Abstract
We show how Bayesian statistics and density-functional theory can be combined to compute reliable values for the interactions in a cluster expansion for adsorbates on a surface. The method is an alternative to the leave-one-out cross-validation method. We show that it easily selects which interactions can be determined even if the total number of possible interactions is very large. We have applied the method to NO/Rh(111) based on the interactions we have determined for this system we have predicted some structures, which have been confirmed by scanning-tunneling microscopy.
Original language | English |
---|---|
Pages (from-to) | 8592-8596 |
Number of pages | 5 |
Journal | Macromolecules |
Volume | 41 |
Issue number | 22 |
DOIs | |
Publication status | Published - 2008 |