Chemicals used was described in the supporting information. Bronze TiO₂ was synthesized by proton exchange in sulfuric acid solution using sodium titanate as precursor.

Synthesis of sodium titanate: The mixture of Na₂CO₃ and P25 (the molar ratio of Na₂CO₃/TiO₂ was 1:4 and 1:2.5, respectively) was fully ground with 2 g of NaCl and 0.5 g of Na₂HPO₄·12H₂O for 20 min, and calcined in Muffle furnace at 825 ℃ for 8 h. The sintered samples were washed and freeze-dried. Finally, two different types of sodium titanate were obtained (Na₂Ti₆O₁₃ and Na₂Ti₃O₇, Figure S1).

Synthesis of bronze TiO₂: First, 30 mL of 0.02 M H₂SO₄ solution with 0.05 g of Na₂Ti₆O₁₃ powder were transferred to a 50 mL polytetrafluoroethylene reactor and heated at 170 ℃ for 96 h. After cooling to room temperature, the supernatant was removed, and the white sediment was washed with deionized water for several times until the solution was neutral. After filtration, the solution was freeze-dried. The obtained sample was named TiO₂(B)-1. Secondly, 500 mL of 0.02 M H₂SO₄ solution with 0.05 g of Na₂Ti₃O₇ powder were placed in a beaker and stirred at room temperature for 72 h. After filtering, repeat washing with deionized water for several times until the solution shows neutral. The freeze-dried sample was further annealed (450 ℃, 2 h), and finally the sample was named TiO₂(B)-2.

The plasma-catalysis experimental apparatus as shown in Scheme 1, which is made up of four important parts: pulsed plasma generator, plasma-catalysis reactor, cooling circulating water and gas. The parameters of the pulsed plasma generator can be adjusted, in this study, the output pulse voltage was set as ± 3 kV, with a pulse frequency of 7 kHz and a pulse width of 3 μs. Two tungsten rods (diameter: φ = 3.0 mm, gap: 0.5 mm) are connected to both ends of the double-layer plasma reactor, which serve both as high voltage electrodes and as gas inflow channels. Prior to turning on the pulsed plasma generator, 100 ml of ultrapure water and 20 mg of catalyst were sonically dispersed in an ultrasonic pulverized for 30 min and then added to the plasma reactor. High purity gas (99.999%) was continuously added to the plasma reactor at a fixed flow rate of 1 L/min throughout the experiment. And the cooling circulating water maintains the temperature inside the reactor at 10 °C. The concentration of H₂O₂ were characterized using a UV-2600 Shimadzu UV spectrophotometer, refer to the support information for detailed testing procedures.

The concentration of H ions of solution was measured with a FE28-Standard Mettler pH meter (range: 0.00~14.00 pH, resolution: 0.01 pH, accuracy: ± 0.01 pH). The conductivity of the solution was collected using a LAQUA ES-71G act conductivity meter. Optical emission spectroscopy signal (OES) of liquid phase plasma were captured with a NOVA-EX 400-001-5685 Ocean Optics' fiber optic spectrometer. Electron spin resonance (ESR) signals of 5,5-dimethyl-1-pyrroline-n-oxide (DMPO) spin-trapped radicals were recorded using a Bruker EMXnano spectrometer at 298 K and 9.63 GHz. Using GWINSTEK GDS-2202A hybrid digital oscilloscope to acquire the values of pulse current and pulse voltage of liquid phase plasma at a high-speed sampling rate of 2 GS/s.

First, the TiO₂ catalyst was placed in an in-situ infrared diffuse reflectance cell. Under an argon gas flow, the catalyst was heated to 120°C to remove surface-adsorbed impurities. The infrared spectrum collected at this stage served as the background. Subsequently, high-purity Ar was bubbled through a porous gas washing bottle containing 30% H₂O₂, introducing H₂O₂ vapor into the gas-solid infrared diffuse reflectance cell. After flowing the gas for 30 minutes, UV light was turned on to irradiate the surface of TiO₂, and infrared spectra were collected at various illumination times.

We investigated the structural evolution of sodium titanate precursors and their transformation to bronze TiO₂. Figure 1a shows the XRD pattern of the precursor Na₂Ti₆O₁₃ prepared by the molten salt method. The diffraction peaks at 2θ = 11.84°, 14.10°, 24.48°, 30.11°, 33.48°, 43.28°, 44.25°, and 48.58° correspond to the (200), (-201), (110), (-203), (402), (602), and (020) planes of monoclinic Na₂Ti₆O₁₃ (PDF#73-1398) with lattice constants a = 1.5131 nm, b = 0.3745 nm, and c = 0.9159 nm. In contrast, Figure 1b presents the XRD pattern of the precursor Na₂Ti₃O₇, also prepared by the molten salt method. The diffraction peaks at 2θ = 10.52°, 15.84°, 25.68°, 43.88°, and 66.87° correspond to the (100), (101), (110), (104), and (124) planes of monoclinic Na₂Ti₃O₇ (PDF#72-0148) with lattice constants a = 0.8571 nm, b = 0.3804 nm, and c = 0.9135 nm. After proton exchange and dehydration, these sodium titanate phases undergo topotactic transformation to bronze TiO₂. In Figure 1c, the diffraction peaks at 2θ = 14.19°, 15.20°, 17.44°, 24.93°, 28.61°, 29.70°, 30.67°, 33.19°, 43.50°, 44.50° and 48.53°correspond to the (001), (200), (20-1), (110), (002), (40-1), (400), (310), (003), (60-1), and (020) of monoclinic bronze TiO₂ (JCPDS 46-1237), respectively. The sharp and narrow peaks indicate that TiO₂(B)-1 has good crystallinity and high purity. In contrast, TiO₂(B)-2 exhibits broad peaks and lacks the characteristic peaks at 2θ = 15.20° and 30.67° corresponding to the (200) and (400) planes of bronze TiO₂. Additionally, the TiO₂(B)-2 shows an unknown diffraction peak at 2θ=12.14°, indicating the presence of impurities. Therefore, XRD characterization indicates that we synthesized and controlled the crystal phase and crystallinity of bronze TiO₂. The Na₂Ti₆O₁₃ precursor can be used to synthesize bronze TiO₂ with higher crystallinity and purity.

The difference between the two bronze TiO₂ samples can be attributed to the crystal structures of the precursors, Na₂Ti₃O₇ and Na₂Ti₆O₁₃. Na₂Ti₃O₇ contains three TiO₆ octahedra connected by shared edges to form strip, which in turn share corners to create a serrated (Ti₃O₇)^2- layer with an open lamellar structure ^[34]. In contrast, Na₂Ti₆O₁₃ features TiO₆ octahedra sharing edges along the c-axis, forming serrated layers that create larger quasi-rectangular tunneling spaces ^[35]. As illustrated in Scheme 2, the larger tunneling structure of Na₂Ti₆O₁₃ can accommodate more amount of H₂O and H₃O⁺, and its stable physical structure can further improve the exchange rate of protons during the hydrothermal reaction. We conducted ICP elemental analysis on the sodium residue in the two bronze TiO₂ samples (Table S1) and found that 98.87% of the sodium in the Na₂Ti₆O₁₃ precursor could be exchanged, whereas only 69.76% of the sodium in the Na₂Ti₃O₇ precursor could be exchanged. The ICP results further confirm that the residual sodium cations in TiO₂(B)-2 lead to non-homogeneous dehydration reactions and partial distortion of the domain structure, resulting in the low crystallinity of TiO₂(B)-2.

The Raman spectra exhibited sharp and intense peaks, indicating a well-ordered long-range structure and regular lattice arrangement. The Raman signals for Na₂Ti₆O₁₃, shown in Figure 1d, match previously reported spectra ^[36]. A vibration peak at 137 cm^-1 corresponds to the O-Ti-O bond, while characteristic peaks at 194 and 275 cm^-1 are associated with the Na-O-Ti bond. Within the range of 400 cm^-1 to 900 cm^-1, various stretching vibration peaks of the TiO₆ octahedron are observed, including a peak at 480 cm^-1 corresponding to the Ti-O-Ti bond. The peak at 871 cm^-1 is attributed to the stretching vibration of the short Ti-O bond coordinated with sodium ions and non-bridging oxygen. Figure 1e displays the Raman data for Na₂Ti₃O₇, with a vibration peak at 147 cm^-1 corresponding to the O-Ti-O bond, and characteristic peaks at 192 and 269 cm^-1 for the Na-O-Ti bond. Similarly, within the 400 cm^-1 to 900 cm^-1 range, various stretching vibration peaks of the TiO₆ octahedron are present, including a peak at 401 cm^-1 for the Ti-O bond and a peak at 835 cm^-1 for the short Ti-O bond coordinated with sodium ions and non-bridging oxygen ^[37]. Notably, a Raman signal at 101 cm^-1 in Na₂Ti₃O₇ is absent in Na₂Ti₆O₁₃, highlighting structural differences between the two materials.

After proton exchange and dehydration, all Raman bands in Figure 1f of both samples are consistent with the vibrational modes Ag (144, 363.4, 405.4, 435, 472, 554, 634.7 cm^-1) and Bg (122, 198.7, 238.5, 253, 663.4 cm^-1) of bronze TiO₂ ^[38]. Compared to TiO₂(B)-1, the Raman peaks of TiO₂(B)-2 are broader and less defined, further confirming the superior crystallinity of the TiO₂(B)-1 material.

We further characterized the morphology and structure of the obtained bronze TiO₂ using SEM, TEM, and HRTEM. Figures 2a-d show SEM images of the prepared bronze TiO₂, revealing a one-dimensional structure the same to that of precursor (Figure S2). Both bronze TiO₂ exhibit a nanobelt morphology, consistent with their precursors. This consistency is attributed to the topotactic transformation from sodium titanate to TiO₂, where the monoclinic phase of sodium titanate favors the formation of monoclinic bronze TiO₂^[32]. The diameter of TiO₂(B)-1 nanobelts ranges from 20 to 100 nm, with lengths extending to several micrometers. In contrast, the TiO₂(B)-2 nanobelts have diameters ranging from 50 to 500 nm and lengths up to 30 μm. By comparison, the TiO₂ nanobelts derived from Na₂Ti₃O₇ display less uniformity compared to those from Na₂Ti₆O₁₃. This discrepancy can be partially explained by the incomplete proton exchange in Na₂Ti₃O₇ and the slightly lower lattice match between Na₂Ti₃O₇ and bronze TiO₂, leading to uneven internal stress during the topotactic transformation and resulting in less ordered TiO₆ octahedral arrangements. High-resolution TEM results corroborate these observations. As shown in Figure 2e-h, TiO₂(B)-1 exhibits a clearer lattice arrangement with a better-defined shape and smoother surface, indicative of its higher crystallinity. The lattice spacings of 0.64 nm for TiO₂(B)-1 and 0.62 nm for TiO₂(B)-2 correspond to the (0 0 1) crystal plane of bronze TiO₂. The slight differences in observed lattice spacings are likely due to variations in the unit cell parameters of the sodium titanate precursors. The element mapping images of TiO₂(B)-1 in Figure 2i,j confirms uniform distribution of Ti and O along the nanobelts. In Figure 2k, the well aligned spots in SAED pattern of an individual TiO₂ nanobelts proves the highly crystalline nature of the TiO₂(B)-1 nanobelt. In contrast, Figure 2l displays bright spots and a series of faint rings, suggesting that TiO₂(B)-2 is partially crystallized. This crystallinity difference between TiO₂(B)-1 and TiO₂(B)-2 agrees with the XRD and Raman results in Figure 1.

We optimized the plasma discharge parameters for H₂O₂ synthesis via solution plasma by varying the gas environment (no gas (Wo), Ar, O₂). As shown in Figure 3a, the presence of gases significantly increased the yield of H₂O₂. Specifically, the total yield increased by 5.42 times with O₂ and 3.21 times with Ar compared to the absence of gas. Additionally, the yield with O₂ was 1.68 times higher than with Ar. Accordingly, O₂ bubbling significantly enhances the production of H₂O₂ through liquid-phase plasma discharge. Figure 3b shows a photo of the discharge spark during the liquid-phase plasma system. In the absence of gas, weak filamentary discharge channels form between the electrodes, resulting in a narrow plasma discharge area. However, the presence of gas significantly expands the plasma area. The discharge channel range is notably enhanced by the rupture of gas bubbles. Water molecules, being highly electronegative, absorb many electrons in the initial discharge stage, making it difficult to form an electron avalanche. Consequently, a large breakdown voltage is required to directly breakdown water and produce plasma. Gas bubbles, however, greatly reduce the breakdown voltage, resulting in a more intense high-energy plasma.

We further characterized and analyzed the effect of the gas on the plasma combining a digital oscilloscope and a fiber optic spectrometer to acquire the pulse current, pulse voltage signal, and emission spectrum of the atoms during the discharge. A 10 Ω resistor was connected in series within the plasma output circuit, and the actual pulse voltage and current generated during the discharge process were monitored using an oscilloscope. As shown in Figure 3c-e, the discharge of the system without gas is unstable, with significant fluctuations in the current value. After introducing gas, both current and voltage values increased, with voltage rising from ±1600 kV to ±2200 kV and current from 800 A to 2000 A. This indicates that the increase in H₂O₂ production after gas injection is partly due to changes in the discharge. The higher peak pulse voltage increases the energy injected into the electrode gap, raising the density of high-energy electrons and thus the frequency of electron collisions with ground-state neutral species, leading to the excitation of more active species and promoting H₂O₂ production.

We diagnosed the active species in the excited state within the plasma by optical emission spectroscopy (OES). To quantitatively compare the intensity of the emission spectra, we normalized the data based on the sodium atomic emission intensity of 100 μL of 10 M NaOH. As shown in Figure 3f, in the absence of gas, the plasma primarily ionizes to excite H₂O molecules, with Hα atomic emission lines at 656.6 nm, O (3P⁵P⁰→3S⁵3S⁰) at 777.5 nm and 845.2 nm (3P³P⁰→3S³S⁰), and OI atomic emission lines at 927.1 nm ^[39]. Figure 3g shows that when Ar is added, atomic emission spectra belonging to the 4P-4S orbitals of Ar appear in the range of 700-900 nm, along with increased intensity of the Hα and OI atomic emission spectra ^[40]. The presence of a large number of metastable Ar species with high excitation energy and long lifetime in the liquid increases the collision frequency of neutral gas molecules with free electrons, further promoting water ionization. As shown in Figure 3h, the most intense emission of oxygen atoms is observed when the system is filled with O₂ compared to Ar, indicating a significant increase in the content of oxygen-active species within the solution. Additionally, the collision reaction of water molecules with excited-state oxygen atoms (O(ID)) can generate ∙OH, further promoting the synthesis of H₂O₂.

Besides, we have explored the effects of other gases, such as N₂ and CO₂, on H₂O₂ synthesis during the plasma discharge process. For the case of N₂ atmosphere, discharge in N₂ introduces nitrate and nitrite ions into the solution. The accumulation of these ions increases the solution conductivity, which destabilizes the plasma discharge and makes it unsustainable for H₂O₂ production. For the case of CO₂ atmosphere, similarly, CO₂ discharge leads to the formation of carbonate ions, which also raise the solution conductivity, resulting in unsustainable plasma discharge conditions.

We have explored several plasma parameters on the production of H₂O₂, as shown in Figure S3. Increasing the pulse voltage enhances the energy input into the plasma, which promotes water dissociation and radical formation, thereby increasing H₂O₂ yield. However, when voltage is higher than 3 kV, the yield of H₂O₂ is hardly increased. The frequency of the pulse affects the rate at which radicals are generated and recombined, optimizing the conditions for H₂O₂ synthesis. Adjusting the pulse width balances the plasma discharge duration and relaxation period, impacting the stability and accumulation of H₂O₂. Through systematic optimization, we identified 3 kV, 7 kHz and 3 μS as the optimal conditions for maximizing H₂O₂ yield.

We further constructed a bronze TiO₂ mediated plasma-catalytic system to produce H₂O₂ under oxygen bubbling conditions. As shown in Figure 4a, TiO₂ nanobelts can be uniformly dispersed in water to form a slurry, ensuring sufficient contact between the plasma and TiO₂. The introduction of oxygen bubbles further promotes the dispersion of the catalyst and plasma discharge, forming a solid-liquid-gas-plasma four-phase interface that maximizes the reaction efficiency. Evaluation of the H₂O₂ yield is shown in Figure 4b, where all TiO₂ catalysts effectively promoted H₂O₂ synthesis compared to a single plasma system. Among them, the plasma-catalysis system using TiO₂(B)-1 nanobelts had the highest yield, increasing by 352.2 μmol/L (P25), 642.6 μmol/L (TiO₂(B)-2), and 696.8 μmol/L (TiO₂(B)-1), respectively. We found that the high crystallinity TiO₂(B)-1 nanobelts could effectively synergize with the liquid-phase plasma discharge technique to prepare H₂O₂. Remarkably, the maximum H₂O₂ concentration reached 3.5 mmol/L, which is 1-3 orders of magnitude higher than conventional TiO₂ photocatalytic H₂O₂ production methods.

The activity of as-synthesized high crystallinity TiO₂(B)-1 nanobelts outperforms TiO₂(B) nanosheets and nanotube synthesized according to the literature ^[41,42], as well as single phase of anatase or rutile TiO₂ (Figure S4). Moreover, the H₂O₂ concentration achieved by gas-phase plasma discharge is lower than that of solution plasma discharge. Even with array electrodes in gas-phase plasma discharge, the accumulation rate of H₂O₂ only reached 0.95 mM per hour, which is still below the performance of our solution plasma discharge method (Figure S5). We also conducted five long-term stability tests on the TiO₂ (B) system, and all results showed no significant decrease in activity. This indicates that the bronze TiO₂ system exhibits stable H₂O₂ accumulation activity during catalytic processes (Figure S6). These results reveal the efficiency of photocatalyst-mediated solution plasma for H₂O₂ synthesis, demonstrating its potential as a highly effective method for H₂O₂ production.

To analyze the main reaction pathways of the plasma-catalysis system and the effect of solution changes on products, we conducted comparative experiments using TiO₂(B)-1 nanobelts at different pH levels. As shown in Figure 4c, alkaline solutions were more favorable for H₂O₂ synthesis. The yield of H₂O₂ at pH = 10.5 is 1.28 times higher than at pH = 7 and 1.7 times higher than at pH = 3.5. Since the alkaline solution provides larger amount of OH^-, we believe that the production of H₂O₂ is partially contributed by OH^- excited species.

We further examined the role of hydroxyl radicals (·OH) in the production of H₂O₂, serving as a crucial indicator for evaluating the production mechanism of H₂O₂. To quantify the ·OH content in the liquid phase, we employed DMPO as a trapping agent. We designed a micro plasma reactor (Figure 4d) to test the instantaneous radical content. Initially, 100 mL of deionized water and 0.020 g of catalyst were uniformly dispersed using an ultrasonic homogenizer. A 2 mL aliquot of the homogenized sample was then placed in the reactor. During discharge, 55 μL of DMPO was added, and the liquid was subsequently placed into a capillary tube to measure the trapped ·OH content. Figure 4e presents the ESR spectra of DMPO-·OH for different samples introduced into the plasma system. In a pure water environment, a characteristic ·OH signal with an intensity ratio of 2:2:1 was detected ^[43]. The trend in ·OH content across different systems is consistent with the H₂O₂ yield results. The introduction of catalysts enhanced the ·OH signal intensity, with the trend in ·OH content across different systems consistent with the H₂O₂ yield results. Furthermore, the pH-dependent catalytic activity suggests that the higher H₂O₂ yield under alkaline conditions likely originates from the abundant OH^- ions, which combine with high-energy electrons in the plasma to generate a substantial amount of ·OH radical.

Figure 4f illustrates the OES spectra of the plasma-catalysis system. The spectral intensity of the plasma was significantly enhanced after the addition of P25, TiO₂(B)-1 and TiO₂(B)-2 compared to the catalyst-free system. Notably, the plasma-catalysis system using TiO₂(B)-1 nanobelts exhibited the strongest atomic emission intensity (Table S2). The Hα and O atomic emission spectral lines indicated the densities of H and O species, respectively. The enhanced Hα signal peak confirmed that the increased discharge intensity led to greater dissociation of water molecules, while the enhanced OI signal peak indicated that more oxygen atoms were excited and further participated in the synthesis of H₂O₂^[44]. This luminescence enhancement is associated with the promotion of plasma discharge by catalyst.

We also monitored the electrical parameters of the plasma generated during the discharge. Figure S7 shows that the voltage increased from approximately ±1900 V to about ±2100 V, and the current increased from 1600 A to 2000 A with the addition of the catalyst. The TiO₂(B)-1 nanobelts induced the largest current and voltage during the solution plasma discharge, which correlated with the strongest luminescence observed. Therefore, bronze-phase TiO₂ enhances plasma discharge, which significantly promotes water dissociation. The primary radical generated during water dissociation is ·OH. While plasma discharge also activates oxygen molecules, our comparative tracking of singlet oxygen and superoxide radicals showed no significant increase in their capture signals. Therefore, we conclude that ·OH are the dominant reactive species, consistent with most reports on plasma-catalytic H₂O₂ synthesis ^[45].

To evaluate the passive photocatalytic performance of the catalysts in solution plasma, a comparative experiment was conducted. The experimental conditions included a UV light intensity of 55 mW/cm², 0.020 g of catalyst, an O₂ flow rate of 1 L/min, and 20 mL of methanol solution as the sacrificial agent. As shown in Figure 4g, the photocatalytic H₂O₂ yields were highest for TiO₂ (B)-1 nanobelts (111.42 μmol/L), followed by TiO₂ (B)-2 nanobelts (94.101 μmol/L), and P25 nanoparticles (68.14 μmol/L). Bronze TiO₂ exhibited superior catalytic effects compared to other crystalline phases, with TiO₂ (B)-1 nanobelts achieving the highest H₂O₂ yield, 1.63 times greater than that of P25 nanoparticles. We compared the light absorption properties of different TiO₂ samples, as shown in Figure S8. We found no consistent correlation between light absorption capacity and catalytic activity. That is, highly crystallized bronze TiO₂ exhibited moderate light absorption intensity-lower than P25 but higher than weakly crystallized bronze TiO₂-yet demonstrated the highest catalytic activity. Thus, we believe that light absorption is only one factor influencing photocatalytic effects, while surface H₂O₂ adsorption capacity and interaction with plasma discharge play more critical roles.

When H₂O₂ does not dissociate immediately from the catalyst surface, electrons migrating to the surface can reduce H₂O₂ to H₂O. Additionally, TiO₂ surfaces form Ti-OOH complexes with H₂O₂, promoting its decomposition under visible light irradiation. To investigate the diffusion behavior of H₂O₂ across different crystalline phases and exposed surfaces, we designed experiments to observe H₂O₂ decomposition under UV light. Figure 4h shows the decomposition effect of H₂O₂ with the catalyst under UV irradiation. Upon UV exposure, H₂O₂ decomposed gradually. In pure water, the decomposition rate was 10.52%, while both TiO₂-based catalysts exhibited slow decomposition rates. Notably, P25 showed the highest H₂O₂ decomposition ability, achieving an 81.27% decomposition rate under UV light. These differences are attributed to the exposed (001) crystal plane of bronze TiO₂, which facilitates excellent H₂O₂ desorption, leading to a decrease in the decomposition rate. This can be also supported by the results in the literature ^[30]: On one hand, the diffusion coefficient of H₂O₂ on the surface of bronze-phase TiO₂ is significantly higher than that on anatase-phase TiO₂. This indicates that H₂O₂ desorbs more readily from the bronze-phase TiO₂ surface, which is beneficial for suppressing its decomposition. On the other hand, the adsorption heat of H₂O₂ on bronze-phase TiO₂ (11.3 kJ/mol) is notably lower than that on bronze/anatase TiO₂ heterojunctions (19.4 kJ/mol). A higher adsorption heat corresponds to stronger adsorption of H₂O₂ on the TiO₂ surface. Therefore, the lower adsorption heat of H₂O₂ on bronze-phase TiO₂ compared to anatase-phase TiO₂ suggests weaker adsorption of H₂O₂. Besides, we conducted a radial distribution function analysis of the O-O distance distribution from the AIMD results of H₂O₂ on the TiO₂ surface. The molecular dynamics simulation of H₂O₂-TiO₂(B) and H₂O₂-anatase TiO₂ configuration are depicted in Figure 5a and c. Figure S9 indicates that their kinetic structural relaxations reach a steady state. The simulation results are displayed in Figure 5b. The peak around 1.3 Å corresponds to the O-O bond length in H₂O₂. By analyzing the g(r)-r relationships, we can effectively compare the stability of H₂O₂ molecules on different TiO₂ surfaces. Upon comparison, we observe that the proportion of stable H₂O₂ molecules on the TiO₂(B) surface is significantly higher than that on the anatase phase of TiO₂, which is revealed by a higher value of g(r) for TiO₂(B). This indicates that after kinetic relaxation, TiO₂(B) retains more H₂O₂ on its surface, which is more favorable for stabilizing H₂O₂.

To further investigate the regulation of H₂O₂ synthesis kinetics by the TiO₂ catalyst, we performed a pre-saturation of H₂O₂ adsorption onto the TiO₂ samples. Following this, ultraviolet light was applied, and in-situ infrared spectroscopy was used to track the H₂O₂ adsorption and desorption capabilities of different samples under UV illumination. As shown in Figure S10a, after the TiO₂ (B) sample adsorbed H₂O₂, it exhibited two distinct types of infrared absorption bands. We regarded the bands near 3450 cm^-1 as a ν(OH) mode of isolated H₂O ^[46]. Additionally, we observed two weak absorption bands around 3600 and 3700 cm^-1, which correspond to the infrared absorption bands of H₂O₂, indicating weak adsorption of H₂O₂ on the TiO₂(B) surface. After UV illumination, the intensity of the H₂O₂ absorption bands rapidly decreased, indicating desorption of H₂O₂ from the TiO₂(B) surface. When the sample was switched to P25 TiO₂, as shown in Figure S10b, we observed two very strong absorption bands at 3600 and 3700 cm^-1, which are characteristic absorption bands of H₂O₂ ^[47]. Notably, these bands did not show significant attenuation after UV illumination, demonstrating strong adsorption of H₂O₂ on P25 TiO₂, with no desorption occurring under UV light. This irreversible adsorption of H₂O₂ on P25 TiO₂ is unfavorable for the accumulation of H₂O₂ in solution. This further confirms our previous observations that P25 TiO₂ readily photocatalyzes the decomposition of H₂O₂, while TiO₂(B) does not facilitate such decomposition.

In our study, significant differences were observed in the photocatalytic decomposition of H₂O₂ on the surfaces of bronze and P25 TiO₂. While the bronze surface inhibited H₂O₂ decomposition, the P25 TiO₂ surface did not exhibit this behavior. To further investigate the unique properties of the bronze surface, we coated both materials with carbon. The carbon coating serves to isolate H₂O₂ from direct contact with TiO₂, thereby inhibiting its decomposition. The carbon coating method is depicted in Figure S11. Raman spectroscopy was employed to characterize the carbon-modified TiO₂ (B) samples. Figure S12 shows the Raman spectra of TiO₂ (B)-1@C nanobelts. The spectra revealed two broad peaks corresponding to carbon nanostructures: the D band at 1364 cm^-1, indicative of graphite carbon edge vibrations, and the G band at 1585 cm^-1, corresponding to the in-plane stretching vibrations of graphite crystals^[29]. XRD analysis confirmed that the crystal lattice structure of TiO₂(B) remained unchanged after hydrothermal carbonization. UV-Vis diffuse reflectance spectra showed that TiO₂(B)-1 turned from white to brown after carbonization, exhibiting strong light absorption in the 400-800 nm visible region. XPS spectra also reveals the existence of carbon related peaks on TiO₂ (Figure S13). Combining the XRD, Raman data and XPS, we successfully demonstrated the loading of carbon on the material surface. This result was also validated through surface analysis of carbonized P25 (Figure S14). Figures 6a and b further confirmed the presence of a carbon layer on the TiO₂ surface. After carbon coating, the H₂O₂ content mediated by TiO₂ in solution plasma catalysis increased further (Figures 6c, d). Photocatalytic experiments revealed that the carbon coating further inhibited the decomposition of H₂O₂(Figures 6e, f). Notably, the photocatalytic decomposition of H₂O₂ on the carbon-coated P25 surface was almost equivalent to that on bronze TiO₂, indicating that the carbon coating mitigated the disadvantage of P25 in H₂O₂ decomposition. We compared the pH and conductivity of the liquid-phase discharge system when using TiO₂@C as a catalyst and observed no significant differences (Figure S15). This aligns with the similar H₂O₂ accumulation capabilities of bronze TiO₂ and P25 after carbon loading. Additionally, this indirectly supports the unique characteristics of the brookite TiO₂ surface. These findings demonstrate the importance of surface crystalline composition in the photocatalytic decomposition of H₂O₂ and provide valuable insights for designing TiO₂ or other catalysts suitable for H₂O₂ production in solution plasma environments.

We also analyzed the carbon loading of two samples and found that both had comparable carbon content (ca. 50 atomic%). However, their specific surface areas showed a significant difference, P25 TiO₂ exhibited a specific surface area of 53.5 m²/g, while TiO₂ (B) had only 28.1 m²/g. This indicates that although TiO₂ (B) has a smaller surface area, its carbon coating is likely thicker. Such carbon coating not only affects the UV absorption characteristics of TiO₂ but may also partially weaken TiO₂ (B)'s ability to inhibit H₂O₂ decomposition. The carbon coating layer could alter the surface electronic structure or light absorption behavior, further influencing its catalytic performance.

Based on the above results, the production and decomposition trend of H₂O₂ over bronze TiO₂, P25, TiO₂(B)@C and P25@C was illustrated in Figure 7. During plasma discharge in water, hydroxyl radicals are generated. When bronze-phase TiO₂ is introduced into the solution plasma (Figure 7a), the TiO₂ nanobelt structure acts as a medium, enhancing plasma discharge and luminescence, thereby generating more hydroxyl radicals and promoting H₂O₂ formation via hydroxyl radical pathways. Additionally, TiO₂ absorbs UV photons from the plasma channel, preventing the decomposition of the generated H₂O₂. Photogenerated electrons and holes in TiO₂ can react with oxygen bubbles and water molecules, respectively, to produce H₂O₂, thus improving the utilization of plasma energy. Additionally, the decomposition of H₂O₂ is particularly problematic in complex fields, especially under high-energy plasma conditions. Here, bronze-phase TiO₂ not only facilitated the photocatalytic generation of H₂O₂ but also inhibited its decomposition. We improved passive photocatalysis in the solution plasma channel through three routes. Firstly, by adjusting the crystalline phase of TiO₂ (Figure 7a,c), it was found that mixed-phase TiO₂ (anatase/rutile) significantly promoted the photocatalytic decomposition of H₂O₂, attributed to the stronger adsorption of H₂O₂ on the anatase surface. In contrast, bronze-phase TiO₂ significantly inhibited the photocatalytic decomposition of H₂O₂. Secondly, by optimizing the crystallinity and purity of the bronze-phase TiO₂, higher-quality growth resulted in enhanced photocatalytic activity, favoring H₂O₂ generation. Thirdly, a uniform carbon layer is coated on the surface of TiO₂ (Figure 7b,d). The carbon coating can further inhibit the adsorption of H₂O₂ on the catalyst surface, which also facilitates the desorption of H₂O₂ from the surface. The moderate strength of adsorption and desorption mediated by the carbon layer promotes the accumulation of H₂O₂ concentration during plasma discharge catalysis.

Last, note that the enhancement of plasma catalytic activity by TiO₂ compared to pure water is not expected several times in other catalytic system, which can be attributed to the inherent constraints of oxide systems. High concentrations of H₂O₂ tend to react with oxide surfaces, forming peroxides and reducing catalytic efficiency. Additionally, the intense energy of plasma is difficult to control, meaning the improvement in catalytic activity is less pronounced than in other catalytic processes, such as electrocatalysis or photocatalysis, which often achieve orders-of-magnitude enhancements ^[48,49]. In plasma-catalyst interactions, even a 50% increase in catalytic activity is considered a significant improvement due to the challenges of synergistic effects between plasma and catalysts. Additionally, the plasma catalysis method currently faces challenges such as high energy consumption. However, embracing renewable energy sources and utilizing off-grid options like surplus wind, solar, and electricity can mitigate these issues. Moreover, modular power combinations can facilitate scaling up for industrial production. In terms of H₂O₂ production, plasma catalysis achieves concentrations several orders of magnitude higher than single photocatalysis. While it still lags behind electrocatalysis, plasma catalysis eliminates the need for electrolytes and subsequent separation processes. Therefore, despite current challenges like energy consumption, its advantages-such as eliminating electrolytes and achieving high H₂O₂ concentrations-highlight its potential as a scalable and eco-friendly solution, particularly when integrated with renewable energy systems.