Ultrasonic cavitation-driven hierarchical pore engineering of activated carbon for enhanced bilirubin adsorption: Quantitative structure–function relationships Authors: Jiameng Xu, Xinyun Wu, Yuxin Yan, Tao Wu, Manickam Sivakumar, Cheng Heng Pang Published: 2026 ABSTRACT Graphical abstractKeywords: Ultrasonication, Activated carbon, Bilirubin adsorption, Pore size distribution, Hemoperfusion Introduction Bilirubin, the final product of heme catabolism, is physiologically metabolised and cleared by the liver [1]. In cases of liver dysfunction or failure, impaired clearance leads to its accumulation in the bloodstream, resulting in hyperbilirubinemia, a condition strongly associated with increased mortality. Clinically, serum bilirubin concentrations exceeding 3 mg dL−1 and 33 mg dL−1 have been correlated with mortality rates of approximately 30 % and 61 %, respectively [2], [3]. Free (unconjugated) bilirubin can readily cross the blood–brain barrier, leading to severe and potentially irreversible neurological damage [4], [5]. Although liver transplantation remains the definitive treatment, its widespread application is constrained by donor scarcity and the inherent risks associated with major surgery [6]. In this context, hemoperfusion (HP), an extracorporeal detoxification technique that directly removes circulating bilirubin from the bloodstream, serves as an important bridging or supportive therapeutic strategy. However, the activated carbon (AC) and polymeric resin adsorbents currently employed in clinical practice exhibit limited adsorption capacity, which prolongs treatment duration and limits their ability to rapidly reduce bilirubin levels under critical conditions, thereby failing to meet the increasing demand for high-efficiency detoxification systems [7]. To overcome these limitations, recent research has explored various materials engineering strategies, primarily focusing on three key directions. First, the construction of materials with high specific surface areas, such as three-dimensional porous graphene and hierarchical macro- and mesoporous reduced graphene aerogels, has been widely explored to enhance adsorption capacity by increasing the density of accessible active sites [8], [9]. Second, surface chemical functionalization, particularly the introduction of amino groups, as exemplified by UiO-66-NH2 metal–organic frameworks, enhances adsorption performance through electrostatic interactions with bilirubin carboxyl groups, thereby significantly increasing uptake capacity [10]. Third, the development of advanced porous materials, such as supramolecular organic frameworks (SOFs), has achieved exceptionally high adsorption capacities of up to 609.1 mg g−1 along with rapid kinetics, reaching equilibrium within 20 min [11]. Despite these advances, existing strategies remain constrained by fundamental limitations. Bilirubin adsorption is governed by two primary mechanisms: surface interactions and physical confinement. Surface interactions are mediated by the chemical functionality of the adsorbent, encompassing π–π stacking between aromatic domains and the pyrrole rings of bilirubin, hydrogen bonding between surface oxygen-containing groups and bilirubin carboxyl moieties, electrostatic attraction between positively charged sites and the negatively charged carboxyl group (–COO−), and hydrophobic interactions with the nonpolar tetrapyrrole core and side chains, whereas physical confinement arises from molecular sieving within the pore network and is strongly dictated by the pore size distribution [12]. While surface interactions primarily dictate the adsorption affinity of bilirubin for the adsorbent, physical confinement governs molecular accessibility within the pore network and the associated mass transfer efficiency. Therefore, effective adsorption requires a synergistic interaction between these two mechanisms, with pore structure serving as a prerequisite for utilising surface chemistry functionalities. Therefore, elucidating pore size-dependent effects is essential for the rational design of adsorbents, enabling the controlled distribution and optimal density of functional groups within accessible pore domains. This approach ensures that chemical modifications are strategically localised within transport-accessible channels, rather than being confined to the external surface or inducing pore blockage. Therefore, elucidating the role of pore size in HP adsorbents is crucial for achieving enhanced adsorption performance. However, current studies on pore size effects remain limited, as they often fail to distinguish between physical confinement and surface chemistry or focus on narrow pore size ranges. For example, Zhang et al. modified activated carbon via nitric acid treatment. They reported that pores in the ranges of 2.4–4.1 nm and 8.0–30.0 nm were favorable for the adsorption of free bilirubin and albumin-bound bilirubin, respectively [13]. However, this modification concurrently increased the surface concentration of carboxylic groups from 0.14 to 0.23 mmol g−1, thereby introducing additional chemical adsorption sites and precluding the isolation of the independent contribution of pore size. Similarly, a recent study employing ordered mesoporous silica (OMS) systems demonstrated that a pore size of 2.5 nm yields optimal bilirubin adsorption performance through precise synthetic control [14]. Although this study demonstrates the importance of precise control over a single pore size, it does not address the potential influence of hierarchical or hybrid pore size distributions on bilirubin adsorption. However, it provides valuable insight into the relationship between pore size and adsorption. Importantly, the rational design of an optimal pore size distribution capable of simultaneously enhancing mass transfer and adsorption efficiency represents a critical yet underexplored strategy for advancing high-performance bilirubin adsorbents. Tao et al. reported a carbon monolithic material featuring ordered small mesopores (∼3.3–4 nm) interconnected with large mesopores (∼30 nm) [15]. By leveraging the synergistic roles of macropores, which facilitate mass transfer, and mesopores, which provide adsorption sites, the material achieved a high adsorption capacity of 613 mg g−1. While this work highlights the overall advantages of hierarchical pore architectures, it does not establish a quantitative relationship between specific pore size distributions and adsorption performance, which remains an important gap. Similarly, hierarchical pore structures have been widely utilised in fields such as catalysis, separation, energy storage and resource recovery to enhance mass transfer efficiency and the utilisation of active sites [16], [17], [18]. However, most of these studies have been limited to qualitative descriptions of the overall structural advantages and have not yet quantitatively assessed the independent contributions of specific pore size domains. As traditional modification methods typically alter both the pore structure and surface chemistry simultaneously, it remains challenging to distinguish the contributions of each independently. Ultrasound represents a versatile and non-destructive technique that has been extensively exploited in materials synthesis, surface engineering, and structural modification [19], [20]. In this study, a strategy is proposed to largely quantitatively decouple the respective contributions of physical confinement and surface adsorption. In contrast to conventional chemical modification approaches, such as nitric acid treatment, which can induce substantial increases in surface carboxyl group content up to 64 %, ultrasonic treatment is employed to independently tailor the pore size distribution while largely retaining the intrinsic surface chemistry. Ultrasonication in aqueous media primarily restructures the pore network by removing blockages and enhancing pore connectivity through cavitation-induced microjets and shear forces, while inducing only minimal alterations to surface chemical properties [21], [22], [23]. Consequently, the associated changes in surface oxygen content are significantly lower than those observed in conventional chemical oxidation treatments. In this work, ultrasonic power was systematically varied (10–100 %) to tune the pore size distribution while minimising alterations to surface chemistry. By largely decoupling physical confinement from surface interactions, a quantitative relationship between specific pore size domains and bilirubin adsorption performance was established, thereby elucidating the dominant mechanistic contribution of physical confinement relative to the largely retained surface chemistry. These findings provide a robust framework for the rational integration of physical confinement and chemical affinity in adsorbent design, offering both theoretical insight and practical design guidelines for the development of next-generation high-performance HP adsorbents. Materials and methods Chemical reagents and materials Commercial activated carbon (AC, 150–200 mesh) was obtained from Suzhou Carbon Cyclone Activated Carbon Co., Ltd. (Suzhou, China). Sodium hydroxide, phosphate buffer solution (PBS, pH 7.4), bilirubin, and bovine serum albumin (BSA) were obtained from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). Analytical instruments Pore structure was characterised using a 3FLEX high-performance adsorption analyser (Micromeritics, USA). Surface chemical states were examined by X-ray photoelectron spectrometer (XPS, Kratos Axis Supra+, UK) equipped with a monochromatic Al Kα source, and by attenuated total reflectance Fourier transform infrared spectrometer (ATR-FTIR, Bruker Optik GmbH, Germany). Bilirubin concentrations were quantified using an ultraviolet–visible spectrophotometer (UV–Vis, Thermo Fisher Scientific, USA). Ultrasonic treatment was conducted using a probe-type ultrasonic generator (600 W, 20 kHz, Jiangyin Weiheng Industrial Technology Co., Ltd., China). Preparation and characterization of activated carbon Ultrasonic treatment of activated carbon for pore size distribution modulation To tailor the pore size distribution while minimising alterations to surface chemistry, 2 g of activated carbon (AC) was dispersed in 400 mL of deionized water in a jacketed reactor, with the temperature maintained at 25 ± 2 °C via continuous cooling water circulation. The suspension was subjected to ultrasonication at varying power levels (10–100 % of the 600 W output, in 10 % increments) for 0.5 h. Following treatment, the samples were recovered by vacuum filtration, thoroughly washed with deionised water, and dried at 60 °C to a constant weight. The resulting materials were denoted as AC-X, where X represents the applied ultrasonic power percentage (e.g., AC-10, AC-50, and AC-100). Characterization of pore structure and surface chemistry Nitrogen adsorption–desorption isotherms were recorded at 77 K using the 3FLEX analyser. Before measurement, the samples were degassed under vacuum at 300 °C for 8 h. The specific surface area was determined using the multipoint Brunauer-Emmett-Teller (BET) method over a relative pressure range of P/P0 = 0.05–0.30. The pore size distribution was derived from the adsorption branch of the isotherms using the Barrett-Joyner-Halenda (BJH) method, together with density functional theory (DFT) and non-local density functional theory (NLDFT) models. The total pore volume was estimated at a relative pressure of P/P0 ≈ 0.99. The micropore volume was determined using the t-plot method, while the mesopore volume was obtained by subtracting the micropore volume from the total pore volume. Surface chemical states were analysed by attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FTIR) over the range of 400–4000 cm−1 at a resolution of 4 cm−1, as well as by elemental analysis for the determination of the C/O mass ratio, and by X-ray photoelectron spectroscopy (XPS), including survey spectra and high-resolution C 1 s scans. The XPS binding energy scale was calibrated with reference to the C 1 s peak at 284.8 eV. Post-adsorption samples were additionally characterised by nitrogen adsorption–desorption isotherms following the identical procedures. Selected samples were subjected to brief ultrasonication in deionised water, subsequently vacuum-filtered and dried to evaluate the reversibility of pore blockage. Bilirubin adsorption experiments Bilirubin was initially dissolved in 0.1 M NaOH and subsequently diluted with phosphate-buffered saline (PBS, pH 7.4) to a final concentration of 40 mg L−1. Albumin-bound bilirubin was prepared by combining bilirubin with bovine serum albumin (BSA) at a molar ratio of 2:1. All solutions were protected from light and used within 2 h to minimise photodegradation. Before adsorption experiments, the treated AC samples were pre-equilibrated by dispersing 20 mg of adsorbent in 50 mL of PBS and shaking at 25 °C for 12 h to ensure stabilisation. Following equilibration, the suspensions were vacuum-filtered and thoroughly washed with deionized water to remove residual salts. For adsorption experiments, the pre-equilibrated AC was added to 50 mL of bilirubin solution (free or albumin-bound; 40 mg L−1) in a light-protected conical flask. The resulting suspension was agitated at 37 °C under controlled conditions. At predetermined time intervals, aliquots were withdrawn, filtered through 0.22 μm membranes, and analysed using UV–Vis spectrophotometry at 438 nm for free bilirubin and/or 460 nm for albumin-bound bilirubin. All experiments were conducted in triplicate to ensure reproducibility. The adsorption capacity (q, mg g−1) was calculated according to Eq. (1). Where C0 (mg L−1) is the initial bilirubin concentration, C (mg L−1) is the concentration at a given time, V is the volume of the solution, and m (g) is the mass of activated carbon. To evaluate adsorbent reusability, the used AC50 was recovered by vacuum filtration, regenerated by brief ultrasonication in NaOH solution, thoroughly washed with deionised water to neutral, dried at 60 °C to a constant weight, and subjected to repeated adsorption cycles under identical conditions. Adsorption kinetics and isotherm analysis Three kinetic models and three equilibrium isotherm models were employed to elucidate the adsorption mechanism, surface heterogeneity, and underlying energetic mechanisms. For kinetics, the pseudo-first-order (PFO), pseudo-second-order (PSO), and Weber–Morris intraparticle diffusion models were applied to identify the dominant rate-controlling steps and characterise the underlying mass transfer processes. The adsorption kinetics were analysed using the PFO and PSO models, as described by Eqs. (2), (3), respectively. where qe (mg g−1) is the equilibrium adsorption capacity, qt (mg g−1) is the adsorption capacity at time t, k1 (min−1) is the rate constant of the pseudo-first-order (PFO) model, and k2 (g mg−1 min−1) is the rate constant of the pseudo-second-order (PSO) model. Finally, the Weber-Morris intraparticle diffusion model (Eq. (4) was applied to elucidate the diffusion-controlled stages of the adsorption process. where kp (mg g−1 min−1/2) is the intraparticle diffusion rate constant, and C (mg g−1) is the intercept associated with the boundary layer thickness. For equilibrium isotherm analysis, the Langmuir, Freundlich, and Dubinin–Radushkevich (DR) models were employed to evaluate the adsorption capacity, surface heterogeneity, and adsorption energy, respectively. The Langmuir isotherm (Eq. (5) assumes monolayer adsorption onto a homogeneous surface. where qm (mg g−1) is the maximum monolayer adsorption capacity, and KL (L mg−1) is the Langmuir constant related to adsorption affinity. The Freundlich isotherm (Eq. (6) describes multilayer adsorption on a heterogeneous surface. where KF [mg g−1 (L mg−1)1/n] is the Freundlich constant indicative of adsorption capacity, and n is the heterogeneity factor. The Dubinin–Radushkevich (DR) isotherm (Eq. (7) was employed to characterise the adsorption energy and discriminate between physical and chemical adsorption. where ε denotes the polanyi potential, defined as RT ln(1 + 1/Ce). β, R, and T stand for the adsorption energy-related constant, the gas constant (8.314 J/mol·K), and the absolute temperature (K), respectively. E (kJ/mol) refers to the mean free energy of adsorption for each adsorbate molecule as it moves from the bulk solution to the solid surface. Results and discussion Effect of ultrasonication on the pore structure of AC Specific surface area and total pore volume analysis Nitrogen adsorption–desorption isotherms revealed a pronounced increase in pore volume following ultrasonic treatment. As illustrated in Fig. 1a, both AC0 and AC100 exhibit combined Type I/IV isotherm characteristics, indicative of a hierarchical micro-mesoporous structure. The steep uptake at low relative pressures (P/P0 < 0.1) indicates the presence of abundant micropores, while the H4-type hysteresis loop suggests the existence of slit-shaped mesopores, typically associated with carbonaceous materials [24], [25], [26]. AC100 exhibited a significantly higher N2 uptake over the entire relative pressure range, indicating an overall enhancement in porosity. Notably, the increased adsorption at low relative pressures (P/P0 < 0.1) suggests the reopening of previously inaccessible micropores, whereas the enhanced uptake at higher relative pressures (P/P0 > 0.4) reflects the development of meso- and macroporous connectivity induced by ultrasonic treatment. Furthermore, Fig. 1b shows a significantly increased filling rate for AC100 at very low relative pressures, indicating an increased micropore population. Nitrogen adsorption–desorption isotherms of activated carbon (AC) before (AC0) and after ultrasonic treatment (AC100), illustrating the evolution of hierarchical pore structure. (a) Full-range nitrogen adsorption–desorption isotherms. (b) Enlarged view of the low-pressure region (P/P0 = 10-7 to 10-2, logarithmic scale) of the nitrogen adsorption isotherms. The pore structure was characterized using Brunauer–Emmett–Teller (BET), t-plot, and Barrett–Joyner–Halenda (BJH) analyses. The BET method was employed to determine the specific surface area (SBET), while the t-plot method was used to quantify the micropore volume (Vmicro) and distinguish it from the mesopore volume (Vmeso). In addition, the BJH model, applied to the adsorption branch of the isotherms, was used to estimate the average pore diameter (DBJH). As shown in Fig. 2, the pore structural parameters of AC exhibit a non-monotonic evolution with increasing ultrasonic power, characterised by an initial pore expansion, followed by partial collapse and subsequent structural reorganization. From AC0 to AC50, a substantial enhancement in porosity is observed, with the specific surface area (SBET) increasing from 642.66 m2 g−1 to 986.13 m2 g−1 and the total pore volume (Vtotal) rising from 0.73 to 1.10 cm3 g−1. An ultrasonic power of 50 % (300 W) yielded the most favourable pore characteristics, with a balanced development of micropores (Vmicro = 0.40 cm3 g−1) and mesopores (Vmeso = 0.70 cm3 g−1). At higher power levels (60–80 %), the pore structure exhibited pronounced degradation. For example, at 60 % power, both the specific surface area (SBET) and total pore volume (Vtotal) decreased sharply to 696.27 m2 g−1 and 0.74 cm3 g−1, respectively. Although a slight and transient recovery was observed at 70 % power (SBET = 731.97 m2 g−1, Vtotal: 0.78 cm3 g−1), this improvement was not sustained, as evidenced by the subsequent decline at 80 % power (SBET: 691.20 m2 g−1). This non-monotonic behaviour indicates structural instability at elevated ultrasonic intensities. The observed deterioration is attributed to excessive cavitation-induced microjets and shear forces that exceed the mechanical integrity of the carbon framework, resulting in pore collapse and partial structural degradation. At the highest ultrasonic intensities (90–100 %), SBET exhibits a notable recovery, reaching 881.81 m2 g−1 at 100 % power. This behaviour is likely associated with the partial reorganisation of the carbon framework following cavitation-induced collapse at intermediate power levels, leading to the formation of newly accessible pore structures. Despite these fluctuations, BJH analysis indicates that the average pore diameter (DBJH) remains relatively constant at 4.22–4.71 nm across all samples. Moreover, the BJH model is primarily sensitive to the mesopore range (2–50 nm) and underestimates micropore contributions. Nevertheless, the t-plot data in Fig. 2 already reveal a marked redistribution between Vmicro and Vmeso, confirming that ultrasonic cavitation enhances accessible porosity mainly by unblocking channels and improving connectivity rather than by directly etching pore walls. Evolution of pore structural parameters of activated carbon (AC) as a function of ultrasonic power. Evolution of micropore and mesopore size distributions To comprehensively characterise the full-range pore size distribution under varying ultrasonic powers, density functional theory (DFT/NLDFT) analysis and cumulative pore volume distributions were employed. As shown in Fig. 3, AC0 was used as the reference material to evaluate the evolution of pore size distribution under varying ultrasonic powers. As a representative sample within the low-power regime, AC30 lies near the midpoint of the vibration-dominated range (10–40 %). As shown in Fig. 2, the moderate increases in both micropore and mesopore volumes at this stage reflect the early phase of pore development, characterised by effective pore opening with minimal structural disruption. In this low-power regime, the applied ultrasonic energy remains insufficient to activate and open the pore network fully. In contrast, AC50 exhibits the highest specific surface area and pore volume, indicating an optimally developed pore structure. However, whether this structural optimisation directly translates into superior adsorption performance remains to be validated. AC80 represents a transitional state within the high-power regime (60–90 %), where a dynamic competition exists between cavitation-induced structural degradation and concurrent pore restructuring. In contrast, AC100 reflects the predominance of structural reorganisation under maximum ultrasonic intensity. These observations collectively reveal a three-stage evolution of the pore architecture with increasing ultrasonic power: initial pore expansion, followed by partial collapse, and ultimately structural reconstruction. Pore size distributions of activated carbon (AC) under varying ultrasonic powers. (a) DFT/NLDFT-derived differential pore size distribution curves for representative samples AC0, AC30, AC50, AC80, and AC100. (b) Cumulative pore volume distributions of all samples, highlighting the contribution of different pore size ranges to total porosity. AC0 exhibits broad pore size distribution peaks centred at ∼ 0.5 nm (micropores), ∼1.3 nm (small micropores), ∼3.5 nm (small mesopores), and 7–10 nm (medium mesopores), indicative of a heterogeneous pore structure with limited connectivity. The pronounced peak broadening suggests pore blockage, likely due to amorphous carbon residues and ink-bottle-type pore geometries [27], [28]. Following ultrasonic treatment, the pore size distributions became progressively sharper and more well-defined, reflecting improved pore accessibility and structural regularity. The pore size distribution of AC30 becomes more uniform. However, a noticeable reduction in peak intensity is observed in the 0–1 nm and > 20 nm regions. This suggests that low-power ultrasonication primarily facilitates the removal of pore blockages, enhancing accessibility without significantly increasing the overall pore volume. In contrast, AC50 exhibits the most well-defined and intensified peaks across the entire pore size range, with particularly pronounced development in micropores (1–2 nm) and small mesopores (3–5 nm). These results indicate that an ultrasonic power of 50 % effectively optimises pore accessibility, yielding a more uniform and interconnected pore network without significant structural collapse. At higher power levels, AC80 exhibits broadened and flattened distributions in the > 10 nm region, indicative of pore wall damage and partial collapse induced by excessive ultrasonic intensity, consistent with the marked reduction in surface area observed in Fig. 2 at 60–80 % power. AC100, while retaining a similarly broadened and flattened > 10 nm profile reflecting residual damage from the intermediate-power collapse, simultaneously displays re-intensified peaks in the micropore and small mesopore ranges (Fig. 3a). This micropore recovery accounts for the notable SBET rebound observed in Fig. 2:, supporting the interpretation that partial reorganization of the carbon framework generates newly accessible pore structures at the highest ultrasonic intensities. Similar trends have been reported in previous studies, where excessive cavitation was shown to hinder the development of porosity in activated carbon [29]. The cumulative pore volume distributions further quantify these structural variations. As shown in Fig. 3b, pores smaller than 20 nm (particularly those in the 1–5 nm and 10–20 nm ranges) contribute predominantly to the total porosity, whereas pores (>30 nm) constitute only a minor fraction. Notably, discrepancies are observed between the pore volume trends derived from the t-plot method (Fig. 2) and those obtained from DFT/NLDFT analysis (Fig. 3) in the low-power ultrasonic regime (AC10–AC40). Specifically, the t-plot analysis suggests an increase in pore volume relative to AC0, whereas the DFT/NLDFT results indicate a lower cumulative pore volume. As highlighted by Ania et al., such discrepancies arise from fundamental differences in the underlying assumptions of these methods when characterising activated carbon pore structures [30]. In the low-power ultrasonic regime, the pore network undergoes competing effects of enhanced connectivity and redistribution of pore sizes. Consequently, the t-plot and DFT/NLDFT approaches exhibit different sensitivities to these structural modifications, leading to the observed divergence in pore volume trends. At 50 % ultrasonic power, all pore size fractions exhibit a pronounced increase, reaching maximum values in both microporous and mesoporous regions, indicative of optimal pore development. At higher power levels (AC60–AC100), the pore structure undergoes progressive collapse followed by partial reconstruction. Specifically, AC60 shows a marked reduction in pore volume, while samples in the AC70–AC90 range display pronounced structural fluctuations. These variations reflect an unstable reconstruction process at elevated ultrasonic intensities, in which pore blockage arising from structural fragmentation dynamically competes with the formation of new pore structures. Surface chemical composition analysis Surface functional groups were characterised using ATR-FTIR, elemental analysis, and XPS to verify that ultrasonic treatment predominantly alters the pore structure while preserving surface chemical properties. A comparative analysis between AC0 and ultrasonicated samples (AC10-AC100) was conducted to assess potential variations in functional groups relevant to bilirubin adsorption. The FTIR spectra show a high degree of similarity across all samples, particularly for functional groups associated with bilirubin adsorption, indicating minimal chemical modification following ultrasonic treatment. As shown in Fig. 4, all samples exhibit characteristic absorption bands corresponding to the weak alkyne vibration (∼2098 cm−1), aromatic C=C stretching and C-H bending (approximately 1456 cm−1), and out-of-plane C-H bending (approximately 876 cm−1). Although minor variations in peak intensities are observed, these changes are likely attributable to slight surface oxidation or local structural rearrangements induced by increasing ultrasonic power. Notably, the peak positions remain essentially unchanged across all samples, indicating that no major changes occurred in the dominant surface functional groups following ultrasonic treatment. These aromatic domains provide the π-electron systems necessary for π-π stacking interactions with the tetrapyrrole rings of bilirubin, which represent a key chemical adsorption mechanism [12], [31]. The absence of detectable peak shifts or significant intensity changes in these adsorption-relevant bands further supports that ultrasonic treatment retains the chemical affinity sites essential for effective bilirubin adsorption. ATR-FTIR spectra of activated carbon (AC) samples subjected to different ultrasonic power levels, demonstrating the preservation of surface functional groups following ultrasonication. Elemental analysis and XPS analysis further corroborate that ultrasonic treatment induces only minor variations in the bulk oxidative state and surface chemistry of AC. As shown in Fig. 5a, the C/O mass ratio fluctuated narrowly across all samples, ranging from 1.84 to 2.01 (CV = 3.3 %). Notably, a slight downward trend is observable at the highest power levels, with the C/O ratio decreasing from 1.99 (AC0) to 1.84–1.87 (AC80–AC100), suggesting a modest increase in oxygen content under extreme cavitation. However, this variation remains minor and falls within the overall narrow fluctuation range, indicating that substantial oxidation of the carbon framework is absent under most ultrasonic intensities. XPS analysis further validates the preservation of surface chemistry. As shown in Fig. 5b, the total oxygen-containing functional groups (C–O + C=O + O–C=O) remained essentially invariant across all samples, ranging from 20.02 % to 22.18 % (CV = 3.8 %), with no discernible power-dependent trend. The largely invariant surface chemistry, with only slight oxidation at the highest power as evidenced by both bulk C/O and surface XPS data, ensures that chemical adsorption contributions, including π–π stacking and hydrogen bonding, remain comparable across the samples. Collectively, the FTIR, elemental analysis, and XPS results demonstrate that ultrasound treatment predominantly modifies the pore structure while largely retaining surface functional groups relevant to bilirubin adsorption. This establishes a predominantly controlled chemical background for subsequent adsorption studies, enabling a comparative assessment of the dominant role of physical confinement. Verification of surface chemical stability of activated carbon (AC) samples under varying ultrasonic power levels. (a) C/O mass ratio of AC samples as a function of ultrasonic power, determined by CHNS/O elemental analysis. Values represent the mean of triplicate measurements. (b) Total oxygen-containing functional groups (C–O + C=O + O–C=O) content of AC samples as determined by XPS under varying ultrasonic power levels, demonstrating minimal variation in surface oxygen-containing groups. Bilirubin adsorption on ultrasonicated AC Adsorption capacity The adsorption capacity for free bilirubin (qB) is presented in Fig. 6. A clear non-monotonic dependence on ultrasonic power is observed. The AC50 sample exhibits the highest adsorption capacity of 84.23 mg g−1, representing increases of 16.1 % and 10.8 % compared to AC0 (72.55 mg g−1) and AC100 (76.04 mg g−1), respectively. In contrast, samples treated at low (10–40 %) and intermediate-high (60–90 %) ultrasonic powers show adsorption capacities comparable to or lower than that of AC0, ranging from 70.07 to 75.59 mg g−1. These variations closely track the evolution of specific surface area and pore volume (Fig. 2), suggesting that adsorption capacity is primarily governed by the availability of accessible pore domains for bilirubin confinement. At low ultrasonic power, pore opening is incomplete. At 50 % power, the optimal balance between pore development and structural preservation is achieved. At higher power levels, the marked reduction in BET surface area and total pore volume (Fig. 2), together with the broadened pore-size distribution indicative of pore-wall collapse (Fig. 3), reduces the accessible adsorption volume and the internal surface area available for bilirubin, thereby lowering qB. Effect of ultrasonic power on the adsorption capacity of activated carbon (AC) for free bilirubin (qB). In contrast, the adsorption capacity for albumin-bound bilirubin (qAB) shows a distinct trend. As shown in Fig. 7, at low ultrasonic power levels (AC10-AC20), qAB decreases markedly from 45.26 mg g−1 (AC0) to 33.85–34.09 mg g−1, indicating that partially blocked pore networks hinder the diffusion and accommodation of the larger bilirubin–albumin complexes. As the ultrasonic power increases beyond 30 %, qAB progressively recovers, reaching 44.36 mg g−1 for AC100, suggesting improved pore accessibility and transport pathways for macromolecular adsorption. This recovery is attributed to the generation of additional transport-relevant pores, the reopening of previously obstructed or inaccessible pores, or enhanced interparticle channel connectivity, as evidenced by the increase in cumulative pore volume, which facilitates the diffusion of bilirubin–albumin complexes. The results suggest that the formation of larger pores at elevated ultrasonic powers partially compensates for local structural collapse, thereby restoring adsorption capacity to near-baseline levels. The distinct adsorption behaviours of free bilirubin and albumin-bound bilirubin highlight the relationship between pore accessibility and molecular size. Free bilirubin adsorption is primarily governed by the accessibility of micropores and small mesopores, resulting in a non-monotonic response to ultrasonic power due to competing effects of pore opening and structural degradation. In contrast, adsorption of the bilirubin–albumin complex is strongly influenced by steric constraints, with performance largely dependent on the availability of larger transport pathways and interparticle pores. Effect of ultrasonic power on the adsorption capacity of activated carbon for albumin-bound bilirubin (qAB), illustrating the dependence of macromolecular adsorption on pore accessibility and transport pathways. Adsorption kinetics and isotherms To comprehensively understand the adsorption behaviour, both kinetic and equilibrium experiments were conducted. To elucidate the kinetic mechanisms underlying the distinct adsorption behaviours, the PFO, PSO, and Weber-Morris intraparticle diffusion models were employed. AC50 was selected as the representative sample for kinetic analysis due to its highest adsorption capacity (84.23 mg g1) among all tested samples, thereby providing a robust basis for distinguishing between diffusion-controlled and surface-controlled adsorption processes. As shown in Fig. 8a-b and Table 1, the PSO model consistently provides a superior fit to the experimental data (R2 > 0.99) than the PFO model, suggesting that the rate-limiting step involves surface-site occupation rather than external diffusion alone. Adsorption kinetics and isotherm analysis of AC samples. (a) Comparison of pseudo-first-order (PFO) and pseudo-second-order (PSO) model fits with experimental data for AC50. (b) Pseudo-second-order kinetic model fit for AC50. (c) Weber–Morris intraparticle diffusion plots showing three diffusion stages for AC0, AC50, and AC100. (d) Langmuir isotherm model fit for AC50. (e) Freundlich isotherm model fit for AC50. (f) Dubinin–Radushkevich (DR) isotherm model fit for AC50. Table 1Parameters of the PFO and PSO models for bilirubin adsorption onto AC50.ModelParameterUnitAC50Pseudo-first-orderqe, calmg g−183.370k1min−10.101R2–0.955Pseudo-second-orderqe, calmg g–197.100k2g mg−1 min−10.003R2–0.997 Parameters of the PFO and PSO models for bilirubin adsorption onto AC50. This behaviour is attributed to non-covalent interactions between bilirubin molecules and the adsorbent surface, including π–π stacking between aromatic domains and the tetrapyrrolic rings of bilirubin, as well as hydrogen bonding and electrostatic interactions involving oxygen-containing functional groups. Physical confinement within the pore network plays a critical role in governing the accessibility and retention of bilirubin molecules. While surface interactions provide the intrinsic driving force for adsorption, mass transfer through the pore network ultimately controls the rate and extent of adsorption. Therefore, diffusion within the pore structure represents the critical transport barrier that determines the overall adsorption capacity. Because the surface chemistry remains largely invariant across all samples, the contribution of chemical adsorption can be considered as a constant baseline. Consequently, the observed differences in adsorption capacity with varying ultrasonic power predominantly reflect variations in mass transfer resistance, which is governed by the pore structure (Section 3.1) and controls access to the underlying chemical adsorption sites. PFO and PSO were performed on AC50 alone because the underlying adsorption mechanism is dictated by surface properties that remain essentially constant across all samples. In contrast, the Weber-Morris intraparticle diffusion analysis was conducted across AC0, AC50, and AC100 to capture the pore-structure-dependent evolution of diffusion behaviour with varying ultrasonic power. This analysis reveals a characteristic three-stage linear diffusion behaviour, corresponding to distinct mass transfer regimes. The diffusion rate constants obtained for each stage reflect the underlying evolution of pore-size distribution. Specifically, AC0 represents the untreated baseline, AC50 corresponds to the optimally developed pore structure with maximum porosity, and AC100 reflects the structural reorganization under the highest ultrasonic power. As shown in Fig. 8c and Table 2, in the initial stage (t0.5 < 3 min0.5), AC100 exhibits the highest rate constant (18.66 mg g−1 min−0.5), followed by AC50 and AC0, indicating that increasing ultrasonic power enhances external mass transfer by improving surface accessibility and reducing diffusion resistance at the solid–liquid interface. The intercept values for AC0 and AC50 are positive (7.81 and 7.48 mg g−1, respectively), indicating the presence of a boundary-layer effect. In contrast, AC100 exhibits a negative intercept (−2.75 mg g−1), suggesting a diminished or altered boundary layer contribution. This shift may reflect enhanced surface accessibility and reduced external mass transfer resistance at higher ultrasonic power. Overall, these results indicate that ultrasonic-induced modifications to the pore structure significantly influence boundary layer diffusion behaviour. The second stage (3 < t0.5 < 10 min0.5) corresponds to intraparticle diffusion, with the rate constants following the trend AC0 ≈ AC100 > AC50. The relatively lower K2 value for AC50 is associated with its highly developed microporous structure (Vmicro = 0.40 cm3 g−1), where increased micropore volume can impose greater diffusion resistance. This behaviour suggests that, although micropores contribute significantly to adsorption capacity, their restricted dimensions may limit mass transfer rates within the pore network. In contrast, AC100 exhibits a higher K2, likely reflecting reduced diffusion resistance associated with its larger, more open pore structure, whereas AC0 shows comparable K2 values, attributable to its simpler pore network. Despite this diffusion limitation observed in AC50, it displays the highest intercept (56.62 mg g−1) at this stage, significantly exceeding those of AC0 and AC100. This elevated intercept indicates a greater contribution from intraparticle adsorption, suggesting that the well-developed internal pore network of AC50 provides a high density of accessible adsorption sites, thereby sustaining superior adsorption capacity. Table 2Parameters of the Weber–Morris intraparticle diffusion model for bilirubin adsorption onto activated carbon samples.StageParameterUnitAC0AC50AC100Stage I (Boundary layer diffusion)K1mg g−1 min−1/213.33616.65418.659C1mg g−17.8117.477−2.755R2–0.9580.9870.923Stage II (Intraparticle diffusion)K2mg g−1 min−1/23.4423.0263.337C2, mg g−1mg g−140.54556.61745.483R2–0.9920.9590.982Stage III (Equilibrium stage)K3mg g−1 min−1/20.1830.2920.181C3, mg g−1mg g−171.25881.42674.271R2–0.8620.9480.922 Parameters of the Weber–Morris intraparticle diffusion model for bilirubin adsorption onto activated carbon samples. The advantage of AC50 becomes most evident in the third stage approaching equilibrium (t0.5 > 10 min0.5). Although the diffusion rate constants of all samples decrease significantly, AC50 maintains higher values than both AC0 and AC100. In particular, AC50 exhibits the highest intercept C3 value (81.43 mg g−1) alongside a sustained diffusion rate constant K3 (0.29 mg g−1 min−0.5), consistent with its enhanced pore volume. This behaviour indicates that the well-developed micropore-small mesopore network in AC50 provides continuous access to adsorption sites, thereby supporting sustained adsorption during the final stage. In contrast, AC0, characterised by a limited pore volume (Vtotal = 0.73 cm3 g−1), exhibits early saturation, accompanied by lower K3 and C3 values. Although AC100 demonstrates enhanced mass transfer in the initial stages, it experiences pronounced diffusion limitations at later stages, likely due to structural instability and reduced effective pore connectivity. Based on the broadened and flattened pore size distribution in the > 10 nm region (Fig. 3), together with the reduced BET surface area at high ultrasonic power (Fig. 2), this behaviour is attributed to structural fragmentation of the pore network. Such fragmentation can generate partially inaccessible micropores, thereby limiting effective utilisation of internal adsorption sites. Overall, AC50 achieves an optimal balance at equilibrium. Its well-developed microporosity provides a substantial adsorption capacity, while the preserved pore connectivity ensures sustained mass transfer throughout the equilibrium phase. To characterise the equilibrium adsorption behaviour, equilibrium adsorption experiments were conducted on AC50 at initial concentrations ranging from 10 to 120 mg L−1. The experimental data were fitted using the Langmuir, Freundlich and DR models (Fig. 8d–f, Table 3). The Langmuir model exhibited a relatively low coefficient of determination (R2 = 0.683), indicating that the assumptions of energy-homogeneous equivalent adsorption sites and monolayer coverage were insufficient to describe the equilibrium behaviour of bilirubin on ultrasonically modified activated carbon. In contrast, the Freundlich model exhibited the highest coefficient of determination (R2 = 0.895), with a 1/n value of 0.141, suggesting favourable adsorption on a heterogeneous surface with multilayer coverage characteristics. The high R2 value suggests that the process is better described by multilayer coverage on a heterogeneous surface rather than ideal monolayer adsorption. This is consistent with the structural heterogeneity of activated carbon, which exhibits a broad pore-size distribution and a diverse set of surface functional groups. The DR model yielded a theoretical saturation capacity of 154.78 mg g−1 and an average adsorption free energy E of 14.56 kJ mol−1. Although its coefficient of determination (R2 = 0.734) was lower than that of the Freundlich model, the E value exceeds the conventional threshold of 8 kJ mol−1, suggesting that the adsorption energy falls within the realm of chemical interactions, thereby supporting the chemisorption-controlled mechanism inferred from the PSO kinetic analysis. Overall, the kinetic and isotherm analyses indicate that the adsorption of bilirubin onto AC50 is governed by surface-mediated chemical interactions on a structurally heterogeneous surface. The favourable PSO fit reflects chemisorption-controlled adsorption kinetics, whilst the Freundlich isotherm describes multilayer coverage behaviour on a chemically heterogeneous surface. The DR model further corroborates this interpretation, yielding an adsorption energy that exceeds the physical adsorption threshold, thereby confirming that bilirubin adsorption onto AC50 is predominantly governed by chemical interactions. Table 3Fitting parameters of Langmuir, Freundlich, and Dubinin–Radushkevich (DR) isotherm models for bilirubin adsorption on AC50.ModelParameterUnitAC50Langmuir ModelKLLmg-10.108qmmgg-1183.648R2–0.683Freundlich ModelKFmgg-1(Lmg-1)1/n80.8311/n–0.141R2–0.895Dubinin-Radushkevich (DR) ModelKDRmol2kJ-22.360*10-3qmmgg-1154.776R2–0.734EkJmol-114.563 Fitting parameters of Langmuir, Freundlich, and Dubinin–Radushkevich (DR) isotherm models for bilirubin adsorption on AC50. Regeneration, reusability, and evaluation To evaluate the reusability of the optimised adsorbent, three adsorption-regeneration cycles were carried out on AC50. As shown in Fig. 9, the removal efficiency remained at a high level of 97 % after the first cycle, but gradually decreased to 88.9 % and 78 % after the second and third cycles, respectively. This gradual decline in removal efficiency suggests that some adsorption sites were occupied or that the pore channels were blocked by residual bilirubin, which could not be completely removed during the regeneration process. Nevertheless, the material retained a considerable adsorption capacity after three cycles, indicating that it possesses a certain potential for reusability in HP applications. Removal efficiency of AC50 for bilirubin over three consecutive adsorption–desorption cycles. The adsorption performance of AC50 was further compared with that of recently reported bilirubin adsorbents, with the results summarised in Table 4. The capacity of AC50 (84.23 mg g−1) is comparable to or higher than that of conventional activated carbon-based materials, such as commercial coconut shell activated carbon (48.2 mg g−1) and date pit-derived carbon (2.72 mg g−1). However, it remains lower than that of advanced materials, including glucose/MAPTAC-derived carbon (141.00 mg g−1) and 3D porous graphene (126.10 mg g−1). This capacity gap is to be expected, as the latter typically involves fine chemical functionalisation (such as amino grafting or polymer coating) or ordered mesoporous structures to maximise surface accessibility. It is worth noting that this strategy achieves a 16.1 % improvement over untreated AC through physical ultrasonic treatment alone, without the need for additional chemical modifiers or complex synthetic steps. This highlights the practical advantages of ultrasonic-driven pore engineering as a green, low-cost post-treatment strategy for enhancing the performance of clinically relevant carbon-based adsorbents. Table 4Comparison of maximum bilirubin adsorption capacities (qm) of various carbon-based adsorbents reported in the literature.MaterialMethodCarbonization Temperature (℃)qm(mg/g)RefDate pitsCO2 activationCarbonization: 600 ℃, Activation: 900 ℃2.72[32]Coconut shellHNO3 reflux modification, 4 h− (commercial activated carbon)48.20[13]Glucose and MAPTACOne-pot hydrothermal carbonization180 ℃, 6 h141.00[33]Three-dimensionally porous grapheneThermal exfoliation of graphite oxide followed by CO2 activationThermal exfoliation: 800 ℃, CO2 activation: 850 ℃126.10[8]Activated carbonUltrasonication− (commercial activated carbon)84.23this work Comparison of maximum bilirubin adsorption capacities (qm) of various carbon-based adsorbents reported in the literature. Relationship between pore size and adsorption capacity Linear correlation analysis reveals that the adsorption capacity of free bilirubin (qB) is closely associated with specific pore size domains. As shown in Fig. 10a-b, a pronounced positive correlation is observed for pores in the 1–5 nm and 10–20 nm ranges, indicating their dominant contribution to adsorption performance. Further quantitative analysis (Fig. 10c) demonstrates a strong linear relationship between the pore volume within the 1–5 nm range and qB, confirming that these pores play a critical role in governing adsorption capacity. This pore size domain, characterised by a high slope and a strong correlation coefficient, indicates that increases in the 1–5 nm pore volume are most reliably associated with increases in adsorption capacity. Further analysis (Fig. 10d) reveals that, within this range, a narrower sub-domain of 2–3 nm exhibits a particularly strong correlation with qB (R2 = 0.807). Pore size-dependent adsorption behaviour of bilirubin. (a) slope heatmaps of linear correlations between pore volume and adsorption capacities (qB, qAB), (b) corresponding R2 of heatmaps indicating the strength of these correlations, (c) representative linear relationships between pore volume in the 1–5 nm range and qB and > 30 nm range and qAB, (d) detailed subdivision of the 1–5 nm (1–2, 2–3, 3–4, 4–5 nm) correlated with qB, (e) subdivision of the 5–10 nm range (5–6, 6–7, 7–8, 8–9, 9–10 nm) correlated with qAB, and (f) subdivision of the > 20 nm range (20–30, 30–40, 40–50, 50–100 nm) correlated with qAB. Statistical significance is indicated as *p < 0.05, **p < 0.01, and ***p < 0.001. To further clarify the functional roles of these pore domains, the pore structure of AC50 was characterised before adsorption, after adsorption, and after subsequent ultrasonic treatment. As shown in Fig. 11, the pore volumes of both the 2–3 nm and 10–20 nm ranges decreased significantly following bilirubin adsorption, by 52.1 % and 42.5 %, respectively. This concurrent reduction indicates that bilirubin molecules can exhibit synergistic contributions of confinement-driven capture and volumetric storage. Variation of pore volume in the 2–3 nm and 10–20 nm ranges for AC50 before adsorption, after bilirubin adsorption, and after sonication treatment. By contrast, the correlation between pore size distribution and the adsorption capacity for qAB is weaker. As shown in Fig. 10a-b, no pore size range below 30 nm can reliably predict qAB (all R2 < 0.4). However, pores larger than 30 nm exhibit the highest R2 value (0.473). Fig. 10e further divided the 5–10 nm range and confirms the absence of significant correlations for qAB across all sub-domains (5–6, 6–7, 7–8, 8–9, and 9–10 nm). Fig. 10f further subdivides the > 20 nm range into four sub-intervals (20–30, 30–40, 40–50, and 50–100 nm), revealing similarly weak correlations across most of this range. Notably, within these finer subdivisions, only the 9–10 nm and 40–50 nm sub-intervals exhibit marginally higher R2 values relative to their domains, though the overall correlations remain weak (R2 < 0.5). These results further support the conclusion that pore size distribution is a weak independent predictor of albumin-bound bilirubin adsorption capacity, suggesting that pore accessibility is a necessary but insufficient condition. The sharp decline in qAB at low ultrasonic power (AC10–AC20) suggests that the absence of large transport pores (>30 nm) severely limits adsorption by preventing complexes from reaching the carbon surface. Conversely, the plateau in qAB at approximately 44 mg/g (AC50–AC100) demonstrates that, despite significant restructuring of the pore architecture, the adsorption capacity never exceeded the initial AC0. This suggests that once surface accessibility is restored, the ultimate adsorption limit depends on the chemical affinity required for the adsorbent to compete with albumin for the release of bilirubin, rather than on pore volume. Consequently, the linear correlation between qAB and pore size is weak, as pore optimisation can only influence the transport stage, whilst the competitive displacement stage is governed by surface chemistry, which remained essentially constant in this study. Overall, the data suggest that for albumin-bound bilirubin, pore engineering is effective only until surface accessibility is achieved. Beyond this point, enhancing adsorption requires targeted chemical functionalisation to strengthen the ability to displace bilirubin from albumin. Future adsorbent designs should therefore retain a broad hierarchical pore distribution spanning 2–3 nm, 10–20 nm, and > 30 nm domains to ensure sufficient internal surface area and transport pathways, while incorporating appropriately oriented functional groups (e.g., carboxyl, pyridinic, or amino moieties) to enhance π–π interactions, hydrogen bonding, and moderate electrostatic attraction. Such an integrated strategy would address both the physical accessibility and chemical affinity requirements, enabling the synergistic coupling of transport and competitive binding for the efficient removal of albumin-bound bilirubin. Conclusion In this study, ultrasonic treatment (10–100 % power) was used to systematically modulate the pore size distribution of activated carbon (AC). Cavitation-induced microjets and shear forces were utilised to open blocked pores and reorganise the pore network, while preserving the inherent stability of the surface chemistry. This approach enabled the evaluation of the contribution of pore structure to bilirubin adsorption. AC50 exhibited the highest specific surface area (986.13 m2 g−1) and adsorption capacity (84.23 mg g−1), corresponding to increases of 53.5 % and 16.1 %, respectively, compared with AC0. These results demonstrate that ultrasonic modulation simultaneously optimises pore structure and adsorption performance. Quantitative correlation analysis identified optimal pore size domains for free bilirubin adsorption. Pores in the 2–3 nm range exhibited the strongest linear correlation with adsorption capacity (R2 = 0.807) and served as size-matched confinement zones that promote initial molecular trapping through enhanced near-wall confinement. In parallel, mesopores in the 10–20 nm range (R2 = 0.808) acted as adsorption chambers, providing sufficient volume for molecular accommodation. These complementary pore domains synergistically enhance adsorption performance through the combined effects of efficient molecular access and confinement-driven stabilisation. For albumin-bound bilirubin, the correlation between pore size and adsorption capacity is markedly weaker (R2 < 0.5 across all pore size domains, with a maximum of 0.473), significantly lower than that observed for free bilirubin. This behaviour indicates a transition from confinement-influenced adsorption to transport-limited behaviour. These results suggest that the effective removal of albumin-bound bilirubin depends on a sequential process involving transport through larger pores and subsequent capture within specific pore domains, highlighting the synergistic relationship between transport pathways and confinement-enabled adsorption. By employing ultrasonic treatment to decouple and quantify the contribution of physical confinement to a significant degree, this study establishes a mechanistic foundation for the rational design of next-generation hemoperfusion adsorbents. Maintaining 2–3 nm size-matched confinement zones and 10–20 nm adsorption chambers, while selectively introducing functional groups (e.g., carboxyl or pyridinic moieties) within these pore domains and preserving > 30 nm pores as transport pathways, enables the synergistic optimisation of physical confinement and surface affinity. This design strategy enables the efficient simultaneous removal of both free and albumin-bound bilirubin, offering practical guidance for the development of advanced adsorbent materials for the treatment of liver failure. CRediT authorship contribution statement Jiameng Xu: Writing – original draft, Validation, Methodology, Formal analysis, Data curation, Conceptualization. Xinyun Wu: Writing – review & editing, Validation, Methodology, Investigation, Formal analysis. Yuxin Yan: Writing – review & editing, Project administration, Methodology, Formal analysis. Tao Wu: Supervision, Project administration, Investigation, Formal analysis. Manickam Sivakumar: Writing – review & editing, Project administration, Investigation, Formal analysis. Cheng Heng Pang: Writing – review & editing, Visualization, Validation, Supervision, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis. Declaration of competing interest The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Given his role as executive editor, Sivakumar Manickam had no involvement in the peer review of this article and had no access to information regarding its peer review. Full responsibility for the editorial process for this article was delegated to another journal editor. Other authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgements The authors acknowledge support from the Research Fund for International Excellent Young Scientists (Grant No. W2532017) from the Natural Science Foundation of China (NSFC), the Ningbo Commonweal Programme (Grant No. 2022S122), and the Ningbo Youth Science and Technology Programme (Grant No. 2024QL007) funded by the Ningbo Science and Technology Bureau, China. The authors also acknowledge support from the Ningbo Municipal Key Laboratory on Clean Energy Conversion Technologies (2014A22010) and the Zhejiang Provincial Key Laboratory for Carbonaceous Wastes Processing and Process Intensification Research funded by the Zhejiang Provincial Department of Science and Technology (2020E10018). References