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

Graphitic carbon nitride, known chemically as g-C3N4, is a polymeric semiconductor composed of carbon and nitrogen arranged in layered sheets reminiscent of graphite. It has attracted intense scientific interest because it is chemically stable, non-toxic, inexpensive, and remarkably easy to synthesize from common nitrogen-rich precursors. Crucially, it absorbs visible light, meaning it can in principle be driven by ordinary sunlight rather than expensive ultraviolet lamps. When photons strike the material, they excite electrons from the valence band to the conduction band, leaving behind positively charged holes. These photogenerated charge carriers can then react with water, oxygen, and dissolved species to produce highly reactive oxidizing agents capable of shredding organic pollutant molecules into harmless fragments.

Yet pristine g-C3N4 suffers from well-documented weaknesses that have kept it out of real-world deployment. Photogenerated electrons and holes recombine rapidly, releasing energy as heat before they can do useful chemical work. The material’s surface is comparatively inert, limiting its ability to adsorb and react with target pollutants, and its visible-light absorption, while real, does not extend deeply enough into the solar spectrum. The Hanoi review systematically examines how researchers have overcome these limitations by constructing hybrid photocatalysts, or CN-HPs, in which g-C3N4 is coupled with other materials to form sophisticated junctions that separate charges, extend light harvesting, and boost surface reactivity.

The strategies catalogued in the review are strikingly diverse. Heterojunctions pair g-C3N4 with metal oxides such as ZnO, TiO2, WO3, and BiVO4, or with more exotic partners including MXenes, layered double hydroxides, perovskites, and metal-organic frameworks. Z-scheme and S-scheme architectures are particularly elegant: they mimic natural photosynthesis by selectively retaining the most energetic electrons and holes on different components while allowing the useless carriers to recombine, thereby preserving strong oxidation and reduction power simultaneously. Doping with elements such as sulfur or phosphorus tunes the electronic structure, while carbon quantum dots and noble metal nanoparticles act as electron sinks and co-catalysts. Magnetic components such as Fe3O4 even allow the catalyst particles to be retrieved from treated water with a simple magnet, addressing one of the classic headaches of nanoparticle-based treatment.

The performance figures reported across the literature are impressive, at least under laboratory conditions. CN-HPs have achieved high removal efficiencies for a wide range of pharmaceutical pollutants, including tetracycline, ciprofloxacin, levofloxacin, oxytetracycline, amoxicillin, sulfamethoxazole, ibuprofen, naproxen, acetaminophen, and the anticancer drug 5-fluorouracil. Many systems operate under visible light or even natural sunlight, and some have been tested in continuous-flow configurations and real hospital or mariculture wastewater matrices. The review emphasizes that these materials also retain their activity over repeated catalytic cycles, a critical requirement for any technology hoping to move from the bench to a treatment plant.

Beneath the headline efficiency numbers lies a more nuanced chemical story. The review details how degradation proceeds through the concerted action of three principal reactive species: superoxide radicals, hydroxyl radicals, and photogenerated holes. Each attacks drug molecules at characteristic bonds, cleaving rings and side chains to produce a cascade of smaller intermediate compounds. Ideally, these intermediates are then progressively mineralized into carbon dioxide, water, and inorganic ions. However, the authors highlight an important caveat that recurs throughout the literature: total organic carbon reduction and complete mineralization typically require substantially more time than the initial disappearance of the parent pollutant. In other words, a water sample may test clean for the original drug while still carrying a cocktail of partially oxidized fragments.

That gap between degradation and detoxification is where the review delivers its most consequential warning. Toxicity assessments show that the by-products formed during photocatalytic degradation can carry toxicity profiles different from, and in some cases potentially greater than, those of the original compounds. High removal efficiency alone, the authors argue, does not demonstrate environmental safety. A treatment system that rapidly converts an antibiotic into unidentified fragments without verifying their ecological impact may simply be trading one hazard for another. The review therefore calls for toxicity evaluation to become a standard, non-negotiable component of photocatalysis research, rather than an optional add-on performed in only a minority of studies.

The authors are equally candid about the obstacles standing between laboratory success and field deployment. Most published experiments use purified water, single pollutants at concentrations far above environmental levels, and idealized light sources. Real wastewater presents a hostile environment: competing organic matter scavenges radicals, suspended solids scatter light, mixtures of drugs interact unpredictably, and pH fluctuates continuously. Long-term environmental safety of the nanomaterials themselves, including potential release of catalyst components into treated water, remains insufficiently characterized. Energy efficiency and the engineering challenge of scaling thin-film or slurry reactors to the volumes handled by municipal plants are also unresolved. The review frames these not as reasons for pessimism but as a well-defined research agenda.

One of the most forward-looking threads in the review concerns artificial intelligence. The authors highlight the potential of combining experimental photocatalysis data with machine learning models to optimize treatment conditions and predict how CN-HP systems will perform against specific target pollutants in the field. Given the enormous design space of possible hybrid materials, dopants, junction types, and operating parameters, data-driven approaches could dramatically accelerate the search for formulations that balance efficiency, stability, cost, and safety. Similar AI-assisted strategies are already gaining traction across membrane design and pollution monitoring, and photocatalysis appears poised to follow the same trajectory.

What emerges from the Hanoi analysis is a technology at an inflection point. The fundamental chemistry works: metal-free, sunlight-driven catalysts can genuinely dismantle the pharmaceutical residues that conventional treatment leaves behind, and they can do so repeatedly and, in many configurations, using abundant and benign raw materials. The remaining task is to prove that this chemistry survives contact with the messy, variable, and economically constrained reality of full-scale water treatment, and that the end products of degradation are demonstrably safe for the ecosystems that receive them. If researchers can close the gap between breaking molecules and proving detoxification, graphitic carbon nitride hybrids may become a cornerstone of the next generation of water purification infrastructure.

Subject of Research: g-C3N4-based hybrid photocatalysts for photocatalytic degradation of emerging pharmaceutical pollutants in water and wastewater

Article Title: g-C3N4-based hybrid photocatalysts for removal of emerging pharmaceutical pollutants from water: recent advances, degradation pathways, toxicity assessment,and future perspectives

Article References: Hanh, D. T. M., & Hoai, P. T. T. (2026). g-C3N4-based hybrid photocatalysts for removal of emerging pharmaceutical pollutants from water: recent advances, degradation pathways, toxicity assessment,and future perspectives. Environmental Geochemistry and Health, 48(16), Article 625. https://doi.org/10.1007/s10653-026-03530-z

Image Credits: AI Generated

DOI: 10.1007/s10653-026-03530-z

Keywords: graphitic carbon nitride, photocatalysis, pharmaceutical pollutants, water treatment, heterojunctions, advanced oxidation processes, degradation pathways, toxicity assessment, antibiotics, mineralization, artificial intelligence, wastewater remediation