Hydrolytically stable nanosheets of Cu–imidazolate MOF for selective trapping and simultaneous removal of multiple heavy metal ions

Abstract

The global water security crisis necessitates the pursuit of cost-effective and sustainable water purification solutions that can be deployed at large scales. The current class of adsorbents are predominantly tailored towards highly specific unary metal ion systems, largely constraining their applicability to realistic scenarios. Thus, to create effective decontamination processes, it is essential to understand the adsorbent's potential in multicomponent systems. Here, hydrolytically stable copper–imidazolate (Cu–Im) metal–organic framework (MOF) nanosheets are presented as molecular sieves for trapping multiple heavy metal contaminants. These nanosheets are synthesized rapidly (within 30 minutes), at ambient temperature and atmospheric pressure, obviating the necessity for post-modification or functionalization. Adsorption experiments targeting four heavy metal ions revealed remarkable adsorption capacities: 492 mg g⁻¹ for Pb(II), 327 mg g⁻¹ for Cd(II), 233.1 mg g⁻¹ for Mn(II) and 72.6 mg g⁻¹ for Ni(II). Notably, in multi-component and industrial effluent settings, Cu–Im exhibited superior removal efficiency compared to other non-functionalised MOFs and classes of porous materials such as zeolites. Furthermore, Cu–Im demonstrated robust recyclability, maintaining removal efficiencies of 99% for Pb(II), 90% for Mn(II), 60% for Cd(II) and 70% for Ni(II) over three adsorption cycles. Compatibility assessments using a model aquatic organism cell model, the zebrafish ZF4 cell line, revealed no significant reduction in cell viability up to 48 hours. Cu–Im's stability is benchmarked against a global spectrum of 125 MOFs using machine learning algorithms. Moreover, Cu–Im complies with seven of the twelve principles of ‘Green Chemistry’, establishing itself as a sustainable, scalable, stable and reusable solution to a pressing global challenge.

---

Environmental significance

Toxic metal contamination poses a grave threat to global water security, adversely affecting aquatic ecosystems and human well-being. Conventional strategies fail to perform in realistic scenarios and largely lack scalability. Here, we introduce a metal–organic framework-derived nanosheet, rapidly synthesized in just 30 minutes at room temperature and pressure. This material exhibits remarkable selectivity, capturing Pb(II), Cd(II), Mn(II), and Ni(II) from water, with high adsorption capacities. Notably, it is non-toxic and adheres to seven key ‘Green Chemistry’ principles, showcasing its sustainability. This scalable, stable, sustainable and robust material offers a promising outlook for future decontamination strategies. Its attributes offer hope for mitigating the multifaceted challenges presented by toxic metal contamination, ensuring a safer aquatic environment and healthier communities worldwide.

1 Introduction

A significant proportion of the world's population (∼80%) is exposed to serious threats related to water security.¹ Due to the massive investments that water management infrastructures (such as wastewater treatment) require, developing nations and their associated bio-diversities are disproportionately affected from an economic, healthcare and social perspective.¹ Most contaminants exist in trace quantities, making their detection and subsequent removal a paramount need,^2,3 with effective solutions demanding a synergy between policymakers, practitioners and researchers.⁴ Particularly challenging contaminants in this regard are heavy metal ions due to the health hazards associated with their exposure, even in trace quantities. Human exposure to these contaminants can lead to carcinogenicity, vascular complications, contact dermatitis and respiratory infections, among a range of other hazards.⁵ While a majority of heavy metal water contamination is of anthropogenic origin, some natural sources such as weathering can contribute to the problem through the interactions of rocks with the surrounding environment.^5,6 In their landmark review on water purification technologies, Shannon and colleagues³ recognise that “it is extremely challenging to develop robust, low-cost, effective means of chemical sensing relevant to the water contamination problems of developing nations”, with most existing solutions leveraging sophisticated laboratory instrumentation, making them difficult to scale. From a heavy metal removal perspective, there are adsorption-based separation techniques, membrane-based filtration techniques, chemical-based separation, and electrically induced techniques. Each of them have unique advantages as summarised by Qasem et al.⁷ Among the several techniques developed in recent decades, adsorption-based processes stand out owing to their relative simplicity, cost-effectiveness, ease of function and absence of secondary pollutant production.⁸ While the ideal adsorbent should maximise interactions with adsorbates, have good stability in working conditions and have low production costs to ensure the feasibility of large-scale synthesis, the majority of them fail to satisfy most of these criteria, suffering from low adsorption capacity, limited reversibility and poor selectivity.^9,10 Specifically with regards to selectivity, biosorbents show promise, due to a varienty of functional groups present on their surface. For instance, Cao et al. immobilised Saccharomyces cerevisiae cells on a natural regenerated cellulose nanofibers network for Pb removal in the human body.¹¹

Among the array of contemporary materials serving as adsorbents, metal–organic frameworks (MOFs) stand out as a compelling paradigm. Their noteworthy characteristics, including high surface area, surface functionality, modularity, and structural diversity, position MOFs as an intriguing choice for applications centred around adsorption.¹² The diversity of the chemical landscape of organic linkers gives rise to tuneable host–guest chemistry, favouring MOFs for heavy metal capture due to a chelation effect between the MOF components and the heavy metal ions.¹³ MOFs and their modified forms have shown exceptional removal capabilities for a broad spectrum of pollutants such as heavy metals,¹⁴ anionic pollutants¹⁵ and organic contaminants.¹⁵ Traditionally, MOFs have struggled with long-term stability, often failing under operating conditions. However, strategies such as doping¹⁶ and functionalisation¹⁷ have sufficiently addressed these problems, making these materials fit for investigation in biological and environmental applications. Specifically with regards to heavy metal capture, over the past decade, there have been notable developments. For instance, Zhong et al.¹⁸ proposed a broad-spectrum heavy metal trap (BS-HMT) composed of ethylenediaminetetraacetic acid (EDTA) incorporated into a robust MOF (MOF-808). EDTA is a non-specific chelating agent with a high coordination number, allowing it to bind with various heavy metal ions. EDTA was immobilized onto MOF-808 to facilitate the easy recovery and reusability of the adsorbent. Similar developments have been the subject of several excellent reviews.^5,9,13

While impressive metrics have been reported for the adsorption of heavy metals from water using MOFs, most studies do not take into account crucial considerations that would make them applicable in practically relevant scenarios. In particular: (i) most reported studies focus on the removal of a single metal ion and show poor performance for capturing other metals. In industrial activities, it is often the case that multiple metal contaminants are introduced into the ecosystem in parallel. In such scenarios, materials that can tackle multiple contaminants are ideal.¹⁸ (ii) To enhance removal efficiency (RE), complex functional groups are introduced to MOFs.^18–20 Functionalisation often adds an extra step to the synthesis procedure, increasing the associated costs and timescales, making these formulations difficult for large-scale production and deployment. (iii) The sustainability of the MOF (and its functional groups) is often ignored in such discussions, including requirements for water, energy, and solvents. (iv) To utilise these materials for environmental remediation applications, the material itself must not be harmful to the ecosystem. (v) To be applicable in realistic scenarios, the material should retain its performance over multiple cycles, in the presence of co- and counter-ions in real effluents.²¹

This study focuses on the fabrication of 14 nm thick non-functionalized copper–imidazolate-based molecular sieves, denoted as Cu–Im. These sieves demonstrate outstanding stability under operating conditions,²¹ sustainability, reusability, and cost-effectiveness, particularly when viewed from a large-scale production standpoint. The synthesis of Cu–Im is a single-step process, requiring mild conditions and ambient pressure, taking only 30 min (before activation), making it feasible to be produced at a large scale. With a predicted pore size²² of 2.4–4.4 Å, ideal for heavy metal capture,^8,18 the developed nanomaterial is stable in water up to 48 h, across acidic and basic conditions, and at higher temperatures (up to 42 °C). Cu–Im nanosheets are capable of removing Pb(II), Mn(II), Cd(II) and Ni(II) (consistently ranked among the most prevalent and harmful water contaminants worldwide)^23,24 with high adsorption capacities (492.6 mg g⁻¹ for Pb(II), 327.8 mg g⁻¹ for Cd(II), 233.1 mg g⁻¹ for Mn(II) and 72.6 mg g⁻¹ for Ni(II)), demonstrated in real effluents, and across multiple cycles. A chemical landscape analysis among 125 MOFs indicates that Cu–Im shares chemical similarities with high kinetic stability MOFs, suggesting further applications in humid conditions. According to the ISO guidelines for Biological Evaluation of Medical Devices,²⁵ Cu–Im is non-toxic during the timeframe of adsorption, while also adhering to 7 of the 12 principles of green chemistry²⁶ and is thus emerging as a sustainable solution to heavy metal ion pollution in a practically relevant context.

2 Experimental

2.1. Synthesis of Cu–Im MOF nanosheets

Cupric nitrate trihydrate (Cu(NO₃)₂·3H₂O) (Merck), and 2-methyl imidazole (2-MI) (Sigma-Aldrich) of analytical grades were used for the synthesis of Cu–Im MOF. Cu(NO₃)₂·3H₂O (1.619 g) was dissolved in 12 mL of ultrapure water and 2-MI (3.319 g) was dissolved in 48 mL ultrapure water and stirred separately for 30 min at 25 °C. Following this, both solutions were mixed, magnetically stirred vigorously, and washed three times with ultrapure water. The obtained precipitate was vacuum-dried at 60 °C in an oven. Subsequently, the obtained Cu–Im MOF powder was activated in a vacuum oven for 12 h at 110 °C to evaporate the trapped solvents from the pores of the MOF.

2.2. Characterization of MOF nanosheets

Phase identification of Cu–Im was performed using X-ray diffraction (XRD) (Bruker D8-XRD) (Cu-Kα, 40 kV, 30 mA, 2θ range of 5–70°). The morphology and size of the synthesized MOF particles were determined using scanning electron microscopy (SEM) (JEOL JSM7900F). To calculate the thickness of the MOF nanosheets, atomic force microscopy (AFM) analysis was performed in the tapping mode (Bruker Nano wizard Sense AFM). Energy-dispersive X-ray (EDX) analysis was performed to check the elemental composition and distribution. To ascertain the degradation temperature of the Cu–Im MOF, thermo-gravimetric analysis (TGA) (PerkinElmer TGA4000) was carried out from 25 °C to 500 °C, at a heating rate of 10 °C. Fourier-transform infrared (FTIR) spectrum (Spectrum Two, PerkinElmer) was collected in the attenuated total reflectance (ATR) mode in the range of 400–4000 cm⁻¹, with 4 cm⁻¹ resolution to identify the characteristic molecular bonds and functional groups present in Cu–Im. Cu content in the MOF powder (n = 3) was measured using inductively coupled plasma–optical emission spectroscopy (ICP-OES) (AVIO 200, Perkin Elmer) after complete digestion of the powder in aqua regia at 80 °C. Isoelectric point (IEP) measurement was carried out for the nanosheets at 1000 mg L⁻¹ using Zetasizer (NanoZS, Malvern instrument). The surface chemistry and oxidation state of elements present in the MOF were determined using X-ray photoelectron spectroscopy (XPS) (Thermofisher Scientific, Nexsa base). Al-Kα (1486.6 eV) source was used as the incident X-ray source. High-resolution XPS spectra of C 1s, O 1s, N 1s, Cu 2p, Cd 3d, Mn 2p, Pb 4f and Ni 2p were captured at constant analyser energy (CAE) with pass energy and step size of 50 eV and 0.100 eV, respectively. XPS data were acquired with 700 s dwell time and 0.1 eV step size. The binding energies for all the elements and samples were corrected as per an adventitious carbon reference at 284.6 eV to compensate for the shift caused by the charges on the sample surface.

2.3. Hydrolytic stability studies

Stability studies of Cu–Im nanosheets were conducted in aqueous medium of varying pH and temperature. Cu–Im MOF powder (1000 mg L⁻¹) was immersed in ultrapure water for different exposure times (2 h, 24 h, and 48 h), at varying pH (3, 6.5 and 9) and temperatures (25 °C and 42 °C). After exposure, the aliquot was centrifuged, and the supernatant was acidified using 2% HNO₃ and then analysed using ICP-OES to measure Cu release from the MOF. The dried MOF powders were analysed using SEM and XRD to observe morphological and structural changes. Zeta potential (Malvern NanoZS) measurements were performed for the nanosheets to identify the IEP.

2.4. Multi-element removal studies and reusability studies

Stock solutions were prepared using salts of the respective elements: MnCl₂·4H₂O, Ni(NO₃)₂·6H₂O, Cd(NO₃)₂, and Pb(NO₃)₂. For adsorption studies and kinetics measurements, Cu–Im nanosheets (1000 mg L⁻¹) were placed in glass vials, each containing 5 mL aq. 100 mg L⁻¹ (for Mn(II), Ni(II) and Cd(II)) and 500 mg L⁻¹ for Pb(II) solution of individual heavy metals ions. The dispersion of Cu–Im nanosheets and the contaminant (individual metal ions aqueous solution) were kept at 25 °C and stirred at 230 rpm. At specific time points, aliquots (2 mL) were taken and centrifuged. The supernatant from the centrifuged solution was acidified using 2% HNO₃ and analysed using ICP-OES to measure the individual metal concentrations (Mn, Ni, Cd, Pb and Cu) to assess any release during exposure. For ICP-OES analysis, calibration standards (0.1 mg L⁻¹ to 10 mg L⁻¹) were prepared from 1000 mg L⁻¹ stock of standard solution of individual metal ions.

The adsorption isotherm measurements were performed separately for each of the metal ions using 1000 mg L⁻¹ Cu–Im and 1–500 mg L⁻¹ of heavy metals (Mn(II), Ni(II) and Cd(II)) and 1–1000 mg L⁻¹ for Pb(II). The MOF powder was dispersed in individual metal ion aqueous solutions (with different initial concentrations) and stirred for 24 h at 25 °C. Subsequently, aliquots (2 mL) were taken, and the metal concentrations were determined using ICP-OES. The effect of adsorbent concentration on individual heavy metal removal was investigated using a range of MOF concentrations (10–1000 mg L⁻¹) while keeping the heavy metal concentrations fixed at 100 mg L⁻¹ (for each metal ion). The adsorption experiments were conducted at three different temperatures (25 °C, 35 °C, and 40 °C) to study the temperature-dependent adsorption by Cu–Im and at different values of pH (ranging from 3–7).

The heavy metal adsorption capacity of Cu–Im was also evaluated in the presence of counter ions (Mg²⁺, Ca²⁺, and Na⁺) for a period of 24 h by using a powder concentration of 1000 mg L⁻¹, and adsorbate concentration of 100 mg L⁻¹ (1.71 mM of Na⁺, 0.99 mM of Ca²⁺, 1 mM of Mg²⁺) for each metal ion. All the equations used for batch adsorption experiments, kinetics studies, adsorption isotherms and adsorption thermodynamics studies are mentioned in the ESI† (Section S1). Cu–Im nanosheets were re-analysed post-adsorption to determine the changes in their morphology, structure, and chemistry with the help of SEM–EDS, XRD, and XPS.

2.5. Acellular Cu release and in vitro cytotoxic assay

To investigate the cytotoxic profile of Cu–Im, a cell counting kit-8 (WST-8) assay was performed on ZF4 cells which are embryonic fibroblast cells from zebrafish. Briefly, 8 × 10³ cells were seeded in a 96-well plate and were incubated at 28 °C overnight for the growth of the cells. Then, the cells were exposed to Cu–Im at a range of concentrations (0–500 μg mL⁻¹) for up to 72 h. PBS was used as a negative control and 2% H₂O₂ was applied to cells as a positive control. After 24 h, 48 h, and 72 h of incubation with the treatments, the cells were washed with PBS to remove the Cu–Im followed by an addition of 100 μL fresh cell medium per well. Later, 10 μL CCK-8 test solution (WST-8 solution) per well was added and incubated for 3 h under dark conditions. Post incubation, the 96 well plates were subjected to analysis at 460 nm. The percentage of cell viability was calculated with respect to the negative control. Confirmation of the source of the toxic response induced by Cu–Im requires a clear demonstration of Cu release from the framework structure. To assess the release of dissolved Cu during the exposure, 100 and 200 μg mL⁻¹ of Cu–Im MOFs were exposed to complete media (DMEM, 10% FBS) for a duration of 72 h. Briefly, Cu–Im at each concentration were placed inside a dialysis bag (MWCO: 3.5 kDa) with 50 mL of complete media outside the bag (sink). Aliquots were withdrawn from outside the bag at specific time points and digested using an ashing mixture (nitric acid and peroxide) to obtain a clear solution. The digested samples were then subjected to ICP-OES (Avio 200, Perkin Elmer, USA) to measure the concentration of Cu in the samples.

2.6. Machine learning for benchmarking Cu–Im MOF hydrolytic stability relative to other MOFs

A database of metallic centres and linkers was compiled for 125 single metal, single linker MOFs, with their corresponding water stability according to the framework described by Burtch et al.²⁷ The data was featurized using an open-source cheminformatics library to accurately capture chemical information across length scales and dimensions.²⁸ For the metal centre, its corresponding electronic configuration was used, while for the linker, we calculated 197 molecular descriptors spanning electro-topological states, accessible surface areas and general features such as molecular weights. This high-dimensional data was visualized in two-dimensional maps using the t-distributed stochastic neighbourhood embedding (t-SNE) technique which preserves pairwise distances in two dimensions, thereby helping to identify clusters of MOFs sharing chemical similarity. These clusters were assigned a hue according to their water stability.

3 Results and discussions

3.1. Characterisation of Cu–Im nanosheets

Cu-based MOFs are broadly appealing owing to the relative ease of generating numerous geometric complexes through simple interactions with various organic ligands.²⁹ Cu–Im was prepared by reacting cupric nitrate trihydrate with 2-methyl imidazole (2-MI). The synthesis was a single-step process, performed at room temperature under ambient pressure. The MOF was activated under vacuum at 383 K to remove trapped guest species. The crystallinity and phase purity of the material (following activation) was confirmed using powder X-ray diffraction (PXRD) (Fig. 1a), which showed characteristic diffraction peaks of Cu–Im at 14.7°, 29.8°, 33.2° and 47.7°. SEM revealed a uniform nanosheet-like structure with an average lateral dimension of 378 ± 128 nm (Fig. 1b), and AFM revealed a thickness of 14 ± 4 nm for individual nanosheets (Fig. 1c). FTIR was used to analyse the structure of the MOF by confirming the presence or absence of specific functional groups (Fig. 1d). Peaks at 1423 cm⁻¹ and 1307 cm⁻¹ are attributed to C–N stretching, and a peak at 1150 cm⁻¹ is attributed to C–H stretching, thus confirming the presence of the 2-MI linker.³⁰ Peaks at 1570 cm⁻¹ and 1618 cm⁻¹ confirm the presence of –C=N, while the adsorption band observed at a lower wavenumber of 431 cm⁻¹ evidences the formation of Cu–N,³¹ confirming interactions between Cu and 2-MI. Fig. S1† shows the thermogravimetric analysis (TGA) of Cu–Im, where three distinct regions of behaviour are observed. Within the range of 30 °C to 240 °C, the material demonstrates stability, followed by a notable sudden mass loss (≈20%) at 250 °C. This initial weight reduction likely corresponds to the removal of unreacted or weakly bonded 2-MI, potentially accompanied by the breakdown of certain functional groups within the MOF structure. Furthermore, a secondary degradation phase becomes evident around 390 °C, primarily attributed to the degradation of the organic linker. This decomposition results in the generation of carbonaceous residues while marking the breakdown of the structural framework.³²

Fig. 1 Characterization of Cu–Im nanosheets. a. PXRD diffraction pattern of the synthesized Cu–Im nanosheets. b. SEM image showing the size and morphology of the Cu–Im nanosheets. c. AFM images showing the thickness of the nanosheets. d. FTIR spectra showing the functional groups present in the Cu–Im nanosheets. e. Elaboration of BET surface area using the BETSI approach. Pchip stands for ‘Piecewise Cubic Hermite Interpolating Polynomial’, used for interpolating values between measured points. For details into the calculation of the plot, readers are referred to Osterrieth et al.³³ f. A schematic illustration of pore limiting diameter as calculated using the machine learning approach proposed by Pétuya et al.²²

Since the discovery of MOFs, the Brunauer–Emmett–Teller (BET) method and associated surface area have emerged as a central characterisation requirement.^34,35 However, a recent inter-laboratory study conducted by Osterrieth et al.³³ expressed concern over the reproducibility of BET areas, especially for microporous materials. For a less ambiguous assignment of BET area, we followed the method proposed in the same study (called ‘BET Surface Identification’ or ‘BETSI’), which first fits a particular region of the isotherm using an ordinary least-squares regression, checks the validity of the fit against Rouquerol criteria,³⁶ and repeats the analysis for all combinations of points on the isotherm. Based on the optimal fit using BETSI, Cu–Im showed a BET-specific surface area of 18 m² g⁻¹ (Fig. 1e, further elaborated in Fig. S2†). Post Pb(II) adsorption, Cu–Im showed a BET surface area of 22 m² g⁻¹. A machine-learning model²² was utilised to predict the pore-guest accessibility and pore-limiting diameter (PLD) of MOFs. The model predicted Cu–Im being guest accessible (having a pore diameter > 2.4 Å) with 80.5% certainty, having small pores with sizes ranging from 2.4–4.4 Å with 76.4% certainty (Fig. 1f), indicating microporosity. These findings agree with a previous report on the synthesis of Cu–Im.³⁰

3.2. Hydrolytic stability of Cu–Im MOF nanosheets

The hydrolytic stability of the nanosheets was studied under varying environmental conditions (pH, temperature) using ICP-OES, SEM, XRD and IEP measurements. SEM indicates that the morphology of the framework is perfectly intact for up to 2 h of water exposure, transforming into a nano-spindle form at 24 h and 48 h (Fig. 2a). Despite morphological transformations, XRD indicates that the structure and crystallinity of the MOF are largely retained during the timescales of the experiments (Fig. 2b) suggesting structural stability, while ICP-OES reveals an ∼0.2% and 4% Cu release in ultrapure water at 2 h and 24 h, respectively (Fig. 2c) indicating high compositional stability. In comparison, other common Cu-based MOFs (such as HKUST-1) show significantly higher rates of Cu leaching (∼18%) across similar periods and media, under similar conditions.^16,21 Lower Cu release throughout the metal ion binding experiments is indicative of better sustainability, as lower amounts of ion leaching serve to reduce the associated environmental burden. Similar morphological changes were observed upon exposure to acidic (pH 3) and basic (pH 9) conditions (Fig. 2a), while compositional stability was maintained, with 5.6% Cu release under acidic conditions and 5% release under basic conditions (Fig. 2c). Notably, no changes in morphology, crystallinity or Cu release rates were observed at a higher temperature (42 °C) (Fig. 2a–c). The observed stability of the Cu–Im nanosheet is in line with the hard soft acid base (HSAB) theory, which deliberates that ligands with strong basicity (such as 2-MI, pK_a ≈ 14.2) form strong metal–ligand bonds with divalent soft acids (such as Cu(II)).^17,37 In a previous report, a positive correlation between the N-content in the linker, and the strength of the bond between the metal and the linker was shown.²¹

Fig. 2 Stability of Cu–Im nanosheets. a. SEM analysis showing the morphological transformation by 24 and 48 h in water, and at acidic and basic pHs. b. PXRD analysis showing the structural integrity of Cu–Im nanosheets over time and at different pHs and elevated temperature. c. ICP-OES analysis showing the release profile of Cu in the different environments tested (in ultrapure water, acidic (pH 3), basic (pH 9) and warm water (42 °C)). d. Isoelectric point measurements (Cu–Im nanosheets concentration: 1000 mg L⁻¹).

3.3. Heavy metal adsorption performance

3.3.1 Effect of pH and kinetics, isotherm and thermodynamics of metal adsorption. A multi-element batch adsorption study was performed to investigate the adsorption characteristics of the Cu–Im MOF in water containing practically relevant concentrations of Pb(II), Mn(II), Cd(II) and Ni(II). Since the pH of the system influences the competition between H⁺ and OH⁻ for the available sorption sites and affects metal speciation and the point of zero charge (pH_pzc), pH optimisation studies were performed. The RE of the metal ions was measured across pH values ranging from pH 3 to pH 7 (an industrially relevant range). No significant differences in RE were observed in this pH range, further validating the utility of the Cu–Im MOF across diverse industrial scenarios. Cu–Im has an IEP of 10.65 (Fig. 2d), implying that at a pH below this value, the framework has a positive surface charge, while in the pH range of 3–7, metal ions would be in the form M²⁺ or M(OH)⁺ [M = Pb, Mn, Cd, Ni]. The data (Fig. 3e) suggests that the metal ion adsorption process favours M²⁺ ions existing in a more complex R–NH₂M²⁺ form, implying that the process is dominated by covalent complex formation between the amine group present in the framework and the metal ions, as opposed to electrostatic interactions.^38–40 To understand the affinity of the different metal ions to the Cu–Im MOF, distribution coefficients (K_D) were calculated and Log K_Dversus adsorbent concentration plots were generated (Fig. 3a–d).

Fig. 3 Effect of Cu–Im nanosheet dosage on a. Pb(II) b. Mn(II) c. Cd(II) and d. Ni(II) adsorption and affinity of metal ions with Cu–Im nanosheets shown using distribution coefficient (K_d) (adsorbent dosage: 10 mg L⁻¹ to 1000 mg L⁻¹). e. Effect of pH on heavy metal ion adsorption. f. R_L values as calculated from the Langmuir equation ([M²⁺] = 100 mg L⁻¹, adsorbent dosage: 1000 mg L⁻¹).

A positive correlation between Cu–Im dosage and affinity was observed, which was attributed to increased adsorption sites.⁴¹ A similar trend was observed on increasing the contact time, with higher RE at higher contact times. After 48 h of exposure, RE of 99.9%, 98.7%, 83% and 61% for Pb(II), Mn(II), Cd(II) and Ni(II), respectively was shown by the MOF nanosheet (Fig. 4a). This wide range of RE (61–99.9%) indicates a generally strong affinity of adsorption of these metal ions, a consequence of pore accessibility and inter-molecular interactions.⁸Table 1 shows the q_e (adsorption capacity) after fitting pseudo-first-order and second-order kinetics models. On comparing the regression coefficients, for Cd(II), the pseudo-second-order model is the best fit (Fig. S3b†), while for the remaining ions, the first-order kinetic model is the best fit (Fig. S3a†). Since equilibrium was attained at 24 h, the subsequent adsorption isotherms and thermodynamic studies were performed at the same time point.

Fig. 4 Cu–Im as a molecular sieve for heavy metal trapping. Removal of Pb(II), Mn(II), Ni(II) and Cd(II) individually from deionized water, a. RE vs. time plot. b. Adsorption isotherm study. c. Effect of coexisting ions on the removal of metals using Cu–Im MOF (1.71 mM of Na(I), 1 mM of Mg(II) and 0.99 mM of Ca(II)). d. Simultaneous removal of Pb(II), Mn(II), Cd(II) and Ni(II) from deionized water and in effluent (initial metal concentration: 100 mg L⁻¹). e. Reusability measurement of Cu–Im MOF for metal ions adsorption over 3 cycles (MOF concentration: 1000 mg L⁻¹, initial metal ion concentration: 1–500 mg L⁻¹ for Mn(II), Ni(II) and Cd(II) and 1–1000 mg L⁻¹ for Pb(II), at pH 5 and temperature: 25 °C) (statistical analysis done using a two sample t-test where p ≤ 0.05: *; p ≤ 0.01: ** and p ≤ 0.001: ***).

Table 1 Adsorption kinetics parameters for metal ions adsorption

	Pseudo-first-order			Pseudo-second-order
	K 1 (min^−1)	q e (mg g^−1)	R ^2	K 2 (g mmol^−1 min^−1)	q e (mg g^−1)	R ^2
Pb(II)	0.0029	479.3	0.947	0.003	526.3	0.763
Mn(II)	0.00101	82.92	0.9866	4.05 × 10^−5	106.95	0.9734
Cd(II)	0.00141	74.52	0.8810	4.21 × 10^−5	81.56	0.9564
Ni(II)	6.38 × 10^−4	56.80	0.9537	5.58 × 10^−5	58.99	0.9385

---

To understand the adsorbate–adsorbent interactions and to calculate the maximum adsorption capacity, equilibrium adsorption studies were conducted (Fig. 4b). It was observed that there was an increase in the adsorption capacity of the nanosheets with an increase in metal ion concentration from 1–1000 mg L⁻¹, suggesting the metal ion concentration gradient to be the driving force for metal adsorption. The adsorption isotherm data was fitted with Langmuir and Freundlich models (Table 2); adsorption data for all ions showed a higher correlation with the Langmuir model compared to the Freundlich model, suggesting monolayer adsorption of metal ions on energetically identical and active sites of the MOF surface (Fig. S3c and d†). With the Langmuir isotherm, the dimensionless constant separation factor (RL) can be used to gauge the strength of interactions between the adsorbate and adsorbent. Based on RL values, adsorption isotherms can be categorised as unfavourable (R_L > 1), linear (R_L = 1), favourable (0 < RL < 1) and irreversible (R_L = 0). The data suggests that all four metal ions show a favourable adsorption isotherm (Fig. 3f) with Cu–Im. From the Langmuir model, the theoretical adsorption capacity was calculated to be 492 mg g⁻¹ for Pb(II), 327.8 mg g⁻¹ for Cd(II), 233.1 mg g⁻¹ for Mn(II) and 72.6 mg g⁻¹ for Ni(II). According to Nieboer and Richardson,⁴² the propensity of covalent bond and complex formation between a metal and ligand is positively correlated to a covalent index (Xm²/R). Trends in covalent index [Pb(II) (6.41) > Cd(II) (2.71) > Ni(II) (2.52) > Mn(II) (1.99)]^37,43,44 partially match our experimentally observed trends [Pb(II) > Cd(II) > Mn(II) > Ni(II)]. Contradicting observations for Mn(II) and Ni(II) may be attributed to the higher hydration energy of Ni(II) (−2105 kJ mol⁻¹) compared to Mn(II) (−1841 kJ mol⁻¹), making dehydration more difficult for Ni(II).⁴⁵ To gauge the feasibility of the adsorption process, the adsorption behaviour was studied at varying temperatures (Fig. S3e†). Thermodynamic parameters were extracted from linearly fitted Van't Hoff plots (Fig. S3e†), shedding light on the adsorption feasibility and spontaneity. Thermodynamic parameters (Table 3) indicate that Pb(II) adsorption is exothermic, while Cd(II), Mn(II) and Ni(II) adsorption is endothermic.

Table 2 Adsorption isotherm parameters for metal ions removal

	Langmuir model			Freundlich model
	q max (mg g^−1)	K L (L mg^−1)	R ^2	K f (L mg^−1)	n	R ^2
Pb(II)	492.61	0.863	0.999	110.15	3.52	0.655
Mn(II)	233.1	0.0464	0.9907	13.87	1.9350	0.8970
Cd(II)	327.8	0.0123	0.9085	5.37	1.3465	0.9240
Ni(II)	72.6	0.1161	0.9906	3.94	1.8441	0.9120

---

Table 3 Thermodynamic parameters for metal ions removal

	ΔH° (kJ mol^−1)	ΔS° (J mol^−1 K^−1)	ΔG° (kJ mol^−1)
			25 °C	35 °C	40 °C
Pb(II)	−75.54	−231.62	−6.51	−5.36	−3.04
Mn(II)	29.50	86.38	3.872	3.012	2.582
Cd(II)	28.47	96.19	−0.136	−1.096	−1.576
Ni(II)	31.011	97.66	1.90	0.930	0.442

---

3.3.2 Selectivity, reusability, and removal mechanisms. Removal studies performed in the presence of equal concentrations (100 mg L⁻¹) of co-existing ions (Ca(II), Mg(II) and Na(I)) showed no significant change in the RE of Pb(II), Mn(II) and Ni(II), while a slight decrease was observed in the RE of Cd(II) (from 62% to 56%), implying high selectivity (Fig. 4c). To gauge the practical applicability of the Cu–Im MOF, simultaneous removal experiments of all four ions from ultrapure water and industrial effluents were performed. From Fig. 4d, it is clear that Pb(II) showed the highest RE (≈99%), which was unaffected due to the presence of Cd(II), Mn(II) and Ni(II) even in the industrial effluent system. On the contrary, Cd(II) showed a 63% reduction in RE, Mn(II) a 9% reduction and Ni(II) an 18% reduction from the mixed ion solution compared to their individual binding efficiencies. This reduction is due to a higher selectivity for Pb(II), due to its smaller hydration energy and hydration radius compared to the other ions. Pb(II) ions are therefore able to swiftly diffuse onto the framework surface and into the pores due to lower diffusion resistance.⁴⁵ In addition, Pb(II) has a higher electronegativity than Cd(II), Mn(II) and Ni(II), making it more strongly attracted to the adsorption sites in the MOF to form complexes.⁴³

From a practical viewpoint, to be economically sustainable, adsorbents need to be reusable over multiple cycles. To study the reusability of Cu–Im, the adsorption performance of the framework in mixed-ion systems across multiple cycles was performed. The framework showed RE of 99% for Pb(II), 90% for Mn(II), 60% for Cd(II) and 70% for Ni(II) up to 3rd cycle (Fig. 4e). Data for Pb(II) and Cd(II) removal till 5 cycles is shown in ESI† (Fig. S5). The decrease in RE for Cd(II) and Ni(II) can again be attributed to the higher selectivity for Pb(II), which occupies most of the adsorption sites. The Cu–Im framework reported here showed a higher adsorption capacity in multi-component systems compared to other MOFs reported in the literature (Table S1,†Fig. 6c). Some of the recently developed MOF hybrids (e.g. silk nanofibers-ZIF hybrid membranes) showed more than 99% removal for Cu(II), Cd(II), Pb(II) and Hg(II) within 5 min of treatment time.⁴⁶

Potential mechanisms of metal adsorption and the stability of the framework post-adsorption were studied using XPS, SEM–EDX and XRD. While EDX mapping showed the presence and distribution of Pb(II), Cd(II), Mn(II) and Ni(II) on the surface of the framework post-adsorption (Fig. S6†), XRD revealed that the structure and crystallinity of Cu–Im was retained, re-enforcing its reusability (Fig. 5a). For in-depth insights into the possible adsorption mechanisms, XPS spectra for each element in the framework were measured. Fig. S7a† shows the pre-adsorption spectrum of the framework, with no additional signals except for those of Cu 2p, O 1s, N 1s and C 1s. Post-adsorption signals for Pb(II), Mn(II), Cd(II), and Ni(II) show respective peaks for Pb 4f, Mn 3d, Cd 3d and Ni 2p (Fig. S7b–d†). The spectrum of the Pb(II) adsorbed framework showed two main peaks at 137.6 (Pb 4f_7/2) and 142.5 eV (Pb 4f_5/2) belonging to the divalent states of Pb (Fig. S8a†). The Mn adsorbed framework showed peaks for Mn 2p at 642.2 eV and 653.2 eV corresponding to the Mn³⁺ species of Mn 2p_3/2 and Mn 2p_1/2 (Fig. S8b†). Similarly, the deconvoluted spectrum for the Cd adsorbed framework showed Cd 3d peaks at 405.3 eV and 411.9 eV, assigned to Cd 3d_5/2 and Cd 3d_3/2 respectively (Fig. S8c†). For the Ni adsorbed framework, upon deconvolution, the spectrum showed two strong peaks assigned to Ni 2p_3/2 and Ni 2p_1/2, which separated into two spin–orbit doublets at 855.5 eV and 873.1 eV indicating the presence of Ni³⁺, and at 856.8 and 875.2 corresponding to the Ni²⁺ states of Ni 2p (Fig. S8d†). All peaks post-adsorption are well identified and not present in the pristine Cu–Im samples, confirming the successful adsorption.

Fig. 5 Post-adsorption studies. a. XRD analysis of Pb(II), Mn(II), Ni(II) and Cd(II) adsorbed onto Cu–Im nanosheets. XPS high-resolution N1s spectra of b. pristine Cu–Im nanosheets, c. Pb adsorbed Cu–Im, d. Mn adsorbed Cu–Im, e. Cd adsorbed Cu–Im and f. Ni adsorbed Cu–Im.

Fig. S9† shows Cu2p peaks before and after metal adsorption. Characteristic Cu peaks, centred at 933.1 eV and 952.9 eV with a spin–orbit splitting of 19.8, are attributed to Cu⁺ in Cu₂O-specific Cu 2p_3/2 and Cu 2p_1/2 states, respectively. The peaks present at 934.8 eV and 955 eV correspond to Cu(II) in the Cu 2p_3/2 and Cu 2p_1/2 states, respectively, while the remaining small peaks are satellites. Similarly, the deconvoluted local XPS spectra for O 1s in Cu–Im showed peaks at 529.5 eV and 531.6, ascribed to metal oxide and lattice oxygen species (Fig. S10a†). However, the O 1s peak post-adsorption showed an extra peak around 532 eV, denoting the presence of surface hydroxyl species of oxygen (Fig. S10b–d†). In the deconvoluted peaks of N 1s, two peaks at 398.6 and 399.7 were observed, denoting the presence of –N= and –NH– of imidazole (Fig. 5b–f). For Pb(II), Mn(II), Cd(II) and Ni(II) adsorbed samples, an additional peak was identified at 400.1 eV, 400.9 eV, 400.6 eV and 401 eV, respectively, attributed to protonated N groups of the amine (Fig. 5c–f). The active sites for metal adsorption (Pb(II), Mn(II), Ni(II) and Cd(II)) in nanosheets is the –NH– group. Through XPS analysis it could be observed that the atomic percent of N is 23.61%, suggesting its contribution as active sites for adsorption.

Establishing the direct correlation between the content of functional groups and adsorption capacity would involve other parameters of the different metal ions such as the covalent index, electronegativity, hydration energy and hydration radius. Metal ions could also accumulate inside the pores of the MOF nanosheets. Post-adsorption analysis of nascent Cu–Im and metal ion adsorbed Cu–Im suggests that adsorption is likely due to the formation of metal complexes (strong coordination) between the metal ion and the imino group of 2-MI. A similar mechanism and strong coordination between metal ions and the imino group of 2-MI has been reported in the literature.^47,48

3.4. In vitro cytotoxicity assay: Cu–Im MOF sustainability

Sustainability is a crucial consideration while designing materials for environmental remediation applications. Hence, the Cu–Im framework was exposed to a ZF4 cell line (purchased from ATCC) derived from the embryo of the zebrafish, to test the MOF's compatibility with the environment. In the timescale of the adsorption experiments (24–48 h), there was no significant reduction in cell viability (<10%) (Fig. 6a). However, at a longer timescale of 72 h, we witnessed a ∼35% reduction in ZF4 cell viability at each exposure concentration (Fig. 6a). The observed toxicity can be positively correlated to the release of Cu(II) ions from the framework (Fig. S11†). Based on the data (Fig. S11†), it is also deduced that the leaching of ions occurs in a time-dependent manner, in line with experimental observations of stability. Our findings are consistent with those of a recent report,⁴⁹ wherein it was inferred that higher bioaccumulation of Cu and in situ dissolution of ionic Cu from Cu-based nanomaterials induces higher toxicity compared to direct exposure to ionic Cu. A crucial point to note here is that ZF4 cells have a competent lysosomal exocytosis mechanism, along with a better ability of GSH buffering, which significantly alleviates the toxic response of released Cu ions.⁴⁹ From a safety and sustainability perspective, the framework is non-toxic to the environment along the timescales of adsorption, however, at higher timescales, the framework is moderately toxic, possibly due to the ability of Cu(II) ions to overpower the defence mechanisms of ZF4 cells. Various chemical agents, like trientine and penicillamine, offer the potential to chelate and form complexes with Cu(II). These agents, when applied at specific concentrations, can effectively chelate Cu released from Cu Im MOFs, mitigating their potential hazards.

Fig. 6 Sustainability studies and benchmarking of Cu–Im. a. Cell viability studies of Cu–Im against the ZF4 cell line (n = 3). Statistical analysis was done using a two-sample t-test where ***p ≤ 0.001. b. Compliance of Cu–Im with the 12 principles of green chemistry for membrane technologies where those indicated in blue were fully met, while additional work is needed to achieve those in orange and red. c. Benchmarking of Cu–Im against a non-exhaustive set of adsorbents. To note, most reported adsorbents remove single contaminants, are functionalized, and are non-reusable. d. Benchmarking Cu–Im on a non-exhaustive global landscape of MOFs with varying water stability to capture chemical similarities using t-SNE. Raw data for this plot is provided in the ESI.†

Most studies about adsorption-based processes and water treatment tend to focus purely on binding metrics and associated mechanisms. A crucial aspect that is often ignored is the sustainability of the material or technology. Considering the entire lifecycle of the material, several aspects may be considered non-sustainable, such as the manufacturing processes or raw materials that rely on petroleum-based products and the use of hazardous solvents. Many of these remediation materials pose burdens of disposal and challenges to the well-being of workers and in a broader context, to the environment.²⁶ Under these considerations, guided by the philosophy of ‘green chemistry’,⁵⁰ Cu–Im satisfies 7 of the 12 principles of green chemistry for membrane-based technologies (Fig. 6b) as described by Xie et al.²⁶ Namely, (i) we used benign chemical reagents and products, (ii) we did not use any catalyst for the synthesis of the material, (iii) the material has weak noncovalent interactions (as is evidenced by the data on stability), (iv) we used low toxicity, inert, and abundant solvents (ultrapure water), (v) we employed atom-, step- and solvent-economical processes (minimal waste), (vi) we chose the material with the intent of maximising function and minimising hazards, and (vii) we attempted to maximise performance with the minimal use of material (high adsorption capacity, multi-element removal, non-functionalisation of MOF) with the broader goal of increased commercial viability, without increasing environmental burden. In this regard, we made an initial attempt to create MOF-magnetic nanoparticle hybrids which could be used to effectively separate the MOF from the suspension using an external magnetic field (Fig. S12†). This would ensure that the MOFs themselves do not become an environmental pollutant through dispersal into the environment and subsequent uptake by biota.

3.5. Machine learning approach: benchmarking Cu–Im on a global landscape of MOFs

Through experimental approaches, machine learning and computational algorithms for the elaboration of the BET area, we demonstrated the scalable synthesis, stability, heavy metal adsorption capacity and environmental compatibility of Cu–Im. The building blocks of this MOF have a useful set of chemical features that endow it with superior functionality. We thus conducted an exercise to benchmark this material on a global landscape of MOFs with a varying range of water stabilities. Given that this Cu–Im MOF is being proposed for use in a water-remediation application, this exercise would help us understand whether Cu–Im has an explainable set of chemical features given its stability, helping us recommend rational design choices for the development of MOFs with similar or improved stability and functionality. We first compiled a list of 125 single metal, single linker MOFs from a larger published database,²⁷ found the respective building blocks (metal centre and linker), and featurised these building blocks based on the chemistry of the metal and the linker, using standard cheminformatics packages (a full list of features and the corresponding raw data has been provided as ESI†).²⁸ Upon featurization, t-distributed stochastic neighbourhood embedding (t-SNE) was performed to visualise these high-dimensional features on a 2D map.⁵¹ Since t-SNE preserves pairwise distances, ensuring similar structures are mapped close to each other on the map, we were able to visualise the chemical neighbourhood of Cu–Im, to deduce whether it stands out compared to MOFs with similar features.

In their criteria for the classification of MOFs based on water stability, Burtch et al.²⁷ categorised MOFs as having (i) thermodynamic stability, if stable after long-term exposure to aqueous solutions, (ii) high kinetic stability, if stable after exposure to high humidity conditions, but decomposing after short-exposure times in liquid water, (iii) low kinetic stability, if stable under low humidity conditions, and (iv) low stability, if it quickly breaks down after any moisture exposure. Based on the t-SNE plot (Fig. 6d), Cu–Im lies in a cluster of high kinetic stability MOFs based on its chemical features. This observation agrees with our experimental measurements, wherein the Cu–Im MOF shows water stability for up to 48 h. A notable observation from this plot however is that there are no observable clusters of thermodynamically stable and unstable MOFs based on experimentally reported data, suggesting that it is difficult to define a set of chemical features that correspond to these classes of MOFs. On the other hand, clusters of high and low kinetic stability MOFs imply that it is easier to offer design criteria for the development of MOFs in these categories. Moreover, purely based on where Cu–Im lies on this landscape of chemical features, it can be deduced that the framework would find utility in applications with high humidity conditions. These inferences, however, have certain associated shortcomings; the dataset of MOFs used is modest, the classes are imbalanced as there are very few MOFs reported with exceptionally high and exceptionally low water stability, and the visualisation technique itself is highly dependent on the parameters provided.

4 Conclusions

Guided by the overarching principles of sustainability and green chemistry, we developed low-cost, stable, scalable, and reusable Cu–Im nanosheets serving as effective molecular sieves for trapping multiple heavy metal contaminants from water. The Cu–Im MOF nanosheets demonstrated stability under varying conditions pertaining to pH and temperature, making them relevant to diverse industrial scenarios. Cu–Im showed high single-component adsorption capacities (492–72.6 mg g⁻¹) for Pb(II), Cd(II), Mn(II) and Ni(II). Moreover, during the timescales of adsorption, the material is non-toxic and adheres to relevant principles of sustainability and green chemistry. The material is built on relatively inexpensive precursors and has an extremely simple synthesis protocol, requiring mild conditions and short timescales. All aspects put together Cu–Im emerges as a highly scalable, practical solution to address a leading global concern of toxic metal contamination in water.

There are however some limitations due to the small crystal sizes that prevented the resolution of the crystal structure. Structure resolution would have provided additional mechanistic insights into MOF–guest interactions, opening the door to computational approaches such as molecular dynamics and grand canonical Monte Carlo simulations. The current experimental focus lies on controlling these crystal sizes to resolve the structure of the material. Advances in artificial intelligence and machine learning provide another avenue for crystal structure prediction. The machine learning algorithms leveraged in this study aligned well with experimental observations. Moreover, unsupervised algorithms helped benchmark the material on a global landscape, agreeing with experimental observations. A key point to convey through this study is compatibility with industrial expectations. Most studies reporting excellent metrics are rarely scalable due to complexities in synthesis, costs and timescales involved, and challenges with inherent material safety and sustainability. Materials discovery, design and optimization philosophies going forward should consider these aspects to ensure that the environmental remediation strategies developed can be readily employed at large scales.

Author contributions

The manuscript was written through the contributions of all authors. PB, PM and PG: conceptualization, methodology, writing – original draft, review & editing. SC, BKS and LW: methodology, experimentation, writing – review & editing. DM, IL, SKM: conceptualization, methodology, writing – review & editing.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

The authors are thankful to Prof. Abhay Gautam and Prof. Vimal Mishra for their constructive suggestions and experimental troubleshooting. The funding acquired from the Gujarat State Biotechnology Mission, India (GSBTM/BR-1/292051) and Royal Society Fellowship, UK (IES\R2\232158) were instrumental in executing this project and the authors gratefully acknowledge it.

References

1. C. J. Vörösmarty, P. B. McIntyre, M. O. Gessner, D. Dudgeon, A. Prusevich, P. Green, S. Glidden, S. E. Bunn, C. A. Sullivan, C. Reidy Liermann and P. M. Davies, Nature, 2010, 467, 555.
2. D. S. Sholl and R. P. Lively, Nature, 2016, 532, 435.
3. M. A. Shannon, P. W. Bohn, M. Elimelech, J. G. Georgiadis, B. J. Marĩas and A. M. Mayes, Nature, 2008, 452, 301.
4. K. Bakker, Science, 2012, 337, 914.
5. P. A. Kobielska, A. J. Howarth, O. K. Farha and S. Nayak, Coord. Chem. Rev., 2018, 358, 92.
6. F. R. Siegel, Environmental geochemistry of potentially toxic metals, Springer Berlin, Berlin, ISBN: 978-3-540-42030-9, 2001.
7. N. A. A. Qasem, R. H. Mohammed and D. U. Lawal, npj Clean Water, 2021, 36, 4.
8. X. Zhao, X. Yu, X. Wang, S. Lai, Y. Sun and D. Yang, Chem. Eng. J., 2021, 407, 127221.
9. E. Svensson Grape, et al. , Nat. Water, 2023, 1, 433.
10. M. Mon, R. Bruno, J. Ferrando-Soria, D. Armentano and E. Pardo, J. Mater. Chem. A, 2018, 6, 4912.
11. S. Cao, S. Wu, X. Dong, M. Long, H. Lin, F. Liu, Y. Wu, Z. Zhao, C. Chen and H. Deng, Adv. Funct. Mater., 2023, 33, 2215059.
12. H. Furukawa, K. E. Cordova, M. O'Keeffe and O. M. Yaghi, Science, 2013, 341, 1230444.
13. F. Zadehahmadi, N. T. Eden, H. Mahdavi, K. Konstas, J. I. Mardel, M. Shaibani, P. Chakraborty Banerjee and M. R. Hill, Environ. Sci.: Water Res. Technol., 2023, 9, 1305–1330.
14. N. Sharma, A. K. Dey, R. Y. Sathe, A. Kumar, V. Krishnan, T. J. Dhilip Kumar and C. M. Nagaraja, Catal. Sci. Technol., 2020, 10, 7724.
15. H. Kaur, S. Walia, A. Karmakar, V. Krishnan and R. R. Koner, J. Environ. Chem. Eng., 2022, 10, 106667.
16. P. Goyal, A. Paruthi, D. Menon, R. Behara, A. Jaiswal, K. V. A. Kumar, V. Krishnan and S. K. Misra, Chem. Eng. J., 2022, 430, 133088.
17. M. Ding, X. Cai and H. L. Jiang, Chem. Sci., 2019, 10, 10209.
18. Y. Peng, H. Huang, Y. Zhang, C. Kang, S. Chen, L. Song, D. Liu and C. Zhong, Nat. Commun., 2018, 9, 187.
19. B. Aguila, Q. Sun, J. A. Perman, L. D. Earl, C. W. Abney, R. Elzein, R. Schlaf and S. Ma, Adv. Mater., 2017, 29, 1.
20. M. Mon, J. Ferrando-Soria, T. Grancha, F. R. Fortea-Pérez, J. Gascon, A. Leyva-Pérez, D. Armentano and E. Pardo, J. Am. Chem. Soc., 2016, 138, 7864.
21. P. Goyal, D. Menon, P. Jain, P. Prakash and S. K. Misra, Sep. Purif. Technol., 2023, 318, 123941.
22. R. Pétuya, S. Durdy, D. Antypov, M. W. Gaultois, N. G. Berry, G. R. Darling, A. P. Katsoulidis, M. S. Dyer and M. J. Rosseinsky, Angew. Chem., Int. Ed., 2022, 61, 1–6.
23. P. B. Tchounwou, C. G. Yedjou, A. K. Patlolla and D. J. Sutton, Heavy metal toxicity and the environment, Exper. Suppl., 2012, 101, 133–164,  DOI:10.1007/978-3-7643-8340-4_6 , PMID: 22945569; PMCID: PMC4144270.
24. V. Kumar, R. D. Parihar, A. Sharma, P. Bakshi, G. P. Singh, A. S. Bali, I. Karaouzas, R. Bhardwaj, A. K. Thukral, Y. Gyasi-Agyei and J. Rodrigo-Comino, Chemosphere, 2019, 236, 124364.
25. Food and Drug Administration, 2016. Use of International Standard ISO 10993-1, “Biological Evaluation of Medical Devices Part 1: Evaluation and Testing within a Risk Management Process”. Guidance for Industry and Food and Drug Administration Staff. Food and Drug Administration, Center for Devices and Radiological Health.
26. W. Xie, T. Li, A. Tiraferri, E. Drioli, A. Figoli, J. C. Crittenden and B. Liu, ACS Sustainable Chem. Eng., 2021, 9, 50.
27. N. C. Burtch, H. Jasuja and K. S. Walton, Chem. Rev., 2014, 114, 10575–10612.
28. G. Landrum, RDKit: A software suite for cheminformatics, computational chemistry, and predictive modeling, Components, 2011.
29. A. Burg and D. Meyerstein, Adv. Inorg. Chem., 2012, 64, 219.
30. M. F. Navarro Poupard, E. Polo, P. Taboada, A. Arenas-Vivo, P. Horcajada, B. Pelaz and P. Del Pino, Inorg. Chem., 2018, 57, 12056.
31. M. Kim, S. Hwang and J. S. Yu, J. Mater. Chem., 2007, 17, 1656.
32. S. Loera-Serna, M. A. Oliver-Tolentino, M. D. López-Núñez, A. Santana-Cruz, A. Guzmán-Vargas, R. Cabrera-Sierra, H. I. Beltrán and J. Flores, J. Alloys Compd., 2012, 540, 113.
33. J. W. M. Osterrieth, et al. , Adv. Mater., 2022, 34, 2201502.
34. D. A. Gómez-Gualdrón, P. Z. Moghadam, J. T. Hupp, O. K. Farha and R. Q. Snurr, J. Am. Chem. Soc., 2016, 138, 215.
35. M. Thommes, K. Kaneko, A. V. Neimark, J. P. Olivier, F. Rodriguez-Reinoso, J. Rouquerol and K. S. W. Sing, Pure Appl. Chem., 2015, 87, 1051.
36. J. Rouquerol, P. Llewellyn and F. Rouquerol, Stud. Surf. Sci. Catal., 2007, 160, 49.
37. Z. Mahdi, Q. J. Yu and A. El Hanandeh, Sep. Sci. Technol., 2019, 54, 888.
38. L. Jin and R. Bai, Langmuir, 2002, 18, 9765.
39. N. Esfandiar, R. Suri and E. R. McKenzie, J. Hazard. Mater., 2022, 423, 126938.
40. S. M. Ali, M. A. El Mansop, A. Galal, S. M. Abd El Wahab, W. M. T. El-Etr and H. A. Zein El-Abdeen, Sci. Rep., 2023, 13, 9466.
41. C. Yu, Z. Shao and H. Hou, Chem. Sci., 2017, 8, 7611–7619.
42. E. Nieboer and D. H. S. Richardson, Environ. Pollut., Ser. B, 1980, 1, 3–26.
43. Y. Li, L. Jiang, X. Li, Y. Hu and J. ya Wen, Chem. Res. Chin. Univ., 2013, 29, 568.
44. Q. Kong, S. Preis, L. Li, P. Luo, C. Wei, Z. Li, Y. Hu and C. Wei, Sep. Purif. Technol., 2020, 232, 115956.
45. D. W. Smith, J. Chem. Educ., 1977, 1, 1–3.
46. D. Sun, G. Meng, S. Sun, W. Guo, J. Hai, J. Su, Y. Song, F. Chen and B. Wang, Environ. Sci.: Nano, 2021, 8, 3408.
47. D. Roy, S. Neogi and S. De, J. Hazard. Mater., 2021, 403, 123624.
48. K. Zhu, C. Chen, H. Xu, Y. Gao, X. Tan, A. Alsaedi and T. Hayat, ACS Sustainable Chem. Eng., 2017, 5, 6795.
49. X. Wang and W. X. Wang, Environ. Pollut., 2022, 292, 118296.
50. P. Anastas and N. Eghbali, Chem. Soc. Rev., 2010, 39, 301.
51. L. V. D. Maaten and G. Hinton, J. Mach. Learn. Res., 2008, 9, 2579–2605.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3en00754e

---