Packed bed optofluidic microreactors with Au decorated TiO2 nanoflowers for visible light photocatalytic water purification

Introduction

Water pollution is a pressing global concern, necessitating the development of efficient and sustainable water purification technologies¹. Traditional biochemical water treatment processes are often time-consuming and complex², with limited ability to remove emerging contaminants like antibiotics and aromatics, resulting in residual pollutants in the treated wastewater^3,4. As discharge standards become more stringent, there is an urgent need for advanced water purification methods that can target these residuals in already-treated effluents, where traditional methods fall short.

Photocatalytic degradation has emerged as a promising solution for this task due to its ability to efficiently degrade a wide range of contaminants and deactivate bacteria in already-treated wastewater under solar irradiation⁵. Among the available photocatalysts, titanium dioxide (TiO₂) is widely recognized for its cost-effectiveness and robust ability to generate reactive oxygen species (ROS)⁶. However, its practical application faces several challenges, including limited solar light absorption (< 5% of solar light utilization)⁷, rapid recombination of electron-hole pairs⁸, and susceptibility to photo-corrosion during prolonged use⁹. To address these issues, TiO₂ nanoflowers (TNFs), composed of self-assembled TiO₂ nanosheets, have been developed. These structures provide increased surface area, improved charge separation and optimized light harvesting^10,11,12. Further performance improvements have been achieved by incorporating plasmonic nanoparticles, such as gold nanoparticles (AuNPs), which broaden the absorption spectrum through localized surface plasmon resonance (LSPR), boosting the separation of charge carriers, reducing recombination, and enhancing the structural stability^{13,14,15,16,17,18,19}.

In addition to optimizing the photocatalyst itself, the design of photocatalytic reactors also plays a critical role in determining overall performance^{20,21,22,23,24}. Recent studies have demonstrated that TiO₂-based photocatalysts perform significantly better in microreactors compared to traditional slurry-mode reactors due to precise reaction control, improved mass transfer, uniform light distribution, and simplified separation of products from catalysts^{25,26,27,28,29}. However, most of these studies utilize film-mode microreactors, where photocatalysts are only coated on the inner surfaces of reactors. In contrast, packed-mode or packed-bed optofluidic microreactors, where photocatalysts are packed into the reactor chamber, offer several advantages over film-mode designs, including a higher surface area, shorter diffusion distances, and enhanced light harvesting through multiple scattering effects^30,31,32. These features lead to superior overall performance. While some studies on TiO₂-based photocatalysis have investigated packed-mode designs, they often focus on tubular configurations with photocatalysts coated on supporting materials^30,33. Despite their effectiveness, these tubular packed-mode reactors face limitations such as high pressure drop, limited surface area-to-volume ratio, non-uniform catalyst distribution and uneven light irradiation. Such issues ultimately restrict the reactor efficiency, especially in water purification. To address these limitations, our work introduces a novel approach by directly packing photocatalysts into a planar optofluidic microreactors. This planar packed-mode configuration offers several potential advantages, including uniform catalyst distribution, increased surface area, and consistent light exposure across the reaction surface, making it a promising solution for enhancing water purification efficiency.

In this work, we developed a novel planar packed-mode/packed-bed optofluidic microreactor (PPOM) by directly packing AuNPs-decorated TNFs (Au/TNFs) and evaluated its performance in visible light-driven water purification. Methylene blue (MB) served as the model chemical pollutant for a pilot study. We systematically compared the performance of the PPOM with other reactor configurations, including a planar film-mode optofluidic microreactor (PFOM) and a conventional slurry reactor, to assess the contributions of AuNPs-induced LSPR effects and the reactor design. Numerical simulations and action spectrum analysis were conducted to further clarify the photocatalytic mechanisms and confirm the enhancements induced by the AuNPs. This work highlights the synergistic benefits of combining optofluidic technology with plasmonic photocatalysis, advancing the development of efficient water purification systems that address both photocatalyst and reactor design optimizations. The potential of this system for a wide range of applications has also been discussed, including energy conversion, chemical synthesis, and drug delivery.

Results

Fabrication and characterization of photocatalysts

The fabrication procedures of TNFs and Au/TNFs are illustrated in Fig. 1a. To create TNFs, a simple solvothermal method was employed using a mixed solution of titanium tetraisopropanolate (TIP), diethylenetriamine (DETA), and isopropanol (IPA). The solution was then mixed with tetrachloroauric(III) acid trihydrate (HAuCl₄∙3H₂O), sodium borohydride (NaBH₄) and thioglycolic acid (TGA) to obtain Au/TNFs. The characterizations of TNFs and Au/TNFs were performed to elucidate their structural and optical properties. X-ray diffraction (XRD) patterns showed five characteristic peaks of TiO₂, assigned to (101), (004), (200), (105), and (211) planes for both TNFs and Au/TNFs (Fig. 1b). These patterns were typical for pure phase TiO₂ (JCPDS 21-1272), confirming their anatase phase without any impurity diffraction peaks³⁴. The absence of discernible XRD diffraction peaks for the AuNPs suggested their limited concentration³⁵. X-ray photoelectron spectroscopy (XPS) and inductively coupled plasma-mass spectrometry (ICP-MS) analysis were also conducted to confirm the successful incorporation of AuNPs into TNFs. The presence of Au 4f peaks in Supplementary Fig. 1a, b along with the approximated 4.5 wt% of Au content indicated by ICP-MS, confirmed the effective incorporation of AuNPs into the TNFs. The optical and electrical properties of TNFs and Au/TNFs were further investigated using ultraviolet-visible diffuse reflectance spectra (UV-DRS), which were closely related to their photocatalytic properties. Au/TNFs exhibited enhanced absorbance in visible spectral range (400–800 nm) compared to TNFs (Fig. 1c). Notably, the absorption band near 554 nm in the Au/TNFs spectra could be attributed to the LSPR mode of AuNPs. Subsequently, the band gaps of TNFs and Au/TNFs were calculated using the Kubelka-Munk method^36,37. Figure 1d illustrates the dependents of (αhν)² on photon energy (hν) for the direct transitions, revealing the direct band gaps of 3.23 eV for TNFs and 3.10 eV for Au/TNFs. This reduction in the band gap of TNFs after the decoration of AuNPs could be ascribed to a synergistic effect stemming from the increased abundance of oxygen vacancies induced by Au³⁸, coupled with the potent built-in electric field generated by the interaction between Au and TiO₂¹⁹.

Packed bed optofluidic microreactors with Au decorated TiO2 nanoflowers for visible light photocatalytic water purification — **Fig. 1: Fabrication, structural, and optical characterizations of TNFs and Au/TNFs.**

The microstructures of TNFs and Au/TNFs were characterized using transmission electron microscopy (TEM) (Fig. 2a–f) and field emission scanning electron microscopy (FE-SEM) (Fig. 2g, j). The TEM images (Fig. 2a,h) revealed the flower-like structure of TNFs, with diameters ranging from 400–500 nm. High-resolution TEM images (Fig. 2b, c) further showed that the TNFs were composed of aggregated ~10 nm thick TiO₂ nanosheets, which consist of ~6 nm diameter TiO₂ nanoparticles. After the AuNPs loading, the flower-like structured TNFs remained intact (Fig. 2d), with AuNPs uniformly distributed on TNFs (dark dots in Fig. 2e). The average diameter of AuNPs were estimated to be ~4.5 nm with the size ranging from 2 to 6.5 nm (Supplementary Fig. 2). High-resolution TEM images in Fig. 2c,f revealed characteristic lattice fringe spacings of 0.35 and 0.25 nm, corresponding to the (101) crystallographic planes of anatase TiO₂ and the (111) crystallographic planes of AuNPs, respectively. High-angle annular dark field-STEM (HAADF-STEM) images, TEM elemental mapping images and energy dispersive X-ray analysis (EDAX) spectra (Fig. 2h–2m) confirmed the existence and uniform deposition of AuNPs on the surface of TNFs.

**Fig. 2: Microscopic and elemental analysis of TNFs and Au/TNFs.**

Fabrication of photocatalyst-loaded optofluidic microreactors

After the materials characterization, the synthesized photocatalysts were integrated into the 1 × 1 cm reaction chamber positioned at the center of the optofluidic microreactor for visible light-driven water purification, as illustrated in the three dimensional (3D) diagram of the photocatalyst-loaded optofluidic microreactor in Fig. 3a. The fabrication procedures are detailed in the Methods section and Supplementary Fig. 3, with dimensions of microchannels provided in Supplementary Table 1. The cross-sectional views of the PPOM and PFOM are depicted Fig. 3b, c, respectively. In the packed-mode configuration, the photocatalysts densely occupied the entire reactor chamber. In contrast, in the film-mode configuration, the photocatalysts formed a ~2 μm thick film layer solely on the bottom surface of the glass slide. Detailed visual representations of the fabricated PPOM and PFOM are provided in Supplementary Fig. 4.

**Fig. 3: Schematics of optofluidic microreactors for visible light-driven water purification.**

Dark reaction of photocatalyst powders

To evaluate the photodegradation efficacy of TNFs and Au/TNFs in optofluidic microreactors, MB was used as the model chemical for the photodegradation experiments under visible light. Prior to light exposure, the container was kept in a dark environment to ensure complete saturation of MB molecules onto the photocatalyst particles (Supplementary Fig. 5a). The slurry-mode reactors with TNFs and Au/TNFs exhibited rapid saturation of MB adsorption, completing the dark reaction in < 15 min (Supplementary Fig. 6a). In contrast, the PFOMs coated with TNFs and Au/TNFs required ~50 min (Supplementary Fig. 6b) and the PPOMs filled with TNFs and Au/TNFs showed the saturation time > 120 min (Supplementary Fig. 6c). The notably shorter saturation time observed in slurry-mode reactors could be attributed to continuous stirring, which maximized the contact between the photocatalysts and MB molecules, facilitating faster adsorption. In contrast, photocatalysts in the PFOM and the PPOM relied on the flowing and diffusion of MB molecules for adsorption, resulting in longer saturation times. Particularly, the prolonged saturation time in the PPOM as compared to the PFOM was ascribed to the larger amount of photocatalysts in the PPOM, providing more effective adsorption sites.

MB degradation by photocatalysts in different modes of microreactors

After achieving adsorption saturation, the loaded optofluidic microreactors were exposed to visible light for MB photodegradation (Supplementary Fig. 5b). In the slurry-mode reactors, both TNFs and Au/TNFs exhibited an increasing trend in MB degradation with prolonged residence time (Fig. 4a). TNFs showed approximately 23% MB degradation after 120 min of visible light irradiation, primarily due to their visible-light-driven activity under the self-photosensitization of MB^39,40. In stark contrast, Au/TNFs achieved 90% degradation of MB within the same period of visible light irradiation, highlighting a substantial enhancement in photodegradation performance upon the incorporation of AuNPs. This significant improvement in photodegradation efficiency could be attributed to the formation of Schottky junctions between AuNPs and TNFs that significantly reduced the recombination rate of electron-hole pairs⁴¹. Furthermore, the LSPR of AuNPs, which was excited under resonant light illumination, promoted charge separation and enhanced the photocatalytic activity of TNFs in the presence of AuNPs⁴². The hot electrons generated through LSPR excitation of AuNPs were subsequently injected into the conduction band of TiO₂ to produce O₂^•− radicals, acting as highly potent oxidizing agents for MB degradation^43,44.

**Fig. 4: Measured methylene blue (MB) photodegradation in different photocatalyst-loaded reactors.**

After loading the photocatalysts into the optofluidic microreactors, the photodegradation of MB was substantially improved. In the PFOM, TNFs exhibited a remarkable 30% degradation of MB within 2 min (Fig. 4b). The enhanced photodegradation was also evident upon the addition of AuNPs to the photocatalysts in this mode. Within the same 2-min residence time, the PFOM with Au/TNFs achieved 73% degradation, which was 2.4 times that of TNFs alone. In the PPOM, the impact of AuNPs on MB photodegradation was less pronounced, with Au/TNFs reaching 95% degradation compared to 85% with TNFs. This discrepancy could be attributed to the full packing of photocatalysts having a more prominent impact on the MB degradation than the addition of AuNPs. Notably, employing the same TNFs, the PPOM achieved 2.8 times the MB degradation of the PFOM (i.e., 85% vs. 30%). Overall, the PPOM demonstrates superior performance in MB photodegradation, benefiting from its shortest diffusion length from MB to the photocatalysts and enhanced mass transfer in microreactors.

To gain a deeper understanding of photodegradation efficiency, the pseudo-first-order kinetics of MB degradation were analyzed (Fig. 4c). The PPOM with TNFs and Au/TNFs can be described by the reaction rate constants (k) of 0.64 min⁻¹ and 1.15 min⁻¹, respectively, indicating a 1.8-fold enhancement due to the plasmonic effect. Notably, an empty optofluidic microreactor showed a small reaction rate constant of 0.09 min⁻¹, demonstrating self-degradation of MB under visible light. Table 1 summarizes the reaction constants of TNFs and Au/TNFs in different reactor modes, highlighting the influences of both reactor configuration and the plasmonic effect. More specifically, in the slurry-mode, Au/TNFs achieved an enhancement factor of 46.4 relative to TNFs, affirming the plasmonic effect of AuNPs in improving photocatalytic efficiency. Similarly, the film-mode and the packed-mode showed the plasmonic enhancement factors of 3.2 and 1.8, respectively. These factors were relatively modest because TNFs already exhibited high-rate constants in both modes. Regarding the influence of photocatalyst loading mode, the enhancement factors of the packed-mode to the slurry-mode were 1,517 and 58.7 for TNFs and Au/TNFs, respectively. In contrast, the packed-mode enhancement compared to the film-mode was more modest: only 3.8 for TNFs and 2.1 for Au/TNFs, respectively. These results proved the packed-mode reactor superior efficiency owing to the enhanced mass transfer and the improved light penetration^31,32. Moreover, the packed-mode reactor with Au/TNFs benefited from the synergistic enhancement of both the plasmonic and reactor design effects, showing remarkable enhancement factors of 2,725 relative to the slurry-mode with TNFs and 6.8 relative to the film-mode with TNFs. Overall, these findings highlight the crucial contributions of both the plasmonic effect and reactor configuration in improving photocatalytic efficiency, demonstrating that the PPOM filled with Au/TNFs achieves the highest performance.

Table 1 Comparison of the photodegradation reaction rate constants of TNFs and Au/TNFs in the slurry-mode, the film-mode, and the packed-mode

Full size table

Reusability of photocatalysts in PPOM for MB degradation

Subsequent to the photodegradation experiments accessing the efficacy of TNFs and Au/TNFs in different modes, additional investigations were performed to evaluate their reusability and stability under visible light conditions, which are essential for their practical applications. Specifically, the reusability and stability of TNFs and Au/TNFs in PPOM were investigated by conducting three consecutive MB photodegradation tests under visible light (λ > 420 nm) irradiation. TNFs in PPOM achieved an initial 84% MB degradation in 2 min, with a subsequent decline to 75% after three cycles, corresponding to a performance decline of 10.7% (Fig. 4d). This decline might be attributed to photo-induced hole corrosion and electron-hole pair recombination during the prolonged light exposure^9,45. The stability of TNFs over time could also be affected by the potential loss of TNFs due to the repeated flushing. In contrast, Au/TNFs in PPOM demonstrated superior stability, degrading over 95% of MB in 2 min with only a marginal decrease to 93% after three cycles, representing only a 2.1% decrease in efficiency. This enhanced photodegradation stability could be attributed to the mitigated photo-corrosion facilitated by the addition of AuNPs, which acted as hole traps and preserved the structural integrity of TNFs⁴⁶. In addition, AuNPs facilitated efficient charge separation and transfer with the LSPR effect and Schottky junction formation, reducing the recombination of electron-hole pairs, therefore ensuring sustained photocatalytic activity over multiple cycles. Moreover, Au/TNFs degraded higher percentage of MB in a shorter time compared to TNFs, implying that Au/TNFs experienced diminished stress and degradation, resulting in improved stability over repeated cycles. Compared to other reported film-mode and slurry-mode reactors^47,48,49, the Au/TNFs-PPOM platform offers distinct advantages by providing a densely packed catalytic environment to prevent issues like photocatalyst loss, sedimentation, and surface deactivation observed in PFOM and slurry reactors. In summary, our Au/TNFs-PPOM platform exhibits superior stability after repeated use, making it a promising candidate for continuous water purification in industrial applications.

Action spectrum analysis of photocatalysts in PPOM

To elucidate the mechanism behind the enhanced photodegradation of MB facilitated by the integration of AuNPs into TNFs, an action spectrum analysis was conducted by illuminating the PPOM loaded with TNFs and Au/TNFs with light of different wavelengths (Fig. 5a). The efficiency of Au/TNFs exhibited a distinct peak around 525 nm, closely aligning with the absorption band of Au/TNFs (Fig. 1c). This suggested that the plasmonic resonance of AuNPs significantly enhanced the photocatalytic process at this wavelength. Moreover, Au/TNFs demonstrated high photocatalytic activity in the spectral range of 500–600 nm, whereas the efficiency of TNFs gradually decreased after 400 nm. This indicated that Au/TNFs can effectively utilize visible light, making them beneficial for practical applications.

**Fig. 5: Action spectrum analysis and finite-difference time-domain (FDTD) simulations.**

Finite-difference time-domain (FDTD) simulations of LSPR effect by AuNPs

To further verify that the enhanced photocatalytic degradation was a result of the plasmonic effect, we performed numerical modeling using the FDTD simulations in order to understand the far-field spectrum and the electromagnetic near-field distribution in the vicinity of TNFs and Au/TNFs (Fig. 5b–d).

To simplify the complex geometry of the porous TNF, the Bruggeman’s effective median approximation (EMA) model was employed to represent TNF as a 450 nm diameter nanosphere composite with an effective permittivity equivalent to 75% water and 25% amorphous TiO₂ (insets of Fig. 5c, d)⁵⁰. Unlike solid TiO₂ nanospheres⁵¹, the highly porous structure of TNFs resulted in both experimental (Fig. 1c) and numerical extinction (Fig. 5b) spectra of TNF showing no features of Mie resonances in the visible range, as should be expected for these nanoflower structures.

For Au/TNFs, approximately two thousand 5 nm-diameter AuNPs were randomly dispersed onto the TNF surface (inset of Fig. 5c). The simulated absorbance spectra showed a broad LSPR of Au/TNF at 550 nm (Fig. 5b), while the TNF spectrum showed no resonant absorption in this range, consistent with the experimental absorption spectrum in Fig. 1c. The electromagnetic field distribution around Au/TNF and TNF at the wavelength of 550 nm was then simulated (Fig. 5c, d). The results clearly demonstrated a strongly enhanced local field concentrated around the AuNPs due to their LSPR mode, confirming the presence of strong field confinement. This field confinement favored hot-electron generation, which was important for improving photodegradation efficiency in the visible spectral range. Considering that Au/TNFs exhibited an absorption peak over 525–575 nm (Fig. 5b), we concluded that the enhanced photodegradation efficiency between 500 and 600 nm for Au/TNFs was attributable to the LSPR effect induced by the addition of AuNPs. It was important to note that the enhanced resonant absorption of Au/TNFs under LSPR effect (Fig. 5b) included not only improved photon absorption by TiO₂ in close proximity to AuNPs but also strong direct photon absorption by the AuNPs themselves. This latter effect contributed to the generation of hot carriers and induces a photothermal effect, both of which further enhanced the overall photocatalytic performance.

Plasmonic hot electron-involved photocatalytic reaction mechanism

Based on the previously discussed LSPR enhancement discussed above, the mechanism of plasmonic hot electron involvement in the photocatalytic degradation process by Au/TNFs under visible light is qualitatively explained and schematically illustrated in Fig. 6a. Effective separation and the rapid transfer of photogenerated electrons (e^–) and holes (h⁺) are the crucial factors for high photocatalytic efficiency. Upon irradiation with visible light, electrons in the AuNPs are excited, followed by rapid thermalization of these electrons through electron-electron interactions within ~ 10 fs⁵². A fraction of these “hot” electrons are subsequently transferred to the conduction band (CB) of the adjacent TiO₂ across the Schottky barrier, with the transfer rate highly dependent on the interface properties⁵³. These CB electrons migrate to the TiO₂ surface, where they are captured by the dissolved O₂ molecules to produce superoxide radical anions (O₂^•−). Subsequently, the degradation of MB occurs at the active centers through oxidative species like superoxide radicals and hydroxyl radicals. The key reactions in the photocatalytic redox process are listed as follows^24,54,55:

$${{rm{O}}}_{2}+{e}^{-}to {text{O}}_{2}^{{{bullet }}-}$$

(1)

$${{rm{H}}}_{2}text{O}leftrightarrow {text{OH}}^{-}+{{rm{H}}}^{+}$$

(2)

$${text{O}}_{2}^{{{bullet }}-}+{text{H}}^{+},to {}^{{{bullet }}}{rm{O}}{rm{OH}}$$

(3)

$$2{}^{{{bullet }}}{rm{O}}{rm{OH}}to {text{H}}_{2}{text{O}}_{2}+{text{O}}_{2}$$

(4)

$${text{H}}_{2}{text{O}}_{2}mathop{to }limits^{hv}2{}^{{{bullet }}}{text{OH}}$$

(5)

$${text{H}}_{2}{text{O}}_{2}+{e}^{-}to {}^{{{bullet }}}{text{OH}}+{text{O}}{text{H}}^{-}$$

(6)

$${text{h}}^{+}+text{O}{text{H}}^{-}to {}^{{{bullet }}}{text{OH}}$$

(7)

$${text{Reactant}}+{text{O}}_{2}^{{{bullet}}-}({text{or}};{}^{{{bullet}}}{text{OH}})to {text{degraded}}; {text{products}}$$

(8)

**Fig. 6: Mechanism of visible light-driven photodegradation by Au/TNFs and optical paths in PFOM and PPOM reactors.**

For the remaining hot electrons in the AuNPs, their energy is eventually transferred to the lattice through electron-phonon interactions on a time scale of ~ 100 fs, generating localized heating⁵². This localized heating effect may diffuse into nearby TiO_2, facilitating the catalytic reaction in a manner similar to conventional thermal-assisted catalysis. Both the hot-electron-driven catalytic reactions and the localized heating-assisted catalysis can be further investigated through ultrafast pump-probe spectroscopy, though this is beyond the scope of the present study.

In summary, the enhanced visible light photocatalysis by Au/TNFs primarily arises from three key factors: (1) the strong plasmonic resonance of AuNPs, leading to enhanced light absorption of TNFs nearby AuNPs, (2) the efficient transfer of hot electrons from AuNPs to TiO₂, facilitated by the ultrafast migration rate^56,57, and (3) the localized heating effect generated by the plasmonic AuNPs.

Photocatalytic enhancement due to reactor configuration

In addition to the photocatalytic reaction mechanism, several factors associated with the reactor configuration may contribute to the superior photocatalytic efficiency in the PPOM.

Firstly, densely packing the photocatalysts significantly increases the surface area (SA) and surface-area-to-volume ratio (SA:V) compared to the film-mode configuration, where the photocatalyst is only deposited on the bottom surface. To estimate the SA and SA:V of different microreactor configurations, we assumed close-packing of photocatalyst particles within the microchamber. Detailed calculations for these estimations are provided in the Supplementary Information in “Estimation of surface-area-to-volume ratio for different configurations” Section. The SA of PPOM was estimated to be 20 times that of PFOM, and the SA:V of PPOM was estimated to be 74.2 times that of PFOM. This increased SA and SA:V in the PPOM configuration greatly facilitated interaction between the photocatalyst and the MB solution, thereby enhancing catalytic efficiency.

Secondly, dense packing promotes a more uniform distribution of photocatalysts within the reaction chamber, which shortens the diffusion path and promotes the mass transfer of reactants to catalytic sites. Therefore, the overall reaction rate in this packed-mode configuration is enhanced, offering a clear advantage over both slurry-mode and film-mode configurations.

Thirdly, the packed photocatalysts inside the optofluidic reactors increase the optical path due to light scattering within the densely packed environment, as depicted in Figs. 6b and 6c. This extended optical path enhances light interaction with the packed photocatalyst, promoting more efficient electron-hole pair generation and consequently improving photocatalytic efficiency.

Fourthly, the close packing of Au/TNFs in the PPOM configuration amplifies the LSPR effects due to the higher concentration of AuNPs. These enhanced LSPR effects facilitate more efficient electron transfer processes, further boosting visible-light photocatalytic activity.

In summary, the packed-bed mode configuration creates a densely packed photocatalyst environment with increased surface area, enhanced mass transfer efficiency, extended optical paths, shorter diffusion distances, and higher AuNPs concentration. Together, these factors improve overall system efficiency and amplify the benefits of LSPR-enhanced photocatalysis.

Discussion

Supplementary Table 2 provides a comparative overview of our Au/TNFs-PPOM system against established water purification technologies, highlighting the unique advantages and limitations of each method. Our Au/TNFs-PPOM system demonstrated high photocatalytic efficiency under visible light and holds potential for low operational costs, particularly when utilizing solar energy. While the initial setup involved costs related to material synthesis and microfluidic fabrication, the Au/TNFs-PPOM offered substantial long-term savings due to minimal maintenance and reduced energy requirements. Anticipated advancements in 3D printing and microfabrication technologies are expected to further reduce these initial costs, making large-scale production more feasible and economical⁵⁸. In contrast, traditional water purification methods, such as UV disinfection and filtration, often incur higher operational costs due to dependence on artificial light sources, the need for secondary treatments to eliminate by-products, and the maintenance of large-scale facilities⁵⁹, although these methods are well-established and effective for large-scale industrial applications. Our study focuses on a lab-scale prototype, but we anticipate that the Au/TNFs-packed system can be scaled up for industrial applications by integrating multiple units in parallel or series to enhance throughput⁶⁰. However, scaling up poses practical challenges, such as maintaining high surface-area-to-volume ratios, ensuring uniform flow distribution, optimizing mass transfer, and managing pressure drops across larger systems^61,62,63.

This study demonstrated the effectiveness of Au/TNFs within optofluidic microreactors for degrading MB under visible light. Combining experimental and numerical analyzes, we revealed how the incorporation of AuNPs and the utilization of the PPOM synergistically enhanced photodegradation performance by leveraging LSPR effect, rapid hot-electron transfer, and improved mass transfer. Specifically, Au/TNFs exhibited over 46 times the photocatalytic efficiency of TNFs alone, with an additional nearly 60-fold enhancement when employing the PPOM configuration. Furthermore, the photodegradation efficiency of PPOM with Au/TNFs was more than 2 times that of PFOM, underscoring the importance of optimizing SA:V and optical path length.

Beyond water purification, our findings on this plasmonic-optofluidic-photocatalytic system open avenues for diverse applications. This optofluidic platform could be adapted for environmental remediation efforts such as air purification or volatile organic compounds (VOCs) removal from industrial emissions⁶⁴. Additionally, the platform is promising for solar-driven energy applications like hydrogen production and carbon dioxide reduction, where photocatalytic efficiency and scalability are key concerns^65,66. In high-value chemical synthesis, the ability to precisely control of reaction conditions within microreactors can enable enhanced selectivity and efficiency through fine-tuning reaction conditions at the microscale⁶⁷. Furthermore, the platform adaptability suggests potential in biomedical applications, such as targeted drug delivery, where photocatalysis could trigger localized release of therapeutic agents⁶⁸.

Future improvements could focus on achieving uniform packing of photocatalyst within microchannels while addressing issues such as clogging and photocatalysts stability. For industrial-scale applications, further studies should investigate the feasibility of scaling up optofluidic microreactors through multi-unit integration to enhance throughput⁶⁰. Additionally, incorporating supplementary functionalities, such as combining photocatalysis with electrochemical or biocatalytic processes, could broaden the platform applicability. Such enhancements could enable complex reactions across environmental, energy, and medical applications, including air purification, solar-driven chemical processes, and targeted drug delivery. Continued advancements in this field are essential to address contemporary environmental challenges and support sustainable practices, paving the way for a cleaner and healthier future.

Methods

Chemicals

Tetrachloroauric(III) acid trihydrate (HAuCl₄∙3H₂O, AR), sodium borohydride (NaBH_4, 98%), thioglycolic acid (TGA, AR, 90%), diethylenetriamine (DETA, 99%), and titanium tetraisopropanolate (TIP, 98%) were purchased from Aladdin Reagent Company. Ethanol (AR), isopropanol alcohol (IPA, AR) was purchased from Beijing Chemical Reagent Company. All chemicals were used without any further purification.

Synthesis of TNFs

TNFs were synthesized using a solvothermal method. A mixture containing 350 mL of IPA, 300 µL of DETA and 12.5 mL of TIP was stirred for over 5 min at room temperature to ensure homogeneity. The mixture was then transferred into a Teflon-lined stainless-steel autoclave and heated at 200 ^oC for 48 h. After the reaction, the autoclave was allowed to cool naturally to room temperature.

The resulting precipitate was separated by centrifugation at 9000 rpm for 10 min and redispersed in ethanol using ultrasonication to break up any agglomerates. This washing procedure (centrifugation and ultrasonication) was repeated three times to remove any organic residues and unreacted materials. The final product was dried and subjected to annealing at 400 ^oC for 1 h in a muffle furnace to crystallize the TNFs.

Synthesis of Au/TNFs

To decorate TNFs with AuNPs, 100 mg of the synthesized TNFs were dispersed in 20 ml of ethanol, followed by the addition of 200 µL of TGA. This mixture was stirred for over 3 h to ensure sufficient surface modification of the TNFs with TGA. Afterward, the solution was centrifuged at 9000 rpm for 10 min, and the precipitate was redispersed in ethanol by ultrasonication. The centrifugation and ultrasonication processes were repeated twice to remove any unbound TGA molecules.

Subsequently, 20 mL of 1 mM HAuCl₄∙3H₂O was added to the TGA-modified TNFs dispersion. This mixture was stirred in the dark for over 3 h to allow the adsorption of gold ions onto the TNFs. After the adsorption, the solution was centrifuged at 9000 rpm for 10 min to remove unbound gold ions.

AuNPs were then reduced onto the TNFs surface by adding 100 mg of NaBH₄ dissolved in 5 mL of water dropwise to the dispersion. The solution was stirred for 10 min to complete the reduction process. The resulting Au/TNFs were collected by centrifugation and washed with deionized water multiple times to remove excess reducing agent and by-products. Finally, the samples were dried at 70 ^oC for 12 h before further characterization and testing.

Photocatalyst characterization

The morphologies and structures of the TNFs and Au/TNFs were investigated using a combination of electron microscopy techniques. FE-SEM was conducted using a JEOL JSM-6490 microscope to observe surface morphology and overall particle shape. To obtain detailed information on internal structures and particle sizes, TEM and HRTEM were performed using a JEOL JEM2011 microscope. In addition, HAADF-STEM was used to further confirm the distribution and morphology of the AuNPs on the surface of the TNFs. TEM elemental mapping provided spatial distribution data for the constituent elements within the composite materials. Furthermore, TEM coupled with EDAX was employed for quantitative elemental analysis, confirming the presence and uniform deposition of AuNPs on the TNFs.

The crystalline structure and phase composition of the samples were analyzed by an XRD, performed on a Rigaku SmartLab diffractometer equipped with Cu K_α radiation at an operating voltage of 45 kV. The scan ranging from 20^o to 60^o and a step size of 0.02^o were set to capture key diffraction peaks in the samples.

XPS was performed to analyze the chemical composition of the composites. The measurements were conducted using a Thermo Scientific K-Alpha XPS system equipped with a monochromatic Al Kα X-ray source (1486.6 eV). A spot size of 400 μm was used for analysis. High-resolution XPS spectra were acquired for Ti 2p, O 1s, and Au 4f regions to confirm the chemical states and composition of the elements. The base pressure of the analysis chamber was maintained at around 5 × 10⁻⁹ mbar. The binding energy scale was calibrated using the C 1s peak of adventitious carbon at 284.8 eV as a reference.

The elemental composition of Au/TNFs was analyzed using ICP-MS (Agilent 7800). Approximately 50 mg of each sample was accurately weighed and dissolved in a mixture of concentrated nitric acid (HNO₃) and hydrofluoric acid (HF) under controlled heating to ensure complete dissolution. The resulting solution was then diluted with deionized water for ICP-MS analysis. A standard calibration curve was prepared using certified reference materials to quantify the concentration of Au in Au/TNFs. Data was collected in triplicate to ensure accuracy and reproducibility.

To evaluate the optical properties of the photocatalysts, UV-DRS were recorded using a Shimadzu UV-2459 spectrophotometer equipped with an integrating sphere. Measurements were conducted across the wavelength ranging from 300 nm to 800 nm, with a resolution of 1 nm. The samples were prepared by suspending 0.2 g of TNFs or Au/TNFs in 2 mL of deionized water according to previously reported method²⁸. To promote controlled aggregation and prevent agglomeration, 6.67 µL of acetylacetone and 3.33 µL of Triton X-100 were added, respectively. Additionally, 0.02 g of polyethylene glycol (PEG) was introduced, and the mixture was stirred overnight. The suspension was then evenly spread onto the quartz glass slide using a glass roller. Before the measurement, the film was dried at 80 °C in air and subsequently calcined at 500 °C for 2 h to stabilize the photocatalyst film.

Planar optofluidic microreactors fabrication

The planar optofluidic microreactors were fabricated by bonding a patterned polydimethylsiloxane (PDMS) slice to a glass slide using oxygen plasma. This patterned PDMS slice was prepared via standard soft photolithography using the Sylgard 184 elastomer kit (Dow Corning Corporation) and an SU-8 mold (SU-8 50, MicroChem Corp.). The soft photolithography process involved the design of mask for microfluidic channels, the fabrication of an SU-8 mold via photolithography, and the subsequent molding and bonding of the optofluidic microreactor.

The schematic design of the microfluidic channels features an inlet, a central reaction chamber, and an outlet (Supplementary Fig. 3a). Tree-like branching microchannels at both the inlet and outlet ensure a uniform distribution of the solution throughout the central chamber, which has dimensions of 1 cm × 1 cm.

To create the SU-8 mold (Supplementary Fig. 3b), a 4-inch silicon wafer was firstly cleaned by acetone, IPA and deionized water in an ultrasonic bath. After drying, the wafer was treated in a plasma cleaner. A 40-μm-thick layer of SU-8 50 negative photoresist was then deposited through two successive steps of spin coating, first at 500 rpm for 30 s and then at 1500 rpm for 60 s. The coated wafer was baked at 65 ^oC for 5 min, followed by a hard bake at 95 ^oC for 15 min to solidify the photoresist. UV exposure (45 s with an exposure energy of 180 mJ/cm²) was then performed using a photomask with the designed pattern to harden the specific parts of the photoresist. After a post-bake at 65 ^oC for 5 min and 95 ^oC for 15 min, the silicon wafer was developed in SU-8 developer for 3 min and rinsed with IPA for 10 s. After drying, the mold was hard baked at 150 ^oC for 5 min to finalize its structure, achieving a thickness of approximately 40 µm.

Next, the PDMS compound, formed by mixing PDMS and its curing agent at 10:1 weight ratio, was poured onto the SU-8 mold and cured on a hot plate at 85 ^oC for 30 min. The cured PDMS was then peeled off and bonded to a cleaned glass slide with 30 s of oxygen plasma treatment (Harrick Plasma, USA) after the inlet and the outlet were punched. The fabricated planar optofluidic microreactors were then ready for testing.

Planar film-mode optofluidic microreactors fabrication

The planar optofluidic microreactors were loaded with photocatalysts in two modes for MB degradation: film-mode for PFOM and packed-mode for PPOM.

In the PFOM, the planar microreactors were loaded with a thin film of photocatalysts for the degradation of MB. First, 0.2 g of ground TNFs or Au/TNFs were suspended in 2 mL of deionized water, along with 6.67 µL of acetylacetone, 3.33 µL of Triton X-100 and 0.02 g of PEG. This mixture was stirred overnight to ensure uniform suspension. The suspension was then uniformly rolled onto the glass slide using a glass roller, with the edges protected by Scotch tapes, leaving a 1 cm × 1 cm square for photocatalyst deposition. After air drying at 80 °C, the Scotch tapes were removed, and the glass slide was calcined at 500 °C in air for 2 h to stabilize the photocatalyst film. Subsequently, the film was sandwiched between the molded PDMS slice (top) and the glass slide (bottom) to form the planar film-mode optofluidic microreactor.

Planar packed-mode optofluidic microreactor fabrication

The PPOM was fabricated by directly filling the as-prepared TNFs or Au/TNFs into the microreactor chamber. To achieve this, 0.1 g of TNFs or Au/TNFs powders were finely ground and suspended in 1 mL of ethanol (10 wt%) as a dispersing agent. The mixture underwent ultrasonic treatment for 1 h to ensure homogeneous dispersion and was then manually injected into the microreactor chamber using a syringe. The injection process was repeated several times until the microreactor exhibited a uniform milky white appearance for TNFs or a dark pink color for Au/TNFs, indicating successful packing. A 0.22 µm pore-size filter was placed over the outlet to prevent the loss of photocatalyst particles during operation. Afterwards, the microreactor was left to air-dry until all ethanol had evaporated, leaving behind a densely packed layer of photocatalysts. These packed reactors were then ready for photocatalytic performance evaluation.

Photocatalytic degradation of MB

The photocatalytic efficiencies of TNFs-loaded and Au/TNFs-loaded PPOM and PFOM were evaluated using MB as a model compound. A simulated solar source (AM 1.5 G, 300 mW/cm²) equipped with a UV-cutoff filter (λ > 420 nm) was employed to provide visible light illumination. The microreactors were positioned at a distance of 15 cm from the light source to ensure uniform exposure. A 5 × 10⁻⁵ M of MB solution was injected into the photocatalyst-loaded optofluidic microreactor through the inlet, with flow rates controlled by a syringe pump (Longer) to regulate the residence time of the solution in the reaction chamber. The reaction time, t_r, which represents the effective residence time of the MB solution in the reaction chamber, was calculated as follows:

$${t}_{r}=frac{V}{Q}$$

(9)

where V is the volume of reaction chamber and Q is the flow rate of the injected MB solution. The degraded MB solution was collected from the outlet, and the change in concentration was monitored using a UV-vis spectrophotometer by recording the absorbance at 664 nm. For comparison, control experiments were conducted in a conventional slurry-mode reactor, where a fixed amount (0.5 g) of TNFs or Au/TNFs was suspended in 5 mL of MB solution and stirred continuously.

Kinetic analysis

The photodegradation kinetics of MB were analyzed using a pseudo-first-order model⁶⁹. The degradation rate constant k was calculated based on the following equation:

$${rm{In}}left(frac{{C}_{0}}{C}right){={kt}}_{r}$$

(10)

where C₀ is the initial MB concentration, C is the concentration after treatment, and t_r is the residence time.

Prior to the photodegradation experiments, dark adsorption tests were performed to determine the adsorption capacity of the photocatalysts. For these tests in optofluidic microreactors, MB solution was injected into the microreactors at a flow rate of 5 µL min^-1 in darkness. The solution exiting the reactors was collected every 15 min for concentration determination to assess the time required for absorption equilibrium. Similarly, for the slurry-mode reactors, the photocatalyst suspensions were stirred with MB solution in darkness, and the MB solution was monitored at 15-min intervals to determine the absorption saturation time.

Reusability of PPOM configuration

The reusability of TNFs and Au/TNFs in the PPOM configuration was assessed to evaluate their potential for industrial applications. After each photodegradation experiment, the microreactors were thoroughly rinsed with deionized water under UV-light irradiation to remove any absorbed MB molecules from the photocatalyst surfaces. To ensure consistent performance, a dark adsorption step was performed before each subsequent cycle of MB photodegradation. Three consecutive cycles of MB photodegradation were conducted with both TNFs and Au/TNFs in the PPOM configuration to evaluate the photocatalyst stability and reusability. Changes in photodegradation efficiency across the cycles were recorded and analyzed. The procedures for the reusability experiments are illustrated in Supplementary Fig. 5.

Action spectrum analysis of photocatalyst in PPOM

The action spectrum analysis, also known as the wavelength dependence of apparent quantum efficiency⁷⁰, was performed to assess the influence of AuNPs addition on the photocatalytic activity for MB degradation. This analysis was performed for both TNFs-loaded and Au/TNFs-loaded PPOM by evaluating the MB photodegradation efficiency per photon flux density (denoted as η). Monochromatic light at different wavelengths (i.e., 430, 450, 475, 500, 520, 550, 600, and 650 nm) were generated using bandpass filters (full width at half maximum of 20 ± 3 nm) with a xenon lamp as the light source. The PPOM reactors were exposed to the monochromatic light, and MB solutions (5 × 10⁻⁵ M) were injected into the microreactors at a controlled flow rate of 20 µL min⁻¹. The MB solution exiting the microreactor was collected, and the concentration of MB was measured to assess photodegradation. The apparent quantum efficiency (η) was calculated using the following equation:

$$n=frac{{n}_{{MB}}}{{n}_{p}}=left[frac{({C}_{0}-C){V}_{0}{N}_{A}}{{t}_{r}}right]/left(frac{Plambda }{{hc}}right)=frac{{{hcN}}_{A}({C}_{0}-C)Q}{Plambda }$$

(11)

where n_MB is the number of degraded MB molecules, n_p is the number of incident photons per unit time, C₀ is the initial injected MB concentration, C is the MB concentration after photodegradation, V₀ is the volume of reaction chamber, t_r is the residence time of the solution in the reaction chamber, Q = V₀/t_r is the flow rate, N_A is the Avogadro’s constant, P is the incident light power, λ is the wavelength of the incident light, h is the Planck’s constant, and c is the speed of light in vacuum. The efficiency η represents the number of degraded MB molecules per incident photon. For comparison purposes, the highest η value was normalized to 1 for both TNFs and Au/TNFs.

FDTD simulations

FDTD simulations were performed using Ansys Lumerical 2023 R1.2 software to model the optical behavior of TNFs and Au/TNFs under visible light irradiation. Due to the complex geometry of the porous TNF structure, an EMA was applied. TNF was modeled as a 450-nm-diameter nanosphere with an effective permittivity equivalent to a 75% water and 25% amorphous TiO₂ mixture. For the Au/TNFs system, the model was extended by randomly dispersing thousands of 5-nm-diameter AuNPs on the surface of TNFs. The simulation was then used to calculate the local field distribution around the TNFs and Au/TNFs at the wavelength of 550 nm. Permittivity values for amorphous TiO₂ and Au were sourced from Devlin et al.⁷¹ and Johnson and Christy⁷², respectively.