Re: [QE-users] Symmetry dependent convergence

[Paolo Giannozzi]

On Thu, Apr 7, 2022 at 3:21 PM Bidault, Xavier  wrote:

Just my 2 cents.
>

I appreciate your 2-cent contribution, but none of your observations is
relevant to this case.

About the original post: while I think that options "noinv" and "nosym" are
a disgrace, I agree that one should obtain the same results with or without
inversion symmetry. I was able to reproduce the problem and it's definitely
a bug

Paolo


--
> *From:* users  on behalf of
> Franklin T Cerasoli via users 
> *Sent:* Wednesday, April 6, 2022 5:52 PM
> *To:* users@lists.quantum-espresso.org 
> *Subject:* [QE-users] Symmetry dependent convergence
>
> Hello,
>
> I've been using QE for several years and recently came across an
> interesting issue when try to converge certain large systems.
>
> Attached is a link to an inputfile for a 216 atom system. With QE-6.8
> (also tested on v6.3), the structure converges only if nosym=.f. and
> noinv=.f.. There are two symmetry operations found, the identity and
> inversion.
>
> Setting nosym=.t. causes the estimated scf accuracy to thrash around in
> the hundreds of thousands of Ry, while nosym=.f. allows strong convergence
> (1e-6 Ry) within 30 iterations.
>
> Pseudopotentials are all pslibrary PBE. Below is a link to the inputfile
> and relevant parts of the scf output.
>
> I would appreciate any advice for understanding what the heck is going on
> here.
>
> -Frank Cerasoli
> UC Davis Department of Chemistry
>
> From the outputs:
>
> With nosym=.f.
>  Program PWSCF v.6.8 starts on  4Apr2022 at 16:30:24
> ...
>   2 Sym. Ops., with inversion, found
> ...
>  iteration #  1 ecut=41.00 Ry beta= 0.10
> ...
>  total energy  =  -84091.83957606 Ry
>  estimated scf accuracy<   7.57588335 Ry
> ...
>  iteration #  2 ecut=41.00 Ry beta= 0.10
>  total energy  =  -84090.73018546 Ry
>  estimated scf accuracy<   5.24675479 Ry
>
>  iteration #  3 ecut=41.00 Ry beta= 0.10
> ...
>  total energy  =  -84091.16546184 Ry
>  estimated scf accuracy<   0.84010802 Ry
>
>
> With nosym=.t.
>  Program PWSCF v.6.8 starts on  4Apr2022 at 18: 9:20
> ...
>  No symmetry found
> ...
>  iteration #  1 ecut=41.00 Ry beta= 0.10
> ...
>  total energy  =  -83948.30076200 Ry
>  estimated scf accuracy<  364559.66452101 Ry
>
>  iteration #  2 ecut=41.00 Ry beta= 0.10
> ...
>  total energy  =  -84392.52517827 Ry
>  estimated scf accuracy<  209530.42525106 Ry
>
>  iteration #  3 ecut=41.00 Ry beta= 0.10
> ...
>  total energy  =  -85690.00953396 Ry
>  estimated scf accuracy<  338759.15202062 Ry
>
>
> And the inputfile is here:
>
> https://drive.google.com/file/d/1jaF7drV996sg4ckxYunyHPpE79SRDdjZ/view?usp=sharing
> 
>
> ___
> The Quantum ESPRESSO community stands by the Ukrainian
> people and expresses its concerns about the devastating
> effects that the Russian military offensive has on their
> country and on the free and peaceful scientific, cultural,
> and economic cooperation amongst peoples
> ___
> Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
> users mailing list users@lists.quantum-espresso.org
> https://lists.quantum-espresso.org/mailman/listinfo/users



-- 
Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
Univ. Udine, via delle Scienze 206, 33100 Udine, Italy
Phone +39-0432-558216, fax +39-0432-558222
___
The Quantum ESPRESSO community stands by the Ukrainian
people and expresses its concerns about the devastating
effects that the Russian military offensive has on their
country and on the free and peaceful scientific, cultural,
and economic cooperation amongst peoples
___
Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
users mailing list users@lists.quantum-espresso.org
https://lists.quantum-espresso.org/mailman/listinfo/users

Re: [QE-users] Symmetry dependent convergence

[Bidault, Xavier]

Hello Franklin,

IMO, the cutoff of 41 Ry is too short for a proper and stable structure 
convergence. How does the stress tensor compare at the last step of vc-relax, 
and at the verification step (latest stress tensor written in the output file)?

I'm more familiar with organic systems. But I think it doesn't require much 
more time to select conv_thr = 1e-6. Then, I find your mixing_beta really 
small. Default is 0.7, meaning 70% of the old guess is mixed with 30% of new, 
if I'm correct. But your value is 0.1, so you barely mix the old guess.

Finally, would there be known showstoppers for In, Cs and or Sb and the related 
PSlib PAW PP? I don't think so, but you can check it up on PSlib website.

Just my 2 cents.

Xavier Bidault


From: users  on behalf of Franklin T 
Cerasoli via users 
Sent: Wednesday, April 6, 2022 5:52 PM
To: users@lists.quantum-espresso.org 
Subject: [QE-users] Symmetry dependent convergence

Hello,

I've been using QE for several years and recently came across an interesting 
issue when try to converge certain large systems.

Attached is a link to an inputfile for a 216 atom system. With QE-6.8 (also 
tested on v6.3), the structure converges only if nosym=.f. and noinv=.f.. There 
are two symmetry operations found, the identity and inversion.

Setting nosym=.t. causes the estimated scf accuracy to thrash around in the 
hundreds of thousands of Ry, while nosym=.f. allows strong convergence (1e-6 
Ry) within 30 iterations.

Pseudopotentials are all pslibrary PBE. Below is a link to the inputfile and 
relevant parts of the scf output.

I would appreciate any advice for understanding what the heck is going on here.

-Frank Cerasoli
UC Davis Department of Chemistry

From the outputs:

With nosym=.f.
 Program PWSCF v.6.8 starts on  4Apr2022 at 16:30:24
...
  2 Sym. Ops., with inversion, found
...
 iteration #  1 ecut=41.00 Ry beta= 0.10
...
 total energy  =  -84091.83957606 Ry
 estimated scf accuracy<   7.57588335 Ry
...
 iteration #  2 ecut=41.00 Ry beta= 0.10
 total energy  =  -84090.73018546 Ry
 estimated scf accuracy<   5.24675479 Ry

 iteration #  3 ecut=41.00 Ry beta= 0.10
...
 total energy  =  -84091.16546184 Ry
 estimated scf accuracy<   0.84010802 Ry


With nosym=.t.
 Program PWSCF v.6.8 starts on  4Apr2022 at 18: 9:20
...
 No symmetry found
...
 iteration #  1 ecut=41.00 Ry beta= 0.10
...
 total energy  =  -83948.30076200 Ry
 estimated scf accuracy<  364559.66452101 Ry

 iteration #  2 ecut=41.00 Ry beta= 0.10
...
 total energy  =  -84392.52517827 Ry
 estimated scf accuracy<  209530.42525106 Ry

 iteration #  3 ecut=41.00 Ry beta= 0.10
...​
 total energy  =  -85690.00953396 Ry
 estimated scf accuracy<  338759.15202062 Ry


And the inputfile is here:
https://drive.google.com/file/d/1jaF7drV996sg4ckxYunyHPpE79SRDdjZ/view?usp=sharing

___
The Quantum ESPRESSO community stands by the Ukrainian
people and expresses its concerns about the devastating
effects that the Russian military offensive has on their
country and on the free and peaceful scientific, cultural,
and economic cooperation amongst peoples
___
Quantum ESPRESSO is supported by MaX (www.max-centre.eu)
users mailing list users@lists.quantum-espresso.org
https://lists.quantum-espresso.org/mailman/listinfo/users