Catalysis by the Numbers

Take a snapshot of catalysis research in action, and the picture will show scientists working with reagents, benchtop reactors, gas chromatographs, and other analytical instruments. Until recently, those elements would have filled the frame. But that picture leaves out a key component of modern catalysis research: computation.

The goals of computational investigations in heterogeneous catalysis can be stated simply. "Our aim is to understand what makes a particular surface a good catalyst for a given reaction," says Jens K. Nørskov, a physics professor at Technical University of Denmark, Lyngby. "Of course, that has been the aim of catalysis research for a long time," he notes. "But I think we may finally be getting there for the simplest systems."

According to Mark A. Barteau, a professor of chemical engineering at the University of Delaware, Newark, powerful theoretical methods developed in the past decade have shown promise--not just for identifying catalytic intermediates and reaction pathways that are accessible to experiments, but for providing quantitative predictions about elementary processes that, in large part, are inaccessible to experiments.

The advances in computational techniques have armed theoreticians with detailed information about the energetics, structures, transition states, and other key features of catalytic reaction mechanisms and enabled them to formulate new concepts and develop chemical models. The models provide a basis for understanding catalytic behavior and for predicting the types of materials and conditions that catalyze reactions efficiently and selectively. The predictions, in turn, lead to the design and development of new catalysts.

The alternative to rational design is the conventional approach to discovering new catalysts--trial and error. That approach has been around for quite some time. A famous example dating back to the early 1900s comes from the work of Alwin Mittasch and coworkers at BASF in Germany. Searching for an ammonia-synthesis catalyst, the researchers conducted some 6,500 experiments involving 2,500 candidate materials and eventually discovered an effective iron-based catalyst.

NO DOUBT, the time required to carry out the multitude of experiments could have been cut back using modern high-throughput synthesis and testing methods--had they been available in Mittasch's day. But today's theoretical methods for screening would-be catalysts may have shortened the search even more markedly by zeroing in on just a small handful of promising-looking catalysts, possibly including some materials that were never studied.

That kind of thinking led Nørskov, Claus J. H. Jacobsen, and their coworkers at Danish catalyst manufacturer Haldor Topsøe to use computational techniques to search the periodic table for materials with just the right characteristics for converting nitrogen and hydrogen to ammonia. In the study, which was conducted a few years ago, the team examined properties of molybdenum, iron, ruthenium, and other metals and looked for trends in the catalytic performance of those elements.

By plotting the calculated turnover frequency (a measure of the rate of catalytic activity) as a function of nitrogen adsorption energy on the surfaces of the various metals, the team found a familiar-looking result: a so- called volcano curve. As its name implies, the curve, which is common in catalysis research, rises, plateaus, then falls. The implication is that metals that bind nitrogen too weakly or too strongly are poor catalysts. The ideal catalyst possesses an optimum binding energy that maximizes catalytic activity. This concept is referred to as Sabatier's principle, in recognition of the 1912 chemistry Nobel Laureate, Paul Sabatier.

The calculations indicate that, among the metals examined, ruthenium and osmium would be the most effective ammonia-synthesis catalysts. Third-ranked on the list is iron, which is much less expensive than the other two metals. The predictions are backed up by experimental results, the researchers point out, but the theoretical study goes a step further and suggests a way to design a novel catalyst that can outperform the pure metals.

Combining cobalt, which binds nitrogen too weakly, with molybdenum, which binds the gas too tightly, should result in an alloy catalyst with a nitrogen binding energy that corresponds roughly to the maximum of the volcano plot. "That's exactly what was found experimentally," the team notes. Guided by insights derived from the theoretical results, the Danish scientists prepared Co-Mo catalysts and found the alloys to be more active for ammonia synthesis than pure cobalt, molybdenum, ruthenium, or iron.

According to the researchers, the alloy is more active than its constituents because the properties of the mixed material fall between the properties of the individual elements. It seems counterintuitive. But the group determined through additional calculations that the interaction energy between an adsorbed molecule and a multicomponent surface can be approximated by interpolating between the interaction energies of the components. Interpolating places Co-Mo near the maximum of the volcano plot.

Similarly, Fe-Ru and Fe-Co alloy catalysts and the nitrides of Ni₂Mo₃ and Co₃Mo₃ were predicted theoretically and confirmed experimentally to outperform their components. The group contends that the finding is rather general and applies throughout the periodic table [J. Am. Chem. Soc., 123, 8404 (2001)].

In another example of perusing the periodic table computationally, Nørskov's research group teamed up with Peter Strasser, W. Henry Weinberg, and coworkers at Symyx Technologies, Santa Clara, Calif., to compare theoretical and experimental high-throughput methods for screening materials. The study focused on identifying new carbon monoxide-tolerant alloy catalysts for fuel-cell anodes.

On the experimental side, parallel synthesis and testing methods were used to prepare a collection of alloys consisting of platinum, ruthenium, and a third metal--cobalt, nickel, or tungsten--and evaluate their catalytic properties. Computational procedures were employed to calculate adsorption energies, activation barriers, and catalytic activities of the same types of alloys.

The comparative study shows that experimental and theoretical approaches to materials screening reveal similar trends in the relationship between electrocatalytic activity and alloy composition. In addition, both types of investigation point to similar three-component alloys as candidates for improved fuel-cell anode catalysts [J. Phys. Chem. B, 107, 11013 (2003)].

Each approach has strengths. For example, computation can be used to analyze thermodynamic and kinetic parameters that are inaccessible to experiment or cannot be measured easily. And high-throughput laboratory measurements provide information about the properties of test specimens under real experimental conditions. Nørskov and coworkers emphasize that the screening approaches are complementary and that the real power of high-throughput techniques can be tapped by combining theoretical and experimental screening methods.

ANOTHER TYPE of reaction being scrutinized by computational chemistry methods is ethylene epoxidation. The University of Delaware's Barteau studies the process using theoretical and experimental techniques to elucidate details of the reaction mechanism and design catalysts that are more selective than the silver-based materials used commercially.

A couple of years ago, Barteau's research group identified oxametallacycles--species with an -O-C-C- structure attached at both ends to a metal surface--as key intermediates in the epoxidation reaction. The species were identified on the basis of surface science measurements combined with quantum mechanical calculations.

Barteau and coworkers Suljo Linic and Jerome T. Jankowiak determined that oxametallacycles can go on to form the desired product, ethylene oxide, or the intermediates can form acetaldehyde, which leads to unwanted combustion products. Unfortunately, the two processes have similar activation barriers, which compromises selectivity.

So the Delaware researchers employed computational screening methods to search for a catalyst that would boost selectivity in the epoxidation reaction relative to traditional single-metal silver catalysts. They focused on identifying bimetallic alloys of silver that accentuate the difference in activation barriers for the desired and undesired reactions.

The theoretical results indicate that a bimetallic surface with roughly 25% copper atoms should do the trick. Indeed, the predictions were verified experimentally in tests of alumina-supported Cu-Ag catalysts. The group reports that some of the bimetallic catalysts are 1.5 times more selective than pure silver catalysts under identical conditions [J. Catal., 224, 489 (2004)]. Barteau points out that the goal of previous rational-design investigations was to improve catalyst activity or stability. Now another key catalysis attribute--selectivity--has been enhanced through computational methods.

In related work, the Delaware group turned to computational analysis to examine the mechanistic role of cesium, which is used as a catalyst additive to promote ethylene epoxidation. This promoter has been included in commercial processes for years, yet details of its function aren't well understood.

The theoretical work zeroed in on the energies of the adsorbed oxametallacycle intermediate and the pair of transition-state structures that lead to ethylene oxide and the combustion by-products. The calculations reveal that in its role as a promoter, cesium stabilizes the transition state that leads to ethylene oxide relative to the competing pathway [J. Am. Chem. Soc., 126, 8086 (2004)].

Alloy catalysts also figure into theoretical work being conducted at the University of Wisconsin, Madison. Manos Mavrikakis, an assistant professor in the department of chemical and biological engineering, applies computational tools to investigate catalytic properties of transition metals in an effort to analyze complex reaction pathways occurring on their surfaces. Also, like other practitioners of quantum chemistry, he uses theoretical methods to discover trends that may aid in developing new catalysts.

In a recent study, Mavrikakis and graduate student Jeff Greeley investigated hydrogen chemistry on a large collection of near-surface alloys. These kinds of alloys are solid mixtures in which the concentration of a solute metal near the surface of a host metal differs from its bulk concentration. The Madison team focused on two types of alloys--those in which a solute is present just below the host surface and others in which the solute resides as a thin layer on top of the host metal.

Typically, metals that weakly bind atomic hydrogen to their surfaces aren't very effective in dissociating molecular hydrogen. In other words, a low atomic-hydrogen binding energy generally goes hand in hand with a large barrier to dihydrogen dissociation. The Madison study, however, turned up surprising results. "We find that some near-surface alloys offer an exciting exception to this rule," Mavrikakis explains.

The team's calculations show that some subsurface alloys--for example, alloys consisting of platinum or palladium hosts and tantalum, tungsten, or vanadium solutes--bind hydrogen even more weakly than do gold and copper. Yet those alloys are predicted to dissociate dihydrogen much more readily than gold and copper [Nat. Mater., 3, 810 (2004)]. The group points out that materials possessing this unusual combination of properties may be useful for selective low-temperature production of chiral pharmaceuticals and as catalysts in hydrogen storage applications.

UNLIKE HYDROGEN, oxygen exhibits more conventional behavior. Mavrikakis and coworkers Ye Xu and Andrei V. Ruban applied quantum mechanical methods to study adsorption and dissociation of oxygen on platinum-iron and platinum-cobalt alloys and related materials. The simple reaction plays a central role in catalytic oxidation and ot her processes.

The study provides basic information about alloys that may serve as less expensive substitutes for platinum catalysts. It also reveals an important trend in oxygen surface chemistry: The more weakly a material binds atomic oxygen, the less effective it will be in dissociating dioxygen. The upshot is that oxygen dissociation kinetics on alloy catalysts can be estimated from atomic binding energies, which are much simpler to calc ulate than transition-state properties [J. Am. Chem. Soc., 126, 4717 (2004)].

Mavrikakis remarks that the trend seems to be fairly general and appears to extend beyond platinum and its alloys. But an exception such as the one found for hydrogen cannot be excluded, he says. "We are learning equally by discovering trends and by finding outliers to those trends."

A different sort of trend, one that highlights progress in computational capabilities, is the transition toward investigating chemical systems of increasing complexity. Rutger A. van Santen, who is university rector and a professor in the Schuit Institute of Catalysis at Eindhoven University of Technology, in the Netherlands, notes that there has been a shift in focus among some theoreticians from gas-solid interfaces, which have been studied for decades, to the interfaces between liquids and solids. Interactions between liquids and solids play a central role in electrocatalysis and other processes.

Catalytic reactions occurring on electrode surfaces are often monitored using voltammetric techniques. Several factors influence the shape of voltammograms, but it's difficult to tease apart the causes and effects that give rise to the complex shape of the curves. So van Santen, graduate student Chrétien G. M. Hermse, and their coworkers use Monte Carlo methods to simulate voltammograms in model systems.

TO DEMONSTRATE the procedure, the Eindhoven researchers simulated voltammograms of anion adsorption on single-crystal electrodes with a hexagonal surface structure. In agreement with experimentally measured data from sulfate ion solutions, the simulated voltammograms show two major features. As the voltage is scanned from slightly negative to slightly positive, a broad peak is observed. The calculations show that the broad feature arises from adsorption of anions on the electrode in a disordered phase. A much sharper peak appears at slightly higher voltage. The group notes that this feature corresponds to a change in the structure of the anion layer from disordered to ordered at a surface coverage of roughly 0.2 monolayers [Surf. Sci., 572, 247 (2004)].

The researchers acknowledge that the exact nature of the layer of anions and water molecules adsorbed on the electrode surface is more complex than the simple model used in the simulations. They suggest that more detailed information can be deduced by combining surface spectroscopy with additional calculations.

Meanwhile, at the University of Virginia, Charlottesville, Matthew Neurock, a professor of chemical engineering and chemistry, also applies computational methods to investigate complex catalytic reaction environments. For example, Neurock and coworkers studied vinyl acetate synthesis on palladium surfaces. The reaction involves a large number of elementary steps and competing pathways that lead to unwanted by-products.

The calculations reveal a wide range of mechanistic information. For example, activation of ethylene to form surface vinyl species requires access to a surface site with a minimum of four palladium atoms available for bonding. And the transition state for the coupling reaction between vinyl and acetate species requires a six-atom ensemble. By feeding quantum mechanically derived reaction energies and rate constants into a Monte Carlo simulation, Neurock was able to model the reaction and predict the effect that adjusting partial pressures, temperatures, and other reaction conditions would have on reaction yield and selectivity [J. Catal., 216, 73 (2003)]. In addition, he showed that certain arrangements of gold and palladium atoms could improve catalytic activity and selectivity, thereby providing a path toward atomic-scale catalyst design.

Another complex reaction environment that has given up some of its secrets to theoretical probing by Neurock and coworkers is the anode of bimetallic fuel cells. Platinum is often alloyed with another metal such as ruthenium to control the adverse effect of CO poisoning by oxidizing the contaminant to CO₂. Bimetallic catalysts can be very effective, but researchers are uncertain of the oxidation mechanism.

Neurock and graduate student Sanket Desai found that the ruthenium component of the alloy, together with the presence of liquid water at the anode, provide favorable conditions for formation of surface hydroxyl species. The team's results also show that subsequent oxidation of CO by hydroxyl species on the alloy surface is energetically favorable [Electrochim. Acta, 48, 3759 (2003)].

Earlier this year, James A. Dumesic, a professor of chemical engineering at the University of Wisconsin, Madison, reported that gold nanotubes can mediate room-temperature oxidation of carbon monoxide with oxygen. Dumesic also observed that the reaction is enhanced in the presence of water [Angew. Chem. Int. Ed., 43, 1140 (2004)].

To gain insight into the oxidation mechanism, Neurock studied the gold nanotube system using theoretical methods. He finds that in the presence of water, it is energetically feasible for CO to be oxidized by hydroxyl intermediates or by water via a heterolytic path that forms carbon dioxide, protons, and electrons.

The picture of catalysis research continues to change as new tools and techniques become available to scientists. Whereas in decades past, basing a catalysis research program entirely on computation was unimaginable, nowadays excluding theory entirely is equally unimaginable.

"Are we on the verge of catalysis by design?" Barteau asks. If "by design" means starting strictly from quantum chemistry, then catalysis is not there yet, he says. But with continued advances in computational methods and the benefit of ever-improving experimental techniques, the field continues to move closer to that ideal.