Channels Narrower Than a Virus Could Stretch Fuel Cells Further

A heating holder inside an electron microscope carried a speck of catalyst powder from room temperature to 1,000 degrees Celsius, and the detector watched the whole way up. At 25 degrees there were no particles to see, only platinum and cobalt spread across the carbon as single atoms. By 400 degrees those atoms had pulled together into visible clusters. What happened over the next 600 degrees depended entirely on which carbon they were sitting on.

On an ordinary carbon support, the clusters swelled to more than 9 nanometers across by the time the holder reached 1,000 degrees. On a support that Gang Wu‘s group at Washington University in St. Louis had built as a hollow sphere threaded with channels running out from the center like spokes, the same particles grew to 4 nanometers and stopped there.

That gap matters because the inside of a hydrogen fuel cell is where platinum earns its keep, and platinum is scarce: roughly 70 percent of the world’s mined supply comes out of South Africa. Fuel cells are being pushed at exactly the loads that are growing fastest. U.S. data centers consumed about 4.4 percent of national electricity in 2023, a share the U.S. Department of Energy expects to roughly double or triple by 2028. So far, fuel cells have entered that world as backup power rather than primary supply, as in a 1.5 megawatt demonstration that stood in for diesel generators through a simulated two day outage at a Microsoft site in Wyoming.

Making platinum go further means grinding it into particles a few nanometers wide, because the smaller the particle, the more of the metal’s surface is exposed to do the work. Mixing in cobalt helps too, but only if the two metals settle into a neat alternating pattern rather than jumbling together, and getting them to line up that way means heating them until the atoms have enough energy to shuffle into place. The trouble is that the same heat sets the whole particle moving. Particles wander across the support, bump into each other, and merge, and once they are inside a running cell they also dissolve and rebuild themselves onto their larger neighbors. Most labs settle for lower temperatures, accept a sloppier atomic arrangement, and live with it.

The Particles Went Into the Channels

To build the support, the researchers grew a carbon shell around tiny grains of silica, baked the whole thing, then washed the silica out with a hot caustic solution. What is left is a hollow sphere less than a thousandth of a millimeter across, riddled with pores about 11 nanometers wide. That is narrower than the smallest virus known.

They then soaked the spheres in platinum and cobalt salts, freeze dried the mixture, and baked it again. Slicing the spheres open afterward showed the metal particles were not in the hollow center or on the outer skin. More than 97 percent of them were lodged inside the channels. That fits with earlier work finding that porous carbons keep metal particles from clumping in a way that solid ones cannot, and it points to the channel walls as the thing doing the holding.

Wu, who conceived the project with first author Lei Gao and has filed a patent on the material through Washington University, described the effect in terms of the compromise it avoids. “Traditionally, there would be a trade-off between size and stability, but with the ordered carbon nanochannel host, platinum cobalt nanoparticles can be confined and remain stable at very small particle size even at high temperatures,” he said in a university release.

Order Improved Until the Particles Got Too Big

X-ray measurements taken after each heating step confirmed what the researchers were after: the hotter the bake, the more neatly the platinum and cobalt atoms arranged themselves. Under a powerful microscope, a single particle from the 1,000 degree batch showed the two metals stacked in alternating columns, wrapped in a platinum coating three atoms thick.

The channel walls appear to be what keeps those particles from wandering. The carbon is not smooth. It is rough and grooved, and calculations suggest a groove of the right size grips a particle far more tightly than a flat surface would.

More heat is not simply better, though. Pushing the bake to 1,150 degrees bought only a little extra atomic order while letting the particles swell to roughly 7 nanometers, and the catalyst performed worse for it. The sweet spot at 1,000 degrees turned out to depend less on the metal than on the pores: when the researchers made the pores narrower or wider than 11 nanometers, the particles came out bigger either way.

The Cell Ran to 150,000 Cycles

Built into a small test cell and run under conditions meant to mimic a heavy truck, the catalyst produced about a third more power than a commercial platinum cobalt catalyst did. Then the researchers put it through the electrical equivalent of hard mileage, swinging the voltage up and down 150,000 times. It came out still delivering 82.5 percent of its original output. The commercial benchmark had already given up more than half of its performance after 30,000 swings. Particles recovered from the aged cell measured 4.5 nanometers, against 4.4 at the start.

What a cycle count can’t supply is hours. The U.S. Department of Energy’s testing protocol for heavy-duty catalysts calls for 90,000 voltage swings, and the group that wrote it is still working out how many hours of real service those swings stand in for. Turning 150,000 cycles into years on a truck or in a server rack is not yet a calculation anyone can make with confidence. The tests also ran in cells about the size of a postage stamp, and the largest batch of the new carbon the group reports making is 10 grams.

And the cobalt is still leaving. When the researchers pulled apart a cell that had run 90,000 cycles, they found the neat alternating pattern had loosened slightly and the platinum coating on each particle had thickened, both signs of cobalt working its way out of the particles and into the cell. The channels stopped the particles from moving. They did not stop the atoms inside them from rearranging.

FAQ

Why does a fuel cell need platinum at all? The cathode reaction pulls oxygen from air and splits its double bond, which is sluggish. Platinum lowers the barrier better than any cheaper metal tested at scale, so the practical goal is to use less of it rather than replace it.

What does “ordered” mean here? In a disordered alloy, platinum and cobalt atoms occupy lattice sites at random. In the ordered intermetallic phase they alternate in a fixed pattern. That arrangement tunes how strongly the surface binds oxygen and holds the cobalt more tightly against dissolution.

Why not just skip the high temperature step? Below about 700 degrees the atoms lack the energy to rearrange into the ordered pattern, and the resulting catalysts are typically less than 40 percent ordered. The heat is what buys the ordering.

Does 150,000 cycles mean the catalyst would last 150,000 hours? No. Voltage cycling is an accelerated proxy that compresses years of on and off operation into days of testing. Converting cycles to service hours requires a validated protocol, and the standard for heavy-duty fuel cells is still being finalized.

Could this power a data center? Not on the evidence here. The study measures catalyst performance in a 5 square centimeter cell. Data center deployments to date are megawatt scale backup systems, and reaching them would require stack integration, a hydrogen supply, and manufacturing at a scale the group has not yet attempted.

What is still unresolved? Cobalt continues to migrate out of the particles during operation, thickening the platinum shell and reducing the ordered fraction. How far that goes over a full service life, and whether the confinement that stops physical coarsening also slows the chemistry, has not been measured.