TiO₂-based heterojunctions for photocatalytic hydrogen evolution reaction

Conventional semiconductor heterojunction photocatalyst

Conventional heterojunctions can be categorized into type-I, type-II, and type-III [Figure 3]. In straddling gap type-I heterojunction, the CB of semiconductor A and VB of semiconductor B are more negative than the CB of semiconductor B and VB of semiconductor A. In staggered gap type-II heterojunction, the CB and VB of semiconductor A are more negative than those of semiconductor B, respectively. In type-III heterojunction with a broken gap, both CB and VB of semiconductor A are more negative than CB of semiconductor B.

Figure 3. Schematic illustration of (A) type-I, (B) type-II, and (C) type-III heterojunctions. Quoted with permission from Low et al.^[30].

In type-I heterojunction, the electrons of both semiconductors transition from VB to CB and leave holes in the VB under the photoexcitation [Figure 3A]. Based on the characteristics of the band edge position, the h⁺ will gather in the VB edge of semiconductor B, and photoelectrons will accumulate in the CB of semiconductor B. Among the three conventional heterojunctions, Type-II is one of the most popular photocatalysis heterojunctions studied due to its suitable energy band structure, band-edge position, and high photogenerated carrier separation efficiency. As shown in Figure 3B, when the two semiconductors are in contact, under light excitation, the e^- on the CB of semiconductor A will migrate to the CB of semiconductor B, and the holes on the VB of semiconductor B will gather on the VB of semiconductor A. The oxidation-reduction reaction will be conducted in the CB of semiconductor B and the VB of semiconductor A, respectively, so that the photogenerated e^-/h⁺ pairs will be effectively separated, thus improving the photocatalysis activity^[31]. Figure 3C demonstrates the structural features of type-III heterojunction photocatalysts. Obviously, in type-III heterojunction, both CB and VB of semiconductor B are more positive than those of semiconductor A. Such energy band relationship is not conducive to separating photogenerated electron-hole pairs^[32], and the photocatalytic performance is inferior. Therefore, this review will not further elaborate on the TiO₂-based type-III heterojunction.

Table 1 summarizes the efforts and progress of TiO₂-based photocatalysis type-I and type-II heterojunctions for photocatalysis hydrogen evolution from 2019 to 2023. Cao et al. prepared Cu₂O/D-TiO₂ type-I heterojunctions, and a series of composites with Cu₂O contents ranging from 1% to 10% were synthesized^[33]. Among them, the heterojunction with 5 wt% Cu₂O loading showed the highest H₂ evolution rate of 4.81 mmol^-1h^-1 under UV-visible light [Figure 4A-C]. The work functions of D-TiO₂ and Cu₂O were 4.69 and 5.46 eV, respectively. Only the e^- of the CB of D-TiO₂ flowed to the CB of Cu₂O when contacted in the dark environment. Under photoexcitation, both photogenerated e^-/h⁺ pairs of D-TiO₂ transfer towards CB and VB of Cu₂O, respectively. Luo et al. prepared CdS quantum dots and ultrathin TiO₂(B) nanosheets by photodeposition and hydrothermal methods^[34], respectively, where the heterojunction (PCT) prepared by photodeposition belongs to the type-II heterojunction and the heterojunction (HCT) prepared by hydrothermal method belongs to the type-I heterojunction [Figure 4D]. Under visible light excitation, 10-HCT will be excited only by CdS. However, under full-spectrum irradiation [Figure 4E], the photogenerated e^- and h⁺ on the CB and VB of B-TiO₂ will be transferred to the CB and VB of CdS, respectively, which will improve the photocatalytic activity. Whereas in 5PCT [Figure 4F and G] and 5PCT(PVP) [Figure 4H and I], which satisfy the type-II energy band arrangement, the e^- on the CB of B-TiO₂ will be transferred to the CB of CdS under the visible-light excitation, while the holes are retained in the VB of B-TiO₂, and the full-spectrum irradiation will stimulate the hole transfer on the VB of CdS to the VB of B-TiO₂ at the same time based on the visible-light excitation, thus effectively separating the photogenerated carriers and improving the photocatalytic activity. The photocatalysis hydrogen evolution rates of 10-HCT and 5-PCT reach 11,317 and 34,937 μmolg^-1h^-1, respectively. This shows that type-I heterojunction can improve the photocatalysis hydrogen evolution efficiency but is much less efficient than type-II heterojunction.

Figure 4. (A-C) Cu₂O/D-TiO₂ type-I heterojunction. Quoted with permission from Cao et al.^[33]. (D-I) photocatalysis mechanism diagrams of 10-HCT, 5-PCT and 5-PCT (PVP). Quoted with permission from Luo et al.^[34].

Type-I heterojunction effectively reduces the band gap and accelerates the photoexcited electron and hole transfer, However, it is rarely reported due to the difficulty of type-I heterojunction to separate the photogenerated e^-/h⁺ pairs.

TiO₂/UiO-66-NH₂ type-II heterojunction was constructed by Kuang et al. by photo-induced activation treatment [Figure 5A and B]^[27]. Before contact, the CB and VB of UiO-66-NH₂ were more negative than those of TiO₂, respectively. After contact, under photoexcitation, in situ X-ray photoelectron spectroscopy (XPS) characterization revealed that UiO-66-NH₂ acted as an electron donor to transfer photogenerated electrons from the Lowest Unoccupied Molecular Orbital (LOMO) to CB of TiO₂, while h⁺ of TiO₂ was transferred to the Highest Occupied Molecular Orbital (HOMO) of UiO-66-NH₂, which led to the separation of the photogenerated electron-hole pairs, the inhibition of the combination of photogenerated carriers, and the improvement of photocatalytic performance, with the photocatalytic hydrogen evolution rate reaching 593.53 μmolg^-1h^-1. Park et al. synthesized In₂S₃/S, N-doped TiO₂ nanostructures by a hydrothermal method^[35]; it has been observed that S, N doping significantly decreases the work function of TiO₂, enhances the electron concentration, effectively reduces the distance between CB and Ef, and increases the carrier concentration. Furthermore, the formation of In₂S₃/TiO₂ heterojunctions can effectively inhibit carrier combination, and its carrier lifetime is increased by a factor of 20 compared with that of the monomer. Specifically, the carrier lifetimes of TiO₂, S- and N-doped TiO₂, and S, N-doped TiO₂/In₂S₃ are 72.4, 50.3 ns, and 1.16 µs. The band edge positions [Figure 5C] show that the VB and CB of the S, N-doped TiO₂ are more positive than those of In₂S₃. This conforms to the type-II energy band arrangement, and the bandgap of the heterojunction is only 2.27 eV. Under the photoexcitation [Figure 5D], h⁺ will accumulate on the VB of In₂S₃, while e^- will gather on the CB of S, N-doped TiO₂. Ultimately, the oxygen evolution (OER) and hydrogen evolution reactions will be performed on the VB of In₂S₃ and the CB of S, N-doped TiO₂, respectively, thus realizing photocatalytic water splitting.

Figure 5. (A and B) Schematic diagram of UiO-66-NH₂ heterojunctions. (A and B) is quoted with permission from Kuang et al.^[27]. (C) Energy band position of Pristine TiO₂, S, N-doped TiO₂, In₂S₃ NPs on TiO₂, and In₂S₃, (D) S, N-doped TiO₂/In₂S₃ heterostructure and photocatalytic mechanism diagrams. (C and D) is quoted with permission from Park et al.^[35].

In addition, the excellent photocatalysis properties of binary type-II heterojunctions have led to great interest in ternary and even multicomponent type-II heterojunctions. Recently, Zhou et al. have designed a Type-II/Type-II energy band arrangement of g-C₃N₄ quantum dots/a-TiO₂/r-TiO₂ heterojunctions for photocatalysis total water splitting^[36]. Figure 6A-F shows the Type-II/Type-I and Type-II/Type-II energy band arrangement and photocatalytic mechanism diagrams. In addition, its photocatalysis hydrogen and oxygen evolution efficiencies reached 1,526.4 and 198.8 μmol h^-1, respectively, and it can be extended to the decomposition of methylene blue (MB). Yang et al. prepared Mxene-derived anatase-TiO₂/rutile-TiO₂/In₂O₃ heterojunctions to enhance the hydrogen evolution rate to 1,488 times that of In₂O₃^[37].

Figure 6. (A-F) Energy band arrangements and photocatalytic mechanisms diagrams of T-II/T-II, T-II/T-I, T-II/T-II Stairs, T-II/T-II Concave and T-II/T-II convex. Quoted with permission from Zhou et al.^[36].

Although type II heterojunctions can effectively inhibit the compounding of photogenerated electron-hole pairs, their study is limited by the fact that the original electrons inside the semiconductor will impede the foreign electron transport, which results in a significant reduction of their oxidation-reduction capability.

p-n heterojunction photocatalysts

The reaction mechanism of p-n heterojunction photocatalysis differs from that of type-II heterojunctions. In p-n heterojunctions, photocatalytic efficiency enhancement primarily stems from energy band engineering and improved carrier separation driven by internal electric fields (IEF)^[31]. As illustrated in Figure 7A and B, p-n heterojunctions have two different type-II energy band arrangements^[31]. Both generate an IEF pointing from the n- to the p-type semiconductor in the heterojunction formation^[49]. In Figure 7A, the CB and VB of p-type semiconductors in the first type of p-n heterojunction are higher than those of n-type semiconductors, respectively, and e^- will be transferred from the CB of the p-type semiconductors to the CB of the n-type semiconductors and h⁺ will be transferred from the VB of the n-type semiconductors to the VB of the p-type semiconductors^[50]. Since the direction of the IEF is the same as that of charge transfer, the separation of e^-/h⁺ pairs is accelerated by the IEF. However, in the second type of p-n heterojunction [Figure 7B], the type-II energy band arrangement is opposite to that of Figure 7A, and the e^- will be transferred from CB of an n-type semiconductor to CB of a p-type semiconductor, and the h⁺ will be transferred from VB of a p-type semiconductor to VB of an n-type semiconductor. In this case, we will find that the direction of IEF is opposite to that of the charge transfer, which will result in the separation decelerated by restricting the movement of the e^-/h⁺ pairs^[31]. In addition to the above two types of p-n heterojunctions, some researchers have also suggested that the mechanism of the second type of p-n heterojunctions should be consistent with the direct Z-scheme heterojunctions^[51-53]. The distinction and categorization of p-n and direct Z-scheme heterojunctions has been controversial. Therefore, their difference should be thoroughly verified in future heterojunction studies.

Figure 7. Type-II energy band arrangement of (A) the first type and (B) the second type of p-n heterojunction. (A, B, I) is quoted with permission from Ding et al.^[49]. (C) catalysis mechanism of Co₃O₄/TiO₂. (D and E) G_H of TiO₂ and Co₃O₄ and (F and G) work function for the TiO₂ and Co₃O₄ facets. (C-G) is quoted with permission from Wang et al.^[54]. (H) band structure diagram for Ag₃PO₄/b-N-TiO₂. (H) is quoted with permission from Zhou et al.^[15]. (I) The charge segregation mechanism of Cu₂O/Ni(OH)₂/TiO₂.

The above two p-n heterojunctions with different action mechanisms have been reported and synthesized experimentally in many cases [Figure 7C-G]. Wang et al. synthesized 2D/1D and 3D/1D Co₃O₄/TiO₂ composites by a hydrothermal method with a hydrogen evolution rate of up to 3,460 µmolg^-1h^-1[54], and DFT calculations were used to explore the reason for their high activity. Theoretical calculations show that the Fermi level of Co₃O₄ is significantly higher than that of TiO₂ [Figure 7D and E], resulting in a greater e^- transfer to TiO₂ and a significant increase in electron concentration. Additionally, due to the difference in Fermi energy levels, an IEF is generated within TiO₂ towards Co₃O₄, accelerating the transfer of e^- from the CB of Co₃O₄ to that of TiO₂ and h⁺ from the VB of TiO₂ to that of Co₃O₄. This effectively promotes photogenerated carrier separation, increases carrier concentration, and improves catalytic activity. Figure 7F and G shows the Gibbs free energies after heterojunction formation on TiO₂ and Co₃O₄ of 0.23 and -0.35 eV, respectively. As reported in Figure 7H, Zhou et al. reported p-n heterojunctions have opposite band edge positions^[15]; the CB of Ag₃PO₄ is lower than the CB of b-N-TiO₂, and the VB of Ag₃PO₄ is higher than the VB of b-N-TiO₂. An IEF pointing from b-N-TiO₂ to Ag₃PO₄ will be generated at the interface, which balances the Fermi energy levels, leading to a photogenerated e^-/h⁺ pairs separation by moving the e^- at the CB of b-N-TiO₂ towards the CB of Ag₃PO₄, while the h⁺ move in the opposite direction of the e^-. Overall, the p-n heterojunctions composed of two different type-II energy band arrangements both can effectively improve the transfer and separation of electron-hole pairs and the photocatalytic activity.

Table 2 summarizes some of the applications of p-n type heterojunctions for photocatalytic hydrogen evolution from 2019 to 2023. Most current research has focused on synthesizing a single p-n heterojunction, which is certainly effective, but the utilization efficiency of photogenerated e^-/h⁺ pairs is still poor. Accordingly, researchers have begun to turn their attention to exploring dual p-n type heterojunctions. As presented in Figure 7I, Ding et al. adopt chemical precipitation and ultrasound-assisted glucose reduction to synthesize Ni(OH)₂/TiO₂,Cu₂O/TiO₂ and Cu₂O/Ni(OH)₂/TiO₂ photocatalysts^[49]. It is found that the hydrogen evolution rate of Cu₂O/Ni(OH)₂/TiO₂ is 6,145 µmolg^-1h^-1 which is much higher than that of Ni(OH)₂/TiO₂ (3,265 µmolg^-1h^-1) and Cu₂O/TiO₂ (2,285 µmolg^-1h^-1). The excellent photocatalytic activity is attributed to the synthesized Cu₂O/Ni(OH)₂/TiO₂ being a fully depleted p-n junction with IEF. Under photoexcitation, the e^- on the CB of Ni(OH)₂ and Cu₂O will be transferred to the CB of TiO₂ because the CB of TiO₂ is more positive than the CB of Ni(OH)₂ and Cu₂O, while the h⁺ will be transferred from the VB of TiO₂ to the VB of Ni(OH)₂ and Cu₂O, respectively, because the VB of TiO₂ is more positive. The direction of the IEF is directed from TiO₂ to Ni(OH)₂ and Cu₂O, respectively, which is opposite to the motion direction of the e^- and the same as that of the h⁺, thus promoting the carrier migration and effectively improving the separation of the photogenerated e^-/h⁺ pairs, which effectively improves the photocatalytic activity. In addition, Chen et al. prepared a 3D g-C₃N₄-Cu₂O-TiO₂ by a sacrificial template strategy and photo-deposition with a maximum hydrogen evolution rate of 12,108 μmol g^-1h^-1[55]. Overall, p-n heterojunction plays an indispensable role in heterojunction catalysts due to their unique energy band properties and excellent photocatalytic activity, but it does not overcome the disadvantages of the type II heterojunction.

Z-scheme heterojunction photocatalysts

Comparing the limited carrier oxidation-reduction capacity of type-II and p-n heterojunctions, the Z-scheme can not only effectively promote carrier separation but also enhance the carrier oxidation-reduction capacity. Z-scheme heterojunctions can be classified into three main categories: liquid-phase, all-solid-state, direct, and dual.

The liquid-phase Z-scheme heterojunction [Figure 8A] is characterized by a pair of electron acceptor-donor (A-D) between the two semiconductors as a transport medium for e^-[65]. At this time, the e- on CB of semiconductor 1 and the h⁺ on VB of semiconductor 2 will react within A and D, respectively. After the formation of the heterojunction, the oxidation reaction is carried out on the VB of semiconductor 1, and the reduction reaction is conducted on the CB of semiconductor 2, which effectively separates the photogenerated carriers. However, the existence of the reverse reaction and the light shielding effect will reduce the number of e^- and h⁺. In addition, the liquid medium is easy to deactivate, and all of these uncertainties will affect the photocatalytic activity to a certain extent^[17].

Figure 8. (A) liquid-phase Z-scheme heterojunction. Quoted with permission from Bai et al.^[65]. (B) Au@TiO₂-12%TrTh all-solid-state Z-scheme heterojunction. Quoted with permission from Han et al.^[66]. (C) photocatalytic mechanism of TiO₂-ZnIn₂S₄ nanoflowers. Quoted with permission from Zuo et al.^[71]. (D) physical structure of Pt/g-C₃N₄/TiO₂/IrO_x (PCTI). (E) The charge-transfer process within PCTI upon light irradiation. (F) H₂ evolution rate. (D-F) is quoted with permission from Moon et al.^[70]. (G) Photocatalytic mechanism diagram of double Z heterojunction TMOP. Quoted with permission from Sun et al.^[72].

The all-solid-state Z-scheme heterojunction is proposed to circumvent the disadvantages of liquid-phase Z-scheme heterojunction. Its main feature is to insert a piece of conductor between two semiconductors to promote carrier migration and improve photocatalytic activity. For example [Figure 8B], Han et al. reported that Au was inserted as the most electron-mediated medium between TiO₂ and TrTh^[66]. At this time, the e^- of CB of TiO₂ would combine with the h⁺ on the VB of TrTh through Au, and the oxidation and reduction reactions would take place on the VB of TiO₂ and the CB of TrTh, respectively. The selection and design of electron mediators is a particularly important aspect in all-solid-state Z-scheme heterojunctions. However, the high price of the electronic medium seriously limits the development of the all-solid-state Z-scheme heterojunction.

In 2009, Wang et al. addressed the challenge of costly all-solid-state Z-scheme heterojunctions by introducing electron-mediator-free, direct-contact Z-schemes^[67]. In 2013, Yu et al. further proposed direct Z-scheme heterojunctions enhanced by IEF^[68]. The key distinction of this design from liquid Z-schemes lies in its elimination of the need for an electron transfer medium. As illustrated in Figure 8C, when TiO₂ and ZnIn₂S₄ form a heterojunction, their energy bands align in a staggered configuration due to disparities in Fermi energy levels and work functions, causing band bending at the contact interface. IEF direct from ZnIn₂S₄ towards TiO₂, and coupled with band bending, facilitates electron transfer from the CB of TiO₂ to the VB of ZnIn₂S₄ to recombine with h⁺, while simultaneously impeding the reverse flow of electrons from the CB of ZnIn₂S₄ and holes from the VB of TiO₂. This mechanism both guarantees sufficient charge carriers for redox reactions and spatially separates photogenerated carriers, thereby contributing to enhanced photocatalytic performance.

The direct Z-scheme has been widely studied in academia due to its unique energy band structure and efficient photocatalytic ability. Consequently, we will focus on the direct Z-scheme heterojunctions in photocatalysis hydrogen evolution. Ran et al. used liquid exfoliation, precipitation-hydrothermal methods to prepare the ReSe₂/TiO₂ direct Z-scheme heterojunctions^[69]. These prepared heterojunctions exhibited a photocatalysis hydrogen evolution rate of up to 2,081 µmolg^-1h^-1 and were combined with DFT to show that the presence of an IEF promotes the electron transfer from TiO₂ to ReSe₂. Moon et al. successfully synthesized the Pt/g-C₃N₄/TiO₂/IrO_x heterojunctions [Figure 8D] through the classical Stöber method^[70]. The formation of Z-scheme heterojunctions [Figure 8E] accelerated the surface charge separation and reaction kinetics, resulting in catalysts with apparent quantum yields, hydrogen evolution rates, and oxygen evolution rates as high as 24.3%, 8.15 mmolg^-1h^-1 [Figure 8F], and 443.9 μmolg^-1h^-1, respectively. Photogenerated e^- was reduced to H₂ on the CB of Pt/g-C₃N₄ and h⁺ was oxidized to oxygen on the VB of TiO₂/IrO_x to achieve total water splitting.

Double Z-scheme heterojunction can further enhance the separation of photogenerated e^-/h⁺ pairs and improve the photocatalytic activity. As shown in Figure 8G, MP-1, MP-2, and MP-3 are contacted to form a double Z-scheme heterojunction. After photoexcitation, the e^- on the CB of MP-2 will combine with the h⁺ on the VB of MP-1 and MP-3, respectively. The remaining photogenerated e^- on the CB of MP-1 and MP-3 will participate in the reduction reaction, and the h⁺ on the VB of MP-2 will engage in the oxidation reaction. It should also be noted that the double Z-scheme heterojunction is not a simple superposition of the three components but rather utilizes the synergistic effect among them to accelerate the electron transfer and the separation of the photogenerated carriers to extend the lifetime of the carriers, thus improving the photocatalytic activity.

Table 3 demonstrates some research results of TiO₂-based Z-scheme heterojunctions in photocatalysis hydrogen evolution from 2019 to 2023.

S-scheme heterojunction photocatalysts

Xu et al. reported the first S-scheme heterojunction photocatalysts in 2019^[29]. Zhang et al. extended the S-scheme heterojunction which can only be of an n-n type to n-p, p-n, p-p type in 2022^[86]. However, the condition is that the CB position and Fermi energy level of the reduced semiconductor (RP) should be higher than those of the oxidized semiconductor (OP) simultaneously at this time; the OP and RP can be either n- or p-type semiconductors [Figure 9A-D]. As in Figure 9E-G, taking the n-n junction as an example, when both the CB and E_f of RP are higher than those of OP, due to the difference in E_f, an IEF pointing from RP to OP will be formed between the interfaces, which will lead to the energy band bending, and the photogenerated e^- on the CB of OP and the photogenerated h⁺ on the VB toward RP will be combined under the action of the IEF. The h⁺ on the VB of OP and the e^- on the CB of RP will stay on the energy band due to the bending of the energy band, which promotes the separation of the photogenerated e^-/h⁺ pairs. In the photocatalytic process, oxidation-reduction reactions will be carried out by the e^- and h⁺ retained in RP and OP, respectively^[87]. Overall, S-scheme heterojunctions greatly enhance photocatalytic efficiency due to their strong oxidation-reduction capacity and unique carrier migration mode. However, currently, S-scheme heterojunctions are mostly focused on powder catalysts, and their kinetics need to be well studied and understood. Finally, Table 4 summarizes the 2020 to 2024 S-scheme TiO₂-based heterojunctions for photocatalytic hydrogen evolution.

Figure 9. (A) n-n junction, (B) p-p junction, (C) n-p junction, (D) p-n junction (E-G) Photogenerated carrier migration process and photocatalytic mechanism before and after contact. Quoted with permission from Li et al.^[87].

Other TiO₂-based photocatalytic heterojunctions

In addition to the common type-II, p-n, Z-scheme, and S-scheme heterojunctions discussed above, researchers have also devoted themselves to studying 2D van der Waals heterojunction formed by the IEF dominated by van der Waals forces, phase heterojunction formed by the same semiconductor that exists only in several different crystalline phases, facet heterojunction dominated by exposed surfaces, and Schottky heterojunction formed by semiconductor interacting with metal^[104]. However, these heterojunctions are mostly used for the degradation of organic matter and the photoreduction of CO₂, and fewer of them have been applied to the photocatalytic hydrogen evolution [Table 5], so they will not be specifically discussed in this review.

Application of DFT calculation to investigate TiO₂-based photocatalytic heterojunctions

In recent years, theoretical calculations based on DFT can provide accurate predictions of the electronic structure, optical properties, and photocatalytic activity of materials^[19]. The first is the binding energy of the heterojunction. In DFT calculations, the positive and negative values of the binding energy of the heterojunction can determine whether the two semiconductors need external energy injection in the coupling process. Next, the value of the binding energy can qualitatively determine the difficulty of forming heterojunctions for different semiconductors. Secondly, DFT calculations can confirm the amount of electron transfer in the heterojunction and the direction of charge transfer. Wang et al. constructed 0D/2D Co₃O₄/TiO₂ Z-scheme heterojunctions, and DFT calculations demonstrated that the work functions of Co₃O₄ and TiO₂ were 5.69 and 4.84 eV, respectively^[51]. The difference in the Fermi energy levels drove the charge transfer from TiO₂ to Co₃O₄. Moreover, the hybrid function calculations can accurately obtain the bandgap, HOMO, and LOMO of semiconductors. Density of states (DOS) can analyze the coupling of impurity orbitals in the forbidden bands of semiconductors after modification in a more detailed way and explore the essential reasons for the catalytic performance enhancement of semiconductor catalysts more precisely. For example, Li et al. constructed defective RuO₂/TiO₂ heterostructures [Figure 10A-C]^[107]. DFT calculations showed that Ru defects were surrounded by more positrons, which favors water decomposition, and the Gibbs free energy after the formation of the heterojunction decreased from 0.32 to -0.12 eV, which accelerates the separation of photogenerated e^-/h⁺ pairs by the interfacial effect of the heterostructures, decreases the adsorption energy of H₂, and accelerates the precipitation of hydrogen. Di Liberto et al.^[108]. proposed the SrTiO₃/TiO₂ [Figure 10D-E] heterostructure and found that the CB and VB of SrTiO₃ became more positive after the heterojunction formation and conformed to the type-II energy band arrangement with TiO₂ by hybrid functional theory calculations. Li et al. put forward a TiO₂/ZnS heterojunction and investigated the electronic properties and excitation electron dynamics heterostructure by using Vienna ab-initio simulation package (VASP) and NAnoscale Molecular Dynamics (NAMD)^[99]. It was found that [Figure 10F-H] the work functions of TiO₂, ZnS, and TiO₂/ZnS were 6.89, 5.74, and 5.80 eV, respectively. The differences in the work functions and the Fermi energy levels resulted in the transfer of e- from ZnS to TiO₂ until the charge equilibrium at the interfaces. The plane-averaged and differential charge densities demonstrated in Figure 10I proved the existence of strong interactions between the interfaces. In addition, [Figure 10J-K] the light absorption is significantly enhanced, and the STH conversion efficiency is as high as 23.46% both in acidic, neutral, and alkaline environments. Although DFT calculations have matured and provided much effective guidance for experiments, providing effective support for catalytic mechanisms, they still have many limitations. First, the bandgap of semiconductors is often underestimated in Perdew-Burke-Ernzerhof (PBE) calculations. While DFT + U, DFT-1/2 and Hybrid functional methods have been proposed to correct the bandgap value, they encounter multiple constraints, such as the lack of self-consistency and the fact that hybridized functional calculations are unsuitable for calculating crystal structures with a large number of atoms due to their high accuracy and computational effort^[109].

Figure 10. (A) D-RuO₂/TiO₂ nano-heterostructure, (B) Electron localization function analysis mapped, (C) ΔG_H of TiO₂, RuO₂/TiO₂, Pt/C and D-RuO₂/TiO₂. (A-C) is quoted with permission from Li et al.^[107] Band offsets of (D) SrTiO₃ and TiO₂ slabs (E) SrTiO₃/TiO₂ heterostructure. (D and E) is quoted with permission from Di Liberto et al.^[108] work function of (F) TiO₂(101), (G) ZnS (110), (H)TiO₂/ZnS. (I) Planar-averaged charge density difference for the TiO₂/ZnS heterojunction. (J) Absorption spectra, (K) STH efficiency of the TiO₂ (101), ZnS (110), and TiO₂/ZnS heterojunction. (F-K) is quoted with permission from Li et al.^[99].

On the other hand, with the rapid development of DFT calculation and machine learning (ML) in the field of materials, DFT calculation can predict new heterostructures, provide new heterostructures for experiments, and allow rational design of efficient photocatalytic heterojunctions. For example, Li et al. used it to obtain the Gibbs free energy of RuO₂/TiO₂ as only -0.1 eV and then guided the experimental synthesis of RuO₂/TiO₂ for photocatalytic hydrogen evolution^[107]. However, due to the large number of atoms (> 200) in TiO₂-based heterojunctions, the traditional DFT calculation for heterostructure prediction is highly costly and inefficient, and thus, less research has been carried out so far. Conversely, ML has the characteristics of high efficiency, strong model generalization ability, and self-learning ability, which solves the problem of low efficiency of DFT prediction, and can be used for high-throughput screening of heterostructures according to the set conditions^[6,108]. Currently, ML is in the stage of rapid development; in future research, it can reveal the physicochemical properties of materials, quickly screen the materials that meet the conditions for the construction of heterostructures, optimize the design of catalysts, quickly find the location of active sites and factors affecting the activity of catalysts, and modulate the composition of the material, so as to guide the experiments to synthesize the higher-efficiency heterostructures.