New design paradigm for heterogeneous catalysts

Both scientific discovery and technological development are at some point faced with the question of how to progress from a trial-and-error approach to a highly controlled design process. In heterogeneous catalysis, the search for the optimal active site of a catalyst for a given chemical reaction has been the central objective of research for almost a century. In 1925, Taylor put forward the idea that on a solid catalyst ‘there will be all extremes between the case in which all the atoms in the surface are active and that in which relatively few are so active’ [1]. Ever since the formulation of the Taylor concept of active sites, the quest for observing, identifying, modifying, and designing active sites of heterogeneous catalysts has been on. Heterogeneous catalysis involves an extremely complex set of phenomena, and in order to develop catalyst design strategies, an identification of key parameters, that are principally responsible for the catalytic rate and selectivity (lumped together as ‘activity’ in the following) is needed. A simple approach in this direction has been developed recently for transition-metal surface catalysis [2,3]. The central concept is known as energy scaling relations [4], which together with activity maps and the d-band model have made it possible to develop a quantitative understanding of trends in transition-metal catalysis and enabled prediction of new catalysts [3,5]. These scaling relations are correlations between surface bond energies of different adsorbed species including transition states. Correlations between activation energies and reaction energies, Brønsted–Evans–Polanyi relations, are known throughout chemistry [6,7] and have for long been assumed in heterogeneous catalysis [8]. With the advance of computational methods, it has been discovered that scaling relations are much more general and for a number of reactions over transition-metal surfaces they have been shown to include essentially all intermediates and transition states. The scaling relations enable a mapping of the many energetic parameters determining the rate of a catalytic reaction onto a reduced phase space spanned by a few energy parameters known as descriptors [2,3]. As a result, a catalytic activity map can be constructed defining the rate and/or selectivity in the relevant descriptor space i.e. the relevant bond strength between one or more of the reactants or intermediates and the catalyst surface. The activity map exhibits at least one maximum for the optimal bond strength(s), which defines the optimum catalyst. This approach is illustrated for a simple model of the ammonia synthesis process in Fig. 1a. Following [9], the rate of ammonia synthesis at industrial conditions is calculated in a model that assumes N2 dissociation to be rate limiting and that adsorbed N atoms are the main intermediate covering the surface. In such a model, the rate can be calculated as a function of the transition-state energy for N2 dissociation, EN-N, and the N adsorption energy, EN. The two energy parameters are seen to scale very well, and hence a single parameter, EN, is a good descriptor of the catalytic activity (see Fig. 1b). This is a very simple example of complexity reduction from two to one descriptor. The quantitative aspect of the descriptor approach provides new possibilities in catalyst design. This is also illustrated in Fig. 1b. Knowing which descriptor value defines the highest activity allows for searches for new catalysts with close to optimum properties. While the concept of scaling relations has proven extremely useful by providing an understanding of trends as well as new catalyst design criteria, it has also helped us identify some of the limitations on the performance of large classes of catalysts [2]. Fig. 1a illustrates the point. Clearly, a much better catalyst for ammonia synthesis could be devised if we could find effective ways of circumventing the scaling relations that we know for (stepped) transition-metal surfaces, or, equivalently, find catalysts with active site motifs that obey a different lower lying scaling relation than the so far identified ones. There are several other cases where scaling relations have been suggested to impose limitations on the performance of catalysts. Fig. 1c and d show two such examples: it has proven very difficult to find electrocatalysts that bring down the overpotential for O2 reduction significantly below that for Pt, making low-temperature fuel cells less efficient than we would like, and similarly, it has proven difficult, so far to find electrocatalysts that can reduce CO2 to form hydrocarbons and alcohols without substantial overpotentials. In both cases, this can be traced back to scaling relations limiting our ability to properly optimize catalysts. (a) Ammonia synthesis rate as a function of nitrogen adsorption energy and N2 dissociation barrier with energetics for stepped transition-metal sites and the corresponding scaling relation line (dashed black–white) showing that EN-N is a linear function of EN, which prevents the metal steps to be in the optimum catalyst region (red area). The alkali-promoted scaling relation is given by the magenta line. (b) Calculated rate as a function of a single descriptor of EN. A data point of an intermetallic compound with active sites consisting of both Co and Mo is indicated. Experiments show such a material to activity close to that of Ru. [10] (c) and (d) calculated limiting potential (the potential where an overall electrocatalytic reaction becomes endergonic) for the O2 and CO2 reduction reactions. The subfigures (a and b), (c), and (d) are based on data adapted from [2,11,12], respectively. Hence, a new design paradigm is needed that focuses on overcoming or circumventing the energy scaling relations. To make major breakthroughs, it may be necessary to find ways to stabilize one adsorbed state or transition state without stabilizing others. It is the purpose of the present perspective to discuss a broad palette of general strategies how this might be achieved. First of all, we should semantically define what we mean by circumventing the scaling relations. Let us point out that it is not clear whether it is possible to break scaling relations as they were defined in [4]. It may be that every imaginable active site structure will show scaling relations between energies of different intermediates, one for each structure. However, what is clear at this point is that we have to find materials that do not follow the best scaling relations that we have found so far. We therefore suggest that the term ‘circumventing scaling relations’ is used to refer to any departure from the known scaling relations. Fig. 2 is an attempt to systematize the different approaches that one can imagine taking in order to circumvent scaling relations. This 2D scheme is a simplified version of a vast parameter space. The ordinate of Fig. 2 represents intrinsic versus extrinsic effects, while the abscissa spans over electronic versus structural effects. Schematics of approaches that might be used as possible routes to circumvent energy scaling relations. The routes are positioned in a space spanned by electronic to structural effects and intrinsic to extrinsic effects. We have chosen this representation as it is inclusive and illustrates some of the major directions available in heterogeneous catalysis for tuning activity. We distinguish between intrinsic properties i.e. material-specific properties that are independent of the surroundings, and extrinsic properties that are defined by the relationship of the catalyst with its environment. The intrinsic properties can be tuned, for example, by chemical composition or structure of the material. The extrinsic properties require another material or geometrical structuring, for example, an interface of some kind either with another solid material, a liquid, or a gas, to influence the host catalyst. Each of the effects considered in Fig. 2 can be characterized by yet another dimension as either electronic or structural. Attempts so far have found it challenging to circumvent scaling relations when different reaction intermediates (including transition states) adsorb on the same surface site. Hence, multisite functionalization, that is, catalysts with several types of active sites or local binding environments for different intermediates is a promising possibility. This might be achieved in a number of ways, for example, by alloying, doping, introduction of defects, coupling between catalyst and support, nanostructuring, confinement, subsurface species, etc. For the multisite functionalization approach to be suited for solving the critical problem of circumventing scaling relations, it might require that the adsorbates are large and flexible enough to allow for multiple site interaction. It has been shown that the same material can have different scaling relations depending on the surface considered [4]. For example, the (111), (100), and stepped (211) metal surfaces obey different adsorbate and transition-state scaling relations. One can imagine more structured surfaces, channels, pores, cavities, or nanoparticles or rods that will provide active site motifs beyond the 2D structure by a surface. An additional approach that provides a 3D active site includes co-adsorbing or tethering molecules. This can create local binding environments in analogy to ligand chemistry in homogeneous or enzyme systems. Co-adsorbed species on the catalyst might be species different from the ones involved in the reaction itself; for example, other tethered molecules attached to the catalyst that can interact geometrically and/or chemically with intermediates. In electrochemical processes, the effect of the electrolyte or additives to the electrolyte can strongly change the surface chemistry by solvating different intermediates differently, thus providing the 3D effect discussed above. We would like to point out that in many catalysts that have been refined in industry over decades, one or more of the strategies outlined above may be in play. One such example is the enhanced ammonia synthesis rate achieved by addition of alkali metals as promoters [13] Adding alkali(oxides) to the surface of a transition-metal induces local electrical fields. This allows one to exploit the fact that the N2 dissociation transition state has a larger dipole moment than adsorbed N. Therefore, the scaling relation line in Fig. 1a is shifted downwards resulting in a circumvention of the pure-metal scaling relation and a considerably better catalyst. These industrial advancements can be contrasted to the naturally occurring systems as enzymes including nitrogenase. Despite that this biological system is known to be able to make ammonia at ambient temperatures and pressure, it is a very inefficient process. Hence, not only do we need to find inorganic catalyst but they have to be scalable and more efficient than the naturally occurring process. The challenge remains to develop a systematic approach to understanding effects like this in order to give us a toolbox of strategies to design radically better catalysts. Support from the DOE Office of Basic Energy Science to the SUNCAT Center for Interface Science and Catalysis and the SLAC National Accelerator Laboratory LDRD program are gratefully acknowledged.