Metal-organic frameworks with photocatalytic bactericidal activity for integrated air cleaning

Air filtration has become an essential need for passive pollution control. However, most of the commercial air purifiers rely on dense fibrous filters, which have good particulate matter (PM) removal capability but poor biocidal effect. Here we present the photocatalytic bactericidal properties of a series of metal-organic frameworks (MOFs) and their potentials in air pollution control and personal protection. Specifically, a zinc-imidazolate MOF (ZIF-8) exhibits almost complete inactivation of Escherichia coli (E. coli) (>99.9999% inactivation efficiency) in saline within 2 h of simulated solar irradiation. Mechanistic studies indicate that photoelectrons trapped at Zn+ centers within ZIF-8 via ligand to metal charge transfer (LMCT) are responsible for oxygen-reduction related reactive oxygen species (ROS) production, which is the dominant disinfection mechanism. Air filters fabricated from ZIF-8 show remarkable performance for integrated pollution control, with >99.99% photocatalytic killing efficiency against airborne bacteria in 30 min and 97% PM removal. This work may shed light on designing new porous solids with photocatalytic antibiotic capability for public health protection.

Overall, the manuscript was very interesting to read and the authors made a nice efforts to understand and characterize the underlying mechanism for photo catalytic properties of ZIF-8. However, there are different examples in the literature on the photocatalytic and antibacterial properties of MOF and particularly ZIF-8, which present different conclusions. In my opinion these papers should be mentioned and the different outcome discussed: 1. The authors claim as, "For the first time, the photocatalytic activity of ZIF-8 was discovered and characterized", I would suggest to rephrase this statement.
2. Biocidal activity of ZIF-8 is recently reported "Self-Cleaning and Antibacterial Zeolitic Imidazolate Framework Coatings" ( Adv. Mater. Interfaces 2018, 5, 1800167), where the authors conclude that the antibacterial activity stems from 2-methylimidazole (H-MeIM) and zinc (Zn2+) ions that were released from ZIF-8 particles. On contrary, in current study, Zn2+and H-MeIM had little antibacterial effect (Fig. 2c). It would be interesting if the authors could explanation their contradictory observation with reference to the given example or any other if they find in the literature. Alternatively the authors should clearly state the different outcome and which experiments confirm their hypothesis.
3. For ZIF-8, the estimated bang gap of 3.3 eV (this study) is different from previously published value of 4.9 eV (Angew. Chem. Int. ed. 2011, 50, 450). It would be helpful if the authors explain their different finding compared to the literature values. A clear input would help to avoid discrepancies in the literature or misunderstanding of the values provided. 4. The core message of this study reports reactive oxygen species mainly contribute to the antibacterial properties rather than the Zn ions, supported and evidenced principally by EPR studies. With respect to the estimated band gap, statement of generation of hydroxyl radical is thermodynamically unfavorable (as also supported via scavenger experiments). This is confusing as another report (RSC Adv., 2014, 4, 54454) describes ZIF-8 photocatalysts exhibiting efficient photocatalytic activity for methylene blue degradation under UV irradiation, which was confirmed through the detection of hydroxyl radicals (.OH) by a fluorescence method.
In my opinion the article is very interesting to the readers of Nature Communications and the study is and conclusions are well supported by experiments. However, in order to understand and correlate the new findings, appropriate and directly related citations are missing (few examples are discussed above), which should be added and discussed.

Responses to referees' comments
Referee 1:

With regard to the comment "UV sterilization is an efficient method in providing antibacterial
functions. On page 2 the authors stated UV…has 'low antibacterial efficiency". Please provide evidence." UV sterilization is indeed an efficient method, especially UVC (200-280 nm) germicidal irradiation from artificial light source. However, in ground-level solar spectrum, UV irradiation (mostly UVA and UVB) only accounts for 4% of the total energy. Under sunlight, sterilization is usually inefficient and time-consuming. In this work we introduce a highly efficient solar-assisted catalytic disinfection as a viable alternative to other disinfection methods. Now we have clarified the above points in the revised manuscript (Page 2, Line 26-28).

Concerning the comment "What are the size information and characterizations of ZIF-8 particles?
When zinc acetate dihydrate and 2-methylimidazole were mixed in PEG and applied onto fabrics, any characterization of the produced particles in the solution and on surfaces of the fibers?" In our original paper, ZIF-8 particles have been characterized by PXRD, UV-vis diffuse reflectance spectra and EPR (Fig. S6, S10 and 3b). The PXRD results confirmed that highly crystalized and phase pure ZIF-8 particles have been prepared. Per request, we now have thoroughly characterized all samples using FT-IR, TG, TEM, SEM, elemental mapping and BET adsorption in the revised supplementary information (Fig. S2). SEM and TEM images reveal that the as-obtained ZIF-8 nanocrystals show typical rhombic dodecahedron morphology. The particle size of ZIF-8 ranges from 70 to 110 nm, and an average diameter of 92 nm was obtained by statistical evaluation of 100 objects.
ZIF-8 coating layers were fabricated in-situ on the non-woven fabrics. They have been characterized by using PXRD (Fig. 4a), SEM and elemental mapping (Fig. S20). The PXRD pattern of MOFilter confirmed the successful formation of a dense layer of ZIF-8 on the substrates, which was further evidenced by SEM and elemental mapping. ZIF-8 nanocrystals grown on the fibers also show well-defined rhombic dodecahedral shape. An average size of 80 nm for ZIF-8 particles on fibers was obtained by statistical evaluation of 100 particles. Accordingly, we have included all the above results and related discussions in the revised manuscript (Page 8, Line 15-17) and the supplementary information (Page 4, Line 1-10; Page5-6, Line 27-30, 1-2 and Fig. S2 and S20).

As to the question "No details on nonwoven fabric used in this work. Which polymer, fiber size, made by which process? Density, porosity…?"
Non-woven fabric employed in this work is the spunlaced non-woven fabric with polyethylene (PE) and polyethylene terephthalate (PET) (1:2). It has a fiber diameter of ca. 12 µm, thickness of 0.190 mm, density of 1.4 g cm -3 , maximum pore size of 116.6 µm, average pore size of 56.1 µm and porosity of 69.7%.
Per request, we have included the above detailed information in the revised supplementary information (Page 3, Line 8-12).
4. As for the comments "… a concern on falling-off of the particles from the surfaces, especially under frictions. Please provide some durability testing results." Per request, we have studied the durability of MOFilter via bending and rubbing test. Specifically, 10 pieces of MOFilter were continuously bended and rubbed for 1000 times, and after that the changes of samples were carefully measured and recorded. As shown in Table S2, there were no significant changes in MOFilter weight, indicating the strong affinity of ZIF-8 nanocrystals to the substrate.
Moreover, no obvious changes were observed for surface morphology of fibers after the abovementioned tests (Fig. S21). The high robustness of ZIF-8 coatings had also been verified by abrasion resistance tests and mechanical stirring experiment in our previous work (Angew. Chem. Int. Ed., 2016, 55, 3419 -3423).
To further clarify this point, we have now added the above results and discussions in the revised manuscript (Page 8, Line 18-19) and the supplementary information (Table S2 and Fig. S21).

5.
Regarding the suggestion "ZIF-8 needs 120 min to provide 99.9999% reduction of E. coli in solution containing 5mg/10mL, pretty high concentration." Per suggestion, we also achieved 99.999% reduction of E. coli in solution by lowering the catalyst dosage to 300 mg L -1 (Fig. S3). More importantly, here we do need to use sufficient amount of ZIF-8 to achieve high PM aerosol capture efficiency while maintaining the excellent surface bactericidal performance (Fig. 4d, 4e and 4f). Comparing with other filters with inorganic sorbents (eg. PP/PET non-woven with activated carbon), MOFilter has a drastically lower MOF loading (0.15 mg cm -2 ). It is noteworthy that the volumetric concentration of the photocatalyst in aqueous solution is more relevant in water purification. In this work, we mainly focus on the application of these bifunctional MOFilters with impressive self-cleaning and anti-biofouling effects for integrated air cleaning.
We have now included the above results and discussions in the revised manuscript (Page 5, Line 7-9) and the supplementary information (Fig. S3). During the dark period bacteria reduction was inhibited obviously since no more H 2 O 2 was produced in such condition. We have now included the above results in the revised manuscript (Page 7, Line 24-25) and the supplementary information (Fig. S19).

As for the comments "The UV lamp of 300W Xe shines a light in intensity similar to direct sunlight.
It could already kill a significant amount of E. coli." In this work, a 300 W Xe lamp coupled with a AM 1.5 filter (300 nm < λ < 1100 nm) was employed as light source with optical power density fixed at 100 mW cm -2 (1 sun) to simulate sunlight irradiation. Under this light illumination, a certain amount of E. coli can indeed be killed without photocatalyst as seen in Fig. 2b. However, only less than 77% inactivation efficiency (0.6, -log 10 (C/C 0 ) efficiency) was achieved within 120 min. In contrast, > 99.9999% efficiency (6.1, -log 10 (C/C 0 ) efficiency) was obtained with the introduction of ZIF-8 under identical sunlight irradiation.
To further clarify this point, we have presented a detailed discussion in the revised manuscript (Page 5, Line 1-2).

In response to the question "In Fig 2d, what are the sizes of TiO 2 and ZnO particles? Here, what is the size of ZIF-8 again?"
Per request, we have included the SEM and TEM images of TiO 2 , ZnO and ZIF-8 along with the size distribution of these particles in the revised supplementary information (Fig. S8). A statistical evaluation shows that TiO 2 , ZnO and ZIF-8 particles are about 15-35 nm, 20-100 nm and 70-110 nm in diameter respectively, with an average size of about 25 nm, 50 nm and 92 nm, respectively.

Regarding the suggestion "How to assign the broad absorption at 350nm of ZIF-8 as LMCT? I couldn't find any related information from the reference #46."
ZIF-8 shows strong absorption at 227 nm and a broad absorption at longer wavelength of 350 nm (Fig. S10). The strong absorption centered at 227 nm could be attributed to intraligand charge transfer, and shows a slight red shift relative to 2-methylimidazole (H-MeIM). After illuminating ZIF-8 with 300 nm < λ < 1100 nm light, an obvious signal emerged in EPR spectra (Fig. 3b) due to the generation of trapped electrons on Zn + sites via LMCT process. This indicates that the broad absorption at 350 nm is responsible for such electron transfer to the metal center. As previously reported for MOF-5 (J. Phys. Chem. B., 2006, 110, 13759-13768;Chem. Commun., 2004, 20, 2300-2301, the presence of structural Zn 4 O 13 clusters was responsible for ligand to metal charge transfer centered at 350 nm. We have included the above discussions and related figures in the revised manuscript (Page 6, Line 26-31) and the supplementary information (Fig. S10).

Concerning the comment "Based on XRD, ZIF-8 is a crystal, but Tauc plot is often used for amorphous materials not for crystal."
Indeed, Tauc plot has been used for the evaluation of the band gap of some crystalline materials (J. Am. Chem. Soc., 2016, 138, 13822−13825;Nature Commun., 2018, 9, 1660-1668. Per suggestion, we have also calculated the optical band gap of ZIF-8 by using simple equation (band gap = 1240 / wavelength). A value of 3.3 eV was obtained, which is the same as the result derived from Tauc plot.

As for the comments "Is Zn 2+ or ZIF-8 serve as photosensitizer?"
ZIF-8 with optical absorption centered at 350 nm due to LMCT serves as the photosensitizer.
Zn 2+ with d 10 electronic configurations has no obvious absorption in the UV and visible region, and could not harvest light.

As to the inquiry "If H 2 O 2 is produced in the system, antibacterial tests should use a quenching chemical to remove it during tests."
We did conduct quenching experiment to determine the bactericidal contribution for each reactive oxygen species (ROS) (Fig. 3e and Fig. S18). The scavengers used was sodium chromate (Cr(VI)) for  S7 and Table S1).
In the reference (Adv. Mater. Interfaces., 2018, 5,1800167), the ZIF-8 nanocomposite exhibited antibacterial activity against E. coli due to the releasing of Zn 2+ and H-MeIM. This is actually consistent with the results and discussion as mentioned above. Unfortunately, the authors did not offer inactivation efficiency and the detailed experimental data, such as the released Zn 2+ and H-MeIM concentrations, amount of ZIF-8, initial concentration of E. coli cells and dilution ratio of the sample.
It is thus difficult for us to make direct comparison.
Remarkably, the incorporation of ZIF-8 with light irradiation could decrease the count of bacterial colonies by six orders of magnitude, due to the high effectiveness of ROS disinfection mechanism.
For clarity, we have added the above discussions and results in the revised manuscript (Page 5, Line 16-22) and the supplementary information (Page 7, Line 1-12, Fig. S7 and Table S1).

Concerning the comment "For ZIF-8, the estimated bang gap of 3.3 eV (this study) is different
from previously published value of 4.9 eV (Angew. Chem. Int. ed., 2011, 50, 450 According to the equation for optical band gap calculations (band gap = 1240 / wavelength), the band gap is closely related to the spectra location of absorption edge. As mentioned in response 9 to referee 1, there are two apparent absorption peaks at 227 nm and 350 nm in the UV-vis diffuse reflectance spectra. If a wavelength edge of 250 nm was employed, a value of 4.9 eV would be obtained based on the above the equation. In our work, the absorption edge should be at 375 nm due to the LMCT transition at 350 nm for effective photocatalysis. Therefore under simulated sunlight (300 nm < λ < 1100 nm), a band gap of 3.3 eV is more relevant, as also evidenced by the results from others (J. Alloy. Compd., 2017, 693, 543).

4.
With regard to the suggestion "… generation of hydroxyl radical is thermodynamically unfavorable (as also supported via scavenger experiments). This is confusing as another report (RSC Adv., 2014, 4, 54454) Fig. S15. Time-dependent fluorescence spectra of (a) generated 7-hydroxycoumarin in ZIF-8 reaction system for •OH detection. 7-hydroxycoumarin could emit fluorescence at 455 nm when excited at 332 nm. (b) generated 2-hydroxyterephthalic acid in ZIF-8 reaction system containing 2 mM NaOH for •OH detection. 2-hydroxyterephthalic acid could emit fluorescence at 425 nm when excited at 315 nm.