A nature-inspired hydrogen-bonded supramolecular complex for selective copper ion removal from water

Design, synthesis, characterization, and adsorption tests

We sought to design a HOIF through a new, simple, and scalable synthetic approach. In particular, the reaction of Zn(NO₃)₂, 2-methylimidazole (Hmim), and the copper chelator salicylaldoxime (H₂salox) in water around 50 °C results in the formation of a zinc imidazole salicylaldoxime supramolecule, hereafter referred to as ZIOS. The material forms rapidly as small crystals in relatively high yield (76%) following the combination of separate mixtures of H₂salox/Zn(NO₃)₃ and H₂salox/mim. Synchrotron single-crystal X-ray diffraction (SC-XRD) data (Fig. 1a) revealed that the ZIOS structure features Zn₃(Hsalox)₄(mim)₂ (i.e., Zn₃(C₆H₄CHNOHO)₄(CH₃C₃H₂N₂)₂) trinuclear units, wherein H₂salox and 2-methylimidazole are deprotonated and bonded directly to Zn²⁺ nodes through tetrahedral and pentahedral coordination by oxygen and nitrogen, respectively. Intermolecular hydrogen bonds between the hydroxyimino oxygen of H₂salox and the pyrrolic nitrogen of 2-methylimidazole lead to the formation of a two-dimensional supramolecular network. The ZIOS structure is essentially nonporous, in contrast to MOFs and ZIFs²⁸, with a Brunauer–Emmett–Teller (BET) surface area of ~12–14 m² g⁻¹ confirmed by both BET measurements at different activation conditions, e.g. 100 and 200 °C and grand canonical Monte Carlo simulations (Fig. 2c and Supplementary Fig. 2; c.f. BET surface area ~1200–1492 m² g⁻¹ for ZIF-8 determined in this work, see the Supporting Information). Even still, we find that the Cu²⁺ uptake capacity of ZIOS is 40% higher than that of a sample of ZIF-8 prepared in our hands, and the supramolecular structure exhibits unprecedented rapid adsorption kinetics (via infra). This surprising performance suggests the role of a chemisorption-based uptake mechanism, based on the coordination of Cu²⁺ by the aldoxime and histidine-inspired imidazole pendant groups present in the host structure. Powder X-ray diffraction characterization of ZIOS revealed that the bulk, as-synthesized material is highly crystalline (Fig. 2d). Thermogravimetric analysis data (Supplementary Fig. 6) collected under a N₂ atmosphere revealed minimal (<5%) weight loss at ~50 °C, likely due to traces of uncoordinated water, followed by a two-step decomposition beginning at 250 °C, as a result of ligand degradation. Differential scanning calorimetry (DSC) analysis of ZIOS (Supplementary Fig. 7) further revealed exothermic peaks between ~220 and 250 °C corresponding to salicylaldoxime degradation.

Motivated by how nature performs complex functions based on the propagation of hydrogen-bonding networks, we examined whether ZIOS could function effectively in the capture of copper ions from water. Adsorption tests were performed on as-synthesized samples of ZIOS and ZIF-8 (see the Supporting Information) to observe their adsorption kinetics, isotherms, distribution coefficient, and selective performance in complex aqueous environments containing copper in the presence of several competing ions. Details of these experiments are listed in the Supporting Information. Samples of ZIOS and ZIF-8 were initially exposed to an aqueous CuCl₂ solution (895 ppm, i.e., 425 ppm Cu²⁺, selected to benchmark the adsorptive performance of our adsorbents against ZIF-8 reported by Zhang et al.¹⁸) at the room temperature, and within the first 30 min both samples underwent a color change, from white to pale blue or green, respectively (Fig. 2a). This change is indicative of different interactions between Cu²⁺ and two frameworks, despite the presence of imidazole ligands in both structures. Scanning electron microscopy (SEM) images of ZIOS revealed that the material partially retains its rod-like structure following 24 h of exposure to Cu²⁺ (Fig. 2a, b), and energy-dispersive X-ray spectroscopy characterization revealed an even distribution of ions across the rod (Fig. 2b). In other domains of ZIOS-Cu, we observed a two-dimensional sheet-like structure (Supplementary Fig. 3) which is likely due to the sequence of sonication, vortex, copper adsorption, centrifugation, washing, and drying that the material underwent. The single-crystal structure of ZIOS-Cu was thus not obtained due to the lack of sufficiently large crystals that satisfy the minimal size requirement of the single-crystal X-ray diffraction technique. In contrast, after 24 h the ZIF-8 particles decrease in size and in general take on a different morphology (Fig. 2a). Crystal-to-crystal transformation due to postsynthetic treatments were previously reported for ZIF-8²⁹. Powder X-ray diffraction characterization revealed that both materials undergo a transition to a new crystalline state upon Cu²⁺ uptake (Fig. 2c).

Adsorption kinetics data for ZIF-8 and ZIOS in the presence of 895 ppm CuCl₂ solution are shown in Fig. 3. In the case of ZIF-8 exposure, the copper concentration drops to ~42 ppm within the first 20 min before increasing slightly and reaching a steady state value of ~115 ppm after 75 min (Fig. 3a). Thus, although ZIF-8 initially adsorbs more than 80% of the copper ions present in solution, some of this copper is then gradually released (Fig. 3a). This behavior and the nonlinear pseudo-second-order adsorption kinetics (Fig. 3b) are indicative of weak interactions between copper(II) and ZIF-8. In contrast, addition of ZIOS to a ~425 ppm Cu²⁺ solution results in a significant drop in ion concentration to a steady state concentration <1.5 ppm in ~30 min (Fig. 3a), close to the 1.3 ppm upper limit for copper ions in drinking water. Further analyses revealed that copper capture by ZIOS occurs with exceptionally fast kinetics, with a pseudo-second-order adsorption rate constant of 155.3 ± 26.5 mg mg⁻¹ min⁻¹ (Fig. 3b). This rate is 50 times faster than that determined by us for ZIF-8 under the same conditions (3.4 ± 1.6 mg mg⁻¹ min⁻¹) and is significantly larger than rates reported for other state-of-the-art copper capture adsorbents (Supplementary Table 4). The positive linear dependence of the (t/q_t) as a function of time shown in Fig. 3b is characteristic of pseudo-second-order adsorption of copper ions in ZIOS, confirming a chemisorption-governed process³⁰. Notably, the fast adsorption kinetics exhibited by ZIOS are accompanied by a substantial adsorption capacity: at equilibrium, the amount of adsorbed copper ions (q_e) was found to be 162.9 ± 9.4 mg g⁻¹. For ZIF-8, a smaller equilibrium adsorption capacity of 116.4 ± 5.7 mg g⁻¹ was determined from our measurements (material surface area = 1492 m² g⁻¹). We note this capacity for ZIF-8 is much lower than the uptake of ~800 mg g⁻¹ reported previously (reported surface area = 1340 m² g⁻¹)¹⁸, although the authors in ref. ¹⁸ determined the material uptake after only 30 min of exposure to a copper ion solution, and our data indicate that equilibrium does not occur until ~75 min. As a comparison at this 30 min mark, the uptake of our sample of ZIF-8 was 151.7 ± 4.3 mg g⁻¹. It is worth to note that the ratio of mass concentration (in mg L⁻¹) of Cu: ZIF-8 used in our study is ~0.17:1, whereas that reported in Zhang and coworkers’ work is 1.26:1. We also note that adsorption capacity can be sensitive to slight differences in material preparation and activation conditions.

We further equilibrated ~2.5 mg ZIOS with solutions featuring a wide range of copper concentrations (~0.45 to ~420 ppm, equilibration time of 24 h). Each solution was then analyzed by inductively coupled plasma-optical emission spectrometry (ICP-OES) to determine the quantity of copper adsorbed, and the resulting data are presented in the adsorption isotherm in Fig. 3c. The data could be fit using both single- and dual-site Langmuir models, and the corresponding fitting parameters and correlation coefficients are tabulated in Supplementary Table 5. Both fits indicate the presence of a strong adsorption site with a moderate saturation capacity q_sat of ~208 mg g⁻¹ (Supplementary Table 5), which is eight times higher than that of a commercially available thiol-functionalized resin, Duolite GT73 (BET surface area of 47.3 m² g⁻¹ and capacity of 25 mg g⁻¹ at an equilibrium concentration of 160 ppm, see Supplementary Table 4)³¹. While this capacity for ZIOS is less than that of PAF-1-SMe (q_sat = 662 mg g⁻¹)⁴, our material is advantageous in that it can be prepared through a much simpler synthetic route. In addition, copper uptake kinetics in ZIOS are ~30 times faster than in PAF-1-SMe⁴. Using the Langmuir constant, K_L, obtained from single-site Langmuir model¹⁶, and Blanchard’s³² original proposed equation (see the Supporting Information), we determined that the Gibbs free energy of copper ion adsorption by ZIOS is −27.23 kJ mol⁻¹, indicating that uptake is also thermodynamically favorable.

We investigated the copper ion selectivity of ZIOS and our ZIF-8 sample by evaluating ion uptake from synthetic water samples containing a mixture of Cu²⁺, Na⁺, Ca²⁺, Fe²⁺, Fe³⁺, Mn²⁺, and Ni²⁺ ions, prepared from chloride and nitrate salts of these cations (10 ppm concentration for each ion, solution pH of 3.94). Notably, even in the presence of these other ions, ZIOS adsorbed ~98.0 % of the copper(II) in solution, nearly double the amount adsorbed by ZIF-8 (52.3%), and uptake of Na⁺, Ca²⁺, Mn²⁺ was less than 10% in ZIOS and ZIF-8. However, we found that ZIOS also adsorbs substantial Fe²⁺/Fe³⁺ and Ni²⁺, namely ~93.9% and ~92.2%, respectively, compared with only ~56.1%, and ~8.3% for ZIF-8, respectively (Fig. 3e). This behavior is perhaps unsurprising given that salicylaldoxime is an excellent chelator for transition metal ions³³. Although these results indicate that ZIOS will exhibit low copper ion selectivity in solutions that also contain iron and nickel ions, they also indicate that ZIOS may have potential for applications that target more than one toxic cation and/or the selective removal of other metal ions from select mixtures. If the limited selectivity of ZIOS is indeed governed by the salicylaldoxime ligand, this behavior should be tunable by changing the solution pH to a range favorable for more selective copper uptake by this chelator (pH < 3)³³. Indeed, when adsorption tests were carried out using metal ion solutions prepared with pH = 2.45, copper uptake by ZIOS remained high (~97.5%) while the material adsorbed negligible iron (~14.3%) and nickel (~2.9%); in contrast, ZIF-8 adsorbed more iron (~30.0%) than copper (~14.6%) (Fig. 3e). Significantly, these results for ZIOS at lower pH may be promising for applications such as electrolytic nickel refining, where the separation of copper and nickel ions remains a substantial challenge⁹. For example, the Cu²⁺/Ni²⁺ separation factor achieved with ZIOS at pH 2.45 is 1281.1, while separation factors of 1.2 (pH = ~5) and 4.1 (pH = 4) were reported for the commercial resin Purolite S984²⁸ and an N-octyl-2-aminomethylpyridine-functionalized chelating resin³⁴ under equimolar concentrations of Ni²⁺ and Cu²⁺. We further evaluated the copper selectivity of ZIOS in an ion mixture having Zn²⁺ as one of the contaminants and found this behavior almost unchanged (Supplementary Table 6).

To further investigate the ion selectivity of ZIOS, we calculated the copper distribution coefficient (K_d, mL g⁻¹), or the ratio of copper in the adsorbed versus the solution phase (Fig. 3d), which represents the extent of adsorbent effectiveness in sequestering a target species and the selectivity of the adsorbent for that species³⁵. A K_d > 5000 is considered very good³⁵, ≥10⁴ mL g⁻¹ is considered excellent¹⁶ and >5 × 10⁴ is considered outstanding^4,35. Notably, ZIOS exhibits K_d values as high as 2.3 × 10⁵ across various Cu²⁺ concentration (Fig. 3d), and the distribution coefficient remains high in the presence of several competing ions and at different pH values (e.g., ~1.34 × 10⁵ at pH 2.45, see Fig. 3e). This performance either matches or exceeds that of state-of-the-art Cu²⁺ adsorbents^4,16,17,35.

Adsorption mechanism via spectroscopic characterization

In order to better understand the mechanisms leading to copper ion uptake in ZIOS and ZIF-8, we utilized X-ray photoelectron spectroscopy (XPS), synchrotron radiation near-edge X-ray absorption fine structure (NEXAFS) spectroscopy, and extended X-ray absorption fine structure (EXAFS) spectroscopy to characterize the materials before and after copper ion adsorption. XPS analysis of ZIOS-Cu and ZIF-8-Cu was carried out in the binding energy regions of the Cu 2p, Zn 2p, N 1s, C 1s, and O 1s orbitals. In-depth analysis is shown in the Supporting Information. We note that for both materials, we observe signatures of copper(I) in addition to copper(II) (see below). All peak positions were referenced to the C 1s peak of adventitious carbon at 284.5 eV. Initial deconvolution of the Cu 2p spectra for both materials (Fig. 4) revealed similarities with spectra of a previously reported mixed-valence Cu⁺Cu²⁺ complex (for ZIF-8-Cu) and a Cu⁺ complex and/or a heteronuclear Cu²⁺Zn²⁺ complex (for ZIOS-Cu), both supported by macrocyclic ligands³⁶. For ZIF-8-Cu, pairs of Cu 2p_1/2,3/2 doublets (at 952.4/932.6 eV and 954.8/934.8 eV) and the charge transfer shake-up satellites (at 962.8 and 941.8–944.8 eV) were observed, indicative of Cu⁺ and Cu²⁺, respectively. For ZIOS-Cu, the Cu 2p spectrum exhibits peaks at 952.78 and 932.98 eV characteristic of Cu⁺ (2p_1/2 and 2p_3/2, respectively), in addition to weaker features at 954.48 and 934.78 eV that can be ascribed to Cu²⁺ (2p_1/2 and 2p_3/2, respectively). The copper(I) peaks could indicate the presence of a five-coordinate Cu⁺ diamagnetic adduct or a four-coordinate Cu⁺ complex with nitrogen macrocyclic ligands, while the copper(II) peaks are indicative of a heteronuclear Cu²⁺Zn²⁺ complex³⁷.

In ZIOS, it is possible that copper chelates with salicylaldoxime by partially transmetalating with zinc nodes originally present in ZIOS and/or with 2-methylimidazole by coordinating with the free pyrrolic nitrogen. As seen in Fig. 4 and Supplementary Fig. 8, chelation of salicylaldoxime to Zn²⁺ in ZIOS results in a downshifting of the O 1s binding energies of free salicylaldoxime from 531.68 and 533.08 eV to 531.28 and 532.48 eV (HO–N and HO–C (phenol), respectively). Similar values of 531.28 and 532.38 eV (HO–N and HO–C (phenol), respectively) are observed following copper adsorption in ZIOS. In general, coordination to a metal ion results in a shifting of binding energies to lower values as a result of the electron donating/accepting ability of the metal ions³⁸. For ZIOS, the 0.1 eV-downshift of the binding energy of the phenolic oxygen O 1s peak, along with the enhanced intensity of this peak after copper adsorption, show that this site plays an active role in scavenging copper ions. Moreover, the N 1s binding energy of ZIOS exhibits a downshift from 398.88 to 398.78 eV at the pyridinic nitrogen following copper adsorption and, interestingly, a significant increase in intensity of the peak at ~400.18 eV for C=N(–O) and/or a pyrrolic nitrogen (Fig. 4 and Supplementary Fig. 8). Often, fluctuations of peak intensity at a given chemical state are related to changes in the number of atoms in that state. The remarkable increase of the intensity ratio of I_(C=N)/I_(C–N) from ~0.65 for ZIOS to ~1.54 for ZIOS-Cu indicates significant translational/rotational movement of C=N(–O) and/or pyrrolic nitrogen bonding from the bulk to the surface sites at which copper ions are coordinated. It remains unclear what imidazole nitrogen participates in copper coordination. Given the tautomeric property of the ionizable imidazole ring, both are possible if the Zn-imidazole bond was cleaved at the pyridinic nitrogen by copper ions. Note that the pH value of the stock copper chloride solution is ~5.5.

The roles of oxime and imidazole in coordinating with copper ions in ZIOS were further elucidated with NEXAFS spectroscopy. At the N K-edge, both ZIOS and ZIOS-Cu exhibit a sharp peak around 399 eV corresponding to a 1s→1π* transition. Both features are lower in energy than the corresponding 1s→1π* transition of the pyridinic nitrogen in the imidazole ring of the ZIF-8 structure (~400.5 eV). As the imino group typically shows a transition at 398–399 eV^39,40, the peak at 399 eV for ZIOS and ZIOS-Cu likely stems from the N 1s→1π* transition of the oximate nitrogen (C=N–O) or the imidazolate pyridinic nitrogen⁴¹. After copper(II) ion adsorption in ZIOS, this peak for the 1s→1π* transition intensifies significantly compared with the broad peaks at 403.2, 406.3, and 410.5 eV, which correspond to the ionization of aromatic N–H and transitions from N 1s→σ*_N–H and σ*_C–N, respectively⁴². This intensity increase suggests that the oxime hydroxyimino nitrogen and/or the imidazolate pyridinic nitrogen play a critical role in interacting with copper ions during adsorption. Given the close atomic radii of Zn (134 pm) and Cu (128 pm), along with the higher stability of complexes formed by copper(II) than zinc(II)⁴³, there could be a transmetalation event where Cu²⁺ partially ion-exchanges with some of the Zn²⁺ nodes in ZIOS. During this event, the zinc–oxime bonds may be partially cleaved, leaving oxime ligands poised to scavenge copper ions. On the other hand, the unbound Lewis basic pyrrolic nitrogen atoms of the imidazole side chains in ZIOS should have a high affinity for Lewis acidic Cu²⁺ ions. These sites could actively coordinate with copper to form a CuZn-based heteropolynuclear complex, reactivity that could potentially mimic that found in Cu–Zn superoxide dismutase, an antioxidant enzyme found in almost all cells²⁶ of the human body.

Zn K-edge NEXAFS profiles (Fig. 5b) and Fourier transformed spectra (Supplementary Fig. 9B) reveal that Cu²⁺ adsorption does not significantly change the local environment around the Zn²⁺ node in the ZIOS structure. This result is indicative of negligible adsorption of Cu²⁺ at sites already occupied by Zn²⁺ and preferential coordination at oxime and/or imidazole ligands. In contrast, a stark difference in the Zn K-edge data for ZIF-8 and ZIF-8-Cu (Supplementary Fig. 9B) indicates that the local environment around the Zn²⁺ nodes in ZIF-8 is altered upon Cu²⁺ adsorption, likely as a result of the observed structural transformation and consequent loss of characteristic ZIF-8 Zn²⁺ sites, observed via PXRD (Fig. 1d). We note that our results are in contrast with earlier reports on copper adsorption in ZIF-8¹⁸ and simple metal exchange reactions in zeolitic imidazolate frameworks⁴⁴. Indeed, rather than maintaining the ZIF structure as a result of simple copper physisorption or metal exchange, all our data suggest the ZIF-8 sample is altered completely upon exposure to copper ions.

At the Cu K-edge (Fig. 5c), both ZIF-8-Cu and ZIOS-Cu show similar pre- and main-edge peaks at ~8978 eV (feature α, corresponding to a dipole-forbidden 1s→3d transition) and ~8985 and 8987 eV, respectively (feature β, predominantly governed by the dipolar transitions from 1s→4p empty states), which confirm the presence of Cu²⁺ in the samples^45,46. However, a more detailed examination reveals some modest differences related to the peak intensities and sharpness between the two samples. The Cu L-edge NEXAFS profiles of the two samples (Supplementary Fig. 6C) also revealed differences in peaks positions. In NEXAFS, the energies and intensities of the peaks are typically sensitive to the local symmetry of metal ions, the metal coordination number, and oxidation state changes^45,46,47. For example, ZIF-8-Cu exhibits a more intense feature α than ZIOS-Cu (Fig. 5c, inset) while the main-edge feature β is slightly less intense relative to that of ZIOS-Cu. This is indicative of 3d–4p mixing^45,48,49, which is typically more pronounced in highly distorted, noncentrosymmetric structures. This property of ZIF-8-Cu is further supported by the broadness of peak β, which can be correlated with low coordination symmetry around copper⁵⁰. Furthermore, the intense rising shoulder at ~8987 eV observed for ZIF-8-Cu, which is impacted by the ligand field symmetry, has been associated with the ligand–copper–ligand angle that is less than 180°^47,51. Distinct patterns, except the first major peak at ~1.5 Å, resulting from Fourier transform of EXAFS data at the Cu K-edge further confirmed the existence of different copper coordination environments for ZIF-8-Cu and ZIOS-Cu (Fig. 5d). The first peak, observed in patterns of both samples, is associated with the copper first coordination shell comprising water oxygens, oxygen atoms of the nitrate anion and/or chloride ions^52,53. However, it is unlikely that there are chloride ions in this coordination shell as at the diluted testing concentration of CuCl₂ (895 ppm), CuCl₂ is prone to fully dissociate in the aqueous solution, rendering the inner hydration shell of copper with water molecules and nitrate anions. ZIF-8-Cu shows another predominant peak at ~2.85 Å which is within range of reported Cu–C bonds⁴⁵. Furthermore, there is no evidence of Cu–N bonding in ZIF-8-Cu, supporting the fact that the parent ZIF-8 structure featuring metal–imidazole bonding is completely transformed following copper uptake. On the other hand, peaks at ~2.01 and 2.41 Å for ZIOS-Cu (averaged from different scans) align well with reported Cu–N_equatorial and Cu–O_axial distances^{45,46,47,48,54}, respectively, in a pentacoordinated complex with a square pyramidal geometry, suggesting some degree of chemisorption. Other peaks observed in the range of 3.2–3.9 Å for ZIOS-Cu correspond to Cu–C bonds^45,47. From these analyses, we infer that copper predominantly chemisorbs in ZIOS by coordinating with the oxime and/or imidazole ligands, while copper ions in ZIF-8-Cu appear to be weakly bound through physisorptive interactions and also perhaps weakly coordinated within the structure. The extent to which oxime and imidazole moieties coordinate with copper in ZIOS-Cu remains undetermined.

From ICP-OES analysis of acid-digested solid samples, we were able to estimate that our samples of ZIF-8, ZIF-8-Cu, ZIOS, and ZIOS-Cu contain ~23.4, 19.9, 18.0, and 10.1 wt% Zn, respectively. Moreover, the ratios of Cu:Zn (by mass) in ZIF-8-Cu and ZIOS-Cu are 0.9:1 and 1.7:1, respectively. In these tests, ZIF-8-Cu and ZIOS-Cu were analyzed after soaking ZIF-8 and ZIOS for 24 h in a ~895 ppm CuCl₂ solution at pH 5.5. These results indicate that a large amount of Zn²⁺ still remains in ZIF-8-Cu, in spite of the structural collapse observed via SEM and PXRD, indicating that ion exchange is not the main mechanism governing copper ion uptake in this material. Rather, our data indicate that copper adsorption in ZIF-8 leads to collapse of the ZIF-8 parent structure and the formation of a new phase that has retained the majority of the zinc ions as well as incorporated physisorbed and also possibly some chemically bound copper ions. In contrast, for ZIOS, a fraction of the Zn²⁺ ions appear to be replaced with copper ions, leading to a structure featuring Cu²⁺/Cu⁺ coordination pockets near the coordinated Zn²⁺ ions in ZIOS.

Although our data indicate that the salicylaldoxime and imidazole ligands in ZIOS contribute to its copper ion coordination capability, the precise mechanism of rapid copper scavenging remains unclear. We hypothesize that, upon immersion into a copper(II)-containing aqueous environment, the framework expands to some extent by gradually inviting water molecules to “assimilate” themselves into the framework through the propagation of a hydrogen-bonding network. This process may then open up channels that can accept copper ions. In addition to spontaneously coordinating with imidazole pendants at Lewis basic pyrrolic nitrogen sites, copper ions may partially exchange with the Zn²⁺ nodes of the trinuclear units, chelating with salicylaldoxime ligands and leading to a partial transformation of the crystalline structure. Adsorption evidently occurs quickly enough such that ZIOS does not have enough time to further interact with water and collapse. This behavior of ZIOS aligns well with the guest-binding property of metallohosts, structures that contain one or more transition metal atoms^55,56. This property is regulated by metal coordination or redox reactions leading to site-selective transmetalation and ultimately to the construction of heteropolynuclear supramolecular structures. The transmetalation is evident by the amount of zinc ion transferred from ZIOS into the supernatant during copper adsorption quantified by ICP-OES (Supplementary Fig. 10).

Behaviors and stability of ZIOS in aqueous environment

To further support this hypothesis, we studied the behavior and stability of the highly crystalline ZIOS framework in an aqueous environment both computationally and experimentally. Molecular Dynamics (MD) simulations were performed using the LAMMPS package with standard three-dimensional periodic-boundary conditions. Further details are listed in the Experimental. Our results show that immersion of ZIOS unit cells in water leads to an expansion of the crystal lattice and the formation of nanochannels that can accommodate water molecules through the propagation of a hydrogen-bonding network. Water incorporation occurs quite rapidly—a snapshot at 15 ns (Fig. 6a) shows a noticeable amount of water molecules intercalating into ZIOS channels. Furthermore, as the number of water molecules introduced into the ZIOS unit cell increases, a gradual lattice expansion occurs in all crystallographic directions—L_x, L_y, and L_z (Fig. 6b, c). When water is removed, the lattice contracts back to its original size within less than 1 ns (Fig. 6b, c), indicating the expansion is reversible. Indeed, powder X-ray diffraction characterization of a sample of ZIOS soaked in water for 52 days, washed with methanol, and dried in vacuum revealed an almost identical pattern to that simulated from single-crystal X-ray diffraction data (Fig. 6d). A slight shift of the peaks is observed due to different diffraction scans temperatures (room temperature for the PXRD data versus 100 K for the SC-XRD data). This result serves as an experimental evidence of the breathing phenomena, suggested by the MD simulation, where ZIOS recovers its original crystalline structure after being excessively exposed to water for ~2 months. Other experimental approaches are however necessary to observe this truly interesting behavior in-operando. Still, to an extent, the breathing behavior of ZIOS is analogous to what was reported for some metal–organic frameworks that can adapt their pore opening to accommodate guest species, leading to dramatic change in cell volume without a loss of crystallinity or bond breaking^57,58. In our cases, the new material does not just behave instantaneously. Instead, it held its structure through vigorous treatment steps. The impressive structural integrity of ZIOS supports the presence of a strikingly robust hydrogen-bonded network which helps the material to reversibly breathe through expansion–contraction cycles before and after exposure to water.

Interestingly, our simulation further shows that, when a few water molecules are first introduced into ZIOS, they move within the framework by hopping between various hydrogen-bonding sites, shown by discrete step changes in Fig. 6e and Supplementary Fig. 12. We define a channel in ZIOS as the simplest two-dimensional stack between two consecutive trinuclear subunits with an average size of 2.7 Å (calculated from our single-crystal X-ray diffraction data). These hopping events were observed for all our simulations, involving 1, 2, 5, 10, or 15 water molecules, especially in the initial time frames. When enough water molecules are introduced into the structure and there is enough time for sufficient lattice expansion, water begins to diffuse within the crystal structure. The transition from a hopping to a diffusion mechanism is best illustrated in the case of simulations involving 10 or 15 water molecules. As it is unclear when the onset of this transition to diffusion occurs, we report simulation data collected before and after equilibrium (Fig. 6e and Supplementary Fig. 12). Note that the step size of each hopping moment varies due to the spatiotemporal randomness of such events, which are dependent on the total average number of water molecules and diverse gaps between hydrogen-bonding sites. In ZIOS, the most active hydrogen acceptor site was determined to be the pyridinic imidazole nitrogen, based on the radial distribution function, g(r) (see Fig. 6f and the Supplementary Information). Meanwhile, the most active hydrogen donor sites are the hydrogen atoms of the hydroxyimino groups. From our simulations, the hydrogen-bonding peak was observed within 2 Å for the most active sites (Supplementary Fig. 13). Similar to previously reported^59,60, we found evidence of the contribution of the benzene rings to ZIOS molecular associations through the hydrogen-bonding network. Although a water-benzene hydrogen bond is weaker than a water–water hydrogen bond, water molecules associated with benzene were believed to have more rotational and translational freedom than their water–water counterparts⁶⁰. Further exploration is essential to confirm this hopping behavior experimentally.

Source: Resources - nature.com