Hello everyone,

I am trying to optimise the length of the C-O bond in carbon monoxide using CASTEP. CO is placed in a supercell with an edge length of 20 Ang. The calculation uses the PBE functional, ultrasoft pseudopotentials and a large cutoff (380 eV). Unfortunately, the bond is quite a bit longer than the experimental value, with 1.154 Ang (CASTEP-PBE) vs. 1.128 Ang (experimental). According to the CCCBDB, the PBE functional should deliver a bond length that is closer to experiment when reasonably sized all-electron basis sets are used. On the other hand, calculations with pseudopotentials reported in the CCCBDB appear to inevitably overestimate the bond length.

Furthermore, I found the following statement regarding calculations on small molecules (not CO, however) in a paper by the CASTEP developers (Milman et al., J Mol Struct THEOCHEM 2010, 954, 22): "We found that the use of different pseudopotentials may change the equilibrium bond length by up to 0.05 Å (...)".

Does anybody have a hint if the prediction of the bond length could be improved by changing the setup of the calculation, or do I have to content myself with this deviation?

Thank you and best regards,

Michael