A Membrane That Is 75% Crystal Passes CO2 at Very High Speed

Frontier Signal · Deep Dive · Energy & Carbon Capture · 8 min read · Monday, 12 October 2026

A power station or cement works breathes out a mix of gases. Most of it is nitrogen. A small slice, often 10 to 15%, is carbon dioxide. Capturing that CO2 means separating one gas from another at huge volume, and today that is mostly done with liquid chemicals called amines that soak up CO2 and are then heated to release it. It works, and it costs energy and money.

A membrane is a different idea: a thin film that lets some gases through faster than others, with no heating cycle. Membranes are simple and compact, but a good one needs two things that normally pull against each other. It must be fast (lots of gas through) and it must be picky (mostly CO2).

On 7 October 2026, Nature Energy published a membrane from a Chinese-Swiss-American team that is unusually fast. It does this by loading a polymer film with 75.5% single-crystal filler by volume, far more than the usual limit. The team calls the trick “weak-interaction inlaying”.

The speed is real on paper. What is missing, in the parts of the paper we could access, is the film thickness, the test conditions and any test on real flue gas. Here is the idea and what must be shown next.

74,800 BARRER CO2 PERMEABILITY · SELECTIVITY ABOUT 20 · 85% KEPT AFTER 6 MONTHS
Reported by the authors. The ageing result is for the polyimide version at room temperature. Selectivity means CO2 passes about 20 times more readily than nitrogen. Barrer measures how easily a gas moves through a material, whatever its thickness. Real membranes are thin films, so the number that sets cost is permeance, and we could not find that figure.

On 7 October 2026, Nature Energy published “All-organic mixed-matrix membranes with single-crystal covalent organic frameworks for carbon capture.” The corresponding authors are Guangwei He and Zhongyi Jiang of Tianjin University. Other authors include Kumar Varoon Agrawal at EPFL in Switzerland, with contributors from USC and several Chinese institutions. The paper is behind a paywall.

Global Times reported a statement from the team on 8 October. It said the permeability reached about 70% of the theoretical value for a pure COF membrane, and that the membrane held up in tests with simulated flue gas containing moisture, sulphur oxides and nitrogen oxides. Humidity levels, concentrations and test length were not given in what we saw.

Independent expert commentary has not appeared. The word “record” comes from the team’s framing as relayed by the press, not from text we could verify in the paper.

The old way: a mixed-matrix membrane is a plastic film with a fast-moving filler mixed in. The filler provides a quick route for CO2, and the plastic holds everything together. The trouble is that filler particles clump and leave gaps at the edges. Past modest loadings the film stops working, so practical designs use a small amount of filler.

The new way: the team uses single-crystal COF particles, which are small, regular, porous crystals built from organic building blocks. They tune the surface so the particles stay evenly spread and bond gently to the plastic without sticking to each other. That lets the filler reach 75.5% by volume. At that level the crystals touch, forming a continuous route for CO2.

The polymer and filler. The filler is described as COF-CH3. The ageing test used a polyimide plastic. We could not confirm which other polymers were tested.

In plain English: instead of dotting a few fast lanes into a slow road, the team has built a mostly-motorway road with just enough tarmac holding it together.

Selectivity matters as much as speed. A selectivity of about 20 is modest. If a membrane lets too much nitrogen through with the CO2, you need extra stages to get a pure stream. That adds equipment and energy. A fair test is how the numbers sit against the CO2/nitrogen trade-off curve known as the Robeson upper bound. We have not checked that position, so we do not claim one.

Thickness drives cost. Barrer is independent of thickness, but a plant buys square metres of film, so what counts is permeance, which does depend on thickness. Without the film thickness, we cannot convert 74,800 Barrer into permeance, and we cannot size a plant.

Compared with the best commercial membranes. Membrane Technology and Research (MTR), a US company, has run its Polaris membrane in pilots, including a roughly one-megawatt unit in Norway. MTR quotes performance as permeance, not Barrer, so the two cannot be compared directly. We did not verify any claim that this new membrane beats Polaris.

Capture is a cost problem. The challenge is not knowing how to separate CO2. It is doing so cheaply at the scale of a power plant or a cement kiln, a problem we have followed in cement’s push to capture CO2 by never letting air in. Amine capture has been quoted at $40 to $80 per tonne in a 2008 trade-press figure, and costs today vary by source.

Membranes have a different cost shape. They have no heating cycle and no chemical solvent to replace. Industry material says membranes are best suited to capture rates of 50 to 80%, not the 90% often targeted, with extra measures such as gas recycling to push higher. That makes them good for part of the job, not all of it.

Treat $38 per tonne as a hypothesis. Jiang called it preliminary and the article gives no detail on how it was calculated. For context, Global Times cites a 2025 Nature Sustainability study of a different, graphene-based membrane that estimated $25 to $50 per tonne for coal and cement and $50 to $100 for gas. Those are model estimates for different materials and conditions, so read them as direction, not proof.

There is no clean listed way to own this result. None of the sources we opened names a company licensing or manufacturing the membrane, and the lead authors are at universities. We are not naming tickers, because the evidence does not support it.

Who could be affected. If membranes win a slice of carbon capture, the potential effects reach amine-capture suppliers, industrial-gas companies with separation businesses and equipment makers. Those are categories, not recommendations, and we have not verified any specific company’s position.

The realistic route to money is licensing or a spin-out, then pilots, then a manufacturing partner. That has historically taken years. Carbon-capture economics also depend on policy: tax credits, carbon prices and local rules. Separation chemistry is a theme across energy, as in the race to pull lithium from brine without a pit.

Membrane Technology and Research (MTR) is the nearest commercial reference point. It has pilots in Norway, a planned larger pilot in Wyoming, and a long track record with its Gen 1 membrane. As far as we know it is privately held, which we have not verified.

Spin-outs from the Tianjin and EPFL groups are worth watching. None is announced in what we found. Agrawal’s EPFL group has a track record of membrane research, so a company is plausible, but that is speculation.

Gaps remain. Secondary reports list them and the paper preview does not contradict them: no large, defect-free membrane area has been shown; there has been no test on real flue gas; humidity, sulphur and temperature swings are still to be validated; and stability data cover six months at room temperature only. Power-plant gas is hot and wet, so room-temperature ageing is a first step, not a verdict.

Single-gas versus mixed-gas. Many high Barrer numbers come from tests on pure gases. Real flue gas is a mix, and CO2 can swell the plastic and cut selectivity. We could not tell which test the headline number comes from.

Making it. A lab-scale film with 75.5% crystal is one thing. A defect-free sheet at module size is another. Jiang says scale-up potential is strong and manufacturing costs controllable, which is a statement of intent rather than a demonstration.

What is the permeance of a defect-free film at flue-gas conditions, with real gas and over years?

Barrer is a lab number. Permeance (gas flow per square metre per unit of pressure), measured on a thin film, hot and wet, with a mixed feed, decides how many square metres a plant needs. Everything else follows from that figure.

A small, regular crystal made of organic building blocks with built-in pores. Here it provides fast channels for CO2.

A plastic film with a second material mixed in to improve performance. The challenge is getting enough of the second material in without clumping.

The team and press say so for permeability. Selectivity of about 20 is modest, and the record claim is not something we could confirm in the paper itself.

No. Tests used simulated flue gas, ageing data cover six months at room temperature, and no large module has been shown.

Treat it as preliminary. Team leader Jiang Zhongyi gave the figure to Global Times without detail, and other estimates are for different materials and settings.

The Nature Energy paper is paywalled and we saw the abstract and preview only. Secondary reports are flagged as such.

Frontier Signal explains frontier technology in plain English. Vendor-reported benchmarks should be independently verified. This is general information, not investment advice.