Microstructural origin of selective water oxidation to hydrogen peroxide at low overpotentials: a study on Mn-alloyed TiO₂

Abstract

One key objective in electrocatalysis is to design selective catalysts, particularly in cases where the desired products require thermodynamically unfavorable pathways. Electrochemical synthesis of hydrogen peroxide (H₂O₂) via the two-electron water oxidation reaction (2e⁻ WOR) requires a +0.54 V higher potential than four-electron O₂ evolution. So far, best-performing electrocatalysts require considerable overpotentials before reaching peak faradaic efficiency. We present Mn-alloyed TiO₂ coatings prepared by atomic layer deposition (ALD) and annealing as a stable and selective electrocatalyst for 2e⁻ WOR. Faradaic efficiency of >90% at < 150 mV overpotentials was achieved for H₂O₂ production, accumulating 2.97 mM H₂O₂ after 8 hours. Nanoscale mixing of Mn₂O₃ and TiO₂ resulted in a partially filled, highly conductive Mn³⁺ intermediate band (IB) within the TiO₂ mid-gap to transport charge across the (Ti,Mn)O_x coating. This IB energetically matched that of H₂O₂-producing surface intermediates, turning a wide bandgap oxide into a selective electrocatalyst capable of operating in the dark. However, the high selectivity is limited to the low overpotential regime, which limits the system to low current densities and requires further research into increasing turn-over frequency per active site.

---

1. Introduction

Hydrogen peroxide (H₂O₂) is an important green oxidant¹ routinely used for chemical and environmental applications² such as pulp bleaching,³ chemical-mechanical actuation,⁴ water treatment, and disinfection.⁵ In addition, H₂O₂ is receiving growing interest as a potential energy carrier, capable of long-duration energy storage.⁶ H₂O₂ is primarily produced by the anthraquinone oxidation process.^2,7 However, this process involves sequential hydrogenation and oxidation steps which require energy-intensive gas transport and product separation—processes only economical at plant-scale, hindering distributed use for water treatment or disinfection. Direct synthesis of H₂O₂ from H₂ and O₂ over heterogeneous catalysts is one route for distributed synthesis,⁸ but recently, (photo-)electrochemical synthesis of H₂O₂ which uses abundant water or air as feedstock and renewable electricity or sunlight as energy inputs has provided an attractive alternative to the industrial anthraquinone-based process.⁹

H₂O₂ can be (photo-)electrochemically synthesized via a two-electron (2e⁻) water oxidation pathway, as shown in eqn (1):

2H₂O → H₂O₂ + 2H⁺ + 2e⁻, E = +1.77 V_RHE (1)

However, existing catalysts often exhibit low faradaic efficiency (FE) due to the competitive four-electron (4e⁻) pathway of O₂ evolution reaction (OER; eqn (2)):

2H₂O → O₂ + 4H⁺ + 4e⁻, E = +1.23 V_RHE (2)

The thermodynamic driving force for OER is always at least +0.54 V greater than that of the 2e⁻ H₂O₂ pathway which can allow the undesired O₂ evolution pathway to overcome any kinetic barriers and dominate. In addition, the produced H₂O₂ can be over-oxidized to O₂ at an excess potential of at least +1.09 V (eqn (3)):

H₂O₂ → O₂ + 2H⁺ + 2e⁻, E = +0.68 V_RHE (3)

Finally, catalytic surfaces should not disproportionate the adsorbed H₂O₂ molecules:

2H₂O₂ → 2H₂O + O₂ (4)

Thus, the selectivity of 2e⁻ water oxidation reaction (WOR) represents a common conundrum in electrocatalysis and strategies to address this challenge can be broadly applicable.^9–11

Various water-oxidation materials have been both experimentally and theoretically investigated, including aluminum porphyrins,¹² carbon-based materials,¹³ BiVO₄,^14,15 WO₃,¹⁶ CaSnO₃,¹⁷ and various composite materials.¹⁸ Several electrocatalytic¹⁹ and photoelectrocatalytic¹⁵ systems have showed faradaic efficiency (FE) values as high as 86%. However, best performing 2e⁻ WOR catalysts ZnO²⁰ and CaSnO₃ (ref. 17) reach their peak FE at +2.6 V_RHE and +3.2 V_RHE, respectively, requiring considerable overpotentials to reach the 81% and 76% FE for H₂O₂ selectivity, respectively. Thus, strategies to achieve high selectivity at low overpotentials should be a major milestone for the field.

MnO_x is a conductive oxide that has been used for H₂O₂ production.²¹ However, MnO_x has not been successfully utilized in aqueous environments as it is capable of 4e⁻ O₂-evolution and has been shown to disproportionate H₂O₂. TiO₂-rich surfaces may be promising for H₂O₂ production, since TiO₂ is ineffective for either O₂ evolution or H₂O₂ decomposition. Typically, TiO₂ undergoes a 1e⁻ pathway to produce OH˙, which then produces H₂O₂ or O₂.²²

In this study, the authors synthesize Mn-alloyed TiO₂via atomic layer deposition (ALD), denoted as (Ti,Mn)O_x. The introduction of Mn 3d electronic states leads to the formation of a highly conductive Mn IB inside the otherwise forbidden bandgap of an oxide coating. Based on materials characterizations, we confirmed the nanoscale mixing of conductive and catalytically active phases otherwise not possible without ALD and post-growth annealing. We then investigated the effects of the electronic tuning of TiO₂ and found that the nanocrystalline microstructure resulted in a multifunctional electrocatalyst for selective water oxidation to hydrogen peroxide.

2. Experimental methods

2.1 Atomic layer deposition

The composite (Ti,Mn)O_x films were grown by using 2, 4, 8, or 16 ALD cycles of TiO₂ per 1 cycle of MnO_x, which are denoted as 2 : 1, 4 : 1, 8 : 1, and 16 : 1 (TiO₂ : MnO_x) films (ESI S1†). Tetrakis-dimethylamido-titanium (TDMAT, Sigma-Aldrich, 99.999%) and bis-(ethyl-cyclopentadienyl) manganese (Mn(EtCp)₂, Strem Chemicals, 99.999%) precursors for ALD were used as the Ti and Mn sources, respectively. The growth rates of MnO_x and TiO₂ were 0.97 and 0.47 Å per cycle, respectively. The MnO_x cycle was placed in the middle of the TiO₂ cycles. For example, 4 : 1 cycle ratio should lead to TiO₂-rich surfaces with subsurface Mn atoms. The sequences for each film composition were repeated until each film reached the desired thickness of ca. 50 nm. All (Ti,Mn)O_x films were annealed in air at 500 °C for 2 hours and oven cooled. For electrocatalysis, fluorine-doped tin oxide (FTO) substrates were used, while fused silica substrates were used for material characterizations. The Mn elemental composition was correlated with, but not equal to, the nominal cycle ratio.

2.2 Electrochemical measurements

The electrochemical behavior of (Ti,Mn)O_x electrodes was measured with a three-electrode setup. A Bio-Logic S200 potentiostat system connected the (Ti,Mn)O_x working electrode, saturated Ag/AgCl reference electrode, and Ti foil counter electrode. The Ti counter electrode was chosen to keep consistent with a previous study.¹⁴ A cation exchange membrane separates the (TiMn)O_x and Ti foil electrodes to confine the H₂O₂ production in the working electrode compartment. Cyclic voltammetry was acquired with a scan rate of 20 mV s⁻¹.

2.3 H₂O₂ quantification

FE for H₂O₂ production was quantified by measuring the molar quantities of accumulated H₂O₂ produced by 4 : 1 (Ti,Mn)O_x as a function of current densities (geometric electrode area), ranging from 0.1 to 5.0 mA cm⁻² (Fig. 2c). H₂O₂ concentration was quantified and validated using multiple titration methods in an H-type membrane cell (S1.5).† High-performance liquid chromatography (HPLC, Fig. S15†) quantification was used only to validate product identity and titration precision. Extended discussion can be found in ESI (S1.5 H₂O₂ quantification).†

2.4 XPS characterization of intermediate-band electronic structures

Valence band XPS was used to characterize the electronic structures of catalytic surfaces. The valence XPS spectra of the (Ti,Mn)O_x films were compared with that of a pure TiO₂ film and used to derive the IB energy levels in a band diagram.²³ Peak deconvolution was conducted under the following assumptions: (1) bandgap measurements for MnO_x-alloyed TiO₂ by UV-Vis spectroscopy (Fig. S9†) showed that (Ti,Mn)O_x bandgaps remain the same as pure TiO₂ (2) extended X-ray absorption fine structure (EXAFS) characterizations (Fig. S10†) show the presence of Mn–Mn feature as the second coordination shell, indicating that MnO_x forms clusters or nanocrystals inside the host oxide of crystallized TiO₂. Therefore, we consider the 4 : 1 (Ti,Mn)O_x, coating has nanoscale mixing of rutile TiO₂ and bixbyite Mn₂O₃ crystallites, and therefore contains two types of oxygen, one bridging neighboring Ti atoms and the other connecting neighboring Mn atoms. Each type has a set of non-overlapping O 2p–π and O 2p–σ peaks. Therefore, the peak deconvolution involved six peaks for 4 : 1(Ti,Mn)O_x.

Valence XPS analysis of pure TiO₂ resulted in two main peaks. This analysis is discussed later in the paper and presented in Fig. 4. One peak occurred at the binding energy (BE) of 7.10 eV and the other peak occurred at 5.03 eV.²⁴ The valence XPS spectra of 4 : 1 (Ti,Mn)O_x has two extra oxygen peaks that are slightly (∼0.3 eV) more negative to these two peaks. In addition to the oxygen peaks, two additional peaks were assigned to the Mn 3d states. The Mn 3d-t_2g and Mn 3d-e_g bands corresponded to the Mn 3d_3/2 and Mn 3d_5/2 peaks, respectively.^25,26 The two primary O 2p peaks were still assigned at 7.10 ± 0.10 eV and 4.90 ± 0.10 eV, while the two Mn 3d peaks were located at 3.03 ± 0.10 eV and 1.14 ± 0.10 eV. All these peaks are referring to Fermi level position (E_F, designated at zero binding energy using Au calibration standards, as shown in Fig. S11†).

3. Results & discussion

As deposited (Ti,Mn)O_x contains atomically mixed Mn and Ti atoms in an amorphous phase due to the 150 °C growth temperature.²⁷ After annealing, diffraction peaks of 4 : 1 (Ti,Mn)O_x (Fig. 1a) indicate mixed nanocrystalline phases of rutile TiO₂ (PDF#01-078-1510) and bixbyite Mn₂O₃ (PDF#00-041-1442). Bixbyite Mn₂O₃ represented the dominant phase for annealed 2 : 1 and 4 : 1 (Ti,Mn)O_x while rutile represented the dominant phase for annealed 8 : 1 and 16 : 1 (Ti,Mn)O_x as no obvious XRD peaks for Mn₂O₃ were observed. However, extended X-ray absorption fine structure analysis at the Ti and Mn edges of the annealed 4 : 1 (Ti,Mn)O_x (Fig. S10†) revealed distinct second coordination shells for both Ti and Mn elements, consistent with the nanocrystallinity of both TiO₂ and Mn₂O₃ in the annealed coating.

Fig. 1 (a) Grazing incident X-ray diffraction (GI-XRD) spectra and (b) plan-view scanning electron microscope (SEM) images of annealed (Ti,Mn)O_x. Samples denoted Mn₂O₃ and TiO₂ represent annealed ALD MnO_x and TiO_x, respectively.

Thus, rutile TiO₂ is present in 4 : 1 (Ti,Mn)O_x, and XRD analysis represents only a starting point for understanding the microstructure of (Ti,Mn)O_x, which is the topic of a further report. Diffraction peaks of annealed ALD TiO_x without Mn alloying indicated anatase TiO₂. Therefore, Mn-alloyed TiO_x results in a tuned electronic structure due to a proportion of common anion mixed-metal bonds, which are distinctive from previously reported mixed Ti–Mn oxides.^28,29

Plan-view (Fig. 1b) and cross-sectional SEM images of (Ti,Mn)O_x (Fig. S1†) indicate a conformal coating with a uniform thickness on the FTO substrate. TEM images (Fig. S4†) showed that Mn₂O₃ crystallites were uniformly distributed across the (Ti,Mn)O_x surface due to the diffusion of Mn atoms and Mn₂O₃ ripening during annealing. Image analysis indicated that the average sizes of Mn₂O₃ grains were 4.5 ± 0.5 and 7.7 ± 0.5 nm for 4 : 1 and 2 : 1 (Ti,Mn)O_x, respectively. The TEM images showed bending contours, suggesting residual strain in the annealed films. Williamson–Hall analysis of the GI-XRD data suggested a strained status of Mn₂O₃ crystallites with 0.5% lattice tension. The fitted crystallite sizes were consistent with the analysis of TEM bright-field images.

XPS core-level spectroscopy was used to quantify the surface composition of rutile TiO₂ and bixbyite Mn₂O₃ (see Fig. S2†) and to verify the Mn³⁺ oxidation state (see Table S2 and ESI Results S2.1†). Angle-resolved X-ray photoelectron spectroscopy (XPS) indicated uniform compositional distribution of Ti and Mn from the surface to the bulk (Fig. S3†). The 4 : 1 (Ti,Mn)O_x has 36.09 at% of Mn within the 3 nm skin depth of the coating surface.

(J–E) behavior of (Ti,Mn)O_x in a phosphate-buffered (PB) solution of pH 7 (Fig. 2a), revealed that current density correlated with increasing Mn concentration. The films, except for 16 : 1 which was resistive due to sparse Mn concentration, share a similar on-set potential of approximately +1.8 V_RHE. Annealed ALD MnO_x showed no activity for H₂O₂ production and annealed TiO₂ did not generate significant currents (Fig. S5†). Numerous O₂ bubbles were observed on the annealed Mn₂O₃ control and at the cell sidewall. The redox feature of annealed MnO_x cyclic voltammogram (Fig. S5b†) agrees with previous reports, which indicated the stable OER window at +1.6–1.8 V_RHE and the formation of Mn(VII) species when the potential is above of +1.8 V_RHE.³⁰ These findings indicate that catalytic activity is dependent on alloying TiO₂ with Mn₂O₃ nanocrystallites. Though 2 : 1 and 4 : 1 have similar J–E behavior, the crystal grain sizes for 2 : 1 (Ti,Mn)O_x of 7.7 ± 0.5 nm made it difficult to maintain a TiO₂-rich surface. As surface MnO_x compounds are known to disproportionate H₂O₂ in aqueous environments, 2 : 1 (Ti,Mn)O_x would pose challenges accumulating H₂O₂,³¹ and thus 4 : 1 (Ti,Mn)O_x was chosen as the focus of this study.

Fig. 2 Electrochemical performance. (a) J–E behavior of (Ti,Mn)O_x in 0.5 M pH = 7 phosphate buffer at a scan rate of 20 mV s⁻¹, (b) onset potentials at the respective J = 0.2 mA cm⁻² (left y-axis) and overpotentials (right y-axis) derived from the (Ti,Mn)O_xJ–E curves, plotted together with the data obtained from references of WO₃, BiVO₄, TiO₂, and SnO₂,¹⁰ ZnO,²⁰ and CaSnO₃.¹⁷ The measurement error for overpotentials is ± 0.02 V. (c) FE vs. electrolysis time using a 4 : 1 (Ti,Mn)O_x coating. (d) FE vs. overpotential at various current densities for the 4 : 1 (Ti,Mn)O_x coating in (c), overlaid with the FEs for 2 : 1, 8 : 1, and 16 : 1 (Ti,Mn)O_x films at J = 0.2 mA cm⁻².

J–E behavior measured in pH 3 and 9 PB solutions (Fig. S6†) showed similar trends, however, the pH 7 PB solution was chosen as the electrolyte for further testing to be compatible with reported membrane-free electrolyte-free H₂O₂ electrolysis designs that minimize transport loss.³²

The onset potentials (at J = 0.2 mA cm⁻²) of 4 : 1 (Ti,Mn)O_x (Fig. 2b) was +1.89 V_RHE, comparable to +1.90 and +1.96 V_RHE for the 2 : 1 and 8 : 1 films, respectively. The onset potentials of the (Ti,Mn)O_x coatings are non-variant in changing pH (Fig. 2b) and follow the trend that 4 : 1 (Ti,Mn)O_x has the lowest onset potential. Onset potentials were lower than the many catalysts, including BiVO₄ (+2.20 V_RHE), TiO₂ (+2.80 V_RHE), SnO₂ (+2.50 V_RHE), and were comparable to WO₃ (+1.90 V_RHE).¹⁴ CaSnO₃ reaches 0.2 mA cm⁻² with an overpotential of 230 mV but requires +3.20 V_RHE to reach peak Faraday efficiency of 76%.¹⁷ However, WO₃ was reported to be unstable due to surface H₂O₂ absorption, which may lead to low efficiency in the long term.¹⁶

Computational work suggests that there are optimal potential windows for H₂O₂ production, due to the competition with one-electron and four-electron oxidation reactions.¹⁴ For 4 : 1 (Ti,Mn)O_x, FE values of over 95% were obtained at overpotentials less than 150 mV, i.e., at +1.9 V_RHE, or at equivalent current densities of 0.1 mA cm⁻². As FE begins near unity at just above +1.77 V_RHE and decreases with increasing current density (Fig. 2c) or with increasing overpotentials, (Ti,Mn)O_x suffers from a small optimal potential window (Fig. 2d). However, this range of FE values was 20% higher than the highest reported FE of 66% for the hydrophobic carbon dark anode measured in NaHCO₃(aq).¹⁴ The average FE of 98% at 0.1 mA cm⁻² was comparable to reported photochemical results, such as a FE of 98% for BiVO₄ photoanodes¹⁴ and a FE of 80% for BiVO₄–Al₂O₃ photoanodes.³³ However, when the current density was increased to 1.0 mA cm⁻², corresponding to an overpotential of 250 mV, the FE dropped to less than 20%, and while this decline in FE is undesirable, it is consistent with other reports which show that selectivity can drop off with increased overpotentials.¹⁴ (Ti,Mn)O_x is unique as the measured FEs start >90% at near zero overpotentials and drop off while FE in other systems such as BiVO₄ and ZnO ramp up, whereas the FEs in other systems such as BiVO₄ and ZnO initially ramp up, reaching comparable selectivity only at >400 mV overpotentials.^14,20

Electrochemical impedance spectroscopy (EIS) was used to correlate electrical conductivity with composition-dependent J–E behavior (Fig. S7 & Table S3†). Analyses showed that the charge-transport resistance through the coatings were 48.6, 47.2, 50.1, 78.0 Ω for the 2 : 1, 4 : 1, 8 : 1, and 16 : 1 films, respectively. The impedance data also provided an estimated roughness factor of 10.7, which is calculated by dividing the geometric area of the electrode by the electrochemical surface area (see ESI Results S2.2).†

Eight-hour electrolysis experiment using 4 : 1 (Ti,Mn)O_x in a batch reactor accumulated 2.97 mM of H₂O₂ in 20 mL, or 59.4 μmol (Table S4†). This accumulation was achieved in a static cell demonstrating that (Ti,Mn)O_x is capable of H₂O₂ desorption without overoxidation. As shown in Fig. 3, the cumulative FE begins near unity then gradually decreases to 85.8% after eight hours. However, this figure represents only the lower bound of FE as it does not account for product decomposition in the cell over the same period. As membranes have been reported to be highly effective at preventing H₂O₂ crossover between chambers, all of the H₂O₂ is accounted as being solely produced via WOR.^34,35

Fig. 3 H₂O₂ accumulation using 4 : 1 (Ti,Mn)O_x-coated FTO electrode operated in the dark. Cumulative H₂O₂ concentration and the respective cumulative FE vs. time. The electrode area was 1.2 cm², and the applied potential was +1.31 V vs. Ag/AgCl (equivalent to an overpotential of 150 mV) in a 20 mL 0.5 M pH = 7.0 PB solution.

We further tested the long-term stability for 4 : 1 (Ti,Mn)O_x in PB solutions of pH 5, 7, and 9 to evaluate its potential as a protective layer. Chronopotentiometry studies at 0.2 mA cm⁻² showed a moderate increase in overpotential, with the electrode potentials increasing from +1.8 V_RHE to less than +2.1 V_RHE after 200 hours (Fig. S8†). The potential increase was partially accelerated by the formation of O₂ bubbles on the surface reducing the total surface area of the catalyst. As the chronopotentiometry experiment varies applied potential to sustain a current, this decrease in surface area led to an increase in required potentials during the constant current testing. As demonstrated in Fig. 2c, the selectivity towards H₂O₂ decreased at higher applied potentials, creating a feedback loop where higher applied potentials resulted in more oxygen generation. This process would slowly increase the required potential, until some breakdown voltage which would eventually lead to the failure of the coating. However, the coating was stable in all three pH conditions for over 200 hours, showing promise as protective coatings for semiconductors in corrosive environments.

The free energies of the relevant intermediates of the one, two, and four electron water oxidation reactions, i.e., OH* and O* free energy have been shown to be key in determining reaction pathways.¹⁴ While experimental evidence for the exact identity of the active site would allow for optimization and improved simulation, as recent literature has highlighted,^36,37 it can be challenging identifying the nanocrystalline active sites, even with advanced microscopy and EXAFS analysis. TiO₂ and MnO_x are on opposite sides of this volcano plot of a OH*-overpotential scaling relationship, with TiO₂ too large a OH* free energy and MnO_x not large enough.^38,39 Neither TiO₂ nor Mn₂O₃ is active for hydrogen peroxide production (Fig. S5a and b†), demonstrating that Mn-alloying is the necessary for activity. TiO₂ was chosen to be the surface termination layer as it was found not to disproportionate H₂O₂. Thus, (Ti,Mn)O_x promises favorable thermodynamics for selective peroxide production.

Valence XPS analyses indicated that a spatially continuous Mn-impurity IB was located within the TiO₂ bandgap due to unfilled Mn 3d electronic states (see ESI Results S2.3†). The red arrows in Fig. 4b represent the relative difference in binding energy (BE) between the valence band maximum (VBM) and the Fermi level, namely E_VBM − E_F, while the blue arrows represent the difference of BE between the IB center and the Fermi level, E_IB − E_F. With the addition of Mn, the position of the E_F shifted towards the VBM. For instance, E_VBM − E_F reduced from 3.20 V for pure TiO₂ (i.e., E_F at 0.14 V_RHE) to 2.31 V for 4 : 1 (Ti,Mn)O_x films (i.e., E_F at 1.09 V_RHE). This E_F lowering occurred because the Mn 3d states created inside the TiO₂ bandgap hosted free carrier electrons.

Fig. 4 Electronic structure characterizations of (Ti,Mn)O_x catalytic coatings. (a) Valence XPS spectra and (b) respective band diagrams (in the potential scale versus Reversible Hydrogen Electrode, RHE) of 4 : 1 (Ti,Mn)O_x and “leaky” TiO₂ surfaces. The positions of valence band maximum (E_VBM), intermediate band center (E_IB), and the Fermi level (E_F) are indicated in (a), and their respective energetics are drawn in (b). E_VBM − E_F and E_IB − E_F are labelled and are listed in Table S1.† The height of colored boxes indicates the full width at half maximum (FWHM) of Ti³⁺-defect band (0.88 ± 0.10 eV) of “leaky” TiO₂ quoted from ref. 36, and of the Mn intermediate bands (Mn 3d_5/2) derived from the valence spectra shown in (a). The energetic error from XPS analyses is ±0.10 eV.

Based on the valence XPS analyses, we assigned the potentials of band edges, Fermi levels, and IBs with respect to the RHE scale (right vertical axis), shown under flat-band conditions (Fig. 4b). At the liquid interface, the CB and VB edges of TiO₂ were located at −0.05 and 3.29 V_RHE, respectively.⁴⁰ For (Ti,Mn)O_x the Mn IB center was 1.17 ± 0.11 eV above the VB edge or at 2.31 ± 0.11 V_RHE. Under applied potentials, their relative positions were considered fixed.

Fig. 5a (“leaky TiO₂”) was derived from a recent report (see ESI Discussions S3.1†), which was supported by direct in situ ambient-pressure XPS of the liquid junction interface of “leaky” TiO₂ and KOH(aq).⁴⁰ The band bending in the TiO₂ is due to the positive and immobile space charge of their oxygen vacancies. The charge density of Mn₂O₃ crystallites, due to the point defects in Mn₂O₃ bixbyite, was at the degenerately p-type doping level due to the Mn vacancy defects in this class of p-type oxides. The Mn IB was always partially filled with electrons that were transferred from nearby rutile TiO₂ crystallites. The high charge density of partially-filled Mn IB (>10²¹ cm⁻³) was supported by the 47–50 Ω resistance of (Ti,Mn)O_x for a range of Mn composition (26–51 at% Mn). The stoichiometric TiO₂ achieved by ALD²³ and the lower formation energy of Mn₂O₃ than TiO₂ (ref. 41 and 42) ensured the metallic nature of the Mn IB states after annealing. The Mn₂O₃ nanocrystallites formed a conductive network throughout the (Ti,Mn)O_x coating.

Fig. 5 Interfacial band energetics of semiconductive electrocatalysts. (a) “Leaky” TiO₂ and (b) 4 : 1 (Ti,Mn)O_x films. The left panel of each diagram was drawn under flat band conditions, and the right panel of each diagram was drawn where the electrode was poised at +1.8 V_RHE during water oxidation. Box height was drawn based on the estimated top edge of the Ti³⁺-defect band or the Mn³⁺ intermediate band.

Under bias, the band bending for TiO₂ and Mn₂O₃ follow their respective properties of charge density and band positions at the liquid interface. As shown in Fig. 5a, applying +1.8 V_RHE creates a band bending potential profile. This is because the ionic charge density of 10²¹ to 10²³ cm⁻³ is typically observed at the TiO₂-liquid interfaces.⁴³ For the “leaky” TiO₂ with a Ti³⁺-defect band, the positive space-charge density was calculated to be 2.1 × 10²⁰ cm⁻³, but its free-electron concentration was less than 10¹⁷ cm⁻³ due to electron trapping by the Ti³⁺-defect band.⁴⁴ In contrast, the metallic Mn₂O₃ should accumulate holes under the applied potential. Accordingly, the Mn IB shifted its potential energy downward to align with the applied potential. As shown in Fig. 5b, at +1.8 V_RHE, the in-gap Mn states shifted to align with the E_F of Mn₂O₃ crystallites of (Ti,Mn)O_x due to their metallic behavior. Higher overpotentials would lead to increased band bending for the TiO₂ host oxide, but a shift in the position of the Mn intermediate band. This may enable other oxidation pathways, resulting in parasitic anodic current and an overall decrease in selectivity.

Under the applied potential of +1.8 V_RHE there exists a small and positive overpotential to the 2e⁻ water oxidation for H₂O₂ formation. Under such a small overpotential, the electron potential of the (Ti,Mn)O_x coatings lies more positive than the formal potential of the H₂O₂/H₂O couple, i.e., 24 mV positive to E(H₂O₂/H₂O). During operation, the energy levels of CB and VB of TiO₂ are fixed, or “pinned”, at −0.05 and 3.29 V_RHE, respectively, and due to the semiconductor-like band bending of the TiO₂ bulk and the “unpinning” behavior of Mn-IB at the liquid-junction interface, the empty states of the Mn IB can precisely align with the energetics of H₂O₂ producing intermediates. Under a higher applied potential, i.e., +2.3 V_RHE, there would exist a moderate overpotential for the 2e⁻ water oxidation, but such a potential also enables the other pathways including 1e⁻ water oxidation to produce radicals, leading to an overall reduction in H₂O₂ production rates due to the reduced selectivity. Thus, the strategy for tuning the intermediate band towards highly selective electrocatalysis is best suited for the desired low-overpotential electrolysis and other strategies are therefore necessary for increasing geometric current density and, eventually, turnover frequency per site.

4. Conclusion

This study offered a theory-supported design of electrocatalytic surfaces that favor a H₂O₂ pathway. The authors design a surface that does not exhibit favorable kinetics for O₂ evolution, over-oxidize H₂O₂, nor disproportionate the as-produced H₂O₂. The mixed TiO₂/Mn₂O₃ nanocrystallites are chemically stable and protected by the TiO₂-rich surface. (Ti,Mn)O_x coatings have been shown to be stable over 200 hours in modestly acidic and modestly basic solutions suggesting that those (Mn,Ti) sites are not as labile and soluble as the ordinary MnO_x sites. Their acid stability is comparable to the ordered ternary phases of recently reported Mn-based acid stable catalysts.⁴⁵

This study elucidates a new strategy of introducing charge-transport IBs and led to the design of H₂O₂-producing surfaces that oxidize water to H₂O₂ with high selectivity at very low overpotentials. The Mn³⁺-impurity IBs were created within the otherwise forbidden bandgap of TiO₂ which enhanced the electrical conductivity across the coating and matched the electronic states for H₂O₂-producing surface intermediates. Given the growing literature related to H₂O₂ selective catalysis, this approach is expandable to a range of host oxides that prevent H₂O₂ disproportionation to favor 2e⁻ H₂O₂ electrochemistry. However, this strategy is limited to the low overpotential regime, which limits the system to low current densities and requires further molecular or materials discovery for increasing turn-over frequency per active site.¹¹

Author contributions

S. H. conceived the project. J. L., S. H., and Q. Z. designed the experiments. X. S. fitted and analyzed the valence XPS. J. L., G. C., and D. S., performed catalytic reactions and titration experiments. J. L., D. S., and Q. Z. performed XRD, XPS, and TEM characterizations and data analyses. J. L., S. H., X. S., Q. Z., and D. S. analyzed the data and co-wrote the manuscript. All co-authors discussed the results, finalized the conclusions, and edited the manuscript.

Conflicts of interest

Yale University has filed a provisional U.S. patent application directly related to the work described in this paper (patent application no. 62/843856, filed on 6 May 2019).

Acknowledgements

The authors would like to thank Dr Min Li at Yale's Materials Characterization Core (MCC) for his invaluable help with SEM, GIXRD, and XPS characterizations. The authors would like to acknowledge the start-up support from the Tomkat Foundation. J. L. would like to acknowledge the China Scholarship Council (CSC) for its financial support and the Donghua University Doctoral Innovation Fund Program (17D310606, 106-06-0019058).³

References

1. R. Ciriminna, L. Albanese, F. Meneguzzo and M. Pagliaro, ChemSusChem, 2016, 9, 3374–3381.
2. J. M. Campos-Martin, G. Blanco-Brieva and J. L. G. Fierro, Angew. Chem., Int. Ed., 2006, 45, 6962–6984.
3. R. Hage and A. Lienke, Angew. Chem., Int. Ed., 2006, 45, 206–222.
4. M. Wehner, R. L. Truby, D. J. Fitzgerald, B. Mosadegh, G. M. Whitesides, J. A. Lewis and R. J. Wood, Nature, 2016, 536, 451–455.
5. K. Kosaka, H. Yamada, K. Shishida, S. Echigo, R. A. Minear, H. Tsuno and S. Matsui, Water Res., 2001, 35, 3587–3594.
6. S. Fukuzumi, Y. Yamada and K. D. Karlin, Electrochim. Acta, 2012, 82, 493–511.
7. G. Goor, J. Glenneberg, S. Jacobi, J. Dadabhoy and E. Candido, Ullmann's Encycl. Ind. Chem., 2019, 1–40,  DOI:10.1002/14356007.a13_443.pub3.
8. D. W. Flaherty, ACS Catal., 2018, 8, 1520–1527.
9. S. Siahrostami, A. Verdaguer-Casadevall, M. Karamad, D. Deiana, P. Malacrida, B. Wickman, M. Escudero-Escribano, E. A. Paoli, R. Frydendal and T. W. Hansen, Nat. Mater., 2013, 12, 1137.
10. Z. W. Seh, J. Kibsgaard, C. F. Dickens, I. Chorkendorff, J. K. Nørskov and T. F. Jaramillo, Science, 2017, 355, eaad4998.
11. J. Tang, T. Zhao, D. Solanki, X. Miao, W. Zhou and S. Hu, Joule, 2021, 5, 1432–1461.
12. S. Mathew, F. Kuttassery, S. N. Remello, A. Thomas, D. Yamamoto, S. Onuki, Y. Nabetani, H. Tachibana and H. Inoue, ChemPhotoChem, 2018, 2, 240–248.
13. Y. Ando and T. Tanaka, Int. J. Hydrogen Energy, 2004, 29, 1349–1354.
14. X. Shi, S. Siahrostami, G.-L. Li, Y. Zhang, P. Chakthranont, F. Studt, T. F. Jaramillo, X. Zheng and J. K. Nørskov, Nat. Commun., 2017, 8, 701.
15. K. Zhang, J. Liu, L. Wang, B. Jin, X. Yang, S. Zhang and J. H. Park, J. Am. Chem. Soc., 2020, 142, 8641–8648.
16. J. C. Hill and K.-S. Choi, J. Phys. Chem. C, 2012, 116, 7612–7620.
17. S. Y. Park, H. Abroshan, X. Shi, H. S. Jung, S. Siahrostami and X. Zheng, ACS Energy Lett., 2019, 4, 352–357.
18. Y. Jiang, P. Ni, C. Chen, Y. Lu, P. Yang, B. Kong, A. Fisher and X. Wang, Adv. Energy Mater., 2018, 1801909.
19. C. Xia, S. Back, S. Ringe, K. Jiang, F. Chen, X. Sun, S. Siahrostami, K. Chan and H. Wang, Nat. Catal., 2020, 3, 125–134.
20. S. R. Kelly, X. Shi, S. Back, L. Vallez, S. Y. Park, S. Siahrostami, X. Zheng and J. K. Nørskov, ACS Catal., 2019, 9, 4593–4599.
21. A. Izgorodin, E. Izgorodina and D. R. MacFarlane, Energy Environ. Sci., 2012, 5, 9496–9501.
22. J. Zhang and Y. Nosaka, J. Phys. Chem. C, 2014, 118, 10824–10832.
23. P. Nunez, M. H. Richter, B. D. Piercy, C. W. Roske, M. Cabán-Acevedo, M. D. Losego, S. J. Konezny, D. J. Fermin, S. Hu, B. S. Brunschwig and N. S. Lewis, J. Phys. Chem. C, 2019, 123, 20116–20129.
24. T. Umebayashi, T. Yamaki, H. Itoh and K. Asai, J. Phys. Chem. Solids, 2002, 63, 1909–1920.
25. V. E. Henrich and P. A. Cox, The Surface Science of Metal Oxides, Cambridge University Press, Cambridge, 1994.
26. R. J. Lad and V. E. Henrich, Phys. Rev. B: Condens. Matter Mater. Phys., 1988, 38, 10860–10869.
27. G. Siddiqi, Z. Luo, Y. Xie, Z. Pan, Q. Zhu, J. A. Röhr, J. J. Cha and S. Hu, ACS Appl. Mater. Interfaces, 2018, 10, 18805–18815.
28. K. L. Pickrahn, A. Garg and S. F. Bent, ACS Catal., 2015, 5, 1609–1616.
29. R. Frydendal, E. A. Paoli, I. Chorkendorff, J. Rossmeisl and I. E. L. Stephens, Adv. Energy Mater., 2015, 5, 1500991.
30. A. Li, H. Ooka, N. Bonnet, T. Hayashi, Y. Sun, Q. Jiang, C. Li, H. Han and R. Nakamura, Angew. Chem., Int. Ed., 2019, 58, 5054–5058.
31. A. Plauck, E. E. Stangland, J. A. Dumesic and M. Mavrikakis, Proc. Natl. Acad. Sci. U. S. A., 2016, 113, E1973.
32. S. Hu, Sustainable Energy Fuels, 2019, 3, 101–114.
33. X. Shi, Y. Zhang, S. Siahrostami and X. Zheng, Adv. Energy Mater., 2018, 1801158.
34. E. Lobyntseva, T. Kallio, N. Alexeyeva, K. Tammeveski and K. Kontturi, Electrochim. Acta, 2007, 52, 7262–7269.
35. Q. Zhu, M. Hinkle, D. J. Kim and J.-H. Kim, ACS ES&T Engg, 2021, 1, 446–455.
36. K. Feng, H. Zhang, J. Gao, J. Xu, Y. Dong, Z. Kang and J. Zhong, Appl. Phys. Lett., 2020, 116, 191903.
37. L. Liu and A. Corma, Nat. Catal., 2021, 4, 453–456.
38. S. Siahrostami, G.-L. Li, V. Viswanathan and J. K. Nørskov, J. Phys. Chem. Lett., 2017, 8, 1157–1160.
39. V. Viswanathan, H. A. Hansen and J. K. Nørskov, J. Phys. Chem. Lett., 2015, 6, 4224–4228.
40. M. F. Lichterman, S. Hu, M. H. Richter, E. J. Crumlin, S. Axnanda, M. Favaro, W. Drisdell, Z. Hussain, T. Mayer, B. S. Brunschwig, N. S. Lewis, Z. Liu and H.-J. Lewerenz, Energy Environ. Sci., 2015, 8, 2409–2416.
41. N. Birkner and A. Navrotsky, Proc. Natl. Acad. Sci. U. S. A., 2017, 114, E1046.
42. M. R. Ranade, A. Navrotsky, H. Z. Zhang, J. F. Banfield, S. H. Elder, A. Zaban, P. H. Borse, S. K. Kulkarni, G. S. Doran and H. J. Whitfield, Proc. Natl. Acad. Sci. U. S. A., 2002, 99, 6476.
43. P. N. R. L. J. Lipkowski, The Electrochemistry Of Novel Materials (Frontiers In Electrochemistry, Wiley-VCH, 1993.
44. S. Hu, M. R. Shaner, J. A. Beardslee, M. Lichterman, B. S. Brunschwig and N. S. Lewis, Science, 2014, 344, 1005.
45. L. Zhou, A. Shinde, J. H. Montoya, A. Singh, S. Gul, J. Yano, Y. Ye, E. J. Crumlin, M. H. Richter, J. K. Cooper, H. S. Stein, J. A. Haber, K. A. Persson and J. M. Gregoire, ACS Catal., 2018, 8, 10938–10948.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: 10.1039/d1ta05451a

‡ Jiahui Li and Devan Solanki contributed equally.

---