I find that the easiest way to think about why our chemistry is different is to ask what Nafion was actually designed for. It's a question that gets skipped a lot, because Nafion has been the default for so long that people treat it as a baseline of nature rather than a design choice. But it's a design, and like all designs it was optimized for a specific set of constraints, and those constraints aren't the ones that matter today.

Nafion was designed in the 1960s for two applications. The first was the chlor-alkali industry, which uses electrolysis to make chlorine and caustic soda. Chlor-alkali cells operate at moderate current densities, with high concentrations of salt water on one side and high concentrations of caustic soda on the other. The membrane's job is to keep the two streams separated while letting sodium ions through. The chemistry had to handle a brutal corrosive environment and run for a decade without degrading, but it didn't have to do anything fancy — no high current density, no rapid cycling, no extreme pressure differential, no integration with intermittent renewables.

The second application was NASA fuel cells, specifically the Gemini program. These were small cells running at low current density, in a sealed vehicle, for two-week missions. The astronauts breathed the byproduct water. The constraints were strict but narrow: don't poison the crew, don't catch fire, work in zero gravity, run reliably for two weeks. The cells didn't have to last for thirty years. They didn't have to cycle on and off thousands of times. They didn't have to handle the pressure and current density of modern electrolyzers. They didn't have to be cheap.

Nafion was brilliant for both of these. Walther Grot designed a polymer that did exactly what those two customer sets needed. The chlor-alkali industry adopted it, NASA flew it, and within a decade it had become the default. By the1980s it was the standard for every emerging PEM application, and by the 1990s the small optimizations were done. The chemistry has been frozen at roughly that performance level ever since.

Now think about what modern applications actually need the membrane to do. A PEM electrolyzer running on intermittent renewable input cycles its membrane through huge swings in current density, with shocks at startup and shutdown. A fuel cell in a heavy-duty truck runs through forty thousand cold starts and stops over a million-mile life. An electrolyzer in a high-pressure hydrogen production system has to hold a sharp pressure differential without letting hydrogen cross over to the oxygen side, which would be a safety problem. A data center backup fuel cell runs at base load for years without interruption. A flow battery cycles its electrolyte through the membrane tens of thousands of times. A CO₂ reduction reactor needs the membrane to be selective against specific anions that didn't matter in chlor-alkali.

These are very different jobs than the ones Nafion was designed for, and the consequence is that the entire industry has been working around the membrane's limitations rather than designing the membrane for its applications. Stack architectures have evolved to compensate for membrane crossover. Catalyst loadings have been kept artificially high to compensate for membrane resistance. System designs have included buffer steps and recovery cycles to compensate for membrane sensitivity to input variation. The membrane has been the constraint, and the rest of the system has been bent around it.

What we did, conceptually, was to start from the other direction. Take a list of the things the membrane needs to do for the next several decades of industrial use, design an architecture to do those things, and accept that this would require rethinking the molecular structure and membrane architecture rather than tweaking the side chains of Nafion.

The architecture isn't a single change. It's a stack of changes. A different backbone chemistry for higher thermal and mechanical stability. A different reinforcement strategy for handling pressure differentials. More layers, so that properties of different materials can be combined. A different surface treatment for better catalyst adhesion at lower loadings. A different manufacturing approach so that the whole thing can be produced on roll-to-roll equipment that already exists. Each change isincremental in isolation. Together they're a redesign.

The numbers that come out the other end are the ones that matter at the system level. Five times the durability. Seven times the current density. Eighty percent less iridium per stack. These aren't features of the membrane in some abstract benchmarking sense. They're what the membrane lets the rest of the system do.

I think the reason this redesign didn't happen earlier isn't that the chemistry was impossible. It's that the field had organized itself around a substrate that worked well enough for chlor-alkali and NASA and never went back to question whether the same substrate was the right one for everything else. Once you ask the question, the answer is fairly clearly no.

The thing I find satisfying about this work, even apart from the commercial implications, is that it answers a question that should have been asked thirty years ago. The membrane wasn't a mystery. It was a design choice. The design was for the 1960s. The world is using it in the 2020s, and the world is not the 1960s.

We designed for the next fifty years instead. That's the part I think is worth talking about.

‍