Mixed-Phase (2H and 1T) MoS 2 Catalyst for a Highly Efﬁcient and Stable Si Photocathode

: We describe the direct formation of mixed-phase (1T and 2H) MoS 2 layers on Si as a photocathode via atomic layer deposition (ALD) for application in the photoelectrochemical (PEC) reduction of water to hydrogen. Without typical series-metal interfaces between Si and MoS 2 , our p -Si/SiO x /MoS 2 photocathode showed efﬁcient and stable operation in hydrogen evolution reactions (HERs). The resulting performance could be explained by spatially genuine device architectures in three dimensions (i.e., laterally homo and vertically heterojunction structures). The ALD-grown MoS 2 overlayer with the mixed-phase 1T and 2H homojunction passivates light absorber and surface states and functions as a monolithic structure for effective charge transport within MoS 2 . It is also beneﬁcial in the operation of p - i - n heterojunctions with inhomogeneous barrier heights due to the presence of mixed-phase cocatalysts. The effective barrier heights reached up to 0.8 eV with optimized MoS 2 thicknesses, leading to a 670 mV photovoltage enhancement without employing buried Si p - n junctions. The fast-transient behaviors via light illumination show that the mixed-phase layered chalcogenides can serve as efﬁcient cocatalysts by depinning the Fermi levels at the interfaces. A long-term operation of ~70 h was also demonstrated in a 0.5 M H 2 SO 4 solution.


Introduction
The photoelectrochemical (PEC) splitting of water into oxygen and hydrogen offers green fuel from solar energy. It requires a semiconductor to absorb photons of visible light and to generate excitons, fast charge separation within the depletion layer followed by efficient charge transfer to the electrolyte. Employing thin layers of cocatalysts onto semiconductor surfaces can alter the surface energetics by bending the degree of energy bands and/or charge transfer kinetics [1]. By choosing suitable catalytic systems, moreover, the semiconductor can be protected from the solution environment [2]. Silicon (Si) has been spotlighted as a promising earth-abundant light absorber due to its small band gap of 1.12 eV and its viability in the electronic device industry, especially in the field of solar cells. Since the available photovoltage obtainable from a single Si junction is about 0.5 V and not enough for water splitting, which requires 1.23 V of minimum thermodynamic potential, additional bias should be required during the PEC operation. In the unbiased PEC configurations, therefore, multiple Si-based solar cells and tandem structures with wide-gap semiconductors are needed to boost the water splitting reactions [3]. In this study, we focus on the properties of photocathode as half-cell in the three-electrode configuration. To adapt Si as Recently, we developed an ALD chemistry that can directly produce as-grown crystalline MoS 2 thin films on various substrates at low temperatures (250-300 • C) using inexpensive precursors (i.e., MoCl 5 + H 2 S). ALD techniques have shown an exceptional capability for studying the critical thickness required for optimal HER operation [29]. Here, we describe that mixed-phase (1T and 2H) MoS 2 layers can be formed on Si as photocathodes via ALD for the PEC reduction of water. Note that we did not employ the extra phase-transition procedure from semiconducting 2H to metallic 1T phase. Without utilizing conventional series-metal interfaces, our simple p-Si/SiO x /MoS 2 photocathode showed an efficient and stable operation during HERs. The present devices have spatially genuine architectures in three dimensions (i.e., laterally homo and vertically heterojunction structures). The effective barrier heights reached 0.6 eV with optimized MoS 2 thicknesses, leading to a V ph enhancement of 670 mV without employing buried Si p-n junctions. A long-term operation of 70 h was also demonstrated. We suggest that in general, 1T-2H mixed-phase layered chalcogenides could serve as efficient cocatalysts by depinning Fermi levels at the interfaces, resulting in efficient electron transfer mechanisms.

Results and Discussion
The unique device configuration of our p-Si/SiO x /MoS 2 photocathode contrasts typical MIS photocathodes reported by other researchers (Figure 1). The electron-hole pairs are generated in the depletion region, mainly inside p-Si, and then transferred through the tunnel oxide to the thin metal interfaces toward cocatalysts in the electrolyte. The monolithic homojunctions consist of mixed-phase 1T and 2H MoS 2 that develop during the growth of MoS 2 via ALD. Depending on the degree of distortion in the S-Mo-S atomic planes, 2H and 1T phases exhibit different structural and electronic properties. The 2H phase has semiconducting properties and is more stable compared to other phases (e.g., 3R and 1T), whereas the 1T phase is metallic and unstable. A key mechanism may be the inclusion of an excess chlorine moiety along the MoS 2 basal plane upon incomplete reactions at low temperatures as probed by elemental analysis, X-ray photoelectron spectroscopy, and X-ray diffraction patterns [30]. We also demonstrated that as a result of the presence of Cl, the 1T phase was stable upon thermal annealing at 400 • C. Since the 1T phase of MoS 2 is metallic, the resulting MoS 2 layers as cocatalysts on the Si photocathode exhibit distinctively monolithic homojunctions with changes in the Φ B laterally along the interfaces, as illustrated in Figure 1b. Our p-Si/SiO x /MoS 2 photocathode makes the electron transfer pathway effective compared to a conventional MIS electrode, eliminating additional processes that are required for complicated interface engineering. Note that additional phase-conversion (2H-to-1T) and film-transfer processes were not necessary, thanks to the direct growth of the mixed-phase films on to the observer. MoCl5 + H2S). ALD techniques have shown an exceptional capability for studying the critical thickness required for optimal HER operation [29]. Here, we describe that mixed-phase (1T and 2H) MoS2 layers can be formed on Si as photocathodes via ALD for the PEC reduction of water. Note that we did not employ the extra phase-transition procedure from semiconducting 2H to metallic 1T phase. Without utilizing conventional series-metal interfaces, our simple p-Si/SiOx/MoS2 photocathode showed an efficient and stable operation during HERs. The present devices have spatially genuine architectures in three dimensions (i.e., laterally homo and vertically heterojunction structures). The effective barrier heights reached 0.6 eV with optimized MoS2 thicknesses, leading to a Vph enhancement of 670 mV without employing buried Si p-n junctions. A long-term operation of ~70 h was also demonstrated. We suggest that in general, 1T-2H mixed-phase layered chalcogenides could serve as efficient cocatalysts by depinning Fermi levels at the interfaces, resulting in efficient electron transfer mechanisms.

Results and Discussion
The unique device configuration of our p-Si/SiOx/MoS2 photocathode contrasts typical MIS photocathodes reported by other researchers (Figure 1). The electron-hole pairs are generated in the depletion region, mainly inside p-Si, and then transferred through the tunnel oxide to the thin metal interfaces toward cocatalysts in the electrolyte. The monolithic homojunctions consist of mixed-phase 1T and 2H MoS2 that develop during the growth of MoS2 via ALD. Depending on the degree of distortion in the S-Mo-S atomic planes, 2H and 1T phases exhibit different structural and electronic properties. The 2H phase has semiconducting properties and is more stable compared to other phases (e.g., 3R and 1T), whereas the 1T phase is metallic and unstable. A key mechanism may be the inclusion of an excess chlorine moiety along the MoS2 basal plane upon incomplete reactions at low temperatures as probed by elemental analysis, X-ray photoelectron spectroscopy, and X-ray diffraction patterns [30]. We also demonstrated that as a result of the presence of Cl, the 1T phase was stable upon thermal annealing at 400 °C. Since the 1T phase of MoS2 is metallic, the resulting MoS2 layers as cocatalysts on the Si photocathode exhibit distinctively monolithic homojunctions with changes in the Φ laterally along the interfaces, as illustrated in Figure 1b. Our p-Si/SiOx/MoS2 photocathode makes the electron transfer pathway effective compared to a conventional MIS electrode, eliminating additional processes that are required for complicated interface engineering. Note that additional phase-conversion (2H-to-1T) and film-transfer processes were not necessary, thanks to the direct growth of the mixed-phase films on to the observer.  The overall morphologies (Figure 2c-e) and the resulting thicknesses (Figure 2a,b,f,g,k) of MoS 2 on Si surfaces were reproducible as previously reported [29]. The layered-structural nature was shown by the Raman spectra (Figure 2j), and the mixed-phase (1T and 2H) structures were also observed via XPS spectra (Figure 2h,i). Initial cycles up to 100 ALD cycles resulted in several MoS 2 nanoflakes (diameter of~10 nm). Most of the nuclei layers have parallel basal planes to the substrate surfaces, as shown in Figure 2a,f. The TEM of 300-cycled MoS 2 shows more vertically grown MoS 2 layers on the p-Si, as shown in Figure 2f. This growth mechanism was observed in thicker layers (1000 cycles in Figure 2g) and was identical to earlier work by some of the authors [29]. The overall morphologies (Figure 2c-e) and the resulting thicknesses (Figure 2a,b,f,g,k) of MoS2 on Si surfaces were reproducible as previously reported [29]. The layered-structural nature was shown by the Raman spectra (Figure 2j), and the mixed-phase (1T and 2H) structures were also observed via XPS spectra (Figure 2h,i). Initial cycles up to 100 ALD cycles resulted in several MoS2 nanoflakes (diameter of ~10 nm). Most of the nuclei layers have parallel basal planes to the substrate surfaces, as shown in Figure 2a,f. The TEM of 300-cycled MoS2 shows more vertically grown MoS2 layers on the p-Si, as shown in Figure 2f. This growth mechanism was observed in thicker layers (1000 cycles in Figure 2g) and was identical to earlier work by some of the authors [29].  We investigated the HER performance of our p-Si/SiO x /MoS 2 photocathodes as a function of the MoS 2 thickness under illumination (100 mW·cm −2 ), as shown in Figure 3. For a bare p-type silicon treated with HF, a photoelectrochemical reduction process begins at an onset potential of −0.12 V (which is defined when the photocurrent reaches 0.1 mA·cm −2 ) with a J ph of 24 ± 2 mA·cm −2 , which is consistent with other studies [31]. The corresponding measured V ph is shown in Figure 3b, which is 0.11 ± 0.04 V and is much less than the theoretical maximum V ph of Si of 0.48 to 0.5 V [32]. The gradual increase in J ph near the turn-on potential (shown in Figure 3a (solid black line)) indicates the slow kinetics of the charge transfer processes. The decreased V ph and the kinetics are attributed to the presence of surface states, followed by Fermi-level pinning (hereafter called FLP ss ) [33,34] at the surface/electrolyte interface and by the build-up of a surface oxide layer [23] and/or by electron traps by the complex surface chemistries that are associated with the formation of Si-OH bonds [35]. The photocathode produced via a few tens of ALD cycles of MoS 2 (data now shown) deposited on p-Si, which corresponds to less than 5 nm in thickness, showed negligible improvements in the PEC performance. This is partly due to the incomplete coverage of the silicon surface area since, from the initial ALD cycles, a number of seed layers of less than a few nanometer-thick nanoflakes randomly grow on Si. The photocathode produced with 100 ALD cycles, which is~7 nm in average thickness, achieved an enhanced onset potential of 0.27 ± 0.02 V with an increased J ph of 27.5 ± 0.5 mA·cm −2 . As shown in Figure 3b, the difference in the V ph from the bare p-Si corresponds to the anodic onset voltage shift value of the 100-cycle ALD photocathode. The increase in V ph is attributed to the surface states being reduced by the MoS 2 layers. The increase in a J ph compared to bare p-Si is due to the reduced surface recombination, which is one of the limiting factors for photocurrents [36].
As the number of ALD cycle increased to 300, an onset potential shift to 0.24 ± 0.03 V was observed with an increase in the V ph of 0.48 ± 0.01 V. As explained in our previous studies, after a certain number of flakes, additional MoS 2 layers grow in the vertical direction, covering the remaining exposed silicon surface to passivate from the electrolyte. Consequently, the Fermi level effect was almost alleviated, as V ph reached the maximum value for p-Si. The maximum V ph value of 0.80 to 0.67 V, which is much higher than the maximum V ph limit (V 0 ) of p-Si and the J ph of 30 mA·cm −2 , was attained at 500 ALD cycles (30 to 35 nm in average thickness). This is mainly attributed to the formation of a p-i-n heterojunction, including MIS between the p-Si/1T-MoS 2 partially within the bulk films, which will be discussed later in more detail.
The kinetic improvement that was partially due to an exposure of the edge sites is depicted in Figure 3c. The Tafel slope variations of 56 to 124 mV·decade −1 with the function of the number of ALD cycles were obtained by fitting the Tafel plots. Note that the Tafel slope of the 500-ALD-cycled photocathode, 90 mV·decade −1 , was larger than that of the 300-ALD-cycled photocathode, which is 56 mV·decade −1 ; 56 mV·decade −1 for 300-ALD-cycle photocathode indicates the faster charge transfer rate. This value is in between the value of 2H-MoS 2 nanodots (61 mV·decade −1 ) [37] and the value of 1T-MoS 2 (43 mV·decade −1 ) [29] in terms of HER as electrocatalysts.
Our results revealed that the PEC performance with the 500-ALD-cycles photocathode does not necessarily exhibit the highest charge transfer kinetic property. Rather, surface energetics-i.e., a capability of electron/hole charge carrier's separation at near photoelectrodes' interface changed by the presence of a catalyst-can also contribute the PEC performances [38]. Increasing the number of ALD cycles from 600 to 800 degrades the PEC performance and the kinetics as shown in Figure 3a-c, which mostly came from the reduced light transmittance through the thicker MoS 2 layers. As mentioned in previous studies [39,40], the light intensity is necessary to obtain a surface V ph , which turned out to be a function of the penetration depth. As the thickness increases to more than 40 nm, the penetration of light degradation induces a lower generation rate of electron-hole pairs. After 1000 ALD cycles, a 0.26 ± 0.01 V V ph , which is similar or even less than for bare p-Si, was absorbed. An increased electrical and interfacial resistance degrades the PEC performance with the V on voltage of −0.17 V was observed. The summary of important parameters for HER in PECs is shown in Table 1.
Note that our observation of the reduced PEC performance in thicker Cl-doped MoS 2 differs from recent works with MoS x Cl y grown by chemical vapor deposition. −0.17 V was observed. The summary of important parameters for HER in PECs is shown in Table 1. Note that our observation of the reduced PEC performance in thicker Cl-doped MoS2 differs from recent works with MoSxCly grown by chemical vapor deposition.  To understand the junction characteristics further, we carried out the Mott-Schottky C-V analysis of our p-Si/SiOx/MoS2 photocathode (Figure 3d). Generally, C-V measurement can be a reliable method for the determination of built-in voltage (Vbi) and the doping density of p-n junction for Schottky junctions interface [41,42]. The C-V characteristics of 100-and 500-ALD-cycle photocathodes have been measured in the dark. Dotted lines are fitted to the experimental data. Vbi  To understand the junction characteristics further, we carried out the Mott-Schottky C-V analysis of our p-Si/SiO x /MoS 2 photocathode (Figure 3d). Generally, C-V measurement can be a reliable method for the determination of built-in voltage (V bi ) and the doping density of p-n junction for Schottky junctions interface [41,42]. The C-V characteristics of 100-and 500-ALD-cycle photocathodes have been measured in the dark. Dotted lines are fitted to the experimental data. V bi of the photocathodes can be estimated from the famous Mott-Schottky equation, where the intercepts of the straight lines yield 0.36 and 0.80 V at 200 kHz for 100-ALD cycle and 500-ALD-cycle photocathode, respectively. The slopes also indicate the photocathode of p-type conductivity. Kenny et al. pointed out the catalyst layer thickness dependence on the MIS junction characteristics utilizing partial screening charges and Debye length [43]. Another report by Fujii et al. stated that the thickness of the deposited semiconducting layer on another semiconductor of p-n heterojunction affects the depletion width, thus resulting in the V bi variation [44]. The change in the build-up of the depletion region at the p-Si/MoS 2 (100-and 500-ALD cycles) interface junctions results in the increase of V bi .
Indeed, our p-Si/SiO x /MoS 2 photocathodes regulated the overall energy barriers and impedances of the total junctions. Accordingly, the stable operation under 1 sun illumination at a −0.3 V reverse bias and 0 V vs. RHE is shown in Figure 4a. The stable operation is due to the secure chemical contact formed between the p-Si/MoS 2 interface due to the ALD deposition method, which conformally deposited the layer on the silicon surfaces. Ding et al. experimentally confirmed the effect of a higher quality interface for 1T-MoS 2 /Si by comparing the PEC characteristics of drop-casted 1T and CVD-grown 1T MoS 2 films, showing improved performance with a 0.235 V vs. RHE onset potential for the CVD-grown MoS 2 film on Si [45]. The metallic nanojunctions in our mixed-phase (1T and 2H) MoS 2 layers grown via ALD improved the stability for HER by reducing the contact resistance to Si. Upon stable operation of the p-Si/SiO x /MoS 2 photocathode for 70 h under light, the resulting samples were analyzed by XRD, as shown in Figure 4b. Accordingly, the overall structures remained unchanged.  [43]. Another report by Fujii et al. stated that the thickness of the deposited semiconducting layer on another semiconductor of p-n heterojunction affects the depletion width, thus resulting in the Vbi variation [44]. The change in the build-up of the depletion region at the p-Si/MoS2 (100-and 500-ALD cycles) interface junctions results in the increase of Vbi. Indeed, our p-Si/SiOx/MoS2 photocathodes regulated the overall energy barriers and impedances of the total junctions. Accordingly, the stable operation under 1 sun illumination at a −0.3 V reverse bias and 0 V vs. RHE is shown in Figure 4a. The stable operation is due to the secure chemical contact formed between the p-Si/MoS2 interface due to the ALD deposition method, which conformally deposited the layer on the silicon surfaces. Ding et al. experimentally confirmed the effect of a higher quality interface for 1T-MoS2/Si by comparing the PEC characteristics of drop-casted 1T and CVDgrown 1T MoS2 films, showing improved performance with a 0.235 V vs. RHE onset potential for the CVD-grown MoS2 film on Si [45]. The metallic nanojunctions in our mixed-phase (1T and 2H) MoS2 layers grown via ALD improved the stability for HER by reducing the contact resistance to Si. Upon stable operation of the p-Si/SiOx/MoS2 photocathode for 70 h under light, the resulting samples were analyzed by XRD, as shown in Figure 4b. Accordingly, the overall structures remained unchanged. While the overall electronic structure of the photocathode can be depicted as a p-i-n heterojunction, inside the MoS2 film, the islands of embedded 1T phase nanostructures form a Schottky junction with p-Si. In Figure 5, the band alignments of p-Si/SiOx/MoS2 and the electrolyte in the nanoscale domain are drawn to explain an injected electron transport path that is energetically favorable, where Vo denotes the bend bending at 0 bias, Vr is the applied reverse bias, is the Schottky barrier height, Vph is the photovoltage, EF is the Fermi level, and EFn is the quasi-Fermi level for electrons. For the band alignment at p-Si/MoS2 interface, all the necessary parameters in terms of vacuum level and normal hydrogen electrode level in electrolyte of pH = 0.3 are shown in Figure S1, where the p-Si doping level is ~1.6 × 10 15 cm −3 . The injected electron passes from 1T-MoS2 to 2H-MoS2, considering that 1T-MoS2 exhibits a work function of approximately 4.2 eV, [46] which is not much different from that of the electron affinity of 2H MoS2 (4.3 eV), and higher electrical conductivity compared with the 2H phase. Bai et al. studied the 1T/2H MoS2 contact with different contact types and concluded that the edge-contact model exhibits a low tunneling barrier of 0.1 eV and even Ohmic contact in the case of having excess in-plane dangling bonds at the edge of 2H MoS2 [47]. An important interpretation from the J-V characteristics of the PEC performance is that the conditions for bulk junction properties are fulfilled after 500 ALD cycles and the MoS2 layer growth is already in the vertical direction. Therefore, in a situation of a high portion of 1T phase with an ~10 cm While the overall electronic structure of the photocathode can be depicted as a p-i-n heterojunction, inside the MoS 2 film, the islands of embedded 1T phase nanostructures form a Schottky junction with p-Si. In Figure 5, the band alignments of p-Si/SiO x /MoS 2 and the electrolyte in the nanoscale domain are drawn to explain an injected electron transport path that is energetically favorable, where V o denotes the bend bending at 0 bias, V r is the applied reverse bias, φ B is the Schottky barrier height, V ph is the photovoltage, E F is the Fermi level, and E Fn is the quasi-Fermi level for electrons. For the band alignment at p-Si/MoS 2 interface, all the necessary parameters in terms of vacuum level and normal hydrogen electrode level in electrolyte of pH = 0.3 are shown in Figure S1, where the p-Si doping level is~1.6 × 10 15 cm −3 . The injected electron passes from 1T-MoS 2 to 2H-MoS 2 , considering that 1T-MoS 2 exhibits a work function of approximately 4.2 eV, [46] which is not much different from that of the electron affinity of 2H MoS 2 (4.3 eV), and higher electrical conductivity compared with the 2H phase. Bai et al. studied the 1T/2H MoS 2 contact with different contact types and concluded that the edge-contact model exhibits a low tunneling barrier of 0.1 eV and even Ohmic contact in the case of having excess in-plane dangling bonds at the edge of 2H MoS 2 [47]. An important interpretation from the J-V characteristics of the PEC performance is that the conditions for bulk junction properties are fulfilled after 500 ALD cycles and the MoS 2 layer growth is already in the vertical direction. Therefore, in a situation of a high portion of 1T phase with an ∼ 10 23 cm −3 density of Cl ions, within the 2H phase, there exists the 1T/2H MoS 2 phase as an edge-contact type, indicating that this electron transfer path model is valid.  Since the non-ideal growth characteristic of our ALD MoS2 layers is well understood in the context of strong anisotropy in its covalency, the effects of active edge sites could be correlated in PEC operations. The initial number of layers forms in a nanoflake-like structure, mostly parallel to the substrate up to several tens of nm, followed by growing in the vertical direction. The layers progressively passivate the surface states of Si and decrease FLPss, which takes a significant portion for electron-hole pair recombination at the Si interfaces while maintaining the electrochemical activation sites for the HERs. The amount of surface recombination reduction leads to an increased Vph [48], as observed in our experiments. The quantitative analysis of the surface energetics can be understood using Equation (1), 22 where is the surface electron concentration at light, is the surface electron concentration at dark equilibrium, and is a constant at the interface, which is the ratio of energy states to the bulk, and Vo is the amount of band bending at the junction interface. From Equation (1), we can deduce that the MoS2 catalyst layer's thickness range is in the nanoscale domain, changes the surface charge concentration (here, electrons) of the Si, and modifies the surface energetics. The induced band bending also affects the photocurrent increment with the MoS2 layer thickness to the certain point, which can be analyzed using the Butler−Volmer relation in Equation (2) Here, is the electron current, is the electron charge transfer coefficient, and is the overpotential. Based on the equation, the induced band bending modification explained above is an influencing factor for the exchange current and for the overpotential, . The current induced within the bulk region is the sum of the current induced in the junction (depletion region) and the current from the diffusion outside the depletion quasi-neutral region, which can be expressed by Equation  Since the non-ideal growth characteristic of our ALD MoS 2 layers is well understood in the context of strong anisotropy in its covalency, the effects of active edge sites could be correlated in PEC operations. The initial number of layers forms in a nanoflake-like structure, mostly parallel to the substrate up to several tens of nm, followed by growing in the vertical direction. The layers progressively passivate the surface states of Si and decrease FLP ss , which takes a significant portion for electron-hole pair recombination at the Si interfaces while maintaining the electrochemical activation sites for the HERs. The amount of surface recombination reduction leads to an increased V ph [48], as observed in our experiments. The quantitative analysis of the surface energetics can be understood using Equation (1) [22], where n s is the surface electron concentration at light, n 0 s is the surface electron concentration at dark equilibrium, and γ is a constant at the interface, which is the ratio of energy states to the bulk, and V o is the amount of band bending at the junction interface. From Equation (1), we can deduce that the MoS 2 catalyst layer's thickness range is in the nanoscale domain, changes the surface charge concentration (here, electrons) of the Si, and modifies the surface energetics. The induced band bending also affects the photocurrent increment with the MoS 2 layer thickness to the certain point, which can be analyzed using the Butler−Volmer relation in Equation (2) [49].
Here, i n is the electron current, α n is the electron charge transfer coefficient, and η is the overpotential. Based on the equation, the induced band bending modification explained above is an influencing factor for the exchange current and for the overpotential, η. The current induced within the bulk region is the sum of the current induced in the junction (depletion region) and the current from the diffusion outside the depletion quasi-neutral region, which can be expressed by Equation (3) [50], where i ph is the photocurrent density, q is the electronic charge, I 0 is the monochromatic photon flux incident on the semiconductor, α is the light absorption coefficient of silicon, W is the depletion region, and L is the bulk diffusion length of Si. Since the depletion region variation was confirmed from the Mott-Schottky analysis in Figure 3d, we can conclude that for a certain MoS 2 thickness, the solid junction characteristics between the Si and the MoS 2 thin film play a role in enhancing the photocurrent. As a result, the spatially genuine 3D architectures (i.e., laterally homo and vertically heterojunction) with mixed-phases of 1T and 2H of MoS 2 mainly offer synergetic functions in PEC HERs. The ALD deposited MoS 2 layers gradually alleviate the FLP ss effect, and at the same time forming a bulk heterojunction interface. Considering the semiconductor p-n heterojunction utilized in PEC process, there are two types of V ph values. One is at the p-n junction interfaces, and the second is on the surfaces, which are MoS 2 /electrolyte interfaces [36]. It is confirmed in our previous study that our MoS 2 is insensitive to light (i.e., high doping concentrations). It is concluded then that the driving potential for the generated electron-hole pair separation and p-n junction interface-dominated V ph . Now, considering the p-n heterojunction part in the p-Si/MoS 2 , the built-in potential can be expressed by conventional solid-state physics in Equation (4) [51], where V bi is the built-in potential, T is the temperature (294 K), q is the charge of an electron (1.6 × 10 −19 C), k is Boltzmann's constant (1.38 ×10 −23 JK −1 ), N a is the acceptor concentration of p-Si, N d is the donor concentration of MoS 2 , χ is the electron negativity for each type of semiconductor, E G for the band gap, and N CN and N V p are the density of states of the conduction band and valence band, respectively. The values of MoS 2 except for the doping density were referred from the literatures [52][53][54][55].
The theoretical built-in voltage is 0.6 V. For the case of a Schottky barrier, it can be expressed as in Equation (5), where φ B is the Schottky barrier height, φ m is the work function of 1T-MoS 2 . The theoretical barrier height is therefore 0.87 to 0.97 V. The open circuit potential (OCP) difference between dark and light directly indicates the V ph observed, the amount of the band bending of the junction interfaces ( Figure 6). Upon illumination, the amount of the Fermi level shift (V m ) is 0.8 V by observing a very sharp increase in the OCP as shown Figure 6a. This is distinctive in that typical PEC cells exhibit gradual increments when light turns on. The subsequent saturation is attributed to the recombination at the electrode/electrolyte interface, which results in 0.67 V (V ss ). Considering the physical dimension of metallic 1T MoS 2 of approximately 10 nm, which is much smaller than the depletion width, environmental Fermi level pinning (FLP nS ) is expected by nano-Schottky junctions of p-Si/1T MoS 2 although the surface states (FLP ss ) were passivated. Moreover, FLP nS competes with both FLDP ox and FLDP ox followed by pinch-off, enabling effective tuning of the Schottky barrier, which results in the final value of the effective barrier height of 0.8 V (V m ). By observing a sharp increase in OCP, indeed, fast transient behavior occurred when light turns on, implying that FLP is negligible in the present electrochemical reactions. The OCP decay after turning off the light provided information on the generated electron-hole pair recombination at the electrolyte interface and the carrier lifetime as shown in Figure 6b,c. The hole lifetime from the voltage decay is given by Equation (6) [56].  Figure 6c shows the hole lifetime determined from the results. The long lifetime of 60 s for the photocathode confirms that the p-Si/SiOx/MoS2 structure effectively prevented surface recombination with the electrolyte. The fast charge transfer kinetics could be further understood by combining the EIS measurements of our p-Si/SiOx/MoS2 photocathode shown in Figure 6d. The charge transfer characteristics can be analyzed by means of an equivalent circuit model ( Figure S2b), and the circuit parameters are derived from fitting Nyquist plots. In Figure 6d, EIS measurements on 500-ALD-cycle photocathodes under dark and illumination are presented. The overall charge transfer resistance was dramatically reduced under the illumination condition (red inserted) compared to the dark equilibrium condition. Under illumination, for 500-ALD-cycle photocathode, the charge transfer resistance (RMoS2) of 22.38 Ω·cm 2 at the MoS2/electrolyte interface, and the p-Si/MoS2 junction resistance (Rjc) of 27.38 Ω·cm 2 were obtained, confirming the facile kinetics of the photocathode and are catalytically active. The Nyquist plot shows the two distinctive semicircles, which represent the frequency-dependent resistancecapacitance (RC) characteristics. The equivalent circuit, which is depicted in Supporting Figure S2b consists of constant phase elements coupled with the charge transfer resistance. Measured parameters are summarized in Table S1. R refers to the charge transfer resistance at the p-Si/MoS2 junction, R is the charge transfer resistance between the MoS2 and the electrolyte, respectively. Note that under illumination, the R and R significantly decreased more than an order of magnitude compared with that of the dark equilibrium condition, which depicted the induced voltage drop across the. R , charge transfer resistance can be derived from the impedance scan of the low-frequency range, the charge transfer resistance at junction Rjc, and bulk resistance Rbulk from   Figure 6c shows the hole lifetime determined from the results. The long lifetime of 60 s for the photocathode confirms that the p-Si/SiO x /MoS 2 structure effectively prevented surface recombination with the electrolyte. The fast charge transfer kinetics could be further understood by combining the EIS measurements of our p-Si/SiO x /MoS 2 photocathode shown in Figure 6d. The charge transfer characteristics can be analyzed by means of an equivalent circuit model ( Figure S2b), and the circuit parameters are derived from fitting Nyquist plots. In Figure 6d, EIS measurements on 500-ALD-cycle photocathodes under dark and illumination are presented. The overall charge transfer resistance was dramatically reduced under the illumination condition (red inserted) compared to the dark equilibrium condition. Under illumination, for 500-ALD-cycle photocathode, the charge transfer resistance (R MoS2 ) of 22.38 Ω·cm 2 at the MoS 2 /electrolyte interface, and the p-Si/MoS 2 junction resistance (R jc ) of 27.38 Ω·cm 2 were obtained, confirming the facile kinetics of the photocathode and are catalytically active. The Nyquist plot shows the two distinctive semicircles, which represent the frequency-dependent resistance-capacitance (RC) characteristics. The equivalent circuit, which is depicted in Supporting Figure S2b consists of constant phase elements coupled with the charge transfer resistance. Measured parameters are summarized in Table S1. R jc refers to the charge transfer resistance at the p-Si/MoS 2 junction, R MoS2 is the charge transfer resistance between the MoS 2 and the electrolyte, respectively. Note that under illumination, the R jc and R MoS2 significantly decreased more than an order of magnitude compared with that of the dark equilibrium condition, which depicted the induced voltage drop across the. R MoS2 , charge transfer resistance can be derived from the impedance scan of the low-frequency range, the charge transfer resistance at junction R jc , and bulk resistance R bulk from the mid-frequency range to high frequency. Q jc is the junction space charge region capacitance and Q MoS2 is the MoS 2 /electrolyte interface capacitance. The MoS 2 /electrolyte resistance was the lowest for the 300 ALD cycle photocathode, which is due to the vertical layers and exposure of the edge sites. The resistance values for 500 ALD cycles photocathode are somewhat higher than those of the 300 ALD cycle photocathode. The second semicircle, which corresponds to the capacitance-resistance (RC) through the p-i-n junction, indicates that the illumination increases the junction voltage.
Finally, we systematically propose the degree of band bending qualitatively when employing mixed-phase metal chalcogenides on Si for application in PEC water splitting. First of all, the issues of FLP ss could be cleared by ALD passivation. The other would be Fermi-level pinning (FLP nS ) by nanoscale Schottky junctions (p-Si/1T MoS 2 ) [57]. When it comes to the metallic 1T MoS 2 contacted with Si, which forms the increased Schottky barrier height by environmental Fermi-level pinning. Moreover, it has trade-off effects by Fermi-level depinning (FLDP ox ) by the presence of oxide interfaces, i.e., SiO x [58]. We suspect that these mechanisms cancel out each other in V ph enhancements. Because of the discontinuous nature of the deposited layers (based on metallic 1T MoS 2 ), electrical charges are screened only partially, inducing discontinuous junctions. This is also responsible for V ph increase although the junction properties are not the same as the bulk counterpart. Such a conclusion is plausible as studies on the thickness dependence of catalyst layers on the junction characteristics, insisting that the catalyst layer thinner than the Debye length only partially screen charges with nonideal characteristics. Note that in such PEC systems, the currents flow not only through the nanoscale Schottky (metal-semiconductor) junctions but also through semiconductor-electrolyte interfaces. This differs from conventional solid-state devices and was systematically studied by Rossi and Lewis [59]. Therefore, pinch-off effects by Fermi-level depinning (FLDP inhm ) operates from inhomogeneous Schottky junctions at the nanoscale.

Atomic Layer Deposition
Atomic layer deposition of the MoS 2 films. The MoS 2 film was grown on a p-type silicon wafer (Boron doped, 1-30 Ωcm, 525 ± 25 µm thickness) using a custom-designed ALD system. The silicon wafer was cleaned with piranha solution, a 3:1 mixture of concentrated sulfuric acid (H 2 SO 4 ) with hydrogen peroxide (H 2 O 2 ), for 20 min to remove any organic residues on the surface, followed by immersion in buffered oxide etch or BOE (HF:NH 4 F 7:1) to remove the native oxides. The chamber was heated and stabilized for 30 min before reactants were introduced. The MoS 2 thin film was deposited at a temperature range of 250-300 • C using MoCl 5 (99.6%, Strem Chemicals, Inc., Newburyport, MA, USA) and H 2 S (3.99%, balanced N 2 , JC Gas, Daegu, Korea) and a carrier gas of Ar (5N, JC Gas, Daegu, Korea). The pulse time was controlled to be 0.2 s followed by 15 s of purging with Ar at a gas flow of 50 sccm.

Electrochemical Characterization
All electrochemical measurements were performed using a three-electrode system with Ag/AgCl as the reference electrode, a Pt wire (diameter of 0.5 mm) as the counter electrode and MoS 2 on a silicon substrate as the working electrode. Cyclic voltammetry and linear sweep voltamperometry were performed using a VMP3 Potentiostat from Bio-Logic at a scan rate of 50 mV/s in 0.5 M of H 2 SO 4 (pH 0.3) as the electrolyte. For illuminated OCP measurements, 5-10 samples from different batches for each cycle were taken and a solar simulator (Oriel Sol 301A, Newport) with a Keithely 2400 source meter was used under AM 1.5 illumination (100 mW cm −2 ). Electrochemical impedance spectra were measured over a frequency range of 10 6 to 0.1 Hz at 0 vs. the reversible hydrogen electrode (RHE). The Mott-Schottky impedance was performed at 300 kHz to 1 kHz at a bias of −0.6 V to 0 V vs. RHE at 1 sun illumination. The electrode area was 0.785 cm 2 . To make an ohmic contact, the back side of the silicon was scratched with a diamond cutter, followed by a Gallium-Indium eutectic alloy and stuck to the copper electrode using silver paste epoxy. Magnetic stirring at 500 rpm was performed during the experiment.

Conclusions
Our chlorine-rich, ALD-grown MoS 2 was deposited with different cycles on lightly doped p-Si. Neither a high-temperature sulfurization process nor any interface engineering was necessary due to the mixed-phase characteristics of the MoS 2 thin film. Thickness-dependent PEC characteristics were theoretically demonstrated along with experimental results. Our photocathodes showed the anodic shift of 0.47 V compared to bare p-Si with a saturation J ph of 30 mA·cm −2 and excellent stability (75 h under 1 sun illumination). ALD-grown MoS 2 film can induce energy band bending, resulting in a V ph enhancement, which coincides with a bulk p-i-n junction model. The energetically favorable charge transfer mechanism is introduced for the embedded metallic 1T phase MoS 2 .
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4344/8/12/580/s1, Figure S1: Schematic energy level diagram of the band alignment at p-Si/MoS 2 interface in terms of vacuum level and normal hydrogen electrode level in electrolyte of pH = 0.3. The p-Si doping level is~1.6 × 10 15 cm −3 , Figure S2: (a) Nyquist plot of 700-ALD-cycles and 300-ALD-cycles on p-Si photocathode under 1sun illumination at 0 bias. (b) Equivalent circuit corresponding to the EIS measurement. R bulk is bulk resistance of the silicon, Q jc is constant phase element (CPE) of p-Si/MoS 2 junction along with junction resistance R jc , CPE of MoS 2 /electrolyte interface is denoted as QMoS 2 and the resistance of R Mos2 , Table S1: Measured charge transfer resistances.