Platinum single-atom and cluster catalysis of the hydrogen evolution reaction

If you’ve ever looked at a platinum nanoparticle under a microscope and thought, "Wow, that’s a lot of precious metal sitting in the middle doing nothing," you’re not wrong. Platinum is the gold standard for making hydrogen from water. However, in the classic "nanoparticle on carbon" setup, most platinum atoms are buried in the core and can’t participate in the reaction. The dream has been clear for years: place every atom where it can function. One by one, spread them across a support like thumbtacks on a cork board. The challenge is keeping them from wandering, clumping, and becoming inactive under real electrochemical conditions. Cheng, Sun, and colleagues tackle that problem directly. They utilize atomic layer deposition, or ALD, which grows films one molecular layer at a time, to position isolated platinum atoms and tiny clusters onto nitrogen-doped graphene nanosheets. The nitrogen is crucial. The nitrogen atoms in the carbon lattice create chemical "seats" that capture and hold platinum. Using Me C p P t Me three as the platinum source and oxygen as the co-reactant, they grow at two hundred fifty degrees Celsius. They can adjust the amount of platinum simply by changing the number of ALD cycles they perform. Fifty cycles result in about 2.1 weight percent platinum; doubling to one hundred cycles increases that to 7.6 percent. Same method, different dosage. Now, if you are going to claim "single atoms," you need to show them. They do. Aberration-corrected scanning transmission electron microscopy—the technique where a single heavy atom glows like a pixel-sized lighthouse—reveals what you would hope to see after fifty cycles: lots of lone platinum atoms, a scattering of two or three atom clusters, and only a few larger particles. By pushing to one hundred cycles, both single atoms and small clusters are still present, but more sizeable clusters start to emerge. In other words, the nitrogen-doped graphene is doing its job as an anchor, but if you load it heavily enough, some sintering occurs. There’s also a spatial aspect: deposition prefers the doped edge sites on the nanosheets, where nitrogen’s lone pairs create particularly sticky docking points for platinum. So, does this atomic strategy actually pay off at the electrode? They test hydrogen evolution in acid using the classic 0.5 molar sulfuric acid in a rotating-disk setup, and the answer is a strong yes. At 0.05 volts versus the reversible hydrogen electrode, the mass activity of the fifty-cycle sample measures in at 10.1 amperes per milligram of platinum. Under the same conditions, commercial platinum on carbon achieves just 0.27. That’s the power of utilizing nearly every atom. If you care about current per geometric area, the results are similar: the low-loading ALD sample produces more current than platinum on carbon at the same tiny overpotential without requiring a large number of nanoparticles. Kinetically, the story aligns. When they fit the polarization curves to the Tafel relation—which is the straight-line log current versus overpotential plot that indicates the speed of the mechanism steps—the ALD single-atom catalyst shows a slope of 29 millivolts per decade, just slightly below the 31 millivolts per decade achieved with platinum on carbon. That may seem like a minor detail, but in electrochemistry, every millivolt per decade is hard-earned. It suggests slightly faster reaction steps on those isolated platinum sites and tiny clusters. Now, durability. Single atoms sound delicate, and in many instances, they are. Here, the nitrogen-doped graphene proves its worth. After one thousand accelerated cycles sweeping between negative and positive potentials, the fifty-cycle ALD material retains about 96 percent of its initial current at 0.05 volts, while the commercial platinum on carbon decreases to roughly 81 percent. In a separate constant-potential hold at a modest overpotential, the ALD catalyst continues to perform well, but the platinum on carbon current declines. When they return to the microscope, the ALD sample looks much like it did at the start—perhaps a bit of growth but no islands merging into larger particles. On the commercial carbon, the nanoparticles clearly coarsen, showing bigger grains, fewer active surface atoms, and poorer performance. It’s the classic Ostwald ripening story, only now you can observe how a chemical tether—a platinum-nitrogen bond—can slow that process. Next, spectroscopy connects structure to electronics. X-ray absorption near-edge structure, or XANES, is sensitive to empty states in the platinum 5d band, and a simple trend emerges here: the white-line intensity, which tracks those unoccupied 5d states, is highest for the fifty-cycle ALD catalyst, lower for the heavier-loading ALD sample, lower still for platinum on carbon, and lowest for a platinum foil. That’s a signature of platinum that’s more electron-poor and more reactive when it’s bound atom-by-atom to the doped graphene. Moving into the extended fine structure, or EXAFS, you can literally hear the neighbors. In metallic platinum, the main platinum-platinum shell is around 2.6 angstroms; in platinum on carbon, it shifts to about 2.5; in the ALD samples, it stretches out to roughly 2.7. That shift and attenuation indicate reduced platinum-platinum coordination—fewer same-metal neighbors and more isolation. There’s also a clear low-radius feature around 1.7 angstroms that signifies platinum bound to lighter atoms like carbon or oxygen, exactly what you’d expect if single atoms and tiny clusters are anchored into a graphitic support at or near nitrogen sites. The most gratifying part of this story comes from theory because it helps explain why the pictures and spectra look the way they do. Density functional calculations quantify the stickiness of those nitrogen sites. When a single platinum atom binds at the most stable position next to a nitrogen dopant, the adsorption energy is 5.171 electron volts—a very deep well for a metal atom on a carbon sheet. Bader charge analysis shows that charge flows from platinum to the nitrogen-doped lattice, about 0.257 electrons in that favored configuration, leaving platinum slightly positive and nitrogen slightly enriched. The key thermodynamic comparison predicts whether platinum will remain isolated or cluster together. On the nitrogen-doped graphene, the energy difference favors the isolated state by 1.145 electron volts; on pristine graphene, the preference flips—clustering is favored by over a volt. Cheng and colleagues also see strong hybridization between the discrete platinum 5d orbitals and nitrogen’s 2p states near the Fermi level. That mixing is not just a bonding detail; it reshapes how platinum binds hydrogen. Hydrogen is the other half of the interaction. To form hydrogen gas, protons and electrons first must land on the surface—the Volmer step—and then combine, either through two adsorbed hydrogens meeting (the Tafel route) or by a proton from solution reacting with an adsorbed hydrogen (the Heyrovsky route). On a single platinum atom anchored to doped graphene, the first hydrogen sticks with a binding energy around 0.8 electron volts—not too strong and not too weak—while the binding of subsequent hydrogens becomes slightly softer due to repulsion and orbital crowding. When they assess the barriers for hydrogen gas formation, the numbers are impressively low: about 0.664 electron volts when two hydrogens surround the platinum site, and 0.571 when four are present. For reference, performing the same chemistry on a clean platinum one one one terrace costs around 0.85 electron volts. This molecular-level insight relays why you observe a slightly smaller Tafel slope and much greater mass activity in the experiment: the single-atom electronic structure, optimized by the nitrogen-doped support, aligns the platinum-hydrogen bond just right and creates a lower-energy pathway for hydrogen gas to escape. Take a moment to connect the pieces. ALD allows exquisite control over how much platinum you place and where it goes. Nitrogen-doped graphene supplies anchoring sites that are both chemically strong and electronically transformative. The microscope shows those atoms sitting in their seats. The X-rays indicate their d-bands are perturbed and their platinum-platinum coordination is low. The electrochemistry demonstrates they are working diligently and holding up under stress. Finally, the theory brings it all together, showing that the act of bonding platinum to a nitrogen-doped lattice pulls electron density in a way that both stabilizes the single atoms and optimizes hydrogen binding and release. It’s also insightful to examine what doesn’t work as effectively. If you perform the same ALD on undoped graphene, you see more platinum-platinum bonding in the spectra, larger particles in the images, and poorer retention of activity during cycling. This corresponds with the calculated thermodynamics: on pristine graphene, clustering is more energetically favorable. In other words, "single atom" isn’t merely about a clever deposition technique; it’s about sculpting the support so that atomic metal genuinely wants to remain isolated. There’s an underappreciated nuance in the ALD control as well. At fifty cycles, you’re in a sweet spot where the anchoring sites can accommodate many isolated atoms and occasional tiny clusters. When you push to one hundred cycles, while many single atoms persist, the statistical probability of platinum finding other platinum increases. The performance mirrors that transition: while the heavier-loading sample still surpasses commercial platinum on carbon in geometric terms, its mass activity decreases in comparison to the leaner, single-atom-rich catalyst. That difference—10.1 amperes per milligram versus 0.27 under the same potential—is not just a nice statistic. It directly reflects how many platinum atoms are engaged with the chemistry. If you’ve been following the broader movement toward single-atom catalysis, you’ll recognize the approach: isolate the active metal, engineer a support that binds it securely, and tune the local electronic structure so the key bond in the reaction—here, the platinum-hydrogen interaction—falls into the optimal range. What distinguishes this work is how seamlessly these aspects align across techniques. The scanning transmission electron microscopy doesn’t conflict with the spectroscopy, the spectroscopy corroborates the theory, and the electrochemistry corresponds to the energetics. As Cheng and colleagues demonstrated, this cohesion enables them to assert, confidently, that nearly every atom you paid for is contributing effectively. Could this extend beyond platinum and hydrogen? Likely. Atomic layer deposition is flexible with respect to the metal, and nitrogen isn’t the only heteroatom that can interface with a carbon sheet. The key point, however, isn’t to fantasize about the periodic table. It’s to recognize the framework: define the binding site in the support, employ a deposition technique that adheres to that definition, and verify the electronic fingerprint to ensure success. When those elements come together, you can develop catalysts that don’t just appear different—they perform differently, influencing everything from an atom’s charge to a lab’s current trace. And when considering costs, remember where we started. In a world where platinum’s price and availability cast a shadow over clean hydrogen solutions, increasing utilization is far from an academic exercise. It distinguishes between viable technology and one that falters. Single atoms on the appropriate support don’t just conserve metal; they redefine what that metal can accomplish. As this research underscores, sometimes the smallest unit—one atom, one seat—is the most powerful leverage you possess.