[Wannier] system size limits for the transport calculation with Wannier90

Nick Papior nickpapior at gmail.com
Thu Apr 6 14:45:49 CEST 2017


Dear Farideh,

If you need to perform really large scale transport calculations I can also
refer to TBtrans (part of the SIESTA DFT suite).
TBtrans is now a "stand-alone" program capable of calculating transport of
user-defined tight-binding parameter sets (and also phonon transport).
It implements highly efficient algorithms (only sub-diagonal matrix
operations, no full matrix operations) and is furthermore capable of
calculating bond-currents among many other things.

Currently you can use sisl (https://github.com/zerothi/sisl) to read in the
tight-binding hamiltonian from Wannier90 output (*_hr.dat), and from there
write it to a readable format for TBtrans.

TBtrans and its capabilities are recently published in this paper (see Sec.
4):
http://www.sciencedirect.com/science/article/pii/S001046551630306X
which also highlights that it is capable of N>=1 electrode calculations, if
so desired. Thus using Wannier90 to generate tight-binding sets for
N-electrode transport calculations is possible.

Due to TBtrans efficiency we are periodically doing calculations of
tight-binding systems exceeding 400,000 orbitals (the capabilities are
system-shape dependent) for bond-current calculations, and >1,000,000
orbitals for transport-only calculations.
Also, TBtrans is hybrid parallelised using MPI + OpenMP.

Disclaimer, I am the developer of the latest TBtrans and sisl.

Hope this may help you if needed.


2017-04-06 12:18 GMT+02:00 Mostofi, Arash <a.mostofi at imperial.ac.uk>:

> Dear Farideh,
>
> There are several points at which limitations may apply:
>
> + The underlying electronic structure calculation: a 500 atom DFT
> simulation would need significant computational resources
>
> + Generating the Wannier function basis: for transport calculations you
> would most likely want to use the disentanglement algorithms to obtain an
> atom-centred WF basis; for a 500 atom system I expect that this would need
> some careful thought and effort to do successfully (see my earlier post
> about projections and energy windows)
>
> + Computing the quantum conductance: the matrix operations (in particular
> the matrix inversions) in the Landauer transport code are not parallelised
> yet so the amount of memory attached to your processor will determine
> whether this is possible. A 500 atom “conductor” region, assuming 4 Wannier
> functions per atom, would give you a 2,000 x 2,000 matrix to be inverted
> for example.
>
> Best wishes,
>
> Arash
>
>> Arash Mostofi — www.mostofigroup.org
> Director, CDT in Theory and Simulation of Materials
> Imperial College London
>
>
>
> On 5 Apr 2017, at 14:08, Faride Hajiheidari <hajiheidari.faride at gmail.com>
> wrote:
>
> Dear all,
>
> I would like to start the transport calculations with Wannier90. My
> systems are double-walled carbon nanotubes (DWCNTs) with the maximum 500
> atoms per unit cell.
> My question is if there is a limit size for the system that can be studied
> by Wannier90 approach in the most efficient way.
>
>
> Many thanks in advance,
> Farideh
>
>
>
> ---------------------------------
> Farideh Hajiheidari
> RWTH Aachen University
> Institute for theoretical solid state physics
>
> _______________________________________________
> Wannier mailing list
> Wannier at quantum-espresso.org
> http://mailman.qe-forge.org/cgi-bin/mailman/listinfo/wannier
>
>
>
> _______________________________________________
> Wannier mailing list
> Wannier at quantum-espresso.org
> http://mailman.qe-forge.org/cgi-bin/mailman/listinfo/wannier
>
>


-- 
Kind regards Nick
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/wannier/attachments/20170406/1c181883/attachment.html>


More information about the Wannier mailing list