Recent advances in defect-engineered molybdenum sulfides for catalytic applications

Yunxing Zhao ^ab, Xiaolin Zheng *^c, Pingqi Gao *^a and Hong Li *^bde
aSchool of Materials, Sun Yat-sen University, Guangzhou 510275, China. E-mail: gaopq3@mail.sysu.edu.cn
^bSchool of Mechanical and Aerospace Engineering, Nanyang Technological University, 639798, Singapore. E-mail: ehongli@ntu.edu.sg
^cDepartment of Mechanical Engineering, Stanford University, California 94305, USA. E-mail: xlzheng@stanford.edu
^dCINTRA CNRS/NTU/THALES, UMI 3288, Research Techno Plaza, 637553, Singapore
^eCentre for Micro-/Nano-electronics (NOVITAS), School of Electrical and Electronic Engineering, Nanyang Technological University, 639798, Singapore

First published on 5th July 2023

---

---

Wider impact Molybdenum sulfide catalysts have gained widespread acceptance in industrial petroleum hydrodesulfurization, and they are expanding to the electrochemical reactions. Recent advances have witnessed defect engineering as a powerful tool to effectively activate structures. Herein, we systematically summarize the recent advances in defective molybdenum sulfides for catalytic applications. Compared to previous articles centered around a single application with limited defect types, this article reviews the various defect types of molybdenum sulfides for a range of catalytic applications, emphasizing the correlation between the defect structure, theoretical mechanism, and performance. Besides elaborating the defects in crystals, this article also fills a gap in similar reviews by systematically addressing defects in both amorphous and clustered molybdenum sulfides. The review is highly interdisciplinary, relating to materials science, electrochemistry, thermal catalysis, heterogeneous catalysis, semiconductor materials, and nanotechnology; so, we anticipate that this review will be of great interest to a wide variety of researchers with extensive backgrounds. Moreover, the focus of this review is on defects in molybdenum sulfide, where defect screening, structural design, fabrication methods, and characterization techniques are important elements in catalysis and materials science, and thus this review contributes to the advancement of materials methodology and the development of defect-based industrial catalysts.

1. Introduction

Molybdenum disulfide (MoS₂) has gained great popularity as a well-established industrial catalyst for the petroleum hydrodesulfurization (HDS) reaction for oil upgrading, and it is hard to prevent its aggressive expansion into cutting-edge applications in electrochemical catalysis. With its high abundance (10000 times that of platinum) and excellent hydrogen evolution reaction (HER) catalytic properties, MoS₂ has been emerging as one of the most promising HER catalysts to replace platinum (Pt) in proton exchange membrane (PEM) electrolyzers, signifying its potential value in the field of hydrogen energy. Moreover, MoS₂ extends its catalytic applications in other reactions such as the nitrogen reduction reaction (NRR), carbon dioxide reduction reaction (CO₂RR), oxygen reduction reaction (ORR), oxygen evolution reaction (OER), nitric oxide (NO) reduction reaction, nitrate ion (NO₃⁻) reduction reaction, and organic molecule conversion. MoS₂ also showcases advantages as an electrode material with catalytic functions in energy storage devices such as metal–sulfur batteries and metal–oxygen/air batteries. Broader applications based on MoS_x catalysts are yet to be discovered. Bearing this in mind, this paper will review the catalytic applications of defective MoS₂ materials in two-dimension (2D) comprising 2H, 1T, and 3R phases and non-2D (amorphous and cluster phases) morphologies, mainly for the HER, as well as for the NRR, the CO₂RR, metal–sulfur batteries, metal–oxygen/air batteries and HDS reactions. Before specifically reviewing its applications in catalysis, we will give a brief overview of the structural classification and catalytic applications of MoS_x.

1.1. Introduction of molybdenum sulfide

2D MoS₂ crystals are constructed by stacked layers, in which an infinitely extended planar monolayer composed of Mo and S elements binds to adjacent layers with interlayer van der Waals forces. Three crystalline phases including phases 2H, 1T, and 3R are defined by different periodic arrangements within and/or between the layers and their unit cell structures are shown in Fig. 1a, where 1, 2, and 3 layers are displayed in the unit cell for phases 1T, 2H, and 3R, from left to right, respectively.¹ For the 2H phase, the adjacent layer is rotated 180° and stacked in a direction perpendicular to the expanded basal plane; and for the 3R phase, the adjacent layer is slightly displaced from each other; while for the 1T phase, the adjacent layer remains unchanged. As summarized in Table 1, 2H and 3R phases possess the same trigonal prismatic coordination but with different hexagonal and rhombohedral symmetries, as well as distinct point and space groups, while the 1T phase has an octahedral coordination with trigonal symmetry along with point group D_3d and space group Pm1. Furthermore, in terms of stability, the semiconducting 2H phase is dominant in natural minerals, followed by the semiconducting 3R phase and its mixture, however, the metallic 1T phase could not be found in nature on account of its metastable properties in thermodynamics, transforming into 2H or 3R phases when heated above 100 °C or just upon aging.²

---

With respect to its electronic structure, different atomic arrangements of the MoS₆ unit lead to distinct splitting of Mo 4d orbitals according to crystal field theory. In the case of 2H or 3R phases, the d orbitals of the D_3h-MoS₆ unit split into three groups including d_z², d_xy and d_x²–y², d_xz and d_yz; and two electrons occupy the d_z² orbital (Fig. 1b). In contrast, the d orbitals of the O_h-MoS₆ unit in the 1T phase split into triply degenerate t_2g (d_xy, d_xz, d_yz) and doubly degenerate e_g (d_x²–y², d_z²) orbitals, with two electrons occupying two out of the three t_2g orbitals without considering the Jahn–Teller effect.^2–4 Engineering of unoccupied orbitals by injecting extra electrons from metals (e.g., Li/Re/Mn) can be used in a wide range of fields, such as crystal phase tuning, ferromagnetism, photocatalysis, superconductivity, etc.^2–7 The band structure is formed by the hybridization of Mo and S states, where the Mo 4d orbital contributes to the upper valence band edge and lower conduction band edge while the S 3p orbital mainly contributes to the valence band. For the 2H phase, the valence band maximum (VBM) is located at the Γ point in the Brillouin zone, but the conduction band minimum (CBM) is located approximately halfway along the Γ–K direction (Fig. 1c), which leads to an indirect band gap of 1.29 eV in bulk 2H-phase MoS₂. In contrast, the Mo 4d states form a broad band hosting the Fermi level in 1T-phase MoS₂, suggesting its metallic behavior.^3,8–12 Furthermore, as the number of layers decreases, the band gap increases (Fig. 1d).¹³ Meanwhile, the valence and conduction band states near the Γ point are strongly influenced by the interlayer interactions because of the large contribution of p orbitals of the S atoms to the electronic states, while the states near the K point are marginally affected because they consist mainly of localized d orbitals of Mo atoms, leading to an indirect-to-direct band gap transition of MoS₂ from the multilayer to monolayer,^13–15 and the monolayer MoS₂ with a direct band gap of ∼1.8 eV can be utilized as an advanced material for optoelectronic applications such as phototransistor and field-effect transistor.^16,17

We here classify 2H-, 1T-, and 3R-phase crystalline MoS₂ as 2D molybdenum sulfides because they are usually prepared in several layers to ensure a larger specific surface area when used as catalytic materials. We classify amorphous and clustered MoS_x as non-2D molybdenum sulfides because they do not consist of atomically thin layered units. We consider a perfect MoS₂ crystal with n (n ≥ 1) monolayers stacked according to the standard crystal pattern. In the perfect crystal, the defective structure is absent, e.g., no difference in the number of stacked layers, only an infinitely small ratio of edge positions, without discontinuous layers, no voids, lack of distortion, and the absence of dislocations. In fact, edge defects, atomic vacancies, heteroatom adsorption sites, and other types of defects are usually present in natural MoS₂ crystals. From the perspective of controllable defect modulation, we focus here on artificial preparation of defect types, including porous structure formation, atomic doping, designing strained structures, dislocation and distortion formation, atom vacancy formation, alloying, hybridization, phase modulation, unsaturated coordination, bonding regulation, surface decoration, heterointerface formation, etc.

The electronic structure of MoS₂ can be tuned by internal defects, such as grain boundaries (Fig. 1e) and S-vacancies (Fig. 1f), both of which can introduce mid-gap states that altering the optical, transport, or adsorption properties.^18,19 The electronic structure can be further tuned by external factors, for instance, via applying elastic strain on monolayers to obtain a narrower energy gap and decreased exciton energy with increasing tensile strain (Fig. 1g), providing opportunities for next-generation optoelectronic or photovoltaic technologies.²⁰ In addition, a sizable spin–orbital splitting occurs in monolayer MoS₂, which breaks the degeneracy of the valence band maximum (VBM) and the conduction band minimum (CBM), displaying its potential for application in spin electronics and quantum information fields.¹²

1.2. Catalytic application of molybdenum sulfide

2D MoS₂ has already been established as an industrial desulfurization catalyst for gasoline production and is being actively extended to electrocatalytic applications. With 2D MoS₂ as an example, we briefly illustrate the electrocatalytic applications of molybdenum sulfide. Electrocatalysis comprises a transport process, which relies on electrons and holes in an external circuit and ions in the solution to transfer electrons/holes to the adsorbent before obtaining the target chemicals through the oxidation reaction at the anode and the reduction reaction at the cathode; and a catalytic process, which improves the conversion efficiency of electrical energy to chemical energy by overcoming the kinetic barrier. With the development in the field of electrocatalysis in the last decade, MoS₂ has shown great potential for catalytic applications in the HER, OER, ORR, CO₂RR, NRR, nitric oxide (NO) reduction reaction, nitrate ion reduction reaction, organic small molecule conversion, and so on, by virtue of its excellent performance, decent electrical conductivity and stable structure in harsh environments.^19,21–27Table 2 summarizes a few representative applications of MoS₂ in the HER, OER, ORR, CO₂RR, and NRR including half-reaction, equilibrium potentials E⁰, media, and catalysts. MoS₂ has shown excellent catalytic performance for the HER in acidic media, but encounters difficulties in catalyzing the HER and the OER in basic media due to its inferior activity compared to Fe–, Co–, and Ni-based materials. Fortunately, engineering of intrinsic MoS₂, such as carbon doping, substitutional doping of Ru, construction of heterostructures, and design of biaxially strained MoS₂ nanoshells, can generate competitive activity for the catalytic HER in alkaline media.^28–31 Fabrication of quantum dots or construction of hetero-interfaces renders MoS₂ enhanced OER activity.^21,32 Moreover, it also exhibits remarkable activity towards the CO₂RR compared to Ag particles, delivering a much higher current density of 65 mA cm⁻² at −0.764 V_RHE and an exceptional faradaic efficiency (FE) up to 98%.²³ With regard to the NRR, intrinsic MoS₂ gives a NH₃ yield of 8.08 × 10⁻¹¹ mol s⁻¹ cm⁻² at −0.5 V_RHE in 0.1 M Na₂SO₄ with a FE of 1.17%; and further optimization of the catalytic system, e.g., upgrading electrolytes or making composites, could yield an order of magnitude increase in activity and selectivity.^33,34 In addition, defective MoS₂ is also very useful when serving as a catalyst for batteries. Being an electrode material, it can catalyze the conversion of Li₂S_n (3 ≤ n ≤ 8), Na₂S_n (4 ≤ n ≤ 8), Li₂O₂ and O₂/OH⁻, in Li–S, Na–S, Li–O₂ and Zn–air batteries, respectively.^32,35–37

---

Many studies have focused on defective MoS₂, especially on the methods of defect generation (i.e., direct synthesis of defective MoS₂ and creation of defects based on the as-prepared crystals) and on the understanding of the effects of defects on activity and stability, to achieve controlled design, fabrication, and use of defective MoS₂ towards catalysis. As a special class of catalytically active sites, defects lead to very different properties from conventional crystals due to their structural diversity, high distortion energy, and complex atomic arrangement, which are significant for catalysis. It should be noted that we focus on reviewing the catalytic applications of defective MoS_x materials in 2D (2H, 1T, and 3R phases) and non-2D (amorphous, cluster) morphologies. Fig. 2 summarises the catalytic applications of various defective MoS_x materials in the HER, NRR, CO₂RR, metal–sulfur batteries, metal–oxygen/air batteries, and HDS reactions, and provides navigation for subsequent discussion. Among these electrocatalysis fields, MoS₂ has made the most successful breakthrough in acidic HER applications from our perspective, paving the way for the development of platinum-free catalysts to replace the precious platinum in PEM electrolyzers. We will therefore firstly detail the progress of MoS₂-based electrocatalysts for HER application and then we discuss the catalytic performance of defective MoS₂ in other electrocatalytic fields, including the NRR, CO₂RR, metal–sulfur batteries, and metal–oxygen/air batteries and finally we will briefly review its application in hydrodesulfurization for petroleum upgrading.

1.3. Preparation of defective molybdenum sulfide

The preparation method of defects determines their structure and thus plays a key role in the catalytic performance. Various novel defect preparation methods have emerged over a long period of time, and the design ideas behind them are worth reviewing. Controlled preparation methods confer reproducibility to the material, which is crucial for its practical applications. In this section, we provide a brief review of several methods for the preparation of defective molybdenum sulfide based on some representative cases.

The porous structure breaks the continuity of the material and provides a large number of defect sites such as edge and unsaturated coordination atoms. Abundant porous structures can be prepared using a template method, chemical exfoliation method, etc. For instance, Skrabalak and coworkers prepared porous nanostructured molybdenum sulfide using an ultrasonic spray pyrolysis (USP) method.³⁸ Specifically, they obtained a MoS₂/SiO₂ composite by nebulizing and pyrolyzing a mixture of ammonium tetrathiomolybdate and colloidal silica and then cleaned the SiO₂ template with HF to obtain a highly porous MoS₂ network with a surface area of up to 250 m² g⁻¹. Yin and coworkers employed a liquid-ammonia-assisted lithiation (LAAL) strategy to exfoliate 2D-MoS₂.³⁹ This process includes the lithiation reaction that occurs when liquid ammonia contacts with lithium, the desulfurization reaction between lithium and MoS₂, and the exfoliation of Li-intercalated MoS₂ by ultrasonication in water. The sample obtained using this method is mesoporous 1T-phase MoS₂ nanosheet (P-1T-MoS₂), which can be converted into P-2H-MoS₂ by annealing. In addition to the pore structure, P-1T-MoS₂ also has a plenty of edge sites and numerous S-vacancies and the contents of S-vacancies gradually increase with the ratio of lithium to MoS₂.

For the fabrication of S-vacancies in crystalline MoS₂, breakthroughs have been achieved using physical or chemical methods, with the former typically involving direct disruption of the Mo–S bond by kinetic energy and the latter allowing S removal by altering electron transfer or chemical bonding. Plasma treatment techniques have been extensively investigated for the preparation of S-vacancies on the basal plane of MoS₂. For instance, Li and coworkers created S-vacancies by exposing CVD (chemical vapor deposition)-synthesized monolayer 2H-phase MoS₂ to a mild argon plasma atmosphere, where the plasma was generated by dispersing small RF power in a vacuum chamber, which results in a short mean path for the radicals in the plasma and thus ensures a cold and mild Ar plasma on the MoS₂ surface, thus making the desulfurization process more controllable.¹⁹ Different vacancy concentrations can be realized by the duration of the treatment. Wu and colleagues used a chemical route to prepare MoS₂ with in-plane S-vacancies.⁴⁰ Commercial bulk MoS₂ and Zn powder were mixed by grinding, then dried and degassed under an inert atmosphere and subsequently heated at a high temperature (700 °C). During the heating process, Zn as a reductant reacted with MoS₂ to produce Mo^d+ (d < 4), thereby excluding S from the basal plane and producing defective MoS₂ nanoparticles of about 25 nm. By adjusting the ratio of Zn to MoS₂, the density of S-vacancies and the electrical conductivity can be tuned simultaneously. Additionally, more chemical pathways for creation of S-vacancies have been investigated because of their advantage in large-scale production of defective MoS₂.

Strain in MoS₂ introduces many defects, such as dislocations or mismatch of atoms within the layer, folding or warping of the two-dimensional layer structure, resulting in the modulation of the electronic structure as well as the catalytic properties. The generation of strain depends on force effects such as substrate-induced strain,¹⁹ lattice mismatch-induced strain,⁴¹ and shaping-induced strain.⁴² For instance, Li and coworkers applied strain to the MoS₂ monolayer based on a nanostructured substrate.¹⁹ Specifically, the CVD-grown MoS₂ monolayer was wet-transferred onto a SiO₂ nanocone substrate pre-patterned using nanosphere lithography. After removing PMMA with chloroform, the SiO₂ nanocone with transferred MoS₂ was soaked in ethylene glycol in vacuum environment, ensuring that both sides of the MoS₂ sheet were immersed. The sample was then dried in ambient air to evaporate ethylene glycol, when the capillary force pulled the MoS₂ against the SiO₂ nanocone. The MoS₂ region on the tip of the nanocone is most strained and the region between nanocones is less strained.

Doping defects include substitution doping, interstitial doping, surface doping, etc., which regulate the electronic structure, electrical conductivity, and defect density of catalytic materials through the synergistic effect between elements. Although there are many methods of doping, such as the hydrothermal method, co-precipitation, CVD method, etc., they can be simply divided into in situ doping and post-synthesis doping according to the sequence of the doping procedure. For example, Zhang and colleagues employed a one-step hydrothermal method at 220 °C to prepare Ru, O-co-doped MoS₂ using H₃₂Mo₇N₆O₂₈, CH₄N₂S, and RuCl₃ as precursors, where the ruthenium content was adjusted by the ratio of the feed.⁴³ This method yielded molybdenum sulfide with a nanoflower structure, in which Ru and O are doped into the lattice. Sun and colleagues used a two-step method to achieve post-synthesis doping of molybdenum sulfide.⁴⁴ Briefly, Mn-doped MoS₂ nanosheets (Mn–MoS₂) were first synthesized in a hydrothermal reactor at 180 °C using ammonium molybdate, manganese chloride, and thioacetamide as precursors. Subsequently, N and Mn-co-doped MoS₂ nanosheets (N, Mn–MoS₂) were obtained by treating Mn–MoS₂ in a tube furnace under a NH₃ atmosphere at 300 °C.

MoS₂ with mixed crystalline phases forms an abundant phase heterojunction that differs from single-phase crystals in the electronic structure and increases the density of defect sites, offering an opportunity to reduce the catalytic kinetic barrier and to increase the catalytic activity. Both top-down exfoliation and bottom-up synthetic routes have been extensively studied. For instance, Ke and colleagues prepared MoS₂ in the 2H/1T phase using an exfoliation method.⁴⁵ Briefly, the MoS₂ powder was immersed in a n-butyllithium solution, placed in a glove box under an inert atmosphere and washed with hexane after 3 days, then it was moved to an air environment. After dispersion in water, sonication in ice water and collection of the powder, monolayer nanosheets were obtained with 51.7% yield of the 1T-phase, which were transformed from the 2H-phase by the charge donation from Li atoms. Ren and colleagues instead used a one-pot hydrothermal method to synthesize MoS₂ in mixed phases.⁴⁶ They used ammonium molybdate tetrahydrate and thiourea as the molybdenum and sulfur sources, respectively, and performed the hydrothermal reaction at 200 °C. They obtained 1T/2H-MoS₂ and 1T/3R-MoS₂ when the molar ratios of molybdenum and sulfur were controlled at 1:5 and 1:25 and the solutions occupied 45% and 75% of the reactor volume, respectively.

Intercalation could be considered as an interlayer defect with the function of stabilizing metastable phases (e.g., 1T, 1T′) as well as modulating the properties of the guest. Similarly, both in situ intercalation during synthesis and post-synthesis intercalation based on MoS₂ have been investigated. For example, Deng and colleagues prepared (N, PO₄³⁻)-MoS₂/VG using an in situ route.⁴⁷ Briefly, aqueous solution of sodium molybdate and thioacetamide and vertical graphene (VG) films were transferred to a Teflon-lined steel autoclave and held at 200 °C for 12 h. Subsequently, an MoS₂/VG core/shell array with N doping (28%) and PO₄³⁻ intercalation (10%) was obtained by annealing the films at 300 °C. Feng and coworkers used a polymer-direct-intercalation (PDI) strategy to construct a MoS₂/N-doped carbon (MoS₂/NC) heteroaerogel.⁴⁸ Specifically, PEI (polyethyleneimine) and MoS₂ were ultrasonically mixed to form a suspension, during which the PEI molecule with positively charged NH²⁺ group was adsorbed on or inserted into the surface or the interlayer of the negatively charged MoS₂ nanosheets. The suspension was then freeze-dried to obtain aerogel-like 3D MoS₂-PEI due to the linkage of PEI molecules between MoS₂ nanosheets. Black MoS₂/NC was obtained by further annealing MoS₂-PEI at 800 °C under an argon atmosphere, in which the heteroaerogel shape was well preserved.

Single atoms anchored on the surface of molybdenum disulfide realize a synergistic catalysis with the carrier by modulating the local electronic structure. Methods such as interfacial reaction adsorption and one-step synthesis are effective options.^49,50 For instance, Li and coworkers prepared an Fe_SA/MoS₂ catalyst via a spontaneous reduction reaction of Fe³⁺ in defect-containing MoS₂.⁴⁹ Specifically, MoS₂ nanosheets were firstly synthesized by heating a mixture of ammonium tetrathiomolybdate and N,N-dimethylformamide in a Teflon-lined autoclave at 220 °C. The obtained MoS₂ was sonicated in H₂O₂ solution (3 wt%), during which abundant defects were introduced by H₂O₂ etching. The sample was then dispersed in an FeCl₃ solution before stirring, during which the defect-rich MoS₂ induced the reduction of Fe³⁺ due to its redox properties and individual Fe anchored to its surface. The sample was further dispersed in (3-mercaptopropyl)trimethoxysilane/dichloromethane, where the thiol group of the molecules refilled some S-vacancies and was bound to Mo. Finally, dissociation of the alkyl group of the molecules was realized by heating the sample in H₂/Ar at 300 °C. The single-atom Fe (loading of 0.31 wt%) was confirmed by AC HAADF-STEM as well as by X-ray absorption spectroscopy, i.e., each Fe is coordinated with three adjacent S and located on top of Mo.

For the preparation of MoS₂-based alloy materials, an in situ preparation technique and a post-synthesis conversion technique are both applicable. Wang and coworkers used a one-step CVD method to synthesize MoS_2(1−x)Se_2x/SnS_2(1−y)Se_2y heterojunction nanosheets.⁵¹ The solid sources include MoO₃/SnSe in the center of the furnace, and S powder at the upstream. The furnace was ramped up to the growth temperature (650 °C) under an N₂ flow (200 sccm) and the vapor-phase reaction was performed for 4 min under another N₂ flow (10 sccm) to grow both types of layers simultaneously. The AFM images as well as Raman spectra showed that SnS_2(1−y)Se_2y was fully or partially grown on the MoS_2(1−x)Se_2x sheets, while Mo–S and Mo–Se bonds coexisted in MoS_2(1−x)Se_2x. Zhang and colleagues obtained a MoS_2−xSe_x alloy/graphene composite by direct selenization of MoS₂/graphene.⁵² In brief, they mixed hydrothermally synthesized MoS₂/graphene with Se powder, then placed the mixture in a tube furnace and heated it at 700 °C for 2 h in a N₂ flow. MoS_2−xSe_x alloy exhibited an expanded interlayer spacing and an S:Se ratio of 1.2:0.8.

To prepare amorphous or clustered MoS_x, electrodeposition,⁵³ sputtering deposition,⁵⁴ and wet chemistry methods have been extensively studied.⁵⁵ For example, Escalera-López and coworkers conducted electrochemical deposition to prepare an amorphous MoS_x film on a Si/Ti/Au substrate.⁵³ Deoxygenated solution containing ammonium tetrathiomolybdate and NaClO₄, saturated Ag/AgCl, and a Pt mesh were employed as the electrolyte, reference electrode, and counter electrode, respectively. The Pt electrode was encapsulated in a glass vial with a fritted junction to prevent Pt deposition on the MoS_x films. AE-MoS_x was obtained by anodic electrodeposition at a constant voltage of +0.1 V (versus Ag/AgCl), while CE-MoS_x was obtained by cathodic electrodeposition at −1 V (versus Ag/AgCl). The relationship between film the thickness and charge density was established by measuring the film with a surface profilometer. Kibsgaard and colleagues used a wet chemical method to prepare Mo₃S₁₃²⁻ clusters.⁵⁵ Ammonium polysulfide solution (25 wt%) was added to ammonium molybdate solution and the mixture was left in an oil bath (96 °C) for 5 days until dark-red crystals precipitated out. After sequential filtering and washing, the product was then heated in hot toluene (80 °C) for several hours to remove excess sulfur to obtain (NH₄)₂Mo₃S₁₃·nH₂O.

Putting together, there are often multiple types of defects prepared using a particular method and there are multiple preparation methods for a particular defect type. The above review does not provide a relatively comprehensive coverage of the methods for preparing defective molybdenum sulfide and we will continue our discussion of methods in later sections.

2. Defective two-dimensional molybdenum disulfide for the HER

Hydrogen is considered as an ideal energy source to address the challenges due to environmental pollution and energy shortage. Fossil fuel-based hydrogen production processes generate harmful carbon emissions (7.33–29.33 kg CO₂ emissions for H₂ kg⁻¹ production at 75% system efficiency), while water electrolysis utilizing electricity converted from wind-, hydro- and solar- power could achieve carbon-free emissions.⁵⁶ Proton exchange membrane (PEM) electrolyzers with many advantages, including a high operating current density, high-purity pressurized hydrogen, good partial load range (i.e., compatible with intermittent renewable energy sources), high efficiency and portability/compactness,⁵⁷ rely heavily on scarce catalysts including platinum (Pt) and ruthenium (Ru)/iridium (Ir) for the HER and the OER, respectively. The cost of highly loaded platinum group metals (>3 mg cm⁻² total platinum group metal content) accounts for a quarter of the cost of the PEM cell stack.⁵⁸ Platinum-free catalysts designed for PEM electrolyzers have made rapid progress in the last decade, and have gained a critical advantage in the potential commercial application of electrolytic technologies.

The leading edge of MoS₂ in hydrogen evolution has made it a research hotspot for electrocatalytic HER applications. As a bio-inspired catalyst due to the similarity of its edge site to the metal coordination environment of hydrogenase and nitrogenase enzymes (Fig. 3a),⁵⁹ MoS₂ has been showing attractive hydrogen evolution properties in its two-dimensional (2D) forms, i.e., 2H, 1T, and 3R phases. Experimental and theoretical studies have shown that the intrinsic 2D basal plane of 2H-phase MoS₂ (the most stable configuration) is inert during HER catalysis, while defective MoS₂, including defect types such as edges, vacancies, doping sites, and strain sites, shows moderate hydrogen adsorption strength and thus affords superior HER performance when employed for catalyzing the hydrogen evolution reaction, especially in acidic media. Similar to that in the 2H phase, the defective MoS₂ also shows higher electrochemical activity than the intrinsic structure in the 1T and 3R phases.

2.1. Sulfur vacancies in 2-H MoS₂ as HER active sites

With in-depth studies, many types of active sites have been discovered in MoS₂ for the HER, including edge sites,^59,60 1T-phase metallic structures,^61,62 sulfur vacancies in 2H-phase MoS₂,¹⁹ and so on. For the HER, a general descriptor, the Gibbs free energy change (ΔG_H*), has been used to evaluate the activity by theoretical calculations since ΔG_H* equal to zero indicates optimal adsorption and desorption processes and thus the highest activity. The edge sites show extraordinary activity due to the near-neutral Gibbs free energy (+0.08 eV, Fig. 3b), indicating a near-optimal H adsorption strength; and the catalytic reaction rate was found to be proportional to the density of edge sites;⁶⁰ however, it is a challenge to distribute uniformly edge-rich tiny nanoparticles on a catalyst support.⁶³ 1T-phase MoS₂ is enriched with active sites, but it is metastable and undergoes a phase transition to the 2H phase upon heating/aging.^2,61 The vast basal plane in 2H-phase MoS₂ was long considered inert until previous work revealed the activation and optimization of the basal plane by creating S-vacancies.^{19,39,64–66} In the next section, we will discuss the S-vacancy-based defects in 2H-phase MoS₂, focusing on the underlying mechanism that unveils the critical role of S-vacancies, the reaction kinetics of defective MoS₂ with S-vacancies, the engineering approach for S-vacancies based on 2D MoS₂, and the defect-enhanced HER performance.

In our previous research, we have demonstrated that the S-vacancy introduces gap states around the Fermi level, allowing hydrogen to bind directly to the exposed Mo atoms.¹⁹ Benefiting from the tuning of the electronic structure, the systematic modulation of ΔG_H* can be realized by varying the concentration of S-vacancies ranging from 0% to 25%, while the optimal ΔG_H* value lies between 12.5% and 15.62% of S-vacancies (Fig. 3c). On top of S-vacancy engineering, the binding energy is further fine-tuned by applying external tensile strain (see Fig. 3d for diagram) to achieve an optimal ΔG_H* of 0 eV under a combination of 3.12% S-vacancies with 8% strain or 12.5% S-vacancies with 1% strain (Fig. 3e). The band structure is also finely tuned in the case of S-vacancies and applied strain, leading to a narrowing of the band gap as well as an increase of gap states around the Fermi level (Fig. 3f). Among various S-vacancies, ∼12.5% S-vacancies with a strain of ∼1.35% resulted in enhanced activity with a reduced overpotential (170 mA at 10 mA cm⁻²), a lower Tafel slope value (60 mV dec⁻¹), and a better turnover frequency (TOF_Mo) in acidic media, exceeding the performance of the basal plane of the 1T-phase MoS₂ or the edge sites of the 2H-phase MoS₂.

To elucidate the HER kinetics of the S-vacancy system, we performed a kinetic study using scanning electrochemical microscopy.⁶⁷ The substrate generation-tip collection (SG-TC) mode was chosen to determine the kinetics parameters such as apparent rate constant k⁰, electron-transfer coefficient α, and formal potential E⁰ (Fig. 4a). Under the assumption of a general one-electron transfer model with Butler–Volmer formalism, COMSOL simulations were performed to calculate the time-dependent H₂ (C_R) (Fig. 4b) and H⁺ (C_O) distributions before theoretically modeling the LSV curves. The equations, (where J_v, K_fv, K_bv, and f represent i/nFA, the forward reaction rate constant, the backward reaction rate constant, and F/RT, respectively), establish the connection between the simulated LSV curves and kinetics parameters, which are then extracted by identifying the best fit between the simulated LSV curves and experimental measurements (Fig. 4a). As a result, the unstrained MoS₂ containing S-vacancy shows the same E⁰, α, but a lower k⁰ than the strained one. Moreover, we verified those kinetic rate parameters for the defective MoS₂ electrode through a transient response study, and the results confirmed the reliability of the one-electron reaction model with a Butler–Volmer equation in dealing with the electrocatalytic behavior of MoS₂ with S-vacancies or a MoS₂-like catalytic system.

There is a wide range of methods for S-vacancy generation in 2H-phase MoS₂, including remote hydrogen plasma treatment,⁶⁸ laser ablation,⁶⁹ oxygen plasma treatment,⁷⁰ and helium ion irradiation.⁷¹ However, despite the great success in understanding the role of S-vacancies in the HER process, previous methods for the vacancy generation are confined to treatment with argon plasma and similar atmospheres, which act only on the oriented surface of the catalyst due to the anisotropy of plasma and thus have difficulty in penetrating the interior bulk and are not applicable to powder or bulk materials, restricting the production of industrial-scale catalysts. Chemical methods are considered as the alternative solution to generate S-vacancies in any forms of MoS₂.

We fabricated S-vacancies in MoS₂ by an electrochemical desulfurization process.⁶⁵ The reduced energy after S removal can be obtained according to theoretical study and the defective sites can be further stabilized with lower surface energy after H occupies the missing S atom, thus preventing the recovery of vacancies or the formation of permanent S-vacancies. As for the vacancy concentration, the surface energy per unit cell (defined as γ) continues to decrease with increasing S-vacancies at a potential of −1 V_RHE until a milestone of 12.5% S-vacancies is reached, and the lowest surface energy is obtained for the form of clustered vacancies following a continuous zigzag configuration (Fig. 4c). More than 12.5% of S-vacancies is hard to achieve due to the increased energy cost. Note that the application of a negative potential throughout the electrochemical desulfurization process is extremely necessary. This is further verified by comparing the intermediate steps, where the first protonation (*S + H⁺ + e⁻ → *SH) to form *SH and the subsequent H₂S formation step (*SH + H⁺ + e⁻ → H₂S + *) overcome huge energy barriers, while both steps become exergonic at a potential of −1.24 V_RHE (Fig. 4d). Experimentally, the electrochemical desulfurization method relied on voltage and duration to generate S-vacancies. It was not only applicable to monolayer MoS₂, but also polycrystalline multilayer MoS₂, obtaining a current increment of about 12 times at −0.32 V_RHE. This electrochemical desulfurization was found to be reliable for creating S-vacancies by comparing its electrochemical HER performance with that of other methods (Fig. 4e), indicating that the electrochemical desulfurization approach can effectively replace the argon plasma treatment for creating S-vacancies on the basal plane of the MoS₂ electrocatalyst.

Beyond the electrochemical desulfurization route, other feasible chemical approaches have been recently developed with opportunities for large-scale fabrication. From our perspective, three routes are accessible to generate S-vacancies with excellent catalytic properties, including the “reduction route”, “oxidant etching route”, and “doping route”. Specifically, reductants (e.g., lithium, zinc, hydrogen, and NaBH₄)^40,66,72,73 can react with sulfur to form sulfides under appropriate conditions (e.g., heating) and thus remove sulfur atoms from the basal plane, while oxidants (e.g., H₂O₂, NaClO)^74,75 can also introduce uniformly distributed S-vacancies through a mild etching strategy, and the other doping routes involve the direct chemical bonding between heteroatoms [e.g., palladium (Pd), Ru]^76,77 and sulfur atoms to create vacancy sites.

Wu and coworkers proposed a co-annealing method based on a homogeneous mixture of Zn metal and MoS₂, in which the formation of a Zn–S bond led to S-vacancies and in situ doping of Zn in the MoS₂ lattice, thereby introducing a large number of vacancy defects in the basal plane.⁴⁰ Theoretical calculations showed that the Zn-doped sites result in a significant decrease in the formation energy of S-vacancies, making it easier to form S-vacancies near the Zn sites and the introduction of S-vacancy induces new states in the band gap near the Fermi level, which might be responsible for the enhanced H atom adsorption at the S-vacancies. This strategy led to a reduction of the HER overpotential from 540 mV of pristine MoS₂ to only 194 mV of the defective MoS₂ at a current density of 10 mA cm⁻², while the Tafel slope value was reduced from 199 to 73 mV dec⁻¹ in an acidic environment. Wang and colleagues developed a simple and gentle H₂O₂ chemical etching strategy by immersing pristine MoS₂ in a H₂O₂ solution to obtain a homogeneous distribution of single S-vacancies onto the MoS₂ nanosheet surface, as captured by aberration-corrected scanning transmission electron microscopy (Fig. 5a).⁷⁴ Furthermore, by systematically adjusting the etching duration, etching temperature, and etching solution concentration, the comprehensive tuning of S-vacancies can be achieved to obtain the optimal HER performance with a Tafel slope of 48 mV dec⁻¹ and an overpotential of 131 mV at a current density of 10 mA cm⁻² in acidic media. The superior activity is attributed to more efficient tuning of the surface electronic structure, which leads to a higher electron density around Mo atoms and a deficiency of electrons on S-vacancy sites (Fig. 5b). The delocalized electrons around the Mo atom with single-atom S-vacancies show stronger attraction to hydrogen atoms, and thus favor the adsorption of hydrogen. Wang and coworkers fabricated a single atom Ru-doped 2H-phase MoS₂ catalyst using a two-step hydrothermal method, where Ru replaced some Mo atoms and created S-vacancies, delivering a lower Gibbs free energy of H-adsorption theoretically and a higher activity with an overpotential of 168 mV at a current density of 10 mA cm⁻² in 0.5 M H₂SO₄ experimentally.⁷⁶ Certainly, not limited to these three current routes, more low-cost and easy-to-operate chemical methods need to be further explored to achieve large-scale fabrication of MoS₂ catalysts with S-vacancy defects, which will help in the screening of industrially applicable materials.

Although important breakthroughs have been made for deepening the understanding of the mechanism of S-vacancy-enhanced activity, for opening new defects types (e.g., Frenkel defects caused by S-vacancies in monolayer MoS₂),⁷⁸ for observing the kinetics behavior of the defective catalyst, and for developing advanced fabrication methods of S-vacancies in single-crystalline monolayers or polycrystalline multilayers, a crucial issue to be addressed is the further improvement of activity and stability of MoS₂ with S-vacancies in acidic media, which is critical for practical applications. The inherent structure and properties determine the upper limit of the catalytic activity of MoS₂ with S-vacancies, and the breakthrough in performance can be achieved by combining S-vacancies with other engineering methods.

It is worth noting that several engineering approaches, including but not limited to N-doping, hybridization, vacancy engineering, and strain engineering, can be integrated into a single MoS₂ system for more efficient HER catalysis (Fig. 5c). We have applied tensile strain to MoS₂ with S-vacancies to tune the binding energy and band structures, which leads to an increase in catalytic activity, as evidenced by lower overpotentials and higher TOF values (see Fig. 5d for the TOFs) in an acidic environment.¹⁹ The catalytic properties of S-vacancies can also be tuned by integrating with heteroatoms. Kang's DFT calculations showed that by doping the transition metals, such as Sn, Tc, Ir, Rh, Ru, Re, Os, Pd, and Pt, into MoS₂ with S-vacancies and enabling the doped atom to replace the Mo atom adjacent to the S-vacancies, the HER activity of defective MoS₂ can be improved.⁸² Park and coworkers loaded Co clusters onto S-vacancies to form a Co–Mo preferentially active interface,⁷⁹ and the calculated ΔG_H* value for Co_n–MoS_2−x is closer to the optimal value of zero. Experimentally, the Co addition on the vacancy enabled the desulfurized MoS_2−x multilayer catalyst to attain a better activity with optimized overpotentials at a current density of −10 mA cm⁻² in 0.5 M H₂SO₄ electrolyte (Fig. 5e). Similarly, the interactions between S-vacancies and other processes, such as Ru deposition,⁸³ Zn doping,⁴⁰ Pd doping,⁸¹ and O substitution⁸⁴ have all shown enhanced HER performance. Table 3 lists more HER catalysts built on 2H-phase MoS₂ with S-vacancies in acidic media. Moreover, the combination of S-vacancies and phase modulation is also useful to elevate the catalytic activity, which will be discussed later.

---

Nevertheless, S-vacancies disrupt the crystalline structure of MoS₂, thus increasing the S-vacancy ratios increases the surface energy per unit cell (γ) (Fig. 4c and 5f),^19,65 which makes MoS₂ less stable. It is difficult to maintain the catalytic performance over long periods due to the structural instability caused by S-vacancies. Therefore, several strategies have been proposed to achieve high stability by stabilizing S-vacancies on defective MoS₂ catalysts, such as lowering the systematic energies by means of a stable support,⁸⁰ anchoring a heteroatom such as Pt or P at the S-vacancy sites,⁸⁵ and theoretically stabilizing the S-vacancy sites by Jahn–Teller distortion.⁸⁶ Moreover, the higher activity with smaller polarization also enhances the durability as the catalyst operating at a smaller overpotential can reduce the damage to the structure, although for the practical catalyst there is often a trade-off between activity and stability, e.g., the introduction of support materials or heteroatoms could block the highly active S-vacancies.

Recently, we reported a catalyst that stabilizes S-vacancies of 2H-phase MoS₂ supported on a vertical graphene network.⁸⁰ The introduced S-vacancies into MoS₂ and C vacancies into graphene synergistically modulate the stability and activity. The DFT results showed that adhesion of MoS₂–graphene systems is exergonic and thus these systems are relatively stable. Among the systems with or without defects, the weakest adhesion system is pristine MoS₂ over pristine graphene, while C vacancies or energetically favorable line-shaped S-vacancies drastically improve the adhesion energy between the catalyst and support (Fig. 5g and h). Specifically, 5.56% C vacancies and 12.50% S-vacancies achieve an 8.15% and 12.49% increase in adhesion energies, respectively. The catalyst with 12% S-vacancies showed the optimal activity with a 128 mV of overpotential at 10 mA cm⁻² and achieved stable performance over 500 h in an acidic environment. We also fabricated an ultra-stable electrocatalyst based on MoS₂ and graphene by doping with Pd.⁸¹ Theoretical studies revealed the preferred pathways for substitutional doping, i.e., formation of a Mo vacancy followed by doping of Pd on a Mo vacancy and then the creation of S-vacancies (Fig. 5i, path 1). Thus, the Mo vacancy sites, grains, boundaries, defects, or the edge of the MoS₂ lattice are preferred to be doped firstly. Pd doping generates rich S-vacancies, which increases the H adsorption and thus improves the activity and further increases the interlayer adhesion. The catalyst exhibited an ultra-stable hydrogen evolution operation for up to 1000 h under 10 mA cm⁻² with a much lower degradation rate of 23 μV h⁻¹ and for up to 180 h at a high current density of 80 mA cm⁻², outperforming many HER catalysts operating in acidic media. It therefore offers an efficient method for making stable, highly active and low-cost catalysts with great potential to replace Pt in PEM electrolyzers.

2.2. Defects in 1-T and 3-R MoS₂ as HER active sites

1T-phase MoS₂ with octahedral coordination and an expanded interlayer distance possesses metallic electrical conductivity and superior catalytic activity at both basal plane and edge sites,^87,88 but also has a metastable structure. Engineering of 1T-phase MoS₂ by basic size control, vacancy formation, or grain boundary modulation can enhance HER activity. For example, by synthesizing nanodots to provide high-density active edge sites,^89,90 by fabricating porous metallic 1T-phase MoS₂ nanosheets with S-vacancies,^39,72 or by making 2H-1T crystal boundaries to optimize the adsorption strength of the hydrogen atom.⁹¹ Besides, at least three defect types, including structural distortion,⁹² hetero-element doping,⁹³ and complex engineering,⁴⁷ have been proposed as efficient activation strategies for HER catalysis. In the next section, we will first discuss the HER application of a distorted 1T-phase MoS₂, focusing on its growth strategy and its integration with other defects; then discuss other defect engineering strategies of 1T-phase MoS₂, with emphasis on the construction of defects by in situ or non-in situ doping of metals or nonmetals to promote the HER, followed by the identification of dynamic active sites in doped 1T-phase MoS₂, and finally we present the HER study of defective 3R-phase MoS₂.

A distorted 1T-phase MoS₂, referred to as 1T′-phase MoS₂, features two adjacent Mo atoms with shortened distances, which are connected to form a zigzag chain, while S atoms in the distorted octahedron are structurally deviated from the pristine plane (Fig. 6a), belonging to the point group C_2v and space group Pmn2₁ in the orthorhombic crystal system.⁹⁴ MoS₂ in the 1T′ phase retains its metallic behavior as in the case of 1T-phase MoS₂, benefiting from a similar splitting of the d orbitals into two generated states, while electrons populate the lower energy levels and show a similar distribution of density of states that hosts the Fermi level.⁷ The subtle difference between 1T-phase MoS₂ and 1T′-phase MoS₂ is that the structural distortion of the metal chain period doubling leads to an intrinsic band inversion between chalcogenide-p and the metal-d bands around Γ, as well as the spin–orbit coupling opening a fundamental gap of 0.08 eV at the Dirac point (Fig. 6b).⁹⁵ DFT calculations suggest that the 1T′ phase can be converted spontaneously from the 1T phase without barrier ascribing to its relatively low energy, resulting in improved structural stability of MoS₂.⁹⁶ Controllable fabrication of 1T′-phase MoS₂ is important for catalysis, superconductivity, magnetism, and other fields. A common method for producing 1T′-phase MoS₂ involves intercalation with alkali metals (e.g., Li, Na, and K ions).⁷ Nevertheless, this method produces either incomplete phase transition or excessive intercalation, leading to impurity or decomposition of MoS₂.

More advanced preparation methods have been pursued before exploring the catalytic properties of 1T′-phase MoS₂. Peng, Yu, Nam and their colleagues have all recently developed a method to produce the high-purity 1T′-phase MoS₂ (as evaluated using the Mo 3d XPS spectra for the percentage of the 1T′ phase in the crystalline phases) using K₂MoO₄ and S as precursors and multi-annealing processes prior to post-treatment with I₂ acetonitrile solution.^7,97,98 This method endows MoS₂ with high-purity 1T′ phase in the form of bulk, or an exfoliated monolayer with a phase purity of 97% and an unprecedented lateral size up to tens to hundreds of micrometers. As observed, typical zigzag chains appear along the a-axis and two different interlayer distances alternate along the b-axis (Fig. 6a), and the lattice constant of the a-axis is shorter than that of the b-axis due to the structural distortion. In addition to the structural uniqueness, the 1T′-phase MoS₂ can also be distinguished by electronic or phonon information. The binding energy of the Mo 3d peak on the spectrum of X-ray photoelectron spectroscopy (XPS) is lower (∼0.8 eV negative shift) compared to that of the 2H-phase MoS₂, and only J₁, J₂, J₃ and A_1g peaks are present on the Raman spectrum without some characteristic peaks of the 2H (E_2g) and 1T (E_g, A_g) phases due to the broken structural symmetry, which both contribute to the identification of the 1T′ form.^97,98 In a typical three-electrode measurement (Fig. 6c), the pure 1T′ phase exhibited the best activity with an onset overpotential of 65 mV and a Tafel slope of 100 mV dec⁻¹ (Fig. 6d), which is better than those of the 1T′-2H mixture (128 mV dec⁻¹, obtained by laser-induced transformation as depicted in Fig. 6e) and the 2H phase (180 mV dec⁻¹, obtained by thermal annealing as shown in Fig. 6e) in an acidic electrolyte.⁹⁷ The superior activity is attributed to the inherent active basal plane and metallic charge transport in the 1T′ phase. Moreover, in-plane anisotropic charge transport affects the catalytic performance along different crystal orientations of the 1T′ phase, as revealed by electrocatalytic evaluation based on an eight-terminal device with four pairs of diagonal electrodes, where the conductivity relies on angle-resolved crystal orientation and the better charge transport ability leads to slightly enhanced catalytic activity.⁹⁸

Towards the goal of achieving an even higher purity of 1T′-phase MoS₂, Liu and coworkers proposed a one-step CVD growth route using K₂MoS₄ as a precursor to realize approximately 100% phase purity in the monolayer form.⁹⁴ Theoretical study suggests that potassium ion reduces the formation energy of the 1T′ phase theoretically and the K⁺ concentration exceeding 44% renders the 1T′ phase with a more stable structure than the 2H-phase MoS₂ (Fig. 6f). The synthesis of 1T′-phase MoS₂ undergoes the reaction process, K₂MoS₄ + H₂ → K_xMoS₂ + K + K₂S + H₂S (gas), where the annealing atmosphere (5–12% H₂ concentration in argon), temperature (650–750 °C) (Fig. 6g), and the substrate (fluorophlogopite mica) play crucial roles in the production of the target 1T′ monolayer material. The K⁺ concentration in the product was controlled by tuning the ratio between reductive H₂ and inert argon gases in the experiment, and a 50% K⁺ concentration was observed to stabilize the CVD-grown 1T′-phase MoS₂ flakes instead of the 2H phase. K⁺ ions can be completely removed after rinsing with water prior to evaluating the properties. Metallic 1T′-phase MoS₂ shows better HER performance than the 2H-phase flakes with an onset potential of 205 mV and a Tafel slope of 51 mV dec⁻¹ (Fig. 6h), attributed to its faster electron transport and higher density of active sites. Notably, the current density exhibits negligible loss after 1000 cycles, demonstrating its decent stability in an acidic environment.

Liu and colleagues further created the defects based on the 1T′-phase MoS₂.⁹² A bottom-up colloidal strategy was developed to prepare high-phase-purity, nano-monolayer, and strained 1T′-phase MoS₂ with Mo vacancies. The 1T′-phase MoS₂ nano-monolayer was subjected to ∼3% tensile strain in the xx direction, ∼2% shear strain in the xy direction (Fig. 7a), and 2% compressive strain in the yy direction according to the geometric phase analysis (GPA) using the atomic-scale HRTEM image. The Mo vacancies and strain together tune the band structure and hydrogen adsorption free energy (Fig. 7b). The best HER performance was achieved on the active S_T site at 3% uniaxial tensile strain. Also, the abundant in-plane/edge active sites and high electrical conductivity contribute to its electrocatalytic properties. In the acidic HER measurement, an overpotential of 149 mV at a current density of 10 mA cm⁻² and a Tafel slope value of 42 mV dec⁻¹ were obtained, with the durability for up to 48 h at an overpotential of 165 mV.

Apart from 1T′ phase engineering, doping of 1T-phase MoS₂ is another effective strategy to enhance the electrocatalytic performance. Tang and coworkers used DFT simulations to screen eleven transition metal dopants and found that Mn, Cr, Cu, Ni, and Fe can reduce the ΔG_H* value of the doped 1T phase.¹⁰¹ Experimentally, at least three doping methods have been developed, including one-step doping synthesis, post-synthesis doping, and doping-induced phase transition. Regarding the one-step doping method, Ji and coworkers pointed out that Cu atoms can be doped into MoS₂ by reducing MoS₃ with intermediate Cu⁺ species under hydrothermal conditions to form Cu-bound metal 1T-phase MoS₂, where Cu can be inserted into the S layer and form Cu–S bonds.⁹³ The synergistic effect of 1T-phase MoS₂, doped Cu atoms, and S- vacancies contributes to its elevated activity with a relatively low overpotential (131 mV@10 mA cm⁻²) and a Tafel slope of 51 mV dec⁻¹. Similar to the function of Cu, Jin and coworkers found that the introduced Co atoms also contribute to the HER activity by expanding the interlayer space, enriching the Co–Mo–S active sites (40.9–91.3%) and the doped 1T phase (73.9–79.2%).¹⁰² Specifically, defective MoS₂ was prepared using a hydrothermal approach with high-pressure H₂ utilized for expanding the interlayer space, while Co atoms form Co–Mo–S in the product. Polarization curves indicated that the as-prepared catalyst displayed an outstanding overpotential of 59 mV at 10 mA cm⁻² and a Tafel slope of 32 mV dec⁻¹ in acidic media, implying the obvious function of Co atom doping and H₂ modulation on the defective 1T-phase MoS₂ system.

Beyond metal atom doping, doping with nonmetallic elements into 1T-phase MoS₂ is also effective. Deng and coworkers recently developed a synergistic N-doping plus PO₄³⁻ intercalation strategy to realize the enlarged interlayer distance, high 1T phase proportion (41%), and nitrogen doping into the lattice with the introduction of the Mo–N bond (Fig. 7c).⁴⁷ The d-band center of (N,PO₄³⁻)-MoS₂ is lower than that of 2H-phase MoS₂ (Fig. 7d), leading to more electrons filling the antibonding states and thus weakening the interaction between the valence states of the adsorbent and catalyst. (N,PO₄³⁻)-MoS₂ achieves an optimized hydrogen adsorption free energy (ΔG_H* = 0.07 eV) for the HER and the combination of (N,PO₄³⁻)–MoS₂ with VG skeleton further enhances the electrical conductivity. Together with these advantages, the (N,PO₄³⁻)–MoS₂/VG catalyst exhibited superior HER activity with a small overpotential (85 mV@10 mA cm⁻²) and a low Tafel slope (42 mV dec⁻¹), as well as decent stability for 10 h at a current density of 10 mA cm⁻² in acidic media. Tan and colleagues synthesized 1T-phase (71%) MoSSe single-layer nanodots with Se vacancies by a combination of ball milling and chemical Li-intercalation.⁹⁰ The alloying effect (i.e., alternative H–S or H–Se bond) and Se vacancies (in this case, H adsorbs on the periphery S around Se vacancies) enable MoSSe nanodots to acquire a more neutral ΔG_H*. Hence, the 1T-phase MoSSe nanodots yielded a low overpotential of 140 mV at 10 mA cm⁻² and a Tafel slope of 40 mV dec⁻¹, as well as superior durability for 10000 cycling in an acidic electrolyte.

A post-synthesis doping procedure is performed on the as-synthesized 1T-phase MoS₂ by a gentle process that avoids phase destruction. Attanayake and coworkers reported a precipitation route in which solutions containing Na⁺, Ca²⁺, Co²⁺, or Ni²⁺ ions are added into the 1T-phase MoS₂ nanosheets, which results in 4–5 atomic% intercalation without altering the 1T phase as well as the 1T proportion (∼75%), leading to optimized H adsorption free energy ΔG_H* (Fig. 7e) and elevated HER performance compared to the pristine 1T phase.⁹⁹ Moreover, Lau and coworkers pointed out that sonochemical doping, a simple modulation method under ambient conditions to prevent the reconstruction from the 1T phase to the 2H phase, can achieve the inclusion of transition metal atoms, such as Pt or Pd, at the surface/interface of 1T-phase MoS₂.¹⁰⁰ The sonochemical process with a reduction function towards metal precursors led to the chemisorption of single transition metal atoms on the basal plane of exfoliated 1T-phase MoS₂, as captured by high-angle annular dark-field imaging-scanning tunneling electron microscopy (HAADF-STEM) images (Fig. 7f). It was significantly more active than the pristine 1T phase after doping, reaching an overpotential of 140 mV (3 wt% Pd doping) and 223 mV (3 wt% Pt doping) at 10 mA cm⁻² in 0.5 M H₂SO₄ electrolyte.

Single-atom doped 1T-phase MoS₂ can also be achieved by atom-induced phase transition. Qi and coworkers proposed the formation of the 1T phase induced by cobalt nanodisk addition and subsequent doping with cobalt by sonication and leaching methods (Fig. 8a).¹⁰³ A cobalt nanodisk was found to efficiently trigger the phase transition to distorted 1T-phase MoS₂ (D-1T MoS₂, see Fig. 8b), arising from the strain caused by the lattice mismatch between metallic Co and 2H-phase MoS₂, while the Co–S bonds are also formed with the assistance of sonication (Fig. 8c). The disordered structure in MoS₂ emerged upon the incorporation of Co, which is evenly located on top of the Mo atom in a single-atom site and coordinated to the adjacent three sulfur atoms, while higher Co coverage (Co–Co distance less than 3.10 Å with a strain of ∼3.70%) determines the stable presence of the 1T phase, as suggested by DFT. SA Co-D-1T MoS₂ with 3.70% cobalt coverage shows the lowest ΔG_H* value of 0.03 eV, very close to the ideal 0 eV. In the actual experiment, 3.54% Co loading was introduced into D-1T MoS₂, delivering an excellent HER performance with a 42 mV overpotential at 10 mA cm⁻², a 32 mV dec⁻¹ Tafel slope value that is superior to most nonprecious catalysts (Fig. 8d), and robust stability for over 10 days (at η = 100 mV vs. RHE) in acidic media.

Despite considerable progress in exploring element selection, synthetic methods, electrochemical performance, and DFT theoretical understanding, tracking and identifying the dynamic active sites of defective 1T-phase MoS₂ during HER operation remain ambiguous despite its essentiality for understanding the HER. Pattengale and coworkers recently revealed the dynamic behavior of active sites by capturing the intermediate structure of a doped 1T-phase MoS₂ electrocatalyst.¹⁰⁴ In their Ni@1T-MoS₂ catalytic system (Fig. 8e and f), a shortening of the Mo–S distance at an applied potential of −0.76 V under acidic conditions was observed, while the intrinsic Ni site did not undergo significant changes in the first coordination shell despite a shortening of the Ni–Mo second-shell interaction. Single-atom Ni replacing Mo on the basal edge underwent reduction from the Ni²⁺ oxidation state to lower valence state at externally applied potentials, without significant change of the pristine structure where the single Ni site bonded to surrounding Mo, S, and O (Fig. 8g). Thus, 1T-phase MoS₂ as the active phase and Ni as the active species undergo minimal changes when no potential, 0 V, and −0.76 V are applied, indicating that the intrinsic catalytic structure of Ni@1T-MoS₂ is quite stable during the dynamic evolution of the catalyst under acidic conditions. Inspired by this work, similar work is proposed to identify the active sites of defective MoS₂ during dynamic catalysis to deepen the understanding of the MoS₂-based catalyst systems. Besides, the controllability including defect concentration, defect location, electrical conductivity, and the proportion of 1T phase in defective MoS₂ remains challenging and needs more attention in future studies.

In stark contrast to the defective 1T phase, the semiconductor 3R-phase MoS₂ with R3m space group and a ABC–ABC stacking order structure possesses unique properties such as excellent non-linear optical response, valley-dependent spin polarization benefiting from broken inversion symmetry even in the bulk form, offering huge opportunities for applications in nonlinear optical devices, electronic and optoelectronic devices, etc.^105,106 In the field of HER catalysis, Toh and coworkers reported that a mixed-phase MoS₂ consisting of rhombohedral 3R and the hexagonal 2H (in a ratio of 2.3:1) phases obtained by a flux route with MoO₃ and S as the precursors in carbonate media at 550 °C, shows a Tafel slope of 113 mV dec⁻¹ (see Table 4 for the comparison of the Tafel slopes and other parameters for more defective 1T- and 3R-phase MoS₂), superior to that of pure 2H-phase MoS₂ (170 mV dec⁻¹) and comparable to that of the exfoliated 1T-phase MoS₂ (99 mV dec⁻¹) under acidic conditions.¹⁰⁷ However, more defective 3R-phase MoS₂ with enhanced catalytic properties remains underexplored, although the S-vacancy defects have also been uncovered in the 3R phase and these defects exhibit migration phenomena similar to that of the 2H phase.¹⁰⁸ Therefore, the methods for making vacancy, strain, doping, and other defects that could hold promise in enhancing the catalytic performance of 3R-phase MoS₂ need further development.

---

It is worth reminding that the defective MoS₂ in the 2H/1T/3R phase discussed above is applied to catalyze the HER in acidic media. In contrast, intrinsic MoS₂ generally behaves poorly for the HER in an alkaline environment due to the high activation energy of the Volmer step of water dissociation and the strong adsorption of the formed OH⁻ on the MoS₂ surface.¹⁰⁹ Although S-vacancy or mixed-phase defects can also effectively boost the HER kinetics of MoS₂, the increased activity from these defects is still far below that of the Fe/Co/Ni-based composition.^110,111 Therefore, most of the current studies on MoS₂-based catalysts for the HER in alkaline media have focused on the introduction of Ni or Co components in the form of atom/oxide/hydroxide/sulfide/nitride/phosphide^{109,111–115} and some studies are on the construction of heterojunctions with other compounds¹¹⁶ or on the introduction of noble metals.¹¹⁷

For example, Liu and coworkers employed hydrothermal and sulfidation methods to prepare 1T_0.72-MoS₂@NiS₂, where heterogeneous interfaces of MoS₂ and NiS₂, heterojunctions of 2H and 1T phases, numerous defects, and high 1T ratios were observed.¹¹¹ When catalyzing the HER in an alkaline environment, 1T_0.72-MoS₂@NiS₂ exhibited an overpotential of 128 mV and stability for 30 h at a current density of 10 mA cm⁻². Moreover, it had a Tafel slope value of 68 mV dec⁻¹, which is lower than those of 1T_0.41-MoS₂ (89 mV dec⁻¹), 2H-phase MoS₂ (124 mV dec⁻¹), and 1T-phase MoS₂ (198 mV dec⁻¹). DFT calculations showed that 1T-MoS₂@NiS₂ has a metallic phase with zero band gap, and PDOS (projected density of states) intensity of 1T-MoS₂@NiS₂ is higher than that of 1T-phase MoS₂ or NiS₂, indicating its higher electron transfer efficiency. Moreover, 1T-MoS₂@NiS₂ has a lower ΔG_H2O (water dissociation energy barrier, 0.1 eV) and ΔG_H* (−0.12 eV) than 1T/2H-MoS₂ (ΔG_H2O, 0.16 eV; ΔG_H*, 0.97 eV) and 2H-phase MoS₂ (ΔG_H2O, 0.82 eV; ΔG_H*, 1.49 eV), which favors the cleaving of the HO–H bond and delivers a near-neutral H* adsorption strength. Huang and coworkers constructed a MoO₂/MoS₂ heterogeneous nanorod encapsulated in N and S co-doped carbon (MoO₂/MoS₂@NSC), in which MoO₂ and MoS₂ were tightly coupled at the atomic scale.¹¹⁶ When applied to the HER in alkaline media, MoO₂/MoS₂@NSC exhibited an overpotential of as low as 156 mV at a current density of 10 mA cm⁻² and a Tafel slope of 99 mV dec⁻¹, which exceeded those of MoS₂@NSC (232 mV, 131 mV dec⁻¹) and S-MoO₂@NSC (364 mV, 142 mV dec⁻¹). Theoretical simulations implied that the positively charged Mo sites in MoO₂/MoS₂ exhibited the most negative water adsorption energy and the Mo–O* bond shortens, while the O*–H bond elongates after water adsorption, thus promoting the dissociation of H₂O. In addition, the ΔG_H* value of MoO₂/MoS₂ is significantly lower than that of MoS₂, indicating the enhanced HER activity. Lang and coworkers proposed a strategy of coupling S-vacancies with single atomic Ru to enhance the HER performance of MoS₂.¹¹⁷ Theoretical analysis showed that Ru₁@D-MoS₂ (i.e., atomic Ru dispersed in defective MoS₂), in which Ru replaces the Mo site, has a ΔG_H2O value of 0.55 eV, which is lower than those of MoS₂, MoS₂-S_v, Ru₁(S)@MoS₂, and Ru₁(Mo)@MoS₂, suggesting that the coupling of vacancies and Ru greatly facilitates the dissociation of water. Also, Ru₁@D-MoS₂ has the optimal ΔG_H* of −0.18 eV, indicating its superior activity. In the experiment, the exfoliation of MoS₂ and the introduction of defects were performed via ball milling and annealing. When serving as a catalyst, Ru₁@D-MoS₂ displayed an overpotential of 107 mV at 10 mA cm⁻² and a Tafel slope value of 96 mV dec⁻¹ in 1 M KOH solution, much better than those of D-MoS₂ (∼264 mV, 85.6 mV dec⁻¹) and commercial MoS₂ (364 mV, 111.7 mV dec⁻¹). In conclusion, the above discussion demonstrates that the defects introduced into crystalline MoS₂ also contribute to an increase in HER activity in alkaline media by lowering the reaction energy barrier.

3. Defective non-two-dimensional molybdenum sulfide for the HER

Distinct from 2D crystalline 2H/1T/3R phase MoS₂, non-2D molybdenum sulfides such as amorphous MoS_x (α-MoS_x) and clustered MoS_x feature structural uncertainty, complicated internal linkage mode, lack of long-range order, and varied morphology without two-dimensional layer structure, leading to critical obstacles in unraveling the catalytic functions. Although the complex intrinsic structures sometimes cannot be fully elucidated, a few relatively well-defined cluster units, such as [Mo₃S₁₃]²⁻, [Mo₃S₁₂]²⁻, and [Mo₂S₁₂]²⁻, still provide a partial basis for further studies to explore the macroscopic chain/island/network morphology, underlying electronic structures, and physicochemical properties. Their superb electrocatalytic HER performance and low-cost fabrication processes have stimulated consecutive research on the synthesis of materials, as well as the identification and optimization of catalytic active sites. In the next section, we will review some representative works on amorphous MoS_x and MoS_x clusters for their HER application in an acidic environment. The catalyst structure, the chemical bonding transition upon activation, and the active sites play a critical role in the catalytic performance, so we closely integrate them in our discussion.

3.1. Amorphous MoS_x for the HER

Amorphous MoS_x, with an overpotential range of 100–180 mV at a HER current density of 10 mA cm⁻² in acidic media, exhibits higher catalytic activity than highly crystalline MoS₂, motivating researchers to investigate its structure–activity relationship.^118–120α-MoS_x, whose structure is mainly responsible for its electrochemical properties, is assigned to be a molecule-based coordination polymer consisting of cluster entities such as [Mo₃S₁₃]²⁻, [Mo₃S₁₂]²⁻, [Mo₃S₄]⁴⁺, [Mo₂S₁₂]²⁻, etc.^121,122 Five sulfur ligands (Fig. 9a), including bridging S₂²⁻, apical S²⁻, shared S₂²⁻, terminal S₂²⁻, and unsaturated S²⁻ (broken S₂²⁻), distinguished by their respective Raman and XPS features, stabilize isolated Mo atoms with apical S²⁻ in the center surrounded by bridging S₂²⁻/terminal S₂²⁻, as well as shared S₂²⁻/unsaturated S²⁻ connecting the discrete units. Together, they construct randomly disordered morphologies in the form of unfolded 1D polymeric chains or interconnected branches.¹²² For the detailed characterization of sulfur, the S–S stretching vibrations of terminal S₂²⁻ [ν(S–S)_te at 525 cm⁻¹], bridging/shared S₂²⁻ [ν(S–S)_br/sh at 555 cm⁻¹], and apical S²⁻ [ν(Mo₃-μ₃S) at 450 cm⁻¹] can be assigned separately in the Raman peaks, where ν(Mo–S)_coupled is located at 382–284 cm⁻¹ (Fig. 9b). Moreover, XPS exhibits a lower electron binding energy of 161.8 eV assigned to terminal S₂²⁻/unsaturated S²⁻ and a higher electron binding energy of 163.1 eV assigned to bridging S₂²⁻/shared S₂²⁻ or apical S²⁻ peaks.^118,122,123 In addition, the chemical state and coordination environment of Mo are affected in the amorphous state compared to the crystalline phase by, (1) the presence of a higher Mo valent state due to MoO₃ or MoO excluding common Mo⁴⁺ in Mo₃^IV–S, (2) the coordination number (CN) of Mo–S changes from 4.25 in the 2H phase to 5.29 and 0.60 (short) in the amorphous state, (3) the Mo–Mo distance (2.1 ± 0.7 Å) is shortened from that of the 2H phase (3.1 Å), and (4) the S:Mo stoichiometries extend to a larger range from 1.7 to 4 depending on the fabrication methods, while the oxygen atoms are concurrently preserved in the material.^{118,119,123,124} Based on the above structures, we will next focus on the preparation of amorphous MoS_x and the identification of HER active sites, including unsaturated Mo, S₂²⁻, unsaturated S²⁻, and the sites synergized by Mo, S, or bonding.

For the preparation of amorphous MoS_x, it is important to ensure mild temperatures throughout the process. The reported methods include electrodeposition, atomic layer deposition, magnetron sputtering, wet chemical synthesis, femtosecond laser ablation fabrication, physical vapor deposition, and hydrothermal synthesis.^118,125 Among them, the electrodeposition technique is used as a simple fabrication method to form amorphous films on conductive electrodes by reduction or oxidation pathways using the chemically aqueous tetrathiomolybdenum [MoS₄]²⁻ precursor in the following reactions, i.e., [MoS₄]²⁻ → MoS₃ + 1/8S₈ + 2e⁻ in the anodic step or [MoS₄]²⁻ + 2H₂O + 2e⁻ → MoS₂ + 2SH⁻ + 2OH⁻ in the cathodic step. The above deposition routes avoid the time-consuming annealing or exfoliation operations that are required in producing crystalline MoS₂, endowing it with tremendous excellent scalability.^122,124 Notably, an activation process is needed for the electrodeposited amorphous MoS_x to show optimized activity. The activation steps eliminate the inert S ligands and expose the active species through structural transformation. For instance, an activation method can involve electrochemical cathodic scanning by linear sweep voltammetry (LSV), cyclic voltammetry (CV), or potentiostatic techniques to transform the as-deposited MoS₃ to MoS_2+x that resembles MoS₂ but in an amorphous state and concurrently to reduce the amount of S ligands that affect the exposure and identification of the actual catalysis sites. The representative works of such activation methods are detailed in Table 5.

---

To determine the active site of the activated α-MoS_x, previous work has tracked changes in Mo and S elements by employing in situ or ex situ measurements. To date, the detailed mechanism remains elusive, although unsaturated Mo, S₂²⁻ and unsaturated S²⁻ have been considered as active centers according to individual studies. Tran and coworkers proposed unsaturated Mo to be the active center.¹²² In their study, Mo–□ was created by reductive activation according to the following hypothetical stepwise or concerted processes: Mo(μ-S₂)Mo + 2e⁻ + xH⁺ → Mo(S)_2−x(SH)_x + Mo–□, or Mo(S₂) + 2e⁻ + 2H⁺ → Mo–□ + HS⁻, when a constant potential of −0.71 V_NHE was applied in neutral media. Disulfide ligands including bridging, shared, and terminal S₂²⁻ were significantly decreased or even disappeared during the activation process, indicating that the disulfide would not be able to support a continuous HER. In addition, the unsaturated Mo^IV-□, which is regarded as the catalytic species, can be converted to Mo^V-□ after protonation and H₂ evolution steps, and the hydration of this Mo^V-□ yields a MoO motif which can be determined by the enhanced resonance Raman signal after activation treatment, further suggesting that the active site is indeed from unsaturated Mo (Fig. 9c). Moreover, the much lower H adsorption free energy ΔG_H* (0.108 eV) to form Mo^V-hydride species from Mo^V-□ also favor the mechanistic route via the Mo–H intermediate prior to H₂ evolution. Xi and coworkers also concluded that the unsaturated Mo center is the active site, where a similar S₂²⁻ reduction route of MoS₂ + 2e⁻ + 2H⁺ →Mo–(S–□) + HS⁻ is considered to generate unsaturated Mo dangling bonds.¹¹⁸ This hypothesis is supported by the complete disappearance of the terminal and bridging S₂²⁻ as shown by online Raman spectroscopy and the ratio of Mo⁶⁺ to Mo⁴⁺ state increasing to 1:3.14 due to the susceptibility of unsaturated Mo atoms with dangling bonds to oxidation in air. More specifically, S₂²⁻ has been partially transformed into S²⁻ existing in the emerging MoS₂ nanoislands converted from α-MoS_x, while the other sulfur losses have been observed to be released as H₂S gas, especially in the early stage of the activation process. The exposed Mo–(S–□) active species, together with its Mo–(S–H*) intermediate adsorption state, confer excellent HER activity to the activated catalyst with an overpotential of 180 mV at a current density of 10 mA cm⁻² and a stability test over 10 h in a 0.5 M sulfuric acid electrolyte, demonstrating the significant role of the activation process in this amorphous catalyst. Li and coworkers also concluded that Mo^V defective species are the active sites and the Mo^V species can be controllably generated by temporally shaped femtosecond laser ablation of (NH₄)₂MoS₄ aqueous solution through tuning the laser energy and pulse delay to control reduction/oxidation.¹²⁵ This work revealed a close relationship between the HER activity in acidic media and the Mo^V species content, showing an incremental HER activity with optimized onset potentials (127–105 mV), overpotentials (173–145 mV) at a current density of 10 mA cm⁻² and Tafel slope values (47–40 mV dec⁻¹), with increasing Mo^V state content (18.2–28.3%). However, after the CV activation process, the content of the Mo^V state decreased, while the content of Mo^IV increased; therefore, this work assumed a similar transition from Mo^IV to unsaturated catalytic Mo^IV–□ centers. Unfortunately, the corresponding studies still lack the direct evidence to capture such Mo^IV–□ catalytic species responsible for hydrogen evolution.

The sulfur ligands are also designated as the active site. Among the five types of apical S²⁻, bridging S₂²⁻, terminal S₂²⁻, shared S₂²⁻, and unsaturated S²⁻, terminal S₂²⁻, bridging S₂²⁻, and unsaturated S²⁻ are successively considered as the actual catalytic sites. Li and coworkers proposed that the terminal S₂²⁻ may undertake the catalytic function in the amorphous MoS_x deposited by CVD at low temperatures.¹²⁶ In brief, this amorphous material grown at 300 °C contained abundant apical S²⁻ and/or bridging S₂²⁻, as well as terminal S₂²⁻ and/or unsaturated S²⁻, but after LSV cycling, the apical and/or bridging S₂²⁻ nearly disappeared as shown by XPS. Thus, the high HER activity that was stable during LSV cycling should come mainly from terminal S₂²⁻ and/or unsaturated S²⁻ and the authors particularly emphasized the vital role of terminal S₂²⁻ in the catalyst, as suggested by the DFT results. In contrast, another investigation conducted by Ting and coworkers concluded that bridging S₂²⁻ may be the major catalytic center towards proton reduction.¹²⁴ The bridging S₂²⁻ was generated by an activation method designed to perform CV oxidative stripping to an anodic potential of 0.75 V_RHE, leading to an increase in the proportion of sulfur with higher electron binding energy due to oxidation of sulfur with lower binding energy, based on post-LSV MoS_x-AE films (MoS_x electrodeposited on the anode). To measure the catalytic activity, TOF measurement instead of LSV curves was carried out to exclude the possible influence from different catalyst loadings. At an overpotential of 200 mV, a linear positive correlation was established between the percentage of S atoms with higher electron binding energy (e.g., bridging S₂²⁻ and apical S²⁻) and TOF (see Fig. 9d for details), while apical S²⁻ was excluded from the main catalytic reaction sites due to its weak H adsorption (ΔG_H* > 1 eV, Fig. 9e), suggesting that the bridging S₂²⁻ site in the catalyst facilitated the evolution of H₂.

Additionally, according to Deng and coworkers, unsaturated S²⁻ may play a dominant role in the evolution of H₂ when more negative potentials are applied to the catalyst.¹²³Operando Raman spectroscopy recorded detailed changes during the CV cathodic sweep at a slow scan rate of 0.5 mV s⁻¹, showing the attenuation of apical S²⁻, terminal S₂²⁻, and bridging S₂²⁻ at −0.07 V. This suggests that the S–S bonds of the terminal and bridging S₂²⁻ are cleaved before being converted to possible unsaturated S²⁻ species, consistent with the XPS results showing an increase in the proportion of S atoms with low electron binding energy (terminal S₂²⁻ and unsaturated S²⁻) after the HER. More importantly, a peak at 2530 cm⁻¹ belonging to the S–H stretching vibration was captured at potentials below −0.07 V (Fig. 9f) and its signal increased with more negative potentials and faded away when the potential sweep was reversed to positive values, indicating the appearance of MoS_x–H species under acidic HER conditions. The single S atom bound to H is thought to be responsible for this HER activity although the type of sulfur cannot be assigned accurately. Consistently, the absence of Mo–H signals further pinpoints the sulfur atoms as the active sites.

Accordingly, since the different studies mentioned above point to different active sites and give corresponding experimental and theoretical evidence, we try to believe that the S active sites may depend on the preparation and the activation methods of the material, since the internal structures of amorphous MoS_x prepared by different processes are clearly different. We also see that the accurate identification of active sites remains a challenge due to the lack of techniques to individually identify each sulfur ligand or Mo intermediate during the catalytic reaction.

Beyond the above discussion on individual S or Mo elements serving as active sites, Lassalle-Kaiser and coworkers proposed a synergistic effect arising from Mo and S cooperation.¹²¹ In their study, Mo^IV was reduced to the Mo^III state according to the Mo L-edge X-ray absorption near energy structure (XANES) results, while S was oxidized to the terminal S₂²⁻ ligand according to the S K-edge XANES results, which occurred at the solid–liquid interface of MoS_2+x at −0.3 V (vs. RHE). The Mo^III(S₂) units, considered to be catalytically active intermediates, underwent protonation and reduction steps to form Mo^IV(S)(SH) species prior to releasing H₂ and renewing the intermediate, exhibiting their cooperative involvement in catalysis. Escalera-López and colleagues combined XPS, Raman and electrochemical tests to reveal the pH-dependence of the active sites in anodically electrodeposited amorphous molybdenum sulfide (AE-MoS_x).⁵³ In a very acidic environment (0 ≤ pH ≤ 2), the predominant HER active sites were unsaturated S²⁻ after bridging S₂²⁻cleaving and terminal S₂²⁻ dissolution at the cathodic potential. In contrast, under mildly acidic conditions (3 ≤ pH ≤ 6), the Mo⁵⁺O_xS_y species (i.e., oxygen-containing and S-deficient Mo-sites) with unsaturated Mo⁵⁺ were thermodynamically more stable on the AE-MoS_x surface and thus successfully catalyzed the HER. Thus, relying on the two active species, the decent stability was observed over a wide range of acidic pHs (0–6) in a 12 h (at 10 mA cm⁻²) stability test. Moreover, besides the element itself, the bond structure in amorphous phase MoS_x is also considered to play a key role in determining the activity. According to a recent study by Wu and coworkers, the inherent short Mo–Mo bonds in the amorphous phase compared to the 2H or 1T phases (Fig. 9g), were elucidated to enhance HER activity and maintain its stability, enlightening us to recognize its active center from a systematic perspective combining materials and structures.¹¹⁹ Especially for doped amorphous MoS_2+x, such as N-doped α-MoS_x that displayed a smaller overpotential than α-MoS_x as reported by Ding and coworkers, it poses a new challenge to distinguish the active center since its complicated components require a systematic study of more factors, including the types of elements, the occupied sites of exotic atoms, the ligand types, and the bond structures, to draw conclusions.¹²⁷ Certainly, in addition to the above work on active site identification, it is equally important to focus on the performance breakthroughs. Zhang and colleagues prepared a W-doped α-MoWS_x/N-RGO catalyst, which achieved a low overpotential of 348 mV and 24 h stability at an industrial-level current density of 1000 mA cm⁻² in an acidic electrolyte and the overpotential of this catalyst was lower than that of Pt/C at high current densities, indicating that the excellent performance of α-MoS_x can be obtained through rational structural design.¹²⁸

Undoubtedly, the cost-effective, efficient, and facile synthesis process by electrodeposition or wet chemical routes under ambient conditions makes such amorphous materials attractive for large-scale H₂ evolution in the PEM electrolyzer. Further investigations are worthwhile to, (1) identify the active sites with advanced techniques, (2) balance amorphous and crystalline states for better activity and stability, and (3) scale up controllable manufacturing methods.

3.2. MoS_x clusters for the HER

The molecular monomers of amorphous MoS_x (e.g., Mo₃S₁₃²⁻ cluster) can independently bear hydrogen production, benefiting from the structural motifs resembling edge active sites in MoS₂. MoS_x clusters present a rich variety of cluster types and structures, including the basic Mo₃S₁₃²⁻, Mo₃S₁₂²⁻, MoS₄²⁻, Mo₂S₁₂²⁻, Mo₃S₄⁴⁺, etc., which contain only Mo and S elements. Moreover, the incorporation of exotic elements or molecular ligands results in more complex structures and modifiable HER properties. In the next section, we will first discuss several basic cluster structures and their HER performance, including Mo₃S₁₃²⁻, Mo₃S₁₂²⁻, and Mo₂S₁₂²⁻ clusters, followed by the modification methods of MoS_x clusters, such as oxygen incorporation, ligand linkage, and heteroatom doping; notably, in each cluster, the preparation method, structure, and properties are in focus.

Among these clusters, the Mo₃S₁₃²⁻ cluster is considered to be the most representative entity to show prominence in the catalytic HER. Kibsgaard and colleagues elucidated that Mo₃S₁₃²⁻ consists of three terminal S₂²⁻ with lower binding energy, three bridging S₂²⁻ and one central apical S²⁻ with higher binding energy (Fig. 10a), with a ratio of 7:6 between the doublet with higher binding energy and the other doublet.⁵⁵ The Mo₃S₁₃²⁻ ion cluster could be stored in the synthetic (NH₄)₂Mo₃S₁₃·nH₂O, a highly crystalline dark red bulk that can be easily deposited on the substrate by the wet-drop casting method, and the individual cluster can be imaged by atom resolved STM imaging (Fig. 10a), showing an ordered atomic-level structure. This catalyst with loadings of 10–100 μg cm⁻² exhibited onset overpotentials of 0.1–0.12 V, low overpotentials between 0.18 V and 0.22 V at a current density of 10 mA cm⁻², superb Tafel slope values of 38–40 mV dec⁻¹ under acidic conditions, and comparable or better TOF values than amorphous MoS_x (Fig. 10b).

It is noteworthy that in the above study, the two substrates, HOPG and GP, are loaded with sub-monolayer and bulk catalysts, respectively, resulting in a disparate Tafel slope and TOF values, as well as the underlying reaction mechanism pathways.⁵⁵ In another similar study by Hellstern and coworkers, changing the substrate produced considerable differences in activity even at similarly low coverages of the catalyst on the substrate, where the authors have ensured the proximity of the catalyst to the substrate.¹³⁰ In their study, a comparative study of copper, glassy carbon, gold, and silver substrates was performed to demonstrate the correlation between the calculated ΔG_H* and TOF values (Fig. 10c), where a gold substrate loaded with Mo₃S₁₃²⁻ achieved an optimal TOF of 0.47 H₂ s⁻¹ per Mo atom (in an acidic electrolyte) and also an optimized thermal adsorption/desorption energy. Thus, the choice of substrates has a great influence on the HER activity, which comes from the possible chemical rearrangement step, electrical conductivity, support affinity, active surface atom from substrate, etc., thus providing a perspective to modulate catalytic activity by substrate engineering, especially in the case of low coverage of clusters.

The Mo₃S₁₃²⁻ cluster described above was synthesized via a wet chemical route, followed by deposition on the scaffold by simple drop-casting or spray coating methods, allowing them to be easily scaled up and applied in PEM electrolyzer devices. Holzapfel and coworkers recently conducted a feasibility study on the application of this material in PEM devices, where Mo₃S₁₃²⁻ clusters and a N-doped carbon nanotube (NCNT) hybrid were used as the cathode and IrO₂ as the anode, which were separated by a Nafion membrane.¹³¹ The electrolyzer achieved a high current density of 4 A cm⁻² at a cell voltage of 2.36 V under a catalyst load of 3 mg cm⁻² (Fig. 10d), which surpassed the current density of the corresponding non-precious metal catalyst in the PEM cell, and attained a slight degradation of 83 μV h⁻¹ in a 100 h stability test at 1 A cm⁻². In addition, component dissolution experiments involving cluster structure stability were performed by inductively coupled plasma mass spectrometry (ICP-MS) at downstream of a three-electrode electrochemical setup, and it was found that the Mo element readily dissolved upon contact with the electrolyte, but then the dissolution slowed down rapidly. This slight dissolution could be ascribed to the formation of Mo⁶⁺ oxides during air exposure although the balance between Mo⁶⁺ oxides and Mo₃S₁₃²⁻ is uncertain, however, the Mo₃S₁₃²⁻ clusters originally underneath the oxide layer were capable of maintaining catalytic activity in the PEM electrolyzer after the depletion of the oxide layer.

Similar to Mo₃S₁₃²⁻ in structure and composition, Mo₃S₁₂²⁻, Mo₂S₁₂²⁻, and [Mo₃S₄]⁴⁺ also possess high HER activity, which enriches the structural variety of the clusters. The counterions of these cluster ions are usually NH₄⁺ or SO₄²⁻, whose effects on catalysis remains to be investigated.^132,133 Simple chemical preparation methods are also applicable to them; for instance, Mo₂S₁₂²⁻ can be obtained in as-synthesized crystalline (NH₄)₂[Mo₂(S₂)₆] 2H₂O with orthorhombic space group Pnna, where each Mo center is coordinated to two bridging S₂ and two terminal S₂;¹³⁴ and cubane-type [Mo₃S₄]⁴⁺ could be produced by a simple (NH₄)₂MoS₄ reduction reaction.¹³³ To obtain Mo₃S₁₂²⁻, Lee and colleagues developed a transformation procedure that revealed that the Mo₃S₁₃²⁻, Mo₃S₁₂²⁻, and low-crystallinity MoS₂ could be obtained through the reaction of MoS₃ and ammonium hydroxide solution by eliminating the apical S and the bridging S below and above 280 °C, respectively.¹³⁵ In the case of Mo₃S₁₂²⁻, the current density at each potential is slightly inferior to that of Mo₃S₁₃²⁻ and the bridging S is considered as the active site of Mo₃S₁₂²⁻, as suggested by the decrease in current density after elimination of the bridging S. In another work, Mo₂S₁₂²⁻ with a loading of 150 nmol cm⁻² exhibited a better polarization profile than Mo₃S₁₃²⁻ or other cluster-type catalysts (Fig. 11a) with an overpotential of 161 mV at 10 mA cm⁻² and a Tafel slope of 40 mV dec⁻¹ in acidic solution.^134,135 The complementary theoretical study revealed that Mo₂S₁₂²⁻ preferentially follows the Volmer–Heyrovsky reaction pathway because the activation energy barrier along the Volmer–Tafel reaction is larger (Fig. 11b), supporting the experimental Tafel slope (40 mV dec⁻¹) and thus the limiting step comes from the electrochemical desorption process.¹³⁴

The tuning of the cluster structure confers flexibility and variability to their structures, thus providing many opportunities for breakthroughs in catalytic performance. At least three strategies, namely oxygen incorporation, ligand linkage, and heteroatom doping are effective in optimizing the cluster-based HER activity. McAllister and coworkers fabricated two oxygen-incorporated clusters, [Mo₂O₂(μ-S)₂(S₂)(S₄)]²⁻ and [Mo₂O₂(μ-S)₂(S₂)(S₂)]²⁻, by a wet chemical method.¹³⁶ In these clusters, in addition to the MoO bond, the Mo centers are also mutually connected by two sulfur bridges and each Mo center is coordinated to a terminal disulfide (S₂) or tetrasulfide (S₄) group. The catalyst achieved a very low overpotential of about 114 mV at a current density of 10 mA cm⁻², a marginal change of the observed overpotential after 1000 cycles in 1 M H₂SO₄ media, as well as Tafel slope values of 52–55 mV dec⁻¹ signifying an efficient Heyrovsky pathway. The simulation results revealed that the Heyrovsky step proceeds along the reaction of H₂[Mo₂O₂(μ-S)₂(S₂)₂]²⁻ + H₅O₂⁺ → H[Mo₂O₂(μ-S)₂(S₂)₂]⁻ + 2H₂O + H₂ with a reaction Gibbs free energy (ΔG°) of −7.3 kcal mol⁻¹, where the H[Mo₂O₂(μ-S)₂(S₂)₂]⁻ anion is considered as the catalytic cycle intermediate and H₂[Mo₂O₂(μ-S)₂(S₂)₂]²⁻ is formed via electron addition and proton addition pathways (Fig. 11c).

Except for the coordination of pure oxychalcogenide to the Mo center, organic molecules can also act as the ligands inside cluster-like complexes such as MoO(S₂)₂(2,2′-bipyridine) and [MoO(S₂)₂picolinate]⁻,^138,139 or act as linkers among individual clusters to build larger entities. Ji and colleagues developed a linkage method to organize single molecular clusters into ordered dimers, cages, and chains through sulfido and thiolate linkers.¹³⁷ This process commenced with the [Mo₃S₁₃]²⁻ cluster that underwent bromide ion replacement on terminal S₂ sites followed by further ligand substitution into thiolate [Mo₃S₇Br₃(SR)₃]²⁻ (Fig. 11d), where 1,3-benzenedithiol yields the dimer (NEt₄)₄[Mo₃S₇Br₃(m-BDT)_1.5]₂ (MOS-1) and 1,4-benzenedithiol produces the dimer (NEt₄)₄[Mo₃S₇Br₄(p-BDT)]₂ (MOS-2) or (NEt₄)₂[Mo₃S₇Br₃(p-BDT)_1.5] (MOS-3) depending on the orientation of clusters. MOS-3 delivers a low overpotential of 89 mV in driving a current density of 10 mA cm⁻² (see Table 6 for the comparison of the HER performance of MoS_x clusters), with a Tafel slope of 57 mV dec⁻¹, which are better than those of [Mo₃S₁₃]²⁻ in acidic media, demonstrating the key role of ligand modification. Additionally, Escalera-López and coworkers enhanced the HER properties via a heteroatom doping strategy, such as nickel-doped MoS_x nanoclusters.⁵⁴ Nickel element can be introduced into MoS_x by dual-target magnetron sputtering deposition to obtain (Ni–MoS_x)₁₀₀₀ clusters, which outperform undoped (MoS_x)₃₀₀ or Ni₂₂₀₀ clusters in the proton reduction reaction in acid media, also extending the modification methods and diversity of the clusters.

---

It bears some similarities among various amorphous MoS_x in terms of facile growth conditions, scalable wet chemistry, and high catalytic efficiency of the clusters; moreover, the flexibility and diversity of composition and linkage structures endow it with more opportunities to make superior HER catalysts with high performance, low cost, and great potential for practical applications. However, the identification of active sites in various clusters remains elusive and the stability of cluster structures remains suboptimal. Therefore, we believe that more attention should be paid to the following aspects, including the identification of active sites by advanced characterization techniques, the improvement of the catalyst durability by screening internally stable ligands, the exploration of the conversion/degradation mechanism of clusters during the hydrogen evolution process, and the enhancement of catalyst stability under continuous bubble impingement in acid by optimizing the growing integration of the clusters with the substrate. In conclusion, MoS_x clusters, like defective 2H/1T/3R-phase MoS₂ and amorphous MoS_x, also exhibit superior performance in catalyzing the HER (see Fig. 12 for the comparison of overpotentials of these defective MoS_x) and are potential alternatives to the noble metal catalysts in PEM electrolyzers.

4. Other catalytic applications of defective molybdenum sulfide

Molybdenum sulfide has demonstrated potential for catalytic applications in other areas beyond the HER, such as the NRR, CO₂RR, ORR, OER, nitric oxide (NO) reduction reaction, nitrate ion reduction reaction, organic molecular upgrading, etc. and as a catalytic electrode material for energy storage devices such as the lithium-oxygen batteries, lithium-sulfur batteries, sodium–sulfur batteries, zinc–air batteries, and so on. Recently, He and coworkers proposed the “self-gating” phenomenon, in which the electrocatalytic reactions of a semiconductor catalyst can modulate the surface conductivity itself, where n-type semiconductors support reduction reactions at negative potentials, such as the HER and CO₂RR; p-type semiconductors tend to support reactions at positive potentials, such as the ORR and OER; and bipolar semiconductors typically support both reactions (Fig. 13a).¹⁴⁰ Thus n-type MoS₂ is suitable for the HER and CO₂RR, while p-type MoS₂ such as phosphorus-doped MoS₂ tends to support both ORR and OER reactions, and Ta or Nb modulated bipolar MoS₂ materials are suitable for both the HER and ORR, thus guiding catalytic applications of MoS_x by tuning its semiconducting properties. Certainly, for a catalytic reaction, it is insufficient to consider the conductance of catalysts alone; more aspects such as the geometric structure, electronic state, adsorption energy, reaction energy barrier, and orbital hybridization also profoundly affect the performance of catalysts and need to be considered comprehensively. In the following, we will screen and describe several representative catalytic applications of defective MoS₂ in these electrocatalytic fields.

4.1. Nitrogen reduction reaction

The Haber–Bosch process is developed industrially for the artificial conversion of dinitrogen to ammonia, the main source of nitrogen for the production of fertilizers, whereas the process consumes 1–2% of the world's energy supply and contributes about 1.5% of global greenhouse gas emissions.³³ The electro-synthesis of ammonia offers a promising solution to address these environmental and energy issues by replacing harsh conditions (650–750 K and 50–200 bar)¹⁴¹ with ambient conditions and a renewable supply of clean energy. Breakthroughs toward electrocatalytic N₂ reduction (N₂ + 6H⁺ + 6e⁻ → 2 NH₃) have been made with Mo-based electrocatalysts, particularly MoS₂. Zhang and coworkers proposed that the edges of 2D MoS₂ are active for the NRR that is similar to the HER, however, the inert basal plane does not favor the reaction according to DFT calculations, which further suggests that the positively charged (+ 0.963[e]) Mo edges overcome a lower barrier to polarize and activate the N₂ molecule by forming N–Mo bonds to weaken the solid NN triple bond.²⁴ Intrinsically, two-dimensional MoS₂ with high crystallinity achieves only 1.17% FE and NH₃ yield of 8.08 × 10⁻¹¹ mol s⁻¹ cm⁻¹ in 0.1 M Na₂SO₄ (Fig. 13b), while the defect engineering on MoS₂ greatly facilitates the electrosynthesis of ammonia. To date, strategies such as bulk structural defects, crystalline phase modulation, engineering of S site, exotic atomic incorporation, and hybridization method, which will be discussed sequentially in the next section, have been proven to be effective for activating the NRR process, suppressing the HER reaction, and thus enhancing the product selectivity.

For direct bulk structural defects, Li and coworkers developed a defective MoS₂ nanoflower with dislocations and distortions on the crystal, leading to disordered atomic arrangement and lattice expansion on the defect sites.¹⁴² These defects and disorders with abundant unsaturated atoms boosted the catalytic activity of the NRR with NH₃ yield and FE of 29.28 μg h⁻¹ mg⁻¹ and 8.34%, respectively, much better than those of MoS₂ with fewer defects (13.41 μg h⁻¹ mg⁻¹ and 2.18%) under the same conditions. Such enhancements benefit from the lower energy barrier of the potential determining step (i.e., *NH₂ → *NH₃) along bare Mo atoms on the rim of defects, and also from the optimized d-band center moving closer to the Fermi level that indicates a stronger interaction between the catalyst surface and N₂ molecule. Chen and colleagues designed a porous atomic layered 2H-phase MoS₂ in which porosity induced the generation of adjacent Mo atom pairs.¹⁴³ The Mo atom pairs effectively reduced the thermodynamic overpotential of the NRR and assisted in obtaining a high FE of 44.36% and a sizeable NH₃ yield of 3405 μg h⁻¹ mg_cat⁻¹. Wu and coworkers synthesized a sub-monolayer MoS_2−x structure with in-plane S-vacancies, which tune its catalytic performance through adjusting its affinity and stability with nitrogen intermediate in a dynamic binding paradigm, thus circumventing the scaling relationship and obtaining excellent electrocatalytic NRR performance at ultra-low overpotential.¹⁴⁵

The crystalline phase affects the NRR reaction path and HER reaction free energy; so, the catalytic performance can be tuned by phase modulation. Liu and coworkers prepared 1T/2H mixed-phase MoS₂ with tunable 1T content using a self-sacrificing template method.¹⁴⁶ The current density, NH₃ yield rate, and FE_NRR gradually increased with increasing 1T-phase content (from ∼10% to ∼75%), while the FE_HER decreased. DFT analysis showed that the Mo-edge of 1T-phase MoS₂ had lower N₂ adsorption energy than that of 2H-phase MoS₂. Moreover, the low ΔG_H* value (−1.9 eV) at low hydrogen coverage (<25%) indicates that the adsorbed H readily binds to S atom in the basal plane of 1T-phase MoS₂, while the barrier for H desorption/combination cannot be overcome at the electrode potential of (−0.36 V, critical potential for overcoming the barrier of the potential-determining step). The *H remaining at the edge of the basal plane can be directly transferred to nearby bound N₂ or nitrogen intermediates, which greatly accelerates the NRR process. Therefore, although the 1T-phase MoS₂ can accelerate the NRR and HER, it will reduce the competition between the HER and NRR due to the separation of active sites, thus synchronizing the selectivity and activity for the NRR. Lin and colleagues compared three crystalline phases of 2H, 1T′ (zigzag chainlike clustering of Mo atoms) and 1T′′′ (trimerization of Mo atoms).¹⁴⁷ The Mo–S distortion and Mo–Mo metal in 1T′′′-phase MoS₂ raise the local electron density surrounding the Mo–Mo chains and lead to the uplift of the d band for Mo, which facilitates the acceptance of lone-pair electrons from N₂ and the electron transfer from the d band to NN π antibonding orbital, thus enhancing the adsorption to N₂. In the 1T′′′ phase, the 1T′′′-3 site (the region surrounded by three Mo–Mo bonds) reinforces the electron loss of NN triple bonds and facilitates the NN breakage prior to the formation of a strong Mo–N σ bond. As a result, the 1T′′′-phase MoS₂ catalyst delivered an NH₃ yield rate of 9.09 μg h⁻¹ mg⁻¹, which is 2 and 9 times higher than those of 1T′ and 2H phases, respectively, as well as a FE of 13.6%.

The active sites of the NRR originate from the bare Mo atom, while the HER process arises from the S-edge primarily and also from the Mo-edge sites.³⁴ Competition with the HER is inevitable because the HER bears a lower energy barrier and a faster reaction rate, thereby a higher NRR FE can be obtained by suppressing the HER on S-edge sites. Engineering of S sites such as interaction with Li, or substitution with F, has demonstrated strong HER suppression.^34,144,148 Liu and coworkers disclosed that the in-operando created strong Li–S interaction on S-rich MoS₂ nanosheets during the NNR in a lithium-containing electrolyte led to superior NRR catalytic activity and significant HER suppression, delivering an NH₃ yield of 43.4 μg h⁻¹ mg⁻¹_MoS2 and a FE of 9.81%, which are more than 8 and 18 times higher than those without the Li–S interaction (5.35 μg h⁻¹ mg⁻¹_MoS2; 0.53%).³⁴ Theoretical analysis revealed that the edge sites of MoS₂ are electrochemically active for the NRR compared to the inert basal plane, while the Li–S interaction is thermodynamically more inclined to occur at the S-edge sites. The strong Li–S interaction significantly suppresses the HER by retarding the HER process on S-edge sites due to its uphill ΔG_H* from 0.03 to 0.47 eV and by impeding the Heyrovsky/Tafel process due to downhill ΔG_H* from 0.09 to −0.72 eV on Mo-edge sites. This ensures firm H* adsorption on Mo and also leads to the positively charged Mo edge sites with dramatically increased N₂ adsorption energy (ΔG_N2 changes from −0.32 to −0.70 eV) and a reduced NRR activation energy barrier for the formation of *N₂H and other intermediates (Fig. 13c).

Moreover, Patil and coworkers explored the intercalation of Li on 1T-phase MoS₂ (Li-1T-MoS₂), which exhibits stronger N₂ adsorption compared to Li intercalated 2H-phase MoS₂ (Li-2H-MoS₂), as the former possesses shorter d(Mo–N) and d(N–Li) distances according to structural calculations, which suggests a pseudo-six-membered ring containing the interaction from N–Li–S–Mo–S–Mo.¹⁴⁴ Li-1T-MoS₂ enhances the N₂ adsorption on the Mo active sites with upgraded N₂ adsorption energy to −0.72 eV from −0.28 eV of 1T-phase MoS₂ and suppresses the HER on S atoms since the theoretical ΔG_H* increases from 0.03 eV to 0.49 eV after Li intercalation (see Fig. 13d for the structure diagram). Experimentally, a high FE of 27.66% was realized from Li intercalated low-crystalline 1T-phase MoS₂ formed during the NRR in the LiClO₄ electrolyte. In addition to the Li interaction/intercalation engineering methods mentioned above, the F substitution on the S site can also achieve the suppression of the HER and increase NRR FE, as reported by Liang and coworkers.¹⁴⁸ The smaller size and higher electronegativity of F, and the compression of the MoS₂ interlayer space caused by the introduction of F, play a role in restraining the HER, as evidenced by the decrease in the current density of hydrogen evolution compared to pristine MoS₂. Another aspect is that the Mo sites with empty d orbitals on the MoS₂ edge acquire stronger N₂ adsorption since the free energy change on the F-MoS₂ edge (−0.85 eV) is lower than that of MoS₂ (−0.38 eV) and bears the lower uphill free energy (0.36 eV vs. 0.64 eV) in the first hydrogenation step, which is considered as the NRR rate-determining step (*N₂ → *NNH). As a result, the FE of the NRR on F-MoS₂ increased to 20.6% with a NH₃ yield of 35.7 μg h⁻¹ mg⁻¹.

The incorporation of exotic atoms into the MoS₂ lattice creates many defects such as hetero-interfaces or S-vacancies, and the synergistic effect leads to an increase in catalytic performance. Zhao and coworkers plotted the limiting potentials of the HER and the NRR versus the N₂H adsorption energy (potential-determining step) and predicted that the onset potentials of the NRR increase in the following order, Mo < Ru < V = Rh < Fe ≈ Co < Cr < Ti < Mn < Ni < Sc, based on a computational model in which the transition metal atom is embedded in a MoS₂ monolayer with S-vacancies and occupies the S-vacancy site.¹⁴⁹ A similar conclusion was drawn by Yang and colleagues. In their high-throughput screening model, a single transition metal atom was anchored on the top of the Mo site and stabilized by neighboring S atoms of MoS₂; and the optimal NRR overpotential, as well as selectivity, was realized for the Mo single atom compared to loading of other metal atoms.¹⁵⁰ Dang and coworkers proposed that a transition metal atom supported on graphene and simultaneously adsorbed on the MoS₂ basal plane with S-vacancies acts as a single-atom promoter (SAP) by inducing strong spin polarization on the exposed Mo atoms, which greatly facilitates N₂ adsorption and polarization.¹⁵¹ In their designed system, Sc or Y SAPs possess favorable NRR catalytic activity and selectivity. Many experimental studies have also been conducted to prove that introduced external atoms increase the NRR catalytic activity of the MoS₂ catalyst.

Among the exotic elements, the integration of Fe atoms and MoS₂ stands out in catalyzing the NRR. Chen and coworkers revealed that Fe- and O-decorated edge-MoS₂ (Fe₁-4O@Mo-edge-MoS₂) is responsible for the high performance of the NRR due to its low Gibbs free energy and high selectivity by theoretical simulations.¹⁵² Xie and coworkers theoretically demonstrated that the MoS₂-supported single Fe site induces a local electric field effect that interacts with the dipole moment of NH₂–NH₂ species or MoS₂-supported double/triple Fe sites induce synergistic effects, both of which reduce N–N bond cleavage barriers.¹⁵³ Li and coworkers experimentally reported a protrusion-like Fe single atom immobilized on monolayer MoS₂ (SACs–MoS₂–Fe) on the top of Mo, by coordinating Fe atoms with three nearest S.¹⁵⁴ The Fe atom with high-curvature protrusions induces an electric field that triggers interfacial polarization, and the enhancement of electric field shown by Kelvin probe AFM follows the same trend of improvement in NRR activity (Fig. 14a and b). In detail, the increased electric field shortens the Fe–N bond and reduces the interfacial barrier, allowing electron transfer to N₂ and thus promoting N₂ polarization. In particular, based on the interfacial polarization field, more electrons are transferred from the Fe d-orbital to the N₂ π_2p* antibonding orbital. As a result, the SACs-MoS₂–Fe with 2% Fe single atoms delivered a high ammonia production rate of 36.1 ± 3.6 mmol g⁻¹ h⁻¹ with a FE of 31.6% ± 2% at −0.2 V_RHE, as well as excellent durability for 50 h in 0.1 M KCl aqueous electrolyte. Su and coworkers developed Fe–MoS₂ nanosheets with atomic-level dispersion of Fe into MoS₂.¹⁵⁵ This Fe–MoS₂ exhibits increased catalytic performance with a faster NRR process, thanks to the lower energy barrier (0.37 eV) of the potential-determining step and suppressed HER due to increased energy barrier (0.15 eV), compared to that of pristine MoS₂ (0.50 eV, 0.03 eV). Consequently, the catalyst showed an NH₃ production rate of 8.63 mg_NH3 mg_cat⁻¹h⁻¹ with a FE of 18.8%. He and colleagues introduced both Fe doping and S-vacancies in mixed-phase MoS₂ to obtain V_S-Fe-doped 1T/2H MoS₂/C (V_S-Fe–MoS₂/C), which exhibited an ammonia yield rate of 17.8 ± 0.7 μg h⁻¹ mg⁻¹ and a FE of 9.2% under ambient conditions.¹⁵⁶ The NRR performance of V_S-Fe–MoS₂/C outperforms those of Fe–MoS₂/C and MoS₂/C, showing the advantages of the multi-engineered synergistic approach in improving the NRR.

For other metal atoms, Suryanto and coworkers reported a catalyst of Ru cluster loaded MoS₂ with S-vacancies for NRR catalysis, delivering an NRR FE as high as 17.6% and a NH₃ yield rate of 1.14 × 10⁻¹⁰ mol cm⁻² s⁻¹ at optimized temperature (50 °C) and potential (−150 mV).³³ Compared to Ru/1T-MoS₂, 2H-phase MoS₂ decorated with amorphous Ru clusters (Ru/2H-MoS₂) exhibited higher NH₃ yields, also exceeding those of 2H-phase MoS₂ and 1T-phase MoS₂. The outstanding properties of Ru/2H-MoS₂ in catalytic NRR stem from the synergistic interaction between Ru clusters as N₂ binding sites and nearby isolated S-vacancies as H binding/provider sites for successive hydrogenations, thereby the interface between the Ru cluster and the S-vacancy is expected to play the key role in catalyzing NRR efficiently. Lou and coworkers discovered that the doped Co atoms collaborating with S-vacancies on MoS₂ also facilitate the conversion of dinitrogen to ammonia under electrosynthesis conditions.¹⁴¹ The MoS₂ monolayer shows very limited NRR activity due to the S passivation on both sides, whereas S-vacancy sites form a semi-open-edge configuration and thus result in the enhanced affinity of N₂ molecules on the exposed three Mo atoms, which benefits from back-donation of electrons to dinitrogen's antibonding orbitals, facilitating the activation and subsequent dissociation of the NN triple bond. Further Co doping makes N₂ preferably to deviate from the tri-Mo center and to approach one of the Mo atoms, which leads to lower energy barriers for the adsorption and hydrogenation steps according to DFT simulations (Fig. 14c). The Co-doped MoS_2−x catalyst prepared by a hydrothermal method showed a higher FE (over 10%) and yield rate (over 0.6 mmol h⁻¹ g⁻¹) than those of MoS_2−x (1.7%, 0.32 mmol h⁻¹ g⁻¹), and the NRR activity is mainly from the basal plane according to the localized electrochemical measurements (Fig. 14d and e), indicating the effective modulation of the basal plane by Co doping and the S-vacancies.

The introduction of nonmetallic elements also contributes to the NRR activity. Fei and colleagues used a hydrothermal route to introduce P dopants to synthesize P-doped S-vacancy-rich MoS₂.¹⁵⁷ The introduced P atoms not only induce S-vacancies into MoS₂, which is the active site for adsorption and conversion of N₂ molecules, but also P itself modulates the electronic structure of MoS₂, thus promoting the interfacial electron transfer between the catalyst and N₂ molecules to enhance the adsorption of N₂ molecules on the P-S_V-MoS₂ surface. The optimal catalyst had a superior electrocatalytic N₂–NH₃ conversion efficiency, with a FE value of 12.22% and an NH₃ yield rate of 60.27 μg h⁻¹ mg⁻¹ at −0.6 V in 0.1 M Na₂SO₄. Chen and coworkers synthesized one-dimensional screw-like MoS₂ nanosheets with oxygen partially replacing sulfur (1D-MoS_2−xO_y), delivering a remarkable ammonia yield rate of 5.56 × 10⁻⁸ mol s⁻¹ cm⁻² and a high faradaic efficiency of 9.86%.¹⁵⁸ The enhanced performance is attributed to the partial atom substitution that reduces the Gibbs free energy of the potential-determining step.

Moreover, it is possible to optimize NRR performance by forming hybrids to lower the energy barriers or alter the NRR reaction pathways. Chen and colleagues constructed a 1T′-MoS₂/Ti₃C₂ catalyst by a hydrothermal method.¹⁵⁹ The hybrid catalyst attained a large NH₃ yield rate of 31.96 μg h⁻¹ mg_cat⁻¹ with a high faradaic efficiency of 30.75%. 1T′-MoS₂/Ti₃C₂ presents a lower energy barrier than pure 1T′-phase MoS₂ in the PDS step (*N₂ → *N₂H), indicating that the MoS₂-based hybrid catalyst makes the activation and reduction of *N₂ thermodynamically more favorable. Xu and coworkers developed a 1T-MoS₂ NDs/g-C₃N₄ composite catalyst in which 1T-phase MoS₂ nanodots were anchored on g-C₃N₄ and the composite catalyst showed the optimal reaction path of the NRR with both associative distal and alternating pathways.¹⁶⁰ The 1T-MoS₂ NDs/g-C₃N₄ catalyst achieved a large NH₃ yield of 29.97 μg h⁻¹ mg_cat⁻¹ and a high faradaic efficiency of 20.48% (see Table 7 for the comparison of NRR performance of the defective MoS₂ catalysts). Zi and colleagues designed a S-vacancy-rich 1T-phase MoS₂ monolayer confined on MoO₃.¹⁶¹ In this hybrid, the appearance of depleted electrons in MoO₃ and the aggregation of electrons in MoS₂ indicate charge transfer from MoO₃ to MoS₂. The S-vacancy active sites strengthen the interaction between MoS₂ and N₂; specifically, the antibonding 2π* orbital of N₂ is shifted towards the Fermi level due to the partial occupation of back-donated electrons from Mo d orbitals, resulting in a weakening of the highly inert NN bond and thus an enhancement of the catalytic activity.

---

However, it is worth noting that although the NRR reaction has achieved high FE based on several strategies mentioned above, the ammonia yield rate remains very limited due to the dominance of the HER in the electrochemical process, especially at higher overpotential, and the selectivity of the NRR continues to be an issue for MoS₂-based electrocatalysts. Therefore, there is an urgent need for more attempts on MoS₂ guided by structural defects, crystal phase regulation, S-site interactions, atomic doping, hybrid design, and other engineering approaches to further improve the NRR performance by enhancing N₂ adsorption, reducing the activation energy, and suppressing the HER process.

4.2. Carbon dioxide reduction

The overconcentration of CO₂ accumulated from excessive consumption of fossil fuels has become one of the major contributors to global warming. Replacing fossil fuels with clean energy sources, such as photovoltaic and green hydrogen to prevent carbon emissions and developing CO₂ capture/storage/conversion technologies to reduce the current CO₂ concentration are considered as promising technologies to address the environmental issues. In particular, the conversion of CO₂ into valuable chemicals through direct electrochemical reduction powered by renewable energy sources such as solar, wind, and tidal energy can provide both environmental and economic benefits. Electrocatalysts including Au, Ag, Pd, Zn, Sn, Bi, Cu, carbon materials, metal oxides, etc.¹⁶³ have demonstrated their catalytic performance in CO₂ electroreduction, although they still face different challenges such as high cost (for precious metals), high overpotential (for some metals), low-value products (mainly the C₁ product), low faradaic efficiency, and low selectivity to target products, which hinder their implementation at the commercial scale. To this end, MoS₂ having shown superior catalytic performance in the CO₂ reduction reaction offers a promising alternative.

Francis and coworkers disclosed that MoS₂ single crystal terraces or films with low edge-site densities could convert CO₂ into 1-propanol in aqueous electrolytes with FE values of ∼3.5% or ∼1%, respectively.¹⁶⁷ In contrast, the molybdenum-terminated edges of MoS₂ favor the formation of carbon monoxide (CO) in an ionic liquid. Asadi and coworkers found that the MoS₂ bulk with a Mo-terminated edge, compared to Ag, exhibited a much higher current density of 65 mA cm⁻² at −0.764 V with a FE value close to 98%, when used as the catalyst for the reduction from CO₂ to CO in 4 mol% EMIM-BF₄ aqueous solution. Such a superior performance is attributed to the much higher d-electron density in the vicinity of the Fermi level of Mo-edge atoms so that the Mo d-electron can form metallic edge states and freely supply excess d electrons to the attached reactants at the Mo edge (Fig. 15a).²³ Accordingly, a vertically aligned MoS₂ nanosheet terminated by Mo atoms was also fabricated, further achieving a 2-fold higher current density for the CO₂ reduction than the bulk state (Fig. 15b). With regard to the CO₂–CO pathway (* + CO₂ + H⁺ + e⁻ → COOH*; COOH* + H⁺ + e⁻ → CO* + H₂O; CO* → CO + *), the formation of COOH* is endergonic and thus known as the rate-limiting step on certain metal surfaces (e.g., Pd, Au, Cu, and Ag), whereas it is exergonic on MoS₂ metal edge sites because of the strong binding of the adsorbed intermediate to the edge thanks to its d-band center close to the Fermi level. In addition, more stable CO* can be formed on the metallic edge, which is also an exergonic process that leads to a lower overpotential for CO₂ conversion. However, the desorption of CO becomes a rate-limiting step since the strong binding of the adsorbed CO on MoS₂ (Fig. 15c) suppresses its release.^164,165 Therefore, accelerating the desorption of CO product from the MoS₂ catalyst surface can facilitate CO₂ conversion with a higher turnover frequency and FE. Various defects engineering methods on MoS₂, including doping, alloying, hybridization, and surface decoration, have been proven to be effective in stimulating the CO desorption process, which will be discussed sequentially in the next section and finally we will discuss defect strategies that facilitate the production of high-value gas/liquid products by electrocatalytic or photo(electro)catalytic means.

Abbasi and coworkers achieved decreased binding strength between the Mo edge and CO by substitutional doping of Nb atoms into the MoS₂ structure, accelerating the turnover of CO desorption, while retaining the exergonic behavior of forming COOH* and CO* (Fig. 15d).¹⁶⁵ As a result, in the overpotential range of 100–300 mV, the CO formation TOF of the p-type doped vertically aligned VA-Mo_0.95Nb_0.05S₂ with 5% Nb concentration was about 1 and 2 orders of magnitude higher than that of pristine VA-MoS₂ and Ag nanoparticles, respectively (Fig. 15e). For other metals, Mao and colleagues systematically investigated the electrocatalytic activity of the CO₂RR on the MoS₂ edge doped with transition metal atoms using high-throughput DFT calculations.¹⁶⁸ It was found that V, Zr and Hf dopants in MoS₂ could significantly reduce the binding energies of CO, thus promoting the desorption of CO from the MoS₂ edge and achieving optimal CO₂RR performance. Beyond metal doping, Lv and coworkers suggested that non-metallic nitrogen doping also helps to weaken the adsorption strength of CO* on the Mo atoms of N-MoS₂, where N doping lowered the CO desorption Gibbs free energy (ΔG) from 0.53 eV (undoped MoS₂) to 0.49 eV (Fig. 15f), leading to greatly enhanced catalytic activity towards the CO₂RR.¹⁶⁶ Huang and coworkers combined experimental and theoretical studies to elucidate that compared to single N or P doping, N,P co-doped MoS₂ achieves higher current density and FE and synergistically reduces the binding strength between the Mo sites and CO moieties, resulting in improved performance.¹⁶⁹

In addition to the doping strategy, Xu and coworkers developed an alloyed MoSeS monolayer to weaken its binding strength with CO.¹⁷⁰ Unlike MoS₂ and MoSe₂ where the charge density is mainly distributed on Mo atoms, MoSeS has the shortened Mo–S bond and the lengthened Mo–Se bond, leading to the off-center charge density of the Mo atoms (Fig. 16a). Consequently, the overlap of the d orbitals of Mo atoms and the p orbitals of C atoms in CO* is reduced and thus weakens the CO adsorption strength as verified by the elongated C–Mo bond and facilitates the rate-limiting CO desorption step, as evidenced by the lowest CO desorption temperature/potential on the MoSeS alloy in the experiment. The off-center charge density of the Mo atom also favors the stabilization of the COOH* intermediate on the charge accumulation region between two neighboring Mo atoms, which leads to a higher exothermic and spontaneous COOH* formation process. As a result, the higher FE than MoS₂ (16.6%) and MoSe₂ (30.5%) was obtained for the MoSeS monolayer (45.2%) at −1.15 V, with a current density of 43 mA cm⁻², which was about 2.7 and 1.3 times higher than those of MoS₂ and MoSe₂, respectively.

Li and coworkers proposed a hybridization strategy of edge-exposed 2H-phase MoS₂ with N-doped carbon (NCMSH).¹⁷¹ Attributed to the electron transfer from the electron-rich N-doped carbon to the electron-deficient Mo edge by theoretical calculations, the acquired electrons around the Mo atom are believed to facilitate the potential limiting step (see Fig. 16b for free energy profiles of the CO₂RR and the electron density change in the NCMSH model), i.e., the reduction of COOH* and the recovery of the original Mo site. This catalyst with a large exposed MoS₂ edge at the N-doped carbon site displayed a FE of 92.68% and a high CO production rate of 31.80 mA cm⁻², which corresponds to a threefold enhancement of the MoS₂ without N-doped carbon. Moreover, Lv and coworkers devised a surface modification method to decorate the exfoliated MoS₂ with fluorosilane (FAS).¹⁷² On the basis of changing the electronic properties of the edge Mo atoms, the decorated FAS on the MoS₂ surface brings down the Gibbs free energy ΔG (0.72 eV) of the rate-limiting CO* desorption step from the undecorated MoS₂ (1.32 eV), thereby weakening the adsorption strength of CO with Mo atoms and accelerating the desorption of CO product. Furthermore, the hydrophobic and aerophilic properties of the FAS-decorated MoS₂ surface suppress the HER process and create sufficient three-phase contact points for the CO₂ reaction (Fig. 16c). As a result, the hydrophobic exfoliated MoS₂ (H-E-MoS₂) obtained a higher CO FE% (81.2%) (see Fig. 16d for CO and H₂ FE% of H-E-MoS₂) and a lower onset potential (−0.24 V) than those of pure exfoliated MoS₂ (41.2%; −0.3 V) and bulk MoS₂ (19.8%; −0.43 V). Li and colleagues synthesized amorphous MoS_x hybrids on polyethylenimine (PEI)-modified reduced graphene oxide (rGO-PEI-MoS_x), in which PEI also suppressed HER and stabilized the CO₂˙⁻ intermediate species during CO₂ reduction, thus synergistically improving the catalytic activity of MoS_x, attaining an onset overpotential of 140 mV and a maximum FE of 85.1% (see Table 8 for the comparison of CO₂RR performance of the aforementioned MoS_x catalysts).¹⁷³

---

Besides the pursuit of higher FE and lower overpotentials of CO formation on defective MoS₂ electrocatalysts, liquid products are also desirable. Although MoS₂ single crystals have shown success in the production of liquid 1-propanol, the low FE (∼3.5%) hinders its practical use for the CO₂RR. Engineering MoS₂ through defective strategies can facilitate the production of high-value gas/liquid products by electrocatalytic or photo(electro)catalytic means. From the perspective of electrocatalysis, Ru@1T′-MoS₂, Fe@MoS₂, Co@MoS₂, Ni@MoS₂, MoS₂-IrCl₃ and monolayer MoS₂ with 2S-vacancies are potential catalysts for the production of CH₄ in different theoretical works.^174–177 For the CH₃OH production, Pt@1T′-MoS₂ and MoS₂-IrF₃ are promising.^174,176 For the generation of HCOOH, Re@1T′-MoS₂ and Zn@1T′-MoS₂ are achievable.¹⁷⁴ For the 1-propanol production, V_S-MoS₂ is hopeful.¹⁷⁸ These vacancies or doping defects alter the catalytic free energy of the intermediate species by breaking the scaling relationship, changing the orbital orientation, or exploiting steric hindrance.^176,177,179 For example, Nørskov and coworkers proposed that in transition metal-doped MoS₂, there are two scaling relationships: one for the doped metal and one for sulfur.¹⁷⁹ The metal site favors CO* adsorption, while the S site favors COOH* and CHO*, so that during the reduction of CO* to CHO*, the intermediate switches from the metal site to S site, thus breaking the constraint of scaling relation and leading to the stabilization of CHO* species and the formation of hydrocarbons. In another Jiang's report, the authors proposed a bowl site structure, which can be a 2S-vacancy or a Mo atom substituting a S₂ column and the bowl sites can achieve the selective reduction of CO₂ to CH₄.^177,180 In this configuration, *CO formed by CO₂ reduction is adsorbed in the bowl site and due to the steric hindrance, the adsorbate on the bowl site forms repulsive interaction with external species, thus preventing the hydrogenation of carbon and enabling the hydrogenation of oxygen, resulting in the formation of *COH instead of *CHO. Similarly, in the next step, *COH has difficulty forming *CHOH, but instead dehydrates to form *C, thus providing the basis for subsequent protonation to form CH₄. In contrast, on the molybdenum sulfide surface without a bowl structure as well as steric hindrance, multiple catalytic pathways coexist, leading to a diversity of products. On the basis of the above theoretical work, relevant experimental evidence is expected to emerge in the near future.

The band gap of MoS₂ (1.17 eV) cannot match the CO₂ reduction potential¹⁸¹ and the photocatalytic activity of CO₂ could be promoted by doping or constructing a heterostructure to optimize the energy band structure or facilitate carrier transfer. Peng and coworkers developed Co-doped MoS₂ with an optimized valence band (0.89 V) and conduction band (−0.52 V) to match the reduction potential of CO₂ and the oxidation potential of H₂O, thus obtaining photocatalytic activity.¹⁸² The methanol yield rate in the photoelectrochemical (PEC) reactor reached 35 mmol L⁻¹. Wang and colleagues reported a Z-scheme heterojunction constructed of 3D-SiC@2D-MoS₂.¹⁸³ Thanks to the more positive valence band and higher hole mobility of MoS₂, and the more negative conduction band and higher electron mobility of SiC, the material achieved CO₂ to CH₄ conversion without sacrificial reagents under visible light (λ ≥ 420 nm) irradiation, delivering a CH₄ yield rate of 323 μL g⁻¹ h⁻¹ and stable operation for 40 h. Yu and coworkers synthesized a d-UiO-66/MoS₂ by integrating MoS₂ nanosheets into porous defective UiO-66 for photocatalytic conversion of CO₂ and H₂O to CH₃COOH under visible light irradiation without adducts.¹⁸⁴ In this system, MoS₂ contributes to the reduction of the band gap (E_g), making it possible to generate active carriers under visible light; MoS₂ also facilitates the adsorption of CO₂. Moreover, bridging Mo–O-Zr on the interface promotes the transfer of photogenerated carriers and lowers the energy barrier for C–C coupling. As a result, the evolution rate and selectivity of CH₃COOH obtained by d-UiO-66/MoS₂ reached 39.0 μmol g⁻¹ h⁻¹ and 94%, respectively. Overall, building on the above success, more types of liquid products and higher yields remain to be expected.

4.3. Metal–sulfur batteries

The current energy density of commercial lithium-ion batteries (≈300 mA h g⁻¹)¹⁸⁵ is approaching its theoretical limits, so there is an urgent need to develop next-generation energy storage batteries with high energy density, fast charging, long cycle life, environmental friendliness, and excellent cost-effectiveness to meet the growing demand for industrial electrification. Among the advanced alternatives, metal–sulfur batteries are promising candidates benefiting from their high theoretical capacity of sulfur (1675 mA h g⁻¹),¹⁸⁶ low-cost and abundant sulfur resources, and environmental benignity. Lithium–sulfur (Li–S) batteries with lithium as the anode and S as the cathode have a theoretical specific energy density of 2600 W h kg⁻¹,¹⁸⁷ while the sodium–sulfur (Na–S) batteries with Na as the anode possess an energy density of 1274 W h kg⁻¹,³⁶ making both of them promising candidates. Upon discharging of Li–S batteries (LSBs), the lithium at the anode is oxidized to form Li⁺ ions (Li → Li⁺ + e⁻) that move towards the cathode through a diaphragm. In contrast, the sulfur at the cathode is gradually reduced from cyclic octasulfur (S₈) to soluble long-chain lithium polysulfide (LiPS) intermediates (Li₂S_n, 3 ≤ n ≤ 8) that undergo further reduction to form insoluble short-chain Li₂S₂ and Li₂S (S₈ + 16Li⁺+ 16e⁻ → 8Li₂S) (Fig. 17a).³⁵ On the other hand, the subsequent charging process undergoes a reverse reaction that reconverts Li₂S to elemental sulfur (8Li₂S → S₈ + 16Li⁺ + 16e⁻) and lithium metal (Li⁺ + e⁻ → Li), resulting in a reversible discharge/charge cycle. Several obstacles, including poor conductivity of S, the shuttle effect of LiPSs, volumetric expansion, sluggish kinetics of sulfur transformation, self-discharge, and lithium dendrite, restrict the commercial applications of LSBs.^35,188,189 As one of the key issues, the shuttle effect arises from the diffusion of intermediate LiPSs from the cathode to the anode, which leads to irreversible loss of sulfur-active material, passivation of the anode, and increased internal resistance due to their high solubility in the organic electrolytes, resulting in low Coulombic efficiency, decreased battery capacity, and poor cycling stability.^190,191

The shuttle effect can be suppressed by reducing the solubility of LiPSs in the electrolyte through physical blocking or chemisorption strategies and by accelerating the conversion of LiPSs to solid Li₂S₂/Li₂S using catalysts.^191,192 Similar to the reaction mechanism of LSBs, room-temperature sodium–sulfur batteries (Na–S batteries) face the same shuttle effect and thus also require the adsorption and rapid conversion of soluble sodium polysulfides (NaPSs).¹⁹³ Among the polysulfide conversion catalysts, MoS₂ is considered as one of the powerful contenders, thanks to its strong adsorption to polysulfides (i.e., LiPSs, NaPSs) through chemical binding and its intrinsic high chemical activity.^194,195 MoS₂ edge sites with unsaturated coordination exhibit stronger interaction and catalytic activity, while the basal plane with saturated coordination is relatively inert in catalyzing polysulfide conversion,^187,196,197 similar to HER catalysis. Optimization of MoS₂ catalysts using defect engineering provides a route to suppress the shuttle effect and to improve the performance of metal–sulfur batteries. As for two-dimensional materials, process optimization can be performed by creating structural defects, vacancies, doping and alloying within the layers, by interlayer expansion and interlayer ion intercalation, and by constructing heterojunctions or hybridization outside the layers, which will be discussed in turn in the next section. In addition, for non-2D MoS_x, excellent catalytic properties can be obtained through methods such as amorphization, which will be discussed at the end.

For the within-layer engineering approach, Lin and coworkers reported a sulfur-deficient 2H-phase MoS₂ nanoflake electrocatalyst via liquid phase exfoliation and the removal of sulfur atoms using thermal treatment in hydrogen.¹⁹⁸ The catalytic behavior of MoS_2−x for polysulfide redox reactions was evaluated in Li₂S₆ electrolyte, where MoS_2−x/rGO exhibited high activity and reversibility with four distinct peaks implying two-step lithiation of sulfur to Li₂S and its reversible reaction (Fig. 17b), while MoS₂/rGO exhibited broad redox peaks with remnants. In terms of practical performance, the MoS_2−x/rGO/S composite used as the cathode in coin cells with a lithium metal anode delivered an increased capacity of 826.5 mA h g⁻¹ compared to those of MoS₂/rGO/S (473.3 mA h g⁻¹) and rGO/S (161.1 mA h g⁻¹) at a rate of 8 C (Fig. 17c) and enhanced cycle stability of the sulfur cathode with a capacity decay of 0.083% per cycle over 600 cycles at a 0.5 C rate. The improved rate performance is attributed to the higher affinity of MoS_2−x/rGO and the accelerated conversion kinetics of polysulfides on the sulfur deficiency sites that are regarded as catalytic centers, leading to a reduced sulfide loss and therefore a sustained cycling capacity. Du and colleagues fabricated 1T-phase MoS₂ (∼90%) monolayers by converting Mo-based MXenes (Mo₂CT_x and Mo_1.33CT_x) in H₂S at high temperatures.²⁰⁰ The Mo vacancy in Mxenes promotes the gliding of the sulfur layer to form 1T-phase MoS₂, while Mo_1.33CT_x-derived MoS₂ with more Mo vacancies exhibits intense chemisorption and high catalytic activity for LiPSs, delivering a reversible capacity of 736 mA h g⁻¹, an outstanding rate capability of 532 mA h g⁻¹, and good stability for 200 cycles in Li–S batteries.

LIB performance can also be improved when S-vacancy defects and phase tuning are combined. Zhang and coworkers developed an MXene/1T-2H MoS₂-C catalyst with plentiful positively charged S-vacancies and abundant 1T phase, which was verified to exhibit stronger trapping and adsorption ability of polysulfides (including S₈/S₆²⁻ and S₄²⁻/S₆²⁻) than MXene/2H MoS₂-C and MXene, through visual observation (i.e., color changes of a Li₂S_x solution before and after adsorption) and UV-vis spectroscopy during discharge.²⁰¹ As a result, MXene/1T-2H MoS₂-C achieved a remarkable improvement in battery performance, with a capacity of 1194.7 mA h g⁻¹ at 0.1 C and a fading rate of 0.07% per cycle over 300 cycles at a 0.5 C rate. Moreover, the MXene/1T-2H MoS₂-C cathode showed reliable performance in soft-package batteries, even after the bending test. Cheng and coworkers suppressed the shuttle effect by using sulfur-deficient 1T-phase MoS₂ nanoflower-decorated graphene (FM@G) as both the cathode and separator, where the S-deficient metallic MoS₂ with an accessible catalytic surface at the cathode boosts chemisorption and catalytic conversion of polysulfides, while the separator further suppresses the permeation of LiPSs to the anode.²⁰² The synergistic design resulted in excellent cyclability with a decay rate of 0.058% per cycle for 500 cycles at 1.0 C rate and an initial capacity of 1057 mA h g⁻¹. More similar catalyst systems incorporating defects and phase modulation, such as edge-rich 1T-phase MoS₂ nanodots²⁰³ and CF@2H/1T MoS₂,²⁰⁴ have also been shown to be effective in the adsorption and catalysis of LiPSs.

Heteroatom doping within the MoS₂ layer improves the adsorption capacity and conversion kinetics of polysulfides by inducing transformation into the metallic 1T phase, improving the electrical conductivity, and introducing S-vacancies. Liu and coauthors recently pointed out that the formation energies of the 1T phase and S-vacancy after Co doping are both lower than those of the 2H phase by DFT calculations, and the decomposition energies of Li₂S₄, Li₂S₂, and Li₂S on the Co-doped 1T-phase MoS₂ are lower than those of 1T-phase MoS₂.²⁰⁵ In the experiments, a hydrothermal method and the use of excess Co precursor were adopted to synthesize Co-doped MoS₂, where Co was substitutionally doped. In such a catalyst, abundant S-vacancies were introduced and the 1T content was increased. As a result, the Co–MoS₂ on graphene showed stronger static adsorption of Li₂S₆, more distinct redox peaks in the symmetric tests, a lower fading rate of 0.029% per cycle over 1000 cycles at the 1 C rate, and a higher capacity of 941 mA h g⁻¹ at 2 C rate in Li–S coin cells. Lin and coworkers employed a cobalt and phosphorus co-doping strategy, where cobalt induced the phase change to metallic 1T phase and phosphorus had an affinity to cobalt, to form Co–P coordinated sites with high catalytic activity for polysulfide conversion.¹⁸⁷ Consequently, the sulfur cathode using the Co–P co-doped MoS₂ catalyst (P-Mo_0.9Co_0.1S₂-2) exhibited a higher rate capability of 931 mA h g⁻¹ at the 6 C rate than those of the cathodes using Mo_0.9Co_0.1S₂ (633 mA h g⁻¹) and MoS₂ (338 mA h g⁻¹) catalysts. Moreover, the P-Mo_0.9Co_0.1S₂-2/S cathode provided a capacity decay rate of 0.046% per cycle over 600 cycles at 1 C rate, implying the importance of a dopant in MoS₂ catalysts for polysulfide conversion. Similar to the doping of cobalt and phosphorus, Li and coworkers introduced nickel and phosphorus, which also induced the phase transition of MoS₂ from the 2H phase to 1T phase and facilitated the formation of Ni–P coordination, which improves the rate performance and cycling performance by accelerating the adsorption and conversion of polysulfides, as well as suppressing the shuttle effect.²⁰⁶

In addition to the doping strategy, alloying approaches have also shown great effectiveness in Li–S batteries. Bhoyate and coauthors developed a mixed phase 2D Mo_0.5W_0.5S₂ (2H-1T phases) alloy, fabricated by co-sputtering and sulfurization methods, which was verified to demonstrate strong adsorption to Li₂S₆.²⁰⁷ Compared with MoS₂-CNT-S and WS₂-CNT-S, the Mo_0.5W_0.5S₂-CNT-S cathode achieved low interfacial and charge-transfer resistances, a high specific capacity of 1228 mA h g⁻¹ at 0.1 C, and high cycling stability. Bhoyate and colleagues synthesized a mixed 2H/1T phase Mo_xW_1−xS_2−y alloy (D-MoWS) with tunable defects, which demonstrated a high specific capacity of 1586 mA h g⁻¹ and an excellent areal capacity of 13.5 mA h cm⁻² at high sulfur loading.²⁰⁸ A mechanistic study has proven that the D-MoWS catalyst accelerates the LiPS conversion, and prevents the shuttling effect and Li–metal corrosion.

As for the interlayer engineering approach, Luo and coworkers noted that the interlayer expansion of MoS_2−x nanosheets with S-vacancies enhances the immobilization and catalysis of NaPSs in Na–S batteries, where the S-vacancy increases the electrical conductivity and binding strength by altering the distribution of valence electrons, and the interlayer-expanded structure offers additional active sites that may enhance the adsorption efficiency, leading to strong catalytic conversion towards NaPSs.¹⁹³ More stable interlayer structure can be obtained by intercalation methods, such as carbon, Li⁺, and Na⁺ intercalation. Pan and colleagues fabricated layer-spacing-enlarged MoS₂ (LE-MoS₂) by carbon intercalation.¹⁹⁹ DFT simulations revealed that compared to the unexpanded MoS₂, the intercalated MoS₂ possesses higher adsorption energies for Li₂S₆ and Li₂S₈ (Fig. 17d), stronger electronic coupling between LiPSs and LE-MoS₂, and lower cleavage energies from Li₂S₄ to Li₂S, resulting in stronger chemisorption and better electrocatalytic performance. When used as the electrode during discharge/charge of a Li–S cell, the sulfur-loaded LE-MoS₂ exhibited complete conversion between sulfur and Li₂S (i.e., lithiation of sulfur and reversible oxidation of Li₂S to LiPSs/sulfur) as confirmed by sulfur K-edge XANES (Fig. 17e and f); and no obvious color change from soluble Li₂S₆ was observed after disassembling the cell, indicating its outstanding catalytic effect and adsorption to LiPSs. For these reasons, LE-MoS₂ showed a high initial capacity of 1550 mA h g⁻¹ and a low decay rate (0.06% per cycle) over 500 cycles at the 1 C rate. Liu and coworkers developed a Li⁺ intercalated MoS₂ by in situ electrochemical methods based on the discharged plateaus, where a cut-off voltage below 1.1 V facilitated the Li⁺ insertion process to enlarge the interlayer space and induce phase transition from 2H to 1T phase (Fig. 18a), which was stable and irreversible even undergoing the next charging process.²⁰⁹ According to CV and electrochemical impedance spectroscopy (EIS) analyses, the TiN/1T-MoS₂/S cathode exhibited faster charge-transfer and redox kinetics and thus delivered a better rate performance of 689 mA h g⁻¹ at 2 C (Fig. 18b) and a lower capacity fading rate of 0.051% per cycle at 1C for 800 cycles than those of the TiN/2H-MoS₂/S cathode. Similar to Li⁺ intercalation, Wang and coworkers proposed a Na⁺ intercalation strategy, which was also realized by discharging to a cut-off voltage of 0.8 V in a room-temperature Na–S battery with S/MoS₂ as the starting cathode (Fig. 18c).³⁶ Compared to the S/MoS₂ electrode (cut-off voltage of 1.0–2.8 V), the S/Na_xMoS₂ electrode (cut-off voltage of 0.8–2.8 V) showed better rate capacity (Fig. 18d), less polarization, smaller charge transfer resistance, higher capacity (774.2 mA h g⁻¹) after 800 cycles, and better retention with a fade rate of 0.0055% per cycle for 2800 cycles. The stronger adsorption energies of Na₂S₂ and Na₂S on Na_xMoS₂ contribute to the adsorption of polysulfides and the nucleation of Na₂S, which make the conversion of polysulfides more reversible and thus result in higher cyclability in batteries.

Establishing heterojunctions or hybridization outside the layers is also an effective strategy to suppress the shuttle effect. Li and coworkers reported a Co₉S₈@MoS₂ core–shell heterostructure in which the conductivity of MoS₂ was improved by introducing Co₉S₈ that brings new states near the Fermi level.¹⁹¹ Furthermore, the adsorption of LiPSs and the conversion of LiPSs to Li₂S can be improved due to stronger adsorption and dissociation energies as revealed by theoretical calculations. Compared with the cathodes of Co₉S₈/CNF and MoS₂/CNF, the Co₉S₈@MoS₂ heterostructure-embedded carbon nanofibers (Co₉S₈@MoS₂/CNF) showed a stronger affinity to LiPSs, higher discharge capacity (1106 mA h g⁻¹ after 100 cycles at 0.5 C), and lower polarization in LSBs (Fig. 18e). Moreover, the Co₉S₈@MoS₂/CNF achieved superb cycling stability with a reversible capacity of 794 mA h g⁻¹ at 1 C and a fade rate of 0.091% per cycle over 400 cycles, as well as high capacities at high sulfur loadings. Similar heterostructure catalysts, such as MoS_2−x–Co₉S_8−y with Mo–S–Co heterojunction²¹⁰ and Fe₇S₈–MoS₂,²¹¹ also perform well in the catalytic conversion of LiPSs. Moreover, Wang and colleagues used a MoS₂–MoN heterostructure to accelerate the conversion of polysulfides.²¹² MoS₂ has moderate adsorption energy to polysulfides and a low lithium-ion diffusion barrier, while MoN provides coupling electrons through a redox reaction and has high electrical conductivity, thus MoS₂–MoN provides better adsorption and regulation of long-chain LiPSs, stronger lithium-ion diffusion capability, and faster reaction kinetics. Along the enhanced adsorption–diffusion–conversion process based on MoS₂–MoN/S cathode, 1000 cycles were achieved at 1 C with a decay rate of as low as 0.039% per cycle (see Table 9 for a more comprehensive comparison of defective MoS₂ catalysts).

---

Moreover, the heterogeneous or hybrid structures with metal (e.g., cobalt) or carbon (e.g., carbon spheres, carbon frameworks, and graphene) improve the conductivity of MoS₂ to accelerate electron transport, while the carbon synergistically facilitates the exposure of the catalytic sites, promotes ionic conductivity, and buffers the volume expansion of the electrode during repeated discharging/charging, resulting in higher activity and more stable cycling capability.^{193–195,213,214} Last but not least, non-two-dimensional molybdenum sulfides may contribute to the superb catalytic properties of hybrid catalysts in metal-sulfide batteries. Wang and coworkers constructed a layered C@SnO₂/1T-MoS₂ array electrode, covered by MoS₂ comprising the abundant amorphous state and a weakly crystalline 1T phase, to serve as the sulfur host for Li–S batteries.²¹⁵ The MoS₂ nanosheet and crystalline SnO₂ form a heterostructure, where MoS₂ expedites Li⁺ ion diffusion due to its lower barrier and catalyzes LiPS conversion on its abundant defect sites, while SnO₂ enables the smooth Li⁺ transport and strong adsorption of LiS_x and also suppresses the volume expansion of sulfur electrode with large pores. As a result, a battery with the C@SnO₂/TMS/S cathode delivered a high specific capacity of 1500 mA h g⁻¹ at 0.2 C and exceptional cycling stability with a capacity decay rate of 0.009% per cycle over 4000 cycles at 5 C. These works will inspire more studies on non-two-dimensional MoS₂ catalysts for metal–sulfur batteries.

4.4. Metal–oxygen/air batteries

Beyond the metal–sulfur battery applications, MoS₂ can also be used as catalysts for oxygen or air cathodes in metal–O₂ or metal–air batteries. Li–O₂ batteries (LOBs) display the highest energy density of 3500 W h kg⁻¹ based on the ideal electrochemical reaction of 2Li + O₂ → Li₂O₂,²¹⁶ much higher than those of conventional lithium-ion batteries (100–265 W h kg⁻¹)²¹⁷ and thus LOBs are considered as one of the most promising next-generation energy storage devices. Zn–air batteries (ZABs) are also potential alternative to LIBs because of their high theoretical energy density (1086 W h kg⁻¹), very low cost, and a high degree of safety.²¹⁸ Nevertheless, these metal–O₂/air batteries consisting of a metal anode and an oxygen or air cathode face many challenges such as sluggish conversion kinetics between oxygen and oxides (e.g., O₂²⁻, OH⁻) via OER and ORR reactions. Upgrading MoS₂ catalysts using engineering approaches such as heteroatom doping, hybridization, phase modulation, alloying and vacancy defects is effective for advanced metal–O₂/air batteries.^219–222 In the next section, we will first discuss the challenges in LOBs and solutions with the aid of defective MoS₂ catalysts, where defects related to edge sites, vacancies, phase modulation, amorphous states, hybridization, doping, and alloying structure accelerate the electrode reaction in Li–O₂ batteries and then we will discuss the difficulties of ZABs and strategies by means of defective MoS₂-based heterostructure catalysts.

LOBs face multiple challenges, including large polarization, insufficient capacity, poor round-trip efficiency, and poor cyclability, mainly due to sluggish reactions, parasitic reactions, and poor conductivity.^{216,219,220,223–225} One of the key approaches for boosting battery performance is to enhance the ORR/OER with catalysts during the discharge/charge process (2Li⁺ + O₂ + 2e⁻ → Li₂O₂, E⁰ = 2.96 V), thus alleviating the slow kinetics of Li₂O₂ conversion and suppressing side reactions. MoS₂ is considered one of the promising cathode catalysts for OER/ORR.^37,225 Compared to MoS₂ powder, two-dimensional MoS₂ nanoflakes exhibit better catalytic activity during the discharging and charging of LOBs, with the edge sites responsible for catalytic reactions, while the basal plane favors the deposition of discharge products.²²⁵ Asadi and coworkers fabricated MoS₂ nanoflakes terminated with Mo edges, which had a high density of electrons and exhibited superior catalytic activity for the ORR and the OER.³⁷ The material also showed high round-trip efficiency (∼80%) and up to 50 cycle reversibility when used as a cathode in a Swagelok cell with an ionic liquid (IL, BMIM-BF₄) electrolyte (Fig. 19a). The elevated performance is attributed to the exposed Mo edge active sites binding to O₂, as well as the IL electrolyte partially covering the Mo edge and facilitating the dissociation of O₂⁻ (Fig. 19b) followed by the formation of Li₂O₂via disproportionation of Li₂O (i.e., 2LiO₂ → Li₂O₂ + O₂). The authors further disclosed that O₂ has stronger interactions than Li₂O₂ on the Mo edge, while the Li₂O₂ monomer shows a rather strong interaction with the basal plane by DFT calculations, suggesting the preferable nucleation and growth of Li₂O₂ on the basal plane (Fig. 19c).²²⁶ With the MoS₂ nanoflake cathode, ionic liquid/DMSO electrolyte, and Li₂CO₃-based protective anode, the Swagelok battery experienced a cycle life >500 cycles in an air-like atmosphere (Fig. 19d).

Hybridization of MoS₂ nanosheets with carbon materials (e.g., CNTs, hollow carbon spheres, and graphene aerogel) is an attractive strategy because of the better electrical conductivity, higher mass transport capability, and expanded storage space to accommodate reaction products.^{223,224,227,228} In the hybrid, MoS₂ increases OER and ORR activity, and suppresses side reactions between the carbonaceous material and reactive oxygen species or electrolyte.^216,224,228 Liu and coworkers fabricated sulfur-deficient MoS_2−x nanoflakes on microporous carbon, which delivered a smaller polarization gap (0.59 V), a higher discharge capacity (8851 mA h g⁻¹) at a current density of 500 mA g⁻¹, and better reversibility than MoS₂ (0.87 V; 6312mA h g⁻¹) in Li–O₂ batteries.²²⁰ The enhanced performance is attributable to the introduced S-vacancy that offers abundant active centers and the hierarchical carbon structure that allows free diffusion of ions and gases. Sadighi and coworkers developed a 1T-MoS₂/CNT structure by hybridizing 1T-phase MoS₂ obtained from a Na⁺ intercalation and exfoliation method with functionalized carbon nanotubes.²¹⁹ Thanks to the enhanced adsorption energy of O₂ and Li₂O₂ on the catalytically active basal plane of 1T-phase MoS₂ as revealed by DFT results, as well as the enhanced electrical conductivity and catalytic activity for ORR and OER, the Li–O₂ batteries with 1T-MoS₂/CNT as the cathode, exhibited remarkable performance with a high reversible capacity of 500 mA h g⁻¹ for more than 100 cycles at a current density of 200 mA g⁻¹. Li and colleagues fabricated core–shell MoS_2−x@CNTs composite, where the charge redistribution on MoS₂ nanosheets after the introduction of S-vacancy increases the number of active sites, while the CNT network facilitates the mass transfer and relieves the volume changes during cycling, thus also enabling improved battery performance.²²⁹ Song and coworkers reported a thin amorphous MoS₂ film (≈5 nm) deposited on carbon nanotube forest (GF-CNT/MoS₂) by atomic layer deposition for LOB.²¹⁶ The disordered MoS₂ provides abundant active edge sites for OER/ORR and suppresses parasitic reactions, while the conductive carbon scaffold provides sufficient space for Li₂O₂ and easy access to the electrolyte (Fig. 19e). As a result, these synergistic effects enabled the GF-CNT/MoS₂ electrode to achieve a small overpotential gap (0.58 V) (Fig. 19f) and a high energy efficiency (83%) at 250 mA g⁻¹, a high capacity (4844 mA h g⁻¹) at 500 mA g⁻¹, and a long cycling to 190 cycles.

The hybridization of transition-metal compounds with MoS₂ acts on the OER/ORR by affecting the electronic structure, adsorption energy, and carrier transfer. Li and colleagues synthesized decorated MoS₂/CoS₂ nanorod heterostructures and the coupling between MoS₂ with favorable catalytic activity and CoS₂ with high electrical conductivity accelerated the electron transfer, exposed numerous edge active sites, and increased catalytic kinetics in the ORR/OER cycling.²³⁰ In addition, MoS₂ and CoS₂ show a high level of lattice matching with Li₂O₂, which promotes the heteroepitaxial growth of Li₂O₂. Accordingly, the MoS₂/CoS₂ heterostructure cathode exhibits high ORR and OER current densities, superior discharge capacity (10495 mA h g⁻¹), along with low overpotentials in discharge–charge curves. Wen and colleagues fabricated MoS₂/NiS₂ heterostructures with low-coordination atoms (LCAs) derived from S-vacancy at the hetero-interface.²³¹ The d-band center of MoS₂/NiS₂ (−1.067 eV) is closer to the Fermi level compared to MoS₂ (−1.839 eV) and NiS₂ (−2.437 eV), which is attributed to the strong electronic interaction between LCAs and the adjacent atoms, leading to enhanced adsorption strength to oxygen intermediates and thus elevated catalytic activity for oxygen redox reactions. Combining with other merits, such as improved charge transfer at the interface, the MoS₂/NiS₂-based Li–O₂ batteries exhibited a high discharge capacity of 12377.4 mA h g⁻¹ (at 200 mA g⁻¹) and a cycling lifetime more than 1000 h in the experiment, revealing the great potential of the heterostructure engineering approach for the design of novel electrocatalysts.

Various other engineering strategies, including doping, alloying, and vacancy engineering, have also been proven to be effective. Doped MoS₂ also accelerates the catalytic process and elevates Li–O₂ battery's performance in the absence of carbon hybrids. Cao and coworkers synthesized nitrogen doped N–MoS₂ with S-vacancies by calcining MoS₂ with a nitrogen source.²²¹ As measured by EPR and XPS, the N-to-S substitution resulted in the deficiency of S and the formation of a Mo–N bond, providing more conducive adsorption sites to Li⁺, more exposed active centers, and promoted reaction kinetics of the ORR and OER. As a result, Li–O₂ batteries using the N–MoS₂ electrocatalyst delivered a higher discharge-specific capacity (5205.5 mA h g⁻¹) than that with pristine MoS₂ (2600.3 mA h g⁻¹) at a current density of 100 mA g⁻¹, as well as cycling stability of 224 cycles at a current density of 200 mA g⁻¹. Zhang and coworkers used a metastable MoSSe solid solution to accelerate the OER/ORR process.²²² Poor control of the interlayer vdW forces and different bond lengths of Mo–S and Mo–Se led to high-degree out-of-plane and in-plane lattice distortions in the metastable MoSSe as observed in the HAADF-STEM images (Fig. 20a), from which the calculated strain maps further illustrate the wave-like lattice expansion along the y-direction and slight in-plane lattice distortion along the x-direction of the (002) plane. In addition, the chemical environment is also modulated as shown by XPS, where a decrease in the binding energy of S 2p and an increase in the binding energy of Se 2p were observed. Consequently, Li–O₂ batteries with the MoSSe electrode demonstrated a high discharge capacity of 708 mA h g⁻¹ with a small overpotential gap (0.66 V, Fig. 20b) at a current density of 50 mA g⁻¹, as well as reversibility for 30 cycles. Sun and coworkers reported rich-edge S-vacancy MoS₂ quantum dots (MoS₂ QDs) prepared by Li⁺-clipping of MoS₂ nanosheets using a top-down strategy.²³² Compared to the bulk MoS₂ electrode used for Li–O₂ batteries, the QD electrode (∼3 nm) had a higher discharge capacity (11786 mA h g⁻¹) (Fig. 20c), a smaller overpotential gap (1.04 V), and better cycling stability (230 cycles). DFT calculations revealed that the enhanced performance could be attributed to the edge with S-vacancies, which favored the adsorption and formation of LiO₂ and Li₂O₂ due to the strong adsorption energy and thus accelerated the catalytic ORR and OER processes due to the reduced overpotential. In addition to the above methods, more engineering approaches deserve further exploration, such as hybridization with metals,^233,234 incorporation of reaction intermediates,^225,235 and design of novel electrode structures for achieving higher OER/ORR performance.²²⁶

Zinc–air batteries hold great promise for large-scale energy storage owing to their great cost-effectiveness, excellent intrinsic safety, high-abundance of zinc, and high-energy density (1086 W h kg⁻¹).²¹⁸ The development of bifunctional electrocatalysts to accelerate the sluggish ORR/OER on the air cathode during the discharge/charge process can reduce electrode polarization and improve energy efficiency, increasing capacity and cycling stability. MoS₂-based hetero-interfaces have demonstrated high activity for OER/ORR in ZABs.^32,236 Yan and colleagues crafted a MoS₂@Fe–N–C heterostructure electrocatalyst with atomically dispersed Fe-N₄ coupled to MoS₂ at the interface, where MoS₂ and N could transfer their electrons to the O₂* and OH* intermediates to generate OOH*.²¹⁸ According to the calculated free energy diagrams, MoS₂@Fe–N–C (interface) has the lowest reaction barriers for the ORR and OER than those of MoS₂ and MoS₂@Fe–N–C (surface), indicating that the interface contributes to the enhanced electrocatalytic activity (Fig. 20d and e). The elevated activity of MoS₂@Fe–N–C was also confirmed experimentally by a superior ORR half-wave potential (0.84 V) and OER overpotential (360 mV) at 10 mA cm⁻². Furthermore, the wearable ZAB with MoS₂@Fe–N–C as the air cathode displayed a high specific capacity of 442 mA h g⁻¹_Zn, an improved power density of 78 mW cm⁻², an outstanding cycling stability of 50 cycles at a current density of 5 mA cm⁻², and great rechargeability even with mechanical bending and releasing during the charge/discharge cycle (Fig. 20f). Amiinu and coworkers used a Mo–N/C@MoS₂ electrocatalyst with multifunctional catalytic activity for ZABs, which achieved a high power density (196.4 mW cm⁻²) and voltaic efficiency (≈ 63% at 5 mA cm⁻²), as well as excellent cycling stability of up to 48 h at 25 mA cm⁻².²³⁶ The elevated performance is attributed to the synergy of the exposed edges of MoS₂, highly active Mo–N coupling centers at the interface of N/C and MoS₂, N-induced activated carbon, improved electrical conductivity at the interface, and porous framework with diffusion channels. In addition, when employed as an air electrode, MoS₂ with graphene (e.g., MoS₂/graphene heterolayer),²³⁷ MoS₂ with SnS (e.g., MoS₂–SnS heterostructure),²³⁸ and MoS₂ with Co₉S₈ (e.g., Co₉S₈@MoS₂ core–shell heterostructure)³² also show enhanced ORR/OER activity (see Table 10 for the comparison of catalytic performance of the defective MoS₂ materials), robust cycling stability, high power density, and improved cell voltage, due to the synergistic effect of the composition, interface, and defects. It is also noteworthy that, echoing these success in metal–S and metal–O₂/air batteries, defective MoS₂ is actively broadening its catalytic applications in fuel cells,²³⁹ metal–CO₂^240,241 and other batteries, which also deserves attention.

---

4.5. Hydrodesulfurization reaction

Catalytic desulfurization for petroleum upgrading via a MoS₂ catalyst is an important industrial application that has been commercialized. Sulfur-containing compounds in petroleum, such as mercaptan, thiophene, benzo-thiophene, and dibenzothiophene cause the release of sulfides during combustion. Fuels with a low sulfur content (<10 ppm level in petrochemicals), consequently, are required in many countries to meet the environmental needs,^242,243 driving the development of advanced hydrodesulfurization (HDS) materials. It is worth mentioning that H₂S generated from petroleum desulfurization is transformed into S or S-based chemicals through industrial recovery and conversion plants, eliminating the concern about by-products from HDS. MoS₂ particles and their hybrids, MoS₂-based heterostructures, Co(Ni)-promoted MoS₂, MoS₂ with single-atom anchoring, etc. have emerged as the main catalysts for the HDS processes.^{242,244–248} Among the diverse defect types, here we focus and briefly discuss the defective MoS₂ with S-vacancies because the S-vacancy sites accelerate the adsorption of sulfur-containing molecules and promote the cleavage of C–S bonds in compounds.²⁴³ Much effort revolving around the vacancy formation, catalysis process, and structure maintenance of defective S-vacancies has been made to deepen the understanding of the underlying mechanism.^243,249,250 In the next section, we will first discuss the role of S-vacancies in the catalytic process, then the formation mechanism of the S-vacancies, and finally their structural stability.

For the desulfurization process of petroleum, S-vacancies provide chemisorption sites with a lowered activation energy barrier of the HDS reaction, and S-vacancies also induce the electron-deficient properties of the surrounding Mo atoms to further reinforce the adsorption of S-containing compounds.²⁴⁹ Moreover, the location of S-vacancy sites affects the catalytic reaction pathways and/or the products. Zheng and coworkers pointed out that the HDS reactions follow either the hydrogenation (HYD) route at the S edge or the direct desulfurization (DDS) route at the Mo edge (Fig. 21a).²⁴³ As for the products, the formation of butane bears relatively high reaction barriers, while the production of butene is much more favorable along their respective routes. Salazar and coauthors demonstrated experimentally that the location of S-vacancies (V_s) in the nanoparticle affects the geometric configuration of the generated S-vacancies and also the subsequent adsorption mode of thiophene.²⁴² After the occupation of S-vacancies by thiophene, the atom-resolved scanning tunneling microscopy (STM) images display that thiophene can directly adsorb at the corner (C) site, while its adsorption on adjacent (A) and middle (M) sites is accompanied by S displacement to form the S₂ dimer, and thiophene adsorbs at the double V_S vacancies to reduce the effects of the steric constraints (Fig. 21b and c). Tuxen and colleagues suggested that the size of MoS₂ nanoparticles leads to differences in dibenzothiophene (DBT) adsorption.²⁵¹ Larger Mo-edge-terminated nanoparticles energetically favor the formation of S-vacancies at the edge, while smaller S-edge-terminated nanoparticles favor the formation of S-vacancies at the corner site. During DBT desulfurization, compared with the weak DBT affinity due to steric hindrance at the edge of large-sized MoS₂ nanoparticles, small-sized nanoparticles with S-vacancies at the corner are favorable for DBT adsorption, attributed to the accessible Mo site (Fig. 21d and e).

As the main active sites, the S-vacancies in MoS₂ would be generated directly under the HDS conditions with a high temperature and hydrogen atmosphere and the formation mechanism of vacancies can be deduced from thermodynamic and kinetic perspectives. Shang and coworkers investigated the generation of sulfur vacancies over the Mo edge (−1010) with 50% S coverage and the S edge (10–10) with 100% S coverage and found the formation of S-vacancies on the S edge proceeds spontaneously with a negative reaction Gibbs free energy (ΔG < 0) while it is almost inhibited at the Mo edge (ΔG > 0).²⁴⁹ The kinetic steps for vacancy formation demonstrate that the H₂ molecules prefer homolytic cleavage to form double S–H bonds on two S atoms and the formation of adsorbed H₂S from the dissociated H atoms on the S edge is the rate-controlling process on account of its highest energy barrier. In contrast, Paul and colleague suggested that the extraction of S atoms may occur at the Mo edge, where the activation energy of its rate determining step (<1 eV) is lower than that of the S edge (>1.3 eV).²⁵² The heterolytic dissociation of H₂ molecules to form one S–H and one Mo–H group is supposed to be the rate-determining step at the Mo edge. The intrinsic endothermic reaction of forming S-vacancies at the Mo edge or S edge suggests an unfavorable thermodynamic process, which requires harsh hydrodesulfurization conditions to facilitate the departure of S atoms in the industrial process. In addition, Zheng and coauthors disclosed that H₂ prefers to form S–H and Mo–H species at the Mo edge, whilst to form two S–H species at the S edge after H₂ cleavage and the overall process starting with the scission of H₂ and ending with the departure of H₂S tended to occur at the Mo edge due to its moderate energy barriers.²⁴³ Salazar and coworkers calculated V_S formation energies of three types of S-vacancies on the corner, adjacent corner, and middle positions, respectively, and found their endothermic processes during vacancy generation.²⁴² In conclusion, the mechanism of S-vacancy formation during hydrodesulfurization is still theoretically controversial.

Finally, in terms of structure maintenance, Mom and colleagues investigated the effect of harsh conditions on the catalyst structure by observing the active MoS₂ edge structure under reaction conditions (1 bar, 250 °C).²⁵⁰ Under pure H₂ conditions, the particle edge shrinks but maintains the periodic lattice spacing along the edges to accommodate the ensuing CH₃SH adsorption, while under hydrodesulfurization conditions, the edge structure changes again since a mixture of adsorbed sulfur and CH₃SH occupies the dominant edge structure during the desulfurization of CH₃SH. Furthermore, the phase diagram theoretically depicts the most stable edge structure of Au-supported MoS₂ as a function of Δμ_S and Δμ_H, which are directly related to temperature, H₂S partial pressure, and H₂ partial pressure (Fig. 21f). Thus, the active edge sites adapt their sulfur, hydrogen, and hydrocarbon coverages according to the gas environment, thus adjusting the prevalent structure. Last but not least, it is worth mentioning that, similar to HDS, MoS₂ catalysts can also be used in hydrodeoxygenation (HDO),^253,254 hydrodenitrogenation (HDN),²⁵⁵ CO₂ hydrogenation,²⁵⁶ and so on, and the high catalytic activity undoubtedly relies on the sophisticated design of the defective structure.

5. Conclusions and outlook

As one of the most promising candidates of nonprecious catalysts, molybdenum sulfide has many potential applications. Defects play a vital role in enhancing the catalytic activity of molybdenum sulfide-based materials. This review has summarized ongoing research efforts on defective molybdenum sulfide for electrocatalytic applications, including the use of defective 2D molybdenum disulfide (i.e., 2H, 1T, and 3R phases) and defective non-2D molybdenum sulfide (i.e., amorphous MoS_x, MoS_x clusters) for the HER, NRR, CO₂RR, metal–sulfur battery, and metal–O₂/air battery applications. We review theoretical computational work to unravel the correlation between defects and catalytic performances of molybdenum sulfide electrocatalytic systems from both thermodynamic and kinetic perspectives and to guide the use of defects to tune materials for achieving enhanced adsorption behavior and optimized catalytic performance. We summarize a range of defect types, including vacancies, strained structures, dislocations, distortions, modulated phases, amorphous states, clusters, and exposed edges, as well as various defect generation methods such as plasma treatment, hydrothermal synthesis, heated growth, heteroatom doping, ion intercalation, hybridization, electrochemical deposition, and solid solution routes. Despite these advances, a few challenges remain:

(1) Lack of methods for precise creation of defects

Although much progress has been made in developing defect generation methods, the quest for more precise fabrication techniques, such as precise control of defect types and sites at the atomic level, is still ongoing. Meanwhile, the coupling of defects has been shown to achieve synergistic enhancement effects. Sophisticated designs are still needed to precisely couple different types of defects to achieve breakthroughs in catalytic performance and thus improve energy conversion efficiency. Factors such as structural stability, electrical conductivity, degree of defects, and controllability of the defect preparation method also need to be considered.

(2) Insufficient depth of mechanism study

Theoretical calculations provide an excellent tool for establishing material–structure correlations and therefore a means for high-throughput screening of materials. For the mechanistic analysis of defects, further optimization of computational methods is needed to achieve more accurate predictions and quantification. Meanwhile, the defect structure is often accompanied by the coupling of various defect types, so the study of a single defect after decoupling, as well as a comprehensive understanding of the catalytic effect of coupled defects, is crucial. Moreover, in situ experimental observations are often used for mechanism verification, such as in situ synchrotron radiation characterization of catalytic active sites, and more in situ characterization studies are needed to increase the accuracy and depth of defect mechanism studies.

(3) Limited demonstration of applications

The catalytic applications of defective molybdenum sulfides are not limited to the HER, NRR, CO₂RR, metal–sulfur batteries, and metal–O₂/air batteries; there are many more catalytic applications that can be explored. Mature electrochemical applications are being aggressively pursued in practice, and a broader range of electrochemical applications are emerging at the intersection of various disciplines. The catalytic reduction of NO to ammonia and the conversion of organic molecules via MoS_x catalyst as mentioned above have not yet been systematically studied. In particular, there is a wide scope for innovation in organic electrocatalysis, so catalytic materials based on defective molybdenum disulfide need to be further explored in these emerging fields of electrocatalysis.

(4) Lack of industrial practice

Molybdenum disulfide catalysts for electrochemical hydrogen production have made considerable progress in fundamental research but the development to date is still very far from industrial applications. For practical applications, the following issues need to be addressed, including further improvement of the activity and stability of the catalytic materials; more efforts in simplifying the defect preparation method, controlling costs of materials and preparation processes, and scaling up of the materials; development of advanced membrane electrode assembly techniques for new catalysts applicable to PEM electrolyzers; and more operational parameters of PEM electrolyzers based on molybdenum disulfide materials, which comprise energy efficiency, gas purity, operating voltage, current density, stability, etc. For example, commercial PEM electrolyzers have achieved operational stability for more than 20000 hours, but current stability based on molybdenum sulfide materials is far from this standard. In addition, other potential industrial applications of electrochemistry need to be actively promoted, such as developing electrochemical reduction of carbon dioxide to produce carbon monoxide or liquid products for carbon capture or storage and improving the catalytic properties of electrode materials for metal–sulfur and metal–oxygen batteries to facilitate their commercialization. Moreover, it is also of great significance to integrate sustainable energy sources such as solar, wind, and tidal power and electrocatalytic applications to achieve electrocatalytic conversion and energy storage of renewable green energy.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was supported by Nanyang Technological University under NAP award (M408050000) and Guangdong Basic and Applied Basic Research Foundation (2022A1515110532).

References

1. B. Han and Y. H. Hu, Energy Sci. Eng., 2016, 4, 285–304 CrossRef CAS .
2. Y. Fang, J. Pan, J. He, R. Luo, D. Wang, X. Che, K. Bu, W. Zhao, P. Liu, G. Mu, H. Zhang, T. Lin and F. Huang, Angew. Chem., Int. Ed., 2018, 57, 1232–1235 CrossRef CAS PubMed .
3. A. N. Enyashin, L. Yadgarov, L. Houben, I. Popov, M. Weidenbach, R. Tenne, M. Bar-Sadan and G. Seifert, J. Phys. Chem. C, 2011, 115, 24586–24591 CrossRef CAS .
4. U. Maitra, U. Gupta, M. De, R. Datta, A. Govindaraj and C. Rao, Angew. Chem., Int. Ed., 2013, 52, 13057–13061 CrossRef CAS PubMed .
5. S.-Z. Yang, Y. Gong, P. Manchanda, Y.-Y. Zhang, G. Ye, S. Chen, L. Song, S. T. Pantelides, P. M. Ajayan, M. F. Chisholm and W. Zhou, Adv. Mater., 2018, 30, 1803477 CrossRef PubMed .
6. H. Tan, W. Hu, C. Wang, C. Ma, H. Duan, W. Yan, L. Cai, P. Guo, Z. Sun, Q. Liu, X. Zheng, F. Hu and S. Wei, Small, 2017, 13, 1701389 CrossRef PubMed .
7. J. Peng, Y. Liu, X. Luo, J. Wu, Y. Lin, Y. Guo, J. Zhao, X. Wu, C. Wu and Y. Xie, Adv. Mater., 2019, 31, 1900568 CrossRef PubMed .
8. W. Zhao, J. Pan, Y. Fang, X. Che, D. Wang, K. Bu and F. Huang, Chem. – Eur. J., 2018, 24, 15942–15954 CrossRef CAS PubMed .
9. G. Seifert, H. Terrones, M. Terrones, G. Jungnickel and T. Frauenheim, Phys. Rev. Lett., 2000, 85, 146 CrossRef CAS PubMed .
10. T. Böker, R. Severin, A. Müller, C. Janowitz, R. Manzke, D. Voß, P. Krüger, A. Mazur and J. Pollmann, Phys. Rev. B: Condens. Matter Mater. Phys., 2001, 64, 235305 CrossRef .
11. K. F. Mak, C. Lee, J. Hone, J. Shan and T. F. Heinz, Phys. Rev. Lett., 2010, 105, 136805 CrossRef PubMed .
12. L. Sun, J. Yan, D. Zhan, L. Liu, H. Hu, H. Li, B. K. Tay, J.-L. Kuo, C.-C. Huang, D. W. Hewak, P. S. Lee and Z. X. Shen, Phys. Rev. Lett., 2013, 111, 126801 CrossRef PubMed .
13. T. Li and G. Galli, J. Phys. Chem. C, 2007, 111, 16192–16196 CrossRef CAS .
14. O. V. Yazyev and A. Kis, Mater. Today, 2015, 18, 20–30 CrossRef CAS .
15. A. Splendiani, L. Sun, Y. Zhang, T. Li, J. Kim, C.-Y. Chim, G. Galli and F. Wang, Nano Lett., 2010, 10, 1271–1275 CrossRef CAS PubMed .
16. Z. Yin, H. Li, H. Li, L. Jiang, Y. Shi, Y. Sun, G. Lu, Q. Zhang, X. Chen and H. Zhang, ACS Nano, 2012, 6, 74–80 CrossRef CAS PubMed .
17. H. Li, Z. Yin, Q. He, H. Li, X. Huang, G. Lu, D. W. H. Fam, A. I. Y. Tok, Q. Zhang and H. Zhang, Small, 2012, 8, 63–67 CrossRef CAS PubMed .
18. A. M. Van Der Zande, P. Y. Huang, D. A. Chenet, T. C. Berkelbach, Y. You, G.-H. Lee, T. F. Heinz, D. R. Reichman, D. A. Muller and J. C. Hone, Nat. Mater., 2013, 12, 554–561 CrossRef CAS PubMed .
19. H. Li, C. Tsai, A. L. Koh, L. Cai, A. W. Contryman, A. H. Fragapane, J. Zhao, H. S. Han, H. C. Manoharan, F. Abild-Pedersen, J. K. Nørskov and X. Zheng, Nat. Mater., 2016, 15, 48–53 CrossRef CAS PubMed .
20. H. Li, A. W. Contryman, X. Qian, S. M. Ardakani, Y. Gong, X. Wang, J. M. Weisse, C. H. Lee, J. Zhao, P. M. Ajayan, J. Li, H. C. Manoharan and X. Zheng, Nat. Commun., 2015, 6, 7381 CrossRef CAS PubMed .
21. B. Mohanty, M. Ghorbani-Asl, S. Kretschmer, A. Ghosh, P. Guha, S. K. Panda, B. Jena, A. V. Krasheninnikov and B. K. Jena, ACS Catal., 2018, 8, 1683–1689 CrossRef CAS .
22. C. Tang, L. Zhong, B. Zhang, H. F. Wang and Q. Zhang, Adv. Mater., 2018, 30, 1705110 CrossRef PubMed .
23. M. Asadi, B. Kumar, A. Behranginia, B. A. Rosen, A. Baskin, N. Repnin, D. Pisasale, P. Phillips, W. Zhu, R. Haasch, R. F. Klie, P. Král, J. Abiade and A. Salehi-Khojin, Nat. Commun., 2014, 5, 4470 CrossRef CAS PubMed .
24. L. Zhang, X. Ji, X. Ren, Y. Ma, X. Shi, Z. Tian, A. M. Asiri, L. Chen, B. Tang and X. Sun, Adv. Mater., 2018, 30, 1800191 CrossRef PubMed .
25. L. Zhang, J. Liang, Y. Wang, T. Mou, Y. Lin, L. Yue, T. Li, Q. Liu, Y. Luo, N. Li, B. Tang, Y. Liu, S. Gao, A. A. Alshehri, X. Guo, D. Ma and X. Sun, Angew. Chem., Int. Ed., 2021, 60, 25263–25268 CrossRef CAS PubMed .
26. L. Maachou, K. Qi, E. Petit, Z. Qin, Y. Zhang, D. Cot, V. Flaud, C. Reibel, H. El-Maghrbi, L. Li, P. Miele, D. Kaplan, M. Chhowalla, N. Onofrio and D. Voiry, J. Mater. Chem. A, 2020, 8, 25053–25060 RSC .
27. D. He, H. Ooka, Y. Kim, Y. Li, F. Jin, S. H. Kim and R. Nakamura, Proc. Natl. Acad. Sci. U. S. A., 2020, 117, 31631–31638 CrossRef CAS PubMed .
28. Y. Zang, S. Niu, Y. Wu, X. Zheng, J. Cai, J. Ye, Y. Xie, Y. Liu, J. Zhou, J. Zhu, X. Liu, G. Wang and Y. Qian, Nat. Commun., 2019, 10, 1217 CrossRef PubMed .
29. K. Jiang, M. Luo, Z. Liu, M. Peng, D. Chen, Y.-R. Lu, T.-S. Chan, F. M. de Groot and Y. Tan, Nat. Commun., 2021, 12, 1687 CrossRef CAS PubMed .
30. X. Huang, H. Xu, D. Cao and D. Cheng, Nano Energy, 2020, 78, 105253 CrossRef CAS .
31. T. Zhang, Y. Liu, J. Yu, Q. Ye, L. Yang, Y. Li and H. J. Fan, Adv. Mater., 2022, 34, 2202195 CrossRef CAS PubMed .
32. J. Bai, T. Meng, D. Guo, S. Wang, B. Mao and M. Cao, ACS Appl. Mater. Interfaces, 2018, 10, 1678–1689 CrossRef CAS PubMed .
33. B. H. R. Suryanto, D. Wang, L. M. Azofra, M. Harb, L. Cavallo, R. Jalili, D. R. G. Mitchell, M. Chatti and D. R. MacFarlane, ACS Energy Lett., 2018, 4, 430–435 CrossRef .
34. Y. Liu, M. Han, Q. Xiong, S. Zhang, C. Zhao, W. Gong, G. Wang, H. Zhang and H. Zhao, Adv. Energy Mater., 2019, 9, 1803935 CrossRef .
35. R. Fang, S. Zhao, Z. Sun, D.-W. Wang, H.-M. Cheng and F. Li, Adv. Mater., 2017, 29, 1606823 CrossRef PubMed .
36. Y. Wang, Y. Lai, J. Chu, Z. Yan, Y. X. Wang, S. L. Chou, H. K. Liu, S. X. Dou, X. Ai, H. Yang and Y. Cao, Adv. Mater., 2021, 33, 2100229 CrossRef CAS PubMed .
37. M. Asadi, B. Kumar, C. Liu, P. Phillips, P. Yasaei, A. Behranginia, P. Zapol, R. F. Klie, L. A. Curtiss and A. Salehi-Khojin, ACS Nano, 2016, 10, 2167–2175 CrossRef CAS PubMed .
38. S. E. Skrabalak and K. S. Suslick, J. Am. Chem. Soc., 2005, 127, 9990–9991 CrossRef CAS PubMed .
39. Y. Yin, J. Han, Y. Zhang, X. Zhang, P. Xu, Q. Yuan, L. Samad, X. Wang, Y. Wang, Z. Zhang, P. Zhang, X. Cao, B. Song and S. Jin, J. Am. Chem. Soc., 2016, 138, 7965–7972 CrossRef CAS PubMed .
40. W. Wu, C. Niu, C. Wei, Y. Jia, C. Li and Q. Xu, Angew. Chem., Int. Ed., 2019, 58, 2029–2033 CrossRef CAS PubMed .
41. B. Shang, X. Cui, L. Jiao, K. Qi, Y. Wang, J. Fan, Y. Yue, H. Wang, Q. Bao, X. Fan, S. Wei, W. Song, Z. Cheng, S. Guo and W. Zheng, Nano Lett., 2019, 19, 2758–2764 CrossRef CAS PubMed .
42. L. Lei, Y. Lun, F. Cao, L. Meng, S. Xing, J. Guo, H. Dong, S. Gu, K. Xu, S. Hussain, Y. Li, Y. Sugawara, F. Pang, W. Ji, J. Hong, R. Xu and Z. Cheng, Nanotechnology, 2021, 32, 465703 CrossRef CAS PubMed .
43. Y. Zhang, T. Yang, J. Li, Q. Zhang, B. Li and M. Gao, Adv. Funct. Mater., 2023, 33, 2210939 CrossRef CAS .
44. T. Sun, J. Wang, X. Chi, Y. Lin, Z. Chen, X. Ling, C. Qiu, Y. Xu, L. Song, W. Chen and C. Su, ACS Catal., 2018, 8, 7585–7592 CrossRef CAS .
45. Q. Ke, X. Zhang, W. Zang, A. M. Elshahawy, Y. Hu, Q. He, S. J. Pennycook, Y. Cai and J. Wang, Small, 2019, 15, 1900131 CrossRef PubMed .
46. T. Ren, W. Tian, Q. Shen, Z. Yuan, D. Chen, N. Li and J. Lu, Nano Energy, 2021, 90, 106527 CrossRef CAS .
47. S. Deng, M. Luo, C. Ai, Y. Zhang, B. Liu, L. Huang, Z. Jiang, Q. Zhang, L. Gu, S. Lin, X. Wang, L. Yu, J. Wen, J. Wang, G. Pan, X. Xia and J. Tu, Angew. Chem., Int. Ed., 2019, 58, 16289–16296 CrossRef CAS PubMed .
48. N. Feng, R. Meng, L. Zu, Y. Feng, C. Peng, J. Huang, G. Liu, B. Chen and J. Yang, Nat. Commun., 2019, 10, 1372 CrossRef PubMed .
49. Z. Li, H. Li, Z. Yang, X. Lu, S. Ji, M. Zhang, J. H. Horton, H. Ding, Q. Xu, J. Zhu and J. Yu, Small, 2022, 18, 2201092 CrossRef CAS PubMed .
50. Z. Li, X. Yan, D. He, W. Hu, S. Younan, Z. Ke, M. Patrick, X. Xiao, J. Huang, H. Wu, X. Pan and J. Gu, ACS Catal., 2022, 12, 7687–7695 CrossRef CAS .
51. X. Wang, L. Pan, J. Yang, B. Li, Y. Y. Liu and Z. Wei, Adv. Mater., 2021, 33, 2006908 CrossRef CAS PubMed .
52. Y. Zhang, H. Tao, S. Du and X. Yang, ACS Appl. Mater. Interfaces, 2019, 11, 11327–11337 CrossRef CAS PubMed .
53. D. Escalera-López, Z. Lou and N. V. Rees, Adv. Energy Mater., 2019, 9, 1802614 CrossRef .
54. D. Escalera-López, Y. Niu, J. Yin, K. Cooke, N. V. Rees and R. E. Palmer, ACS Catal., 2016, 6, 6008–6017 CrossRef PubMed .
55. J. Kibsgaard, T. F. Jaramillo and F. Besenbacher, Nat. Chem., 2014, 6, 248–253 CrossRef CAS PubMed .
56. R. Kothari, D. Buddhi and R. Sawhney, Renewable Sustainable Energy Rev., 2008, 12, 553–563 CrossRef CAS .
57. M. Carmo, D. L. Fritz, J. Mergel and D. Stolten, Int. J. Hydrogen Energy, 2013, 38, 4901–4934 CrossRef CAS .
58. K. E. Ayers, J. N. Renner, N. Danilovic, J. X. Wang, Y. Zhang, R. Maric and H. Yu, Catal. Today, 2016, 262, 121–132 CrossRef CAS .
59. B. Hinnemann, P. G. Moses, J. Bonde, K. P. Jørgensen, J. H. Nielsen, S. Horch, I. Chorkendorff and J. K. Nørskov, J. Am. Chem. Soc., 2005, 127, 5308–5309 CrossRef CAS PubMed .
60. T. F. Jaramillo, K. P. Jørgensen, J. Bonde, J. H. Nielsen, S. Horch and I. Chorkendorff, Science, 2007, 317, 100–102 CrossRef CAS PubMed .
61. M. A. Lukowski, A. S. Daniel, F. Meng, A. Forticaux, L. Li and S. Jin, J. Am. Chem. Soc., 2013, 135, 10274–10277 CrossRef CAS PubMed .
62. D. Voiry, R. Fullon, J. Yang, C. de Carvalho Castro e Silva, R. Kappera, I. Bozkurt, D. Kaplan, M. J. Lagos, P. E. Batson, G. Gupta, A. D. Mohite, L. Dong, D. Er, V. B. Shenoy, T. Asefa and M. Chhowalla, Nat. Mater., 2016, 15, 1003–1009 CrossRef CAS PubMed .
63. Y. Li, H. Wang, L. Xie, Y. Liang, G. Hong and H. Dai, J. Am. Chem. Soc., 2011, 133, 7296–7299 CrossRef CAS PubMed .
64. G. Li, D. Zhang, Q. Qiao, Y. Yu, D. Peterson, A. Zafar, R. Kumar, S. Curtarolo, F. Hunte, S. Shannon, Y. Zhu, W. Yang and L. Cao, J. Am. Chem. Soc., 2016, 138, 16632–16638 CrossRef CAS PubMed .
65. C. Tsai, H. Li, S. Park, J. Park, H. S. Han, J. K. Nørskov, X. Zheng and F. Abild-Pedersen, Nat. Commun., 2017, 8, 15113 CrossRef PubMed .
66. L. Li, Z. Qin, L. Ries, S. Hong, T. Michel, J. Yang, C. Salameh, M. Bechelany, P. Miele, D. Kaplan, M. Chhowalla and D. Voiry, ACS Nano, 2019, 13, 6824–6834 CrossRef CAS PubMed .
67. H. Li, M. Du, M. J. Mleczko, A. L. Koh, Y. Nishi, E. Pop, A. J. Bard and X. Zheng, J. Am. Chem. Soc., 2016, 138, 5123–5129 CrossRef CAS PubMed .
68. C.-C. Cheng, A.-Y. Lu, C.-C. Tseng, X. Yang, M. N. Hedhili, M.-C. Chen, K.-H. Wei and L.-J. Li, Nano Energy, 2016, 30, 846–852 CrossRef CAS .
69. C. Meng, M.-C. Lin, X.-W. Du and Y. Zhou, ACS Sustainable Chem. Eng., 2019, 7, 6999–7003 CrossRef CAS .
70. G. Ye, Y. Gong, J. Lin, B. Li, Y. He, S. T. Pantelides, W. Zhou, R. Vajtai and P. M. Ajayan, Nano Lett., 2016, 16, 1097–1103 CrossRef CAS PubMed .
71. J. Yang, Y. Wang, M. J. Lagos, V. Manichev, R. Fullon, X. Song, D. Voiry, S. Chakraborty, W. Zhang, P. E. Batson, L. Feldman, T. Gustafsson and M. Chhowalla, ACS Nano, 2019, 13, 9958–9964 CrossRef CAS PubMed .
72. C. Wang, H. Lu, K. Tang, Z. Mao, Q. Li, X. Wang and C. Yan, Electrochim. Acta, 2020, 336, 135740 CrossRef CAS .
73. S. Geng, W. Yang, Y. Liu and Y. Yu, J. Catal., 2020, 391, 91–97 CrossRef CAS .
74. X. Wang, Y. Zhang, H. Si, Q. Zhang, J. Wu, L. Gao, X. Wei, Y. Sun, Q. Liao, Z. Zhang, K. Ammarah, L. Gu, Z. Kang and Y. Zhang, J. Am. Chem. Soc., 2020, 142, 4298–4308 CrossRef CAS PubMed .
75. P. Zhang, H. Xiang, L. Tao, H. Dong, Y. Zhou, T. S. Hu, X. Chen, S. Liu, S. Wang and S. Garaj, Nano Energy, 2019, 57, 535–541 CrossRef CAS .
76. J. Wang, W. Fang, Y. Hu, Y. Zhang, J. Dang, Y. Wu, B. Chen, H. Zhao and Z. Li, Appl. Catal., B, 2021, 298, 120490 CrossRef CAS .
77. Z. Luo, Y. Ouyang, H. Zhang, M. Xiao, J. Ge, Z. Jiang, J. Wang, D. Tang, X. Cao, C. Liu and W. Xing, Nat. Commun., 2018, 9, 2120 CrossRef PubMed .
78. J. Xu, G. Shao, X. Tang, F. Lv, H. Xiang, C. Jing, S. Liu, S. Dai, Y. Li, J. Luo and Z. Zhou, Nat. Commun., 2022, 13, 2193 CrossRef CAS PubMed .
79. S. Park, J. Park, H. Abroshan, L. Zhang, J. K. Kim, J. Zhang, J. Guo, S. Siahrostami and X. Zheng, ACS Energy Lett., 2018, 3, 2685–2693 CrossRef CAS .
80. Y. Zhao, M. T. Tang, S. Wu, J. Geng, Z. Han, K. Chan, P. Gao and H. Li, J. Catal., 2020, 382, 320–328 CrossRef CAS .
81. Y. Zhao, J. Hwang, M. T. Tang, H. Chun, X. Wang, H. Zhao, K. Chan, B. Han, P. Gao and H. Li, J. Power Sources, 2020, 456, 227998 CrossRef CAS .
82. Y. Kang, Surf. Sci., 2021, 704, 121759 CrossRef CAS .
83. Y. Zhang, Y. Kuwahara, K. Mori and H. Yamashita, Chem. – Eur. J., 2019, 14, 278–285 CAS .
84. J. Pető, T. Ollár, P. Vancsó, Z. I. Popov, G. Z. Magda, G. Dobrik, C. Hwang, P. B. Sorokin and L. Tapasztó, Nat. Chem., 2018, 10, 1246–1251 CrossRef PubMed .
85. Y. Zhao, W. Xin, B. Liu, H. Li, Y. Xu and Z. Zhang, Inorg. Chem. Front., 2022, 9, 3461–3469 RSC .
86. A. M. Z. Tan, C. Freysoldt and R. G. Hennig, Phys. Rev. Mater., 2020, 4, 064004 CrossRef CAS .
87. J. Ekspong, R. Sandström, L. P. Rajukumar, M. Terrones, T. Wågberg and E. Gracia-Espino, Adv. Funct. Mater., 2018, 28, 1802744 CrossRef .
88. X. Geng, W. Sun, W. Wu, B. Chen, A. Al-Hilo, M. Benamara, H. Zhu, F. Watanabe, J. Cui and T.-P. Chen, Nat. Commun., 2016, 7, 10672 CrossRef CAS PubMed .
89. W. Chen, J. Gu, Q. Liu, R. Luo, L. Yao, B. Sun, W. Zhang, H. Su, B. Chen, P. Liu and D. Zhang, ACS Nano, 2018, 12, 308–316 CrossRef CAS PubMed .
90. C. Tan, Z. Luo, A. Chaturvedi, Y. Cai, Y. Du, Y. Gong, Y. Huang, Z. Lai, X. Zhang, L. Zheng, X. Qi, M. H. Goh, J. Wang, S. Han, X.-J. Wu, L. Gu, C. Kloc and H. Zhang, Adv. Mater., 2018, 30, 1705509 CrossRef PubMed .
91. J. Zhu, Z.-C. Wang, H. Dai, Q. Wang, R. Yang, H. Yu, M. Liao, J. Zhang, W. Chen, Z. Wei, N. Li, L. Du, D. Shi, W. Wang, L. Zhang, Y. Jiang and G. Zhang, Nat. Commun., 2019, 10, 1348 CrossRef PubMed .
92. Z. Liu, K. Nie, X. Qu, X. Li, B. Li, Y. Yuan, S. Chong, P. Liu, Y. Li, Z. Yin and W. Huang, J. Am. Chem. Soc., 2022, 144, 4863–4873 CrossRef CAS PubMed .
93. L. Ji, P. Yan, C. Zhu, C. Ma, W. Wu, C. Wei, Y. Shen, S. Chu, J. Wang, Y. Du, J. Chen, X. Yang and Q. Xu, Appl. Catal., B, 2019, 251, 87–93 CrossRef CAS .
94. L. Liu, J. Wu, L. Wu, M. Ye, X. Liu, Q. Wang, S. Hou, P. Lu, L. Sun, J. Zheng, L. Xing, L. Gu, X. Jiang, L. Xie and L. Jiao, Nat. Mater., 2018, 17, 1108–1114 CrossRef CAS PubMed .
95. X. Qian, J. Liu, L. Fu and J. Li, Science, 2014, 346, 1344–1347 CrossRef CAS PubMed .
96. G. Gao, Y. Jiao, F. Ma, Y. Jiao, E. Waclawik and A. Du, J. Phys. Chem. C, 2015, 119, 13124–13128 CrossRef CAS .
97. Y. Yu, G.-H. Nam, Q. He, X.-J. Wu, K. Zhang, Z. Yang, J. Chen, Q. Ma, M. Zhao, Z. Liu, F.-R. Ran, X. Wang, H. Li, X. Huang, B. Li, Q. Xiong, Q. Zhang, Z. Liu, L. Gu, Y. Du, W. Huang and H. Zhang, Nat. Chem., 2018, 10, 638–643 CrossRef CAS PubMed .
98. G. H. Nam, Q. He, X. Wang, Y. Yu, J. Chen, K. Zhang, Z. Yang, D. Hu, Z. Lai, B. Li, Q. Xiong, Q. Zhang, L. Gu and H. Zhang, Adv. Mater., 2019, 31, 1807764 CrossRef PubMed .
99. N. H. Attanayake, A. C. Thenuwara, A. Patra, Y. V. Aulin, T. M. Tran, H. Chakraborty, E. Borguet, M. L. Klein, J. P. Perdew and D. R. Strongin, ACS Energy Lett., 2017, 3, 7–13 CrossRef .
100. T. H. Lau, S. Wu, R. Kato, T.-S. Wu, J. Kulhavy, J. Mo, J. Zheng, J. S. Foord, Y.-L. Soo, K. Suenaga, M. T. Darby and S. C. E. Tsang, ACS Catal., 2019, 9, 7527–7534 CrossRef CAS .
101. Q. Tang and D.-e Jiang, ACS Catal., 2016, 6, 4953–4961 CrossRef CAS .
102. Q. Jin, N. Liu, C. Dai, R. Xu, B. Wu, G. Yu, B. Chen and Y. Du, Adv. Energy Mater., 2020, 10, 2000291 CrossRef CAS .
103. K. Qi, X. Cui, L. Gu, S. Yu, X. Fan, M. Luo, S. Xu, N. Li, L. Zheng, Q. Zhang, J. Ma, Y. Gong, F. Lv, K. Wang, H. Huang, W. Zhang, S. Guo, W. Zheng and P. Liu, Nat. Commun., 2019, 10, 5231 CrossRef PubMed .
104. B. Pattengale, Y. Huang, X. Yan, S. Yang, S. Younan, W. Hu, Z. Li, S. Lee, X. Pan, J. Gu and J. Huang, Nat. Commun., 2020, 11, 4114 CrossRef CAS PubMed .
105. J. Shi, P. Yu, F. Liu, P. He, R. Wang, L. Qin, J. Zhou, X. Li, J. Zhou, X. Sui, S. Zhang, Y. Zhang, Q. Zhang, T. C. Sum, X. Qiu, Z. Liu and X. Liu, Adv. Mater., 2017, 29, 1701486 CrossRef PubMed .
106. R. Suzuki, M. Sakano, Y. Zhang, R. Akashi, D. Morikawa, A. Harasawa, K. Yaji, K. Kuroda, K. Miyamoto, T. Okuda, K. Ishizaka, R. Arita and Y. Iwasa, Nat. Nanotechnol., 2014, 9, 611–617 CrossRef CAS PubMed .
107. R. J. Toh, Z. Sofer, J. Luxa, D. Sedmidubský and M. Pumera, Chem. Commun., 2017, 53, 3054–3057 RSC .
108. S. Zhou, S. Wang, H. Li, W. Xu, C. Gong, J. C. Grossman and J. H. Warner, ACS Omega, 2017, 2, 3315–3324 CrossRef CAS PubMed .
109. J. Hu, C. Zhang, L. Jiang, H. Lin, Y. An, D. Zhou, M. K. Leung and S. Yang, Joule, 2017, 1, 383–393 CrossRef CAS .
110. Y. Ma, D. Leng, X. Zhang, J. Fu, C. Pi, Y. Zheng, B. Gao, X. Li, N. Li, P. K. Chu, Y. Luo and K. Huo, Small, 2022, 18, 2203173 CrossRef CAS PubMed .
111. M. Liu, J.-A. Wang, W. Klysubun, G.-G. Wang, S. Sattayaporn, F. Li, Y.-W. Cai, F. Zhang, J. Yu and Y. Yang, Nat. Commun., 2021, 12, 5260 CrossRef CAS PubMed .
112. G. Wang, G. Zhang, X. Ke, X. Chen, X. Chen, Y. Wang, G. Huang, J. Dong, S. Chu and M. Sui, Small, 2022, 18, 2107238 CrossRef CAS PubMed .
113. J. Liu, J. Wang, B. Zhang, Y. Ruan, H. Wan, X. Ji, K. Xu, D. Zha, L. Miao and J. Jiang, J. Mater. Chem. A, 2018, 6, 2067–2072 RSC .
114. T. Wu, E. Song, S. Zhang, M. Luo, C. Zhao, W. Zhao, J. Liu and F. Huang, Adv. Mater., 2022, 34, 2108505 CrossRef CAS PubMed .
115. C. Gao, Z. Li, H. Wang, Y. Yang, B. Li, Z. Peng, J. Li and Z. Liu, ChemElectroChem, 2020, 7, 355–361 CrossRef CAS .
116. S. Huang, Y. Cao, F. Yao, D. Zhang, J. Yang, S. Ye, D. Yao, Y. Liu, J. Li, D. Lei, X. Wang, H. Huang and M. Wu, Small, 2023, 2207919 CrossRef CAS PubMed .
117. C. Lang, W. Jiang, C. J. Yang, H. Zhong, P. Chen, Q. Wu, X. Yan, C. L. Dong, Y. Lin, L. Ouyang, Y. Jia and X. Yao, Small, 2023, 2300807 CrossRef PubMed .
118. F. Xi, P. Bogdanoff, K. Harbauer, P. Plate, C. Höhn, J. R. Rappich, B. Wang, X. Han, R. Van De Krol and S. Fiechter, ACS Catal., 2019, 9, 2368–2380 CrossRef CAS .
119. L. Wu, A. Longo, N. Y. Dzade, A. Sharma, M. M. Hendrix, A. A. Bol, N. H. De Leeuw, E. J. Hensen and J. P. Hofmann, ChemSusChem, 2019, 12, 4383 CrossRef CAS PubMed .
120. S. C. Lee, J. D. Benck, C. Tsai, J. Park, A. L. Koh, F. Abild-Pedersen, T. F. Jaramillo and R. Sinclair, ACS Nano, 2016, 10, 624–632 CrossRef CAS PubMed .
121. B. Lassalle-Kaiser, D. Merki, H. Vrubel, S. Gul, V. K. Yachandra, X. Hu and J. Yano, J. Am. Chem. Soc., 2015, 137, 314–321 CrossRef CAS PubMed .
122. P. D. Tran, T. V. Tran, M. Orio, S. Torelli, Q. D. Truong, K. Nayuki, Y. Sasaki, S. Y. Chiam, R. Yi, I. Honma, J. Barber and V. Artero, Nat. Mater., 2016, 15, 640–646 CrossRef CAS PubMed .
123. Y. Deng, L. R. L. Ting, P. H. L. Neo, Y.-J. Zhang, A. A. Peterson and B. S. Yeo, ACS Catal., 2016, 6, 7790–7798 CrossRef CAS .
124. L. R. L. Ting, Y. Deng, L. Ma, Y.-J. Zhang, A. A. Peterson and B. S. Yeo, ACS Catal., 2016, 6, 861–867 CrossRef CAS .
125. B. Li, L. Jiang, X. Li, Z. Cheng, P. Ran, P. Zuo, L. Qu, J. Zhang and Y. Lu, Adv. Funct. Mater., 2019, 29, 1806229 CrossRef .
126. Y. Li, Y. Yu, Y. Huang, R. A. Nielsen, W. A. Goddard III, Y. Li and L. Cao, ACS Catal., 2015, 5, 448–455 CrossRef CAS .
127. R. Ding, M. Wang, X. Wang, H. Wang, L. Wang, Y. Mu and B. Lv, Nanoscale, 2019, 11, 11217–11226 RSC .
128. D. Zhang, F. Wang, W. Zhao, M. Cui, X. Fan, R. Liang, Q. Ou and S. Zhang, Adv. Sci., 2022, 9, 2202445 CrossRef CAS PubMed .
129. H. Vrubel and X. Hu, ACS Catal., 2013, 3, 2002–2011 CrossRef CAS .
130. T. R. Hellstern, J. Kibsgaard, C. Tsai, D. W. Palm, L. A. King, F. Abild-Pedersen and T. F. Jaramillo, ACS Catal., 2017, 7, 7126–7130 CrossRef CAS .
131. P. K. Holzapfel, M. Bühler, D. Escalera-López, M. Bierling, F. D. Speck, K. J. Mayrhofer, S. Cherevko, C. V. Pham and S. Thiele, Small, 2020, 16, 2003161 CrossRef CAS PubMed .
132. B. Seo, G. Y. Jung, S. J. Lee, D. S. Baek, Y. J. Sa, H. W. Ban, J. S. Son, K. Park, S. K. Kwak and S. H. Joo, ACS Catal., 2019, 10, 652–662 CrossRef .
133. T. F. Jaramillo, J. Bonde, J. Zhang, B.-L. Ooi, K. Andersson, J. Ulstrup and I. Chorkendorff, J. Phys. Chem. C, 2008, 112, 17492–17498 CrossRef CAS .
134. Z. Huang, W. Luo, L. Ma, M. Yu, X. Ren, M. He, S. Polen, K. Click, B. Garrett, J. Lu, K. Amine, C. Hadad, W. Chen, A. Asthagiri and Y. Wu, Angew. Chem., Int. Ed., 2015, 54, 15181–15185 CrossRef CAS PubMed .
135. C.-H. Lee, S. Lee, Y.-K. Lee, Y. C. Jung, Y.-I. Ko, D. C. Lee and H.-I. Joh, ACS Catal., 2018, 8, 5221–5227 CrossRef CAS .
136. J. McAllister, N. A. Bandeira, J. C. McGlynn, A. Y. Ganin, Y.-F. Song, C. Bo and H. N. Miras, Nat. Commun., 2019, 10, 370 CrossRef CAS PubMed .
137. Z. Ji, C. Trickett, X. Pei and O. M. Yaghi, J. Am. Chem. Soc., 2018, 140, 13618–13622 CrossRef CAS PubMed .
138. B. R. Garrett, S. M. Polen, M. Pimplikar, C. M. Hadad and Y. Wu, J. Am. Chem. Soc., 2017, 139, 4342–4345 CrossRef CAS PubMed .
139. B. R. Garrett, K. A. Click, C. B. Durr, C. M. Hadad and Y. Wu, J. Am. Chem. Soc., 2016, 138, 13726–13731 CrossRef CAS PubMed .
140. Y. He, Q. He, L. Wang, C. Zhu, P. Golani, A. D. Handoko, X. Yu, C. Gao, M. Ding, X. Wang, F. Liu, Q. Zeng, P. Yu, S. Guo, B. I. Yakobson, L. Wang, Z. W. Seh, Z. Zhang, M. Wu, Q. J. Wang, H. Zhang and Z. Liu, Nat. Mater., 2019, 18, 1098–1104 CrossRef CAS PubMed .
141. J. Zhang, X. Tian, M. Liu, H. Guo, J. Zhou, Q. Fang, Z. Liu, Q. Wu and J. Lou, J. Am. Chem. Soc., 2019, 141, 19269–19275 CrossRef CAS PubMed .
142. X. Li, T. Li, Y. Ma, Q. Wei, W. Qiu, H. Guo, X. Shi, P. Zhang, A. M. Asiri, L. Chen, B. Tang and X. Sun, Adv. Energy Mater., 2018, 8, 1801357 CrossRef .
143. J. Chen, C. Zhang, M. Huang, J. Zhang, J. Zhang, H. Liu, G. Wang and R. Wang, Appl. Catal., B, 2021, 285, 119810 CrossRef CAS .
144. S. B. Patil, H.-L. Chou, Y.-M. Chen, S.-H. Hsieh, C.-H. Chen, C.-C. Chang, S.-R. Li, Y.-C. Lee, Y.-S. Lin, H. Li, Y. J. Chang, Y.-H. Lai and D.-Y. Wang, J. Mater. Chem. A, 2021, 9, 1230–1239 RSC .
145. W. Wu, C. Niu, P. Yan, F. Shi, C. Ma, X. Yang, Y. Jia, J. Chen, M. I. Ahmed, C. Zhao and Q. Xu, Appl. Catal., B, 2021, 298, 120615 CrossRef CAS .
146. R. Liu, T. Guo, H. Fei, Z. Wu, D. Wang and F. Liu, Adv. Sci., 2022, 9, 2103583 CrossRef CAS PubMed .
147. G. Lin, Q. Ju, X. Guo, W. Zhao, S. Adimi, J. Ye, Q. Bi, J. Wang, M. Yang and F. Huang, Adv. Mater., 2021, 33, 2007509 CrossRef CAS PubMed .
148. J. Liang, S. Ma, J. Li, Y. Wang, J. Wu, Q. Zhang, Z. Liu, Z. Yang, K. Qu and W. Cai, J. Mater. Chem. A, 2020, 8, 10426–10432 RSC .
149. J. Zhao, J. Zhao and Q. Cai, Phys. Chem. Chem. Phys., 2018, 20, 9248–9255 RSC .
150. T. Yang, T. T. Song, J. Zhou, S. Wang, D. Chi, L. Shen, M. Yang and Y. P. Feng, Nano Energy, 2020, 68, 104304 CrossRef CAS .
151. Q. Dang, S. Tang, T. Liu, X. Li, X. Wang, W. Zhong, Y. Luo and J. Jiang, J. Phys. Chem. Lett., 2021, 12, 8355–8362 CrossRef CAS PubMed .
152. L. Chen, Y. Zhou, Y. Rao, M. Qin, B. Gong, W. Zhang and Z. Li, J. Phys. Chem. C, 2022, 126, 9352–9360 CrossRef CAS .
153. K. Xie, F. Wang, F. Wei, J. Zhao and S. Lin, J. Phys. Chem. C, 2022, 126, 5180–5188 CrossRef CAS .
154. J. Li, S. Chen, F. Quan, G. Zhan, F. Jia, Z. Ai and L. Zhang, Chem, 2020, 6, 885–901 CAS .
155. H. Su, L. Chen, Y. Chen, R. Si, Y. Wu, X. Wu, Z. Geng, W. Zhang and J. Zeng, Angew. Chem., Int. Ed., 2020, 59, 20411–20416 CrossRef CAS PubMed .
156. Z. He, Y. Jiang, X. Cui, Z. Liu, X. Meng, J. Wan and F. Ma, ACS Appl. Nano Mater., 2022, 5, 5470–5478 CrossRef CAS .
157. H. Fei, T. Guo, Y. Xin, L. Wang, R. Liu, D. Wang, F. Liu and Z. Wu, Appl. Catal., B, 2022, 300, 120733 CrossRef CAS .
158. X. Chen, C. Ma, Z. Tan, X. Wang, X. Qian, X. Zhang, J. Tian, S. Yan and M. Shao, Chem. Eng. J., 2022, 433, 134504 CrossRef CAS .
159. X. Chen, S. Zhang, X. Qian, Z. Liang, Y. Xue, X. Zhang, J. Tian, Y. Han and M. Shao, Appl. Catal., B, 2022, 310, 121277 CrossRef CAS .
160. X. Xu, X. Tian, B. Sun, Z. Liang, H. Cui, J. Tian and M. Shao, Appl. Catal., B, 2020, 272, 118984 CrossRef CAS .
161. X. Zi, J. Wan, X. Yang, W. Tian, H. Zhang and Y. Wang, Appl. Catal., B, 2021, 286, 119870 CrossRef CAS .
162. B. H. Suryanto, D. Wang, L. M. Azofra, M. Harb, L. Cavallo, R. Jalili, D. R. Mitchell, M. Chatti and D. R. MacFarlane, ACS Energy Lett., 2018, 4, 430–435 CrossRef .
163. W. Zhang, Y. Hu, L. Ma, G. Zhu, Y. Wang, X. Xue, R. Chen, S. Yang and Z. Jin, Adv. Sci., 2018, 5, 1700275 CrossRef PubMed .
164. M. Asadi, K. Kim, C. Liu, A. V. Addepalli, P. Abbasi, P. Yasaei, P. Phillips, A. Behranginia, J. M. Cerrato, R. Haasch, P. Zapol, B. Kumar, R. F. Klie, J. Abiade, L. A. Curtiss and A. Salehi-Khojin, Science, 2016, 353, 467–470 CrossRef CAS PubMed .
165. P. Abbasi, M. Asadi, C. Liu, S. Sharifi-Asl, B. Sayahpour, A. Behranginia, P. Zapol, R. Shahbazian-Yassar, L. A. Curtiss and A. Salehi-Khojin, ACS Nano, 2017, 11, 453–460 CrossRef CAS PubMed .
166. K. Lv, W. Suo, M. Shao, Y. Zhu, X. Wang, J. Feng, M. Fang and Y. Zhu, Nano Energy, 2019, 63, 103834 CrossRef CAS .
167. S. A. Francis, J. M. Velazquez, I. M. Ferrer, D. A. Torelli, D. Guevarra, M. T. McDowell, K. Sun, X. Zhou, F. H. Saadi, J. John, M. H. Richter, F. P. Hyler, K. M. Papadantonakis, B. S. Brunschwig and N. S. Lewis, Chem. Mater., 2018, 30, 4902–4908 CrossRef CAS .
168. X. Mao, L. Wang, Y. Xu and Y. Li, J. Phys. Chem. C, 2020, 124, 10523–10529 CrossRef CAS .
169. W. Huang, D. Zhou, H. Yang, X. Liu and J. Luo, ACS Appl. Energy Mater., 2021, 4, 7492–7496 CrossRef CAS .
170. J. Xu, X. Li, W. Liu, Y. Sun, Z. Ju, T. Yao, C. Wang, H. Ju, J. Zhu, S. Wei and Y. Xie, Angew. Chem., Int. Ed., 2017, 56, 9121–9125 CrossRef CAS PubMed .
171. H. Li, X. Liu, S. Chen, D. Yang, Q. Zhang, L. Song, H. Xiao, Q. Zhang, L. Gu and X. Wang, Adv. Energy Mater., 2019, 9, 1900072 CrossRef .
172. K. Lv, C. Teng, M. Shi, Y. Yuan, Y. Zhu, J. Wang, Z. Kong, X. Lu and Y. Zhu, Adv. Funct. Mater., 2018, 28, 1802339 CrossRef .
173. F. Li, S.-F. Zhao, L. Chen, A. Khan, D. R. MacFarlane and J. Zhang, Energy Environ. Sci., 2016, 9, 216–223 RSC .
174. Y. Linghu, T. Tong, C. Li and C. Wu, Appl. Surf. Sci., 2022, 590, 153001 CrossRef CAS .
175. Y. Ren, X. Sun, K. Qi and Z. Zhao, Appl. Surf. Sci., 2022, 154211 CrossRef CAS .
176. Z.-Z. Lin, X.-M. Li, X.-W. Chen and X. Chen, Phys. Chem. Chem. Phys., 2022, 24, 3733–3740 RSC .
177. Z. W. Chen, W. Gao, W. T. Zheng and Q. Jiang, ChemSusChem, 2018, 11, 1455–1459 CrossRef CAS PubMed .
178. S. Kang, S. Han and Y. Kang, ChemSusChem, 2019, 12, 2671–2678 CrossRef CAS PubMed .
179. X. Hong, K. Chan, C. Tsai and J. K. Nørskov, ACS Catal., 2016, 6, 4428–4437 CrossRef CAS .
180. X. Yao, Z. W. Chen, G. J. Liu, X. Y. Lang, Y. F. Zhu, W. Gao and Q. Jiang, ChemSusChem, 2021, 14, 2255–2261 CrossRef CAS PubMed .
181. P. Li, H. Hu, J. Xu, H. Jing, H. Peng, J. Lu, C. Wu and S. Ai, Appl. Catal., B, 2014, 147, 912–919 CrossRef CAS .
182. H. Peng, J. Lu, C. Wu, Z. Yang, H. Chen, W. Song, P. Li and H. Yin, Appl. Surf. Sci., 2015, 353, 1003–1012 CrossRef CAS .
183. Y. Wang, Z. Zhang, L. Zhang, Z. Luo, J. Shen, H. Lin, J. Long, J. C. Wu, X. Fu, X. Wang and C. Li, J. Am. Chem. Soc., 2018, 140, 14595–14598 CrossRef CAS PubMed .
184. F. Yu, X. Jing, Y. Wang, M. Sun and C. Duan, Angew. Chem., Int. Ed., 2021, 60, 24849–24853 CrossRef CAS PubMed .
185. S. F. Ng, M. Y. L. Lau and W. J. Ong, Adv. Mater., 2021, 33, 2008654 CrossRef CAS PubMed .
186. M. Wang, H. Yang, K. Shen, H. Xu, W. Wang, Z. Yang, L. Zhang, J. Chen, Y. Huang, M. Chen, D. Mitlin and X. Li, Small Methods, 2020, 4, 2000353 CrossRef CAS .
187. H. Lin, S. Zhang, T. Zhang, H. Ye, Q. Yao, G. W. Zheng and J. Y. Lee, Adv. Energy Mater., 2019, 9, 1902096 CrossRef CAS .
188. C. Li, Z. Xi, D. Guo, X. Chen and L. Yin, Small, 2018, 14, 1701986 CrossRef PubMed .
189. S. Rehman, K. Khan, Y. Zhao and Y. Hou, J. Mater. Chem. A, 2017, 5, 3014–3038 RSC .
190. J. Yan, X. Liu and B. Li, Adv. Sci., 2016, 3, 1600101 CrossRef PubMed .
191. B. Li, Q. Su, L. Yu, J. Zhang, G. Du, D. Wang, D. Han, M. Zhang, S. Ding and B. Xu, ACS Nano, 2020, 14, 17285–17294 CrossRef CAS PubMed .
192. D. Liu, C. Zhang, G. Zhou, W. Lv, G. Ling, L. Zhi and Q. H. Yang, Adv. Sci., 2018, 5, 1700270 CrossRef PubMed .
193. S. Luo, J. Ruan, Y. Wang, J. Hu, Y. Song, M. Chen and L. Wu, Small, 2021, 17, 2101879 CrossRef CAS PubMed .
194. Z. Ma, Y. Liu, J. Gautam, W. Liu, A. N. Chishti, J. Gu, G. Yang, Z. Wu, J. Xie, M. Chen, L. Ni and G. Diao, Small, 2021, 17, 2102710 CrossRef CAS PubMed .
195. T. Yang, B. Guo, W. Du, M. K. Aslam, M. Tao, W. Zhong, Y. Chen, S. J. Bao, X. Zhang and M. Xu, Adv. Sci., 2019, 6, 1901557 CrossRef CAS PubMed .
196. H. Wang, Q. Zhang, H. Yao, Z. Liang, H.-W. Lee, P.-C. Hsu, G. Zheng and Y. Cui, Nano Lett., 2014, 14, 7138–7144 CrossRef CAS PubMed .
197. G. Babu, N. Masurkar, H. Al Salem and L. M. R. Arava, J. Am. Chem. Soc., 2017, 139, 171–178 CrossRef CAS PubMed .
198. H. Lin, L. Yang, X. Jiang, G. Li, T. Zhang, Q. Yao, G. W. Zheng and J. Y. Lee, Energy Environ. Sci., 2017, 10, 1476–1486 RSC .
199. Y. Pan, L. Gong, X. Cheng, Y. Zhou, Y. Fu, J. Feng, H. Ahmed and H. Zhang, ACS Nano, 2020, 14, 5917–5925 CrossRef CAS PubMed .
200. Z. Du, Y. Guo, H. Wang, J. Gu, Y. Zhang, Z. Cheng, B. Li, S. Li and S. Yang, ACS Nano, 2021, 15, 19275–19283 CrossRef CAS PubMed .
201. Y. Zhang, Z. Mu, C. Yang, Z. Xu, S. Zhang, X. Zhang, Y. Li, J. Lai, Z. Sun, Y. Yang, Y. Chao, C. Li, X. Ge, W. Yang and S. Guo, Adv. Funct. Mater., 2018, 28, 1707578 CrossRef .
202. Z. Cheng, Y. Chen, Y. Yang, L. Zhang, H. Pan, X. Fan, S. Xiang and Z. Zhang, Adv. Energy Mater., 2021, 11, 2003718 CrossRef CAS .
203. Z.-L. Xu, N. Onofrio and J. Wang, J. Mater. Chem. A, 2020, 8, 17646–17656 RSC .
204. C. Tian, B. Li, X. Hu, J. Wu, P. Li, X. Xiang, X. Zu and S. Li, ACS Appl. Mater. Interfaces, 2021, 13, 6229–6240 CrossRef CAS PubMed .
205. W. Liu, C. Luo, S. Zhang, B. Zhang, J. Ma, X. Wang, W. Liu, Z. Li, Q.-H. Yang and W. Lv, ACS Nano, 2021, 15, 7491–7499 CrossRef CAS PubMed .
206. Y. Li, M. Chen, H. Liu, P. Zeng, D. Zhang, H. Yu, X. Zhou, C. Guo, Z. Luo, Y. Wang, B. Chang and X. Wang, Chem. Eng. J., 2021, 425, 131640 CrossRef CAS .
207. S. Bhoyate, J. Kim, E. Lee, B. Park, E. Lee, J. Park, S. H. Oh, J. Kim and W. Choi, J. Mater. Chem. A, 2020, 8, 12436–12445 RSC .
208. S. Bhoyate, B. Park, S. H. Oh and W. Choi, J. Power Sources, 2021, 511, 230426 CrossRef CAS .
209. K. Liu, X. Zhang, F. Miao, Z. Wang, S. Zhang, Y. Zhang, P. Zhang and G. Shao, Small, 2021, 17, 2100065 CrossRef CAS PubMed .
210. H. Song, T. Li, T. He, Z. Wang, D. Fang, Y. Wang, X. L. Li, D. Zhang, J. Hu and S. Huang, Chem. Eng. J., 2022, 450, 138115 CrossRef CAS .
211. Y. Li, X. Yan, Z. Zhou, J. Liu, Z. Zhang, X. Guo, H. Peng and G. Li, Appl. Surf. Sci., 2022, 574, 151586 CrossRef CAS .
212. S. Wang, S. Feng, J. Liang, Q. Su, F. Zhao, H. Song, M. Zheng, Q. Sun, Z. Song, X. Jia, J. Yang, Y. Li, J. Liao, R. Li and X. Sun, Adv. Energy Mater., 2021, 11, 2003314 CrossRef CAS .
213. J. He, G. Hartmann, M. Lee, G. S. Hwang, Y. Chen and A. Manthiram, Energy Environ. Sci., 2019, 12, 344–350 RSC .
214. Q. Wu, Z. Yao, X. Zhou, J. Xu, F. Cao and C. Li, ACS Nano, 2020, 14, 3365–3377 CrossRef CAS PubMed .
215. M. Wang, L. Fan, D. Tian, X. Wu, Y. Qiu, C. Zhao, B. Guan, Y. Wang, N. Zhang and K. Sun, ACS Energy Lett., 2018, 3, 1627–1633 CrossRef CAS .
216. M. Song, H. Tan, X. Li, A. I. Y. Tok, P. Liang, D. Chao and H. J. Fan, Small Methods, 2020, 4, 1900274 CrossRef CAS .
217. W. Wang and Y.-C. Lu, Acc. Mater. Res., 2021, 2, 515–525 CrossRef CAS .
218. Y. Yan, S. Liang, X. Wang, M. Zhang, S.-M. Hao, X. Cui, Z. Li and Z. Lin, Proc. Natl. Acad. Sci. U. S. A., 2021, 118, e2110036118 CrossRef CAS PubMed .
219. Z. Sadighi, J. Liu, L. Zhao, F. Ciucci and J.-K. Kim, Nanoscale, 2018, 10, 22549–22559 RSC .
220. Y. Liu, Y. Zang, X. Liu, J. Cai, Z. Lu, S. Niu, Z. Pei, T. Zhai and G. Wang, Front. Energy Res., 2020, 8, 109 CrossRef .
221. X. Fan, Y. Huang, H. Wang, F. Zheng, C. Tan, Y. Li, X. Lu, Z. Ma and Q. Li, Electrochim. Acta, 2021, 369, 137653 CrossRef CAS .
222. S. Zhang, Z. Huang, Z. Wen, L. Zhang, J. Jin, R. Shahbazian-Yassar and J. Yang, Nano Lett., 2017, 17, 3518–3526 CrossRef CAS PubMed .
223. L. Li, C. Chen, J. Su, P. Kuang, C. Zhang, Y. Yao, T. Huang and A. Yu, J. Mater. Chem. A, 2016, 4, 10986–10991 RSC .
224. S. Wu, W. Di, D. Zhang, W. Liu, H. Luo, J. He, Q. Yang, Z. Li and R. Liu, Energy Environ. Mater., 2022, 5, 918–927 CrossRef CAS .
225. A. Jaradat, C. Zhang, S. K. Singh, J. Ahmed, A. Ahmadiparidari, L. Majidi, S. Rastegar, Z. Hemmat, S. Wang, A. T. Ngo, L. A. Curtiss, M. Daly, A. Subramanian and A. Salehi-khojin, Small, 2021, 17, 2102072 CrossRef CAS PubMed .
226. M. Asadi, B. Sayahpour, P. Abbasi, A. T. Ngo, K. Karis, J. R. Jokisaari, C. Liu, B. Narayanan, M. Gerard, P. Yasaei, X. Hu, A. Mukherjee, K. C. Lau, R. S. Assary, F. Khalili-Araghi, R. F. Klie, L. A. Curtiss and A. Salehi-Khojin, Nature, 2018, 555, 502–506 CrossRef CAS PubMed .
227. A. Hu, J. Long, C. Shu, R. Liang and J. Li, ACS Appl. Mater. Interfaces, 2018, 10, 34077–34086 CrossRef CAS PubMed .
228. A. Hu, C. Shu, X. Qiu, M. Li, R. Zheng and J. Long, ACS Sustainable Chem. Eng., 2019, 7, 6929–6938 CrossRef CAS .
229. D. Li, L. Zhao, Q. Xia, J. Wang, X. Liu, H. Xu and S. Chou, Adv. Funct. Mater., 2022, 32, 2108153 CrossRef CAS .
230. D. Li, L. Zhao, Q. Xia, L. Liu, W. Fang, Y. Liu, Z. Zhou, Y. Long, X. Han, Y. Zhang, J. Wang, Y. Wu and H. Liu, Small, 2022, 18, 2105752 CrossRef CAS PubMed .
231. X. Wen, D. Du, L. Ren, H. Xu, R. Li, C. Zhao and C. Shu, Chem. Eng. J., 2022, 442, 136311 CrossRef CAS .
232. Z. Sun, J. He, M. Yuan, L. Lin, Z. Zhang, Z. Kang, Q. Liao, H. Li, G. Sun, X. Yang, R. Long and Y. Zhang, Nano Energy, 2019, 65, 103996 CrossRef CAS .
233. G. Solomon, M. G. Kohan, M. Vagin, F. Rigoni, R. Mazzaro, M. M. Natile, S. You, V. Morandi, I. Concina and A. Vomiero, Nano Energy, 2021, 81, 105664 CrossRef CAS .
234. P. Zhang, X. Lu, Y. Huang, J. Deng, L. Zhang, F. Ding, Z. Su, G. Wei and O. G. Schmidt, J. Mater. Chem. A, 2015, 3, 14562–14566 RSC .
235. X. Wang, Y. Li, X. Bi, L. Ma, T. Wu, M. Sina, S. Wang, M. Zhang, J. Alvarado, B. Lu, A. Banerjee, K. Amine, J. Lu and Y. S. Meng, Joule, 2018, 2, 2381–2392 CrossRef CAS .
236. I. S. Amiinu, Z. Pu, X. Liu, K. A. Owusu, H. G. R. Monestel, F. O. Boakye, H. Zhang and S. Mu, Adv. Funct. Mater., 2017, 27, 1702300 CrossRef .
237. M. K. Puglia, M. Malhotra, A. Chivukula and C. V. Kumar, ACS Appl. Nano Mater., 2021, 4, 10389–10398 CrossRef CAS .
238. C. He, L. Cui, X. Wu, Y. Cai, Z. Ma, X. Zhong, H. Wang, Q. Li, Y. Huang and Q. Pan, ACS Appl. Energy Mater., 2021, 4, 9498–9506 CrossRef CAS .
239. P. Basumatary, D. Konwar and Y. S. Yoon, Appl. Catal., B, 2020, 267, 118724 CrossRef CAS .
240. A. Ahmadiparidari, R. E. Warburton, L. Majidi, M. Asadi, A. Chamaani, J. R. Jokisaari, S. Rastegar, Z. Hemmat, B. Sayahpour, R. S. Assary, B. Narayanan, P. Abbasi, P. C. Redfern, A. Ngo, M. Vörös, J. Greeley, R. Klie, L. A. Curtiss and A. Salehi-Khojin, Adv. Mater., 2019, 31, 1902518 CrossRef CAS PubMed .
241. C.-J. Chen, C.-S. Huang, Y.-C. Huang, F.-M. Wang, X.-C. Wang, C.-C. Wu, W.-S. Chang, C.-L. Dong, L.-C. Yin and R.-S. Liu, ACS Appl. Mater. Interfaces, 2021, 13, 6156–6167 CrossRef CAS PubMed .
242. N. Salazar, S. Rangarajan, J. Rodríguez-Fernández, M. Mavrikakis and J. V. Lauritsen, Nat. Commun., 2020, 11, 4369 CrossRef PubMed .
243. P. Zheng, A. Duan, K. Chi, L. Zhao, C. Zhang, C. Xu, Z. Zhao, W. Song, X. Wang and J. Fan, Chem. Eng. Sci., 2017, 164, 292–306 CrossRef CAS .
244. P. Nikulshin, V. Salnikov, A. Mozhaev, P. Minaev, V. Kogan and A. Pimerzin, J. Catal., 2014, 309, 386–396 CrossRef CAS .
245. S. P. Lonkar, V. V. Pillai and S. M. Alhassan, ACS Appl. Nano Mater., 2018, 1, 3114–3123 CrossRef CAS .
246. X.-z Wang, N. Liu, H. Hu, X.-p Wang and J.-s Qiu, Carbon, 2014, 76, 471 CrossRef .
247. C. F. Weise, H. Falsig, P. G. Moses, S. Helveg, M. Brorson and L. P. Hansen, J. Catal., 2021, 403, 74–86 CrossRef CAS .
248. S. S. Grønborg, N. Salazar, A. Bruix, J. Rodríguez-Fernández, S. D. Thomsen, B. Hammer and J. V. Lauritsen, Nat. Commun., 2018, 9, 2211 CrossRef PubMed .
249. H. Shang, T. Wang and W. Zhang, Chem. Eng. Sci., 2019, 195, 208–217 CrossRef CAS .
250. R. V. Mom, J. N. Louwen, J. W. Frenken and I. M. Groot, Nat. Commun., 2019, 10, 2546 CrossRef PubMed .
251. A. Tuxen, J. Kibsgaard, H. Gøbel, E. Lægsgaard, H. Topsøe, J. V. Lauritsen and F. Besenbacher, ACS Nano, 2010, 4, 4677–4682 CrossRef CAS PubMed .
252. J.-F. Paul and E. Payen, J. Phys. Chem. B, 2003, 107, 4057–4064 CrossRef CAS .
253. G. Liu, A. W. Robertson, M. M.-J. Li, W. C. Kuo, M. T. Darby, M. H. Muhieddine, Y.-C. Lin, K. Suenaga, M. Stamatakis, J. H. Warner and S. C. E. Tsang, Nat. Chem., 2017, 9, 810–816 CrossRef CAS PubMed .
254. K. Wu, W. Wang, H. Guo, Y. Yang, Y. Huang, W. Li and C. Li, ACS Energy Lett., 2020, 5, 1330–1336 CrossRef CAS .
255. X. Bai, Q. Li, L. Shi, C. Ling and J. Wang, Catal. Today, 2020, 350, 56–63 CrossRef CAS .
256. J. Hu, L. Yu, J. Deng, Y. Wang, K. Cheng, C. Ma, Q. Zhang, W. Wen, S. Yu, Y. Pan, J. Yang, H. Ma, F. Qi, Y. Wang, Y. Zheng, M. Chen, R. Huang, S. Zhang, Z. Zhao, J. Mao, X. Meng, Q. Ji, G. Hou, X. Han, X. Bao, Y. Wang and D. Deng, Nat. Catal., 2021, 4, 242–250 CrossRef CAS .

This journal is © The Royal Society of Chemistry 2023