Cellobiose and glucose photorefining over non-noble Bi⁰-modified TiO₂ with oxygen vacancies: unraveling the effects of lignocellulosic derivatives and oxidation mechanism

Abstract

Solar-driven biomass refining holds sustainable potential for producing value-added chemicals under mild conditions. However, the efficiency of current advanced biomass photorefining systems is often hampered by their reliance on precious metal-based cocatalysts and harsh reaction conditions (e.g., alkaline conditions or organic solvents). To address these limitations, we report an in situ constructed non-noble Bi⁰-modified TiO₂ catalyst with abundant oxygen vacancies (OVs) for the efficient photorefining of cellobiose and glucose under mild, neutral conditions. The significantly improved performance stems from the Bi⁰/OV synergy, which enhances visible-light absorption, facilitates charge separation, and promotes the generation of superoxide radicals (˙O₂⁻). We identified that ˙O₂⁻ acts as the driving force for the oxidation pathway and proposed the corresponding ˙O₂⁻-mediated oxidation mechanism. Crucially, we systematically elucidate the effects of inherent derivatives in authentic lignocellulosic hydrolysates (e.g., sugars, phenols, furans, lignin derivatives, and metal ions) on the photorefining of glucose to arabinose, clarifying their roles from a molecular-level perspective. This work not only demonstrates the viability of Bi⁰/V_O-TiO₂ for solar-driven biomass valorization but also establishes a theoretical foundation for designing interference-resistant photorefining processes. It thereby provides new perspectives for enhancing the value of biomass resources under mild, environmentally friendly conditions.

---

Green foundation

1. This work advances green chemistry by designing a non-noble Bi⁰/V_O-TiO₂ photocatalyst for solar-driven sugar refining under mild, neutral conditions, eliminating the need for precious metals and harsh reagents.

2. This system operates in a neutral pure aqueous solution without any organic solvents, achieving >50% glucaric acid selectivity from cellobiose and enabling glucose-to-arabinose conversion.

3. Its green credentials can be further elevated by enhancing catalyst stability to minimize metal leaching, suppressing non-selective oxidation to reduce carbon loss as CO₂, and integrating inhibitor-removal processes to enable the use of real biomass hydrolysates.

Introduction

The depletion of fossil resources and the pressing challenges of climate change necessitate a shift toward green and sustainable chemical production. Lignocellulosic biomass, as the most abundant renewable carbon reservoir on Earth, represents an essential source of feedstock for sustainable biorefineries.¹ Research and technology development related to its energy and resource utilization have emerged as global research priorities. Biomass photorefining (PBR) is a green technology with significant sustainability potential, enabling solar-driven selective valorization of lignocellulosic components into fuels and high-value chemicals under mild conditions. This technology breaks through the thermodynamic constraints of traditional biomass conversion by leveraging reactive species involved in the key process of photocatalysis to reconfigure reaction pathways, delivering distinct green advantages in energy efficiency, process safety, and product value enhancement.^2,3 Cellulose constitutes the architecturally dominant component within lignocellulosic matrices, where its efficient conversion and high selectivity of cellulose represent the pivotal step in holistic lignocellulose valorization, and photocatalysis emerges as a highly promising method.⁴ Nevertheless, the photocatalytic refining kinetics of cellulose are intrinsically constrained by its macromolecular architecture and extensive hydrogen-bonding networks. Therefore, photocatalytic refining systems leveraging model compounds and platform molecules of cellulose have emerged as a pivotal research domain.^5,6 Recent research breakthroughs have confirmed that targeted photoredox catalyst design and reaction engineering enable selective transformation of cellulose-derived sugars into diverse value-added platform chemicals.^7,8 Representative value-added products, such as hydrogen, arabinose, gluconic acid, and glucaric acid, have been efficiently produced from glucose and cellobiose using rationally designed photocatalytic systems based on TiO₂, g-C₃N₄, ZnIn₂S₄, Zn_xCd_1−xS, and related materials.^9–14 Among these, arabinose serves as a distinctive natural regulator of sucrose metabolism, with demonstrated safety and efficacy that underscore its growing commercial potential in health-focused applications.¹⁵ Concurrently, glucaric acid constitutes a premium biomass-derived chemical whose potential application value hinges critically on overcoming high-cost production barriers.¹⁶ While photocatalytic cellulose refining holds promise for selective production of high-value chemicals, its development is hampered by the challenge in concurrently achieving high sugar molecule conversion and product selectivity under mild conditions. Therefore, rational engineering of an effective photocatalytic system provides fundamental strategies for accelerating cellulose photorefining technology evolution.

The performance of selective photocatalytic oxidation systems is critically governed by precisely tailored interactions between reactive oxygen species (ROS), such as hydroxyl radicals (˙OH), superoxide radicals (˙O₂⁻), singlet oxygen (¹O₂), and target molecular substrates.¹⁷ Wang et al. demonstrated that ˙O₂⁻, with its moderate oxidative capacity, serves as a pivotal regulator in effectively steering the photorefining of glucose and cellobiose toward high-value products like arabinose and glucaric acid.^18,19 However, efficient ˙O₂⁻ generation requires specific catalytic sites, such as surface oxygen vacancies (OVs). OVs could facilitate O₂ adsorption and activation, accelerate ˙O₂⁻ generation kinetics, and refine the local interface microenvironment, making them highly effective for photocatalytic oxidation.^20,21 The coordinative unsaturation of OVs also strengthens targeted adsorption of sugar molecules, thereby suppressing non-selective pathways and improving process selectivity.²² Titanium dioxide (TiO₂) possesses strong oxidation capacity, favorable photostability, low cost, and environmental benignity for sugar photo-oxidation, though its performance is hampered by rapid charge recombination and limited visible-light absorption.^23,24 Thus, effective regulation and modification of TiO₂ is essential to enhance photocatalytic oxidation performance. While noble metal co-catalysts (e.g., Pt, Au, and Pd) enhance the selectivity and efficiency of photocatalytic sugar oxidation, their cost and scarcity limit their widespread application in biomass refining, spurring the search for high-performance, Earth-abundant alternatives.^25,26 Zero-valent bismuth (Bi⁰), a non-precious metal, extends visible-light absorption via the localized surface plasmon resonance (LSPR) effect and photogenerated charge recombination by acting as an electron trap.²⁷ Significantly, the concurrent introduction of Bi⁰ and OVs within TiO₂ induces a distinct synergistic effect. Hot electrons from Bi⁰ are captured by OVs to promote O₂ activation, while OVs enhance the SPR response through electronic structure modulation.²⁸ This cooperative interaction can significantly improve both activity and selectivity in sugar conversion, and a similar mechanism has been previously validated in typical metal systems, thereby supporting the feasibility of Bi⁰-OV configurations and presenting a rational design strategy for noble-metal-free catalysts to achieve efficient and selective biomass photo-oxidation.²⁹ Based on this, we assume that reasonable modification of Bi⁰/OVs-TiO₂ can achieve selective photocatalytic conversion of cellobiose and glucose under neutral conditions.

Integrating biomass pretreatment into the photocatalytic cellulose refining process enables preferential lignocellulose deconstruction and cellulose swelling, thereby enhancing mass transfer and light utilization efficiency.³⁰ These improvements contribute to superior photocatalytic performance in the conversion of cellulose to value-added chemicals, albeit with the generation of byproducts such as monosaccharides, furan compounds, organic acids, and lignin-derived fragments.³¹ These compounds may compete for active sites, alter the interfacial properties of the catalyst, or induce surface passivation, thereby influencing the photocatalytic reactions. The current understanding is largely derived from studies employing purified model compounds, which fail to capture the complexity of real biomass-derived matrices and severely hinder the practical application of this technology. Consequently, there is an urgent need to elucidate the molecular-level interactions between photocatalytic surfaces and actual biomass-derived compounds. A systematic investigation into these effects is expected to advance the fundamental mechanistic understanding of photocatalysis in complex media, thereby facilitating the rational design of photocatalytic systems resistant to interference.

In this study, we demonstrate the feasibility of in situ constructing Bi⁰-modified oxygen-rich vacancy TiO₂ catalysts for the photocatalytic oxidation of glucose and cellobiose under neutral conditions. Compared to pristine TiO₂, the obtained photocatalyst exhibited enhanced visible-light absorption, improved charge separation efficiency, and promoted generation of ˙O₂⁻. The ˙O₂⁻-mediated oxidation mechanism was proposed. Using the conversion of glucose to arabinose as a model reaction, we systematically evaluated for the first time the distinct influences of typical lignocellulosic pretreatment derivatives on the photo-oxidation pathway of glucose. The relationship between the structure and activity, as well as their inhibitory or promotional effects, was elucidated. This work not only highlights the potential of Bi⁰/V_O-TiO₂ in cellulose photorefining and offers new insights into the synthesis of glucaric acid and arabinose under neutral conditions, but also reveals, at the molecular level, the differential impacts of biomass-derived inhibitors on photocatalytic performance. The findings are expected to provide experimental support for the optimization of practical cellulose photorefining processes, including the green design of tailored catalysts and pretreatment strategies.

Results and discussion

Characterization

The microstructure and structure of the as-prepared catalysts were revealed by SEM and TEM images. A panoramic view of the as-prepared 35-BVTO catalysts is shown in Fig. 1a. The 35-BVTO catalyst exhibits a disk-shaped morphology with measurable thickness and an undulating microrough surface. Sparse smooth-surfaced microspheres of varying sizes are clearly deposited on the surface. Conversely, Fig. S1 shows that the V_O-poor TiO₂ material shows a block accumulation of different sizes. The morphology of V_O-rich TiO₂ (reference TiO₂) is consistent with that of 35-BVTO, suggesting that 35-BVTO with this particular morphology may possess a sufficient number of surface-active sites, thus facilitating mass transfer steps in the photo-oxidation reaction of glucose and cellobiose. Furthermore, the textural properties of the catalysts are shown in Fig. S2 and Table S1. Both V_O-poor TiO₂, reference TiO₂, and 35-BVTO exhibit Type IV isotherms with H3-type hysteresis loops, confirming their mesoporous structure and thereby favoring mass transport during the reaction. Although the specific surface area of 35-BVTO (93.9 m² g⁻¹) is considerably lower than that of V_O-poor TiO₂ (124.1 m² g⁻¹), its enhanced photocatalytic performance demonstrates that a high surface area does not play a decisive role in the photo-oxidation process. Instead, the key factor lies in the enhanced intrinsic properties resulting from the synergistic electronic effects of Bi⁰ and OVs. HRTEM analysis (Fig. 1b) reveals lattice spacings of 0.353, 0.239 and 0.328 nm, corresponding to the (101) and (001) crystal planes of TiO₂ and the (012) crystal plane of Bi⁰, respectively. This confirms the reduction and deposition of metallic Bi⁰ on the TiO₂ surface.³² Besides, the elemental mapping results (Fig. 1c) indicate the distribution of Ti, O and Bi atoms in the whole 35-BVTO sample. The in situ deposition of Bi nanospheres on anatase TiO₂ is further evidenced by the XRD pattern (Fig. 1d), where the peaks at 25.3° and 27.1° are assigned to the TiO₂ (101) and Bi (012) planes, respectively.³³ These results, therefore, provide powerful evidence for the successful construction of the Bi⁰/TiO₂ catalyst.

Fig. 1 (a–c) SEM, HRTEM and EDS elemental mapping images of 35-BVTO. (d) XRD patterns of all synthesized samples. (e) UV-Vis diffuse reflectance spectra of TiO₂ and 35-BVTO samples with inset of Tauc plots (hv vs. (αhv)²). (f–h) High-resolution XPS spectra of Bi 4f, Ti 2p and O 1s. (i) ESR spectra of V_O-poor TiO₂, reference TiO₂ and 35-BVTO.

The optical absorption properties of the reference TiO₂ and 35-BVTO samples are analyzed using UV-Vis DRS spectra in Fig. 1e. Compared with the reference TiO₂, 35-BVTO displays a notable red-shift with its absorption edge extending to 479.5 nm. This enhancement stems from the LSPR and coloration effects of Bi⁰ nanospheres, which considerably improve the visible-light absorption capacity of TiO₂, broaden the photoresponse range, and suppress the recombination of photoinduced electron–hole pairs. The Bi³⁺ (158.8/164.2 eV) and Bi⁰ (156.9/162.1 eV) characteristic peaks are observed in Fig. 1f, which further confirms the coexistence of these valence states.³⁴ Remarkably, the surface of Bi⁰ nanospheres is oxidized in air to form a dense layer of amorphous Bi₂O₃.³⁵ The shift of the Ti 2p peak to higher binding energy (Fig. S3) reveals strong electron interactions at the Bi⁰/TiO₂ interface. As shown in Fig. 1g, four peaks at 458.6, 464.2, 457.9 and 463.3 eV are associated with Ti⁴⁺ and Ti³⁺, respectively, and the Ti³⁺ characteristic peaks favorably demonstrate the presence of oxygen vacancies.³⁶ Subsequently, the O 1s spectra (Fig. 1h) reveal an abundance of oxygen vacancies (530.4 eV) in 35-BVTO. The peak area of OVs in 35-BVTO increased by 8.8% compared to reference TiO₂ (Table S2), a finding corroborated by ESR spectroscopy (Fig. 1i). The oxygen vacancy defect (g = 2.003) of 35-BVTO is clearly stronger than the reference TiO₂ and TiO₂ precursor, indicating that 35-BVTO exhibits superior capacity for reducing O₂ to ˙O₂⁻. Chemically, both the cationic sites (Ti⁴⁺, Ti³⁺ and Bi³⁺) and electron-deficient OVs on the 35-BVTO surface can function as Lewis acid sites.^37–39 Based on the hard–soft acid–base (HSAB) principle, these Lewis acid sites may preferentially engage in strong interactions with the hydroxyl oxygen of sugar molecules during the reaction.⁴⁰ This effectively promotes substrate enrichment, immobilization, and activation, inducing strain in specific carbon chain configurations of cellobiose and glucose, resulting in the creation of a favorable conformation for subsequent selective attack by ROS.

To investigate the separation efficiency of the photogenerated carriers on the catalysts, the I–t, EIS and PL spectra of the samples were recorded. As is expected, 35-BVTO exhibits the highest photocurrent response under illumination (Fig. 2a), demonstrating the excellent photoinduced charge separation efficiency. The EIS curve (Fig. 2b) also reveals the excellent charge transfer capability of 35-BVTO. Subsequently, 35-BVTO displays lower fluorescence intensity in the PL spectrum (Fig. 2c). Moreover, the peak at 519 nm is attributed to the defective state.⁴¹ Evidently, the decrease in fluorescence intensity stems from the synergistic effect of Bi⁰/OVs, which suppresses the complex behavior of photoinduced electron–holes. To further investigate the transport of photogenerated carriers, the spatial distribution of surface charges was characterized using in situ KPFM. As illustrated in Fig. 2d, AFM-based surface photovoltage measurements comparing TiO₂ and 35-BVTO under dark and illuminated conditions reveal a markedly larger surface potential shift for 35-BVTO (−45.17 mV in the dark vs. −70.76 mV under light), relative to that of TiO₂ (−38.21 mV in the dark vs. −32.04 mV under light). Meanwhile, the difference in work functions (4.19 eV for Bi⁰vs. 4.96 eV for V_O-TiO₂, Fig. 2e and f) demonstrates the injection of LSPR-generated hot electrons from Bi⁰ into the V_O-TiO₂ conduction band upon contact, a process that equilibrates the Fermi levels of the two components. Mott–Schottky curves (Fig. 2g) of the samples are further analyzed to determine the positions of the valence and conduction bands. Ultimately, the synergistic effect of Bi⁰/OVs effectively optimizes electron transport paths and improves the electron–hole separation efficiency. This synergy enables the efficient generation of key reactive oxygen radicals in the system, thereby accelerating the photo-oxidation reaction rate of glucose and cellobiose.

Fig. 2 (a) Transient photocurrent responses of the as-synthesized samples. (b) EIS Nyquist plots; (c) PL spectra of reference TiO₂ and 35-BVTO. (d) 3D surface potential maps and corresponding surface photovoltage spectra. (e and f) Work function values calculated for the Bi⁰ (012) plane and the TiO₂ (101) plane, with insets showing structural models of the materials. (g) Mott–Schottky plots of pure TiO₂ and 35-BVTO.

Photocatalytic performance

To investigate the photo-oxidation performance of the as-prepared catalysts, cellobiose and glucose were used as the simulated substrates for the evaluation of photorefining. 35-BVTO exhibits a higher cellobiose conversion (71.2%) than TiO₂ (24.7%) and achieves about 55% glucaric acid selectivity and 60% glucaric acid yield (Fig. 3a and b). In comparison with other reported systems,¹¹ the catalyst exhibits a selectivity exceeding 50% under neutral conditions, a result of notable significance. As the reaction time increases, the yield of glucaric acid increases while its selectivity continues to decrease to 42.5%. As illustrated in Fig. S4, the production and selectivity of other by-products including glucose, arabinose, and formic acid increase sequentially with time, which leads to a decrease in glucaric acid selectivity and a total selectivity of about 67.3% after 4 hours. Furthermore, photocatalytic oxidation of glucose was carried out under the same reaction conditions as those used for cellobiose. The results clearly demonstrate that 35-BVTO achieves a glucose conversion of approximately 50% after 4 hours of reaction, as shown in Fig. 3c, whereas reference TiO₂ yields only about 10% conversion under identical photo-oxidation conditions. Unexpectedly, as demonstrated in Fig. 3d, the main products of glucose photo-oxidation are arabinose and formic acid rather than glucaric acid, with a selectivity exceeding 60%. This reveals that the reaction paths of the cellobiose and glucose photorefining reaction systems are not identical and the products exhibit selectivity.

Fig. 3 (a) Cellobiose conversion over the as-synthesized catalysts during 10 hours of illumination. (b) Glucaric acid yield and selectivity over 35-BVTO. (c) Glucose conversion over various catalysts; (d) product selectivity during glucose photo-oxidation using 35-BVTO. (e and f) Kinetic analysis of cellobiose and glucose photo-oxidation over 35-BVTO. (g) Photocatalytic conversion of different sugar molecules as substrates. (h and i) Catalyst recycling in cellobiose and glucose photo-oxidation.

To quantitatively elucidate the reaction kinetics, kinetic analysis was performed based on the time-dependent conversion data (Fig. 3a and c), with the corresponding results presented in Fig. 3e and f. The oxidation of both substrates follows a pseudo-first-order kinetic model. On the reference TiO₂, the apparent pseudo-first-order rate constant (k) of cellobiose is greater than that for glucose, suggesting a reaction dominated by physisorption and mass transfer that favors cellobiose with its more abundant binding sites. Compared to the reference TiO₂, 35-BVTO increases the rate constants for cellobiose and glucose oxidation by 2.3 and 6.9 times, respectively, directly confirming that the Bi⁰/OV modification fundamentally accelerates the core photocatalytic oxidation process. Furthermore, the higher enhancement factor for glucose over cellobiose indicates that the synergistic sites exhibit specific high efficiency for the C1–C2 bond cleavage pathway of glucose. In contrast, for cellobiose, the initial glycosidic bond cleavage step partially limits the overall rate enhancement. This kinetic reversal profoundly reflects that the synergy not only accelerates the reaction but also restructures the catalytic pathway, laying a theoretical foundation for subsequent analysis of the substrate-specific oxidation mechanism.

To monitor the dynamic variation in pH during the reaction process, we performed real-time pH measurements. As presented in Fig. S5, the pH in both glucose and cellobiose oxidation systems decreases markedly as the reaction proceeds, declining from an initial value of approximately 6.8 to about 4.2. This trend suggests the gradual formation of an acidic microenvironment during the process. Despite the continuous decrease in pH, neither the overall reaction activity nor the production of glucaric acid and arabinose is significantly suppressed, which suggests that the self-induced acidification does not fundamentally alter the reaction pathway. Instead, the process appears to be driven mainly by photogenerated charges and ROS rather than by the initial alkaline reaction conditions. Moreover, the increasingly acidic environment in the cellobiose oxidation system likely facilitates the hydrolysis of glycosidic bonds, accounting for the detection of glucose among the reaction products.

Furthermore, we examined its performance on other key monosaccharides (mannose and galactose) and polysaccharides (glucan and xylan) to evaluate the substrate adaptability of the 35-BVTO catalyst. As shown in Fig. 3g, under standard photocatalytic conditions, both mannose and galactose are efficiently converted into arabinose and formic acid as the main products, following the same ˙O₂⁻-mediated oxidative cleavage pathway as observed for glucose. In contrast, only formic acid is detected from polysaccharides, suggesting a complex, non-selective photocatalytic cleavage mechanism. These results collectively demonstrate that 35-BVTO possesses broad substrate tolerance and versatile conversion pathways for soluble monosaccharides, but shows limited effectiveness toward macromolecular polysaccharides.

To evaluate the atom economy and green efficiency of the cellobiose and glucose photorefining system, we calculated the carbon balance for each reaction after 4 hours, as shown in Fig. S6. The carbon balance of the cellobiose photorefining process remains above 85% after 4 hours of reaction, while it decreases to 74% after more than 10 h (Fig. S6a). The carbon balance of the glucose photorefining process decreases to 84.6% after 4 hours of reaction (Fig. S6b), and the yield of arabinose remains at 21.5%. Overall, the high carbon balance observed initially indicates that 35-BVTO effectively enhances reaction rates and atom economy by selectively producing glucaric acid and arabinose instead of unproductive mineralization. For the later stages of the reaction, increased carbon loss may be attributed to the deep oxidation of substances such as formic acid and arabinose ultimately generating CO₂, as well as the molecular condensation of reaction intermediates forming undetectable polymeric byproducts. Therefore, strictly controlling reaction time is the most direct means of preventing carbon loss in later stages. Furthermore, the design of catalysts with higher intrinsic selectivity toward the target pathway is also crucial. Moreover, after five consecutive cycles, the conversion rates of cellobiose and glucose reached 43.1% (Fig. 3h) and 40.3% (Fig. 3i), respectively, with glucaric acid selectivity at 52.9%, demonstrating the exceptional durability and stable performance of 35-BVTO. Finally, the post-cycling 35-BVTO sample clearly indicates that Bi⁰ underwent structural degradation (Fig. S7), and both Bi⁰ and Bi³⁺ species remain present on the surface of the recycled 35-BVTO catalyst. Unexpectedly, no crystalline Bi⁰ peaks are detected by XRD, suggesting that metallic bismuth may exhibit a highly dispersed state on the catalyst surface. Such restructuring may fundamentally alter the nature of the active sites and modify the electronic structure, diminishing their efficiency as electron traps. Furthermore, recalcitrant intermediates or carbonaceous deposits may form and accumulate on the catalyst surface, blocking active sites and preventing reactants from accessing catalytic centers.

Based on these findings, the developed catalyst demonstrates a significant enhancement in the photocatalytic oxidation of both cellobiose and glucose, which highlights the potential of Bi⁰ as a viable alternative to precious metals for the photorefining of small-molecule sugars. Consistent with other reported studies,^11,42 the inability of the catalytic system to effectively suppress deep oxidation and mineralization remains a critical bottleneck. Future efforts should focus on enhancing the stability of Bi⁰ species and fine-tuning the band structure of TiO₂ to effectively regulate the species, concentration, and pathways of active radicals in photocatalytic reactions, thereby improving reaction selectivity.

Active species identification

To clarify the reactive species for the photocatalytic oxidation of glucose and cellobiose by 35-BVTO, a combination of EPR, radical trapping and NBT experiments was executed. As shown in Fig. 4a–c, no oxygen active species signals are observed in TiO₂ and 35-BVTO under dark conditions, whereas TiO₂ and 35-BVTO exhibit DMPO-˙O₂⁻, DMPO-¹O₂, and DMPO-˙OH signals under simulated sunlight irradiation, and the signal intensity of oxygen-active species is stronger in the latter than in the former, implying that 35-BVTO produces major active substances to participate in the photo-oxidation reaction process under simulated sunlight irradiation.

Fig. 4 (a–c) ESR spectra of DMPO-˙O₂⁻, DMPO-¹O₂, and DMPO-˙OH adducts for reference TiO₂ and 35-BVTO. (d) Scavenging experiments during cellobiose photo-oxidation over 35-BVTO. (e) Scavenging tests for glucose photo-oxidation using 35-BVTO. (f) NBT assay for reference TiO₂ and 35-BVTO.

Furthermore, benzoquinone (BQ), furfuryl alcohol (FFA), isopropanol (IPA), ethylenediaminetetraacetic acid disodium (EDTA-2Na) and KBrO₃ are used as bursting agents for ˙O₂⁻, ¹O₂, ˙OH, h⁺ and e⁻, respectively. Obviously, as illustrated in Fig. 4d, the conversion rate of cellobiose decreases from 49.4% to about 25% upon the addition of BQ and KBrO₃, and the introduction of EDTA-2Na reduces it to 40.9%, whereas FFA and IPA exert a weaker effect. As displayed in Fig. S8a, the yield of glucaric acid displays the most significant decrease with the addition of BQ and KBrO₃, and the introduction of IPA can promote its generation, indicating that ˙O₂⁻ is the main oxidatively active species in the photocatalytic oxidation of cellobiose, and the photogenerated electrons required for the generation of ˙O₂⁻ are the key drivers of the reaction. Beyond this, h⁺ plays a certain auxiliary oxidizing role, but the contribution is smaller than that of ˙O₂⁻, and ˙OH and ¹O₂ play a minor role in this reaction system, and the effective reduction of the ˙OH in the system promotes the photorefining of cellobiose into glucaric acid, which provides guidance for the rational design of highly efficient catalysts. Similarly, as presented in Fig. 4e, upon introduction of BQ and KBrO₃, the glucose conversion is suppressed to 20.4% and 37.3%, respectively, suggesting that ˙O₂⁻ and e⁻ are the major oxidizing active species. The addition of EDTA-2Na significantly increases formic acid production in the system (Fig. S8b), indicating that it promotes electron reactions by quenching holes, thereby indirectly enhancing the yield of the key species ˙O₂⁻ responsible for formic acid generation. Besides, NBT assay results conclusively evidence substantially reinforced ˙O₂⁻ formation in the 35-BVTO system in Fig. 4f. Our results demonstrate that ˙O₂⁻, rather than h⁺, is the primary oxidative species in the 35-BVTO photorefining process, which drives the efficient degradation of cellobiose and glucose via direct oxidation or radical chain reactions. This finding underpins further analysis of the effects of pretreatment derivatives and the reaction pathway.

Effect of lignocellulosic derivatives

The differential impacts of seven key lignocellulosic pretreatment derivatives, encompassing soluble sugars, furans, organic acids, and lignin-derived species on the selective photo-oxidation of glucose to arabinose were methodically evaluated. The ensuing discussion elucidates structure–activity relationships and derives actionable insights for optimizing the process.

To investigate the differential effects of pretreatment-derived saccharide components on the photocatalytic oxidation of glucose for arabinose, six soluble sugars (cellobiose, glucan, xylan, xylose, mannose, and galactose) were selected to represent major classes and key structural motifs characteristic of sugars derived from lignocellulose pretreatment. The introduction of glucan sharply decreases glucose conversion by 45.3% and arabinose yield by 83.1% (Fig. 5a). Studies indicate that the introduction of high molecular weight glucan increases solution viscosity, thereby reducing the mass transfer rate.⁴³ Consequently, glucan likely inhibits the photo-oxidation by blocking active sites on the catalyst surface. Fig. S9a demonstrates that xylan exhibits strong inhibitory effects within the oxidation system, primarily through physical shielding and diffusion limitation mechanisms analogous to those observed with glucan. Subsequently, Fig. S9b demonstrates that the presence of xylose inhibits the glucose photo-oxidation process, as evidenced by a 24% decrease in glucose conversion. The inhibitory effect of xylose on glucose photocatalytic oxidation can likely be attributed to the combined actions of competitive adsorption onto the catalyst surface and competitive consumption of ROS.⁴⁴ Fig. S9c reveals that cellobiose exhibits a concentration-dependent effect on glucose photo-oxidation. Combined with those from Fig. 3a, cellobiose causes a mild inhibitory effect at low concentrations, primarily attributed to competitive adsorption. Conversely, its strong competitive oxidation predominates at high concentrations. As shown in Fig. 6b, introducing mannose suppresses glucose conversion but enhances the production of arabinose and formic acid. This counterintuitive result arises because mannose, a glucose enantiomer, itself undergoes photo-oxidation to yield these products.⁴⁵ Remarkably, high concentrations of mannose significantly inhibit glucose conversion, which may indicate surface passivation caused by strong adsorption at high concentrations of 35-BVTO. Specifically, sugar molecules form strong coordination bonds with Lewis acid sites on the 35-BVTO surface through multiple hydroxyl groups. Simultaneously, sugar molecules adsorbed on the 35-BVTO surface undergo deep oxidation or polymerization under illumination, ultimately leading to irreversible occupation of active sites. Similarly, galactose exhibits a mannose-like dual effect (Fig. S9d). Mannose and galactose exhibit both inhibitory and promotional effects during glucose photo-oxidation, highlighting the unique influence of diastereomeric isomers in photocatalytic sugar conversion and providing new mechanistic insights for achieving targeted sugar conversion. Collectively, the above results reveal that soluble pretreatment-derived sugars interfere with the photocatalytic oxidation of glucose through physical barrier effects, competitive adsorption, and ROS consumption. For practical biomass processing, the presence of macromolecular polysaccharides should be minimized, and the residual concentration of monosaccharides and disaccharides must be strictly controlled to mitigate their potent competitive adsorption and oxidation competition.

Fig. 5 Effects of various compounds on the photo-oxidative conversion of glucose to arabinose: (a) glucan, (b) mannose, (c) 5-hydroxymethylfurfural, (d) furfural, (e) acetic acid, and (f) vanillic acid.

Fig. 6 Effects of various compounds on the photo-oxidative conversion of glucose to arabinose: (a) alkali lignin, (b) sodium lignosulfonate, (c) 2-phenoxy-1-phenylethanol, (d) vanillin, (e) K⁺, and (f) Mg²⁺.

To investigate the effect of furan derivatives on the glucose photo-oxidation process, 5-hydroxymethylfurfural (HMF) and furfural (FAL), typical furan compounds under acid or high-temperature pretreatment of cellulose and hemicellulose, were chosen as representatives for the investigation. As illustrated in Fig. 5c, low HMF concentrations (≤60 mg L⁻¹) have a minimal inhibitory effect on glucose photo-oxidation, in contrast to the pronounced inhibition observed at 240 mg L⁻¹, suggesting that HMF exhibits a significant concentration dependence in the glucose photo-oxidation process. The photocatalytic oxidation of HMF to 2,5-diformylfuran (DFF) has been extensively well studied in the presence of ROS.^46,47 Together with Fig. 4, the results indicate that HMF (≥60 mg L⁻¹) consumes ˙O₂⁻ in the system to block the arabinose pathway. Beyond this, it has been proven that the introduction of single atoms and OVs on the surface of TiO₂ can strengthen the adsorption, mass transfer, and activation capabilities for HMF, enabling its efficient conversion to DFF.⁴⁸ This reveals that high-concentration HMF strongly suppresses arabinose and formic acid production through strong competitive adsorption and strong competitive oxidation. As shown in Fig. 5d, the addition of furfural reduces the yields of arabinose and formic acid by approximately 20% at 100 mg L⁻¹, while the glucose conversion rate decreases by only 4.6%. This result demonstrates that furfural has a smaller effect on glucose photo-oxidation than HMF, which is mainly due to differences in their molecular structures.⁴⁹ Briefly, the above results indicate that HMF is the main removal target (concentration ≤60 mg L⁻¹) and furfural has a higher allowable concentration (≤500 mg L⁻¹) when applying the actual lignocellulosic pretreatment substrate to the cellulose photorefinery process. Therefore, selective adsorption of HMF or design of hydrophobically modified catalysts can be prioritized when choosing the actual biomass pretreatment process.

To investigate the impact of acidic pretreatment derivatives on the glucose photo-oxidation system, we systematically analyzed eight pretreatment-derived acids: formic acid (FA), acetic acid (AA), phthalic acid (PA), vanillic acid (VA), and others. These acids encompass representative acidic degradation products from the three major lignocellulosic components and feature characteristic functional group combinations (e.g., phenolic hydroxyl, carboxyl, and conjugated double bonds), thereby reflecting the compositional variability found in practical pretreatment liquors. As displayed in Fig. 5e, the oxidative performance is largely unaffected by AA at concentrations ≤500 mg L⁻¹ but plummets at 1000 mg L⁻¹, reducing glucose conversion, arabinose yield, and formic acid yield by 45.6%, 72.9%, and 68.5%, respectively. This effect is attributed to AA forming hydrophobic microregions on the TiO₂ surface. Weak AA adsorption at low concentrations only marginally affects glucose access to 35-BVTO active sites, whereas high concentrations lead to dense monolayers that severely block them.⁵⁰ A high acetic acid concentration could also change the surface charge state of 35-BVTO, thereby promoting its own electrostatic adsorption. When formic acid is added exogenously (Fig. S10a), the glucose conversion and arabinose yield decrease by 25.1% and 48.7%, respectively. Interestingly, the decline in total formic acid implies that the endogenous production pathway is hindered by the presence of externally supplemented formic acid. Therefore, it is essential to develop formic acid separation technology to overcome equilibrium constraints. Notably, levulinic acid (LA) exhibits weak inhibition of glucose photo-oxidation, resulting in only a 7% reduction in the glucose conversion rate (Fig. S10b), suggesting that LA does not block the core oxidation pathway of glucose. As presented in Fig. S10c, glucose photo-oxidation exhibits greater sensitivity to the dicarboxylic acid moiety, as evidenced by phthalic acid (120 mg L⁻¹). It caused substantially stronger suppression (42% in conversion, 78.5% and 69% in yields) than FA, AA, or LA, which may be attributed to the formation of ligand bonds between PA and Ti⁴⁺ sites via chemisorption, thereby blocking the active sites.⁵¹

Vanillic acid, p-hydroxybenzoic acid (HBA), and coumaric acid (CA) represent signature lignin-derived pretreatment compounds featuring phenolic hydroxyl groups. As demonstrated in Fig. 5f, the strong inhibitory effect of VA at high concentrations (>60 mg L⁻¹) results in almost zero arabinose and formic acid production. The phenolic hydroxyl group quenches ROS, while its methoxyl group (–OCH₃) further enhances the efficiency via the electron-donating resonance effect (+R).^52–54 Meanwhile, like AA, its carboxyl group can also block glucose adsorption by chelating Ti⁴⁺ sites on the surface of the TiO₂ catalyst. In contrast, the methoxy-deficient HBA exhibits a relatively smaller conjugated system, thus exhibiting a higher concentration tolerance than VA (Fig. S10d). Compared to VA and HBA, ferulic acid exhibits the weakest inhibitory effect, as indicated by its higher critical concentration tolerance (Fig. S10e). The weaker inhibition arises as a consequence of the steric hindrance introduced by ortho-substituents, which likely impedes the approach of ˙O₂⁻ toward the phenolic hydroxyl group, resulting in a reduction in radical quenching capacity.^55,56 More importantly, CA is introduced into the system with a stronger inhibitory effect (Fig. S10f). Owing to its sterically unhindered and extensive conjugated system, it is reasonable to propose that CA functions as the most kinetically efficient quencher of key reactive species, thereby exhibiting the strongest inhibitory effect. The systematic investigation reveals that their inhibitory potency is closely correlated with the presence of specific functional groups (such as phenolic hydroxyl, dicarboxylic, and conjugated systems), molecular steric hindrance, and adsorption capacity on catalytically active sites. In practical process treatments, priority should be given to removing strongly inhibitory aromatic acids with high conjugation and low steric hindrance (e.g., CA and VA) as well as dicarboxylic acids (e.g., PA), while short-chain aliphatic acids (e.g., AA and LA) and sterically hindered aromatics can be tolerated at higher concentrations.

To investigate the behavior of the characteristic lignin derivatives produced by the main pretreatment processes in the system of arabinose preparation by glucose photo-oxidation, three representative lignin derivatives (alkali lignin (AL), sodium lignosulfonate (SLS), and milled-wood lignin (MWL)) of the dominant pretreatment process were selected in a glucose photocatalytic system. With increasing concentration of AL, as shown in Fig. 6a, the glucose conversion and arabinose yield decrease dramatically (>60%), while the formic acid yield slightly increases. This phenomenon stems from AL suppressing the overall conversion by competitively inhibiting both the adsorption and the oxidation of glucose on the 35-BVTO surface, via a combination of site blocking and radical scavenging. The slight increase in formic acid yield results from a barrier formed by alkaline lignin and its oxidation products, further inhibiting mineralization of formic acid. An alternative explanation is that intermediate radicals generated through alternative oxidation pathways indiscriminately attack glucose molecules, triggering random bond-breaking reactions that ultimately yield formic acid.^57,58 As presented in Fig. 6b, a key observation is that when sodium lignosulfonate is added to the system, the glucose conversion decreases from 48% to 2.6%, while the arabinose and formic acid yields simultaneously plummet by 75% and 47%, indicating that SLS has strong inhibitory effects on the glucose photo-oxidation process. It has been documented that SLS has a strong adsorption capacity owing to its unique structural properties, thereby reducing the active sites through its strong interaction.⁵⁹ In addition, SLS can also block the glucose photo-oxidation process by utilizing h⁺ and ˙O₂⁻ to participate in a competitive reaction to produce vanillin.^60,61 Nevertheless, MWL only weakly inhibits glucose photo-oxidation, as shown in Fig. S11, likely because it retains the natural β-O-4 bond (inactive) and lacks hydrophilic groups for adsorption or radical quenching. Overall, these findings demonstrate that the inhibitory potency of lignin-derived compounds is governed not by lignin origin but by their chemical structure, particularly the abundance and nature of specific functional groups from pretreatment. Meanwhile, the phenolic hydroxyl content and sulfonic acid group content are key indicators for predicting and interpreting their inhibition strength. During the actual pretreatment process, the concentration of AL in the system must be maintained below 300 mg L⁻¹, whereas that of SLS should remain under 100 mg L⁻¹.

To explore the impact of key lignin linkages (β-O-4, α-O-4, and 4-O-5) on the photocatalytic conversion of glucose to arabinose, 2-phenoxy-1-phenylethanol, benzyl phenyl ether (BPE) and diphenyl ether (DPE) were selected in the system. The β-O-4 linkage, as the predominant interunit bond in native lignin, is the main depolymerization target and a characteristic active linkage that survives mild pretreatment.⁶² In contrast, α-O-4 and 4-O-5 linkages predominantly encompass the stable or recalcitrant fragments generated during more severe pretreatment.⁶³ As illustrated in Fig. 6c, 2-phenoxy-1-phenylethanol causes an 8% decrease in glucose conversion, due to its preferential quenching of photogenerated holes and ¹O₂, thereby exerting a minimal effect on the reaction.^64,65 Nevertheless, its concentration needs to be controlled (<500 mg L⁻¹) to avoid significant loss of arabinose and formic acid yields. Fig. S12a demonstrates that the introduction of BPE results in slight inhibition of glucose conversion to arabinose. However, the high concentration of BPE (≥500 mg L⁻¹) might accelerate the deep oxidation of formic acid, as evidenced by a significant decrease in formic acid production. Subsequently, DPE exhibits enhanced inhibition compared to 2-phenoxy-1-phenylethanol and BPE (Fig. S12b), owing to strong adsorption and light absorption properties derived from its unique 4-O-5 bond structure.^66,67 Consequently, the study reveals that the pretreatment products should preferentially remove soluble lignin derivatives of the 4-O-5 bond structure and strictly control the concentration of β-O-4 and α-O-4 bond derivatives, preferentially separating or transforming them to reduce the total amount of soluble lignin to minimize the competitive inhibition of the photocatalytic reaction.

To further elucidate the differential inhibition of the glucose photo-oxidation process by the characteristic degradation products of different structural units of lignin, vanillin, syringaldehyde, and phenol were selected to simulate the characteristic degradation products of lignin G, S, and H units, respectively. The inhibitory severity of the three compounds at 100 mg L⁻¹ varies considerably (Fig. 6d and Fig. S12c and d); vanillin causes reductions of 44.1% (glucose), 97% (arabinose), and 82.4% (formic acid); butyraldehyde leads to a 48.9% decrease in glucose conversion with over 70% loss in both product yields. In contrast, phenol exhibits the weakest inhibition, reducing glucose conversion and product yields by less than 30%. Studies have confirmed that vanillin, syringaldehyde and phenol, as common phenolic antioxidants, are efficient ROS scavengers.^68–70 Among these, vanillin exhibits the strongest radical scavenging capacity due to its highest phenolic hydroxyl electron density, enabling it to most effectively quench ˙O₂⁻ and thus explaining its strongest inhibition of glucose photo-oxidation. At low concentrations, the variation in inhibitory efficacy is mainly determined by the modulation of the electron density of phenolic hydroxyl groups by substituents in their molecular structure. The scavenging ability of the phenolic hydroxyl group is positively correlated with its electron density, whereas substituents (especially the position of the methoxyl group) determine the order of the electron densities. Collectively, these findings further underscore that lignin derivatives in actual biomass hydrolysates are significant and structure-dependent inhibitors in the photocatalytic sugar conversion process. Therefore, the preferential removal of phenolic compounds can enhance the photocatalytic oxidation of sugars derived from pretreated biomass.

Lignocellulosic biomass contains abundant mineral elements including K, Ca, and Mg. These elements are released as ions during pretreatment and persist in the system and potentially affect the glucose photo-oxidation reaction through surface adsorption, photoelectron capture, and microenvironment alteration.^71,72 To mimic the practical pretreatment systems, K⁺, Ca²⁺, Mg²⁺, and Na⁺ were selected as representative ions to investigate their effects on the reaction. As illustrated in Fig. 6e, increasing K⁺ concentration results in an elevated glucose conversion rate and a significant increase in formic acid production, demonstrating a concentration-dependent relationship, while arabinose production exhibits a slight decrease. Similarly, supplementation of the system with the other three metal ions (K⁺, Na⁺, and Mg²⁺) yields analogous experimental outcomes (Fig. 6f and Fig. S13), indicating that the introduction of metal ions generally improves glucose conversion efficiency and promotes formic acid generation. Based on trapping experimental evidence, we propose that metal ions govern reaction selectivity by accelerating the transfer of photogenerated electrons to O₂, via a dual mechanism of electron-bridging and radical regulation. They may also form stable complexes with formate anions (HCOO⁻) to accelerate formic acid desorption from the 35-BVTO surface, thereby effectively preventing their overoxidation to CO₂. Given their weaker inhibition effect on arabinose yield, divalent metal ions are compatible with pH regulators during pretreatment. A key strategy for reducing non-selective oxidation losses is to selectively control the concentrations of other metal ions.

In summary, this study elucidates the differential effects of lignocellulosic pretreatment-derived compounds on the photocatalytic oxidation of glucose to arabinose. Critical findings reveal that: (i) soluble sugars and lignin derivatives act as potent inhibitors via competitive adsorption and ROS scavenging, necessitating stringent control of their concentrations; (ii) HMF and high-phenolic-acidity compounds constitute the core inhibitory factors requiring prioritized removal; and (iii) the dual effect of mannose and galactose and the electron-bridging function of metal ions can be strategically utilized.

Photocatalytic oxidation mechanism

Based on the above results, we propose a charge transfer mechanism (Fig. 7a) elucidating the synergistic roles of Bi⁰ and OVs in enhancing charge separation, migration, and ROS generation in TiO₂. Upon contact in the dark, electrons flow from Bi⁰ to V_O-TiO₂ until their Fermi energy levels align. Under light irradiation, the photoinduced holes participate in the oxidation reaction, which increases the electron concentration in the V_O-TiO₂ conduction band, thereby forming a reverse driving force. Concurrently, Bi⁰ acts as an electron trap, prompting electron flow from V_O-TiO₂ to Bi⁰, thereby greatly suppressing photoinduced charge recombination. Notably, the SPR excitation of Bi⁰ nanoparticles generates high-energy hot electrons and an electron cloud, which cross the interfacial potential barrier and are injected into the conduction band of TiO₂.⁷³ Simultaneously, OVs serve to capture electrons, adsorb and activate O₂ molecules, promoting the rapid transfer of electrons to O₂ and realizing stable ˙O₂⁻ generation. Consequently, the OVs–conduction band–Bi⁰ interface constructs bifunctional sites that break through the kinetic bottleneck of O₂ reduction, realizing efficient and sustained generation of ˙O₂⁻. This provides an ideal reactive oxygen environment for the selective oxidation of cellobiose and glucose. Building on this premise, the plausible photocatalytic oxidation mechanisms of cellobiose and glucose are presented in Fig. 7b. Typically, the reducing end glucose unit of cellobiose retains a free aldehyde group (C1=O), and its C6–OH group forms a β-1,4-glycosidic bond with the C1 carbon of the adjacent glucose unit, rendering this hydroxyl group non-free. In the 35-BVTO system, the electron-rich nature of the β-1,4-glycosidic bond enables its efficient adsorption and preferential activation on the TiO₂ surface through hydrogen bonding or coordination interactions synergistically mediated by bismuth species and OVs. Under light irradiation, ˙O₂⁻ can attack the C1 aldehyde group at the reducing end, initiating cleavage of the β-1,4-glycosidic bond.^18,74 Concurrently, the oxygen atom of the adjacent glycosidic bond may be oxidized by h⁺, thereby activating the C1–O bond, making the C6–OH group more susceptible to oxidation by ˙O₂⁻, ultimately resulting in the formation of glucaric acid and glucose. For glucose, kinetic results demonstrate that the 35-BVTO catalyst more efficiently cleaves its C–C bonds. Since glucose exists predominantly as a cyclic hemiacetal structure in the initial neutral aqueous solution, it has a low concentration of free aldehyde groups, which hinders simultaneous oxidation at C1 and C6. Concretely, the ˙O₂⁻ rapidly attacks the aldehyde group to afford arabinose and formic acid via simultaneous C1–C2 α-cleavage and decarboxylation and further produces erythritol and formic acid by Ruff degradation.⁷⁵ Given this mechanism, it is reasonably inferred that pretreatment-derived compounds inhibit arabinose production mainly through two pathways. First, the dominant reactive species (˙O₂⁻) are readily quenched by small molecules such as xylose, HMF, high-hydroxyl-acid compounds, and lignin-derived phenolic species. Second, the active sites of the 35-BVTO surface are occupied by soluble polysaccharides, carboxylic acids, and lignin-derived macromolecules through physical blocking and competitive adsorption. Together, these effects collectively hinder the oxidative pathway of glucose and significantly inhibit the production of arabinose.

Fig. 7 (a) Proposed mechanism of photoinduced charge transfer on the Bi⁰/V_O-TiO₂ surface. (b) Proposed mechanism for the photorefining of cellobiose and glucose mediated by Bi⁰/V_O-TiO₂.

Conclusions

In summary, we have reported the rational design of a non-noble Bi⁰/V_O-TiO₂ photocatalyst for the efficient, solar-driven refining of biomass sugars. We demonstrate that the synergy between Bi⁰ and OVs is pivotal, as it significantly facilitates efficient separation of photogenerated charge carriers and robust generation of ˙O₂⁻ radicals, which are identified as the driving species for the oxidation pathway under mild conditions. Crucially, the elucidated ˙O₂⁻-mediated mechanism provides a molecular-level framework to rationalize the inhibitory or promoting effects of different pretreatment derivatives. HMF, high-phenolic-acidity compounds, and lignin-derived phenolic species exhibit potent inhibitory effects by scavenging the critical ˙O₂⁻, whereas metal ions enhance ˙O₂⁻ generation by facilitating electron transfer. This understanding guides the design of targeted pretreatment strategies for “targeted removal of inhibitory components while maintaining functional species”, creating a pathway to enhance the efficiency and economic viability of the entire photorefining system from the outset. Therefore, this work demonstrates the application potential of the Bi⁰/V_O-TiO₂ catalyst in cellulose photorefining and, more critically, by transcending idealized model systems, provides a foundation for future green, efficient, and interference-resistant systems.

Author contributions

Tingting Zhang: writing – original draft, formal analysis, investigation, visualization and methodology. Xinyao Zhang: investigation. Suhang Cheng: investigation. Dong Tian: conceptualization, funding acquisition, and visualization. Li Zhao: methodology and formal analysis. Jiufu Chen: formal analysis. Jinguang Hu: methodology and formal analysis. Fei Shen: conceptualization, funding acquisition, project administration, and supervision.

Conflicts of interest

The authors declare that they have no competing interests.

Data availability

The data supporting this article are available as part of the supporting information (SI). Supplementary information: detailed description of the density functional theory (DFT) calculations pertaining to the materials under investigation. See DOI: https://doi.org/10.1039/d5gc05406k.

The crystal structure data for Bi and TiO₂ are available from the Crystallography Open Database (COD) at https://www.crystallography.net/, under CIF numbers 7123352 and 1526931, respectively.

Acknowledgements

This work was financially supported by the National Natural Science Foundation of China (Grant No. 22478266), the Science and Technology Department of Sichuan Province (2025HJRC0053) and the China Education Association for International Exchange (2023261).

References

1. B. Long, F. Zhang, S. Y. Dai, M. Foston, Y. J. Tang and J. S. Yuan, Nat. Rev. Bioeng., 2025, 3, 230–244.
2. S. Feng, P. T. T. Nguyen, X. Ma and N. Yan, Angew. Chem., Int. Ed., 2024, 63, e202408504.
3. H. Zhao, D. Trivedi, M. Roostaeinia, X. Yong, J. Chen, P. Kumar, J. Liu, B. Su, S. Larter, M. G. Kibria and J. Hu, Green Chem., 2023, 25, 692–699.
4. L. Sun and N. Luo, J. Energy Chem., 2024, 94, 102–127.
5. Z. Sun, H. Zhao, X. Yu, J. Hu and Z. Chen, Chem. Synth., 2024, 4, 4.
6. Y. You, S. Chen, J. Zhao, J. Lin, D. Wen, P. Sha, L. Li, D. Bu and S. Huang, Adv. Mater., 2024, 36, 2307962.
7. Q. Yan, Y. Zhou, Y. Zhang, D. Sun, X. Wu, S. Yang, C. Xu and H. Zhang, Adv. Sci., 2025, 12, e08850.
8. X. Li, J. Ma, W. Jiang, R. Cui, Z. Liu, J. Zhang and R. Sun, ACS Sustainable Chem. Eng., 2024, 12, 14810–14820.
9. F. Kang, C. Shi, Y. Zhu, M. Eqi, J. Shi, M. Teng, Z. Huang, C. Si, F. Jiang and J. Hu, J. Energy Chem., 2023, 799, 158–167.
10. Y. R. Tan, C. Yang, B. Zhang, K. Deng, D. Tang and D. Ding, Mol. Catal., 2024, 554, 113818.
11. J. Wang, H. Zhao, L. Chen, J. Björk, J. Rosen, P. Kumar, L. Jing, J. Chen, M. G. Kibria and J. Hu, Appl. Catal., B, 2024, 344, 123665.
12. J. Wang, Q. Zhao, Z. Li, Y. Xiao, X. Zhang, N. Zhong, H. Zhao, L. Jing and D. Di, Appl. Catal., B, 2025, 360, 124526.
13. J. Huang, Y. Ding, J. Li, Z. Hu, S. Saravanamurugan, J. Wang, Y. Su, S. Yang and H. Li, Carbon Energy, 2025, 7, e675.
14. T. Tana, P. Han, A. J. Brock, X. Mao, S. Sarina, E. R. Waclawik, A. Du, S. E. Bottle and H. Y. Zhu, Nat. Commun., 2023, 14, 4609.
15. J. Li, R. L. Smith, S. Xu, D. Li, J. Yang, K. Zhang and F. Shen, Green Chem., 2022, 24, 315–324.
16. Y. Zhang, J. Su, T. Li, L. Yin, H. Xian and K. Wang, Chem. Eng. J., 2025, 508, 160894.
17. Z. Gao, P. Ren, L. Sun, N. Luo and F. Wang, Nat. Synth., 2024, 3, 438–451.
18. J. Wang, Q. Zhao, P. Kumar, H. Zhao, L. Jing, D. D. Tommaso, R. Crespo-Otero, M. G. Kibria and J. Hu, ACS Catal., 2024, 14, 3376–3386.
19. J. Wang, X. Wang, H. Zhao, J. F. Van Humbeck, B. N. Richtik, M. R. Dolgos, A. Seifitokaldani, M. G. Kibria and J. Hu, ACS Catal., 2022, 12, 14418–14428.
20. S. Zhong, D. Yu, Y. Ma, Y. Lin, X. Wang, Z. Yu, M. Huang, Y. Hou, M. Anpo, J. C. Yu, J. Zhang and X. Wang, Angew. Chem., 2025, 137, e202502823.
21. F. Si, G. Wang, M. Lv, J. Bai, Y. Li, T. Hou and Y. Li, Appl. Catal., B, 2025, 365, 124934.
22. Z. Tian, Y. Da, M. Wang, X. Dou, X. Cui, J. Chen, R. Jiang, S. Xi, B. Cui, Y. Luo, H. Yang, Y. Long, Y. Xiao and W. Chen, Nat. Commun., 2023, 14, 142.
23. H. Zhao, X. Yu, C. F. Li, W. Yu, A. Wang, Z. Y. Hu, S. Larter, Y. Li, M. G. Kibria and J. Hu, J. Energy Chem., 2022, 64, 201–208.
24. M. Eqi, C. Shi, J. Xie, F. Kang, H. Qi, X. Tan, Z. Huang, J. Liu and J. Guo, Adv. Compos. Hybrid Mater., 2023, 6, 5.
25. Z. Li, W. Li, J. Wang, L. Jing, P. Li, H. Zhang, Y. Fan, H. Wang, Z. Chen and J. Hu, Adv. Energy Mater., 2024, 14, 2401838.
26. E. Wang, Y. Chu, W. Zhang, Y. Wei, C. Si, R. Palkovits, X. P. Wu and Z. Chen, Chin. J. Catal., 2025, 74, 308–318.
27. S. Zhang, D. Han, Z. Wang and F. Gu, Small, 2024, 20, 2309656.
28. Q. Geng, H. Xie, Y. He, S. Chen, F. Dong and Y. Sun, J. Hazard. Mater., 2025, 485, 136951.
29. H. Zhao, P. Liu, X. Wu, A. Wang, D. Zheng, S. Wang, Z. Chen, S. Larter, Y. Li, B. Su, M. G. Kibria and J. Hu, Appl. Catal., B, 2021, 291, 120055.
30. X. Wu, H. Zhao, M. A. Khan, P. Maity, T. Al-Attas, S. Larter, Q. Yong, O. F. Mohammed, M. G. Kibria and J. Hu, ACS Sustainable Chem. Eng., 2020, 8, 15772–15781.
31. S. K. Bhatia, S. S. Jagtap, A. A. Bedekar, R. K. Bhatia, A. K. Patel, D. Pant, J. R. Banu, C. V. Rao, Y. G. Kim and Y. Yang, Bioresour. Technol., 2020, 300, 122724.
32. H. Wang, J. Zhang, Y. Cheng, P. Zhang, M. Tan, R. Yu, Y. Zang and X. Li, Appl. Catal., B, 2025, 380, 125814.
33. L. Ma, C. Huang, X. Li, X. An, J. Li, L. Zou, X. Zhu, C. Yuan, X. Dai, Y. Zhou, X. Zhu and C. Tian, Sep. Purif. Technol., 2025, 354, 128635.
34. J. Shan, Y. Huang and W. Jiang, J. Cleaner Prod., 2025, 518, 145947.
35. Z. Jiao, M. Shang, J. Liu, G. Lu, X. Wang and Y. Bi, Nano Energy, 2017, 31, 96–104.
36. I. Gul, M. Sayed, F. Rehman, J. Wang, P. Fu, Y. Zhang and M. N. Nadagouda, Chem. Eng. J., 2025, 506, 159992.
37. J. Hao, Y. Zhang, L. Zhang, J. Shen, L. Meng and X. Wang, Chem. Eng. J., 2023, 464, 142536.
38. Q. Qin, H. Jang, X. Jiang, L. Wang, X. Wang, M. G. Kim, S. Liu, X. Liu and J. Cho, Angew. Chem., Int. Ed., 2024, 63, e202317622.
39. D. Yu, J. Zou, L. Zeng, Y. Hou, W. Lin, L. Wu, M. Anpo, J. C. Yu, J. Zhang and X. Wang, Angew. Chem., 2025, 137, e202425551.
40. T. Yu, S. Yin, Z. Zhai, H. Wen, L. Luo and S. Yin, J. Energy Chem., 2025, 103, 858–865.
41. Y. Liu, M. Wang, D. Li, F. Fang and W. Huang, Appl. Surf. Sci., 2021, 540, 148330.
42. J. Wang, H. Zhao, P. Liu, N. Yasri, N. Zhong, M. G. Kibria and J. Hu, J. Energy Chem., 2022, 74, 324–331.
43. S. M. Tosh, P. J. Wood, Q. Wang and J. Weisz, Carbohydr. Polym., 2004, 55, 425–436.
44. J. Zhang, K. Liu, S. Sun, R. Sun and J. Ma, Green Chem., 2022, 24, 8367–8376.
45. J. Wan, P. Chu, X. Wang, Z. Sun, W. Zhang, Y. Wang, Q. Zhang, W. Yan, X. Jin and C. Yang, ACS Sustainable Chem. Eng., 2024, 12, 14940–14951.
46. T. Shan, L. Luo, T. Chen, L. Deng, M. Li, X. Yang, L. Shen and M. Q. Yang, Green Chem., 2023, 25, 2745–2756.
47. S. Jiang, Z. Chen, S. Xiong, H. Zhao, X. Xiao and Z. Shen, J. Energy Chem., 2025, 109, 830–838.
48. D. Shi, J. Zhang, Z. Qi, L. Duan, R. Yu, Q. Yue, D. Meng, T. Kang, L. Liu, K. Lan, W. Li, D. Zhao, L. Wu and Y. Ma, Adv. Mater., 2025, 37, 2510246.
49. Q. Zhang, B. Gu and W. Fang, Green Chem., 2024, 26, 6261–6288.
50. A. S. Raman and A. Selloni, Angew. Chem., Int. Ed., 2025, 64, e202507721.
51. W. Macyk, K. Szaciłowski, G. Stochel, M. Buchalska, J. Kuncewicz and P. Łabuz, Coord. Chem. Rev., 2010, 254, 2687–2701.
52. S. Xu, K. Chen, Y. Cai, S. Ren, W. Chu, M. Song, Y. Xu and C. Tan, Water Res., 2025, 285, 124139.
53. A. Mondal, A. Pal, S. Sarkar, R. Datta and P. De, Biomacromolecules, 2024, 25, 1649–1659.
54. Z. Luo, S. Zhu, H. Xue, W. Yang, F. Zhang, F. Xu, W. Lin, H. Wang and X. Chen, Angew. Chem., Int. Ed., 2025, 64, e202420217.
55. X. Li, T. Dai, P. Hu, C. Zhang, J. Chen, C. Liu and T. Li, Food Hydrocolloids, 2020, 105, 105761.
56. Z. Meng and K. A. Mirica, Chem. Soc. Rev., 2021, 50, 13498–13558.
57. D. Wang, H. Yang, J. Yang, B. Wang, P. Wasnik, B. B. Xu and Z. Shi, Adv. Compos. Hybrid Mater., 2023, 6, 138.
58. J. Liang, A. Labidi, B. Lu, A. O. T. Patrocinio, A. Sial, T. Gao, S. I. Othman, A. A. Allam and C. Wang, Appl. Catal., B, 2025, 367, 125092.
59. S. Gao, T. Luo, Q. Zhou, W. Luo, H. Li and L. Jing, J. Colloid Interface Sci., 2018, 517, 9–17.
60. L. Sun, X. Ye, Z. Cao, C. Zhang, C. Yao, C. Ni and X. Li, Appl. Catal., A, 2022, 648, 118923.
61. J. Qiu, D. Dai, L. Zhang, Y. Zhou, L. Yang and J. Yao, J. Colloid Interface Sci., 2022, 605, 648–656.
62. J. Wang, Y. Li, H. Liu, Z. Ding, R. Yuan and Z. Li, Angew. Chem., Int. Ed., 2024, 63, e202410397.
63. S. Ntakirutimana, T. Xu, H. Liu, J. Q. Cui, Q. J. Zong, Z. H. Liu, B. Z. Li and Y. J. Yuan, Green Chem., 2024, 24, 5460–5478.
64. D. Dai, J. Qiu, G. Xia, Y. Tang and J. Yao, ACS Catal., 2023, 13, 14987–14995.
65. A. Qayyum, D. A. Giannakoudakis, D. Łomot, R. F. Colmenares-Quintero, K. Nikiforow, A. P. LaGrow and J. C. Colmenares, Ultrason. Sonochem., 2024, 104, 106829.
66. J. Xu, P. Zhou, C. Zhang, L. Yuan, X. Xiao, L. Dai and K. Huo, Green Chem., 2022, 24, 5351–5378.
67. R. Ghalta and R. Srivastava, Green Chem., 2024, 26, 7384–7405.
68. K. Wu, M. Cao, Q. Zeng and X. Li, Green Energy Environ., 2023, 8, 383–405.
69. C. Sinsuebpol, K. Burapapadh, V. Chowjaroen and N. Changsan, J. Adv. Pharm. Technol. Res., 2023, 14, 99–104.
70. W. Liu, X. Cui, Y. Zhong, R. Ma, B. Liu and Y. Xia, Pharmacol. Res., 2023, 193, 106812.
71. J. Li, H. Shi, Z. Li, J. Wang, H. Si, F. Liao, H. Huang, Y. Liu and Z. Kang, Appl. Catal., B, 2024, 343, 123541.
72. X. Xiao, R. Guo, Y. Qi, J. Wei, N. Wu, S. Zhang and R. Qu, J. Hazard. Mater., 2024, 461, 132616.
73. S. T. Kochuveedu, Y. H. Jang and D. H. Kim, Chem. Soc. Rev., 2013, 42, 8467–8493.
74. H. Zhao, C. F. Li, X. Yu, N. Zhong, Z. Y. Hu, Y. Li, S. Larter, M. G. Kibria and J. Hu, Appl. Catal., B, 2022, 302, 120872.
75. X. Du, H. Zhang, T. Yao, S. Dong, L. Jing and J. Hu, Surf. Interfaces, 2024, 48, 104283.

---