Cu–Co bimetallic organic framework as effective adsorbents for enhanced adsorptive removal of tetracycline antibiotics

Abstract

In this study, the removal effect of a new MOF-on MOF adsorbent based on Cu–Co bimetallic organic frameworks on tetracycline antibiotics (TCs) in water system was studied. The adsorbent (Cu-MOF@Co-MOF) were synthesized by solvothermal and self-assembly method at different concentrations of Co²⁺/Cu²⁺. The characterization results of SEM, XRD, XPS, FTIR and BET indicated that the MOF-on MOF structure of Cu-MOF@Co-MOF exhibited the best recombination and physicochemical properties when the molar ratio of Co²⁺: Cu²⁺ is 5:1. In addition, the Cu-MOF@Co-MOF have a high specific surface area and bimetallic clusters, which can achieve multi-target synergistic adsorption of TCs. Based on above advantages, Cu-MOF@Co-MOF provided a strong affinity and could efficiently adsorb more than 80% of pollutants in just 5 to 15 min using only 10 mg of the adsorbent. The adsorption capacity of tetracycline and doxycycline was 434.78 and 476.19 mg/g, respectively, showing satisfactory adsorption performance. The fitting results of the experimental data were more consistent with the Langmuir isotherm model and pseudo-second-order kinetic model, indicating that the adsorption process of TC and DOX occurred at the homogeneous adsorption site and was mainly controlled by chemisorption. Thermodynamic experiments showed that Cu-MOF@Co-MOF was thermodynamically advantageous for the removal of TCs, and the whole process was spontaneous. The excellent adsorption capacity and rapid adsorption kinetics indicate the prepared MOF-on MOF adsorbent can adsorb TCs economically and quickly, and have satisfactory application prospects for removing TCs in practical environments. The results of the study pave a new way for preparing novel MOFs-based water treatment materials with great potential for efficient removal.

Introduction

Tetracycline antibiotics (TCs) are one of the most commonly used antibiotics in animal husbandry, with the output and usage ranking second in the world, and TCs residues are mostly concentrated in groundwater and soil, which have been classified as emerging contaminants by environmental protection authorities in various countries¹. The extensive use of TCs in aquaculture and animal husbandry further increases the residual concentration of TCs in water circulation systems, leading to the emergence and spread of antibiotic-resistant genes and antibiotic-resistant bacteria. These resistance genes, which spread to other microbial populations, also lead to the creation of superbugs and increasing bacterial resistance to antibiotics, thus posing a serious public health risk². With the improvement of living standards and the progress of social development, the public is more concerned about the safety of environmental water quality. Governments around the world have invested considerable research into the technology of efficient removal of TCs. However, TCs have complex molecular structures and changeable states, and cannot be completely decomposed through the metabolic processes of humans and animals. Therefore, it is difficult to completely remove TCs, which greatly endangers the living environment and ecological security of human beings³. Using advanced technology to remove TCs from urban environmental water systems and obtain clean and safe water has become an emerging way to alleviate water scarcity. Therefore, it is urgent to develop an efficient, fast and low-cost TCs removal technology.

At present, the most widely used technical approaches include electrochemistry, biodegradation, photocatalysis, flocculation, membrane separation, chemical oxidation and adsorption⁴. For the removal of complex TCs, most of the current conventional water treatment technologies are difficult to achieve satisfactory removal effect. Among the many developed removal technologies for TCs, adsorption has become the most attractive method in water treatment technology because of its convenient operation, fast operation, high removal efficiency. The adsorption of TCs antibiotics by adsorbents is mainly by adjusting their microstructure, surface functional groups such as –OH and –COOH and other electrostatic sites including heteroatoms such as –F and –Cl⁵. Ren et al. obtained humic acid-loaded Fe₃O₄ magnetic nano-adsorbent (AHA–Fe₃O₄) from rice straw by hydrothermal method and then used to remove three tetracycline antibiotics from water, which was an efficient and green water pollutant adsorbent⁶. In the application of adsorption technology, selecting the adsorbent with good performance is the most important step. Therefore, many efficient adsorption materials, such as metal–organic frameworks (MOFs), carbon nanotubes, metal oxide nanoparticles, biochar, and graphene etc., have been used in the wastewater treatment of TCs^7,8. At present, it is still very meaningful and attractive to develop new adsorbents with low cost and satisfactory removal performance to overcome the shortcomings of some traditional adsorbents with poor adsorption capacity.

MOFs, as a class of crystal material with ordered pores composed of metal clusters and organic ligands, exhibited good adsorption properties and had received extensive attention recently. The reversible coordination bonds between the metal cluster nodes and the organic ligands ensure the generation of highly structured framework structures. Benefitting from its high surface area, adjustable structure and pore size that accelerate the selective removal of antibiotics, the research of MOF-based adsorbents in the field of water treatment has grown exponentially. However, the adsorption properties of monomer MOFs still need to be improved, so MOFs are often combined with some functional species, such as nanoparticles, biomolecules, and other MOFs⁹. The combination of monomer MOFs with other functional MOFs to form multi-metal-functional MOFs is a common way to extend the functionality of MOFs. Bimetallic organic framework such as MOF@MOF composites are usually prepared by coating or wrapping another MOFs on the surface of the MOFs, which can combine the physicochemical properties and structural advantages of the two MOFs, further leading to the synergistic effect¹⁰. This versatile MOFs material provides a new way to quickly and efficiently remove environmental contaminants. It is found that multi-component composites can improve the adsorption performance of pollutants in the environment.

In the present work, we have developed a MOF-on MOF structure with copper-cobalt bimetallic organic framework (Cu-MOF@Co-MOF) via self-assembly method that can efficiently and rapidly remove TCs from environmental water systems based on dispersed solid phase extraction (DSPE), as shown in Fig. 1. Besides, the multi-target adsorption characteristics of Cu-MOF@Co-MOF against TCs were systematically studied, and the adsorption mechanism was further revealed through the instrumental characterization of Cu-MOF@Co-MOF before and after the adsorption of TCs. Aiming at the practical application of pollutant removal in the environment, the design idea of integrating single function materials into multiple synergistic adsorbents provides a new way for environmental engineering applications.

Figure 1.

The procedure for synthesis of Cu-MOF@Co-MOF and removal process of TCs over Cu-MOF@Co-MOF adsorbent via DSPE.

Materials and methods

Materials

Tetracycline (TC) and doxycycline (DOX) were obtained from Macklin and the structure was shown in Fig. S1. Cobalt nitrate hexahydrate (Co(NO₃)₂·6H₂O) was purchased from BeiJing Royaltech Co.,Ltd. Polyvinyl pyrrolidone (PVP), 1, 3, 5-benzenetricarboxylic acid (H₃BTC) was purchased from Nanjing Xiezun Medical Technology Co., LTD. Ethanol, methanol were purchased from General-reagent. Cu(NO₃)₂·3H₂O, 2-methylimidazole (2-MIm) and N,N-dimethylformamide (DMF) was obtained from Aladdin. Unless otherwise stated, all aqueous solutions used are prepared from Deionized Milli-Q water.

Real water samples

The removal properties of the synthesized Cu-MOF@Co-MOF nanomaterials were evaluated by analyzing the removal effects of TCs from two real aqueous solutions. The first sample was tap water, which was taken from the laboratory after letting tap water flow for 10 min. The second sample was taken from lake water in Nantong City, Jiangsu Province. Both samples were filtered through a 0.45 μm membrane filter and then stored in a Teflon bottle at 5 °C for later use.

Preparation of Cu-MOF@Co-MOF

0.0215 mol Cu(NO₃)₂·3H₂O was dissolved in 66 mL DMF to form solution A, and then 0.013 mol H₃BTC was dissolved in 66 mL DMF to form solution B. Subsequently, 1.9752 g PVP was added into solution A and stirred evenly. Then, adding solution B slowly to solution A and stirring for 30 min. The mixture was transferred to a Teflon-lined autoclave, and placed in an oven at 80 °C for 24 h. After the reaction, the precipitation was cooled to room temperature, washed with DMF and ethanol for 3 times, and dried in the oven at 80 °C for 24 h to obtain Cu-MOF.

In the preparation of Cu-MOF@Co-MOF, 300 mg Cu-MOF was placed in 75 mL of methanol, dispersed evenly after 5 min of ultrasonication, and then 3.75, 5 and 7 mmol Co(NO₃)₂·6H₂O were added and stirred for 30 min, and then 75 mL of methanol solution containing 15, 25, and 35 mmol 2-MIm were slowly poured into the above solution. After further stirring for 30 min, the mixture was stood at room temperature for 4 h. The precipitate was collected by centrifugation, washed three times using methanol and dried at 80 °C for 24 h. The resulting product was called as Cu-MOF@Co-MOF-3.75, Cu-MOF@Co-MOF-5, Cu-MOF@Co-MOF-7, respectively. The Co-MOF was prepared by not adding Cu-MOF.

Relevant characterization of Cu-MOF@Co-MOF

The morphologies and size of the synthesized samples were obtained by using a field emission scanning electron microscope (SEM, Carl Zeiss, Germany). Fourier transform infrared (FTIR) spectra with the range of 4000–500 cm⁻¹ were obtained by a FTIR spectrometer. The crystal analysis and phase information were confirmed by X-ray diffraction (XRD) patterns on an X-ray diffractometer with high-density Cu-Kα radiation (D8 VENTURE/QUEST, Germany). The X-ray photoelectron spectroscopy (XPS) was applied to measure the chemical state and elemental composition (Thermo Fisher, USA). Thermogravimetric analysis (TGA) was conducted by a thermal analyzer from room temperature to 800 °C under an N₂ atmosphere at a rate of 10 °C/min (TGA, Netzsch, Germany). The N₂ adsorption/desorption isothermal curves were applied to calculate Brunauer–Emmett–Teller (BET) surface areas and pore size distribution (Micromeritics, ASAP 2460, USA). The concentration of TCs after material adsorption was determined by UV–vis spectrophotometer (TU-1901, Beijing Persee, China).

Batch adsorption experiments

The TCs adsorption process based on Cu-MOF@Co-MOF under DSPE method was as follows. 0.1 g TCs powder was dissolved in 100 mL of ultrapure water to prepare 1 g/L TCs stock solution. The different concentrations of TCs solutions required for the experiment were prepared by diluting the stock solution. Cu-MOF@Co-MOF was used as the adsorbent and adsorbed at a speed of 200 rpm/min in a thermostatic oscillator. 10 mg of Cu-MOF@Co-MOF was added to 20 mL of TCs solution and shake uniformly for a certain period of time. After adsorption was completed, a small amount of supernatant was filtered through a 0.22 μm filter membrane, and Cu-MOF@Co-MOF adsorbent was discarded. The concentration of TCs in the supernatant was determined by measuring the absorbance at 357 nm with UV–Vis spectrophotometer. The pH of the solution was adjusted by 1.0 mol/L HCl and 1.0 mol/L NaOH. The effects of different adsorbent doses, pH, coexisting ions, adsorption time and adsorption temperature on the removal efficiency of TCs were investigated. All the adsorption experiments were performed in parallel for three times, and the average values of the three times were evaluated. In each experiment, the removal efficiency (RE, %) was calculated according to Eq. (1):

$$\text{RE}=\frac{\left({\text{C}}_0-{\text{C}}_{\text{e}}\right)}{{\text{C}}_0}\times 100\text{\%}$$ 1

Among them, C₀ (mg/L) is the concentration of TCs in the initial sample solution, and C_e (mg/L) is the concentration of TCs in the sample solution at equilibrium.

To evaluate the removal capacity of Cu-MOF@Co-MOF to TCs, adsorption capacity q_t (mg/g) at any time (t) and q_e (mg/g) at equilibrium were used to calculate through Eqs. (2) and (3), respectively¹¹.

$${\text{q}}_{\text{t}}=\frac{\left({\text{C}}_{\text{o}}-{\text{C}}_{\text{t}}\right)\text{V}}{\text{m}}$$ 2

$${\text{q}}_{\text{e}}=\frac{\left({\text{C}}_{\text{o}}-{\text{C}}_{\text{e}}\right)\text{V}}{\text{m}}$$ 3

In these equations, m (g) represents the weight of the Cu-MOF@Co-MOF and V (L) is the volume of the TCs solution. C_e (mg/L) is the concentration of TCs in the sample solution at equilibrium. q_t and q_e represent adsorption capacity at arbitrary time (t, min) and equilibrium, respectively.

Adsorption isotherms and thermodynamics

A class of adsorption experiments with TCs solution and Cu-MOF@Co-MOF adsorbent were proceeded by adding 10 mg of absorbent to different initial concentrations of TC and DOX solutions (10–450 mg/L), and the adsorption process was performed for 5 and 15 min, respectively. The used absorbent was removed by filtration with a membrane, and the concentration of TCs was determined at 357 nm by UV–vis spectroscopy.

Adsorption kinetics

The Cu-MOF@Co-MOF adsorbent (10 mg) was added to the aqueous solution containing TC and DOX. In a thermostatic water bath vibrator, the adsorption experiments were performed at 200 rpm with the TC and DOX concentrations of 100 and 240 mg/L for 0.5–20 min, respectively. After the adsorption was completed, a small amount of the solution was withdrawn and filtered by 0.22 μm filter membrane. Then the concentration of TCs in the supernatant was measured immediately through UV–vis spectrophotometry.

Results and discussion

Characterization of Cu-MOF@Co-MOF

The morphology of the prepared materials was characterized by SEM. Cu-MOF exhibited an octahedral structure shown in Fig. 2a. Figure S2 showed that Co-MOF was a dodecahedral morphology. Both the Cu-MOF and Co-MOF had uniform particle size and excellent dispersion. Figure 2b–d listed SEM images of Cu-MOF@Co-MOF composites. It was found that Co-MOF was dispersed and wrapped on the surface of Cu-MOF. With the increase of Co(NO₃)₂·6H₂O dosage, the content of Co-MOF attached to the surface of Cu-MOF gradually increase. The Cu-MOF@Co-MOF-5 synthesized in the presence of 5 mmol Co(NO₃)₂·6H₂O showed a uniform coating layer. When the amount of Co(NO₃)₂·6H₂O continues to increase to 7 mmol, Co-MOF could not grow completely on the surface of Cu-MOF, leaving a lot of residual agglomerated Co-MOF around Cu-MOF@Co-MOF-7. To prove the crystal structure of these MOFs materials, XRD experiments were performed. The characteristic diffraction peaks of prepared Cu-MOF were located at 2θ = 6.69°, 9.48°, 11.60°, 13.41°, which were consistent with the simulated Cu-MOF (Fig. 3a). The characteristic peaks of Co-MOF at 7.33°, 10.38°, 12.70°, 18.01°, represented Co-MOF (001), (002), (112) and (222) peak of Co-MOF, respectively, confirming the successful synthesis of Co-MOF¹². There was almost no characteristic peak of Co-MOF in the XRD pattern of Cu-MOF@Co-MOF-3.75, which might be because the Co content was too small to successfully wrap on the surface of Cu-MOF. The characteristic peaks of Cu-MOF and Cu-MOF could be clearly seen in the XRD patterns of Cu-MOF@Co-MOF-5 and Cu-MOF@Co-MOF-7, which confirmed the successful synthesis of the composite bimetal-organic framework.

Figure 2.

SEM images of (a) Cu-MOF; (b) Co-MOF@Co-MOF-3.75; (c) Co-MOF@Co-MOF-5; (d) Co-MOF@Co-MOF-7.

Figure 3.

Characterization results of (a) XRD patterns; (b) FTIR spectra; (c) XPS spectra; and (d) N₂ adsorption–desorption isotherms.

The composition of surface functional groups was characterized by FT-IR. Figure 3b showed the FTIR spectra of Cu-MOF, Co-MOF, Co-MOF@Co-MOF-5, and Cu-MOF@Co-MOF-7. The bands 1444 cm⁻¹ and 1640 cm⁻¹ represented the –O–C–O– group of Cu-MOF, and the bands of 1373 cm⁻¹ and 1568 cm⁻¹ were the C=C stretching vibration of the organic ligand¹³. The bands near 729 cm⁻¹ belonged to the stretching vibration of Cu–O, indicating that the coordination between the oxygen atom and the Cu atom¹⁴. The main absorption peak of the FT-IR spectrum of Co-MOF was related to 2-methylimidazole, and the peak at 600–1500 cm⁻¹ indicated the stretching and bending modes of the imidazole ring. The peaks at 2927 and 3131 cm⁻¹ belonged to the stretching patterns of the aromatic ring of 2-MIm and the C–H bond of the aliphatic chain, respectively. The peak at 1582 cm⁻¹ was caused by the stretching pattern of the C=N bond in 2-MIm¹⁵. The appearance of three new absorption bands at 3433, 2920, and 2850 cm⁻¹ in Cu-MOF@Co-MOF-5, which could be attributed to N–H bond bending vibrations of tertiary amines and C–H stretching vibrations, respectively, confirmed the fixation of Co-MOF on the Cu-MOF surface during Cu-MOF@Co-MOF-5 synthesis. In addition, compared with Cu-MOF, the FT-IR spectrum of Cu-MOF@Co-MOF-5 showed a redshift at the characteristic peaks of 1620, 1435, 1369, and 727 cm⁻¹, which further confirmed the successful recombination of Cu-MOF and Co-MOF. All the unique absorption bands of Co-MOF and Cu-MOF were present in the structure of Cu-MOF@Co-MOF-5 and Cu-MOF@Co-MOF-7 with a reduction in the spectral intensity due to the recombination process.

Bimetallic MOFs with different Cu-Co ratios were studied by XPS analysis. From the XPS spectrum in Fig. 3c, it could be seen that the characteristic peaks of Cu2p, Co2p and C1s coexisted in the spectrum, and the intensity of the characteristic peaks of XPS of Co increased significantly with the increase of Co content, which further confirmed the successful synthesis of Cu-Co bimetallic MOFs. The specific surface area was calculated using the nitrogen adsorption–desorption isotherm of the Cu-MOF@Co-MOF-5 sample combined with the Brunauer-Emmet-Teller method (Fig. 3d). The BET surface areas, pore size and pore volume of the Cu-MOF@Co-MOF-5 were 581.93 m²/g, 2.23 nm and 0.324 cm³/g, respectively, indicating the large specific surface area and mesoporous structure of the material. The above characterization results confirmed the successful synthesis of MOF-on MOF structure, and revealed various functional groups in and on the surface of Cu-MOF@Co-MOF-5 composites, which contributed to the adsorption of TCs pollutants. Therefore, Cu-MOF@Co-MOF was used to refer to Cu-MOF@Co-MOF-5 in the following description.

Adsorption experiment

In order to investigate the adsorption performance of Cu-MOF@Co-MOF materials on TCs, the variables such as the initial concentration, pH value, TC concentration, adsorption time and coexisting ions during the adsorption of MOF materials on TCs were studied and optimized.

Effect of initial concentration on the adsorption of TCs

The effect of the initial concentration on the removal efficiency of TCs was studied at a dosage of 10 mg Cu-MOF@Co-MOF and a contact time of 5 and 15 min. The experiments were tested at initial concentrations of TC and DOX of 10–450 mg/L. As exhibited in Fig. S3a, the adsorption capacity of Cu-MOF@Co-MOF increased with the increase of TC and DOX loading, which might be due to the increasing number of collisions between the Cu-MOF@Co-MOF and TCs molecules, which led to a greater driving force for the diffusion of TCs molecules with higher concentration from solution to the adsorption site. When the concentrations of TC and DOX were increased to 240 and 400 mg/L, respectively, the adsorption capacities of the two TCs reached equilibrium.

Effect of adsorption time and temperature

The high adsorption rate can greatly reduce the cost of water treatment by shortening the water treatment time, which reflects the application potential of a new type of adsorbent. To test the effect of adsorption time on removal efficiency, adsorption experiments from 1 to 25 min were performed at 25 °C. The initial adsorption rate of Cu-MOF@Co-MOF for TC and DOX was very fast, which was attributed to the abundant adsorption sites and strong affinity provided by the Cu-MOF@Co-MOF adsorbent. Then, as more and more adsorption sites were occupied by TCs molecules, the adsorption and removal rate enhanced slowly with the increase of contact time. Finally, the adsorption processes of TC and DOX tended to equilibrium at 5 and 15 min, respectively, and the maximum removal rate reached 95% (Fig. S3b).

Temperature affects the cavitation phenomenon and solubility of the analyte, which in turn affects the mass transfer process and also a major factor affecting the adsorption removal efficiency of TCs. The effect of temperature was investigated at 25 ℃, 35 ℃ and 45 ℃, and the adsorption capacity decreased with the increase of temperature. In contrast, TCs exhibited the best removal performance at 25 °C.

Effect of incipient solution pH

The determining factors of pH value affecting adsorption capacity are surface functional groups and TCs molecular structure, as it will affect the degree of ionization of the Cu-MOF@Co-MOF and the form of the molecule. The effect of Cu-MOF@Co-MOF on adsorption in pH range of 2.0–11.0 was investigated and shown in Fig. S3c. Although the surface of Cu-MOF exhibits a negative charge¹⁶, the Co-MOF attached to its surface in most cases has a positive surface charge, evidenced by the zeta potential values in the PH range of 2–10¹⁷. Therefore, when pH was lower than 5, Cu-MOF@Co-MOF surface was positively charged, and TC and DC molecules were converted into TC⁺ and DOX⁺ ions, respectively. The repulsive force between TCs⁺ molecules and the adsorbent surface limited the adsorption capacity to some extent. When the pH was between 5 and 9, the TC and DOX molecule were neutral, and the adsorption process mainly relied on non-electrostatic attraction. At pH above 10, Cu-MOF@Co-MOF became negatively charged, and the electrostatic repulsion between TCs and Cu-MOF@Co-MOF surface again resulted in a decrease of adsorption capacity. However, even at pH below 5 and above 9, the maximum adsorption capacity was not less than 420 mg/g, which indicated that the Cu-MOF@Co-MOF can efficiently remove TC and DOX from environmental pollution systems over a wide pH range.

Effect of coexisting anions and ionic strength interference

In addition to tetracycline pollutants, various environmental waters and industrial wastewater usually contain various toxic compounds such as salt acid, alkali, and metal ions, which will greatly reduce the removal performance of the adsorbents. In theory, when there is a repulsive electrostatic interaction between the pollutant and the adsorbent, the increase of ionic strength or salt concentration will increase the adsorption capacity of the adsorbent. If the electrostatic interaction is mutually attractive, the effect of increasing the salt concentration on the adsorption capacity will be weakened¹⁸. Therefore, solutions with different ionic strength were obtained by adjusting NaCl, KCl, CaCl₂ and MgCl₂, and based on this, the effect of ionic strength on the adsorption capacity of antibiotics was investigated. Under the same conditions, when 10 mol/L salt solution was added to TCs solution, the adsorption capacity of TC and DOX decreased by 48.9 and 20%, respectively, indicating that the addition of salt ions would reduce the removal rate of TCs.

Adsorption mechanism study

Isotherm adsorption

Adsorption isotherms are a class of correlation curves used to describe the equilibrium distribution relationship between adsorbate and adsorbent material in a solution at a certain temperature. The potential adsorption capacity was evaluated using the Langmuir, Freundlich, Temkin and Dubinbin-Radushkevich (D-R) isotherm models and the adsorption mechanism was further determined^19–21.

$$\frac{{\text{C}}_{\text{e}}}{{\text{q}}_{\text{e}}}=\frac{1}{{\text{q}}_{\text{m}}{\text{K}}_{\text{L}}}+\frac{{\text{C}}_{\text{e}}}{{\text{q}}_{\text{m}}}$$ 4

$${\text{lnq}}_{\text{e}}=ln{\text{K}}_{\text{F}}+\frac{1}{\text{n}}{\text{ln C}}_{\text{e}}$$ 5

$${\text{q}}_{\text{e}}={\text{BlnK}}_{\text{T}}+{\text{BlnC}}_{\text{e}}$$ 6

$$ln{\text{q}}_{\text{e}}=ln{\text{q}}_{\text{m}}-{\text{k}}_{\text{DR}}{\epsilon }^2\left(\epsilon =\text{RTln}\left(1+\frac{1}{{\text{C}}_{\text{e}}}\right)\right)$$ 7

$$\text{E}={\left(2{\text{K}}_{\text{DR}}\right)}^{-1/2}$$ 8

q_m is the maximum adsorption capacity (mg/g); C_e stands for equilibrium concentration (mg/L), K_L represents the Langmuir constant (L/mg), 1/n and K_F are heterogeneity factor and Freundlich constant, respectively, B is the constant related to the heat of adsorption, K_T represents the equilibrium binding constant at the maximum binding energy (L/mg), ε is the Polanyi potential (KJ/mol), K_DR is the D-R isotherm constant (mol²/KJ²), E (KJ/mol) is the amount of free energy change of 1 mol TCs from solution to surface, which is beneficial to estimate the type of adsorption reaction, which can be obtained from k_ad.

The adsorption isotherm can not only describe the adsorption capacity of the adsorbent, but also reveal the molecular layer characteristics when the analyte and the adsorbent reach adsorption equilibrium. The fitting curves and corresponding parameters of four isothermal models were fitted to the experimental data at different concentrations, and the results were exhibited in Fig. 4 and Table 1. The Langmuir model is originally used for gas adsorption on solid surfaces, assuming that adsorption is mainly a uniform monolayer chemical adsorption process. Usually there is no spatial resistance and lateral force between adsorbents and adsorbate is a single layer covered on the surface of the adsorbent. When each adsorption site on the adsorbent surface adsorbs a molecule, its adsorption capacity reaches the maximum value, and the whole adsorption process is in a dynamic equilibrium state. Moreover, the Langmuir model can also evaluate the potential maximum adsorption capacity of the adsorbent. Here, the Langmuir isotherm model was used to determine the maximum saturated adsorption capacity, adsorption mechanism and adsorption driving force. The fitting curves corresponding to the Langmuir model in Fig. 4a showed that the correlation coefficients (R²) between TC and DOX reached 0.999, which was higher than that of the other three models. The maximum adsorption capacity (q_m) of TC and DOX was 434.78 and 476.19 mg/g, respectively, calculated by Langmuir isothermal. Moreover, the adsorption data obtained by the experiment were also closer to the Langmuir isotherm model, indicating that the adsorption of TC and DOX occurred at the homogeneous adsorption site.

Figure 4.

Isotherms fitting curves of TCs on Cu-MOF@Co-MOF: (a) Langmuir model and (b) Freundlich model.

Table 1.

Adsorption isothermal models and its correlation fitting coefficients.

Samples	Langmuir model			Freundlich model
	qm (mg/g)	KL (L/mg)	R2	KF (mg/g)	n (g/L)	R2
TC	434.78	1.769	0.999	340.22	18.14	0.629
DOX	476.19	0.778	0.999	321.34	12.15	0.94

Samples	Temkin			Dubinbin-Radushkevich
	KT (L/mg)	B	R2	KDR (mol2/kJ2)	E (kJ/mol)	R2
TC	1.391	132.92	0.995	300	0.04	0.9947
DOX	9114.3	34.556	0.972	0.08	2.5	0.918

The Freundlich isotherm model is used to represent the multilayer adsorption characteristics and equilibrium data of heterogeneous surfaces under non-ideal conditions. The adsorption amount is the sum of adsorption at all sites, which can explain the experimental results in a wider range. For the process of adsorption of TCs by Cu-MOF@Co-MOF, the correlation coefficient (R²) of the fitting curve of the Freundlich model ranged from 0.629 to 0.94 (Fig. 4b). n is a parameter that can be used to represent the adsorption process, which can be physical (n > 1), chemical (n < 1), or a linear (n = 1) process. The n of this model was about 5, indicating that there was a partial physisorption process. In addition, 1/n is an important indicator of surface heterogeneity. As the surface of the material becomes more heterogeneous, 1/n will be closer to zero. The 1/n value of this model was 0.05–0.08, indicating a certain degree of heterogeneity on the surface of Cu-MOF@Co-MOF²².

The Temkin adsorption isothermal model is a real model, which assumes that the adsorption heat of molecules in the adsorption layer decreases linearly with the increase of coverage, and also takes into account the influence of the interaction between the adsorbate and the adsorbent. The parameters of the Temkin model in this work were the slope and intercept of the curve between q_e and lnC_e. The Temkin fitting model in Fig. S4a,b showed that Temkin also had a relatively good fitting results, indicating that with the increase of the number of layers covered by TCs molecules, the binding force decreased linearly, so the adsorption process between TCs and Cu-MOF@Co-MOF was dominated by uniformly distributed binding force²³.

The D-R isotherm model, proposed in 1947 by Dubinbin and Radushkevich, is an empirical formula based mainly on pore filling theory that can be used to illustrate the effects of the porous structure of adsorbents. This model is more suitable for describing the actual adsorption situation and characteristics. The D-R isothermal adsorption model is used to describe whether the adsorption process is physical or chemical adsorption, which is mainly related to the type of adsorption reaction. The mean adsorption free energy (E) calculated from D-R isothermal model constant k_ad can be used to evaluate the mechanism of adsorption and confirm the adsorption reaction type, such as physical adsorption (1.0 kJ/mol < E < 8.0 kJ/mol), ion exchange and electrostatic attraction (8.0 kJ/mol < E < 16.0 kJ/mol), and chemisorption processes (E > 16.0 kJ/mol). The E values of TC and DOX shown in Fig. S4c,d and Table 1 were 0.79 and 2.5 kJ/mol, respectively, which were lower than 8 kJ/mol, indicating that the adsorption of Cu-MOF@Co-MOF to the two TCs was mainly dominated by physical adsorption^24,25.

The above adsorption isotherms verified that the adsorption process of Cu-MOF@Co-MOF to TCs contained both chemical and physical adsorption, while the heterogeneous adsorption on the surface of the Cu-MOF@Co-MOF comprised monolayer and multilayer adsorption.

Kinetics study

The adsorption kinetics were studied by measuring the adsorption equilibrium time and adsorption rate to elucidate the adsorption process. Adsorption kinetic models are commonly used to describe rapid adsorption processes, and their fitting data are used to analyze the rate-limiting steps and adsorption mechanisms. In order to further explore the adsorption mechanism, the adsorption kinetics of TCs were investigated using Cu-MOF@Co-MOF as adsorbent. Some widely used dynamic adsorption models such as pseudo-first-order, pseudo-second-order, Elovich, liquid film diffusion and intraparticle diffusion models^26–28 were used for fitting, shown in eqn. as follows.

$$\text{ln}\left({\text{q}}_{\text{e}}-{\text{q}}_{\text{t}}\right)={\text{lnq}}_{\text{e}}-{\text{k}}_1\text{t}$$ 9

$$\frac{\text{t}}{{\text{q}}_{\text{t}}}=\frac{1}{{\text{k}}_2{\text{q}}_{\text{e}}^2}+\frac{\text{t}}{{\text{q}}_{\text{e}}}$$ 10

$${\text{q}}_{\text{t}}=\frac{1}{\beta }ln\left(\alpha \beta \right)+\frac{1}{\beta }\text{ln}\left(\text{t}\right)$$ 11

$$ln\left(1-\text{F}\right)={\text{C}}_1-{\text{k}}_{\text{fd}}\text{t}$$ 12

$$\text{F}=\frac{{\text{q}}_{\text{t}}}{{\text{q}}_{\text{e}}}$$ 13

$${\text{q}}_{\text{t}}={\text{k}}_{\text{id}}·{\text{t}}^{0.5}+{\text{C}}_2$$ 14

where k₁ is the rate constant of the pseudo-first-order adsorption (min⁻¹), q_e is the adsorption capacity at equilibrium (mg/g), q_t is the adsorption amount at time (mg/g), k₂ (g/mg/min) is the rate constant of the pseudo-second-order equation, α is the initial adsorption rate of the reaction (mg/g/min), β is the rate constant related to surface coverage and chemical adsorption energy (g/mg), K_id is the intraparticle diffusion rate constant (mg/g/min^0.5), C₁ and C₂ is the constant related to thickness and boundary layer (mg/g), K_fd is the liquid film diffusion constant (min⁻¹).

The fitting curves for several kinetic models were shown in Fig. 5, and the corresponding parameters were summarized in Table 2. The adsorption equilibrium of the TCs basically reached within 15 min, and the adsorption rate was relatively fast in the first 5 min, and then slowed down. The fitting of the pseudo-first-order model indicates that the adsorption is dominated by physical adsorption, while the fitting of the pseudo-second-order kinetic model means that chemical adsorption is the main rate control step in the adsorption process. As shown in Fig. 5a,b and Table 2, the equilibrium adsorption capacity of TCs calculated by the pseudo-second-order model was relatively close to the experimental equilibrium adsorption capacity. The correlation coefficient (R²) of the pseudo-second-order model of the two TCs (≥ 0.995) was higher than that of other models, which confirmed that the pseudo-second-order model was more suitable for understanding the adsorption mechanism, indicating that the adsorption process was mainly controlled by chemisorption.

Figure 5.

The kinetic models fitting curves of the Cu-MOF@Co-MOF for TCs: (a) Pseudo-first-order model; (b) Pseudo-second-order model; (c) Elovich model; and (d) Liquid film diffusion model.

Table 2.

Adsorption kinetic models and its correlation fitting coefficients.

Samples	Pseudo-first-order kinetic			Pseudo-second-order kinetic
	k1 (min−1)	qe (mg/g)	R2	k2 (g/mg/min)	qe(mg/g)	R2
TC	0.208	112.74	0.875	0.015	476.19	0.999
DOX	0.047	119.55	0.929	0.018	370.37	0.995

Samples	Elovich			Liquid film diffusion
	α (mg/g/min)	β (g/mg)	R2	kfd	C	R2
TC	4.62 × 109	0.0475	0.965	0.166	− 1.609	0.907
DOX	9.11 × 104	0.0274	0.978	0.22	− 0.8207	0.944

Samples	First step			Second step
	kid1 (mg/g/min0.5)	C1 (mg/g)	R12	kid2 (mg/g/min0.5)	C2 (mg/g)	R22
TC	54.287	343.13	0.988	24.37	349.2	0.952
DOX	35.19	243.2	0.938	69.93	91.23	0.634

The Elovich model is based on the assumption that the heat of adsorption on the adsorbent surface decreases linearly with the increase of its surface coverage. This model is often used to investigate the existence of chemisorption processes on the non-uniform surface of adsorbents during water treatment. The fitting curve of the model generally presents two stages, which are the stage of rapid surface adsorption and slow diffusion speed, respectively. For Elovich kinetic model, the calculated values for the model-related parameters such as α (adsorption rate) and β (desorption rate) were shown in Table 2. The α-values of TC and DOX were 4.62 × 10⁹ and 9.11 × 10⁴ mg/g/min, respectively, confirming the faster absorption of TC. The β-values of TC and DOX were 0.0061 and 0.0059 g/mg, respectively, indicating that TC also had a faster desorption rate. The linear correlation coefficients of the fitting curves were 0.965 and 0.978 for TC and DOX, respectively, which verified the good fitting of the Elovich kinetic model (Fig. 5c).

The C value of the liquid film diffusion model reflected the external surface or instantaneous adsorption (Fig. 5d). The small C values shown in Table 2 indicated that the liquid film diffusion was not dominant during TCs adsorption²⁹. Intraparticle diffusion models is mainly applied to evaluate the diffusion mechanism during adsorption kinetics. For the adsorption kinetic model, the initial adsorption usually occurs on the outer surface of the adsorbent, and when all the surface adsorption sites reach saturation, the internal adsorption and intra-particle diffusion processes may occur. As shown in Fig. S5, the intraparticle diffusion model was used to fit the adsorption kinetic data, and two stages of adsorption of TCs in Cu-MOF@Co-MOF were obtained. The two stages were fast adsorption and tardiness adsorption, respectively. The first stage of rapid adsorption was external diffusion and the abundant adsorption sites on Cu-MOF@Co-MOF greatly promoted this rapid adsorption. In the second stage, the tardiness adsorption process was mainly dominated by intra-particle diffusion. The fitting data of the intraparticle diffusion models were linear, and the fitting curve did not cross the origin, indicating that intraparticle diffusion models had a significant effect on the adsorption of two TCs on Cu-MOF@Co-MOF, but it was not the only step rate-determing step.

Thermodynamics study

The determination of thermodynamic parameters is important to evaluate whether the adsorption process is exothermic or endothermic, which can reveal the law of adsorption process change with temperature. Thermodynamic studies were carried out by investigating the adsorption process at different temperatures (298 K, 308 K, 318 K). Effective factors that need to be quantified in thermodynamic studies include Gibbs free energy (ΔG°), entropy change (ΔS°), and enthalpy change (ΔH°), which are calculated as follows²⁷:

$${\text{K}}_{\text{c}}=\frac{{\text{q}}_{\text{e}}}{{\text{c}}_{\text{e}}}$$ 15

$$ln{\text{K}}_{\text{c}}=\frac{\mathrm{\Delta }\text{S}_{}^{_{}^{\circ }}}{\text{R}}-\frac{\mathrm{\Delta }\text{H}_{}^{_{}^{\circ }}}{\text{RT}}$$ 16

$$\mathrm{\Delta }\text{G}_{}^{_{}^{\circ }}=\mathrm{\Delta }\text{H}_{}^{_{}^{\circ }}-\text{T}\mathrm{\Delta }\text{S}_{}^{_{}^{\circ }}$$ 17

where, R (8.3145 J/mol/K) stands for the ideal gas constant, T (K) is the absolute temperature during adsorption, K_c represents the thermodynamic equilibrium constant. The C_e and q_e are the equilibrium concentration and adsorption capacity of TCs, respectively. The slope and intercept are calculated by lnK_c-1/T fitting to obtain ΔH° and ΔS°.

With the increase of temperature, the viscosity of the solution and the thickness of the boundary layer around the adsorbent decreases, which is conducive to accelerating the diffusion of the mass transfer process. If the adsorption process is mainly physical adsorption, the adsorption capacity will decrease when the temperature increased. The thermodynamic parameters were obtained by the slope and intercept of van’t Hoff equations and the relevant thermodynamic results were depicted in Fig. S6 and Table 3. It could be seen in Table 3 that the adsorption capacity decreased with the increase of temperature, and the Gibbs free energy (ΔG°) at all temperatures was negative, confirming that the adsorption process of TC and DOX on the Cu-MOF@Co-MOF could be carried out spontaneously. The value of enthalpy change (ΔH°) of the TC and DOX was − 14.507 and − 8.548 kJ/mol, respectively, indicating that the adsorption process of TC and DOX was an exothermic reaction. In addition, the calculated negative entropy change value (ΔS°) of TC and DOX indicated a decrease in irregularity between the adsorbent and solution, which led to a decrease in the removal process of TC and DOX³⁰. In summary, the process of TC adsorption and removal is a spontaneous, entropy-decreasing and exothermic reactions.

Table 3.

Thermodynamic parameters of TC and DOX adsorption onto Cu-MOF@Co-MOF at different temperatures.

Samples	T (K)	∆G° (kJ/mol)	∆H° (kJ/mol)	∆S° (J/mol/K)
TC	298.15	− 3.173	− 14.507	− 38.015
	308.15	− 2.793
	318.15	− 2.413
DOX	298.15	− 0.752	− 8.548	− 26.15
	308.15	− 0.490
	318.15	− 0.229

Possible adsorption mechanism

The study of the adsorption mechanism is of great benefit to the development of novel high-performance adsorbents, and can effectively promote the application of existing adsorbents. The Cu-MOF@Co-MOF before and after adsorption of TCs was characterized by XRD, FTIR, BET and XPS, and the adsorption mechanism was further discussed. As shown in Fig. 6a, compared with Cu-MOF@Co-MOF, the spectra of Cu-MOF@Co-MOF-TC and Cu-MOF@Co-MOF-DOX showed obvious TC and DOX fingerprint peaks (1600–750 cm⁻¹), respectively, indicating that TCs molecules were successfully adsorbed. Moreover, the stretching vibration peak of the N–H bond at 3433 cm⁻¹ shifted significantly after the adsorption of TCs, and all of them moved to the lower wavenumber to 3430 and 3420 cm⁻¹, indicating that hydrogen bonding was involved in the adsorption process. Besides, the transition of C=C aromatic stretching from 1613 to 1618 cm⁻¹ demonstrated Cu-MOF@Co-MOF could adsorb TCs by π-π interaction. The XRD spectra of Cu-MOF@Co-MOF before and after adsorption were shown in Fig. 6b. After TCs adsorption, the intensity of all the XRD characteristic peaks was greatly reduced, which might be due to the formation of coordination bonds between Co and Cu of Cu-MOF@Co-MOF and N and O atoms of TC molecules.

Figure 6.

(a) FTIR spectra; (b) XRD spectra; and (c–f) XPS spectra: (c) XPS Survey, (d) Cu2p, (e) Co2p, (f) C1s of Cu-MOF@Co-MOF before and after adsorption.

The adsorption mechanism was further confirmed by the XPS spectroscopy of Cu-MOF@Co-MOF before and after adsorption of TCs. As can be seen in Fig. 6c, all the spectra showed obvious characteristic peaks around 284.68 eV, 780.98 and 932.58 eV, representing C1 s, Co2p and Cu2p, respectively. After TCs adsorption, the height of characteristic peaks of each element was slightly different, mainly reflected in that the O1s peak was significantly stronger and the C1s, Co2p and Cu2p peak was weaker, which proved that TCs was successfully adsorbed on Cu-MOF@Co-MOF. The strength of O1s increased significantly after TCs adsorption, indicating that hydrogen bond interaction was involved in the adsorption process³¹. It was noteworthy that after TCs adsorption, the XPS spectra of Co2p and Cu2p peaks shown in the Fig. 6d,e exhibited a slight shift, indicating charge transfer between Cu-MOF@Co-MOF and TCs, which could benefit from the strong surface complexation between the Cu and Co unsaturated sites of the adsorbent and the oxygen (nitrogen) elements of TCs³². Cu and Co acquired electrons from the amino and hydroxyl groups of TCs and form Cu(Co)–O–C and Cu(Co)–N–C bonds via coordination bonds. Compared with pure Cu-MOF@Co-MOF, the binding energy of C1s peaks (C–C, C–O and C–O bonds) in Cu-MOF@Co-MOF–TC and Cu-MOF@Co-MOF–DOX was lower, and the XPS spectra showed obvious redshift, which might be caused by the π–π conjugated interaction between TC molecules and the benzene ring of Cu-MOF@Co-MOF (Fig. 6f)⁵.

The pore filling effect also occupied a large proportion in the adsorption process of TCs. The large specific surface area and abundant pore structure of Cu-MOF@Co-MOF can promote the adsorption of TCs²². We compared the BET area, pore volume and pore size of Cu-MOF@Co-MOF before and after TC adsorption to further confirm this adsorption mechanism. Before adsorption, the pore size of Cu-MOF@Co-MOF was mainly about 2.23 nm, which was larger than that of TCs (1.27 nm), indicating that TCs could shuttle to the inside of the pores. As shown in Table 4, the pore size of Cu-MOF@Co-MOF increased to 13.08–17.47 nm and the pore volume decreased by 0.16–0.17 nm after the adsorption of TCs, which confirmed that the TCs entered the adsorption site of Cu-MOF@Co-MOF through the mesoporous channel and filled the inside of the hole. Furthermore, the surface area of Cu-MOF@Co-MOF decreased significantly from 558.25 to 30–45 m²/g after adsorption, verifying that pore filling played a key role in this process.

Table 4.

Pore character of Cu-MOF@Co-MOF before and after adsorption.

	Surface area (m2/g)	Pore volume (cm3/g)	Pore size (nm)
Before adsorption	558.25	0.324	2.23
After adsorption
TC	42.59	0.161	13.08
DOX	30.71	0.148	17.47

From the characterization spectra before and after adsorption combined with the adsorption model data, it could be concluded that the adsorption between TCs and Cu-MOF@Co-MOF was enhanced by the synergistic effect of hydrogen bonding, π–π interaction, coordination, electrostatic interaction and pore filling effect due to the structure of multiple phenolic hydroxyl, carbonyl and amino functional groups, abundant conjugated benzene rings of TCs and the unique superior properties of MOFs (Fig. 7).

Figure 7.

The adsorption mechanism of TCs on Cu-MOF@Co-MOF.

Comparison of adsorption property with that of other adsorbents

The adsorption properties of TCs were investigated based on Cu-MOF, Co-MOF and Cu-MOF@Co-MOF adsorbents, respectively. As shown in Fig. S6, the adsorption capacity of Cu-MOF for TC and DOX was 81.69 and 142.70 mg/g, respectively, while the adsorption capacity of Co-MOF was higher than that of Cu-MOF (282.44 and 229.39 mg/g, respectively), which was much lower than that of Cu-MOF@Co-MOF for TCs under the same conditions. Furthermore, thanks to the high specific surface area of Cu-MOF@Co-MOF and the synergistic effect of multi-target adsorption in the bimetallic center, the Cu-MOF@Co-MOF exhibited superior removal properties for TCs, demonstrating the great advantages of Cu-Co bimetallic MOFs in the field of TCs removal.

In addition, the Cu-MOF@Co-MOF adsorbent used in this work was also compared with other reported adsorbents to further evaluate its application prospects. Table 5 provided a comparison of Cu-MOF@Co-MOF with other adsorbents reported recently for the removal of TC and DOX. It could be seen that compared with most of the adsorption materials reported in 2023, the adsorption time of the prepared Cu-MOF@Co-MOF composites for TC and DOX was greatly shortened, and the adsorption capacity was significantly higher than that of the existing adsorption capacity, indicating that Cu-MOF@Co-MOF composites had great application potential in the removal of TCs in polluted water bodies.

Table 5.

Comparison of the adsorption of TC and DOX by Cu-MOF@Co-MOF and other reported adsorbents.

	Adsorbent	Adsorbent dosage (mg)	Adsorption time (min)	qm (mg/g)	References
TC	Fe3O4@SiO2–MPS-g-P	200	240	18.51	3
	Humic acid-loaded Fe3O4	100	1440	243.05	6
	ZIF-67-derived Co@CNTs	20	180	174.1	33
	Ce-UiO-66	20	180	86.95	34
	TiO2/BiOI	20	90	24.83	31
	Polycationic straw	120	70	30.09	2
	Magnetic functionalized carbon microsphere	40	120	94.63	35
	HEC-GO/Fe-Zn	50	20	22.5	36
	AC@ZIF-8	30	150	35.64	37
	Cu-MOF@Co-MOF	10	5	434.78	This work
DOX	HEC-GO/Fe-Zn	50	20	16.55	38
	Porous hypercrosslinked polymers	40	600	166.82	39
	CuO/textile waste derived biochar composite	20	120	450.15	40
	Polyvinyl alcohol/porous carbon composite hydrogels	70	–	17.07	41
	Magnetic functionalized carbon microsphere	40	120	67.58	35
	MIL-53/Fe3O4	20	30	322	24
	Cu-MOF@Co-MOF	5	15	476.19	This work

TCs removal from actual water

Since the composition of actual wastewater is more complex than that of deionized water, the removal efficiency of adsorbents for simulated wastewater is often different from that of actual wastewater under the same conditions. Therefore, it is valuable to investigate its adsorption properties in actual wastewater in order to investigate the practical application potential of Cu-MOF@Co-MOF. In this work, river water and tap water were selected as actual samples for adsorption experiments. Under the same conditions, the removal rates of TC and DOX from the two actual water samples by Cu-MOF@Co-MOF were 85.49–87.35% and 98.78–99.03%, respectively. The adsorption capacity was lower than that of deionized water, but the satisfactory removal effect was still maintained. The results showed that Cu-MOF@Co-MOF could effectively remove TC and DOX from actual wastewater. It will be a promising direction to construct MOF-on MOF structure to achieve multi-target synergistic, low-cost, rapid and efficient removal of pollutants.

Conclusion

In this work, a novel MOF-on MOF structure was prepared by loading Co-MOF on the surface of Cu-MOF via adsorption and self-assembly method, which were used to remove two representative TCs from aqueous solution. We demonstrated that the prepared Cu-MOF@Co-MOF could be used as excellent TCs adsorbents thank to their large specific surface area, unique porous properties and synergistic coordination adsorption of bimetallic cluster sites. The adsorption behavior results showed that even a small amount of Cu-MOF@Co-MOF adsorbent could achieve rapid and efficient removal of TCs, thereby significantly reducing the economic cost and time of water treatment. The fitting data of the adsorption model showed that there was both chemical and physical adsorption during the adsorption process, and the thermodynamic adsorption data confirmed that it was a spontaneous and exothermic process. It is worth mentioning that the adsorption capacity of Cu-MOF@Co-MOF for TC and DOX in water was up to 434.78 mg/g and 476.19 mg/g, respectively, which was significantly higher than that of mono-metal center MOFs in a wide pH range benefitting from the abundant synergistic effect. The adsorption mechanism was mainly attributed to the coordination, π–π interaction, hydrogen bonding, electrostatic interaction and pore filling effect between Cu-MOF@Co-MOF and TCs. The cyclic structure of TCs and the hydrophobic organic backbone of MOFs promoted π–π interactions, while the hydroxyl and amino groups of TCs were conducive to the formation of coordination with metal clusters. In summary, these superior properties indicate that Cu-MOF@Co-MOF is a potential adsorbent for the removal of pollutant in the water environment system. This work also provides a new idea for the study of efficient and economical MOFs-based removal materials.

Author contributions

H.D. designed the study. J.Y. carried out the experiment. Q.Z., Y.Y., X.Z., L.W. and Y.G. participated in the experiment. J.L., Y.Z., Y.S. and Y.D. prepared tables. S.D., Q.Z., and H.D. wrote the main manuscript text. All authors reviewed the manuscript.

Data availability

The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-024-67986-8.