How can I solve band alignment problem ?

[chorawut]

Dear All VASP user and admin
I would like to ask some question about band alignment. I want to check my idea for this calculation with VASP codes (5.2). I know that there are two way to determine the band off-set of different materials, ex. ZnO and GaN in my case:
(1) Perform GaN:ZnO interface calculation with non-polar surface to determine the difference in average electrostatic potentials deltaV (in LOCPOT not include XC) in the bulk-like regions of ZnO and GaN in ZnO:GaN surface. Then this can use to align the VBM obtained from ZnO and GaN in bulk calculation. --> offset = deltaV + Evbm(ZnO-bulk) - Evbm(GaN-bulk)
(2) Perform each slab calculation (ZnO-with vacuum and GaN-with vacuum) to determine the difference between the average electrostatic potential (AEP) in bulk-like region and that in vacuum regions for ZnO and GaN surface calculations. Then we can find the deltaV which is the difference between [AEP-Vacuum ZnO] minus with [AEP-vacuum GaN]. Then we can use this deltaV to determine band off-set: offset = [AEP-Vacuum ZnO] - [AEP-Vacuum GaN] + Evbm(ZnO-bulk) - Evbm(GaN-bulk)

Interestingly, my results obtained from these two approaches are very different. I have already tested the size of slab and vacuum.
Anyone have idea about this. Did I make some thing wrong ?
<span class='smallblacktext'>[ Edited ]</span>

[tlchan]

An interface dipole is formed between the ZnO/GaN interface, which is different from the ZnO/vacuum and GaN/vacuum interfaces. This leads to different band alignment. In principle, the interface/surface reconstruction needs to be specified to perform band alignment unambiguously.

[chorawut]

The atoms near the interface (at least 3 layers left&right) of ZnO/GaN or surface of ZnO/Vacuum were allowed to freely move until the HF forces became less than 0.05 eV/A. However, the results obtained from these two approaches are still very different. Do I need to set something else ? or How can I set in my calculations ?
PS. non polar surfaces were used in our calculations
<span class='smallblacktext'>[ Edited Mon Oct 08 2012, 06:49PM ]</span>

[chorawut]

I think I know what the problem is.