Physics Fighting Black Magic of Catalysis

By Prof. Dr. Robert Schlögl, Fritz Haber Institute of the Max Planck Society Department of Inorganic Chemistry

The many breakthroughs in performance catalysis were and still are achieved by combining phenomenological knowledge with empirical experimentation. The result, the working catalyst, often comes about like magic when one looks into the details of synthesis, formulation and activation. The origin of this from a science viewpoint still unsatisfactory state of a strategic chemical science arises from the intricacy of thermodynamic and kinetic processes making a catalyst work.

Disentangling the interwoven aspects of geometric and electronic structure and separating kinetic details from the basic phenomena of adsorption, activation, reaction and desorption of the reactants from the solid surface of a catalyst requires a level of abstraction that can consistently be reached by a theoretical framework of catalysis considering both properties of materials and reactants at a similar level of detail.

The physicist Jens Norskov started his independent work by formulating a theoretical framework for the ammonia synthesis process. Inspired by the precise numerical model results of the surface science community led by Ertl and Somorjai and motivated by the careful and well-documented performance data of the Topsoe research he and his team set out on a consistent theory of ammonia synthesis linking thermodynamic data of elementary steps and electronic structure information to a prediction of kinetic high pressure performance bridging the pressure gap over 10 orders of magnitude. The success of this endeavor gave the signal to the whole community that physics-based rigorous data plus a theoretical concept casting the important energetic quantities into computational catalysis science can beat the black magic appeal of performance catalysis.

It is no surprise that Jens applied the methodology to other reactions as well as to other forms of ammonia synthesis. Soon it became clear that the level of detail that is needed to be considered made such projects so heavy that only limited advantage over conventional empirical or the then upcoming combinatorial methods was to be expected.

This prompted Jens to re-consider the proven but specific approach linking thermodynamics with kinetics of a catalytic reaction and to strip it off every detail that would not be needed to allow for generalization to many catalytic reactions with multiple reactants and products over the vast library of metalloid compounds that could be considered as catalysts. The fundamental ideas behind his approach were known and used in different contexts before. It was Jens Nørskov’s merit to combine them and make them quantitative for the specific application of predicting catalytic reactivity.
. A correlation of the heat of adsorption of the critical molecular fragment in a chemical transformation with the associated energy barrier makes large overviews of parameter spaces tractable. The unique idea is thus to combine the physics-based concepts of catalysis such that computational screening of materials and energy profiles becomes doable with sufficient accuracy and moderate resources.

Only a deep understanding of catalysis and the again from physics inspired belief into the universality of basic concepts made possible this first practically useful approach into predictive computational catalysis. Even at the relatively crude level of consideration a workable approach around the weaknesses of the DFT methodology in describing the details of chemical bonding was necessary. The Norskov approach demands similar quality of theoretical results for describing chemical bonding in solids and at their surfaces, in molecules and their interactions with a surface and in breaking and making chemical bonds between reactants. This and the necessary consideration of dispersive interactions make it difficult to apply a single DFT functional. The pragmatic solution found here was a semi-empirical density functional correction using advanced statistical concepts and machine learning tools.

The result is a broadly applicable tool for limiting the experimental parameter space in searches for a performance catalyst in a given reaction with few limits to the type of reaction. The price to be paid for stripping the details in the general approach is that it is still not possible to design theoretically the one best possible solution. Finding this is still left to experimentation where in particular structural details of the catalyst, its material dynamics and the influences of kinetic boundary conditions need to be studied. The immense value of the Norskov approach is to exclude hopeless approaches and point to potentially useful material combinations.

The single possibly most useful result from this approach is the generation of scaling relations. Their positive value is the super facile first order prediction of useful catalysts in a given reaction but also the insight that in reaction networks requiring multifunctional catalysis only compromises in the optimization of energy barriers can be made. Concepts for breaking this scaling relation require multi-functional sites that are not widely targeted as result for rational catalyst design.

The author interacted with Jens ever since his efforts on ammonia synthesis. As experimentalist working towards disclosing the complexity of performance catalyst systems it is fascinating to see the evolution of a numerical physics-based concept of catalysis that can be used to discriminate the important from the interesting. The determination of Jens to remove the “magical” part from catalysis even for the price of a lack of precision in the answer fascinated the author. The clear and sharp discussions about important aspects versus “technical details” guide the interaction with Jens. His fact-oriented but always personally friendly attitude makes interacting with him a pleasure. This is also felt by others making Jens to a leading figure in catalysis science not only in the “Danish Team” but in his world-wide network.