Density functional theory and partial charge transfer

[arrektor]

When density functional theory is used to simulate a molecule adsorbed on a surface, it turns out that due to their interaction, a fraction of an electron is transferred from the surface to the molecule or vice versa.

These interactions are normally categorised in interactions involving covalent bond formation (also termed as chemisorption), or charge-transfer complexes (also termed as physisorption).

The situation where the charge transfer is integer,i.e., the molecule is in an anionic or cationic state doesn't normally exist.

I was wondering what is the physical meaning of this and how one could determine by the nature of these interactions whether there is a chemical bond being formed or whether it is just physisorption of the molecule on the surface?

 

[cgk]

I would say that the question of whether or not a (partial) covalent bond has been formed can be decided by localizing the occupied orbitals of the Kohn-Sham determinant and analyzing the result. If the localized orbitals can be split exactly into two categories: One set of occupied orbitals only on the molecule, and one set only on the substrate, then no covalent bonding is present. However, if some localized orbitals have substantial contributions on both sides[1], then a covalent character exists. Note that since occupied orbitals have integer occupations, a non-integer total charge on the molecule does imply the presence of localized orbitals which are shared between the molecule and the substrate, and thus the presence of some (potentially small) degree of covalent bonding.

However, charge transfer is not the only mechanism by which molecules could be bound. For example, it is entirely possible that the number of electrons on the molecule is integer, that there is no covalent bonding, and that a physisorption results from either static electric interactions (e.g., molecular dipole moments resulting from the actual distribution the integer of charge across the molecule and the surface) or dynamical electric interactions (i.e., London dispersive interactions). These kinds of van der Waals interactions would typically be classified as purely physical in nature with no chemical bonding component.

[1] This interpretation depends on a definition of partial charge, of course, which is not 100% physically observable (although various reasonable definitions have been given, e.g., http://dx.doi.org/10.1002/jcc.10351 or http://dx.doi.org/10.1021/ct400687b )