[CP2K-user] Incorrect spin moment for P donor in Si

[Nicholas Winner]

HSE should give alright levels for P in Si. I have a test calculation I did 
a while ago with Si:P and found the following:

Dopant energy levels (distance from band edge) for Boron, Phosphorus 
Exp:      0.045eV, 0.045eV
PBE0:   0.21eV, 0.28eV
HSE06: 0.12eV 0.19eV

While by experimental standards those energy levels are pretty far off, 
it's pretty darn good by DFT with total energies. 

Using Janak's theorem instead of total energies for the transition levels:
PBE0:   0.104 eV, 0.176eV
HSE06: 0.017 eV, 0.077eV

Which is even better. HSE will help give pretty good localization, so that 
might not be the issue you are facing. You should try some delocalization 
analysis to see if its actually the problem. You can do this with, say, the 
planar averaged electrostatic correction scheme of Freysoldt. You can also 
look at the localization by investigating the cube file for the electron 
density in a program like vesta. The hyperfine constant may require even 
better theory like a GW calculation, but I'm not an expert on that aspect 
of point defect calculations.
On Monday, April 19, 2021 at 10:37:47 PM UTC-7 ma... at gmail.com wrote:

> Hi Nick,
>
> Thank you for your response. Yes, P is a very shallow donor. I calculated 
> this system using VASP at first but it couldn't give good hyperfine 
> constant either which I thought maybe due to the delocalization of 
> pesudopotential used in VASP. So I tried cp2k because I though Gassusian 
> type basis set maybe more localized. I tried HSE too but it couldn't give 
> good results either. Could you please provide some suggestions on improving 
> this issue?
>
> Regards,
> Hongyang
>
> 在2021年4月20日星期二 UTC+10 上午3:31:14<n... at berkeley.edu> 写道：
>
>> The main thing is because you are trying to simulate a shallow donor with 
>> GGA. If you look at The HOMO-LUMO gap in your output file, one of the spin 
>> channels has a gap of 0.6eV, which is the expected band gap of Si using GGA 
>> approximation. The other spin channel, where the extra P electron has gone, 
>> however, has a negative HOMO-LUMO gap. This is because the P state has 
>> delocalized into the conduction band rather than remaining in the gap. DFT 
>> with GGA will have issues of delocalizing defect states, and they will 
>> become more and more pronounced the shallower that defect state is, since 
>> the defect state will couple to the band edge. P is an extremely shallow 
>> state in Si (~0.045eV below the CB in experiment), and so it's very hard to 
>> remedy this situation. Your best bet is to explore hybrid DFT, which opens 
>> the band gap and provides more localized states, but even that will have 
>> trouble for such a shallow state and so you'll have to check it as you go.
>>
>> -Nick
>> On Sunday, April 18, 2021 at 11:23:23 PM UTC-7 ma... at gmail.com wrote:
>>
>>> Dear cp2k developers/users,
>>>
>>> I'm curretly using cp2k 7.1 to calculate the P doped Si system (1P in 
>>> 512 atoms). Theoretically, since the electronic configuration of P is s2p3 
>>> and that of Si is s2p2. A P doping in Si should result in a total spin of 1 
>>> due to the unpaired electron introduced by the P atom. 
>>> In my calculation, the number of total spin is correct, but the spin 
>>> moment of each atoms seems to be wrong because the spin moment of P should 
>>> be 1 (because of the unpaired electron in P) and that of Si should be 0. 
>>> Could someone please provide some help in solving this issue? Thanks.
>>>
>>> Thanks & Regards,
>>> Hongyang
>>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20210421/c393619b/attachment.htm>