A new membrane traps 'forever chemicals' in water and breaks them down

For the past decade, the standard treatment for PFAS (per- and polyfluoroalkyl substances, the "forever chemicals") in drinking water has generally been a relay: pull the compounds out of the stream with activated carbon, ion exchange, or a membrane filter, then ship the concentrated waste somewhere else to be destroyed. A palladium-coated membrane published this week in PNAS collapses that relay into a single ambient-condition step, capturing six of the most common "forever chemicals" and breaking the carbon–fluorine bond that makes them persist.

The system, built by a team led by Bruce Rittmann at Arizona State University's Biodesign Swette Center for Environmental Biotechnology with collaborators at Rice, Tongji, the National University of Singapore, and Nanjing, threads hydrogen gas through hollow fibers whose walls are coated with palladium. The gas lands on the metal's surface as reactive atomic hydrogen (Hads*). Some of those atoms migrate just below the surface and stockpile as a beta-phase palladium hydride (beta-PdHx), a subsurface reservoir that releases the driving force for stepwise hydrodefluorination: cleaving one C–F bond at a time under ambient temperature and pressure, in a paper published October 19 in PNAS (DOI 10.1073/pnas.2613985123). In separate batch runs, the team reports that delivering H₂ directly through the membrane roughly doubled the fluoride-release rate compared with H₂ bubbled above the water, while leaving PFOA removal roughly unchanged. That suggests the membrane's role as a hydrogen delivery vehicle, not just a catalyst support, is part of the gain.

The 58-day continuous test fed a synthetic water stream containing 50 micrograms per liter each of PFOA, PFOS, PFNA, PFHxS, PFBS, and GenX, six compounds the paper describes as PFAS of regulatory concern. Brighter Side's coverage of the results reports compound-specific removals of 95% (PFOA), 97% (PFOS), 99% (PFHxS), 98% (PFNA), 94% (PFBS), and 93% (GenX), with 83% overall defluorination, the share of the carbon–fluorine bonds actually broken, not just removed from the water. Long-chain compounds came apart faster than short-chain analogs; Brighter Side's account attributes the stronger performance of longer chains to their binding more tightly to the catalyst, with short-chain PFBS and ether-containing GenX proving more resistant. The batch runs that compared delivery routes used pH 4 conditions with pressurized hydrogen; the continuous-flow test ran at ambient pressure.

GenX was the holdout. The team's mechanistic read in PNAS pins the gap on its ether linkage, a C–O–C bridge the other five compounds lack, weaker adsorption on the catalyst, and a higher average C–F bond dissociation energy. The trade press roundup on the result frames GenX as the test case that decides whether a hydrodefluorination system is broad-spectrum or only a long-chain specialist.

The honest scale of all those numbers is small. The reactor pushed 0.05 milliliters per minute, or roughly 72 mL per day, through a 12-hour hydraulic residence time, processing about 4 liters of synthetic water over the full 58-day run, according to ZME Science's write-up. That is enough to prove the chemistry, not enough to prove a treatment train. The press positions the system as a downstream destroyer that could sit behind the existing activated-carbon, ion-exchange, or membrane-filtration concentration steps, polishing the concentrate those processes still have to ship off-site.

The cost sits on the catalyst. Palladium trades around $50,000 per kilogram, and the membrane shed roughly 10% of its palladium over the 58-day run, per the trade press. The open question is whether each kilogram of PFAS destroyed earns back the palladium it costs to lose. The team has a bibliographic record on Crossref and a Semantic Scholar entry for the published method, and the durability number at pilot scale is the next thing the deployment math needs.