Now, a team of researchers in China reports a strategy that directly confronts this processing challenge. Writing in Nature Communications, Meidi Wang, Guan-Zheng Zhou, and colleagues describe a dual-drive transformation that converts COF powders into coherent membranes through the interplay of two forces: an internal driver based on framework reconstruction and an external driver based on acid attack. The work, carried out at China Three Gorges University with collaborators at the Harbin Institute of Technology and Huazhong University of Science and Technology, demonstrates that dynamic covalent chemistry, the same reversible bond-forming behavior that allows COFs to self-correct during synthesis, can be harnessed at the macroscopic scale to reshape an entire material.

The central insight is that COF frameworks are not as chemically inert as they might appear. Under the right conditions, the linkages that hold the network together can break and reform, allowing the structure to reorganize. The researchers exploited this plasticity deliberately. By subjecting COF powders to acid attack, they destabilized the existing bonds, while the intrinsic tendency of the framework to reconstruct drove the material to reassemble into a continuous membrane rather than remaining a collection of disconnected crystallites. The result is a transformation that operates at the macroscopic scale, converting bulk powder into a free-standing or supported film without the need for elaborate layer-by-layer assembly or high-temperature processing.

What makes the approach especially striking is its scalability. The team produced both two-dimensional and three-dimensional COF membranes, with the largest reaching an area of 225 square centimeters. For a field where many reported membranes measure only a few square centimeters, that figure represents a meaningful leap toward practical dimensions. The dual-drive strategy also achieved high conversion ratios and fabrication efficiency, outperforming conventional membrane-making methods on several fronts. Because the transformation relies on chemical driving forces rather than energy-intensive deposition techniques, the authors describe it as a green and versatile route to COF membranes, one that could in principle accommodate the wide structural diversity that reticular chemistry already provides.

The performance of the resulting membranes is where the work becomes truly compelling. The team’s optimum membrane, built from the TpPa framework, a combination of a triphenylene-based aldehyde and a phenylenediamine linker that forms imine-linked 2D layers, delivered impressive molecular sieving results. In filtration tests, the membrane rejected 99 percent of dye molecules and 98 percent of pharmaceutical molecules, while maintaining a pure water permeance of 179.0 liters per square meter per hour per bar. That combination of high selectivity and high flux is the holy grail of membrane science, since the two properties usually trade off against each other: membranes that filter more precisely tend to let water through more slowly, and vice versa.

The implications for water treatment are considerable. Dyes from textile manufacturing and active pharmaceutical compounds from municipal and industrial wastewater are among the most troublesome classes of aquatic pollutants, being small enough to slip through many conventional membranes yet harmful at low concentrations. A membrane that can reject these molecules at the 98 to 99 percent level while sustaining fast water transport could reduce the energy and cost of advanced treatment processes. The authors position the technology within chemical engineering separations broadly, suggesting applications that extend beyond pollution remediation to areas such as solvent purification and resource recovery.

To appreciate why the powder-to-membrane problem has been so persistent, it helps to consider the alternatives. Researchers have tried to grow COF membranes directly on supports through interfacial polymerization, to exfoliate 2D COFs into nanosheets and restack them into laminates, and to blend COF particles into polymer matrices as mixed-matrix films. Each route has drawbacks. Direct growth can be slow and difficult to control over large areas, often leaving pinholes that destroy selectivity. Exfoliation and restacking depend on delicate assembly steps that are hard to scale. Mixed-matrix membranes suffer from poor interface compatibility between the inorganic-like framework and the polymer. The dual-drive transformation sidesteps many of these issues by starting from the fully formed, crystalline powder and letting dynamic covalent chemistry do the assembly work.

The strategy also speaks to a broader trend in materials chemistry: the recognition that synthesis and processing should not be treated as separate stages. Reticular chemistry has excelled at designing structures atom by atom, but the history of materials science shows that a material only matters industrially when it can be shaped into the form a device requires. Graphene, metal-organic frameworks, and zeolites have all faced similar processing bottlenecks, and breakthroughs in shaping them, such as scalable graphene films or zeolite membranes, unlocked entire application domains. The COF community has been waiting for its own processing breakthrough, and a powder-to-membrane transformation that is green, efficient, and general enough to span 2D and 3D frameworks may prove to be exactly that.

There are, of course, questions that will shape the road ahead. The reported work is at an early stage, and the authors note that the published version is subject to further editorial refinement. Long-term stability under continuous operation, mechanical robustness, performance in complex real-world feed streams, and the economics of scale-up all remain to be demonstrated beyond the laboratory. The acid attack step, while central to the transformation, will need careful engineering in industrial settings. Still, the fundamentals are encouraging: the membranes are defect-free, the areas achieved are the largest reported for this class of material by a wide margin, and the underlying chemistry is inherently modular, meaning new COF chemistries could be plugged into the same transformation pipeline as they are invented.

The study, published open access on 30 September 2026 with support from the National Natural Science Foundation of China and the Hubei Three Gorges Laboratory, arrives at a moment when demand for advanced separation technology is rising across water purification, chemical manufacturing, and energy storage. If the dual-drive transformation proves as general and scalable as the authors suggest, the vast library of COF powders accumulated over the past two decades could suddenly become a library of usable membranes. That would be a quiet revolution in separation science, achieved not by inventing a new material, but by finding the chemistry that lets an existing one take the shape the world needs.

Subject of Research: Transformation of covalent organic framework powders into 2D/3D membranes for molecular separation

Article Title: Powder-to-membrane transformation affording 2D/3D COF membranes for molecular separations

Article References: Wang, M., Zhou, G.-Z., Zhou, Z., Shao, L., Shu, C., Ma, T.-Y., Wu, X.-Q., & Li, D.-S. (2026). Powder-to-membrane transformation affording 2D/3D COF membranes for molecular separations. Nature Communications. https://doi.org/10.1038/s41467-026-78068-w

Image Credits: AI Generated

DOI: 10.1038/s41467-026-78068-w

Keywords: covalent organic frameworks, COF membranes, molecular separations, dynamic covalent chemistry, water purification, membrane technology, reticular chemistry, dye rejection, pharmaceutical pollutants, framework reconstruction, porous materials, chemical engineering

Metal–organic frameworks, or MOFs, are crystalline materials in which metal clusters are connected by organic linkers into vast, sponge-like lattices. Because their pores can be designed with atomic precision, MOFs have become leading candidates for storing gases at moderate pressures, offering a safer alternative to compressing hydrogen to extreme pressures or chilling it into a liquid. The central challenge is that hydrogen molecules interact only weakly with any surface, so the best adsorbents must combine enormous internal surface area with carefully engineered pore sizes. The most porous MOFs reported to date, such as DUT-60 with 8,890 square metres per gram and NU-1501-Al with 7,241 square metres per gram, have pushed these limits, but each new record demands ever more sophisticated structural design.

The Cambridge team, led by David Fairen-Jimenez together with colleagues at Peking University, Stockholm University, the ISIS Neutron and Muon Source, Diamond Light Source and Synchrotron SOLEIL, approached the problem through reticular chemistry — the discipline of assembling multi-topic organic linkers and inorganic nodes into predetermined network topologies. Their target was the augmented red net, a three-dimensional arrangement theoretically derived from the parent reo net that contains three distinct types of polyhedral cavity: distorted cubes, cuboctahedra and rhombicuboctahedra. Although computational studies had predicted that this topology should be accessible, no one had managed to synthesise it as a pure crystalline phase, largely because the geometry demands that four-connected linkers adopt two different conformations simultaneously — one essentially planar and one noticeably bent out of plane.

Using two different rigid linkers to satisfy these two geometries tends to produce messy mixtures of competing crystalline phases. The researchers’ solution was elegant: a single linker that could switch between both shapes. They selected a diphenylethyne core, in which two phenyl rings are joined by a rigid carbon–carbon triple bond. While the ethynyl spacer itself is linear and stiff, rotation and bending of the adjacent phenyl rings give the molecule enough conformational freedom to accommodate different coordination geometries and framework curvatures. The team designed five candidate linkers of increasing length and evaluated them with density functional theory calculations, plotting total energy as a function of bending angle. Three of the candidates preferred to stay linear, with energy rising monotonically as they bent. But two of them, H4L3 and H4L5, actually became more stable when bent, reaching energy minima at 15 and 21 degrees respectively before the cost of further distortion kicked in.

H4L3, chosen for its simpler synthesis and its potential to yield a denser framework, was combined with six-connected, oxo-centred trinuclear metal clusters. Heating the linker with iron chloride and 4-nitrobenzoic acid in dimethylformamide produced red cubic single crystals of CU-6-Fe. A postsynthetic metathesis reaction, in which iron ions were exchanged for chromium, delivered the green crystals of CU-6-Cr. Synchrotron single-crystal X-ray diffraction revealed that CU-6-Cr crystallises in the cubic space group P-43m with a colossal unit cell edge of just over 50 nanometres — a hundred times larger than a typical molecular crystal. Within the structure, the linkers indeed adopted two crystallographically distinct conformations, bent by roughly 3 and 13 degrees out of plane, exactly as the design logic required.

The resulting architecture is a marvel of hierarchical porosity. Six pairs of cubic building blocks assemble into rhombicuboctahedral cavities with internal diameters of 4.1 nanometres, while stacking of eight neighbouring rhombicuboctahedra generates cuboctahedral cavities of 2.6 nanometres, and the distorted cubic pores themselves measure 1.6 by 2.4 by 2.4 nanometres. Triangular and square windows, framed by the planar and bent linkers respectively, interconnect all three cavity types into a continuous three-dimensional network. Calculations using a probe radius of 1.2 angstroms showed that 89.7 per cent of the crystal volume is accessible to guest molecules. Low-dose scanning transmission electron microscopy, in both annular dark-field and integrated differential phase contrast modes, directly visualised the ordered porous structure despite the crystals’ extreme sensitivity to electron beams, and powder X-ray diffraction confirmed the bulk samples were pure phases.

The measured porosity is staggering. CU-6-Fe achieved a Brunauer–Emmett–Teller surface area of 7,145 square metres per gram, satisfying all five extended-Rouquerol criteria for reliable surface-area determination, while CU-6-Cr reached 6,488 square metres per gram. Total pore volumes were 4.14 and 3.67 cubic centimetres per gram, with bulk densities of just 0.214 and 0.240 grams per cubic centimetre. When the team measured high-pressure hydrogen adsorption at 77 kelvin, CU-6-Fe took up 17.5 weight per cent at 100 bar and CU-6-Cr 15.9 weight per cent, with volumetric uptakes of 45.4 and 45.5 grams per litre. Under a realistic temperature–pressure swing cycle, from 77 kelvin and 100 bar down to 160 kelvin and 5 bar, the materials delivered 16.9 and 15.4 weight per cent gravimetrically and about 44 grams per litre volumetrically — figures that exceed the US Department of Energy’s 2025 system-level targets of 5.5 weight per cent and 40 grams per litre, albeit at cryogenic rather than ambient temperatures.

The methane results were equally striking. At 80 bar, CU-6-Fe stored 0.67, 0.59 and 0.53 grams of methane per gram of adsorbent at 273, 288 and 298 kelvin respectively, with CU-6-Cr close behind. These are among the highest gravimetric methane capacities ever reported and surpass the Department of Energy’s gravimetric target of 0.5 grams per gram at room temperature. The working capacity between 80 and 5 bar reached 0.49 grams per gram for CU-6-Fe at 298 kelvin, comfortably outperforming benchmark materials such as NU-1501-Al and the ultraporous covalent organic framework 3D-TFB-COF-Me. The researchers are candid about the trade-off: because the crystals are so light and porous, their volumetric methane capacities fall short of denser benchmark MOFs, a limitation inherent to ultrahigh-porosity adsorbents.

Computational and neutron experiments revealed exactly where the gas molecules sit. Grand canonical Monte Carlo simulations identified three main hydrogen adsorption regions, and in situ neutron powder diffraction on deuterium-loaded CU-6-Cr confirmed five distinct adsorption sites within the cavities. The dominant site, with an occupancy of 0.73, lies near the trinuclear metal cluster inside the cubic cavity, while secondary sites cluster around the phenyl and ethyne groups of the linkers. Density functional calculations established the energetic hierarchy: adsorption near the metal clusters is favoured at roughly minus 11 kilojoules per mole, compared with minus 2.4 kilojoules per mole at linker sites. This weak-but-preferential binding is precisely what enables both high total uptake and high deliverable capacity, since gas stored too strongly cannot be released on demand.

Stability data add to the material’s practical promise. CU-6-Cr retained its porosity after three months in an inert argon atmosphere and after 24 hours of exposure to ambient air, and its framework remained largely intact after exposure to water vapour — a notable achievement for a MOF of such extreme porosity. CU-6-Fe proved more fragile, losing about 20 per cent of its nitrogen uptake after a month in a glovebox, although X-ray absorption spectroscopy confirmed its iron coordination environment survived the hydrogen storage measurements unchanged. Beyond the headline numbers, the deeper significance of the work lies in its design principle: linker flexibility, long treated as a nuisance in crystal engineering, emerges as an essential, predictable parameter for constructing complex topologies. By matching conformationally adaptive molecules to demanding geometric targets, the Cambridge-led team has opened a route to an entire family of previously unreachable, ultraporous frameworks — and established the red net as a serious platform in the race to store the fuels of a carbon-neutral future.

Subject of Research: Hierarchically porous metal–organic frameworks with red topology for hydrogen and methane storage

Article Title: Hierarchically porous metal–organic frameworks with red topology for hydrogen and methane storage

Article References: Chen, X., Wee, L. H., Alizadeh Kiapi, M. R., Tampaxis, C., Fan, J., MacLeod, H. W., Zheng, Y., Moore, B. J., Manuel, P., Orlandi, F., Keenan, L. L., Zhang, E., Asgari, M., Menon, D., Charalambopoulou, G., Willhammar, T., Yang, S., Shepard, W., Cui, Y., … Fairen-Jimenez, D. (2026). Hierarchically porous metal–organic frameworks with red topology for hydrogen and methane storage. Nature Synthesis. https://doi.org/10.1038/s44160-026-01163-9

Image Credits: AI Generated

DOI: 10.1038/s44160-026-01163-9

Keywords: metal-organic frameworks, hydrogen storage, methane storage, reticular chemistry, red topology, porous materials, clean energy, linker flexibility, gas adsorption, crystal engineering, neutron diffraction, surface area

The new study breaks that limitation through what the researchers call anisotropic reticular chemistry. Reticular chemistry is the design principle behind most modern metal–organic frameworks: rigid molecular building blocks are linked by metal ions into extended, periodic networks whose geometry is dictated by the shape and connectivity of the components. In conventional systems, the ligands are highly symmetric, meaning every reactive site on the molecule behaves identically. When such a ligand meets metal ions in solution, all of its coordinating groups bind in the same way, and the framework that crystallizes is essentially the only one the chemistry allows. The Dresden-led team instead engineered a ligand with deliberately unequal reactivity across its structure: 1,2,5,6,9,10,12,13-octahydroxydibenzo-[fg,op]naphthacene, a large, planar, polycyclic aromatic molecule decorated with eight hydroxyl groups arranged as four catechol units. Catechols, pairs of adjacent hydroxyl groups on an aromatic ring, are classic chelating groups that bind metal ions strongly, but not all catechols on this ligand are created equal.

That inequality is the key to the whole strategy. Because the four catechol sites on the octahydroxy ligand possess different chemical reactivities, influenced by their electronic environment within the fused aromatic core and by the reaction conditions, the researchers could selectively activate some sites while leaving others dormant. When the ligand was combined with copper ions under one set of conditions, only the most reactive catechol pairs participated in coordination, producing a one-dimensional linear chain structure the team labels Cu-1D. Under different conditions, additional catechol sites were engaged, and the ligand stitched together into two-dimensional sheets. Depending on which subsets of sites were activated, the sheets adopted either a square-grid topology, designated Cu-2D-sql, or a honeycomb topology, designated Cu-2D-hcb. Three topologically distinct copper-based frameworks thus emerged from one single-type ligand, something conventional high-symmetry ligands cannot deliver. The structures were solved with remarkable precision using continuous rotation electron diffraction and synchrotron-based single-crystal X-ray diffraction, and the crystallographic data have been deposited with the Cambridge Crystallographic Data Centre.

The most striking demonstration of programmability came from the one-dimensional copper framework. Because Cu-1D retains unreacted catechol sites pointing outward from the linear chains, it behaves as a reactive intermediate rather than a dead end. The researchers found that these dormant sites could be activated in a subsequent step, allowing the introduction of a second metal ion. In a single-crystal-to-single-crystal transformation, a process in which the crystal lattice survives the chemical change intact, the linear chains were cross-linked into two-dimensional square-arranged sheets containing two different metals in defined positions. By choosing the second metal, the team produced four bimetallic frameworks, CuM-2D-sql, where M is nickel, cobalt, zinc or manganese. Each of these materials contains copper and the second metal arranged in an ordered, periodic pattern dictated by the original anisotropic ligand design. This level of control over bimetal node arrangement addresses a long-standing challenge in framework chemistry, where introducing multiple metals usually produces statistical mixtures or phase separations rather than atomically ordered heterostructures.

Why does such atomic-level ordering matter? In bimetallic conductive frameworks, the identity and spatial arrangement of the metal nodes directly shape the electronic band structure, the magnetic interactions and the catalytic properties of the material. Previous studies have shown that alloying different metals into conductive MOFs can continuously tune conductivity, and that heterometallic nanosheets can outperform their single-metal counterparts in both crystallinity and charge transport. But without positional control, it is difficult to attribute property changes to specific structural features or to design them rationally. The new anisotropic strategy turns metal arrangement from an accident of synthesis into a design parameter. The researchers could, in principle, place magnetic ions such as manganese or cobalt at precise intervals within a conducting copper lattice, creating model systems for studying spin ordering in two dimensions, or introduce catalytically active pairs of metals at fixed distances for cooperative electrochemistry.

The electronic consequences of the structural programming were verified through a combination of theoretical modelling and electrical measurements on individual crystals. Density functional theory calculations of the band structures showed that the different topologies and metal compositions produce distinctly different electronic states near the Fermi level, modulating how charge carriers move through the frameworks. The team then fabricated devices from single crystals and measured their charge transport directly, a technically demanding experiment because these framework crystals are small and delicate. The measurements confirmed that the structural variations introduced by anisotropic reticular chemistry translate into measurable changes in charge transport behavior, establishing a clear structure–property relationship across the family of seven materials. Semiconducting behavior, rather than metallic conduction, characterized these frameworks, which is significant for device applications where tunable band gaps and switchable transport are more useful than simple metallic wires.

The synthesis itself involved careful control of coordination chemistry principles that have been understood for decades but rarely combined in this way. The selectivity of metal binding to the different catechol sites reflects differences in their acidity and donor strength, and the researchers monitored the formation of the square-grid and honeycomb copper frameworks using ex situ powder X-ray diffraction, tracking how the crystalline products evolved over the course of the reactions. The single-crystal-to-single-crystal transformation from Cu-1D to the bimetallic sheets was verified crystallographically for each of the four second metals, with independent structure determinations confirming the ordered incorporation of nickel, cobalt, zinc and manganese into the expanding lattice. Computational analysis of reaction energetics and ligand acidity supported the mechanistic picture, explaining why certain catechol sites react first and why particular metal ions are accepted into the remaining coordination sites.

The broader significance of the work lies in what it suggests about the future of framework materials as programmable matter. Reticular chemistry has already given chemists extraordinary control over the pore sizes, shapes and chemical functionality of metal–organic frameworks, enabling applications from carbon capture to drug delivery. Extending that control into the electronic domain, where the exact sequence of metals and the topology of the conducting backbone determine the material’s physics, opens the door to what some researchers have called MOFtronics: electronic devices built from designed crystalline frameworks rather than conventional inorganic semiconductors. Porous field-effect transistors, chemiresistive gas sensors, supercapacitor electrodes and electrocatalysts based on conductive MOFs have all been demonstrated in recent years, and the ability to program both topology and metal ordering from a single ligand should accelerate the rational design of such devices.

There are, of course, challenges ahead. The anisotropic strategy currently relies on a specific, synthetically demanding ligand whose catechol reactivities happen to be distinguishable, and generalizing the approach to other ligand families and metal combinations will require similar molecular engineering. Growing the framework crystals large enough and uniform enough for practical device fabrication remains difficult, as it does for most 2D materials. Nevertheless, the demonstration that seven distinct semiconducting frameworks, spanning one-dimensional chains, two-dimensional square grids, two-dimensional honeycombs and four ordered bimetallic variants, can all be accessed from one molecular building block represents a conceptual advance. It reframes framework synthesis as a programming problem: by encoding unequal reactivity into a ligand, chemists can write different structural outcomes into the same molecule and read them out as different electronic properties. As the vocabulary of such anisotropic building blocks grows, the design space of atomically precise conductive materials expands with it.

Subject of Research: Anisotropic reticular chemistry for programming the topology, metal composition and charge transport of two-dimensional conjugated metal–organic frameworks

Article Title: Programming semiconducting two-dimensional conjugated metal–organic frameworks via anisotropic reticular chemistry

Article References: Zhang, J., Zhou, G., Un, H.-I., Song, J., Petkov, P. S., Iqbal, R., Fu, S., Huang, X., Zhang, G., Lu, Y., Pohl, D., Liu, R., Fu, Y., Li, X., Zhang, W., Rellinghaus, B., Alfonsov, A., Bon, V., Ding, J., … Dong, R. (2026). Programming semiconducting two-dimensional conjugated metal–organic frameworks via anisotropic reticular chemistry. Nature Chemistry. https://doi.org/10.1038/s41557-026-02251-1

Image Credits: AI Generated

DOI: 10.1038/s41557-026-02251-1

Keywords: metal–organic frameworks, reticular chemistry, two-dimensional materials, conjugated frameworks, bimetallic nodes, semiconductors, charge transport, coordination chemistry, crystallography, single-crystal devices, electronic materials, copper catecholate