Zeolitic Imidazolate Frameworks Overview

• ZIFs are metal–organic frameworks defined by tetrahedrally coordinated metal centers linked by imidazolate ligands, forming zeolite-like networks with high microporosity.
• Their structural versatility allows precise tuning of adsorption, hydrolytic stability, mechanical, optical, and electronic properties through adjustments in metals, linkers, and topology.
• Advanced computational and high-throughput experimental studies reveal extensive polymorphism and phase transitions, informing novel applications from energy absorption and sensing to biomedical uses.

Zeolitic imidazolate frameworks (ZIFs) are a subclass of metal–organic frameworks in which tetrahedrally coordinated metal centers, most commonly Zn$^{2+}$ or Co$^{2+}$, are bridged by imidazolate-type ligands. Because the metal–imidazolate–metal angle is approximately $145^\circ$, analogous to Si–O–Si in zeolites, ZIFs adopt zeolite-like nets such as sod, rho, cag, zni, dia, and lta, while retaining high microporosity together with notable chemical and thermal stability. Their unusually broad design space allows changes in metal, linker substituent, topology, pore geometry, and density to propagate across adsorption, hydrolytic stability, mechanics, melting, optics, electronics, and bioactivity (Fu et al., 2024, Atambo et al., 13 Mar 2025, Xu et al., 8 Jan 2026).

1. Structural chemistry, topology, and polymorphism

The canonical ZIF building unit is a tetrahedral metal center coordinated by four imidazolate nitrogens, yielding local environments such as ZnN$_4$ in ZIF-8 and related zinc frameworks. ZIF-8, often written as Zn(mIm)$_2$, is the best-studied archetype; it has sodalite topology, ZnN$_4$ tetrahedra, and well-defined cages and windows. Other prominent members include ZIF-71, built from Zn$^{2+}$ and 4,5-dichloroimidazolate in a RHO topology; ZIF-4, Zn(Im)$_2$, in the cag topology; and mixed-linker ZIF-62, Zn(Im)$_{2-x}$(BIm)$_x$, which is central to the glass literature. Across the family, linker identity and substituents modulate pore aperture, cavity size, framework polarity, and band-edge composition without necessarily changing the underlying net (Fu et al., 2024, Möslein et al., 2021, Li et al., 13 Aug 2025).

This chemistry produces extensive polymorphism. For zinc imidazolate, Zn(Im)$^{2+}$0, at least 24 crystallographically distinct forms are reported in 19 topologies, and hierarchical crystal structure prediction has expanded the accessible landscape far beyond known experimental structures. In a high-throughput exploration of over 3,000,000 randomly generated ZnIm$^{2+}$1 packings, 9,626 unique minima and 1,493 distinct network topologies were identified, including 864 topologies not present in current topological databases; 13 topological matches corresponded to experimentally observed ZnIm$^{2+}$2 polymorphs. That result does not imply that all such structures are experimentally accessible, but it does establish that the Zn–Im connectivity supports a far denser energy–topology landscape than a small set of textbook exemplars would suggest (Xu et al., 8 Jan 2026).

Local chemistry can be tuned even when topology is fixed. Density-functional calculations on ZIF-8 and its 2-position analogues ZIF-8-H, ZIF-8-Cl, and ZIF-8-Br show that the Zn–N coordinative bond remains highly polarized in every case, while halogen substitution shifts charge distribution and frontier states. In that series, Zn carries a Bader charge of +1.17 e in ZIF-8 and about +1.20 e in ZIF-8-Cl and ZIF-8-Br, while Zn–N bond lengths stay near 2.00 Å in the crystal. The major electronic consequence is not a wholesale redefinition of Zn–N bonding, but the emergence of halogen-derived states near the band edges, especially in ZIF-8-Br, where the band gap is reduced by more than 0.5 eV relative to ZIF-8 (Edzards et al., 2023).

2. Water, hydrophobicity, adsorption, and hydrolysis

Water behavior in ZIFs is governed jointly by topology, pore geometry, and linker functionalization rather than by linker polarity alone. Grand Canonical Monte Carlo simulations at 300 K showed that hydrophobic SOD frameworks such as ZIF-8, SALEM-2, and ZIF-Cl display negligible gas-phase uptake and instead exhibit type V water adsorption with liquid intrusion and hysteresis, whereas ZIF-90 and ZIF-65 become hydrophilic in the gas phase because polar groups create strong binding sites. The zero-loading adsorption enthalpy spans a wide range: 13.6 kJ mol$^{2+}$3 for SALEM-2, 15.5 kJ mol$^{2+}$4 for ZIF-8, 17.8 kJ mol$^{2+}$5 for ZIF-Cl, 39.5 kJ mol$^{2+}$6 for ZIF-7, 52.5 kJ mol$^{2+}$7 for ZIF-65 (SOD), 69.5 kJ mol$^{2+}$8 for ZIF-65 (RHO), and 79.7 kJ mol$^{2+}$9 for ZIF-90. A common simplification is therefore misleading: ZIF hydrophobicity is not a linker-only property. ZIF-7, despite nonpolar benzimidazolate chemistry, develops geometric hydrophilic pockets, while ZIF-65 (RHO) combines hydrophilic patches with hydrophobic cavity filling near saturation (Ortiz et al., 2019).

Under forced intrusion, cage-type hydrophobic ZIFs display a distinct nanofluidic regime. In ZIF-8, loading proceeds through elastic compression, an intrusion plateau at $145^\circ$0, and then further elastic compression; unloading shows a lower extrusion plateau at $145^\circ$1. The key observation is strong rate dependence: $145^\circ$2 rises from about 25 MPa in quasi-static tests to about 70 MPa at approximately $145^\circ$3 s$145^\circ$4, while the absorbed fraction of stored mechanical energy increases from about 17% to about 85%, and the specific energy absorption rises from about 3 to about 47 J g$145^\circ$5. Molecular simulations attribute this to nucleation-controlled transport in hydrophobic cages. Water must form a hydrogen-bonded critical cluster of about four molecules in a target cage before intercage hopping through the 6-membered-ring aperture becomes favorable, with an intrinsic nucleation timescale of about 1 ns. Contrary to a frequent analogy with gas adsorption, water does not open the ZIF-8 6MR gate; its presence shifts swing-angle distributions toward a more closed state (Sun et al., 2021).

Hydrolytic stability follows a related logic but on a reactive free-energy surface. Reactive MD plus metadynamics in the low-water limit found that ZIF-90 has a hydrolysis barrier of $145^\circ$6 kcal mol$145^\circ$7 and ZIF-4 has a barrier of $145^\circ$8 kcal mol$145^\circ$9, with strongly uphill reaction free energies, especially for ZIF-90. In both cases, the relevant event is Zn–N bond cleavage with proton transfer to the imidazolate nitrogen. These barriers are far larger than those reported for Zn-based IRMOFs, supporting the broader conclusion that ZIFs are generally more water-stable than iso-reticular carboxylate MOFs. The same study notes, using comparison points from Yang et al. (2023), that nonpolar methyl substitution can raise hydrolysis barriers to about $_4$0 kcal mol$_4$1, which suggests that hydrophobic and steric shielding near the Zn–N coordination sphere is a practical route to humidity tolerance (Yacham et al., 22 Jul 2025).

3. Mechanical response, elastic anomalies, and thermal transport

Finite-temperature stability in ZIFs cannot be inferred from static energy minimization alone. Molecular dynamics on 18 Zn(im)$_4$2 polymorphs showed that some frameworks remain stable up to 400 K, whereas others are stable only at low temperature or collapse even at 77 K after desolvation. Pressure stability is likewise topology-dependent: among room-temperature-stable polymorphs, coi remains stable to 1.0 GPa, SOD and nog to 0.4 GPa, CAN to 0.3 GPa, MER to 0.2 GPa, FAU, cag, and LTL to 0.05 GPa, and AFI and DFT fail below 0.05 GPa. The general instability mechanism is pressure-induced shear-mode softening, i.e. a violation of the Born elastic stability criterion through a softening eigenvalue of the elastic tensor. The same study predicted large topology-dependent thermal expansion anomalies, including volumetric coefficients of $_4$3 MK$_4$4 for LTL and $_4$5 MK$_4$6 for MER near room temperature (Bourg et al., 2014).

Single-crystal elasticity is correspondingly anisotropic. Ab initio calculations for the cag-topology series ZIF-4, ZIF-62, and TIF-4 yielded low isotropic moduli and pronounced tensorial anomalies. Their Voigt–Reuss–Hill bulk moduli are 1.76, 1.92, and 4.20 GPa, respectively; shear moduli are 1.11, 1.33, and 1.83 GPa; and Young’s moduli are 2.75, 3.24, and 4.79 GPa. All three exhibit negative Poisson’s ratio regions and negative linear compressibility. ZIF-62 is especially anisotropic, with a universal elastic anisotropy index $_4$7, $_4$8, and $_4$9 TPa$_2$0 (Xiong et al., 2017).

These framework-level properties persist, with microstructural modifications, in monolithic ZIFs. Sol–gel monoliths of ZIF-8 and ZIF-71 were shown by tip force microscopy to consist of grains below 100 nm, with ZIF-71 displaying more prominent intergranular porosity. Berkovich nanoindentation gave $_2$1 GPa and $_2$2 MPa for monolithic ZIF-8, and an overall $_2$3 GPa with hardness around 227 MPa for monolithic ZIF-71. Finite-element calibration with $_2$4 and $_2$5 yielded approximate yield stresses of 200 MPa for ZIF-8 and 130 MPa for ZIF-71. The identified deformation sequence was grain-boundary sliding at low stress, bond breakage and partial framework failure at higher stress, and local densification under highly confined contact (Tricarico et al., 2021).

Low-frequency vibrational dynamics supply the microscopic counterpart to these mechanical data. For ZIF-71, high-resolution inelastic neutron scattering, synchrotron infrared spectroscopy, and periodic DFT assigned soft collective modes including 8MR gate opening at 9.46 and 23 cm$_2$6, a 6MR gate-opening mode at 37 cm$_2$7, and a shear mode at 10.46 cm$_2$8. The latter was interpreted as a marker of mechanical instability toward shear-driven phase change. Atomic-force nanomechanics on ZIF-71 nanocrystals gave typical Young’s moduli around 2 GPa and hardnesses around 100–300 MPa, consistent with a softer framework than ZIF-8 and with low-energy ring-shearing pathways (Möslein et al., 2021).

Thermal transport in ZIFs is similarly atypical. Molecular dynamics on Linde Type A ZIFs showed that heat is carried predominantly by diffuson-like modes rather than long-lived propagons: phonon mean free paths approach their wavelengths at elevated wavevectors, satisfying the Ioffe–Regel condition $_2$9. For ZIF-lta, the isotropic elastic analysis yielded $_4$0 m s$_4$1, $_4$2 m s$_4$3, an adiabatic compressibility of about $_4$4 Pa$_4$5, and a Debye temperature of about 162 K. Hydrogen adsorption increases thermal conductivity mainly through gas–framework virial coupling: the hydrogen-related contribution inside ZIF-lta at 20 molecules per unit cell is about 0.04 W m$_4$6 K$_4$7, roughly four times the virial contribution of bulk hydrogen at similar density (Oh et al., 3 Aug 2025).

4. Melting, glasses, and phase transformations

ZIFs do not merely collapse under heat; some enter genuine liquid and glassy states while retaining recognizable coordination chemistry. ZIF-4 is the clearest case. Building on prior identification of a melting point near 865 K by Bennett and co-workers, variable-temperature scattering and first-principles molecular dynamics showed that molten ZIF-4 remains a strongly associated liquid in which Zn–N coordination persists, medium-range order is lost, and porosity survives in a fluctuating form. At 1,500 K the self-diffusion coefficients are $_4$8 m$_4$9 s$^{2+}$0 and $^{2+}$1 m$^{2+}$2 s$^{2+}$3, evidencing coupled motion of linkers and nodes. The enthalpy of fusion was estimated as 173 J g$^{2+}$4, and geometric analysis showed average pore volumes of 52 cm$^{2+}$5 kg$^{2+}$6 at 300 K, 49 cm$^{2+}$7 kg$^{2+}$8 at 2,000 K, and 41 cm$^{2+}$9 kg$_2$0 at 2,250 K. Wharmby et al. provided the evacuated starting material, and the melt-quenched product preserved an extended Zn–Im–Zn network rather than recrystallizing under the conditions used (Gaillac et al., 2017).

Deep-learning molecular dynamics has extended that picture to larger length scales. A 7,344-atom, 3%%%%71$_4$72%%%%3 ZIF-4 supercell simulated with a PBE-D3-trained deep potential revealed spatial heterogeneity during heating that is inaccessible to small-cell AIMD. At slow heating, a low-density regime appears before densification and recrystallization, and coarse-grained density maps show contiguous low-density regions nucleating between about 1400 and 1550 K. The transition temperature decreases with slower ramps, from about 1432 K at 1 K ps$_2$3 to 1418 K at 0.5 K ps$_2$4 and 1352 K at 0.2 K ps$_2$5, which suggests that rate control is central to reconciling simulation and experiment. The mechanistic sequence is ligand exchange, transient undercoordination of Zn, medium-range rearrangement of the Zn network, and emergence of spatially heterogeneous low-density regions (Shi et al., 2023).

Glass-forming ability in ZIFs has been parameterized in terms of electronic cohesion and pore geometry. A high-throughput analysis of more than 20,000 MOFs identified the total bond-order density,

$_2$6

together with the largest cavity diameter (LCD), as effective “structural genes” for melt stability in ZIFs. For the 16 ZIFs explicitly heated from 300 to 1000 K, the reported screening window for thermally stable candidates was TBOD $_2$7, $_2$8 eV Å$_2$9, LCD $_{2-x}$0 Å, and PLD $_{2-x}$1 Å. Representative stable cases such as ZnH$_{2-x}$2(C$_{2-x}$3N$_{2-x}$4)$_{2-x}$5-cag and -zni have TBOD values of 0.220 and 0.219 with LCD values of 4.360 and 4.239 Å, whereas thermally sensitive can and afi structures combine low TBOD with very large cavities (Shi et al., 2023).

Glassy ZIFs are not only structural end states; they can also be functional optical materials. Melt-quenched Zn-based ZIF-62 glass exhibits broadband white-light emission that strengthens under annealing above a critical temperature near $_{2-x}$6. The optimally annealed glass reached an absolute internal photoluminescence quantum yield of 12.2% at $_{2-x}$7, and a white LED fabricated from the annealed glass achieved a luminous efficacy of 4.2 lm/W while retaining 74.1% of its initial luminous efficacy after 180 min of continuous operation. The reported mechanism combines strengthened ligand-to-metal charge transfer with enlarged $_{2-x}$8-conjugation produced by structural relaxation in the glass (Li et al., 13 Aug 2025).

5. Computational, data-driven, and high-throughput frameworks

The complexity of ZIF chemistry has pushed modeling beyond conventional one-structure-at-a-time workflows. For adsorption, an automated high-throughput pipeline has been developed around porE pore analysis, reverse-onion sampling of guest positions, Orix symmetry reduction of molecular orientations, KDTree collision detection, and AiiDA-managed Quantum ESPRESSO calculations. Applied to ZIF-CO$_{2-x}$9-1, this approach defines the adsorption energy as

$_x$0

and showed that the largest pore radius is about 2.63 Å, close to a literature value of 2.85 Å. Symmetry reduction substantially pruned the search space: for one sampling resolution, 4,586 realizations were reduced to 1,555, and for a denser sampling 20,476 were reduced to 7,164 (Atambo et al., 13 Mar 2025).

Machine learning has also been used to test whether ZIFs can be treated as coarse-grained AB$_x$1 frameworks. For Zn(Im)$_x$2 local environments, a fully atomistic SOAP/GPR model predicted Zn-centered local energies with RMSE about 0.042 eV and $_x$3, whereas an AB$_x$4 coarse-grained representation that retained both metal and linker beads gave RMSE about 0.065 eV and $_x$5. An A-only model that discarded linkers degraded to RMSE about 0.150 eV and $_x$6. This supports the long-used AB$_x$7 analogy at the level of topology and local energetics, but only if linker geometry is preserved in the descriptor (Beaulieu et al., 2023).

For crystal discovery, hierarchical crystal structure prediction now combines random structure generation, machine-learned interatomic potentials, and DFT refinement. In zinc imidazolate, that strategy recovered known nets such as zni, dia, cag, dft, gis, sod, crb, and neb, while simultaneously exposing many low-energy hypothetical structures. The reported PaiNN-based models reached energy mean absolute errors of 0.00713 eV on the training set and 0.01214 eV on validation, enabling the large-scale screening noted above. A void-adjusted synthesizability descriptor,

$_x$8

with $_x$9 kJ mol$^{2+}$00, reduced 9,626 candidate structures to 982 likely synthesizable targets (Xu et al., 8 Jan 2026).

Phase identification during simulation has likewise become automated. Neural-network classifiers trained on Zn(Im)$^{2+}$01 polymorphs, glasses, and liquids reached 98.6% accuracy with SOAP descriptors and 92.6% with a 12-feature Zn-only Behler–Parrinello symmetry-function representation. Training on configurations generated by both nb-ZIF-FF and MACE largely removed force-field bias. Applied to the ZIF-4-cp $^{2+}$02 ZIF-4-cp-II transition, the classifier resolved sigmoidal product growth, frustrated pre-transition nucleation, and anisotropic cluster growth faster in the $^{2+}$03 and $^{2+}$04 directions than along $^{2+}$05 (Méndez et al., 10 Apr 2026).

6. Functional applications and cross-domain design principles

Because ZIF topology, pore geometry, and local chemistry are unusually tunable, applications span mechanically driven energy absorption, separation and catalysis, sensing, electronics, and biomedicine. In high-rate nanofluidic damping, the absorbed work is quantified by

$^{2+}$06

and the specific energy absorption by $^{2+}$07. Hydrophobic cage-type ZIFs satisfy a distinctive design rule set: hydrophobicity to preserve an interfacial barrier, cage-type topology with LCD $^{2+}$08 PLD to enforce nucleation-controlled transport, sufficiently large apertures for reusability, and large cages for higher energy density. In that framework, ZIF-8 and ZIF-67 are reusable high-rate absorbers, whereas ZIF-7, ZIF-9, ZIF-11, and ZIF-12 can trap water because their PLD is too narrow for facile extrusion (Sun et al., 2021).

In sensing and optoelectronics, ZIFs often act as active nanoconfinement hosts rather than inert containers. Rhodamine B encapsulated in ZIF-71 produces a fluorochromic material with mechanochromic, thermochromic, and solvatochromic responses. For RhBII@ZIF-71, the emission peak shifts linearly with pressure according to $^{2+}$09 over 0–346.6 MPa, and thermochromism follows $^{2+}$10 from room temperature to 200$^{2+}$11C. The mechanistic basis is nanoconfinement-controlled interconversion among RhB monomers and H- and J-type aggregates inside the framework pores (Zhang et al., 2021).

ZIFs also support electronic device functions. A solution-processed Al/ZIF-8/ITO memristor using an approximately 150 nm ZIF-8 layer showed forming-free bipolar resistive switching with an on/off ratio of about 100, retention up to 10,000 s, and operation from $^{2+}$12C to $^{2+}$13C. Low-frequency noise and impedance spectroscopy were interpreted as evidence for Zn-based filamentary switching through the porous framework, and reset-voltage control enabled multiple intermediate resistance states, with up to nine distinguishable states reported over the observation window (Kaushik et al., 3 Jan 2025).

Biomedical use is more recent but already technically specific. In fungal keratitis models, ZIF-8 synthesized from Zn$^{2+}$14 and 2-methylimidazole yielded polyhedral particles averaging about 279 nm with a zeta potential of +32 mV. In vitro, it showed detectable growth inhibition of Aspergillus fumigatus at 16 $^{2+}$15g mL$^{2+}$16 and more than 90% inhibition at 64 $^{2+}$17g mL$^{2+}$18, while also reducing IL-6, IL-16, and TNF-$^{2+}$19 expression after fungal stimulation. In vivo, topical ZIF-8 at 640 $^{2+}$20g mL$^{2+}$21 improved clinical scores relative to PBS and natamycin, reduced corneal fungal load, and diminished neutrophil infiltration, while Draize testing showed no epithelial damage over the observed interval (Fu et al., 2024).

Taken together, these results suggest a unifying design principle for ZIFs: topology and local coordination do not merely set porosity, but couple directly to transport barriers, hydrolytic resistance, elastic soft modes, melting pathways, spectral response, ionic motion, and biological interface chemistry. The field therefore treats ZIFs not as a single material class with one dominant function, but as a chemically coherent platform whose behavior can be shifted, often sharply, by controlled changes in linker substituent, cavity/window geometry, and phase state.