[QE-users] problem with convergence of MnTe altermagnet hexagonal phase

[Nicola Marzari]

Thanks Alpin,

indeed, the challenge in achieving convergence in iterative methods 
(like those used in QE, VASP, Abinit, etc...) can sometimes be 
(randomly) improved by changing some of these technical ingredients.

The problem to me is that iterative methods are very efficient, but when 
they fail to converge it's very difficult to find a cure - and many 
interesting materials, especially magnetic or mixed valence, are 
difficult to converge (even most of the times). Then, one should go back 
to variational methods, that are bound to converge. We found the 
solution to this problem 30 years ago:

https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.79.1337

(also republished identically in
https://journals.aps.org/prb/abstract/10.1103/PhysRevB.79.241103)

Only Castep and SIRIUS implemented this, but Claude now makes it easy to 
implement it also in QE, so hopefully we'll see an end soon to this 
convergency challenge.

			nicola


On 06/07/2026 12:14, Alpin Novianus Tatan wrote:
> Dear Prof. Marzari,
> 
> Thank you for your comments. I agree that if convergence is achieved and 
> the same physical approximation (e.g., PBE) is used, calculations 
> employing ultrasoft, PAW, or norm-conserving pseudopotentials should 
> yield the same physical results.
> 
> My previous comment was intended to address a different issue, namely 
> the practical difficulty of achieving convergence, which Ireneusz 
> encountered. In practice, the ease of obtaining convergence can differ 
> between pseudopotential types and computational implementations. For 
> that reason, I suggested trying different pseudopotential types rather 
> than restricting the calculations to PAW, since this may improve 
> convergence without changing the underlying physical approximation.
> 
> Best regards,
> 
> Alpin N. Tatan
> Department of Physics
> The University of Tokyo
> 
> 
> On Mon, Jul 6, 2026 at 5:37 PM Nicola Marzari <nicola.marzari at epfl.ch 
> <mailto:nicola.marzari at epfl.ch>> wrote:
> 
> 
> 
>     Dear all,
> 
> 
>     just a note on these comments below. All codes (all electron,
>     pseudopotential PAW, ultrasoft, normconserving) should give the same
>     numerical result for a given physical approximation (e.g., PBEsol).
> 
>     If they do not, there is a problem - this is why verification (in
>     "verification and validation") is very important - making sure codes
>     give a numerically exact answer to a given physical theory (the theory
>     needs to be "validate", i.e. good enough to describe the physics at
>     hand
>     - you cannot do DFT PBEsol to study say a Hubbard sideband.
> 
>     There has been a lot of progress on "verification", as you can see in
>     https://www.nature.com/articles/s42254-023-00655-3 <https://
>     www.nature.com/articles/s42254-023-00655-3> , with all-electron
>     and pseudopotential codes giving very close restuls.
> 
>     For QE, there are curated efforts here:
>     https://legacy.materialscloud.org/discover/sssp/ <https://
>     legacy.materialscloud.org/discover/sssp/> - so that's a good
>     start, and you can find PBE and PBEsol pseudos tuned for efficiency
>     (lower cost) or accuracy (higher cost); we'll release very soon a major
>     update with all the data from the paper above.
> 
>     nicola
> 
> 
>      >     Authors of the paper used VASP but it should be possible in
>     Quantum
>      >     Espresso.
>      >
>      >
>      > This is not always the case, do not assume this.
>      > Pseudopotentials are an important part of the calculation, which are
>      > different sets for VASP and QE.
>      > The calculation and the code implementation are also different.
>      >
>     -- 
>     ----------------------------------------------------------------------
>     Prof Nicola Marzari, Chair of Theory and Simulation of Materials, EPFL
>     Director, National Centre for Competence in Research NCCR MARVEL, SNSF
>     Laboratory Head, PSI Center for Scientific Computing, Theory, and Data
>     Contact info and websites: https://theos-wiki.epfl.ch/en/Main/
>     Contact <https://theos-wiki.epfl.ch/en/Main/Contact>
> 


-- 
----------------------------------------------------------------------
Prof Nicola Marzari, Chair of Theory and Simulation of Materials, EPFL
Director, National Centre for Competence in Research NCCR MARVEL, SNSF
Laboratory Head, PSI Center for Scientific Computing, Theory, and Data
Contact info and websites:  https://theos-wiki.epfl.ch/en/Main/Contact