Data centres have an appetite for electricity. Servers need power around the clock, cooling systems need still more, and as artificial intelligence drives demand for computing upwards, finding enough electricity to keep everything running is becoming an engineering problem in its own right.
Hydrogen fuel cells offer one possible answer. Rather than drawing all their electricity from the grid, data centres could generate some of it on site, converting hydrogen directly into electricity. The technology is well established, but making fuel cells efficient, durable and affordable enough for demanding applications remains a challenge.
Researchers at Washington University in St. Louis have now developed a catalyst that tackles one particularly stubborn part of that problem.
At the heart of a low-temperature fuel cell is a catalyst that helps hydrogen and oxygen react to produce electricity, with water and heat as by-products. Platinum is particularly good at doing the job. Unfortunately, it is also a precious metal, so engineers have spent years trying to achieve the same performance with as little of it as possible.
Advertisement
One solution is to use nanoparticles. Breaking platinum into particles just a few nanometres across exposes far more surface area for chemical reactions, meaning a tiny quantity of the metal can do considerably more work. In fuel-cell catalysts, platinum loadings can typically be kept below a quarter of a milligram per square centimetre.
But there is a catch. Over time, these tiny particles can dissolve, migrate and grow larger, gradually reducing the catalyst’s performance.
The Washington University team approached the problem by concentrating not just on the platinum, but on the material holding it in place. Their design uses hollow carbon spheres filled with tiny channels arranged radially through the structure. Inside these channels sit nanoparticles made from platinum and cobalt.
That architecture allows the particles to be packed densely while remaining small and evenly distributed. More importantly, it allowed the researchers to heat the platinum-cobalt catalyst to 1,000°C without the nanoparticles clumping together. They remained below 5 nanometres in size.
The temperature matters. Platinum-cobalt intermetallic catalysts perform best when their atoms form a highly ordered structure, but achieving that normally requires high-temperature treatment. Traditionally, researchers have faced an awkward compromise: increase the temperature enough to create the desired atomic structure and the nanoparticles tend to grow and bunch together. Keep the temperature below around 700°C to preserve their size and the atoms may not become sufficiently ordered.
Advertisement
The carbon nanochannels change that balance. By physically confining the particles, they allow the catalyst to withstand much higher processing temperatures while keeping the platinum-cobalt nanoparticles small and well dispersed.
Tests suggest that makes a substantial difference to durability. After 150,000 severe voltage cycles, the catalyst retained 85% of its performance. The researchers estimate that this could equate to roughly 25,000 hours of operation.
There is another advantage to the unusual carbon structure. Its open channels provide easier routes for protons, oxygen and water to travel through the electrode, while helping the ion-conducting material within it spread more evenly. In other words, the structure isn’t simply stopping the platinum particles from moving around. It is also helping the fuel cell do its job.
The work, published in Nature Nanotechnology, involved researchers from Washington University, Brookhaven National Laboratory, Lawrence Berkeley National Laboratory, Northeastern University and the University of Pittsburgh. A patent has also been filed on the technology.
There is still a considerable distance between a new catalyst and a fuel cell supplying a commercial data centre. Further development and work with industry will be needed before the researchers’ design can make that jump.
The attraction, however, is clear. Data centres could account for as much as 9% of annual US electricity generation by 2030, according to an Electric Power Research Institute estimate cited by the researchers, up from 4% of electricity demand in 2023. Generating electricity locally with fuel cells could reduce some of that pressure on the grid.
And the potential doesn’t stop at data centres. The same catalyst technology could eventually improve fuel cells used for transport and other forms of electricity generation. The engineering problem in each case is much the same: make a small amount of an expensive metal work harder, and make it keep working for much longer.
Research published in Nature Nanotechnology.