Chemically Tuned Hollow Fibers Push Carbon Membranes to New Heights in Gas Separation

Hydrogen has long been heralded as a cornerstone of the low-carbon economy, yet the infrastructure needed to move it from production sites to end users remains one of the biggest obstacles to its widespread adoption. Building a dedicated hydrogen pipeline network would cost billions of dollars, so researchers have increasingly looked at a clever workaround: blending hydrogen into the vast natural gas pipeline systems that already crisscross continents. The problem is that hydrogen concentrations in such blends must be kept low, typically around 10 percent or less, to avoid hydrogen embrittlement of pipeline steel. That means anyone wanting pure hydrogen at the point of use needs an efficient, economical way to extract it from the dilute mixture. A new study published in Advanced Science reports a scalable membrane technology that may finally make this vision practical, achieving hydrogen purities of 99.8 percent directly from a 10 percent hydrogen–methane blend in a single membrane stage.

The research, led by a team working with carbon molecular sieve (CMS) hollow fiber membranes, tackles a challenge that has dogged the field for decades: how to manufacture ultra-selective carbon membranes in a form that can be produced at industrial scale without sacrificing performance. CMS membranes work by pyrolyzing polymer precursors at high temperatures, in this case 900 degrees Celsius, to create carbon structures with ultramicropores so precisely sized that they can discriminate between molecules differing by fractions of an angstrom. Flat-sheet versions of these membranes have posted extraordinary selectivities before, including a cellulose-derived membrane with a hydrogen-to-methane selectivity of 3,498 and a hollow fiber variant reaching 5,506 at 130 degrees Celsius. But translating those laboratory achievements into hollow fibers, the geometry favored for industrial deployment because of its high surface-area-to-volume packing density, has proven stubbornly difficult.

The breakthrough hinges on a chemical pretreatment of the polymer precursor before carbonization. The team started with Matrimid, a commercially available and relatively inexpensive polyimide, spun into asymmetric hollow fibers with a thin selective layer of roughly 150 nanometers atop a porous support. They then treated these fibers with a dilute solution of diethyltoluenediamine (DETDA) followed by trimesoyl chloride (TMC), both dissolved in hexane. This sequence cross-links the entire fiber matrix, including both the selective skin and the porous substructure, transforming the polymer into a network precursor with fundamentally different pyrolysis behavior. Spectroscopic analysis confirmed the chemistry: Fourier transform infrared spectra showed the ratio of amide to imide peaks rising from 0.52 to 0.71 after treatment, while x-ray photoelectron spectroscopy revealed nitrogen concentrations climbing from 5.9 to 8.0 atomic percent and oxygen from 14.9 to 18.5 atomic percent, clear fingerprints of the grafted diamine and acyl chloride.

Crucially, the researchers paired this cross-linking with a second stabilization step using vinyltrimethoxysilane (VTMS) sol-gel chemistry. Without such protection, the porous support layer of the hollow fiber tends to collapse completely during the intense heat of pyrolysis, densifying the entire wall and choking off gas transport. In previous work by the same group, fibers lacking VTMS treatment suffered exactly this fate. With the dual DETDA/TMC and VTMS treatment, the final carbon membrane retains a hierarchical architecture: a dense carbon separation layer only about one micrometer thick, supported by an open, porous substrate. This combination resolves what the authors describe as a long-standing collapse-stability trade-off, and it does so using cheap commodity chemicals rather than the expensive ionic-liquid solvents or fluorinated polyimides that earlier approaches required.

The separation results are striking. In mixed-gas tests using a feed of 10 percent hydrogen and 90 percent methane at 35 degrees Celsius and 100 psi, the dual-treated membranes delivered hydrogen at 99.56 percent purity one day after pyrolysis, with a permeance of 19.9 gas permeation units. After 24 days of aggressive vacuum aging, the same membranes actually improved, reaching a mixed-gas hydrogen-to-methane selectivity of 2,639.8 with 15.6 GPU permeance and a downstream hydrogen purity of 99.8 percent. For comparison, the celebrated cellulose-derived CMS membrane achieved a pure-gas selectivity of 5,506 only at an elevated 130 degrees Celsius, and its performance at practical ambient temperatures would be substantially lower. Both of the new membranes also surpass the 2015 upper bound for polymer membranes in hydrogen-methane separation, a benchmark that has defined the frontier of polymeric materials for a decade.

The same platform proved equally adept at a second, very different separation problem: recovering helium from natural gas. Global helium demand is growing by roughly two million cubic meters per year, driven by applications ranging from MRI magnets to semiconductor manufacturing, yet most helium-bearing reservoirs contain only between 0.2 and 4 percent helium. Conventional cryogenic distillation is energy- and capital-intensive, and ordinary polymeric membranes suffer from an intrinsic trade-off between permeability and selectivity that limits their efficiency. Testing their membranes against a feed of 5 percent helium and 95 percent methane, the researchers found that the dual-treated CMS achieved 98.8 percent helium purity, rising to 98.96 percent after aging as the selectivity climbed from 1,617.8 to 2,000.8. The VTMS-only variant delivered 99.3 percent purity initially. These purities, achieved in a single membrane stage from such a dilute feed, represent a proof of concept that could reshape how helium is harvested.

Perhaps the most intriguing finding is that aging, usually the enemy of membrane performance, actually improved the dual-treated membranes. Over 24 days of vacuum aging at 35 degrees Celsius, the helium permeance of the DETDA/TMC-plus-VTMS membrane fell by only 5.9 percent, compared with a 29 percent drop for the VTMS-only version, while methane permeance fell sharply by 38 percent. The researchers attribute this to the tightening of imperfectly packed carbon structures: cross-linking introduces small defects into the selective Langmuir domains during carbonization, and controlled aging closes those defects into narrower slits that still admit tiny, spherical helium atoms but increasingly exclude the bulkier methane molecule. Probe-gas experiments with molecules of varying kinetic diameters confirmed a molecular sieving mechanism, with a sharp permeance cutoff between small gases like hydrogen and helium and larger ones like carbon dioxide, oxygen, and nitrogen.

The underlying physics of why DETDA/TMC treatment changes the carbon structure so profoundly remains partly hypothetical, and the authors are candid about this. They propose that the higher density of cross-linked sites promotes stress-induced chain scission during pyrolysis, yielding shorter carbon strands that assemble into smaller carbon plates which stack more tightly. This tightening reduces permeance for medium-sized gases and enhances oxygen-nitrogen selectivity, while partially retained cross-linked branches create a small population of larger bypass pores that raise methane permeance. The data cannot yet establish whether the chemical modification distributes uniformly through the dense selective layer, and the team notes that multistage process design and techno-economic analysis will be needed to quantify real-world benefits. Still, the empirical performance under realistic mixed-gas, high-pressure, aged conditions speaks for itself.

What makes this work resonate beyond the laboratory is the paradigm it suggests. Rather than designing an entirely new polymer for every separation, the method takes ordinary, defect-free asymmetric hollow fibers already spun from cheap commercial materials and chemically retunes them into stabilized network precursors with controllable pyrolysis behavior. This unlocks a vast pool of commercially available polyimides that were previously unusable for CMS membranes because highly cross-linked polymers cannot be spun directly into fibers. The researchers envision extending the approach to different diamines and acid chlorides, and to gas pairs far beyond hydrogen-methane and helium-methane. If the vision holds, the natural gas pipeline network itself could become a distributed hydrogen delivery system, with membrane modules sited at points of use quietly sifting ultrapure hydrogen from the flowing blend, and low-concentration natural gas wells becoming viable helium sources. A simple chemical treatment, applied before the fire, may have redrawn the map of what molecular sieves can do.

Subject of Research: Scalable carbon molecular sieve hollow fiber membranes for hydrogen and helium separation from natural gas

Article Title: Transforming Gas Separations with Scalable Carbon Molecular Sieve Membranes Enabled by Chemically Tuned Hollow‐Fiber Precursors

Article References: Cao, Y., Liu, Z., Shomali, Z., Kim, J., Schlosser, S., Lively, R. P., & Koros, W. J. (2026). Transforming Gas Separations with Scalable Carbon Molecular Sieve Membranes Enabled by Chemically Tuned Hollow‐Fiber Precursors. Advanced Science, Article e78223. https://doi.org/10.1002/advs.78223

Image Credits: AI Generated

DOI: 10.1002/advs.78223

Keywords: carbon molecular sieve membranes, gas separation, hydrogen purification, helium recovery, hollow fiber membranes, pyrolysis, Matrimid polyimide, cross-linking, mixed gas separation, natural gas pipelines, membrane aging, molecular sieving