A computational approach and NCSA resources help MIT researchers better understand the process of manufacturing hydrogen.

How do you harvest hydrogen? The abundant proton-neutron-electron combo accounts for about three-quarters of the elemental matter in the universe, but collecting a sufficient supply for industrial uses—from the mundanity of fertilizer production to the potential for hydrogen fuel cells to reduce our carbon footprint—requires a manufacturing process.

A common method of manufacturing hydrogen is steam methane reforming (SMR), which combines methane, steam, and a metal catalyst (such as nickel) in a high-temperature environment. The process has been used for many years and has been studied in detail, but there are still gaps in scientists' understanding.

"The problem for all heterogeneous catalysis is something called the 'pressure gap'," explains Wayne Blaylock, a chemical engineering PhD student at MIT. "Experimental observation of these processes is limited largely to instruments that operate at a very low pressure. And industrially, you can't make any money operating at that low pressure, so you will operate at much higher temperature and pressure."

That means that the experimental picture doesn't match how the process occurs in the real world. That's why the research team led by Blaylock's advisor, MIT chemistry professor William Green, adopts an ab initio computational approach to model the thermochemistry of SMR under industrial conditions. The goal is to obtain a better understanding of the surface pathways and reactions that control SMR in order to design new catalysts and to control the by-products of the reaction (such as catalyst-clogging carbon).

Their analysis of the key Ni(111) catalyst surface was presented at the spring 2008 meeting of the American Chemical Society.

Unexpected pathways

Using density functional theory (DFT) and statistical thermodynamics, the researchers have developed a detailed model of the intermediate species and transition state behavior occurring in SMR over Ni(111), the most predominant and thermodynamically stable phase of the nickel catalyst. Their calculations include the high temperature (800 degrees Celsius) and pressure (10 bar) that are realistic for hydrogen manufacturing, and they investigated 36 reactions involving 26 species.

The team performed flux and sensitivity analysis on their kinetic model to determine which reaction pathways are dominant and which steps control the rate of the reaction on the Ni(111) surface.

Their work relied on the Tungsten and Abe, Dell clusters at NCSA, in addition to a small computer system in their lab at MIT.

"We're just now getting to the point where we can do this type of modeling; in the past five years or 10 years, the hardware has become more capable of handling this sort of computation," Blaylock says. "Because we want to look at all these different pathways, because we don't want to make assumptions about what is important, we have to look at many species, many reactions on many catalyst surfaces. We looked at every feasible combination of one water and one methane molecule on the surface. We didn't just look at one pathway, we looked at many reaction pathways that could lead to the desired products.

"And that's what drives our reliance on the supercomputing resources."

The researchers found that the steps limiting the rate of the overall reaction were the reactions of CH with O and OH to form CHO and CHOH; the dominant pathway was through CHO.

"Most people would assume that the dominant pathway on the surface would be through a carbon and an oxygen, forming CO," Blaylock says. "But what we found through our microkinetic modeling is that there were other pathways that were more favorable. "It turns out that the C species is very unstable on the Ni(111) surface, and that instability makes reactions involving it kinetically unfavorable.

"Really the only way of realizing just how unstable this species is was computation," Blaylock says. "The important message to take away is that this sort of comprehensive modeling can provide insights that might be contrary to popular assumption or popular belief.

"Computation can really bring the surface into view, where before you simply couldn't see it using available experimental techniques."

Of course, experimental techniques are growing more accurate as well, providing the potential to validate the computational model with experiment and to guide experiment with computational insights.

"What that really means," Blaylock says, "is pairing experiment and computation and driving toward a unified microkinetic model. The amount of information we are able to obtain will just continue to increase from experimentation and computation in connection."

Catalysts and carbon

Ni(111) is just one aspect of the nickel catalyst; there are other nickel step sites that the team plans to model as well, determining their surface energies and reaction barriers and constructing a model that combines all of the nickel facets.

"We've only performed comprehensive work for the Ni(111) surface, but we hope to do a more unified model of various catalyst facets going forward," Blaylock says. "This will give us a better idea of what a real catalyst might do under real conditions."

Determining which steps limit the rate of the reaction provides a target for catalyst development. "If you know the bottleneck, you can design a new catalyst to address that," Blaylock explains.

The researchers are investigating nickel alloys, such as nickel doped with a small amount of silver. So far, a catalyst in which one of every four surface atoms is silver seems to slow the SMR process by an order of magnitude. That doesn't seem like a desirable result, but if a slower reaction results in less carbon formation, that trade-off might be worthwhile.

That's because carbon formation is the bane of hydrogen production. Carbon can build up, encapsulating the catalyst and preventing it from reacting with the methane gas. That's usually just a temporary problem that can be solved by stopping the flow of methane, pumping in oxygen, and letting the carbon burn away in the already high heat required for SMR. But a more serious form of carbon buildup, called whisker formation, can actually crush the catalyst. In the experimental membrane reformers that might someday supply pure hydrogen for fuel cells, this could be disastrously expensive. Finding reactions that reduce carbon creation is therefore another important target for the researchers.

Team members
Greg Beran
Wayne Blaylock
William Green, principal investigator
Teppei Ogura