Photocatalytic and Electrocatalytic Generation of Hydrogen Peroxide: Principles, Catalyst Design and Performance

Highlights

• Basic principles of photo- and electro-catalytic hydrogen peroxide generation.

• Recent progresses on the design, performance and mechanisms of photo- and electro-catalysts for hydrogen peroxide generation.

• Scientific challenges and prospects of engineering photo- and electro-catalysts for hydrogen peroxide production.

Abstract

Hydrogen peroxide (H₂O₂) is a high-demand organic chemical reagent and has been widely used in various modern industrial applications. Currently, the prominent method for the preparation of H₂O₂ is the anthraquinone oxidation. Unfortunately, it is not conducive to economic and sustainable development since it is a complex process and involves unfriendly environment and potential hazards. In this context, numerous approaches have been developed to synthesize H₂O₂. Among them, photo/electro-catalytic ones are considered as two of the most promising manners for on-site synthesis of H₂O₂. These alternatives are sustainable in that only water or O₂ is required. Namely, water oxidation (WOR) or oxygen reduction (ORR) reactions can be further coupled with clean and sustainable energy. For photo/electro-catalytic reactions for H₂O₂ generation, the design of the catalysts is extremely important and has been extensively conducted with an aim to obtain ultimate catalytic performance. This article overviews the basic principles of WOR and ORR, followed by the summary of recent progresses and achievements on the design and performance of various photo/electro-catalysts for H₂O₂ generation. The related mechanisms for these approaches are highlighted from theoretical and experimental aspects. Scientific challenges and opportunities of engineering photo/electro-catalysts for H₂O₂ generation are also outlined and discussed.

1 Introduction

Hydrogen peroxide (H₂O₂) is a strong oxidant reagent compared with Cl-containing oxidants such as HOCl. This efficient and environmental-friendly chemical has been selected as one of the 100 most important chemical substances in the world [1, 2]. Owing to its strong oxidation capability, H₂O₂ has been widely used in a wide range of industrial and household applications, including wastewater treatment, chemical synthesis, industrial bleaching, energy storage, medical disinfection and a promising energy carrier in fuel cells [3,4,5,6,7]. It has been pointed out that H₂O₂ has advantages of the highest concentration of active oxygen (47.1 wt%) and the cleanest by-product (H₂O) without carbon emission. These features make H₂O₂ a promising fuel alternative to H₂ and O₂ for one-chamber fuel cells without a membrane [8,9,10]. Although H₂O₂ fuel cells produce a bit lower theoretical output voltage (1.09 V) than H₂/O₂ fuel cells (1.23 V), the energy density of H₂O₂ (2.1 MJ kg⁻¹ for 60% aqueous H₂O₂) is comparable to that of compressed hydrogen (3.5 MJ kg⁻¹) [11,12,13,14,15]. In addition, H₂O₂ is a liquid fuel that can be fully soluble in water. It is thus more secure and facilitated to be transported and stored than the conventional H₂. Accordingly, H₂O₂ has been predicted to be produced on a scale of approximately 5.7 million tons by 2027 in the global market, emphasizing its growing influence in the fields of sustainable energy [16].

The widely used method to synthesize H₂O₂ in the industry is anthraquinone (AQ) oxidation, which accounts for more than 95% of the total production of H₂O₂ [17]. Generally, the AQ oxidation process is a four-step cycle: (i) hydrogenation of AQ using a catalyst, (ii) oxidation of hydrogenated AQ to regenerate AQ and produce H₂O₂, (iii) extraction of H₂O₂ and (iv) purification concentration of H₂O₂. Therefore, this process requires a high energy demandingness and resource consumption and meanwhile produces many harmful by-products in series reactions (Fig. 1a) [18]. In this context, the pursuit of high concentration of H₂O₂ via AQ oxidation technology brings in more environmental pollutants and explosive substances, being threatened to the safety in fuel storage and transportation [19, 20]. To avoid these shortcomings of the AQ process, the direct preparation of H₂O₂ through the reaction from H₂ and O₂ appears more popular. It has been considered as a more energy-efficient and resource-efficient route alternative to the AQ oxidation (Fig. 1b) [21,22,23,24]. This partial hydrogenation reaction can be fully performed at low temperatures, if the highly efficient catalysts are designed and applied. It needs to be pointed out that this process usually selects expensive Pd-based materials as catalysts to limit competing side reactions [25, 26]. More seriously, it requires precise control of the ratio of H₂ and O₂ to leave for the explosive range, which unavoidably leads to low practice yields of the direct synthesis [27,28,29]. Another potential route to synthesize H₂O₂ is through a range of enzymatic catalytic processes in nature. It has been reported that some oxidase enzymes (e.g., glucose oxidase, D-amino acid oxidase and cholesterol oxidase) can selectively catalyze the corresponding native substrates to generate H₂O₂ [30,31,32,33]. Unfortunately, the inferior activity with a low value of turnover frequency occurs in these enzymatic catalytic reactions, which has impeded the rapid development of such biosynthesis of H₂O₂ [34, 35]. Consequently, environmental-friendly, efficient, safe and convenient alternative routes still need to be explored for H₂O₂ production.

Fig. 1

Schematic illustration of a the anthraquinone (AQ) oxidation process and b the direct synthesis from H₂ and O₂

In recent years, photo/electro-catalytic synthesis of H₂O₂ has attracted extensive attention. Comparison with the aforementioned three routes, these new approaches are more economical and environmental-friendly [36,37,38,39]. Belonging to direct synthesis methods, they can be realized using the abundant H₂O and O₂ on the earth as reactants, but are easier to be handled with lower operating risks [40, 41]. In addition, they can take full advantages of clean and renewable energy sources (e.g., solar energy, wind energy) in a sustainable fashion. To obtain efficient photo/electro-catalytic conversion efficiency, high-performance catalyst platforms need to be established. In other words, the design of advanced photo/electro-catalysts plays a vital role in such reaction conversion [42, 43]. From both fundamental and practical viewpoints, the rational design and controllable synthesis of various photo/electro-catalysts are thus of great importance for H₂O₂ synthesis.

In this regard, many efforts have been made in the design and synthesis of effective photo/electro-catalysts for active and selective generation of H₂O₂. It is therefore extremely significant to summarize the progress and state-of-art about design, characterization and application of developed photo/electro-catalysts for active and selective production of H₂O₂. This review article provides a comprehensive account of the development of photo/electro-catalysts toward H₂O₂ generation, covering related synthesis mechanisms, state-of-the-art catalysts/materials together with their performance evaluation criterion and typical engineering strategies for H₂O₂ formation. The future perspectives of photo/electro-catalyst design for H₂O₂ production are discussed and outlined.

2 Mechanisms of Catalytic H₂O₂ Synthesis

2.1 Photocatalytic Mechanisms of H₂O₂ Synthesis

The best advantage of photocatalytic synthesis is the direct utilization of renewable and sustainable solar energy to synthesize various products or degrade series of pollutes [44,45,46]. Most of recent photocatalytic synthesis studies focus on photocatalytic reactions on the semiconductors. During a typical photocatalytic process of H₂O₂ synthesis, three consecutive fundamental steps undergo on the photocatalysts [47]. In the first step, a semiconductor photocatalyst absorbs the excitation light, of which energy is greater than the band gap of the used photocatalyst, to create the negatively charged electron (e⁻) on its conduction band (CB), accompanying positively charged hole (h⁺) on its valence band (VB). In the second step, these photoinduced charge carriers (namely both e⁻ and h⁺) in the interior are separated and diffuse into the surface of the photocatalyst. Finally, they react with H₂O and O₂ to generate H₂O₂ via different redox pathways. Some recombine with each other and do not participate in any chemical reactions.

In the second step of a photocatalytic process of H₂O₂ synthesis, either oxygen reduction reaction (ORR) or water oxidation reaction (WOR) is involved [15]. As for the process of ORR (Fig. 2a), there are two potential mechanisms: indirect two-step single-electron (O₂ → ·O₂⁻ → H₂O₂) and direct one-step two-electron (O₂ → H₂O₂) pathways. For example, the photons are absorbed by the semiconductor photocatalyst to excite the electrons in its VB to its CB, while holes are remained in its VB at the first step. Subsequently, the holes (h⁺) oxidize H₂O to produce O₂ (Eq. 1). In the indirect two-step single-electron pathway, one formed O₂ molecule reacts with one electron (e⁻) to form O₂^·− in the CB of the photocatalyst (Eq. 2a), which spontaneously combines H⁺ to generate a HO₂^·− intermediate (Eq. 2b). This radical can also get one electron to produce HO₂⁻ intermediate via another one-electron reduction reaction pathway (Eq. 2c). Finally, the synthesized HO₂⁻ species reacts with H⁺, leading to H₂O₂ evolution in the form of Eq. 2d. In the direct one-step two-electron pathway, one formed O₂ molecule directly reacts with two electrons (e⁻), resulting in the formation of H₂O₂, as demonstrated in Eq. 3.

$${\text{2 H}}_{{2}} {\text{O }} + {\text{ 4 h}}^{ + } \to {\text{O}}_{{2}} + {\text{ 4 H}}^{ + }$$

(1)

$${\text{O}}_{{2}} + {\text{ e}}^{ - } \to {\text{O}}_{{2}}^{\cdot- }$$

(2a)

$${\text{O}}_{{2}}^{\cdot- } + {\text{ H}}^{ + } \to {\text{HO}}_{{2}}^{\cdot- }$$

(2b)

$${\text{HO}}_{{2}}^{ \cdot- } + {\text{ e}}^{ - } \to {\text{HO}}_{{2}}^{ - }$$

(2c)

$${\text{HO}}_{{2}}^{ - } + {\text{ H}}^{ + } \to {\text{H}}_{{2}} {\text{O}}_{{2}}$$

(2d)

$${\text{O}}_{{2}} + {\text{ 2 H}}^{ + } + {\text{ 2 e}}^{ - } \to {\text{H}}_{{2}} {\text{O}}_{{2}}$$

(3)

Fig. 2

Simplified schematic of photocatalytic H₂O₂ generation from a ORR, b WOR and c dual-channel pathway. Simplified schematic of electrocatalytic H₂O₂ production from d ORR, e WOR and f dual-channel pathway

The mechanism of H₂O₂ synthesis via the WOR (Fig. 2b) can be divided two pathways: direct one-step (H₂O → H₂O₂) and indirect two-step (H₂O → ^·OH → H₂O₂) approaches. In the case of direct one-step process, one H₂O molecule is directly oxidized by two holes to generate one H₂O₂ through the one-step reaction (Eq. 4). Regarding indirect two-step WOR, H₂O is firstly be oxidized and the ^·OH intermediate is formed (Eq. 5a). These intermediates react with each other to produce H₂O₂ (Eq. 5b).

$${\text{2 H}}_{{2}} {\text{O }} + {\text{ 2 h}}^{ + } \to {\text{H}}_{{2}} {\text{O}}_{{2}} + {\text{ 2 H}}^{ + }$$

(4)

$${\text{H}}_{{2}} {\text{O }} + {\text{ h}}^{ + } \to^{ \cdot } {\text{OH }} + {\text{ H}}^{ + }$$

(5a)

$${2}^{ \cdot } {\text{OH}} \to {\text{H}}_{{2}} {\text{O}}_{{2}}$$

(5b)

It is acknowledged that the process of photocatalytic H₂O₂ synthesis might simultaneously occur through both the ORR and WOR mechanisms (Fig. 2c). However, these strategies have the unavoidable drawbacks [48, 49]. Firstly, the solubility of O₂ in liquid phase is low, limiting related reaction rates. Secondly, the recombination of electrons and holes is fast, reducing dramatically the efficiency of electron utilization. Thirdly, thermodynamic feasibility of the 4e⁻ pathway is greater than of the 2e⁻ pathway, leading to low selectivity of such processes. Therefore, it is necessary to design advanced photocatalysts to overcome these shortcomings.

The apparent quantum yield (AQY), defined as the ratio of the electron number in generating H₂O₂ molecule and the incident photons at a given wavelength, is an important indicator to evaluate the photocatalytic efficiency of various photocatalysts. It is frequently calculated using Eq. 6 [20].

$${\text{AQY}}\left( \% \right) \, = \frac{{2n}}{{N_{{aph}} }} \times { 1}00\%$$

(6)

where n is the electron amount of produced H₂O₂, N_aph is the number of incident photons shined on the photocatalysts.

In the section of discussion of photocatalysts (Chapter 3), we summarize recent progress and achievements on representative photocatalysts for H₂O₂ synthesis, in terms of their morphology and electronic band structures as well as their corresponding visible light-driven performance (e.g., activity, selectivity and stability) for H₂O₂ production.

2.2 Electrocatalytic Mechanisms of H₂O₂ Synthesis

Electrocatalytic synthesis of H₂O₂ via the oxygen electrochemistry has been widely considered as an attractive route for on-site production of H₂O₂. Such an approach has effectively addressed the disadvantages of the AQ process and the possibility to be coupled with sustainable energy sources [50, 51]. Note that there are two direct pathways to produce H₂O₂, including two-electron oxygen reduction reaction (2e⁻ ORR, Fig. 2d) and two-electron water oxidation reaction (2e⁻ WOR, Fig. 2e) [52,53,54]. As for the 2e⁻ ORR process, the reduction of O₂ to H₂O₂ proceeds by a two-step consecutive reaction road. This process can be described as an overall electrochemical reaction (Eq. 7), where $E^{\theta }$ represents the standard thermodynamic equilibrium potential, which is calculated to 0.70 V versus reversible hydrogen electrode (RHE). In more detail, one O₂ molecule diffuses to the active surface of an electrocatalyst and gets adsorbed. If it gains one electron from the electrode and further reacts with one H⁺ ion, the OOH* intermediate is generated (Eq. 7a, noted that * represents an unoccupied active site). Since the formed OOH* possesses high chemical reaction activity, it can combine with another H⁺ ion and one electron to produce H₂O_2, accompanying with regeneration of the active site (Eq. 7b). Unfortunately, the OOH* intermediate can undergo a four-electron pathway to decompose and form *O and *OH intermediates, which forms H₂O instead of H₂O₂ [55]. The overall electrochemical reaction for the 4e⁻ ORR process is depicted as Eq. 8, where $E^{\theta }$ is 1.23 V_RHE. In the reaction mechanism, the as-formed OOH* intermediate in the first step (Eq. 8a) reacts with one H⁺ ion and electron to form O^* (Eq. 8b), which is further combined with another H⁺ ion and electron. The formation of OH* intermediate (Eq. 8c) in this step leads to the undesirable 4e⁻ ORR, which forms H₂O, an alternative to H₂O₂ (Eq. 8d). Therefore, the OOH* intermediate plays a vital role in determining the 2e⁻ or 4e⁻ ORR pathways. For example, retaining the O–O bond in OOH*, meaning weaker oxygen binding energy, is favorable to improve the selectivity of H₂O₂ production [54, 55]. On the contrary, the stronger oxygen binding energy facilitates the 4e⁻ ORR pathway, resulting in the electrocatalytic synthesis of H₂O. The 2e⁻ ORR:

$$\text{O}_{2} + 2( {{\text{H}}^{ + } + {\text{ e}}^{ - } } ) \to \text{H}_{2} {\text{O }}_{2} \;\;\;E^{\theta } = 0.7{\text{ V}}_{\text{ RHE}}$$

(7)

$$^{*} {\mkern 1mu} + {\text{ O}}_{2} + {\mkern 1mu} \left( {{\text{H}}^{ + } + {\text{ e}}^{ - } } \right) \to {\text{OOH}}^{*}$$

(7a)

$${\text{OOH}}^{*} \, + \, \left( {{\text{H}}^{ + } + {\text{ e}}^{ - } } \right) \to {\text{H}}_{{2}} {\text{O}}_{{2}} + \, ^{*}$$

(7b)

The 4e⁻ ORR:

$$\text{O}_{2} \; + \;4(\text{H}^{ + } + e^{ - } ) \to 2\text{H}_{2} {\text{O }}\;\;\;E^{\theta } = 1.23\;{\text{ V}}_{\text{ RHE}}$$

(8)

$$^{*} \, + {\text{ O}}_{{2}} + \, \left( {{\text{H}}^{ + } + {\text{ e}}^{ - } } \right) \to {\text{OOH}}^{*}$$

(8a)

$${\text{OOH}}^{*} \, + \, \left( {{\text{H}}^{ + } + {\text{ e}}^{ - } } \right) \to {\text{H}}_{{2}} {\text{O }} + {\text{ O}}^{*}$$

(8b)

$${\text{O}}^{*} \, + \, \left( {{\text{H}}^{ + } + {\text{ e}}^{ - } } \right) \to {\text{OH}}^{*}$$

(8c)

$${\text{OH}}^{*} \, + \, \left( {{\text{H}}^{ + } + {\text{ e}}^{ - } } \right) \to {\text{H}}_{{2}} {\text{O }} + \, ^{*}$$

(8d)

Similar as photocatalytic process of H₂O₂ synthesis, both ORR and WOR might simultaneously occur in electrocatalytic process (Fig. 2f).

To evaluate the catalytic selectivity in the electrocatalytic H₂O₂ synthesis via the ORR approach, the number of electrons transferred (n) needs to be calculated by use of the rotating ring-disk electrode (RRDE) or rotating disk electrode (RDE) technique. For the RRDE measurements, linear sweep voltammetry (LSV) is generally performed on the disk electrode, while a constant potential (e.g., 1.2 V_RHE) is applied on the ring electrode. When H₂O₂ is produced on the disk, it can diffuse to the ring then be detected. Equation 9 has been often employed to calculate the n value, where I_D is the current on the disk electrode, I_R is the current at the ring electrode, and N is the efficiency of H₂O₂ collection. The N value can be obtained using Eq. 10. For the RDE measurements, the Koutecky–Levich (K–L) equation is applied to calculate n (Eq. 11), where I is the measured steady-state current derived from the ORR (mA cm⁻²), I_K is the kinetic current of the reaction with active substances on the electrode surface, F is the Faraday constant (96,500 C mol⁻¹), A is the geometric area of the work electrode, D₀ is the diffusion coefficient of O₂ in the electrolyte (cm² s⁻¹), ω is the angular rotation speed (ω = 2πN, N is the linear rotation speed), ν is the kinematic viscosity of the electrolyte (cm² s⁻¹), ${C}_{{O}_{2}}$ is the saturated concentration of O₂ in the solution (mol cm⁻³).

$$n\; = \;4\;\frac{{I_{D} }}{{I_{D } + {\raise0.7ex\hbox{${I_{R} }$} \!\mathord{\left/ {\vphantom {{I_{R} } N}}\right.\kern-0pt} \!\lower0.7ex\hbox{$N$}} }}$$

(9)

$${\text{ N}}\;{ = }\;{ - }\frac{{I_{R} }}{{I_{D} }}$$

(10)

$$\frac{1}{I}\; = \;\frac{1}{{I_{K} }}\; + \;\frac{1}{{0.620nFAD_{0}^{2/3} {\upomega }^{1/2} {\upnu }^{ - 1/6} C_{{O_{2} }} }}$$

(11)

Using as-obtained n value, the selectivity of H₂O₂ production can be calculated according to Eq. 12.

$${\text{H}}_{2} {\text{O}}_{2}\, \left( {\text{\% }} \right){ } = { }\frac{4 - n}{2}$$

(12)

Another promising route for electrocatalytic synthesis of H₂O₂ is the 2e⁻ WOR pathway, which can be depicted in an overall electrochemical reaction (Eq. 13) with $E^{\theta }$ of 1.76 V_RHE. In this process, one H₂O molecule is firstly electrooxidized by active sites on the electrode to form OH^*, providing one electron and H⁺ ion (Eq. 13a). Another H₂O molecule goes through the same reaction (Eq. 13b). Two OH^* then combine with each other, leading to the production of H₂O₂ (Eq. 13c). Similar with the case in ORR, the 2e⁻ WOR pathway faces the influence from the 4e⁻ WOR pathway (Eq. 14), where the thermodynamically equilibrium potential (1.2 V_RHE) is lower than that in the 2e⁻ WOR pathway [41, 56]. In more detail, the OH^* radicals generated in the first step (Eq. 14a) are further electrooxidized, resulting in the formation of O^* intermediate (Eq. 14b). These O^* intermediates then lose one electron to form OOH^* intermediate (Eq. 14c). Finally, the OOH^* intermediate converts into O₂ molecular to complete this 4e⁻ WOR process (Eq. 14d).

The 2e⁻ WOR:

$$2{\text{H}}_{2} {\text{O}} \to {\text{H}}_{2} {\text{O}}_{2} + 2({\text{H}}^{ + } + {\text{e}}^{ - } )\,{\text{E}}^{\theta } = 1.76\,{\text{ V}}_{\text{ RHE}}$$

(13)

$$^{*} \, + {\text{ H}}_{{2}} {\text{O}} \to {\text{OH}}^{*} \, + \, \left( {{\text{H}}^{ + } + {\text{ e}}^{ - } } \right)$$

(13a)

$$^{*} \, + {\text{ H}}_{{2}} {\text{O}} \to {\text{OH}}^{*} \, + \, \left( {{\text{H}}^{ + } + {\text{ e}}^{ - } } \right)$$

(13b)

$${\text{OH}}^{*} \, + {\text{ OH}}^{*} \to {\text{H}}_{{2}} {\text{O}}_{{2}}$$

(13c)

The 4e⁻ WOR:

$$2{\text{ H}}_{{2}} {\text{O}} \to {\text{O}}_{2} + 4({\text{H}}^{ + } + {\text{ e}}^{ - } )\,{\text{E}}^{\theta } = 1.23\,{\text{ V}}_{\text{ RHE}}$$

(14)

$$^{*} \, + {\text{ H}}_{{2}} {\text{O}} \to {\text{OH}}^{*} \, + \, \left( {{\text{H}}^{ + } + {\text{ e}}^{ - } } \right)$$

(14a)

$${\text{OH}}^{*} \to {\text{O}}^{*} \, + \, \left( {{\text{H}}^{ + } + {\text{ e}}^{ - } } \right)$$

(14b)

$${\text{O}}^{*} \to {\text{OOH}}^{*} \, + \, \left( {{\text{H}}^{ + } + {\text{ e}}^{ - } } \right)$$

(14c)

$${\text{OOH}}^{*} \to ^{*} \, + {\text{ O}}_{{2}} + \, \left( {{\text{H}}^{ + } + {\text{ e}}^{ - } } \right)$$

(14d)

The key factor to dominate the H₂O₂ synthesis via the 2e⁻ WOR pathway is thus the appropriate interaction between the catalytic surface and O intermediates (e.g., ^*O, ^*OH and ^*OOH) [57]. Specifically, too strong OH binding promotively oxidizes ^*OH to ^*O, further to ^*OOH, thereby prompting the undesirable 4e⁻ pathway to generate O₂. On the other hand, too weak binding improves the selectivity of 2e⁻ WOR, but inevitably causes high kinetic barriers to hydrolysis and significantly slows down the reaction rate. It is disappointed that H₂O₂ produced in the 2e⁻ WOR process is not chemically stable. The undesirable disproportionation (2 H₂O₂ → 2 H₂O + O₂) or homolysis (H₂O₂ → 2 ·OH) spontaneously arises, decreasing the selectivity of H₂O₂ production. Therefore, the electrocatalysts with proper binding energies are highly pursued to boost 2e⁻ pathway toward H₂O₂ production.

In general, the selectivity of H₂O₂ generation via H₂O electrooxidation is defined using the term of Faradaic Efficiency (%FE) (Eq. 15).

$$\% {\text{FE }} = \frac{{Q_{{H_{2} O_{2} }} }}{{Q_{total} }}$$

(15)

where ${Q}_{{H}_{2}{O}_{2}}$ is the charge that is used to produce H₂O₂, and ${Q}_{total}$ is the total charge passed during the whole process for water electrooxidation. In the section of the discussion of electrocatalysts (Chapter 4), we will summarize and evaluate research progress on representative electrocatalysts for synthesizing H₂O₂, in terms of the morphology, structure, composition as well as the corresponding activity, selectivity and stability.

It should be pointed out that the detection of the H₂O₂ concentration in the electrolyte is very important for the practical applications. Such a concentration is generally determined by the following methods, including [58]: (1) titration using potassium permanganate (KMnO₄) or cerium sulfate [Ce(SO₄)₂], (2) colorimetric change from Fe²⁺ to Fe³⁺, (3) colorimetric determination using titanium (IV) sulfate and (4) high-performance liquid chromatography (HPLC). To separate the generated H₂O₂ from the electrolyte, the oil–water two-phase systems or distillation can be applied, although few reports have focused on such an issue.

3 Photocatalytic H₂O₂ Synthesis

3.1 Graphitic Carbon Nitride

Graphitic carbon nitride (g-C₃N₄) possesses a graphitic stacking of structure aromatic molecules with an alternating arrangement of earth-abundant carbon and nitrogen elements. Since the pioneering work of g-C₃N₄ for the promotion of photocatalytic hydrogen evolution under visible light irradiation, g-C₃N₄ has received strongly attention [59]. Such a situation is derived from this metal-free polymer photocatalyst features intrinsic advantages, such as its suitable band gap (~ 2.7 eV) for visible light response, facile preparation, relatively higher conductivity, excellent chemical stability in strong acidic and alkali media and special electronic structure [60,61,62,63,64]. In this context, it has been frequently employed for various photocatalytic applications, such as CO₂ reduction, H₂ production, and especially for H₂O₂ production [65,66,67,68,69,70]. Especially, since the CB potential (−1.3 V) of g-C₃N₄ is more negative than the reduction potential of O₂/H₂O₂ (0.695 V), g-C₃N₄ is thermodynamically promising to reduce O₂ for the H₂O₂ production using visible light [71]. In addition, the VB potential of g-C₃N₄ (1.4 V) is lower enough to effectively prevent the oxidative decomposition of H₂O₂ [64]. For example, Shiraishi’s group employed g-C₃N₄ for photocatalytic H₂O₂ production with the selectivity higher than 90%. The generated intermediate generated on the g-C₃N₄ surface 1,4-endoperoxide promotes the two-electron ORR pathway to produce H₂O₂ [68]. However, it must stress that the pristine g-C₃N₄ has many drawbacks that limit its photocatalytic performance toward the H₂O₂ production, covering its small specific surface area, weak ability to capture visible light and low chemical adsorption capacity of O₂ on its surface [72, 73]. To overcome these shortcomings, ultra-thin g-C₃N₄ nanoplates and hexagonal rosettes of g-C₃N₄ have been prepared, which showed remarkable photocatalytic activity in the H₂O₂ production, stemming from their high surface area, rich active sites and strong light-harvesting capability [74, 75]. Unfortunately, the overall photo-conversion efficiencies of these g-C₃N₄ nano-photocatalysts are still unsatisfactory, due to rapid recombination of photogenerated carriers. Therefore, it is necessary to introduce organic scavengers into these nano-photocatalysts (e.g., ultra-thin g-C₃N₄ nanoplates and hexagonal rosettes of g-C₃N₄) to consume holes in the process of H₂O₂ production [70, 76]. In short, the efficiency for visible-light excitation, the separation of photogenerated carriers and photo-reaction kinetics of g-C₃N₄ photocatalysts must be further improved. In this regard, surface engineer of g-C₃N₄ photocatalysts has been proposed and developed to facilitate its performance with the enhanced the photocatalytic production efficiency of H₂O₂ on g-C₃N₄ photocatalysts. To optimize the photocatalytic H₂O₂ synthesis, the reported strategies cover manufacturing of surface defects, loading precious metal nanoparticles, constructing heterojunction composites, polyoxometalate hybridization and metal/non-metal element doping are performed (Table 1). In the following sessions, the details of these strategies are explained.

Table 1 Summary of the g-C₃N₄ photocatalysts for the H₂O₂ production

3.1.1 Manufacturing Surface Defects

Vacancy defects are known to have the ability to capture photogenerated electrons/holes, thereby effective inhibiting the recombination of photogenerated electrons and holes [77,78,79,80]. Meanwhile, they can enhance the adsorption and activation of gas molecules (e.g., O₂), due to their abundant local electrons for the adjustment of the electronic structure, thus facilitating a photocatalytic process [81, 82]. In general, two main types of defects appear in (or on) the g-C₃N₄ photocatalysts: carbon and nitrogen vacancies.

Carbon vacancies The defects derived from carbon element in/on g-C₃N₄ photocatalysts bring in carbon vacancies. They stand at mid-gap states between VB and CB and thus can capture more visible light [46]. Namely, the optical and electronic structures of g-C₃N₄ photocatalysts can be optimized. For instance, the introduction of carbon vacancies to the surface of g-C₃N₄ accompanied the appearance of amino group by simple calcination under argon atmosphere (Fig. 3a). In this case, the generated carbon vacancies (Cv-g-C₃N₄) reduce the band gap of g-C₃N₄, thereby expanding the absorption of visible light range and increasing excitable electron longevity [71]. Moreover, carbon vacancies offer more sites to adsorb molecular oxygen, beneficial for the transfer of electrons from g-C₃N₄ to the surface adsorbed O₂. Furthermore, this Cv-g-C₃N₄ photocatalyst alters the H₂O₂ generation pathway: from a two-step single-electron indirect reduction pathway to a one-step two-electron direct one. This is attributed to generated amino groups around carbon vacancies (Fig. 3b). The activity of Cv-g-C₃N₄ for the reduction of O₂ to H₂O₂ is 14-fold higher than that of a pure g-C₃N₄ photocatalyst under irradiation with visible light (Fig. 3c). Carbon vacancies (C_vs) are then believed to produce more trapping sites, which retards the recombination of photogenerated electrons and holes, eventually resulting in an increased efficiency of the H₂O₂ generation [83]. Another representative case to strengthen the photocatalytic H₂O₂ synthesis was achieved via the creation of more vacancies, where an inverse opal (IO) structure was constructed on g-C₃N₄ accompanying with increased carbon vacancies to improve the separation ability of carriers. This IO g-C₃N₄ photocatalyst showed an outstanding H₂O₂ generation yield. For example, 325.74 μM was realized only after 2 h of visible light irradiation. The pathway of the H₂O₂ production was proposed as following: the electrons on the CB of this photocatalyst are feasible to reduce O₂ to form H₂O₂ because its CB is more negative than that of O₂/·O₂⁻.

Fig. 3

a Schematic illustration of g-C₃N₄ with carbon vacancy. b Mechanism diagram of H₂O₂ production in Cv-g-C₃N₄. c Concentration of generated H₂O₂ in g-C₃N₄ and Cv-g-C₃N₄ systems. Reproduced with permission from Ref. [71]. Copyright 2016 Elsevier. d Schematic illustration of g-C₃N₄ with nitrogen vacancies (CN-ND). e Proposed mechanism for H₂O₂ generation of CN-ND. f H₂O₂ production of different catalysts. Reproduced with permission from Ref. [86]. Copyright 2021 American Chemical Society

Nitrogen vacancies The defects from nitrogen element in/on g-C₃N₄ photocatalysts are called nitrogen vacancies. Similar to carbon vacancies, they can also facilitate the photocatalytic H₂O₂ synthesis. For example, nitrogen vacancies in g-C₃N₄ have been created via the formation of C≡N functional groups by means of thermal reduction treatment [70]. The as-formed nitrogen vacancies narrow down the band gap and make the band edge positively shift, thus enhancing the visible-light absorption and facilitating photocatalytic H₂O₂ synthesis [84]. Nitrogen vacancies in carbon nitride also effectively trap photoelectrons, boosting the reduction of N₂ to ammonium ions [82]. For the photocatalytic H₂O₂ synthesis, nitrogen vacancies have been designed and embedded in the g-C₃N₄ catalyst (PH-CN) using a dielectric barrier discharge (DBD) plasma in H₂ atmosphere. Since this photocatalyst improves the separation efficiency of the photogenerated carriers, it displays a H₂O₂ concentration of 5.2 mmol L⁻¹, 13 times higher than that obtained on a pristine g-C₃N₄ photocatalyst under the identified conditions [85]. In this case, nitrogen vacancies are assumed not only to serve as active centers for the adsorption of reactive oxygen molecules, but also to alter the electronic band structures of g-C₃N₄ to harness more visible light in the photocatalytic reaction process. One recently reported the nanostructured g-C₃N₄ with nitrogen defects was adjusted by cyanuric acid-melamine supramolecular adducts (CN-ND) in the optimum range (Fig. 3d) [86]. Due to the narrowed band gap and the newly formed midgap states, the photocatalytic H₂O₂ production on CN-ND reached 200 μM under 1 h visible light irradiation, much higher than that obtained bulk g-C₃N₄ photocatalyst (35 μM) and the pristine nanostructured g-C₃N₄ photocatalyst (85 μM) (Fig. 3e, f). Again, such improved performance of photocatalytic H₂O₂ generation was attributed to these nitrogen vacancies since they promote the visible-light-harvesting capability and retard the recombination rate of photogenerated carriers. Meanwhile, the reaction mechanism of photocatalytic H₂O₂ production on nitrogen deficiency in carbon nitride has been investigated through coupling experiments with the aid of time-dependent density functional theory (TDDFT) and density functional theory (DFT) calculations [87]. It is disclosed that the introduced bicoordinated nitrogen vacancies play an important role in oxidation, reduction and charge recombination, which are beneficial to the generation of h⁺, ·O₂⁻ and ¹O₂, respectively. Therefore, nitrogen vacancies in carbon nitride can modulate photocatalytic H₂O₂ generation.

3.1.2 Loading Precious Metal Nanoparticles

The loading of precious metal nanoparticles (NPs) on g-C₃N₄ forms strong interaction between metallic NPs and g-C₃N₄, as derived from the delocalization of long electrons in the g-C₃N₄ matrix. This is one of the most effective approaches to strengthen the photocatalytic activity and selectivity of g-C₃N₄ since the photogenerated charge separation in such photocatalysts are accelerated, leading to boost photocatalytic reaction in the special pathways [88]. In addition, these photocatalysts suppress the H₂O₂ decomposition to obtain high yields during photocatalytic H₂O₂ synthesis. For example, Au NPs deposited on the surface of g-C₃N₄ by a carbon-layer-stabilized method exhibited a higher photocatalytic yield of H₂O₂ than the pristine g-C₃N₄ due to the increased removal of the generated H₂O₂ from the active reactive sites [89]. In another case, Au NPs were uniformly dispersed on g-C₃N₄ support (Au/CN) and further utilized to produce H₂O₂ under visible light irradiation [90]. Its outstanding photocatalytic yield of H₂O₂ was believed to result from its inertness to catalyze the H₂O₂ decomposition reaction. The use of 0.01% Au led to the maximum H₂O₂ production activity (2027 μM) with light irradiation for 30 h, superior to that using other NPs (e.g., Au, Ag, Pd and Pt) on g-C₃N₄. This high activity of 0.01% Au/CN was attributed to the rapidly reduced recombination of charge carriers. It was derived from the strong interaction between Au and g-C₃N₄, as testified by obviously quenched PL signal. Photocatalytic H₂O₂ synthesis for this Au/CN photocatalyst has been further revealed using in-situ electron spin resonance (ESR) technique, indicating so-called direct 2e⁻ reduction mechanism. With similar aim to extend light absorption region and effectively suppress the recombination of electron–hole pairs for photocatalytic H₂O₂ synthesis, Ag NPs have also been combined with g-C₃N₄ [91]. For example, Ag NPs have been loaded on ultrathin g-C₃N₄ nanosheets (U-g-C₃N₄-NS) by use of a post gas etching (PGE) technology. These Ag NPs that are uniformly deposited on the surface of single U-g-C₃N₄-NS layers (Ag@U-g-C₃N₄-NS) possess significantly stronger adsorption intensity in the 200–2000 nm range than the counterparts in their UV–vis diffuse reflectance spectrum (UV–DRS, Fig. 4a). It was suggested that such enhanced adsorption of Ag@U-g-C₃N₄-NS was attributed to the special structure of the U-g-C₃N₄-NS and the localized surface plasmon resonance (LSPR) effect from the Ag NPs. Effective separation of photogenerated carriers helped to adsorb light more efficiently, leading to their enhanced photocatalytic H₂O₂ activity. An optimal structure of the Ag@U-g-C₃N₄-NS-1.0 nanocomposite showed excellent photocatalytic activity of H₂O₂ production with a yield of 1.975 × 10^–6 M min⁻¹ under visible-light irradiation, even without any sacrificial carbon-containing organic electron donor (Fig. 4b).

Fig. 4

a UV-DRS absorption spectra of different samples. b H₂O₂ concentration of different catalysts. Reproduced with permission from Ref. [91]. Copyright 2019 Wiley–VCH Verlag GmbH & Co. KGaA, Weinheim. c Proposed mechanism of different sites on all-solid-state Z-scheme heterojunctions for photocatalytic production of H₂O₂. d Concentration of generated H₂O₂ in NCN and PIx-NCN systems. Reproduced with permission from Ref. [97]. Copyright 2017 Elsevier

3.1.3 Constructing Heterojunction Composites

Using g-C₃N₄ photocatalysts that contain N and/or C vacancies to prepare H₂O₂, the organic reagents (such as methanol and ethanol) are frequently added. They usually act as hole scavengers to improve the yield, but contaminate the product of H₂O₂, and further increase the cost. Note that the potential of VB of g-C₃N₄ (+1.57 V) is negative than that of ·OH/OH⁻ (+1.99 V) [92,93,94], the H₂O₂ synthesis is thus hard to be realized through photocatalytic H₂O oxidation mechanism. Instead, a two-channel H₂O₂ generation approach (namely O₂ reduction and hole oxidation OH⁻) on g-C₃N₄ is highly desirable. To overcome these issues, heterogeneous composites of g-C₃N₄ with other components were constructed. The separation of photogenerated carriers of these composites were enhanced by introduced electron traps, increased potential of VB and created additional active sites. On these composites, the H₂O₂ generation was achieved via the H₂O oxidation pathway. More importantly, this strategy spatially isolates the oxidation and reduction reaction sites, bringing in minimized deactivation of catalytic ability of these photocatalysts. It has been reported that Z-scheme heterostructures with a wide range of light-trapping properties promoted the separation of photogenerated carriers but without the introduction of hole scavengers and thus rapidly improved the catalytic H₂O₂ production [95, 96]. For example, the combination g-C₃N₄ nanosheets (NCN) with perylene imides (PI) has been reported to form an all-solid-state Z-scheme heterojunction: a g-C₃N₄-based photocatalyst (PI_x-NCN), where the excited electrons on the CB of PI are transferred to the VB of NCN, leading to retard recombination of photogenerated carriers (Fig. 4c) [97]. In this case, this PI_x-NCN photocatalyst obviously enhanced the charge separation, resulting in an enhanced H₂O₂ yield where more electrons were found to take participate in producing H₂O₂ from O₂ reduction. In addition, since the holes in the VB of PI moiety exhibited more positive potential (2.08 V) than that of NCN (1.63 V), the H₂O₂ generation was assumed to be realized through the direct 2e⁻ WOR approach. In other words, the application of scavengers was avoided here. The optimal PI_5.0-NCN catalyst showed high activity for the H₂O₂ production: the generation of about 120 μmol H₂O₂ in 120 min under visible-light irradiation (Fig. 4d). Meanwhile, other Z-scheme heterojunction catalysts (e.g., Cu₂(OH)₂CO₃/g-C₃N₄ [98] and Bi₄O₅Br₂/g-C₃N₄ [99]) have been employed for the two-channel H₂O₂ production under visible-light conditions, thanks to their suitable CB and VB band structures for realizing completed ORR and WOR pathways.

3.1.4 Polyoxometalate Hybridization

Polyoxometalates (POMs) are commonly constructed using a basic unit of oxo-metal polyhedra of MO_x, in which a hole center (O⁻) and a trapped electron center (Mⁿ⁺) (n = 5, 6) act as the electron acceptor and donor, respectively [100]. Once POMs are excited by visible light irradiation, they exhibit promising photocatalytic activities and stability. The reason behind is that the reduction of photogenerated-charge recombination is derived from the well-defined HOMO–LUMO band gaps. In this context, POMs have been widely applied in the field of photocatalysis, such as for water oxidation, alcohols oxidation, CO₂ reduction and hydrogen evolution [101,102,103,104]. More interestingly, polyoxometalate hybridization strategy has been recently proposed, namely the formation of POM chemical bonds with g-C₃N₄, which was further utilized to improve the photocatalytic H₂O₂ synthesis. For instance, the POM cluster of [PW₁₁O₃₉]⁷⁻ (PW₁₁) has been covalently combined with three dimensionally ordered macroporous graphitic carbon nitride (3DOM g-C₃N₄), applying for the photocatalytic H₂O₂ synthesis (Fig. 5a) [105]. The hybrid catalyst of 3DOM g-C₃N₄-PW₁₁ produced 3.5 μmol H₂O₂ in 60 min irradiated by λ ≥ 320 nm light, where no organic electron donor was needed (Fig. 5b). Such a H₂O₂ yield was superior to that of the pure 3DOM g-C₃N₄ catalyst (1.3 μmol), the K-PW₁₁ catalyst (< 0.1 μmol) and the 3DOM g-C₃N₄-PW₁₁-IMP catalyst (1.5 μmol). It was explained that the covalent bonding of PW₁₁ clusters with 3DOM g-C₃N₄ optimizes the VB and the CB of 3DOM g-C₃N₄-PW₁₁ when compared with 3DOM g-C₃N₄. Such optimal band structures facilitate both the 2e⁻ WOR pathway and the 2e⁻ ORR pathway for the H₂O₂ production, accompanying with the decreased separation of photogenerated holes and electrons.

Fig. 5

a, d Schematic illustration of the synthetic route to 3DOM g-C₃N₄-PW₁₁ and g-C₃N₄-PWO. b, e H₂O₂ production over different catalysts in 60 min. c, f Scheme of energy levels and charge transfer pathways of 3DOM g-C₃N₄ and 3DOM g-C₃N₄-PW₁₁, g-C₃N₄ and g-C₃N₄-PWO. Reproduced with permission from Refs. [105, 107]. Copyright 2017 Elsevier and 2018 Elsevier

On the other hand, 3DOM g-C₃N₄-PW₁₁ hybrid catalyst still has some defects. For instance, the number and reactivity of -NH₂ groups of g-C₃N₄ need to be enhanced to further increase the covalent interaction between g-C₃N₄ and POMs. Moreover, the photoreduction ability of POMs needs to be improved to further enhance the selectivity of the 2e⁻ ORR pathway. In this context, the surface of g-C₃N₄ has been modified by its covalent binding with another POM cluster, [SiW₁₁O₃₉]⁸⁻ (SiW₁₁) via thermal treatment of g-C₃N₄ mixed with urea [106]. The as-prepared g-C₃N₄ catalyst has a larger amount of more highly active-NH₂ groups than 3DOM g-C₃N₄. Via an organic linker bride, the POM SiW₁₁ cluster with more negative CB potential than PW₁₁ was effectively covalently combined with g-C₃N₄. This g-C₃N₄-SiW₁₁ greatly enhanced the adsorption and activation of O₂, thereby promoting the two-electron reduction of O₂ for the H₂O₂ production. It produced 17.8 μmol H₂O₂ in 60 min under simulated sunlight (AM 1.5) irradiation together with the selectivity as high as 80.1% in the presence of methanol. The high selectivity of O₂ reduction to H₂O₂ was attributed to a positive shift of the CB level in g-C₃N₄-SiW₁₁.

The hybrid catalysts of g-C₃N₄ with POM-derived metal oxides were prepared by the thermal decomposition of the mixture of g-C₃N₄ and POM precursors. These POM-derived metal oxides increase the H₂O₂ production by enhancing photoinduced generation of electrons. Using such a calcination method, the g-C₃N₄-PWO [107] and g-C₃N₄-CoWO [108] have been prepared. Take g-C₃N₄-PWO as an example (Fig. 5d), it showed a H₂O₂ yield of 6.3 μmol in 60 min under visible light but in the absence of organic electron donor, outperforming that of the counterparts (Fig. 5e). This is because the incorporation of PWO into the g-C₃N₄ framework results in the negative shift of the CB edge of g-C₃N₄. The H₂O₂ production was achieved formed a two-step single-electron O₂ reduction reaction routine, namely from 1e⁻ reduction pathway (from O₂ to ·O₂⁻), and sequentially followed by another 1e⁻ reduction pathway (from ·O₂⁻ to H₂O₂) (Fig. 5f).

3.1.5 Metal/Non-metal Element Doping

Doping of metal/non-metal element into g-C₃N₄ modulates the bandgap of g-C₃N₄ and further alters the transfer directions of charge carriers, thereby adjusting electronic, optical and other physical properties of g-C₃N₄. Moreover, the heteroelements on g-C₃N₄ change the reaction pathways of the H₂O₂ generation. Therefore, doping of metal/non-metal element into g-C₃N₄ has been proposed as a potential strategy to improve the performance of photocatalytic H₂O₂ generation. In a typical case, Cu-doped g-C₃N₄ microspheres, denoted as Cu-SCN, have been synthesized using a modified template method (Fig. 6a) [109]. The Cu species were inserted into the interstitial position of SCN with a special mesoporous structure, resulting in the formation of the coordinated Cu–N bonds. The Cu-doping not only decreased the bandgap of g-C₃N₄ to enhance its capture of visible light, but also retarded the recombination of photogenerated carriers, as testified from lower PL intensity of Cu(2)-SCN than that of GCN. In addition, Cu(2)-SCN showed a weaker PL intensity in O₂ atmosphere when compared with that in Ar (Fig. 6b). The performance of photocatalytic H₂O₂ production using the Cu-doped g-C₃N₄ photocatalyst was thus superior to that of the g-C₃N₄ photocatalyst_. From the X-ray photoelectron spectroscopy (XPS) of Cu-doped g-C₃N₄ photocatalyst, Cu⁺ was observed in its XPS spectrum of Cu 2p. Such Cu(I)-N site thus served as the adsorption center for molecular O₂ to increase the O₂ adsorption. Moreover, it changes the transfer pathways of photoelectrons from the catalyst to the adsorbed O₂, altering in a two-step single-electron reaction pathway to a one-step two-electron reduction process. In the other study, K⁺ and Na⁺ ions have been co-doped into g-C₃N₄ through a molten salt method [110]. The incorporation of alkali metals in g-C₃N₄ tuned the band gap of g-C₃N₄, increased its specific surface area and reduced the size of g-C₃N₄ layered structure. Therefore, this alkali metals co-doped g-C₃N₄ photocatalyst can enhance the visible-light adsorption and promote the separation of photoelectron-hole pairs to boost photocatalytic performance toward H₂O₂ production. In more detail, the VB potential of GCN(20) is +2.05 V, more positive than that (+1.99 V) of ^·OH/OH⁻. This was derived from alkali metal incorporation. In this context, the reaction process of photocatalytic H₂O₂ generation was changed from a single pathway (O₂ → H₂O₂) to a dual channel pathway (O₂ → H₂O₂ and OH⁻ → ^·OH → H₂O₂), a very attractive method for practical solar-to-chemical applications (Fig. 6c). The MCN(20) produced the H₂O₂ concentration of 4.6 mmol L⁻¹, more than 9 times higher than that produced by GCN (Fig. 6d). Furthermore, a crystalline g-C₃N₄ modified with K⁺, Ni and N-doped carbon (KNiCN) was fabricated to overcome the inherent structural defects and low carrier separation efficiency [111]. The synergistic effects between K⁺, Ni species and N-doped carbon increased the light adsorption, improved charge separation and boosted O₂ adsorption and selectivity for the H₂O₂ generation. A two-electron reduction pathway was suggested (Fig. 6e). This photocatalyst displayed a photocatalytic H₂O₂ yield of as high as 79.6 μM in O₂-saturated pure water (Fig. 6f).

Fig. 6

a Elemental mapping images of Cu(2)-SCN. b PL intensities of GCN and Cu(2)-SCN. Reproduced with permission from Ref. [109]. Copyright 2017 Elsevier. c Band gap structures of various catalysts. d H₂O₂ concentration over catalysts under visible light. Reproduced with permission from Ref. [110]. Copyright 2018 Elsevier. e Proposed mechanism of H₂O₂ production over the KNiCN catalyst. f Photocatalytic H₂O₂ production on the catalysts in pure water. Reproduced with permission from Ref. [111]. Copyright 2021 Royal Society of Chemistry

In addition to metal doping of g-C₃N₄, halogen doping, another classic non-metal doping has been used to synthesize photocatalysts for photocatalytic H₂O₂ production. In these cases, the incorporation of halogen into g-C₃N₄ not only reduces the bandgap (to enhance the visible-light absorption and improve the electrical conductivity), but also boost charge carrier transmission rates (without damaging of main structures of the g-C₃N₄ polymer). These features benefit much photocatalytic H₂O₂ synthesis [112, 113]. For example, halogen doped g-C₃N₄, prepared by means of a hydrothermal method in saturated NH₄X (X = Cl, Br) solution followed with a post-treatment, has more negative CB potential than g-C₃N₄ [114]. This catalyst is thus beneficial for the O₂ reduction into H₂O₂. For example, a Br-doped g-C₃N₄ photocatalyst showed higher photocatalytic H₂O₂ production capacity (1.99 mmol L⁻¹) in comparison with the counterparts. The incorporation of multiple heteroelements into the g-C₃N₄ framework with various configurations was also demonstrated to tune the bandgap of g-C₃N₄ [115]. One co-doped g-C₃N₄ photocatalyst exhibited an outstanding photocatalytic activity for the H₂O₂ generation, which is 17–25 times higher than that of bare g-C₃N₄.

3.2 TiO₂ Photocatalysts

TiO₂ is the most famous photocatalyst since the discovery of its photocatalytic activity by Fujishima and Honda for the first time in 1972 [128]. It possesses many advantages for photocatalytic reactions, such as appropriate site for the CB, good physical and electrical properties, stable chemical structures and outstanding biocompatibility [129]. The potential of the CB of TiO₂ ($E^{\theta }$) is − 0.5 V_NHE, which is negative enough to drive a 2e⁻ ORR for the H₂O₂ production [16]. The generated photoelectrons in their CBs reduce O₂ to form H₂O₂, following the reaction of O₂ + 2 H⁺  + 2 e⁻  → H₂O₂. While these simultaneously generated holes in their VBs oxidize water to generate O₂ and H⁺, following the process of H₂O + 2 h⁺  → 1/2 O₂ + 2 H⁺. The free Gibbs Energy (ΔG) of such a photocatalytic process is 117 kJ mol⁻¹. An uphill reaction indicates thermodynamical unfavorablity [13, 120]. TiO₂ photocatalysts also have been confirmed to feature low cost, low toxicity, high chemical- and photo-stability during the H₂O₂ production [130,131,132].

However, the concentration of H₂O₂ produced by pure TiO₂ photocatalyst has been only confined to the micromolar level (< 0.2 mM) [133]. This is because TiO₂ photocatalyst can be excited only by UV light irradiation to generate photocarriers, due to the large band gap of TiO₂ photocatalyst [47]. Besides, the formed H₂O₂ on the TiO₂ surface is not stable enough and is easy to be converted to Ti-OOH complexes when it is in touch with the Ti–OH groups on the TiO₂ surface. Subsequently, Ti-OOH is decomposed, following the reduction reaction of Ti-OOH + H⁺  + e⁻ → Ti–OH + OH⁻ [134]. In order to address the drawbacks of these TiO₂ photocatalysts, surface modification strategies on TiO₂ photocatalysts have been validated and applied to facilitate the photocatalytic H₂O₂ production, covering loading precious metal nanoparticles, modifying graphene quantum dots and complexing with cation and anion surfaces (Table 2).

Table 2 Summary of TiO₂ photocatalysts for the H₂O₂ production

3.2.1 Loading Precious Metal Nanoparticles

The NPs from individual precious metals have been loaded on the TiO₂ surface to enhance the photocatalytic activity of TiO₂ photocatalysts. In 2010, the synthesis of the Au/TiO₂ photocatalyst has been reported using a deposition–precipitation (DP) method [133]. The as-obtained Au/TiO₂ photocatalyst exhibited a high yield in the photocatalytic H₂O₂ production. The reported H₂O₂ concentration reached 10 mM under UV-irradiation. The mechanism of the H₂O₂ formation on this Au/TiO₂ photocatalyst was explained as following (Fig. 7a). The oxidation and reduction sites on the Au/TiO₂ photocatalyst are separated to TiO₂ and Au NPs, respectively. The activity for photocatalytic H₂O₂ formation can be adjusted by k_f (the formation rate constant) and k_d (the decomposition rate constant). In another work, Au NPs with a bimodal size distribution were loaded on rutile TiO₂ (BM-Au/TiO₂), which was further modified with carbonates to obtain BM-Au/TiO₂-CO₃²⁻ (Fig. 7b) [135]. The UV/Vis absorption spectra of BM-Au/TiO₂-CO₃²⁻ indicated that the loading of Au NPs results in strong and extensive absorption in visible light region because of the LSPR of Au NPs. This BM-Au/TiO₂ photocatalyst yielded 640 ± 60 μM H₂O₂ under visible-light irradiation for 1 h when 4% HCOOH was presented as the sacrificial agent. Such a yield was much higher than that of single small Au/TiO₂ photocatalyst (50 μM) and single large Au/TiO₂ photocatalyst (75 μM) (Fig. 7c). The improvement on the H₂O₂ yield on the BM-Au/TiO₂-CO₃²⁻ was believed to be mainly originated from the occurred long-range charge separation by the visible-light-induced, vectorial interfacial electron transfer in the way of S–Au → CB-(TiO₂) → L-Au. It is known that the generated H₂O₂ on a pure TiO₂ photocatalyst is easily degraded since it can be reduced by the peroxide species (Ti_s-OOH) on the TiO₂ surface. However, the modification of CO₃²⁻ onto the TiO₂ surface efficiently inhibits such reductive degradation of H₂O₂. The concentration of generated H₂O₂ on this BM-Au/TiO₂-CO₃²⁻ thus reached about 1 mM. The mechanism of the H₂O₂ generation on the BM-Au/TiO₂-CO₃²⁻ photocatalyst includes plasmonic effect and photocatalytic O₂ reduction (Fig. 7d). It has been further reported that the photocatalytic kinetics for the H₂O₂ formation is dependent on temperature and pH value for the deposition of Au NPs on anatase TiO₂ when a heating temperature-varied deposition–precipitation technique was applied [136]. The amount of generated H₂O₂ under the irradiation was increased with a decrease of the temperature and pH in the reaction system. The highest H₂O₂ concentration produced on the Au/anatase TiO₂ photocatalyst was about 17 mM at the temperature of 5 °C in the solution with a pH value of 2. In conclusion, these introduced metal Au NPs on the TiO₂ surface prohibit the H₂O₂ decomposition that is induced by the Ti–OH species. Meanwhile, they capture photogenerated electrons in the CB of TiO₂, leading to retarded recombination of photogenerated carriers. The output of the H₂O₂ production on these photocatalysts was obviously increased. The H₂O₂ concentration even reached the millimolar level. Nevertheless, the generated H₂O₂ molecules are inclined to be adsorbed on Au NPs due to their strong interactions. Subsequently, spontaneous decomposition of H₂O₂ occurs: H₂O₂ + e⁻ → ^·OH + OH⁻ [137, 138].

Fig. 7

a Mechanism of the H₂O₂ formation over the Au/TiO₂ photocatalyst. Reproduced with permission from Ref. [133]. Copyright 2010 American Chemical Society. b TEM image of the BM-Au/TiO₂ photocatalyst. c Photocatalytic H₂O₂ production on various photocatalysts. d Proposed mechanism of the BM-Au/TiO₂-CO₃²⁻ plasmonic catalyst in the photocatalytic H₂O₂ generation. Reproduced with permission from Ref. [135]. Copyright 2016 Wiley–VCH Verlag GmbH & Co. KGaA, Weinheim. e HRTEM image of the Au_0.1Ag_0.4/TiO₂ photocatalyst. f Schematic energy-band and mechanism of the H₂O₂ production on the AuAg/TiO₂ photocatalyst. Reproduced with permission from Ref. [139]. Copyright 2012 American Chemical Society

To promote the H₂O₂ formation and simultaneously inhibit its decomposition on the TiO₂ photocatalysts, the NPs from bimetal alloys have been deposited on the TiO₂ surface to enhance the photocatalytic activity. For example, a bimetal alloy catalyst (AuAg/TiO₂) has been designed and applied for photocatalytic H₂O₂ generation (Fig. 7e) [139]. The H₂O₂ concentration produced by the Au_0.1Ag_0.4/TiO₂ photocatalyst was as high as 3.4 mM, approximately 7 times higher than that produced on a pure TiO₂ photocatalyst. This excellent photocatalytic performance has been attributed to the following two aspects (Fig. 7f). Firstly, the work function of the alloy AuAg is located between Au and Ag. The alloy/TiO₂ photocatalyst produces a barrier, of which energy is larger than that of the Ag/TiO₂ photocatalyst but smaller than that of Au/TiO₂ photocatalyst. Effective separation of hole-electron pairs is thus possible. Secondly, electron transfer from Ag to Au takes place, because Au atoms have a higher electronegativity than Ag atoms. In other words, the electron density of Au is increased, suppressing the H₂O₂ adsorption on Au atoms. Such statements were further proved using the DFT simulations. The affinity between the H₂O₂ and Au atoms was confirmed to be weakened via the alloy effect. It was taken place that H₂O₂ is inclined to be adsorbed on Ag atoms, thus inhibiting the H₂O₂ decomposition. In another case, the midgap state in the Ag₂Au₂@TiO₂ (101) photocatalyst was reveal to own a suitable position for the H₂O₂ production via a photocatalytic reaction, as demonstrated by the DFT analysis. More importantly, the photoresponse of these bimetal alloy photocatalytic systems is rapidly promoted in visible and infrared light region, improving the H₂O₂ yield [140].

3.2.2 Modifying Graphene Quantum Dots

Graphene quantum dots (GQDs) feature advantages of high chemical stability, good biocompatibility, large surface area and high extinction efficiencies (that are derived from the atomically thin sp² carbon structure of graphene). GQDs exhibit unique luminescent characteristic. Therefore, GQDs have attracted extensive attention in various photocatalytic applications. Especially, the sensitization of GQDs with other wide band gap photocatalysts (e.g., TiO₂) can extend light-adsorption range from ultraviolet to the visible region [141, 142]. For example, sulfur and nitrogen co-doped graphene quantum dots (SN-GQDs) have been combined with the TiO₂ photocatalyst (SN-GQD/TiO₂) and further applied for photocatalytic H₂O₂ production [143]. As confirmed from related UV–vis diffuse reflectance spectra (DRS), SN-GQD/TiO₂ exhibited obviously enhanced adsorption in visible light region when compared with the pure TiO₂ photocatalyst. On its surface, more than 82.8 μM H₂O₂ was produced under visible light irradiation only for 90 min. Such a yield was about 5.3 and 3.1 times larger than that generated on the GQD/TiO₂ photocatalyst and the N-GQD/TiO₂ suspensions under the same conditions, respectively (Fig. 8a). Theoretical calculation and free energy diagram analysis showed that the H₂O₂ generation on the SN-GQD/TiO₂ photocatalyst followed a proton-coupled electron transfer (PCET) mechanism. Namely, high selectivity of the H₂O₂ generation on the SN-GQD/TiO₂ photocatalyst was realized via a 2e⁻ ORR pathway (Fig. 8b, c). A Nafion layer was further introduced into the SN-GQD/TiO₂ (Nf-SNG/TiO₂) to facilitate the H₂O₂ generation [144]. The GQDs coupled with a Nafion layer were found not only to improve the visible light adsorption, but also to significantly hinder the H₂O₂ decomposition (Fig. 8d). The yield of photocatalytic H₂O₂ production on the Nf-SNG/TiO₂ photocatalyst with 3.5% Nafion content was 141 μM after 120 min visible light irradiation than that obtained on the counterparts (Fig. 8e). Related photocatalytic reaction mechanism of the H₂O₂ formation on Nf-SNG/TiO₂ was testified to follow a two-electron-dominated ORR pathway (Fig. 8f). It was impacted by GQDs sensitization, different from photocatalytic H₂O₂ production on the pure TiO₂ photocatalyst.

Fig. 8

a Photocatalytic H₂O₂ generation on the catalysts. b Proposed mechanism of photocatalytic H₂O₂ formation over the SN-GQD/TiO₂ photocatalyst. c Free energy diagrams of reaction pathways. Reproduced with permission from Ref. [143]. Copyright 2018 Elsevier. d Photocatalytic decomposition and e production of H₂O₂ on different catalysts under visible light. f Photocatalytic reaction mechanism for H₂O₂ formation on the Nf-SNG/TiO₂ photocatalyst. Reproduced with permission from Ref. [144]. Copyright 2019 American Chemical Society

3.2.3 Complexing with Cation and Anion Surfaces

The surface states are important during photocatalytic H₂O₂ generation, since photocatalytic reactions are surface-controlled processes. Therefore, various surface modification approaches have been developed to modulate photocatalytic activities of TiO₂ photocatalysts. Among them, complexing cations or anions onto the TiO₂ surface has been wildly accepted as the simplest one to effectively boost the photocatalytic H₂O₂ production. For example, the inner spherical surface of the metal cations can modulate interfacial electron transfer via suppressing the surface trapping sites for photogenerated carriers during the photocatalytic process. Taking the modification of TiO₂ by Cu²⁺ ion as an example, it has been revealed that complexing with a small amount of Cu²⁺ onto the TiO₂ surface significantly promoted the photocatalytic activity of H₂O₂ production [145]. A 20-fold increment in the amount of generated H₂O₂ was achieved under UV irradiation in comparison with that on the TiO₂ photocatalyst without Cu²⁺ modification. When the concentration of Cu²⁺ was in the range of 30–40 μM, the performance for photocatalytic H₂O₂ generation remained stable. Such enhanced activity of photocatalytic H₂O₂ production was proposed to be dependent on the optimization of the TiO₂ surface state through Cu²⁺ modification, eventually promoting the H₂O₂ formation on the TiO₂ surface via a 2e⁻ ORR pathway. In another case, Zn²⁺ ions were complexed onto the TiO₂ photocatalyst [146]. This Zn²⁺ modified TiO₂ photocatalyst exhibited obviously improved performance of photocatalytic H₂O₂ generation. Since Zn²⁺ ions block the ≡Ti–OH sites, the complexation of peroxide/superoxide species on the TiO₂ surface is thus much limited, resulting in reduced H₂O₂ decomposition.

In addition to cations, complexing with anions onto the TiO₂ photocatalyst has been also applied for the photocatalytic H₂O₂ production. For instance, the fluorinated TiO₂ photocatalyst exhibited a boosted reaction toward photocatalytic H₂O₂ generation. Stemming from surface state modulation via F⁻ modification, this fluorinated TiO₂ photocatalyst suppresses the Ti-OOH formation. It produced H₂O₂ with a concentration of 1.3 mM under UV light irradiation, one of the best activities among the reported TiO₂ photocatalysts for the H₂O₂ generation [147].

3.3 BiVO₄ Photocatalysts

In addition to TiO₂ photocatalysts, some more complex inorganic oxides have also been employed for photocatalytic H₂O₂ generation. Among them, bismuth vanadate (BiVO₄) is one classic representative of bimetallic oxide photocatalyst [156,157,158,159]. Again, it is well known a pure TiO₂ photocatalyst has a wide bandgap and thus can be excited only under UV light irradiation, resulting in insufficient activity for photocatalytic H₂O₂ generation in sunlight [160, 161]. Distinguished from a TiO₂ photocatalyst, BiVO₄ has its appropriate band structure and is thus active in the visible light region. In other words, it can promote the H₂O₂ production via a 2e⁻ ORR pathway under the visible-light irradiation [162]. Unfortunately, pure BiVO₄ is not conducive to the efficient photocatalytic H₂O₂ production, due to its lack of active sites for the 2e⁻ ORR. In this context, nano-Au cocatalyst was introduced onto the BiVO₄ surface. Stemming from the strong interaction between Au and BiVO₄, the activation of the special d-band electrons promoted the selectivity of the 2e⁻ ORR pathway to efficiently generate H₂O₂ under visible light irradiation (Fig. 9a) [163]. This is because of the edge in the CB of BiVO₄ was disclosed to be 0.02 V, more positive than that (−0.13 V) of one-electron reduction of O₂ and more negative than that (0.68 V) of two-electron reduction of O₂ (Fig. 9b). The inhibition of single electron reduction over the Au/BiVO₄ surface was further identified by electron spin resonance (ESR) spectra, where Au_0.2/BiVO₄ exhibits almost no signal, while Au_0.2/TiO₂ displays distinctive signals (Fig. 9c). The latter was ascribed to the DMPO − ·OOH spin adduct formation.

Fig. 9

a TEM image of an Au_0.2/BiVO₄ photocatalyst. b Energy diagram of the Au/BiVO₄ photocatalyst and reduction potential of O₂. c ESR spectra of an Au_0.2/BiVO₄ photocatalyst and an Au_0.2/TiO₂ photocatalyst in an EtOH/water/O₂ system with DMPO. Reproduced with permission from Ref. [163]. Copyright 2016 American Chemical Society. d Energy diagram of a Cu@Au/BiVO₄ photocatalyst. e Mechanism of H₂O₂ generation on a Cu@Au/BiVO₄ photocatalyst. Reproduced with permission from Ref. [166]. Copyright 2021 American Chemical Society. f Energy diagram and the reaction mechanism of a Cu(hfacac)₂/ms-BiVO₄ photocatalyst. Reproduced with permission from Ref. [167]. Copyright 2020 American Chemical Society

Nevertheless, the activity of the BiVO₄ photocatalyst toward photocatalytic H₂O₂ production still remains unsatisfactory. This is mainly attributed to the formed built-in field between Au and BiVO₄, which inhibits the transfer of photogenerated electrons as well as accumulates negative charges of Au to deteriorate the 2e⁻ ORR pathway [164, 165]. In order to overcome this impediment, one Cu@Au/BiVO₄ photocatalyst was designed and obtained using combined photodeposition and galvanic displacement methods [166]. The Cu species was found to facilitate the transfer of photogenerated electrons from BiVO₄ to Au. The accumulation of negative charges on Au was then reduced, resulting in the enhanced activity for the photocatalytic H₂O₂ production (Fig. 9d, e). Specifically, the ohmic contact was expected to be produced between Cu and BiVO₄ since their work functions (Φ) are different: ${\Phi }_{Cu}$= 4.7 eV and ${\Phi }_{{BiVO}_{4}}$= 4.9 eV. Transferring of the photogenerated electrons is thus beneficial. The photogenerated electrons that are transferred from Cu to Au eventually boost the two-electron O₂ reduction pathway, namely the H₂O₂ formation. On the other hand, photogenerated charges have been accumulated on the nano-Au cocatalyst in the Cu@Au/BiVO₄ photocatalyst, leading to the generation of stronger adsorption of O₂ and HOO^* on the Au surface of Cu@Au/BiVO₄. Consequently, O₂ reduction to form H₂O₂ is accelerated. For a bis(hexafluoroacetylacetonato) Cu(II) adsorbed monoclinic scheelite (ms)-BiVO₄ photocatalyst (Cu-(hfacac)₂/ms-BiVO₄), it exhibited an outstanding activity of photocatalytic H₂O₂ generation [167]. An external quantum yield of 0.47% was reported under visible light irradiation (λ_ex = 470 nm). Such performance was attributed to the enhanced charge separation by the interfacial electron transfer from ms-BiVO₄ to the surface complex and the O₂-enriching effect near the surface of ms-BiVO₄, as well as outstanding electrocatalysis for a 2e⁻ ORR pathway (Fig. 9f).

3.4 CdS Photocatalysts

CdS has exhibited promising photocatalytic activities in various reactions under visible light irradiation, since the potential of its VB is positive enough to drive water oxidation and meanwhile the potential of its CB is more negative to promote O₂ reduction. CdS thus has great potential to be applied as a catalyst for photocatalytic H₂O₂ production [168,169,170]. Unfortunately, CdS exhibits relatively low photocatalytic capability toward H₂O₂ generation, originating from its weak adsorption capacity for reactants and its poor photostability. More seriously, its easy aggregation can cause the severe recombination of photogenerated carriers [171].

To improve the performance of CdS for photocatalytic H₂O₂ production, many efforts have been made and several strategies have been proposed. One strategy is to complex organic polymers onto CdS. For example, the hybrids of CdS and reduced graphene oxide (CdS-RGO) prepared through a hydrothermal process significantly boosted the kinetics of photocatalytic H₂O₂ generation when compared with pure CdS photocatalyst (Fig. 10a) [172]. The CdS-RGO photocatalyst with a 20 wt% RGO content (CdS-G2) produced the H₂O₂ concentration of as high as 128 μM under sunlight irradiation for 12 h. Such enhancement was mainly attributed to accelerated separation of photogenerated carriers due to the enhanced electron transfer from the photoexcited CdS to RGO, the increment of visible light absorption and more active reaction sites (as demonstrated by their PL spectra (Fig. 10b)). The mechanism of photocatalytic H₂O₂ production on the CdS-RGO photocatalyst was also discussed (Fig. 10c). In the first step, electron–hole pairs are generated on the CdS-RGO photocatalyst upon its excitation by light irradiation. Water molecules are then oxidized by photogenerated holes, leading to the production of H₂O₂ and protons. Simultaneously, O₂ is reduced to form H₂O₂, following a 2e⁻ ORR pathway (step 1–4). These photogenerated electrons have been demonstrated to play a vital role in the process of photocatalytic H₂O₂ generation. The H₂O₂ yield from a 2e⁻ ORR pathway is the decisive factor when compared with that from the 2e⁻ WOR pathway in the whole photocatalytic process.

Fig. 10

a H₂O₂ concentration produced on different catalysts. b PL spectra of the CdS and CdS-G2 photocatalysts. c Mechanism for H₂O₂ production on the CdS-G hybrid photocatalyst under sunlight illumination. Reproduced with permission from Ref. [172]. Copyright 2016 Elsevier. Schematic illustration of H₂O₂ formation over a CdS photocatalyst and a CdS@S-doped carbon photocatalyst. e H₂O₂ concentration for 24 h with 20% of 2-propanol. Reproduced with permission from Ref. [174]. Copyright 2020 Elsevier

Noble metals have also been complexed onto CdS to improve the photocatalytic performance of CdS catalysts. For example, the CdS-Pt and CdS-Au nanorods exhibited enhanced photocatalytic activity to produce H₂O₂ when compared with a pure CdS photocatalyst [173]. The Au tip with a smaller size was found to be more efficient as reactive site to form H₂O₂ than a Pt tip. This was attributed to different surface reactivity and selectivity related to the metal tip composition. On these CdS-metal photocatalysts, the H₂O₂ production through ORR pathway has been confirmed, instead of the WOR pathway. A CdS/sulfur-doped carbon nanocomposite, synthesized by thermal annealing of Cd(EDDA) MOF, exhibited outstanding activity toward photocatalytic H₂O₂ production [174]. This is due to the presence of sulfur-doped carbon, which can effectively hinder the H₂O₂ decomposition, as experimentally and theoretically testified (Fig. 10d). The H₂O₂ concentration generated on a nanocomposite based photocatalyst (M-CdS-10 h) reached 17.1 mM under visible light irradiation when 2-propanol was used as a sacrificial agent (Fig. 10e).

3.5 Metal–organic Frameworks Photocatalysts

Metal–organic frameworks (MOFs) own unique porous structures and strong metal–ligand interactions [175]. Similar to semiconductors, MOFs can produce electron–hole pairs upon light irradiation. They thus exhibit great potential in photocatalysis [176,177,178,179]. The Ni/MIL-125-NH₂, obtained via depositing NiO on the MIL-125-NH₂ was the first MOF photocatalyst used in the photocatalytic H₂O₂ synthesis [180]. The mechanism of the H₂O₂ production on the Ni/MIL-125-NH₂ (Fig. 11a) was described as following. Nano NiO effectively inhibits the H₂O₂ decomposition through a 2e⁻ ORR pathway under visible light irradiation when benzylalcohol is accompanied as an electron donor. Specifically, MIL-125-NH₂ produces Ti₈O₈(OH)₄^·− and a hole upon its excitation by light irradiation. Once the hole is trapped by triethanolamine (TEOA), O₂ is reduced to O₂^·− by Ti₈O₈(OH)₄^·−, followed by rapid H₂O₂ formation via the O₂^·− disproportionation reaction in the presence of NiO. Unfortunately, in this case the product was a mixture of H₂O₂ and benzaldehyde dissolved in acetonitrile. In other words, further energy-consumption for product separation and purification is required. Later, a two-phase system containing benzylalcohol/water (BA/water), was employed to separate these products [181]. The hydrophilic MIL-125-NH₂ was conversed to hydrophobic MIL-125-Rn (n = 1, 4 and 7) through the growth of the alkyl chains on the catalyst surface, as clarified using water adsorption isotherms and water contact angles measurements (Fig. 11b). A pristine MIL-125-NH₂ was proved to exist in the water phase during the photocatalytic H₂O₂ process, while alkylated MIL-125-Rn is located in the BA phase (Fig. 11c). During the photocatalytic H₂O₂ process (Fig. 11d), O₂ is firstly reduced to O₂^·− on the hydrophobic MOFs in the BA phase. Subsequently, O₂^·− is transferred to the water phase where H₂O₂ is rapidly produced through a disproportionation reaction in the presence of H⁺ or Na⁺. Although this two-phase system inhibited the further reaction of MOFs with H₂O₂, its photocatalytic activity of the two-phase system was not satisfying. This is because the grafting of the alkyl chains blocks the pores of the MIL-125-R7, thus greatly reducing its photocatalytic activity. In this regard, a hydrophobic MOF, namely OPA/MIL-125-NH_2, was developed. In this case, MIL-125-NH₂ retained most of its pores using octadecylphosphonic acid (OPA) treatment [182]. Its surface area was 1242 m² g⁻¹, comparable to that (1500 m² g⁻¹) of MIL-125-NH₂. Under visible light irradiation for 3 h, the concentration of generated H₂O₂ on this OPA/MIL-125-NH₂ photocatalyst was approximately 3 times larger than that of MIL-125-R7. The enhanced activity was attributed to the rapid diffusion of O₂^·− through the unblocked pores of the OPA/MIL-125-NH₂ photocatalyst, which prevents the H₂O₂ decomposition (Fig. 11e). A hydrophobic titanium doped zirconium-based MOF (OPA/Zr_100-xTi_x-MOF) also exhibited a high rate of H₂O₂ production (9.7 mmol L⁻¹ h⁻¹), where the Ti species played a role in effectively promoting electron transfer from photoexcited linkers of the MOF to Ti, inhibiting the recombination of photogenerated electron–hole pairs in the hydrophobic MOF matrix (Fig. 11f) [183].

Fig. 11

a Photocatalytic mechanism of H₂O₂ formation on the MIL-125-NH₂ photocatalyst. Reproduced with permission from Ref. [180]. Copyright 2018 ROYAL SOCIETY OF CHEMISTRY. b Water adsorption isotherms for various catalysts. c Digital photographs utilizing the two-phase system. d Mechanism illustration of H₂O₂ generation by the two-phase system. Reproduced with permission from Ref. [181]. Copyright 2019 Wiley–VCH Verlag GmbH & Co. KGaA, Weinheim. e Photocatalytic processes of MIL-125-R7 and OPA/MIL-125-NH₂ photocatalysts. Reproduced with permission from Ref. [182]. Copyright 2019 ROYAL SOCIETY OF CHEMISTRY. f Mechanism of photocatalytic H₂O₂ production using the Zr₁₀₀-MOF and OPA/Zr_92.5Ti_7.5-MOF photocatalyst. Reproduced with permission from Ref. [183]. Copyright 2020 ROYAL SOCIETY OF CHEMISTRY. g Photocatalytic O₂ reduction to H₂O₂ over a NH₂-MIL-125@ZnS heterojunction. Reproduced with permission from Ref. [187]. Copyright 2021 Wiley–VCH GmbH

The construction of the heterojunction by combining MOFs with other semiconductors also effectively strengthens H₂O₂ production since the separation of photogenerated electron–hole pairs can be promoted [184,185,186]. For example, on the MOF@MS heterostructure (NH₂-MIL-125@ZnS) the photocatalytic synthesis of H₂O₂ was reported via a 2e⁻ ORR pathway (Fig. 11g) [187]. This heterostructure presents a spatially separated architecture, where ZnS nanocages are selectively adhered on the four corners of a cake-like NH₂-MIL-125 body. The coverage of heterojunction was controlled by altering the diameter of ZIF-8/ZnS. In this way, the regulation of the absorption of light, the generation of electron–hole pairs, the charge separation and accessibility were doable, finally leading to adjusted performance of photocatalytic H₂O₂ production. During the photocatalytic process, photogenerated electrons transfer from ZnS to NH₂-MIL-125, while holes migrate from NH₂-MIL-125 to ZnS, resulting in excellent performance. When the heterojunction coverage is ≈45.1%, the NH₂-MIL-125@ZnS photocatalyst displays a H₂O₂ yield of as high as 120 mM g⁻¹ h⁻¹.

4 Electrocatalytic H₂O₂ Synthesis

4.1 Electrocatalytic H₂O₂ Synthesis through a 2e⁻ ORR Pathway

2e⁻ ORR has been considered as one of the most vital routes in electrocatalytic H₂O₂ synthesis [188, 189]. However, its selectivity is limited by the competition from the 4e⁻ reduction pathway, namely the reduction of O₂ to generate H₂O. The fact behind is that O₂ has a more negative standard potential to form H₂O than that for the H₂O₂ generation [52, 190, 191]. Further electro-reduction of H₂O₂ to H₂O is thus thermodynamically favorable, which severely hinders the yield of H₂O₂ from a 2e⁻ ORR pathway. In addition, the spontaneous H₂O₂ disproportionation to produce H₂O and O₂ prohibits the 2e⁻ ORR pathway, lowering the selectivity of H₂O₂ production. Therefore, it is prerequisite to design electrocatalysts to promote the preferential H₂O₂ formation as well as to facilitate rapid diffusion of H₂O₂ away from the reactive interface. The reported electrocatalysts using the 2e⁻ ORR pathway for the H₂O₂ production can be divided into two categories: noble-metal-based and carbon-based electrocatalysts. In the following session, these reported electrocatalysts are systematically classified and their performance toward the H₂O₂ is detailed (Table 3).

Table 3 Summary of electrocatalytic H₂O₂ production via a 2e⁻ ORR pathway

4.1.1 Noble Metal-based Electrocatalysts

According to Sabatier’s principle, the binding energy between an OOH^* radical and an electrocatalyst should be neither too strong nor too weak for efficient H₂O₂ production, since an OOH^* radical is the only intermediate in the process of a 2e⁻ ORR pathway [41]. Meanwhile, the integrity of an O–O bond must be well maintained during the conversion process since the splitting of O–O bond is conductive to form water. It has to highlight here that the manners of oxygen adsorption on a catalyst mainly follow three modes (Fig. 12a), namely Pauling model (“end-on”), Griffiths model and Bridge model (“side-on”) [55, 192]. Only Pauling model is beneficial for the H₂O₂ generation via two-electron reduction of O₂. This is due to its low influence on the O–O bond splitting. Differently, other two models induce the O–O bond splitting and facilitate the 4e⁻ ORR pathway. Note that it is more thermodynamically inclined to the pathway of the 4e⁻ ORR to form H₂O in comparison with H₂O₂ generation via the 2e⁻ ORR pathway. To address this issue, the key is to design and synthesize advanced catalysts with high activity for O₂ reduction, while well remaining O–O bond during the 2e⁻ ORR process. In the next section, the electrocatalysts on the selective reduction of O₂ to H₂O₂ are reviewed in combination with some typical examples.

Fig. 12

a Three typical modes for O₂ adsorption. Reproduced with permission from Ref. [192]. Copyright 2021 Elsevier. b Sabatier volcano plots for electrochemical ORR for pure metal slabs and M–N/C. Reproduced with permission from Ref. [193]. Copyright 2018 American Chemical Society. c Proposed ORR pathways on the pristine Pt/C and carbon-coated Pt electrocatalysts. Reproduced with permission from Ref. [200]. Copyright 2014 American Chemical Society. d H₂O₂ selectivity of the Au_1‑xPd_x electrocatalyst with different palladium contents at potentials of 0, -0.1 and -0.2 V. Reproduced with permission from Ref. [208]. Copyright 2011 American Chemical Society

Single noble metal materials have been applied for electrocatalytic H₂O₂ synthesis. For example, DFT calculations have been firstly used to reveal the relationship between the ORR activity and various metallic catalysts. The Sabatier volcano plots were then established (Fig. 12b), namely the electrocatalytic ORR activity of different metals and metal-nitrogen/carbon (M–N/C) [193]. Taking the volcano curves of single metals (the left-hand side of the volcano diagram) as the example, the limiting potential of a 4e⁻ ORR process is more positive than that of a 2e⁻ ORR process, thermodynamically indicating that the 4e⁻ reduction reaction is more likely to occur to obtain H₂O than the 2e⁻ reduction reaction to generate H₂O₂. While on the right-hand side of the volcano plots, the limiting potential of 2e⁻ and 4e⁻ ORR processes are located at the same position, which illustrates that they simultaneously take place to produce H₂O₂ and H₂O, respectively. In more detail, as the cases that the strong interaction occurs between metals and O₂ molecules, the selectivity of 2e⁻ ORR can be improved once the metal surface is modified, or its structure is regulated. Taking Pt metal as an example, O₂ is adsorbed on the Pt surface mainly through a “side-on” mechanism, instead of the “end-on” mechanism. The H₂O₂ production on Pt catalysts is thus not conducive to be promoted [194]. In literature, the effects of the size and morphology of metal catalysts have been extensively investigated to explore the activity for electrocatalytic H₂O₂ production, namely via altering the O₂ adsorption mode on the Pt electrocatalysts [195,196,197]. Note that the choice of suitable supports also modulates the selectivity of the Pt electrocatalysts toward electrocatalytic H₂O₂ production [198, 199]. For example, an amorphous carbon layer coated Pt catalyst was prepared by use of a chemical vapor deposition method with acetylene as the precursor [200]. This amorphous carbon layer presumably produces steric hindrance on the Pt surface, which induces O₂ to be adsorbed through the end-on configuration, instead of side on configuration over the pristine Pt/C electrocatalyst (Fig. 12c). The 2e⁻ ORR pathway happened dominantly on the C(Pt)/C-3 h and C(Pt)/C-4 h electrocatalysts. Their H₂O₂ selectivity reached 41% at a potential of 0.1 V_RHE. Similarly, sulfur-doped carbon using zeolite-templated was used to support catalyst, where this carbon material supported and dispersed 5 wt% Pt atoms [197]. This Pt atomic electrocatalyst followed a 2e⁻ ORR pathway. Its H₂O₂ selectivity was as high as 96%. Modified Pd species have been also designed and obtained to assess selective ORR [201, 202]. For instance, using a nano-Pd electrocatalyst, obtained through in-situ electrochemical deposition from Pd ions, high selectivity for the 2e⁻ ORR pathway to form H₂O₂ was achieved. Even when Pd is derived from the lowest precursor concentration, related nano-Pd electrocatalyst exhibited high selectivity, which was surpassed 95% [203].

In contrast, the noble metals (e.g., Au) that possess weak adsorption capability toward O₂ molecules have been applied to reduce O₂ for efficient, selective and stable H₂O₂ production [204,205,206]. Unfortunately, the adsorption of O₂ molecules on the crystal plane of noble metals is different. Namely, the ORR pathway on different metal facets is different. For example, the too weak bond between an Au (111) surface and O₂ molecules makes it difficult to form the OOH^* intermediate [207]. The H₂O₂ production on the Au (111) surface via the 2e⁻ ORR pathway exhibits a low efficiency and high energy consumption is associated. In this regard, the crystal plane of noble metals is applied as the reactive interface, since its binding energy with O₂ molecule in a high index crystal plane is stronger than that occurred in an (111) facet. For example, an Au (211) facet effectively promotes the activity of a 2e⁻ ORR pathway for the H₂O₂ production.

Although the H₂O₂ selectivity of single metal electrocatalysts can be effectively adjusted via various means as discussed above, simultaneous improvement of the activity and selectivity for the 2e⁻ ORR pathway on single metal electrocatalysts is still challenging. This is mainly originated from the fact those with weaker interactions with O₂ often need high overpotential to boost the 2e⁻ ORR, while those owning stronger interactions with O₂ are prone to bring the direct the 4e⁻ ORR. Combining such metals (namely those weakly interact with O₂ and those strongly interact with O₂) to form metal alloys is thus a potential strategy to facilitate both activity and selectivity of metal electrocatalysts toward the H₂O₂ generation. For example, the Au_1-xPd_x nanoalloys supported on Vulcan XC-72 exhibited the superior selectivity toward the H₂O₂ production (nearly 95%) when the Pd content in the alloy arrived at 8%, surpassing that using single Au and Pd metal electrocatalysts (Fig. 12d) [208]. The Pd monomer surrounded by Au atoms on the Au–Pd nanoalloy surface was suggested to promote the H₂O₂ formation via an “end-on” O₂ adsorption model. When the percentage of Pd was increased to 15%, the selectivity of the H₂O₂ generation was decreased. This is because the presence of continuous Pd atoms (e.g., two adjacent Pd atoms) altered the O₂ adsorption model and facilitated the H₂O formation. In a theoretical work, the tactics was provided based on DFT calculations to screen for metal alloy electrocatalysts that feature excellent activity toward electrocatalytic H₂O₂ production [5]. These metal alloys are prepared by surrounding individual elements (e.g., Pt, Ir, Pd, Rh and Ru) that are active for the ORR by inert metallic elements (e.g., Hg and Au). Since the active elements exhibit strong adsorption toward O₂ molecule, these metal alloys can effectively reduce O₂ to OOH^*. On the other side, they do not break the O–O bond, because of their difficulty in dissociating OOH^* in combination with the neighboring inert sites. In short, the activity and selectivity of metal alloy catalysts toward the 2e⁻ ORR can be regulated by using the advantages of the isolated active sites. Theoretical models have even predicted that the isolated active site of the Pt-Hg alloy can effectively activate O₂ molecules and boost the H₂O₂ synthesis. Experimentally, several Hg-based bimetallic alloys have shown high activity and selectivity for the H₂O₂ production [209]. Among them, the Pd-Hg alloy exhibited the highest current for the H₂O₂ production at the same overpotential, as predicted from the DFT calculations. In more detail, the activities of Pt-Hg, Ag and Ag-Hg alloys are one order of magnitude higher than that of individual Au electrocatalyst, while the activity of a Pd-Hg alloy is two orders of magnitude higher than that of individual Au electrocatalyst. The AuPd alloy nanoparticles have been used to disclose the relationship between the activity of electrocatalytic H₂O₂ production and molar ratio of Au to Pd. Modulating the composition of alloy catalysts is thus recognized as a new way to promote the 2e⁻ ORR for electrocatalytic H₂O₂ production [210].

4.1.2 Carbon-based Electrocatalysts

It is clear that the noble metals and their alloys feature high activity and selectivity for electrocatalytic H₂O₂ synthesis via the 2e⁻ ORR pathway. Unfortunately, they are expensive and scarce, hindering their large-scale applications. Many advanced electrocatalysts with reduced costs have been thus explored for electrocatalytic H₂O₂ synthesis. Carbon-based electrocatalysts have been considered to be one of the most promising substitutes of noble metal-based catalysts. This is due to their unique surface and structural properties, abundant availability and low cost for their synthesis [211,212,213,214]. Note that the pristine carbon materials only exhibit low catalytic activity for the 2e⁻ ORR, thus the reconstruction of carbon materials or their surface modification (Fig. 13) have been developed to optimize their electronic properties for electrocatalytic H₂O₂ generation [18]. The widely reported approaches include manufacturing porous structures and defects, heteroatom doping and surface (oxygen) functionalization are provided.

Fig. 13

Schematic demonstration of tuning electronic structures of carbon electrocatalysts for electrocatalytic H₂O₂ production via the 2e⁻ ORR pathway. Reproduced with permission from Ref. [18]. Copyright 2020 American Chemical Society

Among diverse carbon materials, porous and defective carbon materials are highly suitable for the 2e⁻ ORR. This is due to their unique advantages, such as their high surface areas, tunable porosities and high electronic conductivities. It has been widely recognized that carbon electrocatalysts with a porous structure can provide abundant active sites and shorten diffusion distance for the reactive species, producing positive contributions for their electrocatalytic H₂O₂ generation. For example, various hierarchically porous carbon (HPC), produced by hydrothermal treatment and subsequent carbonization, have been applied for selective H₂O₂ formation (Fig. 14a) [215]. The HPC-H24 (hydrothermal treatment for 24 h and carbonization under H₂) exhibited the most positive peak potential and the highest selectivity for the 2e⁻ ORR, namely the best performance of selective H₂O₂ formation. The remarkable electrocatalytic performance of HPC-H24 was assumed to be originated from high contents of active sites (e.g., sp³-C bonds, vacancy defects and edge defects), more exposed and accessible active sites and fast mass transport of H₂O₂ inside a porous structure. On the contrary, it was acknowledged that edge-rich graphene presented the typical 4e⁻ ORR mechanism [216]. Later, two different types of porous carbon, namely microporous carbon (denoted as MicroC) and mesoporous carbon (denoted as MesoC) have been developed toward the H₂O₂ production in alkaline conditions (Fig. 14b) [217]. From the broad π^* resonance of near-edge X-ray absorption fine structure (NEXAFS) spectra, the conclusion that both MicroC and MesoC have a larger number of sp² carbon defective sites than HOPG has been drawn. Electrochemical experiments suggested that both MicroC and MesoC had superior activity and selectivity for the 2e⁻ ORR, where the onset potential was very close to the thermodynamic equilibrium potential (0.7 V_RHE), and the selectivity of the H₂O₂ production even outperformed 70% in some cases (Fig. 14c). In addition, DFT calculations have illustrated the sp²-type defects are the active sites for O₂ reduction, which is beneficial for the H₂O₂ generation. Similar optimization of defects and pore size in carbon materials have been reported for electrocatalytic H₂O₂ production. For example, the graphitic carbon with rich edge sites has been applied as an electrocatalyst to promote the H₂O₂ synthesis. Its activity was about 28 times higher than that of a basal plane-rich carbon nanotube (CNT) [218]. This enhanced activity was explained by the fact that graphitic ordered mesoporous carbon (GOMC) has a larger portion of defective carbon and reaction sites (sp²-C) when compared with CNT. Such a statement was solidly supported from the experimental results by means of Raman spectroscopy (Fig. 14d) and XPS (Fig. 14e) characterization. Compared with CNTs, the GOMC showed the H₂O₂ selectivity higher than 90% via the 2e⁻ ORR pathway and excellent stability (e.g., 90% of the initial current is retained) (Fig. 14f).

Fig. 14

a Schematic illustration of the synthetic route to HPC. Reproduced with permission from Ref. [215]. Copyright 2015 Wiley–VCH Verlag GmbH & Co. KGaA, Weinheim. b NEXAFS spectra of the C-K edge in HOPG, MesoC and MicroC. c H₂O₂ selectivity of the MesoC and MicroC electrocatalysts. Reproduced with permission from Ref. [217]. Copyright 2018 American Chemical Society. d Raman spectra, e C 1s XPS spectra and f H₂O₂ selectivity of GOMC and CNT. Reproduced with permission from Ref. [218]. Copyright 2019 Wiley–VCH Verlag GmbH & Co. KGaA, Weinheim

It has been reported that oxygen-containing groups play an important role on the activity and selectivity of carbon catalysts toward the 2e⁻ ORR [219,220,221,222,223,224,225]. For instance, oxygen-doped microporous polypyrene carbons (OMPC) exhibited the enhanced activity and selectivity for electrocatalytic H₂O₂ generation [220]. In basic electrolyte, the OMPC4 (oxygen content is 6.52 at%) showed obviously stronger ability to H₂O₂ production from the I_ring than MPC, as well as the higher limited oxygen reduction current (I_disk) (Fig. 15a). Meanwhile, the selectivity of the H₂O₂ production on OMPC-4 was remarkably increased (e.g., from 40 to 87% at 0.42 V_RHE) and the calculated electron transfer number (n) was close to 2 through oxygen doping on OMPC (Fig. 15b). In addition, the selectivity of MPC with the different oxygen content is displayed in Fig. 15c, where OMPC-4 exhibited the highest value. In another case, oxygenated carbon (namely activated carbon black, denoted as aCB) electrocatalyst showed higher ORR activity meaning the stronger capacity of H₂O₂ generation than CB (Fig. 15d) [222]. The selectivity above 94% was obtained on the aCB catalyst in the potential range from 0.4 to 0.7 V as depicted in Fig. 15e. The DFT calculations further disclosed the activity difference of these oxygen-containing groups toward the H₂O₂ production (Fig. 15f) is originated from the fact that C–O–C and -CHO are the most active sites toward the H₂O₂ synthesis compared to other functional groups.

Fig. 15

a ORR performance, b selectivity and electron transfer number of OMPC4 and MPC in O₂-saturated 0.1 M KOH. c Relationship between the selectivity and oxygen content of various OMPC-samples. Reproduced with permission from Ref. [220]. Copyright 2020 Elsevier. d ORR performance and e selectivity of aCB and CB in 0.1 M O₂-saturated KOH. f Activity volcano plot of the 2e⁻ ORR. Reproduced with permission from Ref. [222]. Copyright 2018 American Chemical Society

In addition to structured carbon materials, heteroatom (e.g., N, S, B) doped carbon materials have been widely studied for the electrocatalytic H₂O₂ production via the 2e⁻ ORR pathway [226,227,228,229,230]. Many studies have confirmed that N-doping of a carbon material can effectively promote the catalytic performance for the ORR, because the higher electronegativity of N atoms than carbon atoms can activate π electrons in carbon system. The charge redistribution is then induced or the electronic properties of carbon materials are changed, leading to varied adsorption capacity of carbon catalysts toward O₂ and generated intermediates [55, 231, 232]. It must be pointed out that common N-doped carbon catalysts are conducive to the 4e⁻ ORR pathway to produce water [233,234,235]. It is thus extremely important to bridge the relationship of a nano-structured carbon catalyst and the content of doped N inside with the ORR selectivity. For example, a simple hard-templating strategy was adopted to obtain mesoporous nitrogen-doped carbon (meso-BMP-800) catalyst and its nitrogen content was 14.2 wt% (as determined from XPS analysis) or 17.17 wt% (as evaluated from elemental analysis) [236]. On this electrocatalyst, selective H₂O₂ production was enhanced (Fig. 16a). This 2e⁻ ORR process has been experimentally and theoretically doable. In such situations, the strategy of N-doping of carbon materials helped to achieve high conductivities of carbon materials and their high selectivity for electrocatalytic H₂O₂ production. It is so-called a pure two-electron mechanism. However, the regulatory mechanism of the structure of a carbon catalyst on its reactivity is still unclear. In this context, the effect of porous structures of carbon catalysts on their catalytic properties toward the H₂O₂ production has been explored [237]. Mesoporous nitrogen-doped carbon (MNC) exhibited higher H₂O₂ selectivity (> 90%, Fig. 16b) than activated nitrogen-doped carbon (ANC). Enhanced mass transport into a mesoporous structure of this catalyst was suggested to be responsible for such high H₂O₂ selectivity of MNC (Fig. 16c). Later, systematic studies have been explored to reveal the correlations between the performance of electrocatalytic H₂O₂ production on carbon catalysts and their structural characters and surface physicochemical properties [238]. The selectivity of the H₂O₂ production on N-doped carbon catalysts is more dependent on the nitrogen doping effect rather than zeta potentials and defect sites of used carbon catalysts. In other words, the tuning of N-doping in carbon materials is very important and even more critical than structuring of carbon materials to obtain the high performance for electrocatalytic H₂O₂ production. In addition to N-doped carbon materials, B/S-doped ones were synthesized for electrocatalytic H₂O₂ generation. For example, B-doped mesoporous carbon materials have been synthesized via a convenient hydrothermal method, where F127 and boric acid were the template and boron source, respectively [239]. The obtained B-MC-F2 exhibited enhanced performance for electrocatalytic H₂O₂ formation in alkaline solution. In another case, a hollow, porous and S-doped carbon spheres presented higher activity and selectivity (> 70%) via the 2e⁻ ORR pathway than carbon spheres without S-doping. The introduced S–S bond is believed to feature low overpotential for the 2e⁻ ORR, leading to high activity for electrocatalytic H₂O₂ production [240]. In addition to the monoatomic-doped carbon materials, multiple heteroatoms co-doped carbon materials have been utilized for electrocatalytic H₂O₂ generation. For example, B,N co-doped (BN-C) carbon materials prepared through a simple heat-treatment (Fig. 16d) showed higher activity and selectivity for the H₂O₂ production via the 2e⁻ ORR pathway than the N-doped carbon materials [241]. Once again, there are various facts (e.g., heteroatom doping) affect the 2e⁻ and 4e⁻ pathways on carbon materials (Fig. 16e). However, the H₂O₂ formation on B,N-doped carbon materials was preferentially driven by 2e⁻ ORR pathway, resulting from unique catalytic behavior at the interface of h-BN domains and the host graphene lattice.

Fig. 16

a The number of electrons transferred based on the Koutecky–Levich plot. Reproduced with permission from Ref. [236]. Copyright 2012 American Chemical Society. b RRDE voltammograms toward ORR over the MNCs and ANC. c Schematic illustration of the ORR on different N-doped carbon. Reproduced with permission from Ref. [237]. Copyright 2014 American Chemical Society. d HRTEM and elemental mapping images of BN-C1. e Schematic illustration of the types of doping motifs and ORR mechanisms on these active sites in the B,N co-doped electrocatalysts (C: gray, N: blue, B: red; O: white). Reproduced with permission from Ref. [241]. Copyright 2018 American Chemical Society

As an interesting type of electrocatalysts, N-doped carbon materials with atomically dispersed transition metals (M–N-C) have been paid close attention in recent years for various catalytic applications, especially as unique catalytic systems for the selective H₂O₂ production via the 2e⁻ ORR pathway [242,243,244,245]. For example, the relation between physicochemical properties of transition metal (e.g., Mn, Fe, Co, Ni and Cu) single-atom anchored in N-doped graphene and their performance of electrocatalytic H₂O₂ production has been systematically explored by the means of theoretical calculations and experimental tools [246]. The DFT calculations indicated that the intermediates of ^*OOH, ^*O and ^*OH that are adsorbed on the top site of the metal (M) atoms are the most stable configurations. According to related activity-volcano plot (Fig. 17a), Ni and Cu single-atom catalysts (SACs) tend to follow the 2e⁻ ORR pathway during O₂ reduction, suggesting their high selectivity toward the H₂O₂ generation. On the other side, a large barrier exists between the reduction of ^*OOH and the activation of O₂ molecule. A large overpotential for the ORR is thus indispensable, suggesting low activity of these SACs for electrocatalytic H₂O₂ production. Differently, the 4e⁻ ORR pathway becomes more favorable on the Mn and Fe SACs during the ORR, resulting in poor selectivity for electrocatalytic H₂O₂ production. Remarkably, only Co SACs exhibit optimal d-band centers near the apex of the volcano plots, indicated the appropriate adsorption for the intermediates to facilitate the 2e⁻ ORR pathway. Its high activity and selectivity for the 2e⁻ ORR were confirmed by electrochemical ORR tests, where this Co-NC catalyst showed high activity and selectivity for the H₂O₂ production in acidic media (Fig. 17b, c). Similarly, the activity and selectivity of M–N–C (M = Mn, Fe, Co, Ni and Cu) SACs for electrocatalytic H₂O₂ production have been investigated [247]. Among these catalysts, the Co–N–C catalyst exhibited the highest ORR selectivity (80%) at 0.1 V_RHE in 0.5 M H₂SO₄ than that of Fe–N–C (28%), Cu–N–C (36%), Mn–N–C (43%), N–C (45%) and Ni–N–C (52%) catalysts under the same conditions. Such high performance was assigned to the fact that the binding free energy of Co–N–C is close to that at the top of the H₂O₂ production in the volcano plot (Fig. 17e). The number of electrons transferred of a Co–N–C catalyst was much closer to 2 (Fig. 17d). This result indicates that the Co–N–C catalyst is more favorable for 2e⁻ ORR pathway to produce H₂O₂ (Fig. 17f). The H₂O₂ selectivity of a Co–N–C catalyst is increased with the potential decreased in acidic solutions, while remained well in the whole potential range under neutral and alkaline electrolyte. One optimal Co–N₄ moiety incorporated in N-doped graphene exhibited excellent H₂O₂ productivity and durability [248]. In the volcano plot of ORR activity, none of M-N₄/graphene catalysts are located at the peak position (ΔG_OOH* = 4.2 eV) for the H₂O₂ production. As for the Co-N₄, its ΔG_OOH* can be modulated once functional groups are attached. For example, when one O* is adsorbed near the Co-N₄ moiety (denoted as Co–N₄(O)), the ΔG_OOH* value increases from 3.9 to 4.1 eV, very close to the optimal ΔG value for the H₂O₂ production. This phenomenon can be clarified according to the difference in the charge states of cobalt atoms. Specifically, the charge states of cobalt atoms in the Co–N₄(O) and Co–N₄(2O) moieties are 0.05 e⁻ and 0.10 e⁻ more positive than Co–N₄. Therefore, a reasonable design strategy of electrocatalytic H₂O₂ synthesis is to slightly enlarge ΔG_OOH* value of the Co–N₄/graphene electrocatalyst by constructing electron-rich oxygen species near the Co–N₄ moiety. In this way, the charge states of cobalt atoms are thus gently increased. With aid of this strategy, the Co₁-NG(O) SAC were prepared. Among them, the Co₁-NG(O) catalyst showed excellent activity and high selectivity for electrocatalytic H₂O₂ generation.

Fig. 17

a Activity-volcano plot of ORR through 2e⁻ or 4e⁻ route. b Linear sweep voltammograms (LSVs) of various catalysts in O₂-saturated 0.1 M HClO₄. c Faradic efficiencies of the H₂O₂ formation at different potentials. Reproduced with permission from Ref. [246]. Copyright 2019 Elsevier. d H₂O₂ selectivity and the number of electrons transferred (n) at +0.1 V_RHE. e Activity-volcano plot for the ORR. f H₂O₂ selectivity and the number of electrons transferred (n) at different potentials. Reproduced with permission from Ref. [247]. Copyright 2019 American Chemical Society

4.2 Electrocatalytic H₂O₂ Synthesis Through a 2e⁻ WOR Pathway

In addition to the 2e⁻ ORR pathway, the 2e⁻ WOR has been considered as another promising approach toward electrocatalytic H₂O₂ production. It is worth mentioning that in a 2e⁻ WOR process only H₂O is involved as the reactant for the electrocatalytic H₂O₂ production. Such systems are thus more convenient to be operated than those using the 2e⁻ ORR pathway. However, the H₂O₂ generation via the 2e⁻ WOR pathway usually displays relatively low Faradaic efficiency. This is because the competitive 4e⁻ WOR (namely O₂ evolution process) is thermodynamically more favorable than the 2e⁻ WOR (namely the H₂O₂ formation) accompanying with the autocatalytic decomposition from H₂O₂ into O₂. In this regard, there are limited studies on this infant field.

4.2.1 Metal Oxides

Noble metal oxides (e.g., RuO₂ and IrO₂) have been frequently applied for various catalytic applications, due to their high catalytic activities and long-term stability at high potentials [257, 258]. However, their prices are high to limit the large-scale applications [259]. Many alternative noble-free metal oxides thus became more popular in the H₂O₂ production via the 2e⁻ WOR pathway (Table 4). For example, the free energies of OH* and O* were identified as key parameters to confirm selectivity and activity of different electrocatalysts toward the 2e⁻ WOR [260]. The metal oxide catalysts that simultaneously meet two conditions for OH* and O* binding energies (ΔG_O ≳ 3.5 eV and ΔG_OH ≲ 2.4 eV) are favorable for selective H₂O₂ production [261]_. In the summarized relationship between the selectivity of 2e⁻ WOR and the binding energies of O/OH for a number of different metal oxides (Fig. 18a), metal oxides (e.g., WO₃, BiVO₄, MnO₂ and SnO₂) are located in the region for weak O adsorption energies (green region). They thus promote the 2e⁻ WOR to form H₂O₂. In verse, metal oxides (e.g., IrO₂, RhO₂ and PtO₂) that feature strong O adsorption energies (blue region) are favorable to generate O₂ as the major product. This study provides a good theoretical guide to rationally design metal oxide catalysts in the process of 2e⁻ WOR to promote the H₂O₂ production. Meanwhile, the activity trends of various metal oxides (e.g., WO₃, ZnO, CaSnO₃, BiVO₄, SnO₂ and TiO₂) for the H₂O₂ production via the WOR pathway were summarized as the activity volcano plots. They were based on the relationship between limiting potentials and the free energy of OH* (Fig. 18b) [54]. It was proved again that the catalysts possessing suitable ΔG_OH* value (e.g., from 1.6 to 2.4 eV) and low overpotential for the 2e⁻ WOR exhibits high selectivity and activity toward the H₂O₂ production.

Table 4 Summary of electrocatalytic H₂O₂ production via a 2e⁻ WOR pathway

Fig. 18

a Phase diagram represented by the binding energies of O* and OH*. Reproduced with permission from Ref. [261]. Copyright 2017 American Chemical Society. b Activity volcano plots for the 2e⁻ and 4e⁻ WOR. Reproduced with permission from Ref. [54]. Copyright 2020 Elsevier

In 2012, a MnO_x electrocatalyst was applied for the H₂O₂ formation via the 2e⁻ WOR pathway [50]_. In this system, 77% of the H₂O₂ production was achieved. Unfortunately, the activity and yield of H₂O₂ were too low. This low efficiency was attributed to the competing process of the 4e⁻ ORR and/or continuous loss of H₂O₂ from the disproportionation. Three BiVO₄ catalysts with different morphologies containing different surface ratios of (−121) and (040) facets (so-called seed, nanoneedles and truncated BiVO₄) have been applied for electrocatalytic H₂O₂ production [263]. The high-index plane (−121) was disclosed to be favorable for the H₂O₂ formation in 1 M KHCO₃ electrolyte rather than the O₂ formation. Unfortunately, BiVO₄ has poor stability in H₂O₂ synthesis [264]. The Bi₂WO₆ catalysts were then selected as alternatives to BiVO₄ for the H₂O₂ generation via the 2e⁻ WOR pathway [265]. Among a series of Bi-based oxides (Fig. 19a), Bi₂O₃, BiVO₄ and Bi₂WO₆ displayed preference for electrocatalytic H₂O₂ production. On the Bi₂WO₆ electrocatalyst, the best performance for H₂O₂ generation was achieved. The current density on the Bi₂WO₆ electrocatalyst was further increased by means of optimizing the morphology and Mo doping of the Bi₂WO₆ electrocatalyst. The faradaic efficiency of a FTO/Bi₂WO₆:5%Mo electrocatalyst was 79% at 3.2 V_RHE and its maximum yield rate of electrocatalytic H₂O₂ generation rate was as high as 300 μmol h⁻¹ cm⁻² at 3.4 V_RHE (Fig. 19b). The durability test of this electrocatalyst is very stable for such electrocatalytic H₂O₂ production (Fig. 19c).

Fig. 19

a Faradaic efficiencies (FEs) and productivity of Bi-based oxides at 2.6 V_RHE. b FE and H₂O₂ generation rate of a FTO/Bi₂WO₆:5%Mo electrocatalyst at various potentials. c Stability measurement of a FTO/Bi₂WO₆:5%Mo electrocatalyst under different potentials. Reproduced with permission from Ref. [265]. Copyright 2020 Wiley–VCH GmbH. d FE, H₂O₂ generation rates and e the densities of different catalysts at different potentials. f Durability test of CaSnO₃ catalyst under various potentials. Reproduced with permission from Ref. [267]. Copyright 2019 American Chemical Society. g Free energy diagram of the WOR for the most active (10 $\overline{1}$ 0) ZnO catalyst. h Volcano plots of 4e⁻ WOR to O₂ and 2e⁻ WOR to H₂O₂ as a function of OH* binding free energies (ΔG_OH*). i Overall current density and FE for H₂O₂. Reproduced with permission from Ref. [268]. Copyright 2019 American Chemical Society

Other metal oxides such as CaSnO₃ and ZnO have been also used for the H₂O₂ production via the 2e⁻ WOR pathway [266,267,268]. For example, CaSnO₃ nanoparticles, prepared via the colloidal synthesis method, have been employed to evaluate H₂O₂ production in 2 M KHCO₃ electrolyte [267]. Its Faraday efficiency (FE) was 76% at 3.2 V_RHE, higher than that of F:SnO₂ and SnO₂ (Fig. 19d). The H₂O₂ generation rates are outstanding at applied potentials (Fig. 19e). Meanwhile, it exhibited excellent stability during this 2e⁻ WOR process at 2.0 V_RHE within 12 h. Slight decrease occurred when a potential of at 2.2 or 2.8 V_RHE is applied (Fig. 19f). In another case, metal oxide of ZnO was identified as an ideal catalyst for electrocatalytic H₂O₂ synthesis. The (10 $\overline{1}$ 0) facet on the ZnO catalyst is more favorable than the (0001) facet, as demonstrated by the DFT calculations (Fig. 19g, h) [268]. Experimental results further indicated that ZnO nanoparticles with a high fraction of (10 $\overline{1}$ 0) facets own outstanding performance for catalytic H₂O₂ generation: the overpotential of 40 mV at a current density of 0.1 mA cm⁻² and an FE of 81% during this 2e⁻ WOR process (Fig. 19i). Mn-alloyed TiO₂ coatings, prepared by atomic layer deposition, also exhibited the high performance of electrocatalytic H₂O₂ generation via a 2e⁻ WOR pathway: e.g., an FE of higher than > 90% at the overpotential smaller than 150 mV [269]. It has to highlight here that the principles in designing the electrocatalysts with high performance toward the 2e⁻ WOR is still unclear at this stage. More explorations on reaction mechanism for WOR need be provided in future.

4.2.2 Boron-Doped Diamond

As discussed, various noble metal oxides have shown the capability to generate H₂O₂ through a 2e⁻ WOR pathway. However, these noble metals are expensive and rare. Up to date, the efficiency of the H₂O₂ production via 2e⁻ WOR is low. On the other side, noble metal-free oxides are not adapted to be operated at the high current densities and high potentials that are frequently required for the large-scale applications. Among the materials that can endure the corrosion in such a harsh conditions for water oxidation, some molecular metal porphyrins (e.g., AlTMPyP, SiTPyP, ZnTMPyP, SnTMPyP) [273,274,275,276] and carbon fibers [57, 277] are promising and thus have been tried for H₂O₂ production via the 2e⁻ WOR pathway, but their performance was still poor.

Boron-doped diamond (BDD), as a unique kind of carbon material with excellent physical–chemical stability [278], has been employed for the H₂O₂ generation via the 2e⁻ WOR pathway. Since the H₂O₂ production through 2e⁻ WOR on the BDD electrode was reported for the first time in 2003 [279, 280], the catalytic performance of electrochemical H₂O₂ formation using synthetic BDD films that were coated on a titanium substrate has stimulated extensive attention [281]. The BDD exhibited a H₂O₂ yield of as high as 29.0 mmol dm⁻³ at 3.47 V_RHE. The H₂O₂ production rate was 19.7 μmol min⁻¹ cm⁻². However, the FE was only 28% at 3.17 V_RHE and total current densities surpassed 120 mA cm⁻², suggesting a high energy consumption process. To promote the electrocatalytic H₂O₂ generation on BDD via the 2e⁻ WOR pathway as well as to decrease the overpotential at a high applied current density, BBD films were optimized by tailoring their boron content, thickness and crystal size [282]. On the BDD-4 (with the B doping level of 12,600 ppm, film thickness of 2 μm), the biggest FE of 87% was reached and the H₂O₂ production rate was 76.4 μmol cm⁻² min⁻¹. This BDD-4 was also superb durable (e.g., electrocatalytic H₂O₂ production for 10 h at a constant current density of 200 mA cm⁻²). An electrogenerated chemiluminescence (ECL) system was developed via in-situ coreactant production at a BDD electrode [283], where peroxydicarbonate (C₂O₆²⁻) was produced by the oxidation of carbonate (CO₃²⁻), subsequently reacted with water in basic media to form H₂O₂. Based on experimental and no-economic examination, the efficiency of anodic H₂O₂ production on a BDD electrode was explored in an electrochemical H₂/H₂O₂ process using sodium carbonate as an electrolyte [284]. The calculated FE of producing H₂O₂ was to up to 31.7% at 2.90 V (vs. SHE) and the corresponding production rate reached 3.93 μmol min⁻¹ cm⁻². This study provided guidance to further implement selective oxidation of water to H₂O₂ on a BDD electrode.

Note that great progress has been made in the H₂O₂ preparation using photo- and photoelectro-catalytic approaches for wastewater treatment at a pilot scale, even an industrial scale. This is because the efficient H₂O₂ production is necessary to guarantee Fenton’s reaction [285, 286]. For example, a carbon-PTFE electro-oxidative system promoted the oxidative power of Fenton and Fenton-like processes based on the enhanced H₂O₂ generation with a higher current efficiency [287]. Solar photoelectro-fenton (SPEF) with a BDD anode and a cathode of the (Co, S, P)-decorated multiwalled carbon nanotubes showed the excellent performance for bronopol removal due to the high efficiency of H₂O₂ production via the 2e^- WOR on a BDD [288].

5 Conclusions and Perspectives

This article provides an overview of current advances on photo/electro-catalytic H₂O₂ production using different pathways. Different photo/electro-catalysts have been detailed. Compared with traditional preparation methods, such as AQ oxidation, based on fossil resources, the direct preparation of H₂O₂ through photo/electro-catalytic processes appears to be more promising in that these processes are simple to be operated, green and safe, sustainable and clean, especially when solar or wind energy is integrated. The used catalyst is the prerequisite and technical core to determine the efficiency of such catalytic H₂O₂ generation among these technologies, especially in the application beyond the laboratory. In more detail, many advanced photo/electro-catalysts must be constructed under theoretic guidance, including the strategies of their synthesis, design of their components/structures. Once their photo/electro-chemical properties are explored and discussed, their catalytic H₂O₂ production is possible to be boosted. However, the mechanism comprehension of various catalysts for their photo/electro-chemical H₂O₂ production is still in the preliminary stage of research, and there is still much room for further improvement.

In the field of the photocatalytic synthesis of H₂O₂ by means of either the 2e⁻ ORR pathway or the 2e⁻ WOR process, it is concluded that an ideal catalyst needs simultaneously satisfy good photoresponse (e.g., an appropriate band gap), superior separation capacity of photogenerated carriers (namely electrons and holes), as well as high catalytic ability for the H₂O₂ yield. However, the reported photocatalysts up to date still suffer from relatively low activity or selectivity of H₂O₂, which remains a great challenge for practical applications. Future studies should propose the innovative ideas for applying cooperative modulation of light adsorption, photogenerated charge separation and catalytic reaction actives in photocatalysts to enhance catalytic H₂O₂ production. In these regards, supramolecular semiconductors, including MOFs, covalent organic frameworks (COFs) and conjugated polymers can be added as the co-catalysts to improve the H₂O₂ yield. This is because the energy band structures of these supramolecular semiconductors can be optimized. In other words, their light absorption, charge separation and redox capability can be tuned during the photocatalytic H₂O₂ production. Another promising strategy is to combine inorganic semiconductors with biomolecules to strengthen the response to the solar light and the selectivity of electrocatalytic H₂O₂ production. It must be pointed out that although standard photocatalysts (e.g., g-C₃N₄ and TiO₂ materials) have achieved selective H₂O₂ synthesis under visible light irradiation, organic sacrificial agents are usually used to improve the separation of photogenerated electrons and holes in these cases. Such an approach inevitably leads to the contamination of H₂O₂. Additional cost is then required to separate and obtain pure H₂O₂. Consequently, the strategies based on electrochemical assistance, oil–water phase solution and all-solid-state Z-scheme photocatalytic systems appear to effectively facilitate the photogenerated charge separation. Note that the output of H₂O₂ produced by the currently reported catalysts is only at the mM or μM level, too far for large-scale and industrial applications. Porous photocatalysts and/or new photocatalytic reaction systems with enhanced mass transfer rates are needed to boost the formed H₂O₂ removal from the active sites and to avoid the H₂O₂ decomposition. A deep understanding of the photocatalytic mechanism will open up new opportunities for the design of high efficiency photocatalysts. Various high-quality and in-situ characterization technologies (e.g., in-situ HRTEM, IR, XANES) are helpful to disclose the distinct dynamic behavior and reaction mechanism of different photocatalytic H₂O₂ formation systems. In short, exploring effective strategies to promote the ability to harvest wide solar spectrum, improving the lifetime of photogenerated electrons and holes and propelling the 2e⁻ ORR or water oxidation via the design of advanced photocatalysts have been becoming a hot issue for photocatalytic H₂O₂ production.

Like photocatalytic H₂O₂ synthesis, electrocatalytic H₂O₂ synthesis includes two pathways: 2e⁻ ORR and 2e⁻ WOR processes, accompanying with significant achievements acquired in the past few years. Various electrocatalysts with high selectivity (or Faraday efficiencies) and stability have been developed. However, more efforts are still needed to explore and disclose the mechanisms behind. Take the 2e⁻ ORR process as an example, previous reports have revealed that noble metal-based electrocatalysts (e.g., Pt- and Pd-based electrocatalysts) exhibited decent performance for electrocatalytic H₂O₂ production. Unfortunately, owing to their high costs and scarce reserves, it is unrealistic to use them for industrial H₂O₂ production. To address this issue, noble-based alloys, noble catalysts with core–shell structures, and porous or single atomic noble metal catalysts are highly desirable for high-performance H₂O₂ production. Alternatively, carbon-based materials are recognized as potential ones to noble catalysts for the large-scale H₂O₂ generation. Note that pristine carbon is generally inert for H₂O₂ generation due to its lack of active sites. Its catalytic performance can be integrated and improved by the modification of surface carbon atoms such as the manufacture of porous structures and defects, the introduction of heteroatom doping and oxygen functionalization. Nevertheless, most carbon electrocatalysts are only applicable for the H₂O₂ production under alkaline conditions. Their further research is thus needed to improve catalytic performance in acidic media. Moreover, the hurdle of their instability that is caused by the H₂O₂ decomposition as well as their low selectivity over a wide potential range inhibits the practical applications of carbon electrocatalysts. Some new electrochemical and photochemical electrode systems need to be designed and developed. For example, BDD, a robust sp³ hybridized carbon material, has exhibited extremely high durability for the H₂O₂ production in harsh conditions. It is thus highly promising for industrial H₂O₂ production once the manufacturing cost of BDD films is dramatically reduced. Moreover, the structure–activity relationship of these electrode systems, especially when they are at nanoscales, is critical for the H₂O₂ production. Only on their optimized interfaces, the highly efficient H₂O₂ synthesis can be realized. To reduce the H₂O₂ decomposition, it is better to construct porous structures or multi-site active centers inside carbon electrocatalysts, which can prohibit the formed H₂O₂ from further reaction. One more promising strategy is to couple external means (e.g., ultrasound and microwave) with electrochemical approaches to firm up the H₂O₂ desorption from the electrocatalysts. To increase the selectivity of the H₂O₂ production, the sizes and exposed planes of the used electrocatalysts should be further optimized. The employment of the 2e⁻ WOR process should be paid more attention as a more effective pathway for the H₂O₂ production where no gas phase is involved. As confirmed from theoretical calculations and experimental operations, metal oxides (e.g., BiVO₄, SnO₂ and ZnO) tend to generate H₂O₂ via the 2e⁻ WOR approach, while others (e.g., IrO₂, PtO₂ and RhO₂) are more favorable for the 4e⁻ WOR approach to produce O₂. In this context, it is highly necessary to further study the mechanisms of the H₂O₂ formation via the 2e⁻ WOR approach. Once advanced metal oxide catalysts are found, electrocatalytic H₂O₂ production on these metal oxide electrocatalysts can be further enhanced. It is worth mentioning that BDD presented good ability to produce H₂O₂ at high potentials. However, it only exhibited low Faraday efficiency. To utilize diamond for industrial H₂O₂ production at high current densities, it is urgently needed to improve its catalytic ability by the formation of its nanostructures or the introduction other catalytic centers (e.g., defects, dopants, other catalysts).

To sum up, recent advances in the design and synthesis of photo/electro-catalysts have improved their activity, selectivity and stability for photo/electro-catalytic H₂O₂ production in past years. To further improve performance of photo/electro-catalytic H₂O₂ production, the design of advanced catalysts with various functions is still highly necessary. For photocatalysts, two-channel reaction pathways for generating H₂O₂ with high efficiency are highly pursued. As for electrocatalysts, their high activity with low cost and strong durability is the key factor moving forward. From a practical application perspective, further understanding the reaction kinetics using theoretical and experimental (especially in situ and operando techniques) as well as the design of new reactors for the H₂O₂ production at an industrial scale must be conducted in future.