Ghost atom in NEGF surface simulation with regarding surface reconstructions?

[ZalmanChu]

Hi, I am facing an issue when dealing with NEGF surface simulations. There are many surface reconstructions for semiconductors, especially for III-V, how to determine the ghost atoms?

For example, GaN 0001 surface has a stable configuration that a Ga atom absorbed at the ideal  surface. If we followed the tutorial, the outmost Ga atom should be set as ghost atom, however, this seems to be completely unphysical, as the the adatom is part of surface reconstruction. 

Thank you for the response! 

[Petr Khomyakov]

When using an LCAO basis set, deciding on ghost atom positions is a pretty trick thing for studying open structures such as semiconductor and related surface. You may first try not using any ghost atoms at all, unless you are interested in work function calculation.

For accurate  calculations, you may consider using the plane-wave approach, also implemented in QuantumATK. In this case, you should use the slab approach with the appropriate number of atomic layers.

[ZalmanChu]

Thank you. Slab model usually presents strong quantum confinement effect, so we prefer NEGF method.
 I still have 2 questions.
1. Does only workfunction depend on ghost atom?  How about electronic structure (surface states, band structure, etc)?
2. Supposing the surface is a heterostrcuture, is band alignment, which usually correlates to the electrostatic potential, at interfacial region sensitive to the ghost atom?

[Petr Khomyakov]

1. Does only workfunction depend on ghost atom?  How about electronic structure (surface states, band structure, etc)?

I suppose the electronic structure may also depend to some extend on how tails of wave functions are described in the vacuum region above the surface. It is hard to say how much, as one should investigate it for a particular system. Before doing LCAO-NEGF, you could still do benchmarks for slabs using PW vs. LCAO to see whether LCAO is adequate for your particular system.  Note that you may try different basis sets such as Medium, High, Ultra if you are using SG15 or PseudoDojo.

2. Supposing the surface is a heterostrcuture, is band alignment, which usually correlates to the electrostatic potential, at interfacial region sensitive to the ghost atom?

In this case, you do not need to introduce ghost atoms, since atoms at the interfaces are chemically bonded, i.e., there is no vacuum.