The interaction between surface atom and reactive molecule is crucial to the catalytic performance, which is dependent on the complex atomic arrangement and electronic structure of the surface and interface. However, using single component, it is hard to meet the above requirements. Therefore, interface engineering is used to develop heterogeneous catalysts by means of atom-site mixing of components (alloying strategy) or constructing special structures.

The most direct way to modify the adsorption behavior is to adjust the outer electronic structure of metal elements. Taking platinum group metals as example, the position of the d-band center reflects the adsorption strength of the platinum group metals. The upshift of the d-band center (i.e., close to the Fermi level) leads to a stronger binding, while the downshift of the d-band center (i.e., far away from the Fermi level) leads to a weaker binding [29]. Therefore, constructing disordered solid solution alloys and ordered intermetallic compounds using exogenous elements have proved to be an effective way to improve the performance of single-component catalysts by optimizing the binding energy [58].

At present, commercial Pt/C catalysts are still the most widely used catalysts for the HER. In order to improve the catalytic performance of Pt, a series of Pt-based alloy catalysts have been developed. Pt can be alloyed with different precious metals such as Pd [59], Ru [60], Au [61] and Ag [62], but the high cost is not favorable for large-scale commercialization. To address the cost problem, alloying Pt with non-precious metals such as Fe [63], Co [64], Ni [65], Cu [66] and Ti [67] has also been extensively investigated. The d-band center theory mentioned above shows that the distance of the d-band center relative to the Fermi level is closely related to the adsorption energy. Alloying Pt with transition metals (Co, Fe, Ni and Cu) will cause the d-band center of Pt to downshift and weaken the adsorption energy of oxygen-containing (OH*) on the surface Pt atom. Nørskov and co-authors have calculated a “volcanic” relationship between the d-band center of a bimetallic alloy and the chemisorption of catalyst surface molecules [29]. Wang et al. [68] prepared bimetallic nanoparticles with different compositions and structures, obtained dealloyed samples using pickling, and then prepared bimetallic samples with platinum “skin” using annealing. Taking the alkaline HER (in 0.1 M and 1 M KOH, respectively) as an example, the linear relationship between the d-band vacancy and the j₀ was established. The results showed that most of the alloy materials with a thin Pt skin showed strong catalytic activity, while the dealloyed PtCo and PtFe did not follow the linear relationship and showed higher activity. Selective optimization of the adsorption properties of H* and OH* intermediates is the key to improving the HER activity, and dealloying to enhance the surface atomic activity is undoubtedly a successful strategy.

Palladium (Pd) is widely found in the Earth’s crust and is about one-fifth of the price of Pt. Due to its similar atomic size and lattice mismatch of 0.77%, Pd is a potential alternative material for Pt [69]. In the latest research progress of Pd-based alloys, Qin et al. [70] introduced Ru to improve the HER catalytic activity in an alkaline electrolyte and prepared Pd₃Ru alloy nanoparticles enriched with Ru on the surface. The Ru atoms/clusters on the catalyst surface weakened the hydrogen bonding energy and promoted the adsorption of OH*, thus reducing the reaction barrier of the HER. The overpotential η₁₀ of Pd₃Ru alloy in a 1 M KOH solution is 104 mV and 6 mV lower than Pd and Pt, respectively. In addition, non-noble metals can also form alloys with Pd, such as Co, Bi, Ni and Cu [71]. Due to the synergistic effect between the PdNi alloy and carbon materials, the dispersion of the Pd-Ni alloy on carbon materials can improve the HER activity [72]. Alloying with different metals can move the d-band center of Pd downshift (i.e., away from the Fermi level). Alloying can adjust the single element to keep the appropriate hydrogen binding energy of the catalyst and make the intermediate state have the optimal Gibbs free energy, thus improving the HER activity.

In addition to Pt-based alloys and Pd-based alloys, Ir-based alloys have attracted attention in the field of water electrolysis due to their superior intrinsic catalytic activity. At present, Ir alloys with both noble and non-noble metals have been successfully designed, especially with non-noble metals (such as Fe [73], Co [74] and Ni [75]). Alloying not only ensures superior catalytic performance but also has a competitive advantage in cost. Taking IrFe alloy as an example, the overpotential is only 850 mV at the current density of 1000 mA cm⁻² in 1 M KOH, which is better than commercial Pt/C (η₁₀₀₀ = 1.17 V). Due to the difference in electronegativity, the incorporation of Fe significantly modulates the Fermi level of Ir through electron transfer, bringing ΔG_H* closer to 0 and thus enhancing the HER activity [73].

Fine-tuning the composition is also crucial to the catalytic activity of the alloy. Researchers usually choose economical iron elements to form an alloy with precious metals. Shan et al. [76] found that the strength of the HER catalytic activity of RuIr alloy doped with iron elements was Co-RuIr > Ni-RuIr > Fe-RuIr ≈ RuIr. In addition, some research groups have synthesized RhCo alloy nanosheets [77] and RuNi alloy multilayer nanosheets [78]. The RuNi alloy shows excellent HER catalytic activity (η₁₀ = 15 mV, Tafel slope: 28 mV dec⁻¹) in 1 M KOH, attributing to the optimization of H₂O dissociation and hydrogen adsorption and desorption during the HER process by RuNi alloying effect [78].

The alloy can reduce the cost of precious metals and improve the catalytic activity of precious metals, but the disordered structure of a solid solution alloy makes its stability poor. Intermetallic compounds mainly refer to alloy compounds that combine metals and metals or metals and nonmetals (such as H, B, N, S, P, C, Si, etc.) according to a certain atomic stoichiometric ratio to form a crystal lattice different from the original single component. Thermodynamically, solid solution alloys have a higher entropy due to their greatly disordered atomic arrangement, which is more favorable than ordered intermetallic compounds at high temperature. However, under certain compositions and temperatures, the strong d–d interaction between transition metal atoms can provide the required enthalpy change for the alloy system, resulting in a decrease in Gibbs free energy and forming ordered intermetallic junction physical properties. These strong d–d interactions have better stability than disordered solid solution alloys with similar composition and morphology, thus attracting much attention [79].

The synthesis of intermetallic compounds can be divided into two types: annealing after liquid phase synthesis and direct liquid phase synthesis. The difference between the above two methods is the annealing step and the formation mechanism of the ordered structure. The annealing after liquid phase synthesis needs high-temperature treatment under a reduced gas atmosphere. When the annealing temperature is higher than 500 °C, the size and shape of nanostructures are unevenly distributed during the transition from disorder to order, which makes it difficult to accurately control the size and shape of the nanostructures of intermetallic compounds [80]. The direct liquid phase synthesis needs a seed to grow material. During this process, it is hard to control the size of the catalyst. Therefore, developing a synthetic strategy that can accurately control the size and shape of intermetallic compounds is a challenge. Kim et al. [81] used a mesoporous silica template to synthesize a well-defined nanomaterial through nanomaterial space limiting. This method allows precise control of the size and shape of nanostructures by converting disordered alloy Pt₃Co nanowires (D-Pt₃Co-NWs) into ordered intermetallic compound Pt₃Co nanowires (O-Pt₃Co-NWs) without agglomeration. O-Pt₃Co NWs have higher HER activity than O-Pt₃Co NWs and Pt/C catalysts in alkaline conditions. The enhancement of the O-Pt₃Co NWs HER kinetic is achieved through alloying and atomic ordering regulation of the hydrogen binding energy.

Although most research on intermetallic compounds has focused on materials with polymetallic components, some researchers have focused on alloying products formed between metals and non-metals. For example, Ai et al. [82] studied the critical role of electron interaction between metal and boron in surface hydrogenation adsorption and catalytic activity using a combination of theory and experiments. The result shows that the adsorption of hydrogen atoms on the surface of intermetallic compounds is weaker than that on the corresponding pure metal surface. This is due to the strong hybridization between the d-orbital of transition metal and the sp-orbital of boron, which changes the d-band properties. By calculating the ΔG_H* on the surface of boron-containing intermetallic compounds (TMB) of hydrogen atoms, several extremely active hydrogen evolution catalysts (e.g., PdB, RuB) of TMB types are predicted.

In order to better understand and compare these precious metal alloy catalysts, Table 1 lists the essential parameters for evaluating their HER properties.

In addition to the alloying strategy, interface engineering can modify the geometric structure of catalytic materials, reasonably control the atomic arrangement of the surface or interface, optimize the adsorption capacity of reactants, intermediates, and products and effectively improve the efficiency of electrocatalysis. Based on noble metal nanomaterials, this part will be discussed according to the interfacial geometry of the catalyst. Firstly, the coated core–shell nanostructured precious metal particles are introduced, then some typical 2D structure-supported precious metals are introduced, then 1D/3D structure-supported noble metals are introduced, and finally, special atomic interface materials such as single atoms are introduced. The above hybrid nanostructures can be constructed from different materials, such as metal–metal, metal–oxide/hydroxide, metal–sulfide, a metal–metal organic framework, metal–MXene and metal–carbon.

Catalysts with core–shell structure are usually composed of core (one composition) and shell (another composition). For the bimetallic compounds of the core–shell structure, their formation is driven by the difference in the surface free enthalpy between the two metals [97,98]. Usually, the component with less enthalpy is going to be the shell. Because the shell has more exposed surface than the core, it needs higher energy. Among them, the precious metal components can exist in the shell to improve the atomic utilization. The noble metal shell can enhance the catalytic activity through electronic coupling with components in the core at the interface or through the strain effect. The common synthesis method of the core–shell structure is to preferentially grow the seed core, and then the noble metal shell grows on the core with chemical reduction or thermal decomposition. At present, numerous efforts have been made in the preparation of core–shell catalysts to achieve the controlled synthesis of composition, morphology, size and shell thickness. In addition to changing the composition, 0D, 1D and 2D core–shell structures have been developed by controlling the morphology.

The 0D core–shell nanostructure mainly consists of nanoparticles and nanopolyhedra. Wang et al. [99] studied Ru-Pt model catalysts with the same ligand effect but different surface geometry. The Ru@Pt core–shell structure with a strain interface has a higher activity than a strain-free alloy (RuPt). The enhancement of HER activity induced by strain in an alkaline electrolyte was larger than that in an acidic electrolyte, ascribing to a more significant interaction between the catalytic intermediates and OH^- caused by compressive strain. Therefore, the basic HER activity can be improved by the strain at the core–shell interface.

The 1D core–shell nanostructure mainly contains nanorods, nanowires and nanoribbons. Zhang and co-works [83] used the Au nanoribbons as templates, on which other noble metals such as Ag, Pd, and Pt with lattice mismatch less than 5% were epitaxially grown. They then obtained a series of core-shell nanoribbons, including Au@Ag, Au@PdAg, Au@PtAg, and Au@PtPdAg. Among them, Au@PdAg nanoribbons show excellent HER electrocatalytic activity (η₁₀ = 26.2 mV, Tafel slope = 30 mV dec⁻¹) and durability (nearly no loss of catalytic activity after 10,000 cycles). Its excellent catalytic activity is due to its abundant atomic steps, nanodendritic surface, multi-component synergistic effect of the shell and distinct crystal structure. This classic work provides a current strategy for the controlled synthesis of the crystal structure of noble metal nanomaterials. In addition, Luo et al. [84] synthesized mesoporous Pd@Ru core–shell bimetallic nanorods composed of fcc Pd and hcp Ru. The DFT calculation shows that Ru/Pd (111) interface structure enhanced the alkaline HER activity (η₁₀ = 30 mV in 1.0 M KOH).

The 2D core–shell nanostructure is mainly covered with nanoplates and nanosheets. Ru-based nanoplates have attracted extensive attention in recent years due to their distinct 2D structure and excellent catalytic performance. However, the synthesis of Ru-based nanoplates remains a challenge, especially the controlled synthesis of 2D structures with atomic thickness. Han et al. [93] successfully synthesized core–shell Pd@Ru nanoplates with fcc structure with a simple solvothermal method using Pd nanoplates as the substrate. The thickness and crystal structure of the Ru shell can be adjusted by merely changing the amount of Ru precursor. With the increase in the Ru shell thickness, the HER activity of Pd@Ru firstly increased and then decreased. This is because, with the increase in the Ru shell thickness, its crystal structure changed from the fcc to the hcp phase.

In fact, shell thickness is critical to catalytic activity. The electron coupling and strain effects are different between the shell and core with different thicknesses. A shell with the thickness of a single atomic layer has the highest utilization rate. However, it is not a balanced and stable configuration, and adverse diffusion will occur on the surface atom after long-term operation or at high temperatures. Therefore, adjusting the thickness of the shell accurately requires higher requirements for the structure synthesis and interface structure.

The ultrathin 2D nanomaterials have attracted broad attention since Novoselov et al. [100] mechanically separated graphene from graphite in 2004. In recent years, researchers have developed many ultrathin 2D materials, such as transition metal dichalcogenides, metal-organic frameworks, hexagonal boron nitride, graphite carbon nitride, black phosphorus, MXenes, layered metal oxides and layered double hydroxides [101].

The electronic confinement of ultrathin 2D nanomaterials without interlayer interaction, especially single-layer nanomaterials, gives them better electronic properties than other nanomaterials. In addition, their large lateral size and atomic layer thickness give them an extremely high specific surface area, exposing more surface atoms, which is ideal for expanding their applications in the field of high surface activity [103,104]. Based on the above advantages, researchers choose 2D materials as the substrate for noble metals, which can effectively prevent the agglomeration of noble metals in the catalytic process and improve the catalytic performance by exposing the surface active sites.

TMDs are an essential semiconductor material, whose chemical composition is MX2 (M is the transition metal, X is S, Se or Te), and the 2D structural units (such as S-Mo-S) are combined with each other through van der Waals forces, which has broad application prospects in electrocatalysis. At present, noble metal nanoparticles such as Pt, Pd, Ru and Au have been grown on TMDs (such as MoS₂, MoSe₂, WS₂, TiS₂, TaS₂, VS₂ and VSe₂) for the HER [105,106]. For example, Huang et al. [85] synthesized epitaxial-grown Pt-MoS₂ composites using the wet chemical method. Under the same Pt loading, the material showed better HER electrocatalytic activity than the commercial Pt catalyst, mainly due to the strong interaction between MoS₂ nanocrystals and Pt nanoparticles, regulating the electronic state of the Pt nanoparticles (Pt⁰ → Pt^δ+).

BP has a layered structure, and the layers are combined by van der Waals forces. When BP and other materials (such as Ni₂P [107], Co₂P [108] and MoS₂ [109]) are assembled into heterostructures, the result can promote the performance of electrocatalytic HER. In addition, BP can also be used as a carrier to regulate the electronic structure of noble metals and synergistically promote the HER performance. Li et al. [94] synthesized the hybrid structure of PtRu nanoclusters (NCs) and BP nanosheets, which showed 10.2 times higher HER activity (η₁₀ = 22 mV; Tafel slope = 19 mV dec⁻¹) than that of commercial Pt/C in 1 M KOH. The DFT calculations showed that the electronic synergy effect resulting from the strong electron coupling between BP nanosheets and PtRu nanoclusters accelerated the water dissociation, optimized the adsorption strength of H* and enabled the PtRu NCs/BP hybrid material to have high HER activity.

As typical 2D functional materials, MXene has attracted wide attention in the fields of renewable energy and catalysis [110]. They are a large class of 2D transition metal carbides/nitrides derived from the selectively etched layered MAX phase [111]. MXene can generally be expressed as M_n+1X_nT_x (n = 1–3), where M is the transition metal (e.g., Ti, Mo, Nb, Ta or V), X is the C and/or N elements and T represents the surface groups, such as -OH, -O, and -F [112]. MXene has abundant surface chemical properties, high hydrophilicity, mechanical stability, metal conductivity related to high electron density near the Fermi level and other favorable properties [113,114]. For example, Cui et al. [86] mixed a H₂PtCl₆ solution with an MXene solution to form MXene@Pt, and adopted single-walled carbon nanotubes (SWCNTs) as a binder and fluid collector to obtain a layered Pt-MXene-CNT heterostructure. The overpotential is 62 mV at 10 mA cm⁻² in 0.5 M H₂SO₄, and the Tafel slope is 66.6 mV dec⁻¹. The abundant negatively charged groups on the surface of MXene are conducive to the adsorption of noble metal cations and stabilize the reduced noble metal nanoparticles, effectively preventing their aggregation.

The general formula of LDHs is [M_1−x²⁺M_x³⁺(OH)₂]A_x/nⁿ⁻·mH₂O, which is composed of a positively charged brucite layered body and exchangeable charge-balanced interlayer anions. A divalent metal ion (e.g., Mg²⁺, Fe²⁺, Co²⁺, Cu²⁺, Ni²⁺ or Zn²⁺) is partially coordinated by a hydroxyl octahedron, which can be replaced by a trivalent metal ion (e.g., Al³⁺, Cr³⁺, Ga³⁺ or Mn³⁺) in a brucite layer. One example of the charge-balanced anion is CO₃²⁻. mH₂O represents the interlayer water molecules [115]. LDHs have attracted extensive attention in the field of electrocatalysis due to their adjustability of metal cations, interlayer anion exchangeability and easy stripping into monolayer nanosheets. However, their poor electrical conductivity restrict their large-scale applications [46]. Researchers have tried to solve this problem by mixing LDH with conductive materials, growing LDH directly on conductive materials (carbon nanotubes, graphene, etc.) or electrode substrates such as nickel foam and carbon fiber cloth [102,116]. For example, Yan et al. [95] grew NiFe LDH in situ on carbon cloth (CC) decorated with ultrafine Pt sub-nanoclusters (average size of 0.59 nm). Highly evenly dispersed Pt can expose more active sites, shorten the electron transport path and considerably reduce the amount of Pt. In addition, the strong interaction between Pt and 2D NiFe LDH can effectively prevent the aggregation of Pt sub-nanoclusters.

Metal–organic framework (MOF) materials consisting of central metals, and organic ligands are receiving increasing attention. In particular, 2D MOF nanosheets have become one of the most popular materials in current research, due to their large specific surface area, high proportion of exposed metal atoms, porous structure and adjustable surface functional groups [117]. By introducing additional substrate to construct 2D MOF-based nanocomposites, its intrinsic catalytic activity can be further improved and the feasibility of material application can be expanded [118]. For example, Guo et al. [96] prepared heterogeneous materials (Pt-NC/Ni-MOF) of 2D nickel MOF(Ni-MOF) and Pt nanocrystalline (Pt-NC). The Ni-O-Pt bond formed at the interface regulated the electron distribution and optimized the adsorption of H* and OH* (H* adsorption energy decreased, and OH* adsorption energy increased). The reaction kinetics of the HER was accelerated. The mass activity of the composite was 7.92 mA μg⁻¹_Pt at the overpotential of 70 mV, which is one of the best catalysts in alkaline medium reported so far.

In addition to the above materials, there are some other 2D materials such as graphene and its derivatives, including graphene oxide (GO) and reduced graphene oxide (rGO), which have the characteristics of large specific surface area, high carrier mobility and good stability. They are ideal substrates for anchoring functional nanomaterials and are conducive to improving the electrocatalytic activity of materials [119]. For example, Cheng et al. [120] prepared a composite material with defective graphene-supported Pt clusters using the chemical deposition method. The obtained catalyst had favorable HER performance, and its mass activity increased by 11.1 times compared with commercial Pt/C catalyst. In addition, some carbon and nitrogen compounds, such as C₂N [87] and C₃N₄ [88], are also excellent materials for Ru, Pt and Pd catalysts supported.

In summary, 2D materials with rich types, high specific surface area and abundant surface functional groups can not only prevent particles from aggregation but also produce synergistic effect with precious metals when they serve as the supporting substrate of precious metals, thus improving the HER catalytic activity and stability.

One-dimensional materials have a high surface area, high roughness factor and high density of active sites. There is a large amount of open space and porosity between adjacent 1D nanostructures, which enables the rapid mass transfer of electrolyte molecules and full contact between chemical substances and the electrode/catalyst surface [121,122]. Recently, Kweon et al. [89] reported that uniform deposition of ruthenium (Ru) nanoparticles on multiwalled carbon nanotubes (Ru@MWCNT) is an efficient HER catalyst. At a current density of 10 mA cm⁻² in 0.5 M H₂SO₄ and 1 M KOH, Ru@MWCNT exhibits an ultralow overpotential of 13 mV and 17 mV, respectively, exceeding commercial Pt/C (16 mV and 33 mV). The Faraday efficiency (92.28%) was higher than that of the benchmark Pt/C (85.97%) in the actual water electrolysis cell. The DFT calculation shows that the Ru-C bond is the most reasonable active site with the most suitable hydrogen binding energy. Three-dimensional materials can be assembled from one-dimensional or two-dimensional materials. The 3D network has superior mechanical stability and effectively prevents the aggregation and restacking of materials. Moreover, the greatly interconnected structure and porosity of 3D materials make the inner surface of the materials thoroughly utilized and the mass transfer becomes more favorable. Based on these excellent properties of 3D materials, researchers assemble some 1D (such as carbon nanotubes) or 2D (such as graphene, MXene, etc.) materials together to form 3D structures, effectively improving the stability of materials. Kumar et al. [123] prepared noble metals (Pt, Pd, Au and Ag)/3D graphene (3D-G) nanocomposites and investigated their catalytic activity of the HER. The high conductivity and porous network of 3D graphene sponges provide favorable charge transfer and ions diffusion, further improving the HER catalytic activity and stability. Xiu et al. [90] prepared 3D multistage hollow MXene-loaded Pt nanoparticles (Pt@mh-3D MXene) and 2D MXene-loaded Pt nanoparticles (Pt@2D MXene). The 3D hollow structure can be regarded as the high curvature seal of 2D nanosheets around a well-defined pore space. These MXene are synergistically coupled with ultrafine Pt to form a unique multifunctional catalytic interface design. In particular, Pt@mh-3D MXene not only has the high stability and atomic utilization of Pt but also comprehensively enhances the H* binding capacity, charge transfer capacity and ion/substance transport capacity, promoting HER catalytic activity in the whole pH range. After the optimization, Pt@mh-3D MXene has better HER performance (η₁₀ = 13 mV, Tafel slope = 24.2 mV dec⁻¹) than Pt@2D MXene (80 mV, 66.6 mV dec⁻¹) in 0.5 M H₂SO₄. Compared with industrial 20% Pt/C in 1 M KOH, the mass activity and durability are improved by 20 times, while the Pt usage is reduced by 8.3 times.

Single-atom doping plays an essential role in catalytic reactions by maximizing atomic efficiency and activating catalytic sites. Atomically dispersed metals are usually anchored to a variety of substrates, including graphene, metal–organic skeleton-derived porous carbon, metal oxides and zeolites, which in most cases can provide pores, vacancy defects or strong interactions [14]. At present, single-atom HER catalysts of the noble metals Pt, Pd, Ru and Ir have been reported, and these single atoms are fixed by coordination with C, N, P, and S atoms on the substrate [124,125]. The reason for improving the activity of a noble metal single-atom catalyst is not only to maximize the utilization of atoms but also to optimize the electronic structure of the noble metal single atom and its heteroatoms by bonding and interacting with the coordination atoms on the substrate, resulting in ΔG_H* being closer to 0 [124]. The coordination structure has a great effect on the catalytic performance. However, it remains a big challenge for researchers to elaborate the design of the coordination number without changing the atomic dispersion [126,127]. Yang et al. [91] prepared the Ru SAs@PN catalyst using amorphous phosphoimide nitrides nanotubes (HPNs) as the substrate of the stabilized Ru single atoms (SAs). There is a strong coordination interaction between the lone pair of electrons of N in the HPN substrate and the d-orbitals of Ru, thus anchoring the Ru single atoms. The Ru SAs@PN catalyst shows high HER catalytic activity (in 0.5 M H₂SO₄, η₁₀ = 24 mV). The DFT calculation shows that the ΔG_H* of Ru SAs@PN is closer to 0 than that of Ru/C, C and C₃N₄-loaded Ru SAs, which is conducive to hydrogen precipitation. However, atom-anchored materials still face a great challenge, namely the difficulty of achieving large-scale stable SAs doping at high loads, which tend to aggregate and agglomerate to form particles. Feng et al. [12] reported a new concept of selective single-atom doping with high loads through lattice mismatch of multicomponent heterogeneous nanostructures (e.g., NiS@Al₂O₃ [12], MoS₂/NiS₂ [92], etc.). Due to the special properties of heterogeneous nanostructures, a large number of vacancy defects or voids are generated on the heterogeneous interface so that more atoms can be trapped.