Starting reagents, solvents, and materials were commercially available and at least of analytical grade. The H₂L ligands were purchased from Jinan Henghua Sci. and Tec. Co. Ltd. PXRD patterns at diffraction angles from 5° to 55° were obtained with a D/MAX-3D diffractometer. Elemental analysis was performed by Perkin–Elmer Elementarvario elemental analysis instrument. Fourier transform infrared spectra (FT-IR) were recorded on Nicolet iS50 (4,000–400 cm⁻¹). Thermogravimetric analysis (TGA) was performed on a SDT 2960 thermal analyzer from room temperature to 800°C at a heating rate of 10°C/min under nitrogen flow. Ultraviolet-visible (UV-vis) absorption spectra were obtained by UV-2600 UV-vis spectrophotometer. Fluorescence detection was carried out on CARY Eclipse Fluorescence Spectrophotometer at room temperature. Energy dispersive spectrometer (EDS) was obtained by JSM-6490LV (JEOL Ltd., Japan) electron microscope.

Single-crystal X-ray diffraction data of Zn-MOF were collected by Oxford Diffraction SuperNova area-detector diffractometer with the program of CrysAlisPro. The crystal structure was solved by SHELXS-2016 and SHELXL-2016 software (Sheldrick 2008). The crystallographic data and structure refinements are shown in Table 1. The CIF file of Zn-MOF (CCDC No. 2143524) can be downloaded free of charge via http://www.ccdc.cam.ac.uk/conts/retrieving.html. Selected bond lengths and bond angles for the Zn-MOF complex are shown in Supplementary Table S1.

Powder samples (120 mg) were dispersed in vials containing 10 ml of aqueous solution, respectively. The vials were kept at room temperature for 1–7 days, and 30 mg samples were taken out at 1, 3, and 7 days, respectively, for PXRD tests. The results demonstrated that the doping of Cu ions improved the water stability of the Zn-MOF. In order to investigate the stability, Cu_0.1/Zn-MOF powder was dispersed in a vial containing 3 ml aqueous solutions with different pH scales (pH = 1–14) for 1 h, and then the PXRD test was carried out. In addition, in order to explore the fluorescence performance of Cu²⁺ doped Zn-MOF at different pH values, the synthesized sample Cu_0.1/Zn-MOF (2.0 mg) was added into aqueous solutions (3.0 ml) with different pH values, and ultrasonicated for 20 min; then the fluorescence spectra were immediately determined.

For the detection of metal ions, 2.0 mg samples were finely ground and added into 3.0 ml of deionized water of M(NO₃)_x (1 mM, M = Na⁺, Ca²⁺, K⁺, Li⁺, Zn²⁺, Mg²⁺, Co²⁺, Mn²⁺, Cd²⁺, Ni²⁺, Ag⁺, Pb²⁺, Cu²⁺, Al³⁺, Cr³⁺, and Fe³⁺). For the detection of antibiotics, 2.0 mg samples were finely ground and added into 3.0 ml of varied selected antibiotics solutions (100 ppm), including tetracyclines (tetracycline TC; nystatin NS), sulfonamides (sulfamethoxydiazine SM), chloramphenicols (chloramphenicol CHL), aminoglycosides (gentamicin GEN; kanamycin KAN; hygromycin HYG; streptomycin sulfate STR; spectinomycin SH), nitrofurans (nitrofurazone NFZ; nitrofurantoin NFT; furazolidone FZD), and β-lactams (cefotaxime CEF). In order to maintain its homogeneity, the mixtures were ultrasonicated for 20 min to form a suspension. The luminescence data of the suspension were monitored under the same conditions.

A mixture of Zn(NO₃)₂·6H₂O (29.8 mg, 0.1 mmol), H₂L (14.1 mg, 0.05 mmol), CH₃CN/H₂O (2 ml/3 ml), and one drop of dilute HNO₃ (1 M) was added to a 10 ml Pyrex vial and stirred for 20 min. Then it was transferred to a Teflon-lined stainless steel vessel and reacted at 120°C for 72 h. After being cooled to room temperature at a rate of 5°C min⁻¹, the light yellow block crystals were obtained by filtration and collection, washed with CH₃CN, and then air-dried naturally with a yield of 83.8% (based on H₂L. Anal. Calc. (%) for C₁₇H₁₁N₃O₄Zn (Mr = 386.68): C 52.80 H 2.87, N 10.87; found (%): C 52.85, H 2.91, N 10.82. IR (KBr pellet, cm⁻¹): 3,468 (w), 3,105 (m), 3,074 (w), 2,246 (w), 1837 (w), 1786 (w), 1,654 (s), 1,589 (s), 1,511 (s), 1,458 (s), 1,426 (s), 1,388 (s), 1,310 (s), 1,239 (s), 1,179 (m), 1,112 (m), 1,043 (w), 1,004 (w), 921 (s), 850 (w), 784 (s), 750 (s), 718 (s), 689 (m), 648 (w), 531 (m), 456 (s).

Synthesis of Cu _x /Zn-MOF (x = 0.01, 0.1, 0.2, 0.5): A mixture of Zn(NO₃)₂·6H₂O (29.5 mg, 0.099 mmol; 26.8 mg, 0.09 mmol; 23.8 mg, 0.08 mmol; 14.9 mg, 0.05 mmol), Cu(NO₃)₂·3H₂O (0.242 mg, 0.001 mmol; 2.42 mg, 0.01 mmol; 4.83 mg, 0.02mmol; 12.1 mg, 0.05 mmol), H₂L (14.1 mg, 0.05 mmol), CH₃CN/H₂O (2 ml/3 ml), and two drops of HNO₃ (62%, aq.) was added to a 10 ml Pyrex vial and stirred for 20 min, and then transferred to a Teflon-lined stainless steel vessel and reacted at 120°C for 72 h. After being cooled to room temperature at a rate of 5°C min⁻¹, the light blue block crystals were obtained by filtration and collection, washed with CH₃CN, and then air-dried naturally with a yield of 81.2, 82.9, 81.1, and 82.7%, respectively (based on H₂L).

Single-crystal X-ray analysis reveals that {[Zn(L)]·CH₃CN} _n crystallizes in the monoclinic system, P2₁/c space group. In Zn-MOF, the asymmetric unit contains one Zn(II) atom, one fully deprotonated L^2- ligand, and one free acetonitrile molecule (Figure 1A). Zn(II) atom is in the five-coordination mode with four oxygen atoms from four carboxylic acid ligands and one nitrogen atom from the benzimidazolium. Zn–O bond lengths range from 2.027(2) to 2.035(2) Å, and Zn–N bond length is 2.021(2) Å. Adjacent two zinc atoms were bridged by four carboxyl groups to form a bimetallic unit. Each bimetallic unit was connected with four nodes in four directions by different isophthalate groups to generate an infinite 2D layer (Figure 1B). Adjacent layers were joined by nitrogen atoms (N1) of imidazole groups to give rise to a three-dimensional structure (Figure 1C). From a topological viewpoint, each dinuclear Zn(II) cluster connects six L^2- ligands, and each L^2- ligand links three dinuclear Zn(II) clusters. As illustrated in Figure 1D, Zn-MOF could be simplified as a 3,6-c network with the Schläfli symbol {4·6²}₂{4²·6¹⁰·8³} with TOPOS (Blatov et al., 2014).

The chemical compositions of Zn-MOF and Cu_x/Zn-MOF were determined by EDS (Supplementary Figures S1–S5), as displayed in Supplementary Table S2. The results showed that the Cu/Zn ratios were obviously higher than the synthesis ratio. This phenomenon may be due to stronger bonds formed between Cu(II) ions and H₂L ligands (Niu et al., 2014). The SEM-mapping of the Cu_0.1/Zn-MOF shown in Supplementary Figure S6 further indicates the existence of copper and the uniform distribution of copper and zinc in the skeleton structure.

The FTIR spectra of Zn-MOF and Cu_x/Zn-MOF (x = 0.01, 0.1, 0.2, 0.5) were carried out and shown in Supplementary Figure S7. The C=O stretching vibration of the free ligand H₂L is 1720 cm⁻¹, while it is not visible in the five complexes, which illustrates the coordination of carboxylic acids with metals. The vibration band at 1,589 cm⁻¹ and 1,389 cm⁻¹ of the five complexes can be associated with the asymmetrical stretching vibration and symmetrical stretching vibration of -COO-, respectively. The observed band at 1,653 cm⁻¹ in the FTIR spectrum of Zn-MOF is assigned as the stretching vibration of the -C=N- bond (Huang et al., 2020), which migrates to the lower band with Cu ion doping (Cu_0.01/Zn-MOF: 1,653 cm⁻¹, Cu_0.1/Zn-MOF: 1,651 cm⁻¹, Cu_0.2/Zn-MOF: 1,647 cm⁻¹, Cu_0.5/Zn-MOF: 1,639 cm⁻¹), probably due to competitive coordination between Cu²⁺ and Zn²⁺. The signals at 455 cm⁻¹ can be attributed to the stretching vibration of Zn–O. For the copper-doped Zn-MOF, slight shifts (Cu_0.01/Zn-MOF: 455 cm⁻¹, Cu_0.1/Zn-MOF: 461 cm⁻¹, Cu_0.2/Zn-MOF: 467 cm⁻¹, Cu_0.5/Zn-MOF: 478 cm⁻¹) were observed, suggesting the coordination of Cu with the carboxylic group of H₂L ligand. The FTIR spectra of Zn-MOF and Cu-doped Zn-MOF are almost the same, which further indicates that the two materials exhibit an isomorphic structure.

The thermal stabilities of compounds Zn-MOF and Cu_x/Zn-MOF (x = 0.01, 0.1, 0.2, 0.5) were examined by TGA under a N₂ atmosphere in the temperature range 25–800°C. As shown in Supplementary Figure S8, in the Zn-MOF complex, the weight loss rate reached 9.82% (calcd.10.61%) at 152–275°C, which is to the loss of one acetonitrile molecule. Then the skeleton of the compound began to collapse at 419 °C. The analysis of TGA curve of Cu_0.01/Zn-MOF confirmed that the weight loss rate was 10.34% (calcd.10.61%) at 167–275°C, and the skeleton of the compound began to collapse at 413°C. For Cu_0.1/Zn-MOF, the weight loss rate was 10.22% (calcd.10.61%) at 166–300°C, and the skeleton of the compound began to collapse at 381°C. For Cu_0.2/Zn-MOF, the weight loss rate was 9.31% (calcd.10.61%) at 167–295°C, and the skeleton of the compound began to collapse at 352°C. For Cu_0.5/Zn-MOF, the weight loss rate was 9.27% (calcd.10.61%) at 170–295°C, and the skeleton of the compound began to collapse at 348°C. We found that the doping of Cu(II) ions have a significant effect on the decomposition temperature of organic frames. With the increase of the Cu(II) ion doping ratio, the thermal stability decreases gradually, which probably due to copper divalent is easier to be reduced than zinc divalent at high temperature.

As shown in Figure 2A, the PXRD spectra of Zn-MOF were in good agreement with the simulation diagram, indicating the high degree of pure phase. When different proportions of Cu^II were doped into Zn-MOF, the positions of diffraction peaks were unchanged, suggesting that copper doping did not change the crystal structure of Zn-MOF. The doped Cu ion may be incorporated into the framework of Zn-MOF (Cao et al., 2018).

In order to test the stability of the skeleton structure, the Zn-MOF powder sample was soaked in water for 7 days, and its PXRD patterns were investigated. As shown in Figure 2B, after soaking in water for one day, the Zn-MOF sample remained basically crystalline, but the peak value was weakened, and an additional peak appeared at 2θ = 11.2°, indicating that the frame began to decompose (Kaye et al., 2007). The peak value was significantly reduced at 3 days, showing an acceleration of decomposition. After one week, the structure changed and the skeleton became unknown. As shown in Figure 2C and Supplementary Figure S9, four Cu-doped bimetal samples were immersed in water for 7 days, and their peak positions matched well, indicating that Cu-doped Zn-MOF had stronger water stability. The enhancement of water stability may be due to the formation of stronger coordination bonds after Cu^II replaces part of Zn^II in the original skeleton structure, which possibly improves the thermodynamic stability of metal clusters (Ding et al., 2019; Qiu et al., 2020). In practical applications, the chemical stability was also important in an aqueous environment. Therefore, the chemical stabilities of Cu_0.1/Zn-MOF were studied by soaking samples in water with varying pH (from 1 to 14), and adjusted using HCl and NaOH. PXRD patterns showed that samples were highly resistant and matched well with the original sample in a pH range of 3–12 (Figure 2D).

It is known that d¹⁰ configuration of Zn^II is difficult to oxidize or reduce; thus, the metal-to-ligand charge transfer (MLCT) or ligand-to-metal charge transfer (LMCT) is difficult to occur (Zhang et al., 2010). However, Zn^II ion coordinated with conjugated organic ligands may cause intraligand charge transfer (LLCT) (Yao et al., 2019). H₂L ligand containing carboxylic acid and aromatic and N-containing heterocycle may exist in π*→π or/and π*→n electronic transitions.

The fluorescence emission of H₂L ligand and various complexes in water were detected. As shown in Supplementary Figure S10, the fluorescence of Zn-MOF was significantly enhanced at 397 nm relative to the ligand in aqueous solution. The increased fluorescence may probably be attributed to LLCT induced by Zn^II ion (Dong et al., 2018), which may improve the rigidity of the structural unit due to the coordination interactions between the flexible ligand and Zn^II, and finally lead to the reduction of the non-radiative relaxation of the excited state (Dong et al., 2019). Subsequently, with the increase of the Cu^II doping ratio, the emission intensity gradually decreases. When the Cu^II doping rate increases to 0.5, the fluorescence emission intensity is almost equal to that of H₂L ligand. It is well known that the unsaturated electronic state of Cu^II ion (3d⁹) may cause MLCT, which causes the fluorescence quenching (Cui et al., 2012; Niu et al., 2014; Zheng et al., 2017). After doping of Cu^II ion, a new electron cloud distribution was formed, which may increase the energy loss of the system through non-radiative d-d transitions, and also weaken the LLCT, ultimately leading to a decrease in fluorescence intensity.

For a qualified fluorescence sensor, the stability of its frame structure, the luminescence intensity, and the sensitivity of fluorescence detection should be fully considered. Based on the aforementioned situation, Cu_0.1/Zn-MOF was selected as the fluorescence sensor for subsequent fluorescence detection. Three factors were mainly taken into account: First, although Zn-MOF has stronger fluorescence intensity than Cu/Zn-MOF, its fragile structural stability in the water environment precludes its possibility of being used as a fluorescence sensor in the water, which can also be verified by its seven-day fluorescence spectra in water (Supplementary Figure S11). From the perspective of structural and fluorescence stability in water, Cu^II ion–doped Zn-MOF is a better choice (Supplementary Figures S10, S12). Second, in terms of fluorescence emission intensity, Cu_0.01/Zn-MOF and Cu_0.1/Zn-MOF are slightly weaker than Zn-MOF, but significantly stronger than Cu_0.2/Zn-MOF and Cu_0.5/Zn-MOF. Therefore, Cu_0.01/Zn-MOF or Cu_0.1/Zn-MOF may be more suitable as fluorescence sensors. Third, Stokes shift is also an important factor to consider. In fluorescence detection, the increase of Stokes shift is beneficial to reduce background interference and enhance the signal-to-noise ratio and detection sensitivity (Kundu et al., 2012; Ren et al., 2018). As shown in Supplementary Figure S10, compared with the Stokes shift of 114 nm (λ_ex = 283 nm, λ_em = 397 nm) of Cu_0.01/Zn-MOF, the Stokes shift of Cu_0.1/Zn-MOF is up to 140 nm (λ_ex = 257 nm, λ_em = 397 nm). As the fluorescence sensing material with large Stokes shift is rare and attractive, Cu_0.1/Zn-MOF was selected as the fluorescence sensor for subsequent fluorescence detection.

Subsequently, the time-varying fluorescence spectra of Cu_0.1/Zn-MOF samples in the water environment were measured. As shown in Supplementary Figure S13, the fluorescence intensity of Cu_0.1/Zn-MOF in the water at different time periods did not change significantly. In addition, the fluorescence spectra of Cu_0.1/Zn-MOF at different pH values were also measured. As shown in Supplementary Figure S14, the fluorescence emission spectra of Cu_0.1/Zn-MOF at a pH range from 4 to 11 did not change significantly, which proved that the fluorescence of Cu_0.1/Zn-MOF also has good acid–base stability.

For extensive purpose, sixteen metal ions, including Na⁺, Ca²⁺, K⁺, Li⁺, Zn²⁺, Mg²⁺, Co²⁺, Mn²⁺, Cd²⁺, Ni²⁺, Ag⁺, Pb²⁺, Cu²⁺, Al³⁺, Cr³⁺, and Fe³⁺ at the same concentration, were utilized for investigating the sensibility of Cu_0.1/Zn-MOF. As shown in Figures 3A,B, compared with other cations, only Fe³⁺ ion showed obvious fluorescence quenching behavior. The fluorescence intensity at 397 nm was impaired by 93.8% compared with the initial value. This indicated that Cu_0.1/Zn-MOF can be used as a visual sensor for selective detection of Fe³⁺ ion in water. The selectivity experiment of Cu_0.1/Zn-MOF was explored. As shown in Figure 3C, the results showed that the interference of other cations can be ignored in the process of Fe³⁺ ion detection, which further confirms the selectivity of Cu_0.1/Zn-MOF for Fe³⁺ detection.

In order to explore the limit of detection (LOD) of Cu_0.1/Zn-MOF as a fluorescence sensor for the detection of Fe³⁺ ion, a fluorescence titration experiment was carried out. As shown in Figure 3D and Supplementary Figure S15, with the concentration of Fe³⁺ ion increased, the fluorescence intensity of the suspension decreased gradually, and the fluorescence quenching can be observed when the concentration of Fe³⁺ ion was 833 μM. The linear Stern–Volmer (S-V) equation, (I₀/I) = K_SV[C] + 1, can be used to explain the quenching efficiency (Li et al., 2016), where I₀ and I represent the fluorescence intensities of Cu_0.1/Zn-MOF suspension at 397 nm before and after addition of Fe³⁺, respectively. K_SV is the Stern–Volmer constant and [C] is the concentration of Fe³⁺. As shown in the Stern–Volmer plot of Cu_0.1/Zn-MOF (Figure 3E) in the concentration range of 0–70 μM, the KSV was calculated to be 2.76 × 10⁴ M⁻¹ (R ² = 0.995). According to the slope and standard error of the fitting curve, the LOD for Fe³⁺ was calculated to be 0.76 μM according to 3σ/K_SV (Qu et al., 2020), where σ is the standard deviation for eleven repeated luminescent measurements (σ = 0.0070). The sensitivity of Cu_0.1/Zn-MOF is a rival to most of the reported MOF-based sensors for Fe³⁺ (Supplementary Table S3), suggesting that it has potential application value in the detection of Fe³⁺.

The study of fluorescence quenching mechanism is of great significance for the exploration of more effective fluorescence materials. In the previous reports on the detection of metal ions, the quenching of fluorescence is usually caused by structure collapse, cation exchange, excitation energy competitive absorption, and energy resonance transfer (Mi et al., 2019). As shown in Supplementary Figure S9, after soaking in Fe³⁺ solution for 4 h, the PXRD patterns were unchanged, indicating that the fluorescence quenching is attributed to neither skeleton collapse nor cation exchange. In addition, the rapid response of Fe³⁺ detection also ruled out the possibility of fluorescence quenching caused by cation exchange because cation exchange cannot occur in a short period of time. As can be seen in Supplementary Figure S16, Fe³⁺ ions have a wide absorption band from 200 to 450 nm, which overlaps with the emission spectra range 310–510 nm of Cu_0.1/Zn-MOF. Therefore, the main reason for fluorescence quenching may be on account of the fluorescence resonance energy transfer (FRET) (Nagarkar et al., 2013). In addition, Fe³⁺ ion has a higher UV absorption at 257 nm, and when the excitation wavelength chose 257 nm, Fe³⁺ ion has a higher excitation energy competitive absorption.

Due to the excellent fluorescence performance and water stability of Cu_0.1/Zn-MOF, the detection performances of Cu_0.1/Zn-MOF on thirteen common antibiotics in six categories were studied. The results showed that NFT, NFZ, FZD, and TC had obvious quenching effect compared with other antibiotics (Figure 4).

In order to explore the sensitivity of Cu_0.1/Zn-MOF as a fluorescence sensor for the detection of nitrofurans, fluorescence titration experiments were carried out. The detection of nitrofuran antibiotics mostly focused on fluorescence enhancement or quenching. In this experiment, with the increase of titrated concentration, the fluorescence emission peak not only showed fluorescence quenching (Supplementary Figure S17) but also displayed obvious redshift (NFT: 397–448 nm, NFZ: 397–456 nm, FZD: 397–440 nm) (Figure 4). Generally, the decrease of fluorescence intensity and the redshift of fluorescence emission peak with the increase of the detection concentrates are rare. The emission wavelength shift may be attributed to a strong interaction between the analyte and the sensor (Hu et al., 2013; Lei et al., 2021). To verify this interaction, three nitrofurans were added separately to Cu_0.1/Zn-MOF solution. As shown in Supplementary Figure S18, the UV absorption peaks were found to be redshifted, which indicated the existence of static quenching, strong interactions, and the formation of ground state complexes (Zhang et al., 2017c; Kardar et al., 2020). In addition, emission wavelength shifts during titration are rare and attractive, which could enhance the detection specificity by adding an additional recognition dimension, in addition to the traditional recognition dimension of fluorescence intensity enhancement or quenching (Hu et al., 2013).

Considering the red-shift factor of fluorescence spectrum, the maximum emission peak corresponding to each titration concentration was selected in the calculation of the quenching rate of antibiotics and the value of fitting curve, instead of the fluorescence emission peak at a fixed wavelength. In this case, when the titrated concentration of NFT and NFZ reached 75 ppm, the fluorescence quenching efficiency rate was 90.9 and 94.1%, respectively. When the titrated concentration of FZD reached 100 ppm, the fluorescence quenching rate was 85.3%. To further understand the luminescent quenching degree, the quenching curves were quantitatively studied by Stern–Volmer equation. As shown in Figure 4, the Stern–Volmer plots illustrated a linear relationship in the low concentration region with a K_SV of 0.2212 ppm⁻¹ (5.27×10⁴ M⁻¹) for NFT, 0.2447 ppm⁻¹ (4.85×10⁴ M⁻¹) for NFZ, and 0.1660 ppm⁻¹ (3.74×10⁴ M⁻¹) for FZD. Based on the K_SV values, the detection limits of Cu_0.1/Zn-MOF toward NFT, NFZ, and FZD were calculated to be 95.0 ppb (0.4 μM), 85.8 ppb (0.43 μM), and 126.5 ppb (0.56 μM), respectively.

In general, the Stern–Volmer equation shows a linear relationship between the concentration of the analyte and luminescent strength. However, in the detection of NFT and FZD, the fitting curve shows a double exponential type. In this case, the Stern–Volmer equation should be expressed as I₀/I-1 = lg[K_SV]+lg[C]. This means that the quenching of luminescence may be the result of both dynamic and static factors (Yang Y et al., 2020). In addition, fluorescence emissions of NFT, NFZ, FZD, and TC were detected at 257 nm excitation wavelength, and it was found that NFT and FZD had weak fluorescence emission in the 400–510 nm region, while NFZ and TC had almost no fluorescence emission (Supplementary Figure S19). By comparing the fluorescence quenching curves of NFT and FZD titration experiments with the fluorescence emission curves of NFT and FZD itself, it was found that the curve shapes were similar. In other words, at 257 nm excitation wavelength, as the concentration of NFT and FZD increases, the fluorescence spectra are gradually weakened and redshifted, and the coincidence degree with the fluorescence emission peak of NFT and FZD itself gradually increases (Supplementary Figure S20). As a result, at higher concentrations of NFT and FZD, the fluorescence quenching degree was weakened by the fluorescence emission of NFT and FZD itself.

Similarly, in order to further explore the quantitative detection ability of TC by complex Cu_0.1/Zn-MOF, the TC solution was dropped into the complex suspension for fluorescence titration (Figure 5A). With the dripping of TC, the fluorescence intensity of the suspension decreases gradually, and the fluorescence quenching occurred when the TC concentration was 89 ppm. The fluorescence quenching rate was 94.8% compared with the initial value. This indicated that Cu_0.1/Zn-MOF can be used as visual sensor for selective detection of TC in water. The quenching curves were quantitatively studied by Stern–Volmer equation. As shown in Figure 5B, the Stern–Volmer plot illustrated a linear relationship in the low concentration region with a K_SV of 0.1237 ppm⁻¹ (5.94×10⁴ M⁻¹) for TC. Based on the K_SV value, the detection limit of Cu_0.1/Zn-MOF toward TC was calculated to be 169.8 ppb (0.35 μM). Compared with the reported MOF-based sensors for the detection of antibiotics, the result implied that Cu_0.1/Zn-MOF has potential application value in the detection of NFT, NZF, FZD, and TC (Supplementary Table S4).

The mechanism analysis of antibiotic detection is as follows: the PXRD patterns of Cu_0.1/Zn-MOF soaked in four antibiotics were basically unchanged compared with those before soaking (Supplementary Figure S9D), which ruled out the possibility of fluorescence quenching caused by skeleton collapse. The excitation spectra of Cu_0.1/Zn-MOF overlapped with the UV absorption spectra of SM, CHL, CEF, NFT, NFZ, FZD, and TC, but hardly overlapped with other antibiotics (Supplementary Figure S21), indicating that there may be competitive excitation energy absorption effect between the seven antibiotics and Cu_0.1/Zn-MOF. All the seven antibiotics have fluorescence quenching effect on Cu_0.1/Zn-MOF, but the fluorescence quenching effect of NFT, NFZ, FZD, and TC is significantly stronger than that of SM, CHL, and CEF, indicating that other factors may be involved in the fluorescence quenching. As shown in Supplementary Figure S21, the emission spectra of Cu_0.1/Zn-MOF overlapped with the UV absorption spectra of NFT, NFZ, FZD, and TC, and hardly overlapped with other antibiotics. To some extent, this may explain that NFT, NFZ, FZD, and TC have stronger fluorescence quenching rates for Cu_0.1/Zn-MOF than SM, CHL, and CEF, and suggested that FRET may also be an important reason of fluorescence quenching. In addition, due to the low LUMO energy levels of antibiotics, photoinduced electron transfer (PET) is also a possible luminescence quenching mechanism (Toal and Trogler, 2006; Nagarkar et al., 2014). The valence band energy levels (VB) and conduction band energy levels (CB) of MOF can be described in a pattern similar to molecular orbitals (MOs) (Wang et al., 2016). Specifically, the electrons in the Cu_0.1/Zn-MOF–occupied orbital (HOMO) are excited to the lowest unoccupied orbital (LUMO), followed by the transfer to the antibiotic with a lower LUMO level. To explain the possibility of this process, we calculated and derived the band structures of Cu_0.1/Zn-MOF and antibiotics. According to Kubelka–Munk, the bandgap is about 3.6 eV (Figure 6A). As shown in Figure 6B, the Mott−Schottky plots show a flat band potential of about −1.2 V versus Ag/AgCl for Cu_0.1/Zn-MOF, while the LUMOs of NFT, NFZ, FZD, and TC were calculated to be −3.86 eV, −3.62 eV, −3.75 eV, and −4.53 eV, respectively (Wang et al., 2016; Malik and Iyer, 2017). Due to the conduction band of MOFs being at a higher energy level than the lowest unoccupied molecular orbital (LUMO) of the antibiotic, electrons were transferred from MOFs to the antibiotics (Figures 6C, D), which eventually leads to the fluorescence quenching of MOFs.

The aforementioned experimental results indicated that Cu_0.1/Zn-MOF displays good water stability and detection sensitivity. In addition, the fluorescence sensor also needs to have good cyclic stability and fast response. After usage, the sensor can be regenerated by centrifugation and acetone washing. After soaking in 5 analytes for 4 h, the main peak of PXRD patterns matched well with that before soaking (Supplementary Figure S9D), and the fluorescence quenching efficiency remained unchanged after 5 cycles (Supplementary Figure S22). The results showed that the sensor exhibited good recyclability. In addition, the sensor showed rapid response to all the five analytes at different concentrations, as shown in Supplementary Figure S22, and the fluorescence intensity decreased within 20 s and remained stable after 120 s.

It has been proved that the doping of Cu^II ions contributes to the improvement of the water stability of Zn-MOF. Subsequently, in order to verify the influence of the doping of Cu ions on the sensitivity of detection, the original Zn-MOF was used as fluorescence sensors to conduct titration experiments on Fe³⁺, NFT, NFZ, FZD, and TC, respectively. As shown in Supplementary Figure S23–S27, the Stern–Volmer plots illustrated a linear relationship in the low concentration region with a K_SV of 1.01×10⁴ M⁻¹ for Fe³⁺, 9.33×10⁴ ppm⁻¹ (2.22×10⁴ M⁻¹) for NFT, 1.12 ×10⁵ ppm⁻¹ (2.23×10⁴ M⁻¹) for NFZ, 4.74×10⁴ ppm⁻¹ (1.07×10⁴ M⁻¹) for FZD, and 4.96×10⁴ ppm⁻¹ (2.39×10⁴ M⁻¹) for TC. The data showed that compared with Cu_0.1/Zn-MOF, the K_SV of Zn-MOF for five pollutants are significantly reduced. In other words, proper Cu^II doping improves the sensitivity of Zn-MOF to detect analytes. The mechanism of the higher sensitivity of Cu_0.1/Zn-MOF in the detection of analytes is still not completely clear. We speculate that it may be related to the following factors: First, the increase of Stokes shift reduces the background interference, which is beneficial to the strong penetration to the sample, and enhances the detection sensitivity. Second, the liquid UV spectrum displayed that the UV absorption of TC was similar at 257 and 283 nm, while the UV absorptions of Fe³⁺, NFT, NFZ, and FZD were stronger at 257 nm than at 283 nm (Supplementary Figure S28). In other words, it might have stronger excitation competitive absorption at 257 nm, leading to the enhancement of fluorescence quenching.