Tailoring plasmonic coupling in black gold-decorated porous polymer membranes via all-in-one synthesis for enhanced photocatalytic peroxymonosulfate activation

Janine Gerlach^a, Nourman Barakat^a and Lukas Fischer*^abc
aLehrstuhl für Technische Chemie II, University Duisburg-Essen, Universitätsstr. 7, 45141 Essen, Germany. E-mail: Lukas.Fischer@uni-due.de
^bCenter for Water and Environmental Research (ZWU), University Duisburg-Essen, Universitätsstr. 2, 45141 Essen, Germany
^cCenter for Nanointegration Duisburg-Essen (CENIDE), University Duisburg-Essen, Carl-Benz-Str. 199, 47057 Duisburg, Germany

First published on 28th July 2025

---

---

1 Introduction

Global water resources are increasingly threatened by pollution from anthropogenic sources, posing a pressing environmental challenge. For instance, the widespread contamination of aquatic environments with antibiotics promotes the proliferation of antibiotic-resistant bacteria.^1–4 Conventional wastewater treatment plants (WWTPs) are currently ineffective at removing such micropollutants, resulting in their accumulation in natural water bodies.^5,6 Advanced oxidation processes (AOPs) offer a promising solution by generating highly reactive active species from oxidants, which are capable of degrading a wide range of organic contaminants.⁷ However, commercial adoption of AOPs for water treatment has been hampered by long treatment times and high operational costs.⁸

The treatment capacity of AOPs can be significantly enhanced by employing heterogeneous metal catalysts to activate oxidants, such as peroxymonosulfate (HSO₅⁻) or hydrogen peroxide (H₂O₂).^9,10 However, state-of-the-art metal catalysts face significant limitations. First, many transition metals exhibit poor stability under the oxidative conditions of AOPs, leading to metal ion leaching during operation.^11–14 This leaching diminishes catalytic efficiency, compromises long-term stability, and introduces secondary metal pollution. Noble metal catalysts (e.g., gold) offer superior stability but present different challenges. Their high redox potentials typically hinder the M(X)/M(X + 1) redox cycles common in transition metals, often leading to lower intrinsic activity for activating oxidants.¹⁵ Consequently, noble metals favour non-radical pathways or direct electron transfer mechanisms rather than radical generation.¹⁶ Furthermore, while photocatalysis using semiconductor materials (like transition metal oxides) is established, leveraging the localized surface plasmon resonance (LSPR) of noble metals like gold for “standalone” photocatalysis (e.g., without heterojunction to a semiconductor) in water treatment is mostly unexplored.^17,18

A second challenge stems from the prevalent use of catalysts in powder form, which complicates recovery and is poorly suited for continuous flow-through water treatment systems.^19–21 Immobilizing catalysts within porous polymer membranes offers a promising solution for achieving continuous processes.^22,23 These membranes are commonly fabricated via liquid non-solvent induced phase separation (NIPS).²⁴ In NIPS, a polymer dissolved in a suitable solvent (the casting solution) is cast as a thin film and immersed in a non-solvent bath (typically water).²⁵ Solvent/non-solvent exchange triggers polymer precipitation, forming a porous membrane structure.²⁶ Conventional methods for incorporating catalysts include blending pre-synthesized metal nanoparticle powders into the casting solution or coating metals onto pre-formed membranes.^27–30 The pre-synthesis of nanoparticles for blending approaches often leads to significant material loss during powder isolation and is plagued by particle agglomeration. This is especially true for noble metals like gold, which are prone to irreversible sintering.³¹ As a result, blending often yields membranes with poor catalyst accessibility.^32,33 Coating methods, which involve immersing the membrane in metal precursor solutions for in situ reduction/solidification, typically suffer from even lower material efficiency.^34–37 Only a small fraction of the metal precursor usually deposits as solid particles onto the membrane, and these particles can readily detach as they are bound solely by adsorption. This inefficient material usage is especially detrimental when employing scarce and expensive noble metals.

To address these challenges, we report a material-efficient, all-in-one synthesis of porous polyethersulfone (PES) membranes decorated with gold nanoparticles (AuNPs). Adapted from methods we recently developed,^38–40 this process integrates the in situ sonochemical reduction of KAuCl₄ to form AuNPs directly within the PES casting solution, followed immediately by membrane fabrication via NIPS. Crucially, we demonstrate a strategy to control the size and spatial arrangement of the immobilized AuNPs, thereby tuning their collective plasmonic properties for enhanced light absorption. We show that the resulting membranes, featuring tailored plasmonic AuNPs, efficiently activate peroxymonosulfate (PMS) in the visible-light-enhanced photocatalytic degradation of ofloxacin (OFL) in water. Furthermore, we elucidate the underlying catalytic mechanism and demonstrate the high performance and stability of these membranes in continuous flow-through water treatment.

2 Materials and methods

2.1 Materials

All materials are listed in Section S1 of the SI.

2.2 All-in-one synthesis of gold-decorated membranes

150 mg PES, 20 mg KAuCl₄, 0.7 or 0 mg CuCl₂ and 10 or 0 mg polyvinylpyrrolidone (PVP) were dissolved in 850 mg N-methyl-2-pyrrolidone (NMP)/dimethyl sulfoxide (DMSO) (70/30). Afterward, a certain mol eq. NaBH₄ (relative to metal) and 50 mg ethanol were added to this casting solution under high-shear mixing (neoLab 72020 vortex mixer at 3150 rpm). Then, the in situ metal reduction was conducted in the casting solution for 1 h under constant ultrasonication in a sonication bath at 30 °C, 45 kHz and 100 W (Transonic TI-H-10 from Elma). Finally, the casting solution was degassed for 15 min in a vacuum oven at 50 °C and 50 mbar.

Directly afterwards, the casting solution was used for film casting on a glass plate with a film thickness of 200 μm and at a casting speed of 10 mm s⁻¹ (at room-temperature of ∼22 °C and at ∼30% humidity). After casting, the film was immersed in a deionized (DI) water precipitation bath at room-temperature.

After fabrication, the final gold-decorated membranes were stored in fresh DI water.

2.3 Material characterization

Two 16 mm samples of each membrane were weighted, and their thickness was determined, both in water-wet and dry state. From this data, their porosity (water-accessible pore volume) was calculated.

For scanning electron microscopy (SEM) and energy-dispersive X-ray spectroscopy (EDX), membrane samples were dried and then frozen in liquid nitrogen, followed by braking to access the cross-section and sputter coating with Pt. SEM and EDX were conducted on an Apreo S LoVac system from Thermo Fisher Scientific at a pressure of 10⁻³ Pa, an accelerating voltage of 5 kV (SEM) or 20 kV (EDX), and a current of 13 Pa.

For dynamic light scattering (DLS) analysis, a 14 mm coupon of each gold-decorated membrane was dried and then dissolved in 5 mL NMP (shaking for 20 min). The dispersion was analysed (90 s measurement time) using a NANO-flex® 180° DLS system from Particle Metrix GmbH. Afterward, the UV-Vis extinction of that same dispersion was measured between 400–800 nm using a Cary UV-Vis spectrophotometer from Varian.

The relative diffuse reflectance (compared to barium sulfate) of the membranes was determined in the wavelength range from 400 to 800 nm via the integrating sphere method using a UV-2600i system from Shimadzu.

2.4 Catalytic OFL oxidation via PMS activation

All batch reactions were conducted with 16 mm membrane coupons immersed in 10 mL reaction solution. The reactions were carried out under constant shaking and at a temperature of 30 °C using a heated shaker plate (Titramax 1000 from Heidolph). The reactions were stopped after 2 h by removing the membrane coupons from the solution, directly followed by analysis of the reaction solution using a Cary UV-Vis spectrophotometer from Varian (see SI, Section S2). Each batch reaction was conducted twice with two different samples of each membrane.

All reactions in flow-through mode were conducted by placing membrane coupons into a transparent dead-end membrane module with a 3 cm² active area (reusable polycarbonate filter holder from Sartorius). The membrane module was connected to a peristaltic pump, which supplied the feed solution at a constant flow rate. The membrane module was placed inside a closed chamber (see SI, Section S3), approximately 10 cm below the visible-light source (OSRAM LED PAR16 4.2 W spotlight). To start the experiment, the feed was connected to the peristaltic pump, and, for photocatalytic experiments, the visible-light source was turned on. Immediately after collecting each permeate sample, it was analysed using a Cary UV-Vis spectrophotometer from Varian (see SI, Section S2).

3 Results and discussion

3.1 All-in-one: particle synthesis and membrane preparation

We recently introduced an all-in-one approach for preparing membranes with immobilized transition metal nanoparticles.^38–40 In this work, we adapted and extended this method to synthesize porous polyethersulfone (PES) membranes decorated with AuNPs, specifically focusing on tuning nanoparticle size and inter-particle spacing to control their collective plasmonic properties for enhanced catalysis (Fig. 1).

The process started with the in situ sonochemical synthesis of zerovalent AuNPs from AuCl₄⁻ with NaBH₄ directly within the PES (15 wt%) casting solution (NMP/DMSO 70:30). Critically, ethanol served as quencher for excess NaBH₄ via ethanolysis, a reaction catalyzed by the in situ formed AuNPs (see schematic in Fig. 1). Immediately after nanoparticle formation, the casting solution containing the suspended AuNPs was cast into a thin film and immersed in a water bath to induce phase separation of the matrix polymer (PES), forming the porous membrane (see Section 2.2). This integrated approach offers two key advantages: (1) it immobilizes the nanoparticles in the membrane matrix directly after their formation, effectively ‘freezing’ their size and dispersion state achieved in solution. (2) It ensures high material efficiency, as the elimination of intermediate purification and handling steps mitigates the metal loss associated with traditional powder catalyst synthesis.

3.2 Basic membrane characterization

This all-in-one method was employed to prepare membranes decorated with gold-only (Au_10Red) and copper-doped (3 wt% of gold) AuNPs (AuCu_10Red), both synthesized using 10 mol equivalents (eq.) of NaBH₄ (relative to total metal precursor).

To investigate the effect of reductant concentration on nanoparticle properties, Cu-doped Au membranes were also prepared using 3 eq. (AuCu_3Red) and 1.5 eq. (AuCu_1.5Red) of NaBH₄. Furthermore, PVP was used as a capping agent during the in situ synthesis with 3 eq. NaBH₄ (AuCuPVP_3Red). Notably, both the Cu-doping and PVP capping of AuNPs was realized by simply adding Cu²⁺ and/or PVP with AuCl₄⁻ to the casting solution. Analysis of the casting solutions also revealed that >99% of the metal precursors were reduced by NaBH₄ after 1 h reaction under all conditions (see SI, Section S4), confirming the effectiveness of the in situ particle synthesis.

Although cast with a doctor blade set to 200 μm, the final dry membrane thicknesses ranged from approximately 90 to 110 μm (Table 1). This thinning compared to the initial wet thickness is typical for the employed NIPS process, resulting from solvent outflow and film shrinkage during phase separation. Nevertheless, a consistently high porosity (around 80%) was observed for all membranes. Furthermore, the membranes displayed similar Au contents (6.0–6.4 wt%, determined via quantitative EDX) and loadings (140–170 μg cm⁻²). This corresponds to an exceptionally high AuNP yield of 93–99% (amount Au immobilized in the final membranes compared to amount added as AuCl₄⁻ to the casting solutions). While copper could not be reliably quantified by EDX due to its low content, the theoretical Cu loading is approximately 5 μg cm⁻² for all membranes, which we verified exemplarily for the AuCuPVP_3Red membrane via copper leaching (see SI, Section S9).

---

3.3 Characterization of gold-decorated membranes

SEM and EDX were used to examine the membrane morphology and the spatial distribution of the immobilized AuNPs (Fig. 2). The membranes exhibit an anisotropic cross-section characteristic of NIPS,^25,26 featuring small finger-like pores near the top surface (exposed first to the water precipitation bath, cf. Section 2.2) that transition into larger macrovoids towards the bottom (Fig. 2, middle panels). This structure arises from the kinetics of solvent/non-solvent exchange in the cast polymer solution film after immersion in the water bath, which results in gradual polymer precipitation from top to bottom.

High-resolution SEM images reveal that the AuNPs are primarily located on the internal pore surfaces of the membranes (Fig. 2, left panels), maximizing their accessibility for catalysis (high-resolution EDX mappings are shown in Section S5 of the SI). This preferential location likely occurs during phase separation: as the hydrophobic polymer-rich phase solidifies, the polymer-suspended hydrophilic AuNPs are expelled towards the forming interface with the water-rich (pore-forming) phase, becoming anchored at the pore surface upon final polymer precipitation.

Notably, higher NaBH₄ concentrations used during synthesis resulted in poorer nanoparticle dispersion within the membrane, characterized by increased agglomeration. This suggests that a higher ionic strength of the casting solution, resulting from more NaBH₄, promotes interaction between in situ formed AuNPs.

Membranes prepared with 10 eq. NaBH₄ (Au_10Red, AuCu_10Red) showed large particle networks (>500 nm in diameter) composed of numerous primary AuNPs (Fig. 2, left panels). Reducing the NaBH₄ amount to 3 eq. (AuCu_3Red) significantly decreased the extent of agglomeration, while using only 1.5 eq. (AuCu_1.5Red) resulted in immobilization of individually dispersed nanoparticles. Crucially, the addition of PVP as capping agent during in situ synthesis with 3 eq. NaBH₄ (AuCuPVP_3Red) also effectively reduced large-scale agglomeration compared to its PVP-free counterpart (AuCu_3Red), although some smaller clusters were still formed. As a result, the AuCuPVP_3Red membrane features a distribution of AuNPs on its pore surfaces, ranging from individual nanoparticles to small clusters of varying sizes (containing up to ∼20 primary particles).

EDX elemental mapping confirmed the presence and co-location of both Au and Cu in the membranes with Cu-doped AuNPs and further demonstrated a relatively homogeneous distribution of nanoparticles throughout the membrane cross-sections (Fig. 2, right panels). Consistent with the SEM images, the EDX maps for membranes prepared with 10 eq. NaBH₄ (Au_10Red and AuCu_10Red) clearly showed the presence of large, macroscopic gold agglomerates.

3.4 Particle size and optical properties

The size characteristics of the AuNPs formed in situ and the optical properties of both the extracted nanoparticles and the final membranes were investigated (Fig. 3).

To assess the size of the primary nanoparticles immobilized within the membranes, pieces of the gold-decorated membranes were dissolved in NMP, and the resulting dispersions were analysed by DLS (Fig. 3a). This analysis revealed that the apparent solvated diameter (d_Max) of the extracted Cu-doped AuNPs decreased significantly from 51 to 26 and 5 nm as the NaBH₄ concentration used for in situ synthesis was lowered from 10 eq. to 3 eq. and 1.5 eq., respectively. This inverse relationship between reductant concentration and final particle size deviates from the classical LaMer model and suggests an autocatalytic growth mechanism.^41,42 In this scenario, initially formed Au(0) nuclei act as catalytic sites for the reduction and subsequent deposition of further AuCl₄⁻ from solution. Consequently, higher NaBH₄ concentrations lead to faster growth rates and larger final nanoparticles. The trend of decreasing particle size (from 51 to 26 to 5 nm) also correlates with a slight decrease in AuNP yield for the corresponding membranes: from 99% for AuCu_10Red to 95% for AuCu_3Red and 93% for AuCu_1.5Red (cf. Table 1). Since we observed full conversion of all metal precursors in the casting solutions (see SI, Section S4), this gold loss suggests that some smaller particles may diffuse into the water phase before polymer precipitation during membrane fabrication.

Interestingly, although SEM revealed large particle networks (>500 nm) in the intact AuCu_10Red membrane (cf. Fig. 2), DLS analysis showed that most of the Cu-doped AuNPs extracted from it exhibit much smaller diameter (51 nm), with only a minor particle population in the 300–700 nm range (Fig. 3a). This indicates that the large-scale agglomeration of the Cu-doped AuNPs observed inside the membranes is physical (e.g., van der Waals forces) and reversible. In stark contrast, the non-doped AuNPs from the Au_10Red membrane remained irreversibly agglomerated after extraction (d_Max = 605 nm). This could be attributed to the sintering of pure AuNPs. Sintering, the irreversible fusion of AuNPs, requires an identical atomic arrangement at the point of contact. Integrating copper atoms may introduces surface heterogeneities that hinder the fusion process.⁴³

Adding PVP during the synthesis with 3 eq. NaBH₄ (AuCuPVP_3Red) also effectively limited the nanoparticle growth rate, resulting in extracted nanoparticles with an apparent diameter of 12 nm, significantly smaller than those from AuCu_3Red (26 nm). Additionally, PVP-capping promoted a more homogeneous particle growth, yielding a narrow size distribution (10–20 nm). In contrast, for the AuCu_3Red membrane we observed a broad distribution with tailing towards 100 nm.

UV-Vis extinction spectra of the nanoparticles extracted and dispersed in NMP correlate well with the DLS and SEM observations (Fig. 3b). Nanoparticles from membranes synthesized with higher NaBH₄ amounts or without PVP (Au_10Red, AuCu_10Red, AuCu_3Red) exhibited low intensity broadband extinction lacking distinct LSPR peaks. This can be attributed to their wide size distribution, which extensively broadens the LSPR spectral range, and their partial agglomeration, which leads to a lower total extinction cross-section due to shadowing effects.⁴⁴ Conversely, the narrow-sized Cu-doped AuNPs extracted from the AuCu_1.5Red and AuCuPVP_3Red membranes displayed distinct LSPR peaks centered around 540 nm and 545 nm, respectively, characteristic of small gold nanoparticles in NMP.⁴⁵

Relative diffuse reflectance (cf. Section 2.3) of the intact gold-decorated membranes revealed distinct optical properties (Fig. 3c). The AuCu_1.5Red membrane, decorated with individually dispersed nanoparticles (cf. Fig. 2), showed a reflectance minimum around 525 nm, relatively close to the LSPR peak position of its extracted nanoparticles (the difference can be attributed to air as medium). Strikingly, the AuCuPVP_3Red membrane displayed very low diffuse reflectance (between 7.5% and 3.0%) across the entire visible spectrum (400–800 nm), effectively acting as a ‘black’ plasmonic absorber.⁴⁶ This strong broadband absorption, which differs from the behaviour of the extracted nanoparticles, is attributed to the collective plasmonic response of the PVP-capped Cu-doped AuNP population within the membrane's porous polymer matrix (cf. Fig. 2).⁴⁷ Specifically, the immobilized particle nanoclusters retain a high extinction cross-section, while their varying sizes and interparticle distances lead to plasmon coupling with a wide distribution of coupling strengths.⁴⁸ This coupling shifts the plasmon resonances relative to the intrinsic LSPR of the individual particles, creating a superposition of absorption peaks that result in near-continuous and intense visible-light absorption. Importantly, AuCuPVP_3Red retained its black absorption after the Cu-dopant was selectively leached from the membrane (see SI, Section S9), demonstrating that plasmonic gold is responsible for this property.

‘Black gold’ has recently attracted significant interest for catalysis in energy and environmental applications. However, previously reported synthesis procedures are often complex and multi-step (e.g., AuNP growth on supports), designed to generate the necessary plasmonic coupling in powder composites.^49,50 In contrast, the novel, all-in-one method presented here provides an efficient and straightforward synthesis of tailored black gold materials, markedly improving material economy and reducing preparation time compared to existing techniques.

3.5 Catalytic degradation of ofloxacin

The catalytic performance of the prepared membranes was evaluated for the degradation of ofloxacin (OFL) in water via peroxymonosulfate (PMS) activation in both batch and flow-through modes (Fig. 4). OFL was chosen as an exemplary contaminant due to being one of the antibiotics of emerging concern listed on the 4th Watch List of the EU's water framework directive.

We quantified OFL removal from water based on the decrease in the integrated area of its characteristic UV-Vis peak at 293 nm (π–π* transition of the aromatic quinolone core) (see SI, Section S2). Although we also evaluated HPLC analysis for quantifying OFL degradation (see SI, Fig. S7), it showed a more rapid removal of the parent compound than the aromatic content measured by UV-Vis. Therefore, UV-Vis was selected as the more relevant method for assessing overall remediation performance, as it also accounts for the degradation of aromatic oxidation intermediates.

As shown in Fig. 4, the Cu-doped membrane (AuCu_10Red) demonstrated consistently higher catalytic OFL degradation rates than the gold-only membrane (Au_10Red). This difference in catalytic activity matched their OFL adsorption rate: Au_10Red exhibited substantially lower adsorption than AuCu_10Red (3 vs. 6 μg OFL per mg metal per h in DI water; 0.5 vs. 2 μg mg⁻¹ h⁻¹ in ion matrix; Fig. 4a). Given that Cu-doping was observed to drastically reduce irreversible nanoparticle agglomeration (cf. Fig. 3a), these findings collectively suggest that copper enhances performance by improving the surface accessibility of the membrane-immobilized AuNPs, not by acting as a catalyst itself.

In the batch configuration in DI water (Fig. 4a, left), OFL removal by the membranes strongly correlated with the size of the immobilized Cu-doped AuNPs (and thus, NaBH₄ concentration used for all-in-one synthesis). Decreasing the NaBH₄ amount from 10 eq. (AuCu_10Red, 51 nm) to 3 eq. (AuCu_3Red, 26 nm) and 1.5 eq. (AuCu_1.5Red, 5 nm) led to a significant decrease in both the OFL adsorption rate (from 6 to 4 to 3 μg mg⁻¹ h⁻¹) and the catalytic degradation rate of OFL with PMS (from 36 to 29 to 20 μg mg⁻¹ h⁻¹).

This inverse relationship between OFL adsorption/degradation and geometric surface area (smaller particles yield higher area) suggests that the intrinsic surface properties of smaller AuNPs hinder binding of OFL, a trend that was similarly reported in literature for adsorption of biomolecules.⁵¹ We hypothesize that the higher surface energy of smaller particles may promote the formation of a more strongly bound hydration layer in aqueous solution, a barrier that OFL must displace to bind to the catalyst.

AuCuPVP_3Red (12 nm particles, 3 eq. NaBH₄, PVP-capped) diverged from the size-dependent activity trend and reached a catalytic OFL degradation rate of 37 μg mg⁻¹ h⁻¹. Crucially, this membrane still exhibited a similarly low OFL adsorption as AuCu_1.5Red (uncapped 5 nm particles). This suggests the PVP-capped nanoparticles have higher PMS activation activity, compensating the low OFL adsorption. We attribute this enhanced activation to the known effect of PVP of increasing the electron density of AuNPs via its electron-donating functional groups (e.g., nitrogen/oxygen lone pairs).⁵² This modification may facilitate the non-radical PMS activation pathway known for noble metals like gold.¹⁵ Specifically, donation of electron density from the gold surface to the antibonding orbital (σ*) of the O–O bond of PMS can weaken this bond, promoting its cleavage and the subsequent formation of catalyst-bound non-radical active species.

Conducting the batch degradation experiments in an ion matrix representative of wastewater (including Na⁺, K⁺, Ca²⁺, Mg²⁺, NH₄⁺, Cl⁻, CO₃²⁻, PO₄³⁻; exact composition in SI, Section S6) yielded similar activity trends among the membranes as in DI water (Fig. 4a, right), although OFL adsorption and degradation levels were lower. This indicates that OFL and PMS compete with dissolved ions for active sites on the membrane-immobilized AuNPs. The AuCuPVP_3Red membrane demonstrated the modest catalytic rate decrease in the ion matrix (only from 37 to 30 μg mg⁻¹ h⁻¹), while its decrease in OFL adsorption was similar to that of the other membranes. This aligns with our hypothesis that PVP capping promotes interaction between the AuNPs and PMS, enhancing the resilience to interference by components of the water matrix. Crucially, we measured the amount of metal ions that leached from the membranes into solution after 2 h batch reaction (see SI, Section S7). All membranes proved highly stable, exhibiting negligible gold (<0.2%) and copper loss (<0.6%).

Visible-light irradiation in continuous flow-through mode further induced photo-enhancement in all membranes (Fig. 4b, see Section S8 in the SI for full Experimental data). The three grey membranes (Au_10Red, AuCu_10Red, AuCu_3Red) showed similar, moderate enhancement (∼35%), suggesting minimal influence of Cu-doping on the photo-response of the immobilized AuNPs. Conversely, AuCu_1.5Red and AuCuPVP_3Red exhibited significantly larger enhancements of 54% (62 to 95 μg mg⁻¹ h⁻¹) and 70% (114 to 194 μg mg⁻¹ h⁻¹), respectively. The substantial photo-enhancement in AuCu_1.5Red, which has a defined LSPR peak (cf. Fig. 3c) but the lowest dark activity, strongly points towards plasmon-mediated photocatalysis as enhancement mechanism.

Furthermore, the superior photocatalytic performance of the black gold-decorated AuCuPVP_3Red membrane can likely be attributed to its ability to harvest virtually all light across the entire visible spectrum.

To rule out direct photolysis of OFL or PMS, control experiments were conducted in flow-through with a metal-free PES-PVP membrane and also without any membrane (see SI, Section S8). These controls showed no significant PMS-mediated OFL degradation under either dark or light conditions, confirming the essential catalytic function of the gold-decorated membranes. Furthermore, copper was selectively leached from an AuCuPVP_3Red sample using HCl (see SI, Section S9), which did neither reduce the catalytic OFL removal rate of this membrane nor its photo-enhancement effect. This demonstrates that copper does not significantly contribute to the catalytic performance of the AuNPs. Instead, its primary role is to control the AuNP formation during the in situ synthesis. Finally, to assess the role of photo-heating, the surface temperature of AuCuPVP_3Red was measured under visible-light irradiation (see SI, Section S10), revealing an increase of only 2 °C in static conditions. This indicates that the 70% photo-enhancement observed for this membrane is dominated by photocatalytic effects, with only a minor contribution of photo-heating.

3.6 Impact of flow-through process parameter

Next, we evaluated key parameters affecting the flow-through catalytic performance of the AuCuPVP_3Red membrane for OFL degradation with PMS under visible-light irradiation (Fig. 5). Adjusting the flux allowed variation of the residence time (τ) in the membrane (Fig. 5a, left). The flow-through degradation exhibits pseudo-first-order kinetics ([OFL] = [OFL]₀ e^−kτ) relative to OFL, as PMS was supplied in excess. Single-pass OFL removal showed strong exponential dependence on τ, rising from approximately 20% at τ = 1 s to over 95% at τ = 9 s. This yields an apparent rate constant (k) of 0.27 s⁻¹, underscoring the high photocatalytic PMS activation and OFL oxidation activity of the black gold membrane. The initial OFL concentration ([OFL]₀) also influenced relative removal (Fig. 5a, right), albeit less strongly than τ. Decreasing [OFL]₀ from 5 mg L⁻¹ to 3 mg L⁻¹ resulted in only a 6% increase in single-pass removal degree. While the ratio of catalyst sites to OFL molecules increases at lower [OFL]₀ (favouring complete reaction), the absolute conversion rate simultaneously decreases due to the pseudo-first-order kinetics, leading to only slightly higher relative OFL removal efficiency. However, the impact of catalyst site availability becomes dominant at very low concentrations. At 0.75 mg L⁻¹ OFL, near-complete removal (∼99%) was achieved in single-pass, indicating that catalyst sites in the membrane were sufficiently abundant relative to OFL molecules under these conditions.

In the presence of the wastewater ion mix, competitive adsorption between PMS and dissolved anions onto the AuNPs became limiting for PMS activation and OFL removal (Fig. 5b). Consequently, increasing the PMS concentration from 0.5 mM (yielding ∼5% OFL removal) sharply increased performance by promoting PMS adsorption. Saturation began around 2.0–2.5 mM PMS, indicating that at higher PMS levels, the intrinsic photocatalytic turnover rate of the gold catalyst, rather than PMS binding, limits OFL oxidation. Under these conditions (ion mix, 2.5 mM PMS), single-pass removal was ∼55% for [OFL]₀ = 5 mg L⁻¹. However, targeting a more environmentally relevant concentration, the membrane demonstrated exceptional efficiency: ∼99% OFL removal was achieved at [OFL]₀ = 0.75 mg L⁻¹ with the complex ion water matrix and a residence time of only 3.5 s in the AuCuPVP_3Red membrane.

3.7 Photocatalytic mechanism of PMS activation

The (photo)catalytic PMS activation pathway was investigated in flow-through experiments with the AuCuPVP_3Red membrane using selective active species probes (Fig. 6): benzoic acid (BA), 4-hydroxybenzoic acid (HBA), and nitro blue tetrazolium (NBT). While BA and HBA both react with ˙SO₄⁻ at similar rates, HBA reacts approximately twice as fast with ˙OH compared to BA.^53,54 NBT was used to probe for reactive oxygen species (ROS), specifically ˙O₂⁻.⁵⁵

In all experiments, prior to evaluating single-pass removal efficiencies, each AuCuPVP_3Red sample was first equilibrated with an OFL solution for 120 min under flow-through conditions to saturate their adsorption capacity (see Section S11 in the SI for full Experimental data). Under visible-light irradiation but in the absence of PMS, virtually no removal of OFL and the three probe molecules was observed (Fig. 6a). This confirms that the membrane-immobilized black plasmonic AuNPs cannot photocatalytically generate active species directly from water or dissolved oxygen (e.g., O₂ + e⁻ = ˙O₂⁻ and OH⁻ + h⁺ = ˙OH does not occur).

Under dark conditions with PMS present, flow-through OFL removal significantly exceeded that of the probe molecules in both DI water and the wastewater ion mix (Fig. 6b and c). As for OFL, the removal efficiencies of the probe molecules reflect their extent of aromatic oxidation, determined by the decrease in their characteristic aromatic absorption peaks (see SI, Section S2). While BA, HBA and NBT react readily with radicals (˙OH, ˙SO₄⁻, ˙O₂⁻), their electrochemical redox potentials are relatively high due to their electron-deficient aromatic systems, implying that they are less susceptible to non-radical oxidation.⁵⁶ Conversely, OFL features electron-donating nitrogen and oxygen heteroatoms (cf. Fig. 4), creating an electron-rich aromatic system that can be oxidized more easily.

This suggests that the AuNPs catalyse a non-radical PMS activation pathway under dark conditions. This non-radical mechanism is further supported by the relation between BA and HBA degradation. In DI water, their removal efficiencies were similar (Fig. 6b), whereas in the wastewater ion mix, BA was degraded more efficiently (Fig. 6c). This pattern contradicts their theoretical radical reactivity, as ˙SO₄⁻ should degrade both similarly and ˙OH should preferentially degrade HBA over BA.^53,54

Therefore, we propose a surface-mediated mechanism initiated by the adsorption of PMS onto gold (eqn (1)). Electron density is shifted from the catalyst to the σ* orbital of the O–O group of PMS, which promotes its non-radical cleavage and yields an Au–oxygen intermediate (eqn (2)). Subsequently, this surface-activated oxygen is transferred to co-adsorbed organic molecules, facilitating the oxidation of the latter (eqn (3)).

Au + HO–O–SO3− = Au–OH–O–SO3−	(1)

Au–OH–O–SO3− = Au–O* + SO42− + H+	(2)

Au–O* + organics = Au + O-organics	(3)

Visible-light irradiation of the AuCuPVP_3Red membrane in presence of PMS resulted in photo-enhancement of the PMS-mediated oxidation of OFL and NBT, without affecting the removal of BA or HBA (Fig. 6b and c). This probe molecule selectivity implies that the immobilized black AuNPs generate specific active species via plasmonic photocatalysis, instead of unselectively enhancing conversion through photo-heating. Furthermore, published rate constants show that while NBT reacts with ˙O₂⁻, ˙SO₄⁻ and ˙OH, both BA and HBA react readily only with ˙SO₄⁻/˙OH.^53–55 Therefore, we can deduce that ˙O₂⁻ was generated as an active species under visible-light irradiation. Additionally, the fact that photocatalytic activity was only observed in the presence of PMS (cf. Fig. 6a) suggests that ˙O₂⁻ is generated via a PMS-mediated pathway.

Based on these findings, we propose that the Au–oxygen intermediates generated from PMS (via eqn (2)) can migrate and form surface-bound dimers (eqn (4)). Subsequently, the black gold catalyst may transfer a ‘hot’ electron from photo-excited LSPR to the dimer, releasing ˙O₂⁻ (eqn (5)).

2 Au–O* = Au + Au–O*–*O	(4)

Au–O*–*O + hν = Au+ + ˙O2−	(5)

Notably, while Au⁺ generation from Au(0) under light irradiation is known,⁵⁷ it faces a significant thermodynamic barrier due to the high redox potential of Au(0). However, sulfur-containing compounds (e.g., HSO₅⁻) exhibit a strong affinity for gold and can stabilize the Au⁺ state, lowering the effective energy barrier for its formation.^58,59 This suggests that HSO₅⁻ may additionally act as photosensitizer, promoting photo-induced electron release from gold by stabilizing Au⁺. Importantly, we further measured whether metal leached from the AuCuPVP_3Red membrane during the photocatalytic flow-through process (via atomic absorption spectroscopy, cf. Section S7 in the SI). The combined PMS/illumination treatment in both DI water and wastewater ion mix caused a negligible initial leaching (first 30 min permeate with 15 mL: ≤0.02 mg L⁻¹ Cu, ≤2% loss; ≤0.1 mg L⁻¹ Au, ≤0.3% loss), with levels dropping below the detection limit (<0.005 mg L⁻¹) in permeates collected after 30 min flow-through. These results show that the proposed transient formation of Au⁺ does not cause any significant gold loss, possibly due to rapid regeneration of the Au(0) state via substrate oxidation (e.g., OFL/NBT).

3.8 Continuous catalytic water treatment

Finally, we evaluated the continuous water treatment performance of the AuCuPVP_3Red membrane in the photocatalytic and PMS-mediated OFL degradation (Fig. 7). For this, water contaminated with 0.75 mg L⁻¹ OFL (feed) was continuously pumped through the membrane and the OFL removal from the permeate was monitored after single-pass; the permeate thus represents the final treated effluent that would be directly released during practical application of this process.

AuCuPVP_3Red maintained virtually complete OFL removal (>99%) over 12 h of continuous operation, processing a total volume of 1200 L m⁻² (Fig. 7, left). This was achieved despite a short residence time within the membrane of approximately 3.5 s.

Analysis of the permeate's UV-Vis spectra provided additional insight into the degradation process (Fig. 7, right). While the characteristic OFL peak at 290 nm (π–π* transition, conjugated aromatic system) disappeared completely, absorbance at shorter wavelengths (<270 nm) increased slightly. Furthermore, the weaker absorption shoulder around 330 nm, associated with n–π* transitions involving the enone moiety, decreased less significantly. This suggests that while the parent OFL was effectively destroyed, degradation byproducts that still contain the enone moiety are less readily oxidized under these conditions and persist in the treated effluent.

Nevertheless, achieving sustained, near-complete oxidation of the fluoroquinolone core, which is responsible for OFL's antibiotic activity, highlights the exceptional catalytic performance of the AuCuPVP_3Red membrane. The potential of this membrane is further underscored by comparing its performance to other reported catalytic membranes for micropollutant degradation (see SI, Section S12).^36,60–64 The maximum calculated turnover frequency (TOF) for OFL degradation by our black gold-decorated AuCuPVP_3Red membrane reached approximately 105 × 10⁻³ h⁻¹ (mol OFL per mol catalyst per h), significantly exceeding the TOFs reported for other membranes by factors ranging from 5 to 100.

While many catalysts developed for oxidative water treatment utilize more abundant and less expensive non-noble metals a review of literature indicates that such non-noble metal-based catalytic membranes often suffer from significant metal leaching during continuous operation, with reported effluent concentrations reaching levels as high as 2 mg L⁻¹ for potentially toxic metals like cobalt (see SI, Section S12).^36,60–64 This comparison highlights a crucial advantage of the highly stable gold-decorated membranes presented here, which exhibit negligible metal leaching, ensuring effluent quality and long-term stability.

4 Conclusion

This work highlights a powerful strategy for designing advanced catalytic materials, combining a resource-efficient synthesis with tailored plasmonic properties to create robust and highly effective membranes for sustainable water purification. The presented all-in-one approach combines in situ particle synthesis within the casting solution with immediate particle immobilization during membrane formation via film casting and polymer precipitation in water. First, this integrated method achieved exceptionally high immobilized AuNP yields, overcoming the significant material loss that often plagues traditional membrane functionalization approaches like blending or coating. Secondly, the rapid immobilization of AuNPs immediately following their formation could effectively prevent agglomeration, which is a common challenge in conventional synthesis protocols for small nanoparticles.

We further demonstrated that this all-in-one method allows for precise control over the effective size (5 to 51 nm) and spatial arrangement of the immobilized AuNPs by tuning reaction conditions (copper doping, NaBH₄ concentration, PVP capping). This control enabled the tailoring of plasmonic coupling, culminating in the AuCuPVP_3Red membrane with its broadband ‘black plasmonic’ absorption properties. This black gold membrane with 12 nm PVP-capped AuNPs displayed the best catalytic performance, likely due to a combination of PVP complexation promoting PMS activation and enhanced photocatalytic activity arising from its effective light harvesting. Mechanistic investigations using probe molecules suggested a dual activation pathway involving non-radical surface-mediated PMS activation dominant in the dark, supplemented by plasmon-enhanced generation of superoxide radicals (˙O₂⁻) under illumination. Crucially, the AuCuPVP_3Red membrane demonstrated outstanding potential for practical application, achieving >99% single-pass ofloxacin removal in flow-through mode, even in a complex ionic water matrix. Our systems performance surpasses that of other state-of-the-art catalytic membranes from literature, both in catalytic turnover frequency and operational stability.

Author contributions

Janine Gerlach: methodology, investigation, formal analysis, writing – review & editing. Nourman Barakat: methodology, investigation, writing – review & editing. Lukas Fischer: project administration, supervision, conceptualization, methodology, formal analysis, writing – original Draft, visualization, funding acquisition.

Conflicts of interest

There are no conflicts to declare.

Data availability

Materials; quantification of organic solute removal; metal precursor conversion; high-resolution SEM and EDX analysis; composition of ionic water matrix; metal leaching quantification; experimental data of all flow-through experiments; comparison with data from literature. The data supporting this article have been included as part of the SI.

Supplementary information is available. See DOI: https://doi.org/10.1039/d5ta03185k.

Acknowledgements

Dr Lukas Fischer is grateful for the funding by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation – grant 530864107), with which this research was financed. The authors further thank Tobias Bochmann and Robin Meya, both at UDE, for conducting the SEM and AAS measurements, respectively.

References

1. S. Thanigaivel, S. Vickram, N. Dey, P. Jeyanthi, R. Subbaiya, W. Kim, M. Govarthanan and N. Karmegam, Chemosphere, 2023, 313, 137475 Search PubMed .
2. D.-P. Häder, A. T. Banaszak, V. E. Villafañe, M. A. Narvarte, R. A. González and E. W. Helbling, Sci. Total Environ., 2020, 713, 136586 Search PubMed .
3. K. C. Jones, Environ. Sci. Technol., 2021, 55, 9400–9412 Search PubMed .
4. D. G. J. Larsson and C.-F. Flach, Nat. Rev. Microbiol., 2022, 20, 257–269 Search PubMed .
5. A. Saravanan, P. Senthil Kumar, S. Jeevanantham, S. Karishma, B. Tajsabreen, P. R. Yaashikaa and B. Reshma, Chemosphere, 2021, 280, 130595 Search PubMed .
6. J. Aravind kumar, T. Krithiga, S. Sathish, A. A. Renita, D. Prabu, S. Lokesh, R. Geetha, S. K. R. Namasivayam and M. Sillanpaa, Sci. Total Environ., 2022, 831, 154808 Search PubMed .
7. D. Ma, H. Yi, C. Lai, X. Liu, X. Huo, Z. An, L. Li, Y. Fu, B. Li, M. Zhang, L. Qin, S. Liu and L. Yang, Chemosphere, 2021, 275, 130104 Search PubMed .
8. I. A. Ike, K. G. Linden, J. D. Orbell and M. Duke, Chem. Eng. J., 2018, 338, 651–669 Search PubMed .
9. S. Wu, L. Shen, Y. Lin, K. Yin and C. Yang, Chem. Eng. J., 2021, 414, 128872 Search PubMed .
10. J. Han, L. Sun, M. Ulbricht, L. Fischer, G. Zhang, W. Gao, L. Lv, T. Ren, X. Liu and Z. Ren, Chem. Eng. J., 2025, 505, 159084 Search PubMed .
11. J. Leichtweis, E. Carissimi, U. Hagemann, M. Ulbricht and L. Fischer, Chem. Eng. J., 2023, 477, 147203 Search PubMed .
12. J. Leichtweis, Y. Vieira, E. Carissimi, M. Ulbricht and L. Fischer, Appl. Surf. Sci., 2024, 669, 160603 Search PubMed .
13. Y. Lei, C.-S. Chen, Y.-J. Tu, Y.-H. Huang and H. Zhang, Environ. Sci. Technol., 2015, 49, 6838–6845 Search PubMed .
14. T. Yu, H. Chen, T. Hu, J. Feng, W. Xing, L. Tang and W. Tang, Appl. Catal., B, 2024, 342, 123401 Search PubMed .
15. Y.-Y. Ahn, E.-T. Yun, J.-W. Seo, C. Lee, S. H. Kim, J.-H. Kim and J. Lee, Environ. Sci. Technol., 2016, 50, 10187–10197 Search PubMed .
16. G. Young Kim, D. Lee, H.-S. Choe, J.-M. Park, S. Jeong, E. Jongwoo Park, J. Won Lee, C. Lee and J.-H. Kim, Environ. Sci.: Nano, 2022, 9, 2510–2520 Search PubMed .
17. M. E. King, C. Wang, M. V. F. Guzman and M. B. Ross, Chem Catal., 2022, 2, 1880–1892 Search PubMed .
18. V. Guerrero-Florez, S. C. Mendez-Sanchez, O. A. Patrón-Soberano, V. Rodríguez-González, D. Blach and F. M. O, J. Mater. Chem. B, 2020, 8, 2862–2875 Search PubMed .
19. B. C. Hodges, E. L. Cates and J.-H. Kim, Nat. Nanotechnol., 2018, 13, 642–650 Search PubMed .
20. G. Maniakova, K. Kowalska, S. Murgolo, G. Mascolo, G. Libralato, G. Lofrano, O. Sacco, M. Guida and L. Rizzo, Sep. Purif. Technol., 2020, 236, 116249 Search PubMed .
21. X. Zheng, X. Niu, D. Zhang, M. Lv, X. Ye, J. Ma, Z. Lin and M. Fu, Chem. Eng. J., 2022, 429, 132323 Search PubMed .
22. N. Li, X. Lu, M. He, X. Duan, B. Yan, G. Chen and S. Wang, J. Hazard. Mater., 2021, 414, 125478 Search PubMed .
23. J. Liang, K. Gao, A. Zhou, Y. Fang, S. Su, L. Fu, M. Sun and X. Duan, Appl. Catal., B, 2023, 327, 122440 Search PubMed .
24. D.-M. Wang and J.-Y. Lai, Curr. Opin. Chem. Eng., 2013, 2, 229–237 Search PubMed .
25. H. Susanto, N. Stahra and M. Ulbricht, J. Membr. Sci., 2009, 342, 153–164 Search PubMed .
26. S. Mazinani, S. Darvishmanesh, A. Ehsanzadeh and B. Van der Bruggen, J. Membr. Sci., 2017, 526, 301–314 Search PubMed .
27. L. Y. Ng, A. W. Mohammad, C. P. Leo and N. Hilal, Desalination, 2013, 308, 15–33 Search PubMed .
28. S. S. Ozdemir, M. G. Buonomenna and E. Drioli, Appl. Catal., A, 2006, 307, 167–183 Search PubMed .
29. Rajlaxmi, N. Gupta, R. P. Behere, R. K. Layek and B. K. Kuila, Mater. Today Chem., 2021, 22, 100600 Search PubMed .
30. J. Yin and B. Deng, J. Membr. Sci., 2015, 479, 256–275 Search PubMed .
31. D. Radziuk, D. Grigoriev, W. Zhang, D. Su, H. Möhwald and D. Shchukin, J. Phys. Chem. C, 2010, 114, 1835–1843 Search PubMed .
32. L. Fischer, A. Volz, U. Hagemann and M. Ulbricht, Chem. Eng. J., 2023, 463, 142437 Search PubMed .
33. L. Fischer, S. A. H. Hesaraki, A. Volz and M. Ulbricht, ACS Appl. Nano Mater., 2023, 6, 14910–14920 Search PubMed .
34. W. Qing, F. Liu, H. Yao, S. Sun, C. Chen and W. Zhang, Adv. Colloid Interface Sci., 2020, 282, 102207 Search PubMed .
35. Z. He, S. Mahmud, Y. Yang, L. Zhu, Y. Zhao, Q. Zeng, Z. Xiong and S. Zhao, Sep. Purif. Technol., 2020, 250, 117266 Search PubMed .
36. Y. Bao, W. J. Lee, T.-T. Lim, R. Wang and X. Hu, Appl. Catal., B, 2019, 254, 37–46 Search PubMed .
37. I. Sanjuán, V. Kumbhar, O. Prymak, M. Ulbricht, C. Andronescu and L. Fischer, ChemSusChem, 2025, 18, e202401228 Search PubMed .
38. S. A. H. Hesaraki, O. Prymak, M. Heidelmann, M. Ulbricht and L. Fischer, ACS Appl. Mater. Interfaces, 2024, 16, 17517–17530 Search PubMed .
39. S. A. H. Hesaraki, M. Ulbricht and L. Fischer, Chem. Eng. J., 2024, 498, 155266 Search PubMed .
40. M. Ropertz, M. Ulbricht and L. Fischer, Chem. Eng. J. Adv., 2024, 20, 100683 Search PubMed .
41. C. B. Whitehead, M. A. Watzky and R. G. Finke, J. Phys. Chem. C, 2020, 124, 24543–24554 Search PubMed .
42. P. N. Njoki, J. Luo, M. M. Kamundi, S. Lim and C.-J. Zhong, Langmuir, 2010, 26, 13622–13629 Search PubMed .
43. E. S. Gnanakumar, J. M. Naik, M. Manikandan, T. Raja and C. S. Gopinath, ChemCatChem, 2014, 6, 3116–3124 Search PubMed .
44. J. M. Zook, V. Rastogi, R. I. MacCuspie, A. M. Keene and J. Fagan, ACS Nano, 2011, 5, 8070–8079 Search PubMed .
45. X. Liu, M. Atwater, J. Wang and Q. Huo, Colloids Surf., B, 2007, 58, 3–7 Search PubMed .
46. Y. Liu, Z. Liu, Q. Huang, X. Liang, X. Zhou, H. Fu, Q. Wu, J. Zhang and W. Xie, J. Mater. Chem. A, 2019, 7, 2581–2588 Search PubMed .
47. K. Bae, G. Kang, S. K. Cho, W. Park, K. Kim and W. J. Padilla, Nat. Commun., 2015, 6, 10103 Search PubMed .
48. E. Ringe, M. R. Langille, K. Sohn, J. Zhang, J. Huang, C. A. Mirkin, R. P. Van Duyne and L. D. Marks, J. Phys. Chem. Lett., 2012, 3, 1479–1483 Search PubMed .
49. D. Liu, F. Zhou, C. Li, T. Zhang, H. Zhang, W. Cai and Y. Li, Angew. Chem., Int. Ed., 2015, 54, 9596–9600 Search PubMed .
50. R. Verma, S. Kundu and V. Polshettiwar, J. Mater. Chem. A, 2024, 12, 27235–27245 Search PubMed .
51. K. Kaur and J. A. Forrest, Langmuir, 2012, 28, 2736–2744 Search PubMed .
52. H. Tsunoyama, N. Ichikuni, H. Sakurai and T. Tsukuda, J. Am. Chem. Soc., 2009, 131, 7086–7093 Search PubMed .
53. G. V. Buxton, C. L. Greenstock, W. P. Helman and A. B. Ross, J. Phys. Chem. Ref. Data, 1988, 17, 513–886 Search PubMed .
54. P. Neta, R. E. Huie and A. B. Ross, J. Phys. Chem. Ref. Data, 1988, 17, 1027–1284 Search PubMed .
55. Y. Wen, J. Yan, B. Yang, Z. Zhuang and Y. Yu, J. Mater. Chem. A, 2022, 10, 19184–19210 Search PubMed .
56. P. Luo, E. C. Feinberg, G. Guirado, S. Farid and J. P. Dinnocenzo, J. Org. Chem., 2014, 79, 9297–9304 Search PubMed .
57. Z. Mao, R. Espinoza, A. Garcia, A. Enwright, H. Vang and S. C. Nguyen, ACS Nano, 2020, 14, 7038–7045 Search PubMed .
58. S. Witzel, A. S. K. Hashmi and J. Xie, Chem. Rev., 2021, 121, 8868–8925 Search PubMed .
59. U. Koelle and A. Laguna, Inorg. Chim. Acta, 1999, 290, 44–50 Search PubMed .
60. S. Wang, J. Tian, Z. Wang, Q. Wang, J. Jia, X. Hao, S. Gao and F. Cui, Chem. Eng. J., 2020, 396, 125289 Search PubMed .
61. A. Liangdy, W. J. Lee, P. Tonanon, R. D. Webster, S. A. Snyder and T.-T. Lim, Chem. Eng. J., 2023, 454, 140075 Search PubMed .
62. W. J. Lee, Y. Bao, A. Liangdy, X. Hu and T.-T. Lim, Chem. Eng. J., 2022, 430, 132921 Search PubMed .
63. S. Sun, H. Yao, W. Fu, S. Xue and W. Zhang, J. Hazard. Mater., 2020, 386, 121955 Search PubMed .
64. J. Liang, K. Gao, A. Zhou, Y. Fang, S. Su, L. Fu, M. Sun and X. Duan, Appl. Catal., B, 2023, 327, 122440 Search PubMed .

This journal is © The Royal Society of Chemistry 2025