October 22, 2013

Global optimization of surface structures

Combinations of constraints.

As catalytic systems under study become more complex, the number of possible surface configurations can become enormous. For example, if we would like to determine the number of ways that just two CO molecules can bind to a simple 3×3 stepped (211) surface, we would manually have to try about 230 configurations to 'exhaustively' search the possibilities, even after removing symmetrically identically configurations. Clearly, systems can become much more complex as coverages increase and adsorbates take on configurational degrees of freedom -- the possible configurations can easily number in the thousands. Trying to enumerate these is a daunting task for a researcher.

To make this problem tractable, we have adapted the minima hopping technique first developed by Stefan Goedecker to surface science configurations. This technique uses a combination of established atomistic techniques -- molecular dynamics and local optimization -- in tandem with self adjusting search parameters. In order for such a technique to work in finding adsorbate configurations, we have developed constraints (penalty functions) based on Hooke's law that force the adsorbates to stay intact, while having no effect on the energetics when the molecular fragments are in reasonable configurations.

We introduced the modified technique, which we call 'constrained minima hopping', in a paper published this month and showed how it successfully found the global optima configurations of a suite of adsorbates that had previously been optimized in a 'brute intuition' approach, finding the global optimum configuration in a computationally efficient manner.

The work was published in a special issue of Topics in Catalysis in honor of the 60th birthday of Jens Nørskov. Read the article. We have implemented the techniques [minima hopping, Hookean constraint] in the open-source Atomic Simulation Environment (ASE). You can find a tutorial here.