Metal-Organic Frameworks

Metal-organic frameworks (MOFs) are crystalline porous materials in which metal nodes — single ions or small clusters — are linked by polytopic organic struts into extended networks. The combination of modularity (the same node can be paired with many different linkers, and vice versa) and crystallinity (the resulting framework has a well-defined pore geometry that survives guest removal) gives MOFs the highest surface areas of any known materials and a designable pore chemistry that zeolites and activated carbons cannot match. The field’s methodological work organises around four axes: function-targeted pore design (engineering specific chemical environments inside the pore to recognise a target gas or molecule), non-crystalline derivatives (extending the toolkit to glasses, films, and composites whose properties go beyond what a bulk crystal can offer), computational acceleration (predicting adsorption from first principles fast enough to screen the combinatorial design space), and data infrastructure (capturing the synthesis knowledge buried in the literature so that machine-learning models have something to train on).

Function-targeted pore design

The defining advantage of MOFs is that the chemistry of the pore is a design variable rather than a fixed property of the host. Chen et al. (2024) exploit this in their study of water-enhanced direct air capture of CO2: by engineering a MOF whose primary-amine-decorated pore preferentially binds CO2 in the presence of co-adsorbed water, they convert what is normally a problem (humidity competing with the target gas) into a feature, with capacity that grows under realistic atmospheric conditions instead of collapsing. The work reframes direct air capture from a humidity-management problem into a co-adsorption design problem. Yang et al. (2023) attack a different separation challenge with the same toolkit, immobilising a polar group inside an ultramicroporous MOF to achieve benchmark inverse selectivity between hydrocarbon isomers — the framework preferentially binds the less polarisable component, inverting the order predicted by surface-area arguments. Together the two papers illustrate the field’s current synthetic granularity: not just “build a porous solid” but “place a specific functional group at a specific position relative to the metal node so that one guest binds and a similar guest does not”.

Beyond the bulk crystal: films and glasses

A MOF as conventionally synthesised is a polycrystalline powder, which limits integration into devices. Luo et al. (2024) report wrinkled MOF thin films with tunable Turing patterns: by controlling the kinetics of film growth on a flexible substrate, they programme the surface morphology into reaction-diffusion patterns whose wavelength can be tuned, and they demonstrate pliable device integration that powdered samples cannot support. The result connects MOF synthesis to the broader pattern-formation literature and opens a route to MOF-based wearable sensors. In a complementary direction, Song et al. (2023) melt a glass-forming cobalt-based zeolitic imidazolate framework and recover a MOF glass — an amorphous solid that retains the local coordination chemistry of the parent crystal but loses long-range order — and show that the glass is a viable photocatalyst. The work extends the MOF concept beyond its founding assumption (crystallinity) and asks whether function survives, and sometimes improves, when the framework loses its periodicity. Both papers fit a broader trend: the most interesting MOFs of the next decade may not be crystals at all.

Computational acceleration and data infrastructure

The combinatorial design space of MOFs — millions of plausible node-linker combinations — is too large for synthesis-driven exploration alone, and screening needs predictions of adsorption energies that classical force fields cannot deliver reliably. Goeminne et al. (2023) close that gap with DFT-quality adsorption simulations enabled by machine-learning interatomic potentials: a neural-network potential trained on density-functional data reproduces DFT energetics at the cost of a classical simulation, making it tractable to screen thousands of framework-guest pairs at near-quantum accuracy. The paper is a worked example of how ML potentials are reshaping the screening pipeline for porous materials. Glasby et al. (2023) attack the other half of the problem with DigiMOF, a database of MOF synthesis recipes extracted from the literature by text mining; structural databases for MOFs exist, but the synthesis conditions that produce them have historically been locked inside experimental sections of papers. DigiMOF turns that knowledge into a structured resource that machine-learning models can use to predict which conditions produce which framework. The combined methodological direction is clear: the next generation of MOF design will couple ML interatomic potentials (for property prediction) with text-mined synthesis databases (for makeability prediction), so that a screening run produces not just a target framework but a plausible recipe to synthesise it. Open methodological frontiers include extending these tools to defect engineering, multivariate MOFs whose linkers vary along a single crystal, and operando structural studies that watch a framework respond to a guest under real conditions.