[Wannier] 回复: Query about wannier90 calculations on disordered supercell

Tue Oct 31 21:15:02 CET 2023

The experience we had with Wannierization using Gamma-point only was 
that we never found local minima if the unit cell was 
cubic/tetragonal/orthorombic. This admittedly was with the real-space 
car-parrinello code cp.x of the QE distribution.

There is an extensive discussion of MLWFs and disentanglement and 
interpolation when using Gamma only in Chap 2 (and Chap 4) of Young-Su 
Lee 2006 PhD thesis: http://dspace.mit.edu/handle/1721.1/37371

nicola

On 31/10/2023 14:08, Stepan Tsirkin wrote:
> Hi Mouyang,
> 
> 
> Tecnically it is ok to wannierise from a 1x1x1 when the unit cell is 
> big. For example, in fig 5a of this paper 
> https://iopscience.iop.org/article/10.1088/1361-648X/ab51ff/meta#cmab51fff05 an example is shown of a 12-atom carbon chain that was wannierised from Gamma-point only and the dispersion is restored with the MDRS method. So, technicall Boltzwann can do the calculation even if the wannierisation started from Gamma-point only. It will correctly interpolate the bands to any grid of k-points and take the derivatives.
> 
> 
> But if it makes sense physically - is another question. The 
> semiclassical equations of motion (underlying the Boltzwann) are defined 
> for  crystals. I guess the condition is that mean-free-path is much 
> larger then the size of the unit cell.
> 
> 
> May be you can use the Wannier model to study quantum transport (See 
> chapter 7 of the user guide 
> https://raw.githubusercontent.com/wannier-developers/wannier90/v3.1.0/doc/compiled_docs/user_guide.pdf) but I am not an expert in that.
> 
> 
> Best,
> 
> Stepan.
> 
> 
> 
> On 31.10.23 02:02, Mouyang Cheng wrote:
>> Dear Stepan,
>>
>> Thanks for the reply, and sorry for my late response. Yes, I did try 
>> to make the k-mesh 1*1*1 and perform the wannierlization at this 
>> accuracy. I think my question is whether this approach gives a 
>> reasonable tight-binding model at the DFT-level? Because I feel 
>> strange for wannier90 only fitting one k-point for each energy level 
>> to get so many localized wavefunctions.
>>
>> Another problem is that at first I tried to perform further transport 
>> simulation in Boltzwann, but Boltzwann calculation requires 
>> information on multiple k-points to do derivatives, (i.e. it needs a 
>> whole band). So I guess the alternative is to wannierlize first, then 
>> carry out tight-binding simulation on transport by other means?
>>
>> Really appreciate your advice and help!
>>
>> Best regards,
>> Mouyang
>> ------------------------------------------------------------------------
>> *发件人:* Stepan Tsirkin <stepan.tsirkin at ehu.eus>
>> *发送时间:* 2023年10月27日 20:42
>> *收件人:* Mouyang Cheng <vipandyc at mit.edu>; 
>> wannier at lists.quantum-espresso.org <wannier at lists.quantum-espresso.org>
>> *主题:* Re: [Wannier] Query about wannier90 calculations on disordered 
>> supercell
>>
>> Dear Mouyang,
>>
>>
>> As I understand, there is no fundamental obstruction to make the 
>> k-mesh 1x1x1, So you will have a Gamma-only sampling. Although I do 
>> not have experience in that. Did you try that, and did you come across 
>> some difficulties?
>>
>>
>> Best Regards,
>>
>> Stepan
>>
>>
>>
>> On 26.10.23 05:15, Mouyang Cheng wrote:
>>> Dear Wannier90 developers,
>>>
>>> This is Mouyang Cheng, a student at MIT. I'm interested in generating 
>>> wannier orbitals for disordered systems, e.g. amorphous 2D graphene 
>>> sheet (large disorder). I've read several articles successfully 
>>> coping with applying generalized Wannier orbitals on disordered 
>>> systems like:
>>> Maximally-localized Wannier functions for disordered systems: 
>>> Application to amorphous silicon - ScienceDirect 
>>> <https://www.sciencedirect.com/science/article/abs/pii/S0038109898001756>;
>>>
>>> However as to my understanding in the user manual of Wannier90, we 
>>> need to specify the Kmesh and number of bands to do wannier fit. But 
>>> for a large supercell (~200 atoms) it is only practical for DFT to 
>>> deal with only one Gamma point for BZ, and there is no concept of 
>>> band in non-crystals.
>>>
>>> *So my question is: can Wannier90 deal with such an amorphous 
>>> supercell and get a tight-binding Hamiltonian? *If not, could you 
>>> give any suggestions on any other code or convenient methods; If yes, 
>>> how does it work?
>>>
>>> Thank you so much for taking your time reading this email and I would 
>>> greatly appreciate any help or clarification.
>>>
>>> Best regards,
>>> Mouyang Cheng
>>>
>>> NSE, Massachusetts Institute of Technology
>>> <https://www.sciencedirect.com/science/article/abs/pii/S0038109898001756>
>>> 	
>>> Maximally-localized Wannier functions for disordered systems: 
>>> Application to amorphous silicon 
>>> <https://www.sciencedirect.com/science/article/abs/pii/S0038109898001756>
>>> We use the maximally-localized Wannier function method to study 
>>> bonding properties in amorphous silicon. This study represents, to 
>>> our knowledge, the …
>>> www.sciencedirect.com <http://www.sciencedirect.com>
>>>
>>>
>>>
>>> _______________________________________________
>>> Wannier mailing list
>>> Wannier at lists.quantum-espresso.org  <mailto:Wannier at lists.quantum-espresso.org>
>>> https://lists.quantum-espresso.org/mailman/listinfo/wannier  <https://lists.quantum-espresso.org/mailman/listinfo/wannier>
> 
> _______________________________________________
> Wannier mailing list
> Wannier at lists.quantum-espresso.org
> https://lists.quantum-espresso.org/mailman/listinfo/wannier

-- 
----------------------------------------------------------------------
Prof Nicola Marzari, Chair of Theory and Simulation of Materials, EPFL
Director, National Centre for Competence in Research NCCR MARVEL, SNSF
Head, Laboratory for Materials Simulations, Paul Scherrer Institut
Contact info and websites at http://theossrv1.epfl.ch/Main/Contact