[QE-users] Issue with scf_elastic_constants Calculation in thermo_pw (frozen_ions=.FALSE.)

[Vinay Maithani]

Dear QE community,

I have struggled for over a year to obtain reliable SCF elastic constants
with thermo_pw when ionic relaxation is enabled. I know this query is more
relevant to the thermo_pw community, unfortunately i haven't got any
response from the themo_pw community so I am reaching out to this forum in
anticipation:

The themo_control file is:
&INPUT_THERMO
  what = 'scf_elastic_constants',
  frozen_ions = .FALSE.
/
With `frozen_ions=.TRUE.` everything converges, but setting `.FALSE.`
aborts with

```


*task # 0from bfgs : error # 1dE0s is positive which should never happen*
```
Lowering thresholds  (etot_conv_thr)  yields elastic constants, yet the
values swing wildly (tens of GPa's) with minute changes in energy
tolerances, given that the system I am working on is of low-symmetry
(triclinic).

Additionally,following adjustments didn't yield any success:

- `trust_radius_max` reduced to 0.05 a.u.

- tighter `conv_thr`, `etot_conv_thr`, `forc_conv_thr`

- `ion_dynamics='fire'` and `'damp'` (BFGS still invoked in thermo_pw)

Has anyone resolved this coupled QE and thermo_pw issue  or identified
safer optimization settings (e.g., alternative algorithms, pre-relax
procedures, or recommended `trust_radius_max`) that stabilize the
calculation?

Any guidance or example inputs would be greatly appreciated, I will be
happy to share my input files if needed.

Thank you for your time.

Best regards,
Vinay Maithani
IIT Kharagpur


On Tue, Mar 4, 2025 at 12:08 PM Vinay Maithani <
vinaymaithani21 at kgpian.iitkgp.ac.in> wrote:

> Dear QE community,
>
> I am currently utilizing thermo_pw to compute the self-consistent field
> (SCF) elastic constants for various materials. While setting
> frozen_ions=.TRUE. in the &INPUT_THERMO namelist yields successful
> results, attempts with frozen_ions=.FALSE. lead to the following error:
>
> *task #         0
>      from bfgs : error #         1
>      dE0s is positive which should never happen*
>
> To address this issue, I have implemented several strategies without
> success:
>
>    1. Reducing the trust_radius_max parameter to as low as 0.05.
>    2. Adjusting convergence thresholds within the input file.
>    3. Modifying ion_dynamics to 'fire' and 'damp'; however, the
>    optimization persists with the BFGS algorithm.
>
> I seek your guidance on resolving this error to facilitate accurate
> elastic constant calculations with ionic relaxation.
>
> Thank you for your assistance.
>
> --
> Regards,
> Vinay Maithani
> Department of Metallurgical and Materials Engineering
> Indian Institute of Technology, Kharagpur
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20250611/57874088/attachment.html>