Net force (by HSE LCAO) is large

[MoseyQAQ]

Describe the bug

I am trying to optimize a structure for a 5-atom unit cell using HSE06 functional in ABACUS+LCAO.

I found the net force is large:

#TOTAL-FORCE (eV/Angstrom)#
 -------------------------------------------------------------------------
     Atoms              Force_x              Force_y              Force_z
 -------------------------------------------------------------------------
       Ba1         0.0177951935         0.0177488828         0.0177028125
       Ti1         0.0158074471         0.0157663094         0.0157253852
        O1         0.0062647778        -0.0199501018        -0.0198983179
        O2        -0.0199337092         0.0063167428        -0.0198983179
        O3        -0.0199337092        -0.0198818332         0.0063684380
 -------------------------------------------------------------------------
                Net force vector (eV/Ang) = [ 0.2373037792, 0.83973197546, -1.0772531237 ]
                     Total drift (ev/Ang) = 1.3863394847

My INPUT file:

INPUT_PARAMETERS
calculation                             relax
relax_nmax                              100
basis_type                              lcao
ecutwfc                                 100
scf_thr                                 1e-7
scf_nmax                                100

gamma_only                              0
kspacing                                0.1

smearing_method                         gauss
smearing_sigma                          0.003

mixing_type                             pulay
mixing_beta                             0.4
dft_functional                          hse

pseudo_dir  .........../pp_orb/apns-pseudopotentials-v1
orbital_dir .........../pp_orb/apns-orbitals-efficiency-v1

I am not sure whether it's expected behavior. Can I reduce the total drift by adujusting parameters?

The attachement is the original input and output:
hse-omp.tar.gz

Expected behavior

No response

To Reproduce

No response

Environment

ABACUS version: v3.11.0-beta8
Compiler: openmpi/5.0.8 + gcc/14.1.0 + scalapack + openblas (both mathlibs were compiled by toolchain

Additional Context

No response

Task list for Issue attackers (only for developers)

• Verify the issue is not a duplicate.
• Describe the bug.
• Steps to reproduce.
• Expected behavior.
• Error message.
• Environment details.
• Additional context.
• Assign a priority level (low, medium, high, urgent).
• Assign the issue to a team member.
• Label the issue with relevant tags.
• Identify possible related issues.
• Create a unit test or automated test to reproduce the bug (if applicable).
• Fix the bug.
• Test the fix.
• Update documentation (if necessary).
• Close the issue and inform the reporter (if applicable).

[mohanchen]

Did you find the same behavior in 3.10.LTS version? It's useful to first confirm this point.