Empty orbitals in Transiesta

Question

I have noticed some difficulties in converging Transiesta calculations when including empty orbitals in the basis set, particularly when a non-zero bias is applied. In particular, I have tried to simply converge graphene structure and I could easily converge the calculation by excluding the empty orbitals. Including them, I could still converge zero-bias case but not the non-zero cases. Since I used quite long tails, I extended the divide region to approx. 6/7 layers. I understand that in the case of graphene, the energy window of interest is well described even by excluding these orbitals. However, I will include some atoms in which I included empty orbitals to properly describe their pristine band structure (in particular, B and N and I reproduced hBN band structure). My question is: is it advised not to include empty orbitals in transiesta calculations?

Answer 1

There is per see nothing to prohibit the use of empty orbitals in TranSiesta calculations. In fact, in some cases they are very much required (vacuum spaces between STM tip and sample for instance).

However, there are a few details that require extra care.

1. Adding empty orbitals in the scattering region, but not in the electrodes, may require an extended region of electrode layers between the actual electrode and the device region.
2. The placement of empty orbitals can greatly affect the self-consistency depending on their relative position and how close they are to other orbitals.

So the answer is no to your question. But extra care should be taken when using empty orbitals in the device region and not in the electrode region.