Dependence of photocatalytic activity on the direction of the intrinsic dipole in GaOI-based Janus structures

Photocatalytic hydrogen production is believed to be an appealing approach to respond to energy crisis and environmental pollution,¹ since the discovery of water photolysis on TiO₂ photoelectrode by Fujishima.² Searching for appropriate photocatalysts is a crucial step to enhance the photocatalytic reaction. Recently, two-dimensional (2D) photocatalysts have been extensively studied due to their large specific surface area, moderate bandgap, and considerable light absorption.^3–17 Bismuth oxyhalide BiOX (X = Cl, Br, I)^12–17 received wide attention because the interlayer electric dipole moment promotes the separation of photo-induced electron–hole pairs. Similar to BiOX, GaOI is also a typical 2D group-IIIA oxyhalide (MOX; M = Ga, In, Tl, X = Cl, Br, I).^18–23 GaOI is a type of p-block-element-based material with an open-layered tetragonal structure. The alternately arranged I⁻/[Ga₂O₂]²⁺ layer induces a strong electric dipole moment.^12–15 The formation energy of GaOI is similar to that of InOCl which has been successfully synthesized by the liquid dissociation method.²⁴ The kinetic and thermal stability of GaOI are also confirmed based on the first-principle calculation.²² 

Constructing Janus structure is an intriguing strategy to release the bandgap limitation and promote the reaction activity in 2D photocatalysts.^25–35 Because of the broken inversion symmetry, the built-in electric field (E_in) and electrostatic potential difference (ΔΦ) are self-induced by the intrinsic dipole (P).^8,34–36 The P vector is affected by charge distribution and is pointed from negative to positive charge. In the dielectric medium, the distribution of surface-bound charge is enslaved to the work function which is determined to be the energy difference between the vacuum level and the Fermi level. The vacuum level, which is related to the electrostatic potential at the surface, refers to the energy of a free stationary electron that is outside of any material.²⁵ On the other hand, the relative electrochemical potentials to the vacuum level are used to determine H⁺/H₂ and H₂O/O₂ potentials based on the SHE model.^37–39 Consequently, H⁺/H₂ and H₂O/O₂ potentials could be modulated by changing the P vector in semiconductor photocatalysts.^31,33,40

Overpotential represents the ability of photo-induced carriers to drive OER and HER processes for water splitting. The overpotentials can be evaluated from the energy difference $[ χ ( O 2 )]$ between the H₂O/O₂ potential and the valence band minimum (VBM) at the oxidation reaction (OER) site, or the energy difference $[ χ ( H 2 )]$ between the conduction band maximum (CBM) and the hydrogen reduction H₊/H₂ potential at the reduction reaction (HER) site. In Janus structures, with the spatial separation of the CBM and VBM, the HER and OER reaction sites react on the S_c surface and the S_v surface. The H⁺/H₂ and H₂O/O₂ potentials should be determined referring to the vacuum level of S_c and S_v, respectively.^31–33 Changing the orientation of the P vector could decide the relative height of ΔΦ between S_c and S_v, modulating the redox potentials and regulating the performance of the Janus photocatalyst for water splitting. However, the mechanism for switching the orientation of the P vector between S_c and S_v and the correlational photocatalytic reaction process have rarely been reported.

To this end, we investigate the effects of the direction of the P vector on H⁺/H₂ and H₂O/O₂ potentials, band alignments, and photocatalytic activity in GaOI-based Janus, based on first-principles calculation. The orientation of the P vector between S_c and S_v can be switched by substituting a metal atom or a halogen atom layer. In GaM′OI (M′ = In and Tl), the P vector points from S_c to S_v, and elevates (reduces) the vacuum level in S_c (S_v), releasing the bandgap requirement, broadening the absorption spectrum of visible light, and enhancing the STH efficiency. While, in GaOX′I (X′ = Cl and Br), the P points from the S_v to S_c surface and suppresses the photocatalytic performance.

First-principles calculations were performed in the Vienna Ab initio Simulation Package (VASP) based on density functional theory (DFT).^41–43 Both structural relaxation and static self-consistency calculations employed a generalized gradient approximation (GGA) with Perdew–Burke–Ernzerhof (PBE).⁴⁴ The interaction of ions with electrons was ascribed by projector-augmented wave potentials (PAWs), in which the semi-core electrons are considered in valence states.⁴⁵ The cutoff energy was set to 500 eV to ensure the accuracy of the convergence force of less than 0.01 eV Å⁻¹ per atom and a total convergence force of 10⁻⁵ eV received. The Brillouin zone was sampled using 7 × 7 × 1 and 9 × 9 × 1 Monkhorst–Pack grids for geometry optimization and electronic structure calculations, respectively. The dipole correction is applied for the whole calculation because of the asymmetry in the vertical direction.^46,47 The vertical vacuum layer was set to ∼20 Å, so that, neighboring layers would avoid interactions in the z-direction. The semi-empirical DFT-D2 method was used to describe the interlayer weak van der Waals (vdW) interactions.^48–50 In addition, by avoiding underestimation of the bandgap in GGA-PBE calculations, all band structures and optical properties were calculated based on the Heyd–Scuseria–Ernzerhof (HSE06) function, which includes an exact 25% non-local Hartree–Fock exchange and 75% semi-local PBE exchange.^51,52 Details on additional calculations of ab initio molecular dynamics (AIMD), phonon diffusion curves, formation energies, light absorption coefficients, carrier mobility, and Gibbs free energy difference (ΔG) in water redox reactions are provided in the supplementary material

The atomic stacking order of GaOI is I–Ga–O–O–Ga–I, in which two orthogonal Ga–O layers are sandwiched by the I atoms. To switch the P orientation between the S_c and S_v surfaces, we construct the Janus structures through two strategies, substituting one Ga layer to the In (Tl) atomic layer [represented as GaM′OI (M′ = In, Tl)] [see Fig. 1(b)] or substituting the I layer to the Cl (Br) atomic layer on a surface [represented as GaOX′I (X′ = Cl, Br)] [see Fig. 1(c)]. To verify the stability, the formation energies are calculated according to Eq. (S1). As shown in Table I, the formation energies are comparable to that of the InOCl monolayer which has been successfully synthesized by a micromechanical exfoliation approach.²⁴ The ab initio molecular dynamics (AIMD) calculations are carried on to further verify the thermal stability. As shown in Figs. S1(a)–S1(d), the potential energy and structure are only vibrated slightly around the equilibrium positions at 300 K for 4 ps. Taking GaInOI as an example, we also evaluate the thermal stability at higher temperatures as shown in Figs. S1(e)–S1(i). One can see that, these structures can hold their integrity without significant distortions up to 700 K temperature. The phonon spectrum is also calculated to confirm the dynamic stability, as shown in Fig. S2. There are no imaginary frequencies throughout the Brillouin zone, indicating the stability of these four Janus structures.

The band structures are calculated based on the HSE06 functional, as shown in Fig. 2. All the Janus structures show indirect bandgaps. CBMs are located at the Г point and VBMs at the Y point (for the GaInOI, GaTlOI, and GaOBrI) or X point (for the GaOClI). Compared with GaOI, GaM′OI reduces the bandgap, while GaOX′I hardly affects the bandgap. The spatial separation of VBM and CBM is found in Janus systems as the band decomposition charge density, as shown in Fig. 2. The distributions of CBM and VBM are closely associated with orbital hybridization in the corresponding atomic layer. In GaM′OI, since the In (Tl) atom has much lower valence electron orbital energy than the Ga atom, the hybridized orbital between In_5s + 5p (Tl_6s + 6p) and I_5p is located at a much lower energy position than that between Ga_4s + 4p and I_5p. As a result, the VBM derives from I_5p states in the Ga–I layer which corresponds to S_v and the CBM mainly consists of antibonding hybridized In_5s + 5p (Tl_6s + 6p) in the M′–I layer, which corresponds to S_c, as shown in Figs. 2 and S3. On the other hand, in GaOX′I, the Cl (Br) atom possesses much more electronegativity than the I atom, resulting in much lower bonding orbitals in the X′-Ga layer. The CBM mainly derives from hybridized Ga_4s + 4p in the Ga–X′ layer, while the VBM derives from I_5p states in the Ga–I layer. The Ga–X′ layer and Ga–I layer correspond to S_c and S_v, respectively.

Based on Yang’s theory,²⁵ the oxidization potential (H₂O/O₂ potential) and reduction potential (H⁺/H₂ potential) are enslaved to ΔΦ, and then, the bandgap restriction for the photocatalyst is reduced from 1.23 to 1.23 eV − ΔΦ.²⁵ ΔΦ and E_in are induced by the intrinsic dipole (P) according to Eqs. (S2) and (S3). P, depending on the charge distribution, is formed from negative to positive charge. In GaInOI (GaTlOI), the 0.13 D (0.33 D) of P is found from the S_c [Tl–I (In–I) layer] to the S_v (Ga–I) layer. From Figs. 3(a) and 3(b), one can see that the vacuum level of the Ga–I surface is 0.327 eV (0.839 eV) higher than that of the In–I (Tl–I) surface, resulting in 0.327 eV (0.839 eV) lower work function on the In–I (Tl–I) surface so that electrons can easily escape to the In–I (Tl–I) surface. Negative bound charges prefer to be located on the In–I (Tl–I) layer. As a result, the P vector is formed in the direction from the S_c [In–I (Tl–I) layer] to S_v (Ga–I) layer. On the contrary, in GaOClI (GaOBrI), 0.23 D (0.14 D) of P points from S_v (Ga–I layer) to S_c (Ga–X layer) because the Ga–I surface possesses a much lower work function than the Ga–X′ surface, as shown in Figs. 3(c) and 3(d).

Based on the standard hydrogen electrode model,^37–39 the potentials of both H₂O/O₂ and H⁺/H₂ are determined to be the relative electrochemical potentials to the vacuum level. In GaM′OI, as shown in Figs. 3(a) and 3(b), the P vector which points from S_c (In–I or Tl–I surface) to S_v (Ga–I surface), induces E_in and ΔΦ in the opposite direction. A much higher vacuum level is formed on the S_v surface. As a result, the P vector drops down the H⁺/H₂ potential on the S_c surface and elevates the H₂O/O₂ potential on the S_v surface, which relieves the bandgap restriction from 1.23 eV to 1.23 eV − ΔΦ and improves the overpotentials. In GaInOI, based on the 0.327 eV of ΔΦ, the VBM is lower by 0.51 eV than the H₂O/O₂ potential which is −5.67 eV to the vacuum level in the S_v surface, while the CBM is higher by 0.40 eV than the H⁺/H₂ potential which is −4.44 eV to the vacuum level in the S_c surface, as shown in Fig. 3(a). The 1.88 eV bandgap not only satisfies the bandgap requirements of water splitting but also efficiently absorbs visible light. In GaTlOI, considering 0.839 eV of ΔΦ, the 0.99 eV bandgap is narrower than 1.23 eV but larger than 1.23 eV − ΔΦ. Both VBM and CBM still straddle the redox potentials of water splitting. However, in GaOClI (GaOBrI), the vacuum level on S_v is lower by 0.659 (0.392) eV than that on S_c, increasing the bandgap restriction from 1.23 eV to 1.23 eV + ΔΦ and reducing the overpotentials. Although GaOClI and GaOBrI possess 2.11 and 2.15 eV bandgaps, VBMs are 0.23 and 0.1 eV higher than the H₂O/O₂ potential, hindering the oxidization reaction.

Optical absorption, which is the key physical quantity for a photocatalyst, is also affected by the orientation of the P vector between S_c and S_v. The absorption coefficients (α) derived from the dielectric function calculated based on the HSE06 functional are obtained based on Eq. (S4). The optical bandgap is the energy barrier that electrons in a substance jump when light interacts with the substance and is determined by the energy band structure of the substance itself. The optical bandgap was estimated based on the absorption spectrum and is shown in Table I. The same as the evolution of the bandgap restriction, the optical absorption range is also enlarged (compressed) when the P points from S_c (S_v) to S_v (S_c). The absorption coefficients are shown in Fig. 4 and the optical bandgap is shown in Table I. One can see a significant red shift in both GaInOI and GaTlOI. The α values are up to the order of 10⁴ cm⁻¹ in the visible region, which is superior to the heterojunction constructed by InSe and GaTe.^53,54 The red shift is also in accordance with the electronic bandgap reduction, as shown in Table I. For GaOX′I, a notable blue shift is found and the absorption coefficients are suppressed in the range of visible light. Distinct anisotropy is maintained in optical absorption, which is suitable for the separation of electron–hole pairs. Anisotropic optical properties can be attributed to its special crystal structure. Different atomic species at the top and bottom break the vertical mirror symmetry in the x and y directions.

High carrier mobility facilitates the photo-induced carriers’ migration to the chemically active sites for water splitting. The carrier mobility (μ) is assessed based on the theory of deformation potential (DP) by Eq. (S7).⁵⁵ Since the DPA theory does not take into account some factors, such as lattice vibrations and the interaction between electrons and phonons, we also verified the carrier mobility by solving the electron scattering matrix using Fermi's golden rule in the framework of the first-principles calculation⁵⁶ to obtain more accurate results. By using the HSE06 and PBE schemes, the VBM and CBM follow similar trends with applied strain;^57,58 however, the HSE06 method is time intensive. Therefore, the results obtained by the PBE scheme are still accurate but less time-consuming. Table II shows the calculated results in the PBE scheme, including the elastic modulus $( C 2 D)$⁠, effective mass $( m ∗)$⁠, DP constant $( E i)$⁠, and corresponding mobilities μ₁ (by DPA) and μ₂ (by Nanodcal). The μ of GaOI is also calculated for comparison. We find that the electron mobility and hole mobility of four Janus systems are superior to the results of GaOI, demonstrating their fast carrier migration capabilities. The mobility of Janus exhibits a notable anisotropy. For instance, $μ 2 e$ is almost ten times than $μ 2 h$ in the x direction of GaInOI, as shown in Table II, which indicates that the x direction of GaInOI prefers transferring electrons to transferring holes. The anisotropy of carriers also facilitates the separation of photoexcited carriers.

Increasing the efficiency of STH is the main purpose for pursuing photocatalysts. The theoretical upper limit of STH efficiency is calculated by Eq. (S8) in ideal conditions, where the efficiencies of carrier separation and light absorbance are all assumed to be 100%. The impact of E_in is also considered because it affects the separation of electron–hole pairs.²⁵ In Eqs. (S9)–(S11), the bandgap $E g$⁠, overpotential of HER $χ( H 2)$⁠, overpotential of OER $χ( O 2)$⁠, and ΔΦ are calculated by the HSE functional. The results are shown in Table S2. The same as the case in optical absorption, an obvious dependence is also found between the P orientation and STH efficiency. One can see that GaM′OI systems possess high $η ′ STH$⁠, but GaOX′I systems exhibit lower $η ′ STH$⁠. The high $η ′ STH$ of GaM′OI systems derives from the broad range of visible light absorption as well as suitable $χ( H 2)$ and $χ( O 2)$⁠, which determines the photon energy used for water decomposition. The $η ′ STH$ (see Table S2) of GaTlOI and GaInOI is up to 35.7% and 19.6% with instinct dipole in consideration, higher than that of the Janus WSSe monolayer (11.68%),³³ meeting the standard for commercial applications of hydrogen production using photocatalytic water splitting for (10%).⁵⁹ 

Based on the above results, we find that GaInOI and GaTlOI are potential photocatalysts for overall water splitting. To achieve the reaction process of HER and OER under light irradiation, the photogenerated carriers should provide sufficient reaction driving force, i.e., sufficient external potential. We calculated the external potential of photogenerated electrons for HER $U e$ (photogenerated holes for OER $U h$⁠) as the energy difference between CBM (VBM) and H⁺/H₂ (H₂O/O₂) potential. Based on the band structures (see Fig. 2), we find spatially separated catalytic behaviors in GaM′OI systems. The reduction and oxidation reactions occur on the M′–I surface and Ga–I surface, respectively. As a result, U_e is determined relative to the H⁺/H₂ potential on the M′–I surface, while U_h is calculated referring to the H₂O/O₂ potential on the Ga–I surface.

$U e$ and $U h$ are then calculated based on $E H + / H 2$ and $E H 2 O / O 2$⁠, respectively. The results are shown in Fig. S4. We find that $U e$ is positive when pH ≤ 6 for GaInOI, but that of GaTlOI is positive only when pH ≤ 2, which means that GaInOI can work in an acid environment but GaTlOI can only work in a strong acid environment. $U h$ is always positive in the whole pH range, indicating that GaM′OI possesses noteworthy activity to drive the oxidation reaction.

To evaluate whether the redox reactions of water splitting could proceed spontaneously, we systematically investigated the reaction mechanism of HER and OER in GaM′OI. Dramatic structure distortions were found in H- or O-adsorbed GaTlOI, as shown in Fig. S5, indicating its unavailability as a photocatalyst for water decomposition. On the contrary, barely any change is found in GaInOI. Hereinafter, we focus on the surface activity of GaInOI for water decomposition. First, the adsorption of H₂O molecules is studied as shown in Fig. S6. The 4 × 4 × 1 of supercells is used to avoid the interaction between the adjacent adatoms. Adsorption energy is determined as the energy difference between the initial and final states in the process of H₂O absorption on the surfaces: $E absorb= E Janus _ H 2 O− E H 2 O− E Janus$⁠. $E Janus _ H 2 O$⁠, $E H 2 O$⁠, and $E Janus$ indicate the cohesive energy in H₂O absorbed Janus, isolated H2O, and isolated Janus, respectively. The adsorption energies (E_adsorb) are calculated to be 0.137 and 0.887 eV on the In–I side and Ga–I side, respectively, and, thus, H₂O is hardly absorbed on both In–I and Ga–I sides.

Inducing vacancy defects has been reported to be an efficient approach to improve the surface activity of catalysts in redox reactions. The HER and OER are evaluated in GaInOI with I vacancy (V_I). The HER can be described by Volmer reaction and Heyrovsky reaction [see Eqs. (S13) and (S14)]. The binding energy of H* (ΔG_H*) is a well-accepted descriptor to evaluate the HER performance in the acid condition. The most stable absorption structure is shown in Fig. S7(b). ΔG_H* was calculated according to Eqs. (S12) and (S19) based on Nørskov's standard hydrogen electrode (SHE) model.⁶⁴ The ΔG_H* value close to 0 eV means optimal hydrogen adsorption and desorption processes, which are beneficial for HER of catalysts. The HER activity is also investigated in the In–I surface with I single vacancy (V_I). The most stable absorption structure is shown in Fig. S8. The ΔG_H* value in Fig. 5(a) is 0.31 eV, which means the absorption of proton is still weak at the equilibrium potential. The ΔG_H* value can be tuned to be close to zero eV by increasing the external potential of electrons. The ΔG_H* value is modified to be −0.03 eV by applying 0.4 V of potential which is just the potential that the photo-induced electrons can provide. Consequently, GaInOI with V_I is an anticipated photocatalyst for the HER in water decomposition.

Then, the OER process is evaluated following a four-electron reaction path. We focus on the Ga–I surface of V_I-doped GaInOI, where H₂O tightly binds (E_absorb is calculated to be −1.52 eV). The structures with absorbed OH*, O*, and OOH* intermediates are given in Fig. S9. The reaction free energy diagrams calculated based on Eqs. (S12) and (S20)–(23) are shown in Fig. 5(b). One can see that at U = 0 V, the first step is exergonic, but all other steps are endothermic. The rate-determining step is the last step in which O₂ is formed and released from the Ga–I surface. According to the typical experimental electrocatalytic conditions, OER performance usually takes place in an alkaline electrolyte. From Fig. S4, we find that the potential supplied by the photo-induced hole is 1.25 V at pH = 7, and it increases as the pH value increases. From Fig. 5(b), one can see that the OER proceeds spontaneously when the potential is 1.25 V. As a result, V_I-doped GaInOI is an outstanding photocatalyst for the OER.

In summary, based on the DFT calculation, we find that the orientation of P between Sc and Sv dominates the band alignment, STH efficiency, and photocatalytic activity in GaOI-based Janus. The direction of P can be overturned by substituting different atomic layers. In GaM′OI, the P orientates from Sc to Sv, and Janus structures exhibit a broad light absorption range, high carriers’ mobility, and STH efficiency. In GaOX′I, the P orientates from Sv to Sc, and Janus structures exhibit terrible photocatalytic performance. The STH of GaInOI is up to 19.6% and that of GaTlOI is up to 35.7% in the full solar spectrum. V_I-doped GaInOI is a potential photocatalyst for overall water splitting.

See the supplementary material for the calculation details and detailed results (such as electronic structures, AIMD, phonon spectrum, absorbed structures, and intrinsic dipole). Calculations for the bands and optical absorption are carried on by using the HSE06 method.

We gratefully acknowledge HZWTECH for providing computation facilities. This work was supported by the National Natural Science Foundation of China (NNSFC) (Grants No. 11974299 and 11874316), the Scientific Research Fund of Hunan Provincial Education Department (Grants. 20K127, 20A503, and 20B582), Provincial Applied Characteristic Discipline Material Science and Engineering Open-end Funds of Hunan Institute of Technology (Grant No. KFB22008), and the Program for Chang Jiang Scholars and Innovative Research Team in University (Grant No. IRT13093).

The authors have no conflicts to disclose.

Manting Li: Data curation (equal); Formal analysis (equal); Methodology (equal). Mengshi Zhou: Conceptualization (equal); Formal analysis (equal); Methodology (equal). Jiewen Min: Formal analysis (equal); Investigation (equal); Methodology (equal); Validation (equal); Writing – review & editing (equal). Chunxiao Zhang: Conceptualization (equal); Formal analysis (equal); Investigation (equal); Validation (equal); Writing – review & editing (equal). Chao Tang: Conceptualization (equal); Formal analysis (equal); Project administration (equal); Software (equal); Supervision (equal); Validation (equal). Jianxin Zhong: Conceptualization (equal); Formal analysis (equal); Project administration (equal); Software (equal); Validation (equal).

The data that support the findings of this study are available from the corresponding authors upon reasonable request.