2</sup> and 3d5/<sup>2</sup> peaks were observed at 231.7 and 228.6 eV, respectively with a spin-orbit splitting energy of 3.1 eV. This is in good agreement with +4 oxidation state of MoSe2 [5]. The corresponding Se 3d5/<sup>2</sup> and 3d3/<sup>2</sup> peaks were found to be located 55.1 and 54.2 eV, respectively. The O 1s XPS peak was observed at 533.1 eV, attributed adsorbed surface oxygen [42]. For MoOx/Si sample, the Mo:O ratio was estimated to be 1:4.1. The overestimated oxygen was plausibly due to surface oxygen such as H2O and OH. Mo:O ratio was estimated to be 1: 2.3 by only considering the lattice O 1s signal. For MoS2/Si sample, Mo:S ratio was calculated to be 1:1.95, very close to MoS2. For MoSe2/Si sample, Mo:Se ratio was calculated to be 1:2.15, close to MoSe2. For the valence band (VB) XPS spectra, the density of states (DOS) near the Fermi level was more discernible for MoS2 and MoSe2 reflecting metallic/semiconducting states.

**Figure 3.** Mo 3d, O 1s, Se 3d, S 2p and valence band (VB) X-ray photoelectron spectra (XPS) for MoOx/Si, MoS2/Si and MoSe2/Si samples.

#### *2.4. Photoelectrochemcial Hydrogen Evolution*

Three different catalyst systems of MoOx/Si, MoS2/Si and MoSe2/Si were tested for hydrogen evolution reaction in 0.1 M H2SO4 electrolyte. Linear sweep voltammetry (LSV) curves (Figure 4) were obtained from +0.2 V to −1.0 V at a scan rate of 10 mV/sec after full nitrogen gas purging under the dark and visible light exposure conditions. Under the dark condition, the current density at -1.0 V (vs Ag/AgCl) showed the order of MoOx/Si (0.08 mA/cm2) < MoS2/Si (0.11 mA/cm2) << MoSe2/Si (0.86 mA/cm2) while the order changed to MoS2/Si (0.51 mA/cm2) < MoOx/Si (0.64 mA/cm2) << MoSe2/Si (4.3 mA/cm2) under the visible light exposure condition. They all commonly showed an increase in HER CD under visible light exposure. The inset in Figure 4 displays LSV curves taken under the light ON-and-OFF condition during the scan. It clearly showed that all three samples have photocatalytic activity.

**Figure 4.** Linear sweep voltammetry curves (voltage range: +0.2~1.0 V) at a scan rate of 10 mV/sec) under visible light exposure for (a) MoOx/Si, (b) MoS2/Si and (c) MoSe2/Si samples. Inset LSV curves were taken under the dark condition. The corresponding light ON-and-OFF LSV curves are displayed in the inset. Inset photo shows the bubble formed on the catalyst surface during LSV.

#### *2.5. Photoelectrochemcial CO2 Reduction*

For CO2 reduction in 0.1 M NaHCO3 electrolyte, the LSV measurements (Figure 5) were obtained from +0.2 V to −1.0 V upon full N2 and CO2 gas purging at a scan rate of 10 mV/sec under the dark and visible light exposure conditions, respectively. Upon full N2 gas purging in the electrolyte under dark condition (inset in Figure 5A), the current densities (CDs) at −1.0 V were observed to be 0.0043, 0.0041 and 0.012 mA/cm2 for MoOx/Si, MoS2/Si and MoSe2/Si, respectively. Under the visible light exposure condition (Figure 5A), the CDs were found to be drastically increased to 0.037, 0.011 and 0.018 mA/cm2 for MoOx/Si, MoS2/Si and MoSe2/Si, respectively. Upon full CO2 gas purging in the electrolyte under the dark condition (inset in Figure 5B), the CDs at −1.0 V were found to be 0.007, 0.009 and 0.014 mA/cm2 for MoOx/Si, MoS2/Si and MoSe2/Si, respectively. Under the visible light exposure condition (Figure 5B), the CDs were found to be increased to 0.043, 0.023 and 0.13 mA/cm<sup>2</sup> for MoOx/Si, MoS2/Si and MoSe2/Si, respectively. The CDs were increased by 6.0, 2.6 and 9.4×, respectively upon visible light exposure. The inset in Figure 5B displays LSV curves taken under the light ON-and-OFF condition during the scan. It clearly showed that all three samples also have photocatalytic activity.

**Figure 5.** Linear sweep voltammetry curves (voltage range: +0.2~1.0 V) in N2- (**A**) and CO2-purged (**B**) 0.1 M NaHCO3 electrolyte at a scan rate of 10 mV/sec under dark (in the corresponding inset Figure) and the visible light exposure condition for (a) MoOx/Si, (b) MoS2/Si and (c) MoSe2/Si samples. The inset (**B**) shows the light ON-and-OFF LSV curves.

#### **3. Discussion**

The photo-electrochemical deposition method was first successfully employed to fabricate MoOx on a Si support. Upadhyay et al. reported MoOx nanoparticles on a stainless steel support prepared by electrodeposition at -1.0 V in a mixed solution of 0.1 M Na2MoO4 and 0.1 M NH4NO3 [46]. They reported oxidation states of Mo6<sup>+</sup> and Mo5<sup>+</sup> for MoOx and the corresponding O 1s peak at 530.6 eV. This is in good agreement with our present results. Liu et al. performed electrodeposition of MoOx films on a Ti (130 nm)/Si substrate in a mixed solution of 0.1 M Na2MoO4, 0.1 M Na2EDTA and 0.1 M CH3COONH4 [47]. They concluded that the as-electrodeposited MoOx film (with oxidation states of Mo4<sup>+</sup> and Mo5<sup>+</sup>) was amorphous. This is also in good agreement with the present result, as discussed above (Raman spectra in Figure 2). Based on the SEM image, the photoelectrochemical deposition of MoOx on a Si support appeared to be occurred through Volmer-Weber island growth process [48]. Small islands were initially formed on the entire surface and then larger islands subsequently were grown. For the SEM images of MoS2 and MoSe2 by the RTA process, the morphology of MoS2 was more similar to that of MoOx, compared with that of MoSe2. This indicates that less energy was required for the formation of MoS2, compared with that for MoSe2. Overall, the RTA process was found to be efficient for the formation of MoS2 and MoSe2 without much impacting the original morphology of electrodeposited MoOx.

For HER in 0.1 M H2SO4, MoSe2/Si showed a much higher electrochemical activity (or current density) than MoS2/Si. For HER mechanism in the acidic condition, adsorbed hydrogen is known to be formed via H3O<sup>+</sup> <sup>+</sup> <sup>e</sup>−→ Had <sup>+</sup> H2O. Then, hydrogen is generated via Had <sup>+</sup> H3O<sup>+</sup> <sup>+</sup> <sup>e</sup><sup>−</sup> <sup>→</sup> H2 <sup>+</sup> H2O or Had + Had → H2 [23]. In the mechanism, hydrogen adsorption Gibbs free energy, ΔGHX (X = S or Se) is known to play a major role in determining the activity [10]. The optimal condition is ΔGHX = 0 eV and ΔGHX of MoSe2 is closer to the optimal condition than that of MoS2. The HER activity of MoSe2 has commonly been reported to be higher than that of MoS2 [49]. This is in good consistent with the present result. For HER of nanoflowers-like MoS2 and MoSe2 materials on GC electrodes in 0.5 M H2SO4, Ravikumar et al. reported that the activity (11 mA/cm2 at 0.3 V vs RHE) of MoSe2 showed a higher than that (7 mA/cm<sup>2</sup> at 0.3 V vs RHE) of MoS2, attributed to higher electrical properties and defects [50], in good consistent with the present result. Evidently, based on the DOS near the Fermi level as discussed in Figure 3, the enhanced electronic conductivity could play an important

role in HER and CO2 reduction performances [22]. Because the Gibbs free energy is also known to be dependent on morphology (e.g., defects and edge sites) the catalyst fabrication methods is important for improving a catalyst activity. Upon visible light exposure on the catalyst surface during the LSV, an increased CD was commonly been observed. The enhancement factors (light ON/OFF CD ratio) for HER at −1.0 V upon visible light irradiation were observed to be 8.0, 4.7 and 5.0 for MoOx/Si, MoS2/Si and MoSe2/Si, respectively (Figure 6). The photocatalytic HER activity under visible light was due to photo-generated electron by absorption of light in the visible region [22,42]. As mentioned above, in HER mechanism electron plays a crucial role in generation of hydrogen. Overall, the photocatalytic activity is all dependent on material nature (e.g., electrical conductivity), morphology, surface natures (e.g., defects) and light absorption efficiency. For the splitting of water, the molar stoichiometric ratio of H2/O2 is ideally 2.0 [51] assuming that no other side electrochemical reactions are involved for MoS2 and MoSe2 [1]. Before further discussion, it should be here mentioned that our conclusion was based only on the CD. The H2/O2 production ratios and CO2 reduction products are needed to be further examined by gas chromatography [51].

For electrochemical CO2 reduction, the current density was commonly been increased in CO2-purged 0.1 M NaHCO3 electrolyte, compared with that in N2-purged 0.1 M NaHCO3 electrolyte. This indicates that CO2 reduction was occurred for all the samples. For CO2 reduction (Figure 6), under the dark condition, the enhancement factors before (only N2 bubbling) and after CO2 bubbling were observed to be 1.6, 2.2, 1.2 for MoOx/Si, MoS2/Si and MoSe2/Si, respectively. Under the visible light-on condition, the enhancement factors before (only N2 bubbling) and after CO2 bubbling were observed to be 1.1, 2.1 and 7.2 for MoOx/Si, MoS2/Si and MoSe2/Si, respectively. The CDs upon light exposure were increased by 6.0, 2.6 and 9.4× for MoOx/Si, MoS2/Si and MoSe2/Si, respectively. Overall, MoSe2 showed the most dramatic photo-electrochemical CO2 reduction efficiency. For MoSe2, photogenerated electrons are created by light absorption in the visible region and electron-hole recombination is suppressed by good electron transport [1]. Consequently, the CD by photoelectrochemical CO2 reduction is enhanced.

**Figure 6.** Enhancement factors for HER CDlight ON/ HER CDlight OFF, CDCO2 bubbling/CDN2 bubbling under light ON condition and CDCO2 bubbling/CDN2 bubbling under the light OFF condition for (**a**) MoOx/Si, (**b**) MoS2/Si and (**c**) MoSe2/Si samples.

#### **4. Materials and Methods**

For photoelectrochemical Mo deposition (MoOx/Si), a three-electrode (Ag/AgCl reference, Pt wire counter and Si working electrode) electrochemical cell was used using a VersaSTAT3 (Princeton Applied Research, Oak Ridge, TN, USA) potentiostat galvanostat. For the preparation of a Si working support electrode before electrodeposition, a single-side polished Si (100) wafer (B-doped p-type, thickness of 525 ± 20 μm, resistivity of 1–10 Ω·cm, 2 cm × 0.5 cm) was used as the support, cleaned in 2% HF solution to remove oxide layer and washed with deionized water by sonication. The electrolyte was a mix of 15 mM Na2MoO4 (99.0% extra pure, Samchun Chem. Co., Seoul, Republic of Korea), 1.0 M NaCl and 1.0 M NH3Cl, where pH was adjusted to 9.2 using NH3OH (28~30%, Samchun Chem. Co., Seoul, Republic of Korea). The photoelectrochemical deposition was performed at an applied potential of −1.5V (versus Ag/AgCl electrode) for 10 sec under visible light exposure onto the working electrode. In the present study, we only showed the samples prepared at fixed parameters among different applied potentials and deposition times. No efficient (or less uniform) Mo deposition occurred under the dark condition although the applied voltages and deposition times were varied. CDs of −0.10 and <sup>−</sup>0.92 mA/cm2 were measured at <sup>−</sup>1.5 V under dark and visible light, respectively. The CD was enhanced by 9–10× in the potential ranges from −1.0 to −2.0 V. For the preparation of MoS2 and MoSe2 on Si support, a rapid-thermal annealing (RTA) method was employed using a LABSYS RTP-1200 (Nextron Co., Ltd., Busan, Republic of Korea). For this, a MoOx/Si substrate was placed on a quartz plate (15 mm × 20 mm) in the RTA chamber. Sulfur (S.P.C. GR reagent, Shinyo Pure Chem. Co., Ltd., Hyogo, Japan) or Selenium (99.5+%, 100mesh, Sigma-Aldrich, St. Louis, MO, USA) powder (~0.02 g) was placed below the substrate. The chamber was maintained in 5% H2 in He balance. The temperature heating rate was 12 ◦C/sec and the time was 6 sec at the maximum temperature of 700 ◦C. The surface morphology of the prepared samples was examined using a Hitachi S-4800 (Tokyo, Japan) scanning electron microscope at an electron acceleration voltage of 10.0 kV. Raman spectra with 514 nm laser line were obtained using a LabRAM HR-800 microRaman spectrometer (HORIBA Jobin Yvon, Kyoto, Japan). X-ray photoelectron spectra were taken using a Thermo Scientific K-Alpha<sup>+</sup> X-ray photoelectron spectrometer with micro-focused monochromatic Al Kα X-ray source and a hemispherical energy analyzer. XPS spectra curve fitting was performed using a XPSPEAK ver. 4.1 software. For XPS element quantification, XPS sensitivity factors of 2.75, 0.66, 0.54, and 0.67 were used for Mo 3d, O 1s, S 2p and Se 3d, respectively [52]. For photoelectrochemical HER and CO2 reduction, a three-electrode electrochemical cell was also used using a VersaSTAT3 potentiostat/galvanostat. For HER, nitrogen gas was fully purged into the electrolyte (0.1 M H2SO4 solution) to minimize an effect of dissolved oxygen. Linear sweep voltammetry (LSV) was carried out at a scan rate of 10 mV s−<sup>1</sup> from +0.2 V to <sup>−</sup>1.0 V under dark and visible light exposure conditions. A white LED USB Flashlight (A-10, Teckmedia) was used for visible light (400~700 nm) [53]. For CO2 reduction experiment in 0.1 M NaHCO3 solution, LSV was conducted after N2 gas purging at a scan rate of 10 mV s−<sup>1</sup> from <sup>+</sup>0.2 V to <sup>−</sup>1.0 V under dark and visible light irradiation conditions. The same LSV experiment was also conducted after CO2 gas purging into the electrolyte to examine the CO2 effect.

#### **5. Conclusions**

In this work, a combined photoelectrochemical deposition and rapid-thermal annealing method was first been employed to fabricate MoS2 and MoSe2 thin films on Si substrates. Photoelectrochemical HER and CO2 reduction were demonstrated for the newly developed catalytic systems. The main results are as follows:


mainly based on the Raman and XPS results. The maximum temperature was achieved by rapid heating to 700 ◦C of S or Se powers on the MoOx/Si and maintained for 6 sec.


The newly developed catalyst preparation method could be very useful for developing thin film catalyst systems for diverse application areas.

**Author Contributions:** Y.S. designed the experiments and wrote the paper; S.H. performed the experiments; C.K.R. provided valuable idea for obtaining the data.

**Funding:** This research was funded by the National Research Foundation of Korea (NRF) grant funded by the Korean government (MEST), grant number NRF-2016R1D1A3B04930123. The APC was funded by the National Research Foundation of Korea (NRF).

**Acknowledgments:** This work was financially supported by the National Research Foundation of Korea (NRF) grant funded by the Korean government (MEST) (NRF-2016R1D1A3B04930123).

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

#### **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* **Band Gap Modulation of Tantalum(V) Perovskite Semiconductors by Anion Control**

**Young-Il Kim 1,2,\* and Patrick M. Woodward <sup>2</sup>**


Received: 11 January 2019; Accepted: 3 February 2019; Published: 7 February 2019

**Abstract:** Band gap magnitudes and valence band energies of Ta5+ containing simple perovskites (BaTaO2N, SrTaO2N, CaTaO2N, KTaO3, NaTaO3, and TaO2F) were studied by diffuse reflection absorbance measurements, density-functional theoretical calculations, and X-ray photoelectron spectroscopy. As a universal trend, the oxynitrides have wider valence bands and narrower band gaps than isostructural oxides, owing to the N 2*p* contribution to the electronic structure. Visible light-driven water splitting was achieved by using Pt-loaded CaTaO2N, together with a sacrificial agent CH3OH.

**Keywords:** perovskite oxynitride; band gap; density-functional theory; water splitting

#### **1. Introduction**

Transition metal oxynitrides are of interest due to their potential as photocatalysts [1–6], pigments [7,8], battery electrodes [9], high-permittivity dielectrics [10], etc. Such a diverse functionality of oxynitrides is derived largely from the coexistence of O2−/N3<sup>−</sup> in the anion lattice. As is well established, the conduction and valence bands of simple perovskites *AMX*<sup>3</sup> are based mostly on the frontier orbitals of *M* and *X*, respectively. In the case of oxide perovskites, O 2*p* orbitals participate in the valence band formation near the Fermi level. However, the inclusion of nitrogen, which brings a higher 2*p* orbital energy level than that of oxygen 2*p*, can effectively shift the top of the valence band upward resulting in the decreased band gap. It is interesting to note that if the energy difference between O 2*p* (−14.1 eV) and N 2*p* (−11.4 eV) orbitals [11] is reflected onto the valence band edge positions, many of the complex oxynitrides containing Ta5+, Nb5+, or Ti4+ would have band gaps falling in the visible light range (3.1~1.8 eV). Such a prospect in optical properties has motivated a number of studies on oxynitride perovskites and related phases, with views to semiconductor developments for visible light-harvesting photocatalysts or non-toxic inorganic pigments [1–8,12–18].

In 2001, Asahi et al. reported the visible light-driven photocatalytic activity of TiO2−*x*N*<sup>x</sup>* [1], which was followed by a number of studies on oxynitride-type photocatalysts. Promising photocatalysts were identified in various structure types such as simple perovskite (CaTaO2N, SrTaO2N, BaTaO2N, LaTiO2N, LaTaON2, CaNbO2N, SrNbO2N, BaNbO2N), complex perovskites (LaMg*x*Ta1−*<sup>x</sup>*O1+3*x*N2−3*x*), spinel (ZnGa2O*x*N*y*), wurtzite (Ga1−*x*Zn*x*N1−*x*O*x*), baddeleyite (TaON), and anosovite (Ta3N5) [3–6]. Among notable examples, Ga1−*x*Zn*x*N1−*x*O*<sup>x</sup>* has an absorption edge at ~500 nm and showed a quantum yield of 5.2% for 410 nm light [19]. BaTaO2N-BaZrO3 solid solution could catalyze H2 evolution from water without sacrificial agents [20]. TaON showed overall water splitting activity with surface modification and appropriate co-catalysts [21]. LaMg*x*Ta1−*<sup>x</sup>*O1+3*x*N2−3*<sup>x</sup>* and CaTaO2N could achieve overall water splitting with a Rh-Cr mixed oxide co-catalyst [22,23]. However, it was apparent that the photocatalytic behavior of a particular catalyst depended not merely on the composition but the morphology, defects, co-catalysts, and the type of photocatalytic reaction.

In this study, we compare the electronic structures of several Ta5+ perovskites having different anion matrices of pure oxide, oxynitride, and oxyfluoride types. The diffuse reflection absorbance spectra for *A*TaO2N (*A* = Ba, Sr, and Ca) are presented along with those of KTaO3, NaTaO3, and TaO2F, revealing a clear dependence of the semiconductor band gap on the electronegativity of anion components. The density-functional theory (DFT) based computations, combined with the valence level X-ray photoelectron spectroscopy (XPS), confirm that the N 2*p* component plays a critical role in extending the valence band edge in oxynitride compounds. We also present the photocatalytic activity of oxynitride samples tested by examining the water splitting under visible light irradiation.

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

Figure 1 displays the diffuse reflection absorbance spectra for simple Ta5+ perovskites where the oxynitride phases are found with markedly smaller band gap energies than the others. The optical band gaps were estimated by Shapiro's method [24]. The linear region of the absorption edge was extrapolated to the wavelength axis, where the intersection (zero absorption) was taken as the band gap value. The estimated band gap energies are in the following order: BaTaO2N (1.8 eV) < SrTaO2N (2.1 eV) < CaTaO2N (2.4 eV) < KTaO3 (3.6 eV) < NaTaO3 (4.0 eV) < TaO2F (4.1 eV).

**Figure 1.** Diffuse reflection absorption spectra for (**i**) BaTaO2N, (**ii**) SrTaO2N, (**iii**) CaTaO2N, (**iv**) KTaO3, (**v**) NaTaO3, and (**vi**) TaO2F.

For the *d*<sup>0</sup> perovskites *AMX*3, it has been well elucidated that the band gap magnitude depends on (i) electronegativity difference between *M* cation and *X* anion, (ii) deviation of *M*−*X*−*M* bond angles away from 180◦, (iii) *M*−*X* bond distance, and (iv) electronegativity of *A* cation [25,26]. The band gap variation among BaTaO2N (cubic), SrTaO2N (tetragonal), and CaTaO2N (orthorhombic) can be explained by the structural distortion factor (ii), in which the more distorted Ta−(O,N)−Ta linkage leads to the narrower band width and the wider band gap. However, the same reasoning cannot be used across distinct anion systems as the cubic KTaO3 has a greater band gap than that of CaTaO2N. In this regard, it can be judged that the control of anion components among N, O, and F, which have well-separated electronegativity values, makes a dominant effect on the resulting electronic structure. The absorbance spectra were also examined by using Tauc plots [27] from which the above oxynitride perovskites were found to be indirect-gap semiconductors.

A detailed aspect of the electronic structural evolution depending on the anion components was studied by band calculations at the DFT level and the XPS measurements. Structural parameters for DFT calculations were taken from the Rietveld refinements for BaTaO2N, SrTaO2N, and CaTaO2N [10], or from the literature data for KTaO3 [28], NaTaO3 [29], and TaO2F [30]. Since the computation codes

cannot handle mixed occupation of any crystallographic site, ordered O/N (or O/F) distributions were assumed for mixed anion phases.

The density of states (DOS) in BaTaO2N, SrTaO2N, CaTaO2N, KTaO3, and TaO2F resulted from the calculations using CAmbridge Serial Total Energy Package (CASTEP) and are compared in Figure 2. As previously observed for similar compounds, the computation tends to underestimate the band gap magnitude. Still, it can be well recognized that the width and position of valence bands vary depending on the anion components. Both of the mixed anion systems have widened valence bands due to the 2*p* orbital mixings between O/N or O/F: extended toward a higher energy side for oxynitrides and toward a lower energy side for oxyfluoride. However, the conduction bands of those five compounds were found at fairly similar energy ranges (not shown) since they have the same octahedral cation, Ta. The net result is the effective band gap reduction in oxynitrides, as compared with oxides. On the other hand, for the oxyfluoride derivative, the band gap itself would not change very much to a first approximation.

**Figure 2.** Valence band density of states (DOS) structures for BaTaO2N, SrTaO2N, CaTaO2N, KTaO3, and TaO2F as calculated using the CAmbridge Serial Total Energy Package (CASTEP) code.

The N 2*p* contribution to band structures of oxynitride compounds can be better viewed by extracting the partial DOS of the component atoms. Figure 3 shows the DOS plots for BaTaO2N as an example, which was obtained by employing linear muffin-tin orbital (LMTO) calculation. Both O *p* and N *p* orbitals were found as the major constituents of the valence band but notably the N character resides primarily at the upper region of the valence band, in agreement with the design concept of these oxynitride perovskites.

Along with the theoretical calculation, an experimental probe was also used to study the valence band structures of *A*TaO2N (*A* = Ba, Sr, Ca), KTaO3, NaTaO3, and TaO2F. The XPS spectra presented in Figure 4 were collected at near the Fermi level and, therefore, depict the DOS of valence bands. After the energy calibration using the C 1*s* peak energy and background subtraction, the tops of the valence bands were determined as indicated on the plots (Figure 4). The valence band edges of oxynitrides were found to be higher in energy by significant margins (≈1 eV) than the oxides' or oxyfluoride's, which is consistent with the electronic structure calculations. It is, therefore, corroborated both experimentally and theoretically that the hybridization of O 2*p* and N 2*p* orbitals are energetically feasible in the extended solid lattice, and that the partial N/O replacement can be a useful means to reduce the band gap size of oxide semiconductors. Based on the measured band gap magnitudes and the valence band widths, simplified band structures can be proposed for the Ta5+ perovskites studied here (Figure 5).

**Figure 3.** Total and partial DOS for BaTaO2N as calculated using linear muffin-tin orbital (LMTO) code.

**Figure 4.** Valence level XPS spectra for (**i**) BaTaO2N, (**ii**) SrTaO2N, (**iii**) CaTaO2N, (**iv**) KTaO3, (**v**) NaTaO3, and (**vi**) TaO2F. Vertical bar on each data indicates the top edge of the valence band.

**Figure 5.** Conduction (unfilled) and valence (shaded) band positions for several simple perovskites with octahedral Ta5+, as deduced from diffuse reflection absorbance and XPS measurements.

The reduced band gaps of oxynitride phases have immediate relevance to the photocatalytic reactivity. In this respect, we tested the water splitting by Pt-loaded oxynitride samples under visible light irradiation. Figure 6 presents the time-dependent H2 evolution from the Pt-CaTaO2N in H2O/CH3OH, along with the result from Pt-TiO2 (P25). Since CH3OH contains carbon with a formal oxidation number of −2, it can act as a reducing agent that removes O2 and expedite water decomposition as follows:

$$2\,\mathrm{H}\_{2}\mathrm{O} + h\mathrm{v} \to 2\,\mathrm{H}\_{2} + \mathrm{O}\_{2} \tag{1}$$

$$2\text{CH}\_3\text{OH} + \text{O}\_2 \rightarrow 2\text{CO}\_2 + 2\text{H}\_2\tag{2}$$

The sacrificial agent CH3OH should boost the generation of H2 according to the Le Chatelier principle, and also help suppress the H2−O2 recombination.

**Figure 6.** Photocatalytic H2 evolutions over Pt-loaded powders of CaTaO2N (filled circles) and P25 TiO2 (open circles). At the beginning and after 30 h had elapsed, the reactor vessel was purged with Ar.

As displayed in Figure 6, Pt-CaTaO2N possesses the photocatalytic activity that can be triggered by visible light photons. Using the irradiation source of λ > 395 nm here, the photocatalytic efficiency of Pt-CaTaO2N is significantly higher than that of Pt-TiO2 (P25), a well-established photocatalyst

system. Certainly, the superior performance of Pt-CaTaO2N is attributed to its narrower band gap. As can be found from Figure 1, CaTaO2N can utilize the photons with λ as long as ≈500 nm, whereas the absorption by TiO2 (P25) is limited to λ < 400 nm. The other oxynitride samples Pt-BaTaO2N and Pt-SrTaO2N and an oxide sample Pt-KTaO3 were also examined under the same experimental condition, but none of them produced discernible amounts of H2. The lack of photocatalytic ability in Pt-KTaO3 is simply ascribable to its wide band gap. However, in the cases of BaTaO2N and SrTaO2N, which possess even smaller band gap energies than CaTaO2N, the inferior photocatalytic property can be due to other factors. As one possibility, the valence band edges of BaTaO2N and SrTaO2N might be higher than the O2−/O2 oxidation level, or the H2−O2 recombination might occur so fast as to disallow the observation of water decomposition. Yet, BaTaO2N and SrTaO2N are regarded as promising candidates for visible light photocatalysts that could be well exploited in deliberately designed reaction systems. In the studies by Domen et al., it was demonstrated that the combination of Pt-*A*TaO2N (*A* = Ba, Sr, Ca) and Pt-WO3 achieves overall water splitting under visible light in the presence of IO3 −/I− as a shuttle redox mediator [2,12].

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

#### *3.1. Sample Syntheses and Crystal Structure*

Polycrystalline oxynitride samples *A*TaO2N (*A* = Ba, Sr, Ca) were prepared by ammonolysis reaction using BaCO3 (J. T. Baker, 99.8%, Phillipsburg, NJ, USA), SrCO3 (Aldrich, 99.9+%, St. Louis, MO, USA), CaCO3 (Mallinckrodt, 99.95%, Phillipsburg, NJ, USA), and Ta2O5 (Cerac, 99.5%, Milwaukee, WI, USA), as described previously [10]. Quantitative mixture of reagents was heated in anhydrous ammonia (99.99%) at a flow rate of ≈50 cm3/min. Each ammonolytic heating cycle consisted of heating/cooling ramps of 10 ◦C/min and a dwell step of 20 h at 1000 ◦C. The heat treatment cycle was repeated 2–5 times to obtain phase pure products. For preparing TaO2F, Ta powder (Alfa Aesar, 99.9%, Karlsruhe, Germany) was dissolved in HF solution (47%) in a Teflon beaker. After evaporating the solvent at 125 ◦C, white precipitate was washed with distilled water, dried at 150 ◦C, and finally heated at 500 ◦C in air for 1 h. Reference compounds KTaO3 (Cerac, 99.9%) and NaTaO3 (Cerac, 99.9%) were used as purchased.

Crystal structure analyses of *A*TaO2N samples used the synchrotron X-ray powder diffraction patterns collected at the beamline X7A of National Synchrotron Light Source, Brookhaven National Laboratory (Upton, NY, USA). Lattice parameters and atomic coordinates for *A*TaO2N phases were refined using the Rietveld method as incorporated in the GSAS-GUI software suite [31,32].

#### *3.2. Electronic Structure and Photocatalytic Property*

Diffuse reflectance data were recorded and converted to absorbance using a spectrophotometer (Perkin Elmer, Lambda 20, Waltham, MA, USA) equipped with a 50-mm Labsphere integrating sphere over the spectral range 200–900 nm. The band gap energies were determined from Shapiro's method [24] of extrapolating the onset of absorption to the wavelength axis.

DFT-based computations were performed using the CASTEP program as embodied in Accelrys Materials Studio [33]. Norm-conserving nonlocal pseudo-potentials were generated using the Kerker scheme with a kinetic energy cutoff of 400 eV. A convergence criterion of 0.02 meV was applied for the energy change per atom. Electron exchange and correlation were described using the Perdew-Wang generalized gradient approximation (PW91-GGA) [34]. For BaTaO2N, the total and partial densities of states were also calculated using a computation code, Stuttgart LMTO version 47, developed by Anderson and co-workers [35,36]. The program employs a TB-LMTO-ASA (tight binding linear muffin-tin orbital atomic sphere approximation) algorithm. Integrations over *k* space were performed using the tetrahedron method with a total of 40 irreducible *k* points from a 6 × 6 × 6 grid of reducible *k* points.

Valence band structures of *A*TaO2N (*A* = Ba, Sr, Ca), KTaO3, NaTaO3, and TaO2F were experimentally studied by XPS at near Fermi energy level, using a V. G. Scientific spectrometer equipped with a Mg *K*α source (1253.6 eV) and operated at 9 kV and 20 mA with a base pressure of ≈<sup>2</sup> × <sup>10</sup>−<sup>9</sup> Torr. Shirley method [37] was used for the data smoothening and background removal from the raw XPS spectra.

Photocatalytic activity of CaTaO2N, in comparison with that of TiO2 (Degussa P25) [38], was examined for the water decomposition using visible light irradiation. To focus on the photocatalytic H2 evolution, Pt was employed as a co-catalyst [39]. For preparing the Pt-impregnated catalyst, sample powder was stirred in an aqueous solution of H2PtCl6·6H2O ([Pt4+] ≈ 0.4 mM) under ultraviolet (UV) irradiation for 24 h, rinsed, and dried at room temperature. Thus, the obtained Pt-loaded catalyst (≈50 mg) was suspended in a mixture of 35 mL H2O and 0.6 mL MeOH contained in a 43.5 mL quartz vessel, which was sealed with a latex septum and filled with ≈1 atm of Ar. The photocatalytic reaction was induced by external illumination with an Oriel Xe lamp (24 V, 7 A) through a liquid filter and a long-pass filter (λcutoff = 395 nm), and was monitored using a gas chromatograph (Shimadzu, GC-14A, Tokyo, Japan) with Ar (99.998%) carrier gas. By using the liquid filter with a circulating water cooler, the reaction vessel was kept from the heating effect of infrared light component.

#### **4. Conclusions**

It was shown, using six simple perovskites with octahedral Ta5+, that the semiconductor band gap can be widely modulated by the electronegativity of anion components. The band gap generally widens from oxynitrides to oxides to oxyfluorides, and in most cases, the *d*<sup>0</sup> oxynitride phases have band gaps corresponding to visible light energy. Band structure calculations by the DFT method and XPS measurements indicate that the N 2*p* component contributes to extend the top of the valence band in oxynitrides, making a principal distinction from the oxides' electronic structures. The photocatalytic H2 generation from H2O was observed by using Pt-CaTaO2N and a sacrificial electron donor CH3OH under visible light.

**Author Contributions:** P.M.W. designed the research and reviewed the draft. Y.-I.K. prepared and characterized the samples and wrote the draft.

**Funding:** This research was funded by the National Research Foundation of Korea (NRF-2015R1D1A1A01056591) through the Basic Science Research Program.

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

#### **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/).

## *Review* **Photocatalytic Hydrogen Evolution via Water Splitting: A Short Review**

**Yifan Zhang 1, Young-Jung Heo 1, Ji-Won Lee 1, Jong-Hoon Lee 1, Johny Bajgai 2, Kyu-Jae Lee 2,\* and Soo-Jin Park 1,\***


Received: 23 October 2018; Accepted: 8 December 2018; Published: 12 December 2018

**Abstract:** Photocatalytic H2 generation via water splitting is increasingly gaining attention as a viable alternative for improving the performance of H2 production for solar energy conversion. Many methods were developed to enhance photocatalyst efficiency, primarily by modifying its morphology, crystallization, and electrical properties. Here, we summarize recent achievements in the synthesis and application of various photocatalysts. The rational design of novel photocatalysts was achieved using various strategies, and the applications of novel materials for H2 production are displayed herein. Meanwhile, the challenges and prospects for the future development of H2-producing photocatalysts are also summarized.

**Keywords:** photocatalysis; H2 generation; water splitting; solar energy

#### **1. Introduction**

The development of renewable green energy sources is a critical challenge for modern society. H2 is environmentally friendly, renewable, and considered to be an ideal candidate for an economically and socially sustainable fuel [1–6], and was previously regarded as an alternative energy source. Interestingly, some researchers also found that H2-rich water has neuron effects owing to its antioxidant properties. Although the deep mechanism is not clear, more and more researchers made an effort to study the biological function of H2 [7–21]. To date, almost all H2 gas production processes in the industry are based on natural gas, coal, petroleum, or water electrolysis. These traditional preparation methods are limited due to the associated CO2 emissions and high energy consumption. Hence, it is urgent to develop a low-cost method for efficient H2 generation and, thus, support the emerging H2 economy.

The sun provides an energy output of ~3 × <sup>10</sup><sup>24</sup> J per year, which is approximately 12,000 times higher than the current energy demand. Therefore, solar energy can act as a sustainable alternative energy source in the future. To date, the transformation of solar energy into H2 via water splitting is deemed as a desirable H2 preparation method to solve the energy crisis [22,23].

The proper use of H2 requires insight into the physical properties of H2 molecules. As we know, the lengths and strengths of hydrogen bonds are exquisitely sensitive to temperature and pressure. Meanwhile, the charges of H2 molecules also vary with temperature [24] because the spin direction of the nucleus in the H2 molecule changes depending on the temperature, and an energy difference occurs between H2 molecules. The *para*-H2 fraction changes with temperature, and it is necessary to understand the characteristics of H2 molecules according to temperature [25]. During

the reaction, hydrogen can be used safely at room temperature; however, it is rather dangerous in high-temperature environments.

As we know, H2 gas, often called dihydrogen or molecular H2, is a highly flammable gas with a wide range of concentrations between 4% and 75% by volume. Meanwhile, H2 is the world's lightest gas. The density of H2 is only 1/14 of that of air. At 0 ◦C, the density of H2 is only 0.0899 g/L at standard atmospheric pressure, which is the smallest-molecular-weight substance; it is mainly used as a reducing agent. The enthalpy of combustion is about −286 kJ/mol, which can be displayed by the following equation: 2H2(g) + O2(g) → 2H2O(l) + 572 kJ (286 kJ/mol). Currently, H2 is the main industrial raw material and the most important industrial gas. It has various applications in the petrochemical, electronic, and metallurgical industry, as well as in food processing, float glass, fine organic synthesis, aerospace, and other fields. At the same time, H2 is also an ideal secondary energy source. Owing to the properties of H2, the aerospace industry uses liquid H2 as fuel. Now, it is common to produce H2 from water gas rather than using high-energy-consuming water. The produced H2 is used in large quantities in the cracking reaction of the petrochemical industry and the production of ammonia. Unfortunately, all H2 production methods are highly energy (thermal and electrical) demanding, which limits their application. Thus, it is crucial to find a new method of H2 production.

Fujishima and Honda first reported photocatalytic water splitting using a TiO2 electrode in 1972 [26]. Research on solar H2 production attracted researchers in various fields, such as (1) chemists for the design and synthesis of various catalysts to investigate structure–property relationships; (2) physicists to fabricate semiconductor photocatalysts with novel electronic structures, as predicted by theoretical calculation; and (3) material scientists to construct unique photocatalytic materials with novel structures and morphologies [27–30]. When photocatalysts are illuminated at wavelengths which are suitable to their band gap energy, after the excitation, the charge carriers will either combine or transfer to the surface of the photocatalysts to participate in photocatalytic reactions. For the generation of efficient semiconductor photocatalysts, long-lived charge carriers and high stability are required [31–33].

Significant developments were made toward H2 generation via water splitting over the last several decades by a number of talented researchers [34–38].

Herein, we attempt to sum up the advances achieved to date. Therefore, we briefly summarize the background related to various photocatalysts for H2 generation and the achievements of high-efficiency photocatalysts. The main synthesis routes and modifications for adjusting the band structure to harvest light and enhance charge separation are also discussed.

#### **2. Principle of H2 Generation via Water Splitting**

In the pioneering study by Fujishima and Honda [27], electrochemical cells were made up for the splitting of the water into H2 and O2, as shown in Figure 1. While the TiO2 electrode was under ultraviolet (UV) light irradiation, water oxidation (oxygen evolution) occurred on its surface, while the reduction reaction (H2 evolution) occurred on the surface platinum black electrode. With this study in mind, semiconductor photocatalysts were later developed by Bard et al. in their design of a novel photocatalytic system.

**Figure 1.** Schematic of a photoelectrochemical cell (PEC). Reproduced with permission from Reference [26]; copyright (1972), Nature Publishing Group.

Figure 2a shows a display of hydrogen evolution by photocatalysts. The photocatalytic reaction occurring on the semiconductor photocatalysts can be divided into three parts: (1) obtaining photons with energy exceeding that of the photocatalyst's band gap, generating electron and hole pairs; (2) separating carriers by migration in the semiconductor photocatalyst; and (3) reaction between these carriers and H2O [39–46]. In addition, electron–hole pairs will combine with each other simultaneously. As shown in Figure 2b, while photocatalysts are involved in hydrogen evolution, the lowest position of the conduction band (CB) should be lower than the reduction position of H2O/H2, while the position of the valence band (VB) should be higher than the potential of H2O/O2 [47–50].

**Figure 2.** Schematic illustration of hydrogen evolution over photocatalysts. Reproduced with permission from Reference [39]; copyright (2014), Elsevier.

Various photocatalysts were reported to decompose water into H2 and O2 (Equation (1)). As we know, the hydrogen evolution reaction can be separated into two parts: oxidation for the evolution of O2 (Equation (2)) and water reduction to produce H2 (Equation (3)) [51–56]:

$$\begin{aligned} \,^1H\_2O \to 2H\_2 + O\_2 \end{aligned} \tag{1} \qquad \qquad \qquad \Delta E^0 = 1.23 \text{ V} \tag{1}$$

$$H\_2O \to 4H^+ + 4e^- + O\_2 \qquad\qquad\qquad E^0 = +1.23\ V\text{ vs.}\,NHE,\;pH = 0\tag{2}$$

$$4H^{+} + 4e^{-} \rightarrow 2H\_{2} \tag{3.1} \tag{3.2} \\ \text{and} \qquad \Delta E^{0} = 0 \text{ V vs.} \text{ N} \\ \text{HE, } pH = 0 \tag{3.3}$$

#### **3. Photocatalysts for Water Splitting**

Many photocatalysts were created as photocatalysts for hydrogen evolution. Based on these species, they can be divided into three major parts: (1) graphene-based photocatalyst; (2) graphitic carbon nitride (g-C3N4)-based photocatalysts; and (3) heterojunction photocatalysts (semiconductor–semiconductor or semiconductor–(metal, element)).

#### *3.1. Graphene-Based Photocatalysts*

Recently, graphene-based photocatalysts attracted significant attention for enhancing photocatalytic H2 production performance. Graphene is used to enhance photocatalytic efficiency owing to its novel structure and electrochemical properties (Figure 3).

**Figure 3.** Proposed mechanism of graphene-based photocatalysts. Reproduced with permission from Reference [56]; copyright (2013), American Chemical Society.

To date, many reports regarding the synthesis of graphene-based photocatalysts with improved photocatalytic efficiency were published. Graphene is a well-known two-dimensional (2D) material, which can improve surface area, and its 2D membrane-like structure imparts unique electrochemical properties [57–60]. Generally speaking, photocatalysts prepared by simple physical mixing with graphene will involve only a bit of direct contact with the graphene sheets. This small amount of contact between the photocatalyst and graphene results in weak interactions and inhibits charge transfer rates. Hence, the synthesis of photocatalysts with more interactions is highly needed.

Previously, Kim et al. synthesized novel graphene oxide (GO)-TiO2 photocatalysts [58] in 2013, comprising a core–shell nanostructure with enhanced photocatalytic efficiency (Figure 4). The improved H2 production activity compared to that of TiO2 revealed that the utilization of the core–shell structure enhanced photocatalytic efficiency. This novel structural design offers three-dimensional (3D) close contact between the materials and provides more active sites, which will enhance the charge separation rate and H2 production efficiency [61–63].

**Figure 4.** Schematic display of synthetic process of graphene oxide (GO)/TiO2 and TiO2/GO. Reproduced with permission from Reference [57]; copyright (2012), American Chemical Society.

Currently, many researchers are more interested in visible-light-driven photocatalysts, which are achieved using band-gap modification or taking graphene as a photosensitizer to broaden the visible-light adsorption range [64–66]. Significant efforts were conducted for building visible-light response systems because of the UV-only response of TiO2, and its nontoxic properties [67]. Recently, it was found that graphene regulating TiO2 involves visible-light adsorption activity. The carbon-layered structure of graphene with enriched π electrons forms bonds with titanium atoms. As a result, this strong interaction will shift the band position and reduce the band gap [68–70]. Lee et al. [71] also achieved a lower band gap using a graphene/TiO2 photocatalyst. The improved photocatalytic efficiency of the graphene/TiO2 composite owes to the band-gap regulation, which consequently promotes charge transfer rates through the graphene sheets.

#### *3.2. g-C3N4-Based Photocatalysts*

Currently, carbon-nitride-based photocatalysts receive significant attention for their photocatalytic H2 generation owing to a unique electronic structure (Figure 5) [72–77]. This section summarizes recent significant achievements in building C3N4-based photocatalysts for H2 evolution. Methods including nanostructure regulation, band-gap modification, dye sensitization, and heterojunction fabrication are highlighted herein.

**Figure 5.** Proposed mechanism of graphitic carbon nitride (g-C3N4)-based photocatalysts. Reproduced with permission from Reference [72]; copyright (2014), American Chemical Society.

Recently, carbon nitride attracted significant attention following the pioneering research of Wang et al. in 2009 for photocatalytic hydrogen evolution [78,79]. The assumed structure of C3N4 is a 2D framework with the tri-*s*-triazine linked by tertiary amines (Figure 6); it is thermally stable and chemically stable. Pioneering studies regarded g-C3N4 as a visible-light-driven phorocatalyst with a band gap of approximately 2.7 eV and an appropriate band position for water splitting [80–85]. Hence, g-C3N4 is an ideal candidate for photocatalytic H2 evolution.

**Figure 6.** Schematic display of the structure of g-C3N4. Reproduced with permission from Reference [72]; copyright (2014), American Chemical Society.

H2 generation performance using g-C3N4 can be promoted with noble-metal particles such as Au or Pd, which obtain electrons in the CB to inhibit the charge recombination rate [86–90]. Many researchers are developing metal-free photocatalysts for H2 evolution, and recent reports involved the introduction of non-noble-metal catalysts into g-C3N4 photocatalysts, displaying enhanced photocatalytic performance compared to noble-metal catalysts [91–95]. Hou et al. [86] synthesized MoS2/g-C3N4 composite photocatalysts (Figure 7) in 2018. MoS2/g-C3N4 increased the surface area and decreased the barrier when the electrons transported, thereby improving the charge transfer rate. The formation of band alignment enabled electron transfer from the CB (g-C3N4) to MoS2. Therefore, the MoS2/gC3N4 nanojunction significantly enhanced H2 evolution efficiency, achieving the highest H2 evolution rate and an optimum quantum efficiency of up to 2.1% (420 nm), which was higher than g-C3N4/Pt.

**Figure 7.** Schematic display of charge transfer on MoS2/g-C3N4 heterostructures during water splitting. Reproduced with permission from Reference [72]; copyright (2014), American Chemical Society.

#### *3.3. Metal-Loading-Based Photocatalysts*

Metal loading is also regarded as a useful method for photocatalytic enhancement. Song et al. [96] constructed Ag-rGO-TiO2 composite photocatalysts (Figure 8) in 2018. In order to analyze the photocatalytic mechanism of the architectural Ag-TiO2 and Ag-rGO-TiO2 composites, their structures with Ag nanocubes for light absorption and TiO2 nanosheets were well displayed. The difference between Ag-TiO2 and Ag-rGO-TiO2 is the interface between Ag nanocubes and TiO2 nanosheets, which enhances the electron transfer capability. For Ag-TiO2, the direct contact between the two materials results in the formation of Ag (100)/(001) TiO2 interface. Meanwhile, for Ag-rGO-TiO2, both Ag(100)/rGO and rGO/(001) TiO2 interfaces are formed by rGO. As mentioned above, the synergistic effect of Ag(100)/rGO and rGO/(001)TiO2 interfaces, rather than the Ag(100)/(001) TiO2 interface, offers quicker electron transfer. As shown in Figure 8, no Schottky barrier is formed between Ag and TiO2, and the hot electrons on the surface of TiO2 flow back to Ag and then recombine with holes. Meanwhile, for the Ag-rGO-TiO2 sample, no barrier is necessary to facilitate the electron transfer. The electrons generated on the surface of Ag nanocubes with smaller work function flow to rGO via a contact so as to equilibrate the electron Fermi distribution on the interface [97,98]. Moreover, the rGO nanosheets can act as conductive channels, further transferring the electron to the rGO/TiO2 interface. Owing to the light absorption of rGO, the transferred electrons within the rGO nanosheets can be further transferred to the CB of TiO2 under light excitation. The proposed photocatalytic mechanism of Ag-rGO-TiO2 is illustrated in Figure 8.

**Figure 8.** Schematic illustrating photocatalytic mechanism for Ag-TiO2 and Ag-rGO-TiO2 samples under visible-light irradiation. Reproduced with permission from Reference [95]; copyright (2018), Elsevier.

#### *3.4. Z-Scheme Photocatalysts*

An illustration of Z-scheme water splitting is shown in Figure 9. During an H2 evolution reaction, the reactions which happen on the surface of photocatalysts include the reduction of protons by CB electrons and the oxidation of an electron donor (D) by VB holes, yielding the corresponding electron acceptor (A), as follows:

$$2H^{+} + 2e^{-} \rightarrow H\_{2} \text{ (photareduction of } H^{+} \text{ to } H\_{2}\text{)}$$

$$D + nh^{+} \rightarrow A \text{ (photonicization of } D \text{ to } A\text{)}$$

On the other hand, the forward reactions on an O2 evolution photocatalyst are as follows:

$$A \,\, + \,\, n\text{e}^- \to \,\, D \,\, (photareduction\,\, of \,\, A \,\, to \,\, D)$$

$$2H\_2O \,\, + \,\, 4h^+ \to \,\, O\_2 \,\, + \,\, 4H^+ \,\, ((photonicization\,\, of \,\, H\_2O \,\, to \,\, O\_2))$$

where the electron acceptor generated by the paired H2 evolution photocatalyst is converted to D, and the water oxidation process occurs via the valence band holes. Thus, the water-splitting process can be achieved.

**Figure 9.** Diagram of photocatalytic water splitting using a Z-Scheme system. Reproduced with permission from Reference [98]; Copyright (2010), American Chemical Society.

Amal et al. reported a Z-scheme photocatalytic water-splitting system using Ru/SrTiO3 and partially reduced GO (PRGO)/BiVO4 (Figure 10) in 2011 [100]. As described in the report, the PRGO/BiVO4 (O2 photocatalyst) and Ru/SrTiO3:Rh (H2 photocatalyst) were attached due to surface charge modification in acidic conditions, as depicted in Figure 10. Under irradiation, electrons are excited from the VB (BiVO4) or an impurity level in Rh (Ru/SrTiO3:Rh) to the CB. We can indicate that the PRGO does not contribute to the electron and hole generation. In other words, the RGO in this work acts as an electron conductor. PRGO transfers the electrons from the CB of BiVO4 to the Ru/SrTiO3:Rh. Meanwhile, the electrons in Ru/SrTiO3:Rh reduce the water to H2 on the surface of the Ru co-catalyst, while the holes left in BiVO4 oxidize the water to O2. Additionally, the PRGO provides a pathway for photogenerated electrons in the BiVO4 photocatalyst. Each reaction can migrate as follows: reduction of water, transfer of electrons to PRGO, and transfer of holes to PRGO for oxidation. Because the majority of the photocatalyst surface is surrounded by water and only relatively small portions are in contact with PRGO [101], most electrons in Ru/SrTiO3:Rh and holes in BiVO4 are used for water splitting.

**Figure 10.** (**a**) Schematic display of a suspension of Ru/SrTiO3 and partially reduced GO (PRGO)/BiVO4 in water. (**b**) Mechanism of water splitting using Z-scheme system consisting of Ru/SrTiO3 and PRGO/BiVO4 under irradiation. Reproduced with permission from Reference [99]; copyright (2018), American Chemical Society.

#### *3.5. Defect Engineering Photocatalyst*

Among the various photocatalyst designs, the defect engineering strategy is regarded as an important way of modifying the photocatalysts. Defects are places where the atoms or molecules in the materials are disrupted, and they greatly influence photocatalytic performance. The defects in the lattice of photocatalysts not only act as an electron–hole recombination center, but also break the electronic structure and display a scattering center for electron and hole travel. Nevertheless, the positive effect of defects in photocatalytic performance enhancement were also recognized with the development of defect photocatalysts and the development of the photocatalytic field.

Chen et al. reported the synthesis of a bismuth subcarbonate (Bi2O2CO3, BOC) with controllable defect density (BOC-X) (Figure 11) in 2018. The BOC-X with defect density displayed a photocatalytic nitrogen fixation of 957 <sup>μ</sup>mol·L–1 under irradiation within 4 h, which was 9.4 times higher than that of pristine BOC. This photocatalytic performance enhancement of BOC-X can be attributed to the surface defects. These defects contribute to the defect levels in the forbidden band, which improves the light harvest percentage. Meanwhile, surface defects can also inhibit the electron–hole recombination rate to promote the separation efficiency of charge carriers. Photocatalytic nitrogen fixation by BOC-X is displayed in Figure 11. If the light energy is higher than the band-gap energy, the electrons on the VB surface of BOC-X are transferred to the CB and react with N2 to form NH3. Moreover, some of the VB electrons are transferred to the defect level and then react with N2. However, if the light energy is lower than the band-gap energy, the electrons of BOC-X are also excited from VB to the defect level and then participate in the reaction. Defects modulate the band gap of BOC-X and improve the light absorption range, thereby enhancing the carrier transport, and leading to photocatalytic enhancement.

**Figure 11.** Mechanism of photocatalytic nitrogen fixation on defective Bi2O2CO3. Reproduced with permission from Reference [101]; copyright (2010), American Chemical Society.

#### *3.6. Heterojunction Photocatalysts*

During the H2 evolution reaction, the formed electron–hole charges are transferred to the surface of the photocatalyst for the next step of the reaction or recombine with each other [102–106]. To better reveal this point, we assumed it as a simple case [107]: the influence of gravity on a man jumping (Figure 12a,b). When a man (electron) jumps from the ground (VB) to the sky (CB), it can return to the floor immediately (recombination of the electron and hole) owing to gravity. In order to let the people rise off the floor (separation of the charge carrier pairs), an instrument (semiconductor B) can be used (Figure 12c,d). Subsequently, the previously mentioned people can drop to the instrument rather than the ground (inhibition of the electron and hole pair recombination). Although the inhibition of electron–hole recombination rate is an urgent issue, it can be achieved via suitable construction of materials. Many methods were conducted to achieve better electron–hole pair separation rate, such as element combining [108,109], metal doping [110,111], or the use of heterojunctions [112,113]. Among these strategies, heterojunctions were proven to be the most desirable method for achieving efficient photocatalysis due to their improved separation ability of electron–hole pairs (Figure 12d).

**Figure 12.** Schematic display of (**a**) the influence of gravity on a person jumping, (**b**) electron–hole pair combination using a photocatalyst, (**c**) utilization of a stool to keep the person from returning to the ground, (**d**) electron–hole pairs separated in composite catalyst. Reproduced with permission from Reference [105]; copyright (2010), John Wiley & Sons, Inc.

A heterojunction is regarded as the connection between two kinds of photocatalysts with different band structures, which leads to a new band arrangement [114,115]. Generally, three kinds of composite photocatalysts are developed (Figure 13). As shown in Figure 13a, the CB and VB of A are a bit over and under the band position of B, respectively [116]. As a result, when the light irradiates, the generated electrons and holes are transferred to the CB and VB of B. Because the generated electrons and holes move to the same photocatalyst, the recombination rate of electron–hole pairs is not efficiently inhibited. The photocatalytic process happens on photocatalyst B with a mild potential requirement; thus, the photocatalytic ability of the photocatalyst using this heterojunction will be lower than others. As dispalyed in Figure 13b, the band positions of CB and VB are over that of photocatalyst B. Hence, during the photocatalytic reaction, the generated electron moves to photocatalyst B, while the holes are transferred to photocatalyst A, which leads to the formation of long lived electron–hole pairs [117–119]. Parallel to Figure 13a, the photocatalytic performance of the type-II composite photocatalysts is inhibited by the redox process occurring on B. Meanwhile, as displayed in Figure 13c, the band structure of type-III composite photocatalysts is parallel to type II, apart from the interlaced gap changing into non-overlapping band gaps [120,121]. Thus, the generated electron–hole pairs cannot be transferred between the two photocatalysts, resulting in them being inappropriate for long lived electron–hole pair separation. We can determine that the type-II heterojunctions are desirable for enhancing redox ability due to their optimum structure for long-lived electron–hole separation. In previous reports, great efforts were conducted to synthesize type-II composite photocatalysts, including g-C3N4/TiO2 [122], WO3/BiVO4 [123], WO3/g-C3N4 [124], and BiPO4/g-C3N4 [125].

**Figure 13.** Schematic display of three kinds of electron–hole pair separation among composite photocatalysts: (**a**) type-I, (**b**) type-II, (**c**) type-III heterojunctions. Reproduced with permission from Reference [105]; copyright (2017), John Wiley & Sons, Inc.

Yu et al. designed CdS/NiS composites photocatalysts using various heterojunctions in 2012, which greatly enhanced the hydrogen evolution performance. As shown in Figure 14a, around 20 nm of NiS particles were loaded onto the CdS uniformly, which supported a close connection between CdS and NiS. The formation of p–n heterojunctions facilitates charge transfer between the NiS and CdS, and inhibits charge-carrier recombination (Figure 14b,c). We can see that the holes left on the n-type catalyst are transferred to the p-type catalyst, providing a negative specie. The electron–hole pair distribution keeps moving until a Fermi-level equilibrium is achieved [126–128]. The generated active species move through the internal electric field of the composite photocatalysts, resulting in long-lived electron–hole pair separation rates. Thus, the electron–hole recombination rate is efficiently inhibited owing to the synergistic effect between the two photocatalysts. The photocatalytic H2 production rate over CdS/NiS composite photocatalysts with 5 wt.% NiS was found to be higher than that of the CdS and 1 wt.% Pt/CdS (Figure 14d). More NiS doping resulted in a reduction in photocatalytic efficiency due to NiS catalysts reducing the number of redox sites during the reaction.

**Figure 14.** (**a**) SEM image of CdS/NiS composite catalysts; (**b**,**c**) illustration of electron–hole pairs with CdS/NiS composite photocatalysts; (**d**) contrast of photocatalytic efficiency of CdS with different NiS content. Reproduced with permission from Reference [105]; copyright (2017), John Wiley & Sons, Inc.

#### **4. Summary and Perspectives**

Over the last several decades, photocatalysis was shown to be a promising method for H2 production. Even though the principles controlling photocatalytic activity in the developed semiconductors were identified, several aspects remain unclear. Therefore, practical applications and the commercialization of photocatalytic H2 production require further research. Meanwhile, the charge transfer among photocatalysts due to the influence of structure and electrochemical properties is also not very clear, while the influence of various preparation methods on the catalytic performance is not well understood. The development of improved photocatalysts will benefit from advances in science. Improved building of novel co-catalysts will arise from using efficient catalysts. Many researches are underway investigating new synthesis methods for sample preparation and novel system construction. Herein, we concluded the most prominent achievements associated with H2 production via photocatalysis. We hope this report will assist further research efforts regarding the development of photocatalysts.

**Funding:** This research was supported by the Korea Evaluation Institute of Industrial Technology (KEIT) through the Carbon Cluster Construction project [10083586, Development of petroleum-based graphite fibers with ultra-high thermal conductivity] funded by the Ministry of Trade, Industry, & Energy (MOTIE, Korea), and the Commercialization Promotion Agency for R&D Outcomes (COMPA) funded by the Ministry of Science and ICT (MSIT) [2018\_RND\_002\_0064, Development of 800 mA·h·g−<sup>1</sup> pitch carbon coating materials].

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

#### **References**


© 2018 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/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Catalysts* Editorial Office E-mail: catalysts@mdpi.com www.mdpi.com/journal/catalysts

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18