Silver single atoms and nanoparticles on floatable monolithic photocatalysts for synergistic solar water disinfection

Abstract

Photocatalytic water disinfection technology is highly promising in off-grid areas due to abundant year-round solar irradiance. However, the practical use of powdered photocatalysts is impeded by limited recovery and inefficient inactivation of stress-resistant bacteria in oligotrophic surface water. Here we prepare a floatable monolithic photocatalyst with ZIF-8-NH₂ loaded Ag single atoms and nanoparticles (Ag_SA+NP/ZIF). Atomically dispersed Ag sites form an Ag−N charge bridge, extending the lifetime of charge carriers and thereby promoting reactive oxygen species (ROS) generation. The photothermal effect of the plasmonic Ag nanoparticles reduces the bacterial resistance to ROS and impairs DNA repair capabilities. Under sunlight irradiation, the synergistic effect of Ag single atoms and nanoparticles enables 4.0 cm² Ag_SA+NP/ZIF to achieve over 6.0 log inactivation (99.9999%) for the stress-resistant Escherichia coli (E. coli) in oligotrophic surface water within 30 min. Furthermore, 36 cm² Ag_SA+NP/ZIF is capable of disinfecting at least 10.0 L of surface water, which meets the World Health Organization (WHO) recommended daily per capita drinking water allocation (8.0 L). This study presents a decentralized and sustainable approach for water disinfection in off-grid areas.

Introduction

Water pollution caused by pathogenic microorganisms poses a serious threat to public health. Millions of individuals in developing nations succumb to mortality each year due to the consumption of water contaminated with pathogenic microorganisms^1,2. Every 2 min a child under the age of five perishes from diarrheal diseases caused by poor sanitation and substandard water quality^3,4. Since the 20th century, the implementation of urban water supply network systems and centralized water treatment facilities has significantly mitigated infections caused by waterborne pathogens. However, effective access to safe drinking water remains scarce in off-grid rural areas, such as in sub-Saharan Africa, North Africa, and the Middle East⁵, often due to lack of electricity and other vital water treatment infrastructures to ensure reliable and safe drinking water access. With the intensification of climate change, it is estimated that 3.6 billion people worldwide currently live in regions that may face potential water scarcity, and this number could rise to 5.7 billion by 2050⁶. Therefore, improving water management and sanitation is critical for communities that have no or limited access to traditional water disinfection methods.

Point-of-use (POU) disinfection at household or community scale is one of the most promising solutions in addressing this issue^5,7. POU disinfection technologies eliminate the need for expensive large-scale water treatment equipment and avoid the complexities associated with transportation and storage procedures. Despite decades of technological advances, existing POU disinfection technologies still face issues with high energy consumption (e.g., boiling water), generation of carcinogenic by-products (e.g., chlorination), and high costs (e.g., reverse osmosis)^8,9,10. Fortunately, among the 64 countries with <80% rural access to basic drinking water services, solar irradiance (5.59 kWh m⁻² per day) significantly exceeds the global average (4.70 kWh m⁻² per day)⁵, offering valuable opportunities for solar energy-driven water treatment. Therefore, the development of advanced materials and processes that can fully harness solar energy to inactivate microorganisms will contribute to the attainment of environmentally friendly and sustainable water disinfection in off-grid communities.

Single-atom photocatalysts (SAPCs) have been reported for disinfection due to their maximum metal atom utilization and efficient charge separation^11,12,13. SAPCs are known to generate ROS that attack cell membranes and intracellular components like DNA¹⁴. However, microorganisms possess innate damage repair system, such as DNA repair process (SOS response), which confer resistance to ROS stresses¹⁵. Such resistance is more pronounced in microorganisms at late stationary phase in natural waters^16,17,18. In oligotrophic surface water, bacteria at this growth phase are usually starved, and their adaptation to starvation enhances their tolerance to ROS and other environmental stresses¹⁹. This is why these bacteria cultured under above specific conditions are usually called stress-resistant bacteria. Consequently, numerous photocatalysts exhibiting exceptional antimicrobial performance in laboratory show limited bactericidal effectiveness in real water environments, posing challenges for their large-scale applications. Nanoparticles with plasmonic effects, such as Ag and Au, exhibit excellent photothermal properties through collective resonance of free charge carriers and non-radiative decay processes^20,21. Previous studies have shown that photothermal ablation can target tumor cell proteins and promote protein aggregation, consequently interfering with DNA damage repair²². Given that the bacterial inactivation by SAPCs relies on ROS-mediated intracellular DNA damage, any potential reduction in bacterial DNA repair capacity due to the photothermal effect would significantly enhance the bactericidal efficiency. Hence, we speculate that SAPCs integrated with plasmonic nanoparticles could effectively inactivate stress-resistant microorganisms in water.

In the synthesis of SAPCs, the prevalent use of powdered supports such as carbon nitride and metal oxide materials to anchor single atoms may interfere their recovery in aqueous reactions^23,24,25, thereby increasing the contamination risk throughout the disinfection process¹⁶. In addition, the widespread presence of suspended sediment and natural organic matter in the natural water significantly interferes with the light absorbance of SAPCs⁵. To improve catalyst recovery and light absorption, previous studies have employed polymer binders to immobilize catalysts on the platforms such as nickel foam and ceramic plate^26,27, resulting in compromised accessibility of the potential active reaction sites and hindered electron transfer^28,29,30. Consequently, there is an urgent need to avoid using polymer binders while strengthening the stability of the catalytic layer on the support. This requirement can be effectively fulfilled by the fabrication of binder-free floatable monolithic photocatalysts via loading metal cocatalysts into the floatable monolithic supports.

Here, we prepared a silver single-atom and nanoparticle-loaded floatable monolithic photocatalyst (Ag_SA+NP/ZIF) via a two-step synthesis method as shown in Fig. 1a. Firstly, a monolithic metal-organic frameworks ZIF-8-NH₂ layer was synthesized in situ by a ligand evaporation deposition strategy, employing 3D melamine sponges with zinc ions as a substrate, and utilizing 2-methylimidazole and 2-aminobenzimidazole as organic ligands. ZIF-8-NH₂ material possesses semiconducting properties and is frequently employed as support for loading metal cocatalysts^31,32. Abundant N atoms with lone pair of electrons in ZIF-8-NH₂ can form strong coordination with Ag metal atoms, effectively preventing their migration and aggregation^23,33. The low density of sponges ensures that the monolithic photocatalyst can float on water. The monolithic photocatalyst was selectively loaded with Ag_SA and Ag_NP using the ice-photoreduction method, wherein the photocatalytic reaction primarily occurs on the submerged surface with sunlight exposure (Fig. 1b). Ag_SA+NP/ZIF can be utilized in a tablet-like form for convenient recyclable water disinfection. Under natural light irradiation, the photothermal effect of Ag_SA+NP/ZIF synergizes with photogenerated ROS to inactivate stress-resistant bacteria in oligotrophic surface water. This monolithic photocatalyst exhibits practical advantages for POU water disinfection, as evidenced by its performance, recyclability, durability, and cost-effectiveness validated in this study.

Fig. 1: Overview of the floatable monolithic photocatalyst.

a Synthesis schematic of Ag_SA+NP/ZIF. b Disinfection operation schematic of Ag_SA+NP/ZIF. Note that Ag species are magnified in size.

Results

Material characterization

The photocatalyst support consists of a 3D melamine sponge and ZIF-8-NH₂ layer. As shown in Fig. 2a, a dense and continuous ZIF-8-NH₂ layer was deposited on the surface of the melamine sponge. The thickness of ZIF-8-NH₂ coating is 1.0-1.4 μm. Cross-sectional scanning electron microscope (SEM) image (Supplementary Fig. 1) shows that some ZIF-8-NH₂ crystals were infused into the inner sponge layer, which potentially enhances the adhesion between the surface ZIF-8-NH₂ layer and substrate. X-ray diffraction (XRD) patterns verified that the surface layer exhibits intact ZIF-8-NH₂ crystal structure (Supplementary Fig. 2)³⁴. Transmission electron microscopy (TEM) images show successful loading of Ag nanoparticles with an average size of 13.1 nm (Fig. 2b and Supplementary Fig. 3).

Fig. 2: Structural characterization of Ag_SA+NP/ZIF.

a Top (i) and (ii) cross-sectional view SEM images of monolithic ZIF-8-NH₂ support. b TEM image of Ag_SA+NP/ZIF. Inset is the statistical size distribution of Ag nanoparticles. c AC-HAAD-STEM image of Ag_SA+NP/ZIF (the bright dots in the red circles are Ag single atoms). Inset is the HAAD-STEM image of Ag_SA+NP/ZIF. d HAADF-STEM with elemental mapping images of Ag_SA+NP/ZIF. e Fourier-transformed EXAFS results of Ag_SA/ZIF, Ag_NP/ZIF, Ag_SA+NP/ZIF, Ag foil and Ag₂O at the Ag K-edge. f EXAFS fitting curve of Ag_SA+NP/ZIF in R space.

Aberration-corrected high-angle-annular-dark-field scanning transmission electron microscopy (AC-HAADF-STEM) images (Fig. 2c and Supplementary Fig. 4) show atomically dispersed Ag (circled bright spots). The elemental mapping images also confirm that Ag atoms and nanoparticles dispersed on ZIF-8-NH₂ (Fig. 2d). Thus, Ag single atoms and nanoparticles were both present on this monolithic photocatalyst. The total Ag loading content was 1.28 wt% as detected by an inductively coupled plasma optical emission spectrometer (ICP-OES), where the proportion of Ag single atoms and Ag nanoparticles was 46.9% and 53.1%, respectively. In addition, the 1.21 wt% Ag single-atom-loaded ZIF-8-NH₂ (Ag_SA/ZIF) and 1.16 wt% Ag nanoparticle-loaded ZIF-8-NH₂ (Ag_NP/ZIF) were also prepared for comparative studies and their AC-HAADF-STEM and TEM images are shown in Supplementary Fig. 5.

A measurement of X-ray absorption fine structure (XAFS) was carried out to further investigate the dispersion and coordination environment of Ag species³⁵. Fourier-transformed (FT) extended Ag K-edge X-ray absorption fine structure (EXAFS) spectra show that Ag_SA/ZIF and Ag_SA+NP/ZIF share a common peak at 1.56 Å, which may be attributed to the coordination of Ag single atoms with N or C (Fig. 2e). The best-fitting result for EXAFS spectrum of Ag_SA+NP/ZIF shows that the dominant peak at 1.56 Å can be attributed to the Ag−N coordination (Fig. 2f and Supplementary Table 1). The Ag−N and Ag−Ag coordination numbers are 1.5 and 4.6 at bond distances of 2.02 Å and 2.83 Å, respectively. In addition, a peak at 2.7 Å was observed in both Ag_SA+NP/ZIF and Ag_NP/ZIF, resembling the characteristic Ag−Ag peak of metallic Ag, thereby confirming the presence of Ag nanoparticles in these samples.

Optical and photothermal properties

Light capture and photogenerated carrier separation play key roles in photocatalytic disinfection reactions^16,36. As shown in Fig. 3a, compared to ZIF-8-NH₂, Ag_SA+NP/ZIF exhibits the strongest light absorption in the visible region, which may result from the introduction of Ag single atoms and nanoparticles^37,38. The light absorption intensity of Ag_NP/ZIF is higher than that of Ag_SA/ZIF, indicating a more significant contribution from the plasmonic Ag nanoparticles to light adsorption capacity. The noise at 300-380 nm originates from the sponge (Supplementary Fig. 6a). The plasmonic peak of Ag nanoparticles is absent at 450 nm, probably due to the low Ag loading (1.16 wt% for Ag_NP/ZIF) and the overlaps of the significant optical absorption of ZIF-8-NH₂ with the weak plasmonic peak³⁹.

Fig. 3: Optical and photothermal properties of the photocatalysts.

a UV-Vis diffuse reflectance spectra of photocatalysts. b Electromagnetic simulations: enhancement factor of electric field on the Ag nanoparticle-ZIF-8-NH₂ surface under visible light (450 nm) irradiation. c, d Steady-state fluorescence spectra (c) and TRPL decay curves (d) of photocatalysts, the excitation wavelength is 375 nm. e Electrochemical impedance spectra of photocatalysts, inset shows the corresponding equivalent circuit diagram. R_s: Solution resistance; CPE: Constant phase element. f Infrared thermal images of photocatalysts under simulated visible light (400-760 nm) irradiation at an intensity of 100 mW cm⁻² for 15 min. g Temperature elevation curves of photocatalysts under simulated visible light (400-760 nm) irradiation at an intensity of 100 mW cm⁻². h Temperature elevation curves of water after adding the monolithic photocatalyst under natural light irradiation. Reaction condition: 2 × 2 cm² catalyst; 0.1 L Yellow River water from upper reaches (Longyang Gorge); 33.0-35.0 mW cm⁻² sunlight intensity.

The increased light absorption by Ag nanoparticles can be attributed to the local electromagnetic enhancement resulting from the localized surface plasmon resonance⁴⁰. Noble metal (e.g., gold and silver) nanoparticles contain large numbers of freely mobile electrons that can strongly interact with light by either absorbing or scattering photons^20,21. The oscillating electric field of the incoming radiation induces coherent collective oscillation of the free electrons on the metal surface^20,21. Supplementary Fig. 6b shows the simulation of the electromagnetic field at the interface between one single Ag nanoparticle (e.g., 13 nm in diameter) and ZIF-8-NH₂ under 450 nm irradiation using COMSOL Multiphysics. The average enhancement factor, defined as P = | E | /E₀, quantifies the amplification of the localized electric field (E) after light irradiation relative to the electric field (E₀) before light irradiation in the near-field region of the Ag nanoparticle. Clearly, along the distance from the core of the Ag nanoparticle, the local electromagnetic field yielded a P-value of >10-fold due to the dynamic charge accumulation at the Ag nanoparticle surface under a 450 nm planewave excitation. In addition, the effects of ZIF-8-NH₂ thickness and Ag nanoparticle size on the electromagnetic field enhancement are evident. The results presented in Fig. 3b demonstrate a proportional increase in the enhancement factor as the radius of Ag nanoparticles ranges from 1.0 to 8.0 nm. It is observed that the thickness of ZIF-8-NH₂ does not significantly impact the enhancement factor when exceeding 4.0 nm. Furthermore, there is no enhancement of the electric field intensity within the near field of pure ZIF-8-NH₂ (Supplementary Fig. 6c), suggesting that the interface between ZIF-8-NH₂ substrate and Ag nanoparticle is essential for enhancing the light absorbance of the photocatalyst²².

Furthermore, the band gap of Ag_SA+NP/ZIF was determined to be 2.78 eV, which is smaller than that of pure ZIF-8-NH₂ (3.47 eV) as compared by the Tauc plots in Supplementary Fig. 7a, b. The strong local electric field generated by Ag nanoparticles or the formation of Ag−N coordination might influence the electronic density distribution in ZIF-8-NH₂, consequently modulating its band gap^41,42. Mott-Schottky plots of Ag_SA+NP/ZIF were employed to determine the conduction band minimum (CBM) (Supplementary Fig. 7c). Higher CBM (−0.43 V vs. NHE) of Ag_SA+NP/ZIF than reduction potential of O₂ to O₂^•−(−0.33 V vs. NHE) implies the feasibility of Ag_SA+NP/ZIF for ROS generation⁴³.

The concentration of ROS production depends on the separation of photogenerated carriers, which was investigated by steady-state fluorescence spectroscopy and time-resolved photoluminescence (TRPL) spectroscopy. As shown in Fig. 3c, both Ag_SA+NP/ZIF and Ag_SA/ZIF exhibited lower fluorescence intensities compared to Ag_NP/ZIF, indicating that the recombination of electron-hole pairs was inhibited. This is because the atomically dispersed Ag provides more reactive sites to capture the photogenerated electrons⁴⁴. As shown in Fig. 3d and Supplementary Table 2, the average lifetime of photogenerated charge carriers in Ag_SA/ZIF (1.67 ns) exceeds those in Ag_SA+NP/ZIF (1.53 ns) and Ag_NP/ZIF (1.10 ns). The Ag content of Ag_SA/ZIF is 1.21 wt%, which is lower than that of Ag_SA+NP/ZIF (1.28 wt%). Nonetheless, Ag_SA/ZIF exhibits a longer charge carrier lifetime, indicating that Ag_SA is more conducive to the capture of photogenerated electrons. Moreover, Ag_SA+NP/ZIF and Ag_SA/ZIF show smaller impedance arc radii in electrochemical impedance spectra compared to Ag_NP/ZIF (Fig. 3e). The charge transfer resistance (R_ct) of Ag_SA/ZIF is 25.5 KΩ, which is lower than that of Ag_SA+NP/ZIF (39.0 KΩ). The atomically dispersed Ag atoms bonded with N atoms facilitates the formation of Ag−N charge bridge. This configuration significantly enhances the metal-support interaction, thereby boosting the charge transfer³⁹. Hence, we conclude that the incorporation of Ag single atoms endows monolithic ZIF-8-NH₂ with intensive photochemical properties.

The photothermal properties of the dry monolithic photocatalysts were investigated by infrared thermal imaging. After exposure to simulated visible light (400-760 nm) at an intensity of 100 mW cm⁻², the dry surface temperatures of Ag_SA+NP/ZIF and Ag_NP/ZIF rapidly increase to 67.1 and 72 °C within 15 min, respectively, followed by a plateau (Fig. 3f, g). In comparison, monolithic ZIF-8-NH₂ and Ag_SA/ZIF demonstrated weak photothermal effects (<36 °C) after 60 min light irradiation. Both Ag_SA+NP/ZIF and Ag_NP/ZIF exhibit greater photothermal effects due to surface plasmon resonance of Ag nanoparticles²². We further evaluated the water temperature elevation induced by monolithic photocatalysts under natural light irradiation of 33.0-35.0 mW cm⁻², which is approximately equivalent to the average daily irradiance (5.59 kWh m⁻² per day) for most off-grid areas⁵. As shown in Fig. 3h, the utilization of Ag_SA+NP/ZIF resulted in a significant elevation of water temperature to 41.8 °C within 15 min light irradiation, whereas the ZIF-8-NH₂ and Ag_SA/ZIF-treated group exhibited only a minor increase in water temperature (<36.7 °C). These results indicate that Ag_SA+NP/ZIF possesses pronounced photothermal heating properties.

Natural water disinfection performance

We investigated the water disinfection performance of Ag_SA+NP/ZIF toward stress-resistant bacteria E. coli at late stationary phase cultivated in the river water from the upper reaches of Yellow River. The chemical parameters of the water are shown in Supplementary Table 3, wherein the total organic carbon (TOC) is 5.49 mg L⁻¹ and primarily originates from natural organic matter that exhibits light absorption properties. However, unlike water-dispersed particulate catalysts, Ag_SA+NP/ZIF floating at the water surface (Fig. 4a) remains unaffected by impurities in the water, thereby ensuring efficient light delivery.

Fig. 4: Photocatalytic disinfection performance under solar irradiation.

a Disinfection operation schematic of Ag_SA+NP/ZIF. b Bactericidal efficiencies of 1, 2 and 4 cm² Ag_SA+NP/ZIF after 30 min solar irradiation in 0.1 L Yellow River water from upper reaches (Longyang Gorge). c Photocatalytic bactericidal performance of 1 cm² Ag_SA+NP/ZIF against E. coli in 0.1 L Yellow River water from upper reaches (Longyang Gorge) and laboratory-derived E. coli under solar irradiation. The lines represent time curves. d Time profiles of E. coli inactivation by 39.8 μg L⁻¹ Ag⁺ and 207.2 μg L⁻¹ Zn²⁺. e Time profiles of E. coli inactivation in 0.1 L Yellow River water from upper reaches (Longyang Gorge) by 4 cm² Ag_SA+NP/ZIF, Ag_SA/ZIF, Ag_NP/ZIF and ZIF-8-NH₂ photocatalysts. f Time profiles of S. aureus and virus MS2 inactivation in 0.1 L Yellow River water from upper reaches (Longyang Gorge) by 4 cm² Ag_SA+NP/ZIF photocatalyst. g Time profiles of E. coli inactivation in 0.1 L Yellow River water from middle reaches (Toudaoguai Station) and lake water (Shichahai) by 4 cm² Ag_SA+NP/ZIF photocatalysts. h E. coli inactivation efficiency in 1.0 L Yellow River water from upper reaches (Longyang Gorge) by four 3×3 cm² Ag_SA+NP/ZIF, corresponding to a total area of 36 cm². The line represents the time curve. i Disinfection efficiency of Ag_SA+NP/ZIF during recycling process. Experiments in b-i were conducted in triplicate, and the error bars represent the arithmetic mean ± standard deviation.

Firstly, we assessed the disinfection activity of different areas of Ag_SA+NP/ZIF against the stress-resistant E. coli in 0.1 L river water under 30 min of 33.0-35.0 mW cm⁻² solar irradiation (Fig. 4b). Four cm² Ag_SA+NP/ZIF achieved a 6.50 log₁₀ reduction of E. coli, implying almost complete inactivation of the bacteria. Under a dark condition, a mere 0.44 log₁₀ reduction was detected, indicating that the inactivation of bacteria was light-driven, thus ruling out adsorptive removal of bacteria by Ag_SA+NP/ZIF. The bactericidal activity of the 1 cm² and 2 cm² Ag_SA+NP/ZIF exhibited a reduction of at least 4 log₁₀ compared to the 4 cm² Ag_SA+NP/ZIF. This decrease can be attributed not only to the lower photocatalyst dosage but also to the inability of the smaller samples to raise the water temperature above 37 °C (Supplementary Fig. 8). In contrast, 4 cm² Ag_SA+NP/ZIF effectively increased the water temperature to over 41 °C (Fig. 3h), surpassing the optimal growth temperature for E. coli at 37 °C.

In addition, we found that the bactericidal efficiency of 1 cm² Ag_SA+NP/ZIF was enhanced by ~4.0 log₁₀ after incubating E. coli to the logarithmic phase with saline (Fig. 4c), specifically referring to E. coli cultured under laboratory conditions without any stress resistance. The saline solution was prepared using the same river water, thus the different bactericidal activity toward stress-resistant bacteria and laboratory-cultured bacteria was not caused by other factors in the river water. This result confirms the stress resistance of E. coli at late stationary phase and highlights the necessity and precision of employing resistant bacteria to study the bactericidal activity of catalysts in actual aqueous environments. After 30 min of light exposure, 4 cm² Ag_SA+NP/ZIF dissolved Ag⁺ and Zn²⁺ in 0.1 L river water, which were detected by an inductively coupled plasma mass spectrometry (ICP-MS) as 39.8 μg L⁻¹ and 207.2 μg L⁻¹, respectively. These released metal ions demonstrated limited inhibitory efficiency against E. coli (Fig. 4d).

The time profiles of E. coli inactivation revealed that 4 cm² Ag_SA+NP/ZIF exhibited superior bactericidal efficiency compared to Ag_SA/ZIF and Ag_NP/ZIF (Fig. 4e). The converted atomic inactivation efficiency of Ag_SA+NP/ZIF within 30 min was 5.69 log₁₀ per mg Ag (Supplementary Table 4), which was over 195 and 2630 times higher than those of Ag_SA/ZIF (3.40 log₁₀ per mg Ag) and Ag_NP/ZIF (2.27 log₁₀ per mg Ag), respectively. The bactericidal efficiency of Ag_SA+NP/ZIF correlates with the temperature rise curve (Supplementary Fig. 9). During the initial 15 min, as the water temperature gradually rises, the bactericidal efficiency increases slowly. When the temperature reaches 41.8 °C after 15 min, there is a marked increase in bactericidal efficiency. This result suggests that the photothermal effect of Ag nanoparticles significantly enhances the bactericidal activity. In addition, the disinfection efficiencies of 4 cm² Ag_SA+NP/ZIF against Gram-positive bacteria Staphylococcus aureus (S. aureus) and virus MS2 were 2.10 log₁₀ and 5.77 log₁₀ (Fig. 4f), respectively, thereby confirming the high efficiency of Ag_SA+NP/ZIF to disinfect a broader range of microorganisms.

To verify the high bactericidal activity of the catalyst under varying water quality conditions, we studied its effectiveness against stress-resistant E. coli cultured in water from the middle reaches of Yellow River (Toudaoguai Station) and lake water (Shichahai). Despite differences in water types (Supplementary Table 3), the 4 cm² Ag_SA+NP/ZIF achieved significant bacterial reductions of 4.35 log₁₀ in the middle reaches of Yellow River and 6.27 log₁₀ in lake water. The bactericidal activity in the middle reaches of the Yellow River was ~2 log₁₀ lower than that observed in the upper reaches (TOC: 5.49 mg L⁻¹), potentially attributed to higher concentrations of natural organic matter (TOC: 8.31 mg L⁻¹) in the middle reaches that effectively quench ROS. Additionally, the 4 cm² Ag_SA+NP/ZIF catalyst inactivated nearly 100% of the microorganisms in untreated real river water (Supplementary Fig. 10), demonstrating its potential for practical disinfection applications.

The application of Ag_SA+NP/ZIF was further extended to a large-scale water disinfection system in 1.0 L river water. As illustrated in Supplementary Fig. 11, four 3 × 3 cm² Ag_SA+NP/ZIF were employed to sterilize water with shaking. To consistently raise the water temperature above 41 °C, the system required a minimum catalyst area of 36 cm² and a velocity gradient of 70.7 s⁻¹ (Supplementary Fig. 12a). This velocity corresponds to an extremely low rotational speed of ~100 rpm, which is insufficient to cause catalyst turnover and thereby fully expose the upper catalyst layer to light irradiation. The water temperature rose to 41.5 °C (Supplementary Fig. 12b) and 6.60 log₁₀ E. coli inactivation was achieved within 1.0 h (Fig. 4h) under solar irradiation of 33.0-35.0 mW cm⁻².

Durability experiments showed that the bactericidal activity of Ag_SA+NP/ZIF barely decreased after disinfecting a cumulative volume of 10.0 L water (Fig. 4i). This demonstrates its potential to meet the WHO recommended daily drinking water allocation of 8.0 L per person. The total loss of Ag compared to the total Ag present in the entire photocatalyst is 5.57% detected by ICP-MS. AC-HAADF-STEM image shows that Ag single atoms remain atomically dispersed and Ag nanoparticles are not depleted after the durability experiments (Supplementary Fig. 13), confirming the stability of Ag_SA+NP/ZIF. In addition, the average concentration of Ag⁺ and Zn²⁺ dissolved by Ag_SA+NP/ZIF for a 1-h cycle was 40.2 μg L⁻¹ and 331.4 μg L⁻¹ (Supplementary Table 5), respectively. The Chinese standard for drinking water quality (GB 5749-2022) stipulates that the concentration of Ag and Zn in drinking water should not exceed 50.0 μg L⁻¹ and 1000.0 μg L⁻¹, respectively, which indicated that our detected levels are complying.

The price of Ag_SA+NP/ZIF-based water disinfection was also evaluated as shown in Supplementary Table 6. Four 3 × 3 cm² Ag_SA+NP/ZIF can disinfect at least 10.0 L water at a cost of $0.27, which is relatively higher compared to conventional treatment technologies such as chlorine dioxide disinfection, which costs $0.11. This higher cost arises from the synthesis process, which is conducted on a laboratory scale utilizing small packages of high-purity reagents (analytical reagent grade). However, the cost could be substantially reduced by scaling production to an industrial level, utilizing more cost-effective, technical-grade reagents. This adjustment could satisfy the economic and practical demands for disinfection applications.

Photothermal effect synergizes water disinfection

Previous studies have demonstrated that the photothermal action of plasmonic nanoparticle promotes protein aggregation in tumor cells and interferes their DNA damage repair capabilities²². However, the impacts of photothermal action on bacterial cells remain unclear. We hypothesize that the bacterial stress-resistant systems may potentially be compromised primarily due to the photothermal heating of Ag nanoparticles, resulting in increased bacterial susceptibility to ROS.

The transcriptomic analysis was employed to elucidate the impact of photothermal effect on the stress-resistant system of E. coli at the gene expression level. Compared to Ag_SA/ZIF treated E. coli, a total of 287 genes exhibited differential expression in E. coli treated with Ag_SA+NP/ZIF for 15 min ( | log₂ fold-change | ≥ 1, p < 0.05), with 121 genes being downregulated and 166 genes being up-regulated (Fig. 5a). Gene Ontology (GO) functional enrichment analysis results shown in Fig. 5b and Supplementary Table 7 indicate that the downregulated genes are enriched in response to hydrogen peroxide, response to reactive oxygen species, and DNA repair items (p-value < 0.05), with the rich factors of 0.16, 0.11 and 0.07, respectively. The expression of downregulated genes in Ag_SA/ZIF-treated E. coli was 2286 (Supplementary Fig. 14). However, no significant enrichment was found in the aforementioned three GO categories (Supplementary Table 8), which suggests that the photothermal effect of Ag_SA+NP/ZIF at 41.8 °C can decrease the bacterial resistance to ROS and DNA repair ability⁴⁵, with a higher proportion of downregulated genes responding to ROS (particularly hydrogen peroxide) compared to those involved in DNA damage repair. This result was further validated through disinfection experiments (Supplementary Fig. 15), wherein the 30-min inactivation of E. coli exhibited a >2.0 log increase when the temperature of the Ag_SA/ZIF system was elevated from 36 °C to 41 °C, approximating the temperature of the Ag_SA+NP/ZIF system. It is worth noting that the sole increase in temperature to 41 °C did not result in a significant impact on bacterial survival.

Fig. 5: Synergistic disinfection mechanism elucidation.

a Volcano plot of RNA sequencing data from E. coli culture exposed to Ag_SA+NP/ZIF for 15 min. b GO functional enrichment of the downregulated genes subset in E. coli exposure to Ag_SA+NP/ZIF for 15 min. Rich factor refers to the ratio of the number of differentially expressed genes to the total number of all annotated genes located in the same term. The dot color represents the p-value, and the dot size represents the number of differentially expressed genes in each term. c Intracellular ROS levels of E. coli exposed to monolithic photocatalysts for 15 min. d O₂^•− generation kinetics indicated by the absorbance of formazan dye that generated by the reaction of NBT with O₂^•−and H₂O₂ generation kinetics of Ag_SA+NP/ZIF. e •OH generation kinetics of Ag_SA+NP/ZIF and time profile of ABMDMA degradation by Ag_SA+NP/ZIF-generated ¹O₂. Reaction conditions: 4 cm² monolithic photocatalysts; 0.1 L Yellow River water from upper reaches (Longyang Gorge); 33.0-35.0 mW cm⁻² sunlight intensity. Experiments in d and e were conducted in triplicate, and the error bars represent the arithmetic mean ± standard deviation.

Given that the bacterial anti-stress system involves the removal of intracellular ROS, the fluorescent probe 2,7-dichlorodihydrofluorescein diacetate (DCFH-DA) was employed in conjunction with flow cytometry for semi-quantitative analysis of the total intracellular ROS accumulation. In the early phase of the reaction, the mean fluorescence intensity across the groups exhibited no significant variation, suggesting a similar accumulation of ROS (Supplementary Fig. 16). After 15 min (Fig. 5c), the mean fluorescence intensity in the Ag_SA+NP/ZIF-treated cells increased 33.6-fold when compared to that in the Ag_SA/ZIF-treated cells. Upon comparing the mean fluorescence intensity between groups following 30 min of reaction (Supplementary Fig. 16), the observed trend remained consistent. This phenomenon can be attributed to the photothermal effect of Ag_SA+NP/ZIF, which effectively down-regulates the cellular response to ROS and consequently hinders their removal⁴⁵.

The mean fluorescence intensity of Ag_SA/ZIF-treated cells was 1.7 times higher compared to that of Ag_NP/ZIF-treated cells (Fig. 5c). This suggests that a sole photothermal effect from Ag nanoparticles is not significant to elevate intracellular ROS levels in E. coli. The reaction system was subsequently treated with 20 μM melatonin to eliminate exogenous ROS, while low levels of endogenous ROS were detected in E. coli as indicated by low 2,7-dichlorofluorescein (DCF) intensity (Supplementary Fig. 17). Therefore, the accumulated intracellular ROS mainly originated from exogenous ROS generated by the photocatalyst. The absence of Ag single-atom sites in Ag_NP/ZIF leads to insufficient generation of exogenous ROS, whose transmembrane import is the main cause of the rise in intracellular ROS levels. Thus, Ag_SA+NP/ZIF, which possesses atomically dispersed Ag sites and plasmonic Ag nanoparticles, could respectively facilitate intracellular ROS generation and accumulation in cells, thereby synergistically inactivating stress-resistant E. coli cells.

To comprehend the disinfection mechanisms associated with extracellular ROS, nitro blue tetrazolium chloride (NBT, O₂^•−-specific indicator), coumarin (Cou, •OH-specific indicator), 9,10-anthracenediyl-bis(methylene)dimalonic acid (ABMDMA, ¹O₂-specific indicator), p-hydroxyphenylacetic acid (HPA, H₂O₂-specific indicator) were applied to monitor specific ROS generated by Ag_SA+NP/ZIF in deionize water. Under solar irradiation, the absorbance of formazan dye generated by the reaction of NBT with O₂^•− increases in a time-dependent manner (Fig. 5d). Similarly, the fluorescent products generated by the reaction of H₂O₂ with HPA were detected, and the concentration of H₂O₂ after 30 min reaction was 5.22 µmol L⁻¹ (Fig. 5d). In the disinfection system, H₂O₂ can be generated through both the oxygen reduction pathway (O₂ + 2H⁺ + 2e⁻ → H₂O₂) and the disproportionation reaction of O₂^•− (2O₂^•−+ H⁺ + H₂O → H₂O₂ + O₂ + OH⁻)⁴³. When investigating the production of •OH and ¹O₂, we found no significant increase in fluorescence intensity of Cou and decrease in absorbance of ABMDM (Fig. 5e), indicating the absence of detectable levels of •OH and ¹O₂. •OH generation is thermodynamically unfavorable as the valence band potential of Ag_SA+NP/ZIF (2.35 V vs. NHE) may not be sufficient to oxidize water (2.68 V vs. NHE)⁴³.

The generation of extracellular ROS was further validated using two independent tests: electron paramagnetic resonance (EPR) spectroscopy and scavenger quenching analysis. A characteristic 5-tert-butoxycarbonyl-5-methyl-1-pyrroline N-oxide (BMPO)-O₂^•− peak indicated the production of O₂^•− (Supplementary Fig. 18a), while no EPR signals were detected for •OH and ¹O₂ (Supplementary Fig. 18b, c). Scavenging H₂O₂ and O₂^•− significantly reduced E. coli inactivation efficiency (Supplementary Fig. 19). Therefore, O₂^•− and H₂O₂ are the primary ROS responsible for Ag_SA+NP/ZIF disinfection, with h⁺ playing a secondary role in E. coli inactivation. It is worth mentioning that O₂^•− and H₂O₂ both exhibit longer lifespans compared to •OH and elicit longer disinfection power^46,47. The direct contact between bacterial cells and photocatalyst is less important for our disinfection system, as H₂O₂ and O₂^•− can effectively diffuse into the bulk solution and inactivate bacterial cells that are not in contact with the photocatalyst⁴⁸.

TEM analysis showed no significant damage to the cell membrane structure of Ag_SA+NP/ZIF-treated bacteria compared to untreated bacteria (Supplementary Fig. 20a, b). Bacteria with intact membrane selectivity remain impermeable to propidium iodide (PI) staining but can be stained by SYTO9, while the bacteria with disrupted membrane selectivity are PI-permeable. Approximately 71.2% of bacteria exhibited disrupted membrane selectivity after 30-min treatment with Ag_SA+NP/ZIF (Supplementary Fig. 20c-e). The disruption interfered with normal cellular transport, as evidenced by the down-regulation of genes related to electron transport coupled proton transport, amino acid transmembrane transport, and anion transmembrane transport (Fig. 5b). Quantitative polymerase chain reaction (qPCR) analysis revealed a 1.27 log reduction in the 16s gene copy number in the Ag_SA+NP/ZIF-treated group after 30 min (Supplementary Fig. 21), indicating significant DNA damage. This result is consistent with transcriptome analysis (Fig. 5b) showing down-regulation of DNA repair genes. While direct structural damage to the cell membrane was minimal due to limited contact, Ag_SA+NP/ZIF effectively compromised membrane selectivity and induced intracellular DNA degradation through ROS generation.

Discussion

In this study, we successfully developed a floatable monolithic Ag_SA+NP/ZIF photocatalyst for water disinfection. Atomically dispersed Ag promotes charge transfer to facilitate the generation of H₂O₂ and O₂^•−. Transcriptomics analysis suggests that the photothermal effect of plasmonic Ag nanoparticles disrupted the anti-stress system of E. coli, thereby reducing their resistance to ROS. Under solar irradiation, the synergistic effect of Ag single atoms and nanoparticles enables 36 cm² Ag_SA+NP/ZIF to disinfect at least 10.0 L of surface water, which meets the WHO’s recommended daily per capita drinking water allocation. The present study demonstrates a potential approach to addressing the challenges of low photocatalytic disinfection efficiency in real-world water and the limited recyclability of traditional granular photocatalysts. It provides a foundation for sustainable and environmentally friendly POU water disinfection. However, the findings are based on the evaluation of a single photocatalyst under controlled laboratory conditions. To apply this approach in real-world settings, particularly in off-grid regions, requires overcoming several critical challenges. These include reducing material costs, improving scalability, and ensuring sustained photocatalytic performance under varying environmental conditions. Furthermore, overcoming practical obstacles such as efficient system integration and optimizing energy consumption is essential for the successful deployment and long-term viability of this technology in large-scale water treatment systems.

Methods

Monolithic ZIF-8-NH₂ support preparation

Commercially available sponges are mainly composed of melamine and have a porous structure with a density of 45.0 mg cm⁻³. The 4×2×1 cm³ three-dimensional melamine sponge was utilized as a substrate to absorb 10 mL 2.0 wt% aqueous sodium alginate solution (99%; Aladdin), followed by immersion in 30 mL 2.0 wt% zinc nitrate hexahydrate solution (99%; Aladdin) for 12 h to form gel-coated sponge. The gel-coated sponge was then placed in an oven and dried at 150 °C for 2 h to facilitate water evaporation and obtain a desiccated sponge. Then 0.279 g 2-methylimidazole (99%; Sigma Aldrich) and 0.08 g 2-aminobenzimidazole (97%; Aladdin) were placed at the bottom of the customized Teflon-lined stainless-steel autoclave, and a 1×1×0.2 cm³ dried sponge situated above the organic ligands⁴⁹. The dried sponge and organic ligand powder were kept separate. The autoclave was heated at 150 °C for ligand evaporation and deposition. After the reaction for 2 h, the obtained monolithic ZIF-8-NH₂ support was washed three times with deionized water and dried in a vacuum oven at 60 °C for 12 h. Larger size monolithic ZIF-8-NH₂ was prepared through proportionally increasing the amount of reaction reagents.

Monolithic photocatalyst preparation

The 1 × 1 × 0.2 cm³ monolithic ZIF-8-NH₂ sponge and 120 µL 67.9 mg mL⁻¹ silver nitrate solution (99.9%; Sigma Aldrich) were added to 10 mL of deionized water and left for 12 h to sufficiently adsorb Ag ions. The mixture was then snap-frozen in liquid nitrogen and the upper surface (1×1 cm²) of monolithic ZIF-8-NH₂ sponge was irradiated under UV lamp (Beijing Perfectlight, PCX-50C Discover) with a peak intensity at 365 nm for 10 min. The light intensity on the reaction solution surface was 52.8 mW cm⁻² as measured using the UV power meter (UV-A, Beijing Normal University Photoelectric Instrument Factory). The ambient temperature is maintained at 25 °C to allow the ice to melt during irradiation. The ice lattice inhibited nucleation⁵⁰, thus reducing Ag ions to single atoms. With the melting of ice, a portion of Ag was transformed into nanoparticles, leading to the formation of photocatalysts that incorporate both Ag single atoms and Ag nanoparticles. The obtained monolithic Ag_SA+NP/ZIF was washed three times with deionized water and dried in a vacuum oven at 60 °C. In addition, the synthesis of Ag_SA/ZIF was carried out under the same conditions except that the ambient temperature was kept at 0 °C. Ag_NP/ZIF was synthesized by the same process without liquid nitrogen freezing.

Material characterization

The microstructure and morphology of photocatalysts were obtained by SEM (FEI Inspect F50) and high-resolution TEM (FEI Talos F200S). Ag single-atom was observed by AC-HAADF-STEM (Titan Themis 60-300) operated at 200 kV, coupled with a probe spherical aberration corrector. Atomic dispersion and coordination environment of Ag species were confirmed by XAFS measurements, which were carried out on the sample at 11S2 X-ray absorption beamline of Aichi Synchrotron Radiation Center. The radiation was monochromatized by a Si (111) double-crystal monochromator. The measurements were carried out with a passivated implanted planar silicon fluorescence detector using Ag foil to calibrate the energy. XAFS data was analyzed by the Athena (version 0.9.26) and Artemis software (version 0.9.26) as shown in Supplementary Table 1 and Supplementary Note. The Ag loading content and dissolved Ag⁺ concentration of photocatalysts were measured by ICP-OES (Agilent 725-ES) and ICP-MS (Agilent 7700), respectively. The UV-Vis diffuse reflectance spectra of the photocatalysts were recorded on a PE Lambda750 spectrophotometer. The steady-state and time-resolved photoluminescence spectra were carried out for the characterization of electron-hole pairs separation using the Hitachi F-7000 spectrophotometer and Edinburgh FLS1000 spectrofluorometer with an excitation wavelength of 375 nm, respectively. The average lifetime was obtained by fitting the luminescence decay curves with a biexponential function. Electrochemical impedance spectroscopy and Mott–Schottky measurements were performed in 0.5 M Na₂SO₄ (99%; Aladdin) solution by using a CHI-660E electrochemical workstation (Shanghai, China). Photothermal properties of photocatalysts were characterized by a digital infrared thermal camera (FLIR E50).

Since Ag single atoms and Ag nanoparticles account for 100% of the silver content, the proportion of Ag nanoparticles can be deduced from the Ag single atoms proportion. The proportion of Ag single atoms can be estimated using the average Ag−N coordination number⁵¹. The Ag−N coordination number of atomically dispersed Ag sites in Ag_SA/ZIF catalyst is 3.2. Therefore, the average Ag−N coordination number in Ag_SA+NP/ZIF catalyst can be calculated by the following equation:

$$n=\frac{{N}_{{Ag}{-}N}}{{N}_{{Ag}}}=\frac{{N}_{Ag} * {X}_{1} * 3.2}{{N}_{Ag}}={X}_{1} * 3.2$$

(1)

where n represents the coordination number of Ag−N in Ag_SA+NP/ZIF catalyst, N_Ag−N represents the number of total Ag−N bond, N_Ag represents the number of total Ag atoms, and X₁ represents the proportion of Ag single atoms in Ag_SA+NP/ZIF catalyst. Therefore, Using the observed coordination number (Supplementary Table 1), the proportion of Ag single atoms in the Ag_SA+NP/ZIF catalyst can be approximately calculated as follow:

$${X}_{1}=n/3.2$$

(2)

Evaluation of the photocatalytic disinfection performance

River water and lake water with various water qualities, including the upper reaches of Yellow River (Longyang Gorge, Qinghai Province, China), the middle reaches of Yellow River (Toudaoguai Station, Inner Mongolia, China), and lake water (Shichahai, Beijing, China), were collected to test disinfection activity. The chemical parameters of the water samples were detected by the water quality analyzer (ProQuatro, YSI) as shown in Supplementary Table 3. Briefly, river water was filtered through cellulose membranes with the pore size of 0.45 μm to remove suspended sediment. The filtered river water was supplemented with 100 mg L⁻¹ of yeast extract and autoclaved to remove background microorganisms. This sterilized water samples were used to grow bacteria for disinfection experiments and RNA extraction.

E. coli BW25113 purchased from the Coli Genetic Stock Center (CGSC, Yale University, USA) was used as the model microorganism. 5 μL of E. coli strain were inoculated in 5 mL of Luria-Bertani (LB) medium at 37 °C for 12 h, then 5 mL of culture was transferred to 50 mL of LB medium and cultured for another 2 h. E. coli culture was collected and washed by centrifugation, and re-suspended in 50 mL sterilized river water to achieve an optical density of 0.06 at 670 nm ( ≈ 1×10⁸ bacteria mL⁻¹). E. coli cells was allowed to stand in the sterilized river water at 25 °C for 24 h to acclimatize to the new growth conditions and collected by centrifugation. Collected E. coli cells were then transferred to 50 mL fresh sterilized river water for 48 h incubation with shaking at 160 rpm at 25 °C to reach the late stationary phase (Supplementary Fig. 22). This growth phase better resembles bacterial cell conditions observed in oligotrophic surface water, where cellular responses are typically activated to counter starvation and other environmental stressors¹⁷. After 48 h of incubation, E. coli cultures were diluted 100-fold by fresh river water for disinfection testing. The monolithic photocatalyst with surface area of 1-4 cm² was plunged into 0.1 L E. coli suspension and irradiated without any agitation under sunlight (14:00-15:00 pm in September in Beijing, China) with a light intensity of 33.0-35.0 mW cm⁻². For large-scale water disinfection, 1.0 L E. coli suspension was placed in a 2.0 L beaker and shaken on a shaker at a constant velocity. The Camp-Stein velocity gradient (G) for a shaker was calculated by the following Eq. (3):

$$G=\sqrt {\frac{P}{{pv}{{{\rm{\forall }}}}}}$$

(3)

where P represents the power input into the control volume (0-14 W), p represents the density of the fluid (997 kg m⁻³), v represents the volume of the fluid (0.001 m³), and ∀ represents volume of the control volume (0.002 m³).

Five mL of the sample was carefully pipetted out at the sampling time and diluted 10, 100, and 1000 times with sterile normal saline, respectively. 100 μL of the diluted E. coli samples were spread evenly on LB agar plates and incubated at 37 °C overnight for colony counting. The concentration of E. coli was calculated by the formula: C = n × dilution times × 10, where C represented the concentration of E. coli in the sample (CFU mL⁻¹) and n represented the number of colonies. The concentration of E. coli without photocatalytic treatment was represented by C₀, thus the bactericidal efficiency was expressed as Log₁₀(C₀/C). Each experiment was performed three times independently.

Transcriptome analysis

To analyze the impact of photothermal stress on biological functions in E. coli, transcriptomics analysis was conducted by extracting RNA from E. coli cells exposed to Ag_SA/ZIF and Ag_SA+NP/ZIF for 15 min, as well as a control group. Total RNA extraction was performed using the Total RNA Extractor kit (Sangon Biotech, Shanghai) for subsequent RNA-seq analysis. The quality and quantity of total isolated RNA in the samples were analyzed by a Qubit 2.0 fluorometer using Qubit RNA Quantification Kit (Invitrogen). The rRNA depletion was carried out using Ribo-off rRNA Depletion Kit (Vazyme, Nanjing) according to manufacturer’s specifications. The rRNA-deleted samples were further purified to obtain purified mRNA, and the RNA libraries of purified mRNA were constructed with VAHTS Stranded mRNA-seq V3 Library Prep Kit for Illumina (Vazyme, Nanjing). The quality of the RNA libraries was checked by 8% polyacrylamide gel electrophoresis. After that, all libraries were sequenced using an Illumina HiSeq 3000 system with paired-end 150 base-pair reads. Significantly differentially expressed genes were analyzed by DESeq2 R-package (version 1. 12. 4) with a screening condition of p-value ≤ 0.05 and |log₂ fold-change | ≥1. GO enrichment analysis was performed by topGO R-package (version 2. 24. 0) to evaluate the biological function enrichment corresponding to the differentially expressed genes sets. In general, GO terms with an adjusted p-value < 0.05 were considered significantly enriched.

ROS measurements

When bacterial cells are exposure to monolithic photocatalyst, ROS may penetrate the cell membrane causing intracellular ROS accumulation, which was detected by using the fluorescent probe DCFH-DA (99.81%; MCE) in conjunction with flow cytometer (ACEA NoveCyte^TM). DCFH-DA can freely penetrate cellular membranes and undergo intracellular hydrolysis by esterase, resulting in the formation of 2,7-dichlorodihydrofluorescein (DCFH) that becomes sequestered within the cytomembrane. Subsequently, intracellular ROS oxidizes DCFH to generate fluorescent DCF, exhibiting maximum excitation and emission wavelengths at 488 nm and 525 nm, respectively. Briefly, E. coli treated for 0-30 min in the photocatalytic disinfection system were incubated with DCFH-DA at a concentration of 25 μM for 30 min at 37 °C. The bacterial cells after incubation were collected and washed with saline (85% NaCl solution) to remove excess fluorescent probe. The flow cytometer laser was selected to be 488 nm, and the number of the collected cells was 100,000. The sample collection speed was adjusted to the low mode (14 μL min⁻¹), and the threshold of FSC-H was set to 1000. The mean intracellular fluorescence intensity of bacteria was analyzed by FlowJo V10 software.

For H₂O₂ concentration detection, a working solution was prepared by adding 2.7 mg of p-hydroxyphenylacetic acid (98%; Aladdin) and 1 mg of horseradish peroxidase (200 U mg⁻¹; Aladdin) to 10 mL of potassium hydrogen phthalate buffer solution (8.2 g L⁻¹). Subsequently, 50 μL of working solution was added to 1 mL sample, diluted with 1 mL deionized water, and the reaction was quenched by adding 1 mL 0.1 M NaOH (99%; Aladdin) after 10 min. The fluorescence spectra of samples were measured with an excitation wavelength of 315 nm and detected at an emission wavelength of 409 nm. For the detection of O₂^•−, 30 µL of 1 mM NBT (98%; Aladdin) dissolved in dimethyl sulfoxide (99.8%; Aladdin) and 3 mL of sample were mixed and incubated for 10 min. The reaction generated formazan dye with maximum absorbance at 680 nm was detected by UV-Vis spectrophotometer. The concentration of •OH generation was determined by the fluorescence method using coumarin as a probe⁴³. One mM coumarin (99%; Aladdin) and Ag_SA+NP/ZIF were mixed in 0.1 L water. Coumarin reacts with •OH to produce 7-hydroxycoumarin, which can emit fluorescence at 451 nm with excitation wavelength of 332 nm. The presence of ¹O₂ was detected by using ABMDMA (90%; MCE) as a sensitive indicator. The bleaching of ABMDMA by singlet oxygen resulted in a decrease in its absorbance at 400 nm. 0.5 mL ABMDMA (2 mM) and Ag_SA+NP/ZIF were mixed in 0.1 L water. The consumption of ABMDMA was monitored by recording the absorbance at 400 nm with a UV-Vis spectrophotometer. The sunlight intensity was 33.0-35.0 mW cm⁻² and Ag_SA+NP/ZIF size is 4 cm².

ROS were detected by EPR spectroscopy on a Bruker A300 spectrometer operating at an X-band frequency of 9.85 GHz. The monolithic catalyst was crushed and prepared as a suspension sample (100 mg L⁻¹). Ten µL of 50 mg mL⁻¹ BMPO (99%; Aladdin) and 10 µL of water were mixed with 30 µL of catalyst sample. The resulting liquid was aspirated using a capillary tube, followed by light exposure for 10 min to detect both •OH and O₂^•−. For the detection of •OH alone, the procedure was modified by substituting 10 µL of water with 10 µL of superoxide dismutase (SOD, 5 U mL⁻¹; Aladdin) to quench O₂^•−. To detect ¹O₂, 25 µL of 10 mM 2,2,6,6-tetramethylpiperidine (TEMP, 99%; Aladdin) was mixed with 25 µL of catalyst sample, and measurements were conducted under both 10-min light exposure and in dark conditions.

The scavengers used for •OH elimination, H₂O₂ neutralization, O₂^•− scavenging, hole scavenger, electron capture, and ¹O₂ quenching were 2.5 mM isopropanol (99.5%; Sigma), catalase (CAT, 300 U mL⁻¹; Sigma), SOD (400 U mL⁻¹; Sigma), 2.5 mM sodium oxalate (99.5%; Sigma), 2.5 mM sodium chromate (Cr(VI)), (99.5%; Sigma), and 2.5 mM furfuryl alcohol (99%; Sigma), respectively. The scavengers applied in this study showed no significant impact on bacterial activity (Supplementary Fig. 23). These scavengers were added to the bacterial suspension prior to illumination. Bacterial concentrations in solution were determined at different time intervals using standard spread plating method. Each sample was serially diluted and plated in triplicate onto nutrient agar before incubation at 37 °C for 24 h.

Membrane damage and DNA degradation

Cell membrane structure damage was analyzed by TEM on sliced bacterial specimen. Briefly, the harvested bacterial cells were first fixed in a solution of 4% paraformaldehyde (95%; Aladdin) and 2.5% glutaraldehyde (50%; Aladdin) for 12 h at 4 °C. The samples were then infiltrated and embedded in Spurr’s resin with propylene oxide (99%; Aladdin) and cured overnight at 70 °C. Subsequently, ultrathin sections were prepared using an ultramicrotome (MT-X, RMC) and examined with a Hitachi H7700 TEM operated at 75 kV. The collected samples were stained with LIVE/DEAD BacLight Viability Kit (Thermo Fisher scientific) and analyzed with flow cytometry (ACEA NoveCyte^TM). This kit contains two fluorescent nucleic acids dyes, namely SYTO 9 and PI. E. coli cells with intact membrane can only be stained green by SYTO 9 (Ex. 488 nm and Em. 530 nm), and E. coli cells with disrupted membrane selectivity can be stained red by PI (Ex. 488 nm and Em. 630 nm). The defined gates include: complete membrane destruction (Q1), disrupted membrane selectivity (Q2), impurities (Q3), and intact membrane (Q4). The proportion of bacteria (P) with disrupted membrane selectivity was determined using Eq. (4):

$${{\rm{P}}}=\frac{{{\rm{Q}}}1+{{\rm{Q}}}2}{{{\rm{Q}}}1+{{\rm{Q}}}2+{{\rm{Q}}}4}$$

(4)

The qPCR amplicon targeting the 16S gene was employed for highly sensitive detection of DNA damage⁵², and the primer sequence can be found in Supplementary Table 9. The qPCR reaction parameters were as follows: pre-incubation at 95 °C for 3 min; amplification of 35 cycles with each cycle being 95 °C for 30 s, 60 °C for 30 s, and 72 °C for 45 s; melt curve generation at 90 °C for 10 s, and 60 °C for 60 s; and a gradual heating to 90 °C at a rate of 0.03 °C s⁻¹.

Electromagnetic simulation

To explain the enhancement local electromagnetic field near the surface between the Ag nanoparticles and the ZIF-8-NH₂, we used COMSOL Multiphysics (version 6.2), employing the Electromagnetic Waves, Frequency Domain physics interface and Wavelength Domain in the optics module. Briefly, the refractive indices of Ag nanoparticles and air were obtained from the COMSOL built-in material library, while the refractive index of ZIF-8-NH₂ substrate was set to 1.38. A series of parallel simulations is designed to show the effect of varying ZIF-8-NH₂ thickness and Ag nanoparticle size on the electromagnetic field at the interface between Ag nanoparticle and ZIF-8-NH₂ under 450 nm irradiation. The variations include ZIF-8-NH₂ thicknesses ranging from 1 to 15 nm and Ag nanoparticle radius of 1, 3.5, 5, 6.5, 8 nm. The geometric model consisted of one Ag spherical particle on the center of a ZIF-8-NH₂ substrate as shown in the Supplementary Fig. 24. Incident light with a wavelength of 450 nm was directed vertically downward pointing the x-y plane. The air domain is confined by the inclusion of a Perfect Matching Layer (PML), which restricts the model to a specific area of interest to conserve computing resources while maintaining accuracy consistent with results from a large, unrestricted solution domain⁵³. The solution within this domain remains unaffected by the introduced PML. The Perfect Electrical Conductor (PEC) boundary conditions were used on the x-y symmetry plane. A Frequency Domain was defined as the initial incident electric field E₀, serving as the background for the scattered field. The distribution of |E | / | E₀| was computed to analyze the electromagnetic field enhancement at the interface of Ag nanoparticles loaded on ZIF-8-NH₂. The COMSOL simulation parameters are provided in Supplementary Method and Supplementary Table 10.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.