BUG: Geometry Optimization badly in simple Pt111 surface

QuantumMisaka commented 10 months ago

Describe the bug

When doing relax calculation in simple 64-atoms Pt111 surfaces using bfgs method, the force grow larger to more than 100 eV/A Like:

(ase) [liuzq@amd41 Pt111]$ grep GRAD abacus.log 
 LARGEST GRAD (eV/A)  : 6.309e-01
 LARGEST GRAD (eV/A)  : 4.029e-01
 LARGEST GRAD (eV/A)  : 1.306e+03
 LARGEST GRAD (eV/A)  : 2.539e+02
 LARGEST GRAD (eV/A)  : 1.460e+03
 LARGEST GRAD (eV/A)  : 7.242e+02
 LARGEST GRAD (eV/A)  : 1.069e+03
 LARGEST GRAD (eV/A)  : 1.317e+03
 LARGEST GRAD (eV/A)  : 1.398e+03
 LARGEST GRAD (eV/A)  : 1.431e+03
 LARGEST GRAD (eV/A)  : 1.446e+03
 LARGEST GRAD (eV/A)  : 1.453e+03
 LARGEST GRAD (eV/A)  : 1.457e+03
 LARGEST GRAD (eV/A)  : 1.459e+03
 LARGEST GRAD (eV/A)  : 1.459e+03
 LARGEST GRAD (eV/A)  : 1.460e+03
 LARGEST GRAD (eV/A)  : 1.460e+03
 LARGEST GRAD (eV/A)  : 1.460e+03
 LARGEST GRAD (eV/A)  : 1.460e+03
 LARGEST GRAD (eV/A)  : 1.460e+03
 LARGEST GRAD (eV/A)  : 1.460e+03

And then

 trust radius is too small! Break down.
 CHECK IN FILE : OUT.Pt111/warning.log

If use CG method whith Gaussian smearing and smearing_sigma 0.002, calculation can be done with 16 ION steps

 LARGEST GRAD (eV/A)  : 6.794e-01
 LARGEST GRAD (eV/A)  : 5.804e-01
 LARGEST GRAD (eV/A)  : 7.134e-01
 LARGEST GRAD (eV/A)  : 5.683e-01
 LARGEST GRAD (eV/A)  : 7.781e-01
 LARGEST GRAD (eV/A)  : 6.223e-01
 LARGEST GRAD (eV/A)  : 5.026e-01
 LARGEST GRAD (eV/A)  : 3.539e-01
 LARGEST GRAD (eV/A)  : 2.371e-01
 LARGEST GRAD (eV/A)  : 9.594e-02
 LARGEST GRAD (eV/A)  : 6.459e-02
 LARGEST GRAD (eV/A)  : 5.225e-02
 LARGEST GRAD (eV/A)  : 5.345e-02
 LARGEST GRAD (eV/A)  : 6.003e-02
 LARGEST GRAD (eV/A)  : 8.091e-02
 LARGEST GRAD (eV/A)  : 4.367e-02

but this structure can be easily relax by VASP with `ISMEAR=1, SIGMA=0.1

(base) [2201110432@wm2-login01 dipole]$ grep FORCE OUTCAR 
  FORCE on cell =-STRESS in cart. coord.  units (eV):
 FORCES acting on ions
 POSITION                                       TOTAL-FORCE (eV/Angst)
  FORCES: max atom, RMS     0.478433    0.254403
  FORCE total and by dimension    2.035220    0.478433
  FORCE on cell =-STRESS in cart. coord.  units (eV):
 FORCES acting on ions
 POSITION                                       TOTAL-FORCE (eV/Angst)
  FORCES: max atom, RMS     0.340290    0.176057
  FORCE total and by dimension    1.408459    0.340290
  FORCE on cell =-STRESS in cart. coord.  units (eV):
 FORCES acting on ions
 POSITION                                       TOTAL-FORCE (eV/Angst)
  FORCES: max atom, RMS     0.245468    0.149148
  FORCE total and by dimension    1.193184    0.245468
  FORCE on cell =-STRESS in cart. coord.  units (eV):
 FORCES acting on ions
 POSITION                                       TOTAL-FORCE (eV/Angst)
  FORCES: max atom, RMS     0.191442    0.132379
  FORCE total and by dimension    1.059034    0.191442
  FORCE on cell =-STRESS in cart. coord.  units (eV):
 FORCES acting on ions
 POSITION                                       TOTAL-FORCE (eV/Angst)
  FORCES: max atom, RMS     0.139927    0.107616
  FORCE total and by dimension    0.860928    0.139927
  FORCE on cell =-STRESS in cart. coord.  units (eV):
 FORCES acting on ions
 POSITION                                       TOTAL-FORCE (eV/Angst)
  FORCES: max atom, RMS     0.111320    0.084793
  FORCE total and by dimension    0.678347    0.111320
  FORCE on cell =-STRESS in cart. coord.  units (eV):
 FORCES acting on ions
 POSITION                                       TOTAL-FORCE (eV/Angst)
  FORCES: max atom, RMS     0.041044    0.026231
  FORCE total and by dimension    0.209851    0.041044

and Pt(111) surface should be a metal system, but when using CG method with mp smearing with smearing_sigma 0.008 , optimization is also hard to be done.

 LARGEST GRAD (eV/A)  : 6.359e-01
 LARGEST GRAD (eV/A)  : 5.287e-01
 LARGEST GRAD (eV/A)  : 6.775e-01
 LARGEST GRAD (eV/A)  : 4.233e+00
 LARGEST GRAD (eV/A)  : 1.677e+01
 LARGEST GRAD (eV/A)  : 1.611e+01
 LARGEST GRAD (eV/A)  : 1.411e+01
 LARGEST GRAD (eV/A)  : 1.326e+01
 LARGEST GRAD (eV/A)  : 1.072e+01
 LARGEST GRAD (eV/A)  : 9.763e+00
 LARGEST GRAD (eV/A)  : 6.893e+00
 LARGEST GRAD (eV/A)  : 6.001e+00
 LARGEST GRAD (eV/A)  : 3.326e+00
 LARGEST GRAD (eV/A)  : 2.709e+00
 LARGEST GRAD (eV/A)  : 8.593e-01
 LARGEST GRAD (eV/A)  : 6.650e-01
 LARGEST GRAD (eV/A)  : 6.637e-01
 LARGEST GRAD (eV/A)  : 6.631e-01
 LARGEST GRAD (eV/A)  : 6.613e-01
 LARGEST GRAD (eV/A)  : 6.603e-01
 LARGEST GRAD (eV/A)  : 6.572e-01
 LARGEST GRAD (eV/A)  : 6.557e-01
 LARGEST GRAD (eV/A)  : 6.509e-01
 LARGEST GRAD (eV/A)  : 6.484e-01
 LARGEST GRAD (eV/A)  : 6.409e-01
 LARGEST GRAD (eV/A)  : 6.370e-01
 LARGEST GRAD (eV/A)  : 6.253e-01
 LARGEST GRAD (eV/A)  : 6.194e-01
 LARGEST GRAD (eV/A)  : 6.018e-01
 LARGEST GRAD (eV/A)  : 5.930e-01
 LARGEST GRAD (eV/A)  : 5.665e-01
 LARGEST GRAD (eV/A)  : 5.539e-01
 LARGEST GRAD (eV/A)  : 5.163e-01
 LARGEST GRAD (eV/A)  : 7.314e-01
 LARGEST GRAD (eV/A)  : 3.935e+00
 LARGEST GRAD (eV/A)  : 9.860e+00
 LARGEST GRAD (eV/A)  : 4.475e-01
 LARGEST GRAD (eV/A)  : 5.035e-01
 LARGEST GRAD (eV/A)  : 1.675e+00
 LARGEST GRAD (eV/A)  : 5.031e-01
 LARGEST GRAD (eV/A)  : 5.037e-01
 LARGEST GRAD (eV/A)  : 8.366e-01
 LARGEST GRAD (eV/A)  : 1.733e+00
 LARGEST GRAD (eV/A)  : 1.323e+00
 LARGEST GRAD (eV/A)  : 1.432e+00
 LARGEST GRAD (eV/A)  : 1.429e+00
 LARGEST GRAD (eV/A)  : 1.421e+00
 LARGEST GRAD (eV/A)  : 1.219e+00
 LARGEST GRAD (eV/A)  : 1.724e+00
 LARGEST GRAD (eV/A)  : 1.697e+00
 LARGEST GRAD (eV/A)  : 1.616e+00
 LARGEST GRAD (eV/A)  : 4.267e-01
 LARGEST GRAD (eV/A)  : 1.787e+00
 LARGEST GRAD (eV/A)  : 1.895e+00
 LARGEST GRAD (eV/A)  : 2.217e+00
 LARGEST GRAD (eV/A)  : 4.229e-01
 LARGEST GRAD (eV/A)  : 1.274e+00
 LARGEST GRAD (eV/A)  : 1.281e+00
 LARGEST GRAD (eV/A)  : 1.445e+00
 LARGEST GRAD (eV/A)  : 9.905e-01
 LARGEST GRAD (eV/A)  : 9.034e-01
 LARGEST GRAD (eV/A)  : 8.726e-01
 LARGEST GRAD (eV/A)  : 8.333e-01
 LARGEST GRAD (eV/A)  : 1.381e+00
 LARGEST GRAD (eV/A)  : 1.458e+00
 LARGEST GRAD (eV/A)  : 1.346e+00
 LARGEST GRAD (eV/A)  : 1.719e+00
 LARGEST GRAD (eV/A)  : 4.074e-01
 LARGEST GRAD (eV/A)  : 1.462e+00
 LARGEST GRAD (eV/A)  : 1.624e+00
 LARGEST GRAD (eV/A)  : 2.109e+00
 LARGEST GRAD (eV/A)  : 4.029e-01
 LARGEST GRAD (eV/A)  : 1.038e+00
 LARGEST GRAD (eV/A)  : 1.091e+00
 LARGEST GRAD (eV/A)  : 2.614e+00
 LARGEST GRAD (eV/A)  : 4.191e-01
 LARGEST GRAD (eV/A)  : 1.180e+00
 LARGEST GRAD (eV/A)  : 1.240e+00
 LARGEST GRAD (eV/A)  : 1.420e+00
 LARGEST GRAD (eV/A)  : 1.390e+00
 LARGEST GRAD (eV/A)  : 1.285e+00
 LARGEST GRAD (eV/A)  : 1.214e+00
 LARGEST GRAD (eV/A)  : 1.001e+00
 LARGEST GRAD (eV/A)  : 8.908e-01
 LARGEST GRAD (eV/A)  : 1.220e+00
 LARGEST GRAD (eV/A)  : 1.209e+00
 LARGEST GRAD (eV/A)  : 1.173e+00
 LARGEST GRAD (eV/A)  : 3.865e-01
 LARGEST GRAD (eV/A)  : 1.286e+00
 LARGEST GRAD (eV/A)  : 1.402e+00
 LARGEST GRAD (eV/A)  : 1.751e+00
 LARGEST GRAD (eV/A)  : 3.830e-01
 LARGEST GRAD (eV/A)  : 7.659e-01
 LARGEST GRAD (eV/A)  : 7.864e-01
 LARGEST GRAD (eV/A)  : 8.478e-01
 LARGEST GRAD (eV/A)  : 4.217e-01
 LARGEST GRAD (eV/A)  : 1.372e+00
 LARGEST GRAD (eV/A)  : 1.692e+00
 LARGEST GRAD (eV/A)  : 1.957e+00
 LARGEST GRAD (eV/A)  : 6.963e-01
 LARGEST GRAD (eV/A)  : 4.373e-01
 LARGEST GRAD (eV/A)  : 3.749e-01
 LARGEST GRAD (eV/A)  : 7.242e-01
 LARGEST GRAD (eV/A)  : 3.738e-01
 LARGEST GRAD (eV/A)  : 1.402e+00
 LARGEST GRAD (eV/A)  : 1.336e+00
 LARGEST GRAD (eV/A)  : 1.137e+00
 LARGEST GRAD (eV/A)  : 5.627e-01
 LARGEST GRAD (eV/A)  : 4.447e-01
 LARGEST GRAD (eV/A)  : 3.709e-01
 LARGEST GRAD (eV/A)  : 4.638e-01

Also, if use ASE to do BFGS using ABACUS calculator, the force can be normally lower

  Loading requirement: openmpi/4.1.5
      Step     Time          Energy          fmax
BFGS:    0 10:42:40  -211246.084060         0.630909
BFGS:    1 10:52:47  -211246.176390         0.608847
BFGS:    2 11:04:45  -211246.659120         1.220412
BFGS:    3 11:18:10  -211246.999820         0.481578
BFGS:    4 11:31:16  -211247.105570         0.512988
BFGS:    5 11:45:03  -211247.496240         0.611267
BFGS:    6 12:00:44  -211247.630610         0.591031
BFGS:    7 12:15:23  -211247.968720         0.479451
BFGS:    8 12:29:13  -211248.112920         0.299274
BFGS:    9 12:45:03  -211248.193780         0.168083
BFGS:   10 12:58:21  -211248.220070        0.029687

Input files and BFGS results are below 64core_bfgs.tar.gz

Expected behavior

The structural relaxation can be done properly and efficiently

To Reproduce

Build ABACUS with intel-icc toolchain
run by OMP_NUM_THREADS=16 mpirun -np 4 abacus
The result can also be reproduced by gcc-toolchain-installed ABACUS with OMP_NUM_THREADS=4 mpirun -np 4 abacus
Tried to unset dipole correction, no use
Tried to tighten scf_thr from 1e-6 to 1e-8, no use

Environment

ABACUS version: 3.3.4
Dependencies: Intel-OneAPI 2023.0 or gcc-toolchain

Additional Context

No response

Task list for Issue attackers

[X] Verify the issue is not a duplicate.
[X] Describe the bug.
[X] Steps to reproduce.
[X] Expected behavior.
[X] Error message.
[X] Environment details.
[x] Additional context.
[x] 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).

QuantumMisaka commented 10 months ago

Tried ABACUS 3.3.3, the problem also exists

QuantumMisaka commented 10 months ago

Tried ASE-ABACUS using ase.optimizer.QuasiNewton: converged quick:ly

BFGSLineSearch:    0[  0] 16:12:24  -211246.084060*       0.6309
BFGSLineSearch:    1[  3] 16:42:53  -211246.861010*       0.6315
BFGSLineSearch:    2[  4] 16:53:20  -211247.372430*       0.4717
BFGSLineSearch:    3[  5] 17:04:44  -211247.973470*       0.5044
BFGSLineSearch:    4[  7] 17:25:51  -211248.218230*       0.0590
BFGSLineSearch:    5[  9] 17:48:26  -211248.219120*       0.0366

with the same INPUT (change calculation from relax to scf, smearing: mp and 0.008 sigma),

QuantumMisaka commented 10 months ago

Tried bfgs with Gaussian smearing, the problem also emerged

QuantumMisaka commented 10 months ago

Tried bfgs relaxation in Cu111-N2 system, this problem still occur:

 LARGEST GRAD (eV/A)  : 3.490e+00
 LARGEST GRAD (eV/A)  : 3.321e+00
 LARGEST GRAD (eV/A)  : 1.747e+02
 LARGEST GRAD (eV/A)  : 2.747e+00
 LARGEST GRAD (eV/A)  : 1.355e+00
 LARGEST GRAD (eV/A)  : 1.707e+00
 LARGEST GRAD (eV/A)  : 8.912e-01
 LARGEST GRAD (eV/A)  : 1.142e+00
 LARGEST GRAD (eV/A)  : 4.499e-01
 LARGEST GRAD (eV/A)  : 1.064e+00
 LARGEST GRAD (eV/A)  : 7.257e-01
 LARGEST GRAD (eV/A)  : 6.735e-01
 LARGEST GRAD (eV/A)  : 8.031e+00
 LARGEST GRAD (eV/A)  : 2.288e+01
 LARGEST GRAD (eV/A)  : 1.067e+02
 LARGEST GRAD (eV/A)  : 4.990e+01
 LARGEST GRAD (eV/A)  : 1.016e+02
 LARGEST GRAD (eV/A)  : 1.049e+02
 LARGEST GRAD (eV/A)  : 1.059e+02
 LARGEST GRAD (eV/A)  : 1.063e+02
 LARGEST GRAD (eV/A)  : 1.065e+02
 LARGEST GRAD (eV/A)  : 1.066e+02
 LARGEST GRAD (eV/A)  : 1.066e+02
 LARGEST GRAD (eV/A)  : 1.067e+02
 LARGEST GRAD (eV/A)  : 1.067e+02
 LARGEST GRAD (eV/A)  : 1.067e+02
 LARGEST GRAD (eV/A)  : 1.067e+02
 LARGEST GRAD (eV/A)  : 1.067e+02
 LARGEST GRAD (eV/A)  : 1.067e+02

But energy seems to be in convergence

 ETOT DIFF (eV)       : 0.000e+00
 ETOT DIFF (eV)       : -1.701e+00
 ETOT DIFF (eV)       : 2.519e+01
 ETOT DIFF (eV)       : -1.155e-01
 ETOT DIFF (eV)       : -2.368e-01
 ETOT DIFF (eV)       : -1.355e-01
 ETOT DIFF (eV)       : -1.274e-01
 ETOT DIFF (eV)       : -1.080e-01
 ETOT DIFF (eV)       : -6.072e-02
 ETOT DIFF (eV)       : -1.077e-01
 ETOT DIFF (eV)       : -5.966e-02
 ETOT DIFF (eV)       : -1.382e-01
 ETOT DIFF (eV)       : 1.029e+00
 ETOT DIFF (eV)       : -7.723e-02
 ETOT DIFF (eV)       : -5.084e+00
 ETOT DIFF (eV)       : 5.797e+00
 ETOT DIFF (eV)       : 5.620e-01
 ETOT DIFF (eV)       : 1.955e-01
 ETOT DIFF (eV)       : 8.475e-02
 ETOT DIFF (eV)       : 3.976e-02
 ETOT DIFF (eV)       : 1.928e-02
 ETOT DIFF (eV)       : 9.495e-03
 ETOT DIFF (eV)       : 4.712e-03
 ETOT DIFF (eV)       : 2.348e-03
 ETOT DIFF (eV)       : 1.172e-03
 ETOT DIFF (eV)       : 5.859e-04
 ETOT DIFF (eV)       : 2.929e-04
 ETOT DIFF (eV)       : 1.468e-04
 ETOT DIFF (eV)       : 2.579e-03

running file:

setting: OMP_NUM_THREADS=16 mpirun -np 1 abacus with v3.3.4

seems relaxation module in ABACUS have problem when tackle with fcc111 system ?

Cu111-N2_wrong.tar.gz

QuantumMisaka commented 10 months ago

In Au111-H2 and Au111-2H, bfgs have the likely problem

[liuzq@amd41 2H]$ grep GRAD abacus.log 
 LARGEST GRAD (eV/A)  : 9.010e-01
 LARGEST GRAD (eV/A)  : 4.952e-01
 LARGEST GRAD (eV/A)  : 1.679e+00
 LARGEST GRAD (eV/A)  : 9.513e-01
 LARGEST GRAD (eV/A)  : 4.968e-01
 LARGEST GRAD (eV/A)  : 4.946e-01
 LARGEST GRAD (eV/A)  : 4.979e-01
 LARGEST GRAD (eV/A)  : 4.947e-01
 LARGEST GRAD (eV/A)  : 4.971e-01
 LARGEST GRAD (eV/A)  : 4.949e-01
 LARGEST GRAD (eV/A)  : 4.962e-01
 LARGEST GRAD (eV/A)  : 4.954e-01
 LARGEST GRAD (eV/A)  : 4.961e-01
 LARGEST GRAD (eV/A)  : 4.957e-01
 LARGEST GRAD (eV/A)  : 4.956e-01
 LARGEST GRAD (eV/A)  : 4.957e-01
 LARGEST GRAD (eV/A)  : 4.953e-01
 LARGEST GRAD (eV/A)  : 4.956e-01
 LARGEST GRAD (eV/A)  : 4.951e-01

While CG method can reach converge

smearing method is gau and sigma=0.002

QuantumMisaka commented 10 months ago

Tried PW method in the advice of @kirk0830 , despite the low speed ( LCAO and PW all run by 16core, LCAO by OMP, PW by MPI ), the largest GRAD still have problem BFGS:

 LARGEST GRAD (eV/A)  : 5.863e-01
 LARGEST GRAD (eV/A)  : 3.839e-01
 LARGEST GRAD (eV/A)  : 1.541e+03
 LARGEST GRAD (eV/A)  : 3.876e+02
 LARGEST GRAD (eV/A)  : 1.759e+03
 LARGEST GRAD (eV/A)  : 8.884e+02

CG:

 LARGEST GRAD (eV/A)  : 5.840e-01
 LARGEST GRAD (eV/A)  : 4.349e-01
 LARGEST GRAD (eV/A)  : 2.985e-01
 LARGEST GRAD (eV/A)  : 7.125e-01
 LARGEST GRAD (eV/A)  : 5.687e-01

QuantumMisaka commented 9 months ago

If transform z-axis vacuum to y-axis vacuum by ATOMKIT function 405, the BFGS and CG will converge, but also need lots of ion steps BFGS: need 149 ion steps

 LARGEST GRAD (eV/A)  : 3.793e-01
 LARGEST GRAD (eV/A)  : 7.951e+00
 LARGEST GRAD (eV/A)  : 6.399e-01
 LARGEST GRAD (eV/A)  : 3.714e-01
 LARGEST GRAD (eV/A)  : 3.783e-01
 LARGEST GRAD (eV/A)  : 3.760e-01
 LARGEST GRAD (eV/A)  : 3.778e-01
 LARGEST GRAD (eV/A)  : 3.750e-01
 LARGEST GRAD (eV/A)  : 3.784e-01
 LARGEST GRAD (eV/A)  : 3.749e-01
 LARGEST GRAD (eV/A)  : 3.777e-01
 LARGEST GRAD (eV/A)  : 3.732e-01
 LARGEST GRAD (eV/A)  : 3.775e-01
 LARGEST GRAD (eV/A)  : 3.717e-01
 LARGEST GRAD (eV/A)  : 3.790e-01
 LARGEST GRAD (eV/A)  : 3.736e-01
 LARGEST GRAD (eV/A)  : 3.802e-01
 LARGEST GRAD (eV/A)  : 3.741e-01
 LARGEST GRAD (eV/A)  : 3.761e-01
 LARGEST GRAD (eV/A)  : 3.742e-01
 LARGEST GRAD (eV/A)  : 3.754e-01
 LARGEST GRAD (eV/A)  : 3.738e-01
 LARGEST GRAD (eV/A)  : 3.736e-01
 LARGEST GRAD (eV/A)  : 3.726e-01
 LARGEST GRAD (eV/A)  : 3.735e-01
 LARGEST GRAD (eV/A)  : 3.731e-01
 LARGEST GRAD (eV/A)  : 3.728e-01
 LARGEST GRAD (eV/A)  : 3.727e-01
 LARGEST GRAD (eV/A)  : 3.722e-01
 LARGEST GRAD (eV/A)  : 3.727e-01
 LARGEST GRAD (eV/A)  : 3.724e-01
 LARGEST GRAD (eV/A)  : 3.692e-01
 LARGEST GRAD (eV/A)  : 3.681e-01
 LARGEST GRAD (eV/A)  : 3.673e-01
 LARGEST GRAD (eV/A)  : 3.664e-01
 LARGEST GRAD (eV/A)  : 3.639e-01
 LARGEST GRAD (eV/A)  : 3.640e-01
 LARGEST GRAD (eV/A)  : 3.652e-01
 LARGEST GRAD (eV/A)  : 3.632e-01
 LARGEST GRAD (eV/A)  : 3.618e-01
 LARGEST GRAD (eV/A)  : 3.606e-01
 LARGEST GRAD (eV/A)  : 3.601e-01
 LARGEST GRAD (eV/A)  : 3.608e-01
 LARGEST GRAD (eV/A)  : 3.624e-01
 LARGEST GRAD (eV/A)  : 3.609e-01
 LARGEST GRAD (eV/A)  : 3.591e-01
 LARGEST GRAD (eV/A)  : 3.520e-01
 LARGEST GRAD (eV/A)  : 3.464e-01
 LARGEST GRAD (eV/A)  : 3.544e-01
 LARGEST GRAD (eV/A)  : 3.634e-01
 LARGEST GRAD (eV/A)  : 3.638e-01
 LARGEST GRAD (eV/A)  : 3.669e-01
 LARGEST GRAD (eV/A)  : 3.830e-01
 LARGEST GRAD (eV/A)  : 4.006e-01
 LARGEST GRAD (eV/A)  : 3.820e-01
 LARGEST GRAD (eV/A)  : 3.868e-01
 LARGEST GRAD (eV/A)  : 3.883e-01
 LARGEST GRAD (eV/A)  : 3.950e-01
 LARGEST GRAD (eV/A)  : 3.938e-01
 LARGEST GRAD (eV/A)  : 3.926e-01
 LARGEST GRAD (eV/A)  : 3.899e-01
 LARGEST GRAD (eV/A)  : 3.873e-01
 LARGEST GRAD (eV/A)  : 3.853e-01
 LARGEST GRAD (eV/A)  : 3.838e-01
 LARGEST GRAD (eV/A)  : 3.810e-01
 LARGEST GRAD (eV/A)  : 3.775e-01
 LARGEST GRAD (eV/A)  : 3.741e-01
 LARGEST GRAD (eV/A)  : 3.699e-01
 LARGEST GRAD (eV/A)  : 3.655e-01
 LARGEST GRAD (eV/A)  : 3.613e-01
 LARGEST GRAD (eV/A)  : 3.561e-01
 LARGEST GRAD (eV/A)  : 3.493e-01
 LARGEST GRAD (eV/A)  : 3.444e-01
 LARGEST GRAD (eV/A)  : 3.365e-01
 LARGEST GRAD (eV/A)  : 3.273e-01
 LARGEST GRAD (eV/A)  : 3.200e-01
 LARGEST GRAD (eV/A)  : 3.108e-01
 LARGEST GRAD (eV/A)  : 2.826e-01
 LARGEST GRAD (eV/A)  : 2.286e-01
 LARGEST GRAD (eV/A)  : 2.427e+00
 LARGEST GRAD (eV/A)  : 8.800e-01
 LARGEST GRAD (eV/A)  : 8.176e+00
 LARGEST GRAD (eV/A)  : 8.781e-01
 LARGEST GRAD (eV/A)  : 8.793e-01
 LARGEST GRAD (eV/A)  : 8.747e-01
 LARGEST GRAD (eV/A)  : 8.795e-01
 LARGEST GRAD (eV/A)  : 8.730e-01
 LARGEST GRAD (eV/A)  : 8.775e-01
 LARGEST GRAD (eV/A)  : 8.719e-01
 LARGEST GRAD (eV/A)  : 8.782e-01
 LARGEST GRAD (eV/A)  : 8.709e-01
 LARGEST GRAD (eV/A)  : 8.762e-01
 LARGEST GRAD (eV/A)  : 8.701e-01
 LARGEST GRAD (eV/A)  : 8.771e-01
 LARGEST GRAD (eV/A)  : 8.693e-01
 LARGEST GRAD (eV/A)  : 8.751e-01
 LARGEST GRAD (eV/A)  : 8.694e-01
 LARGEST GRAD (eV/A)  : 8.744e-01
 LARGEST GRAD (eV/A)  : 8.662e-01
 LARGEST GRAD (eV/A)  : 8.740e-01
 LARGEST GRAD (eV/A)  : 8.669e-01
 LARGEST GRAD (eV/A)  : 8.719e-01
 LARGEST GRAD (eV/A)  : 8.657e-01
 LARGEST GRAD (eV/A)  : 8.718e-01
 LARGEST GRAD (eV/A)  : 8.642e-01
 LARGEST GRAD (eV/A)  : 8.689e-01
 LARGEST GRAD (eV/A)  : 8.625e-01
 LARGEST GRAD (eV/A)  : 8.682e-01
 LARGEST GRAD (eV/A)  : 8.605e-01
 LARGEST GRAD (eV/A)  : 8.647e-01
 LARGEST GRAD (eV/A)  : 8.591e-01
 LARGEST GRAD (eV/A)  : 8.617e-01
 LARGEST GRAD (eV/A)  : 8.554e-01
 LARGEST GRAD (eV/A)  : 8.557e-01
 LARGEST GRAD (eV/A)  : 8.549e-01
 LARGEST GRAD (eV/A)  : 8.536e-01
 LARGEST GRAD (eV/A)  : 8.509e-01
 LARGEST GRAD (eV/A)  : 8.480e-01
 LARGEST GRAD (eV/A)  : 8.453e-01
 LARGEST GRAD (eV/A)  : 8.423e-01
 LARGEST GRAD (eV/A)  : 8.388e-01
 LARGEST GRAD (eV/A)  : 8.334e-01
 LARGEST GRAD (eV/A)  : 8.315e-01
 LARGEST GRAD (eV/A)  : 8.237e-01
 LARGEST GRAD (eV/A)  : 8.225e-01
 LARGEST GRAD (eV/A)  : 8.154e-01
 LARGEST GRAD (eV/A)  : 8.114e-01
 LARGEST GRAD (eV/A)  : 8.037e-01
 LARGEST GRAD (eV/A)  : 7.998e-01
 LARGEST GRAD (eV/A)  : 7.912e-01
 LARGEST GRAD (eV/A)  : 7.843e-01
 LARGEST GRAD (eV/A)  : 6.781e-01
 LARGEST GRAD (eV/A)  : 6.218e-01
 LARGEST GRAD (eV/A)  : 6.141e-01
 LARGEST GRAD (eV/A)  : 5.769e-01
 LARGEST GRAD (eV/A)  : 5.571e-01
 LARGEST GRAD (eV/A)  : 5.349e-01
 LARGEST GRAD (eV/A)  : 5.291e-01
 LARGEST GRAD (eV/A)  : 4.947e-01
 LARGEST GRAD (eV/A)  : 4.609e-01
 LARGEST GRAD (eV/A)  : 2.467e-01
 LARGEST GRAD (eV/A)  : 1.281e-01
 LARGEST GRAD (eV/A)  : 1.634e-01
 LARGEST GRAD (eV/A)  : 1.962e-01
 LARGEST GRAD (eV/A)  : 1.648e-01
 LARGEST GRAD (eV/A)  : 8.909e-02
 LARGEST GRAD (eV/A)  : 1.875e-01
 LARGEST GRAD (eV/A)  : 5.806e-02
 LARGEST GRAD (eV/A)  : 3.949e-02

CG: need 138 ion steps

 LARGEST GRAD (eV/A)  : 3.793e-01
 LARGEST GRAD (eV/A)  : 1.708e+00
 LARGEST GRAD (eV/A)  : 2.563e+00
 LARGEST GRAD (eV/A)  : 2.292e+00
 LARGEST GRAD (eV/A)  : 1.512e+00
 LARGEST GRAD (eV/A)  : 1.087e+00
 LARGEST GRAD (eV/A)  : 2.172e-01
 LARGEST GRAD (eV/A)  : 2.059e-01
 LARGEST GRAD (eV/A)  : 2.030e-01
 LARGEST GRAD (eV/A)  : 1.776e-01
 LARGEST GRAD (eV/A)  : 2.240e-01
 LARGEST GRAD (eV/A)  : 1.721e-01
 LARGEST GRAD (eV/A)  : 5.395e-01
 LARGEST GRAD (eV/A)  : 3.244e-01
 LARGEST GRAD (eV/A)  : 8.697e-01
 LARGEST GRAD (eV/A)  : 8.633e-01
 LARGEST GRAD (eV/A)  : 8.464e-01
 LARGEST GRAD (eV/A)  : 4.065e-01
 LARGEST GRAD (eV/A)  : 6.686e-01
 LARGEST GRAD (eV/A)  : 1.752e-01
 LARGEST GRAD (eV/A)  : 5.221e-01
 LARGEST GRAD (eV/A)  : 4.966e-01
 LARGEST GRAD (eV/A)  : 4.447e-01
 LARGEST GRAD (eV/A)  : 1.941e-01
 LARGEST GRAD (eV/A)  : 7.665e-01
 LARGEST GRAD (eV/A)  : 7.271e-01
 LARGEST GRAD (eV/A)  : 6.244e-01
 LARGEST GRAD (eV/A)  : 3.657e-01
 LARGEST GRAD (eV/A)  : 3.863e-01
 LARGEST GRAD (eV/A)  : 3.251e-01
 LARGEST GRAD (eV/A)  : 1.690e-01
 LARGEST GRAD (eV/A)  : 1.246e-01
 LARGEST GRAD (eV/A)  : 2.769e-01
 LARGEST GRAD (eV/A)  : 1.819e-01
 LARGEST GRAD (eV/A)  : 7.915e-01
 LARGEST GRAD (eV/A)  : 1.976e-01
 LARGEST GRAD (eV/A)  : 3.705e-01
 LARGEST GRAD (eV/A)  : 3.701e-01
 LARGEST GRAD (eV/A)  : 3.769e-01
 LARGEST GRAD (eV/A)  : 2.852e-01
 LARGEST GRAD (eV/A)  : 6.451e-01
 LARGEST GRAD (eV/A)  : 6.061e-01
 LARGEST GRAD (eV/A)  : 5.077e-01
 LARGEST GRAD (eV/A)  : 2.688e-01
 LARGEST GRAD (eV/A)  : 3.843e-01
 LARGEST GRAD (eV/A)  : 3.565e-01
 LARGEST GRAD (eV/A)  : 2.634e-01
 LARGEST GRAD (eV/A)  : 1.392e-01
 LARGEST GRAD (eV/A)  : 6.492e-01
 LARGEST GRAD (eV/A)  : 6.228e-01
 LARGEST GRAD (eV/A)  : 5.523e-01
 LARGEST GRAD (eV/A)  : 4.431e-01
 LARGEST GRAD (eV/A)  : 2.196e-01
 LARGEST GRAD (eV/A)  : 2.087e-01
 LARGEST GRAD (eV/A)  : 1.248e-01
 LARGEST GRAD (eV/A)  : 1.240e-01
 LARGEST GRAD (eV/A)  : 1.231e-01
 LARGEST GRAD (eV/A)  : 1.093e-01
 LARGEST GRAD (eV/A)  : 1.735e-01
 LARGEST GRAD (eV/A)  : 1.058e-01
 LARGEST GRAD (eV/A)  : 2.692e-01
 LARGEST GRAD (eV/A)  : 1.859e-01
 LARGEST GRAD (eV/A)  : 4.541e-01
 LARGEST GRAD (eV/A)  : 1.165e-01
 LARGEST GRAD (eV/A)  : 5.150e-01
 LARGEST GRAD (eV/A)  : 1.267e-01
 LARGEST GRAD (eV/A)  : 2.376e-01
 LARGEST GRAD (eV/A)  : 2.397e-01
 LARGEST GRAD (eV/A)  : 2.518e-01
 LARGEST GRAD (eV/A)  : 2.053e-01
 LARGEST GRAD (eV/A)  : 5.192e-01
 LARGEST GRAD (eV/A)  : 5.101e-01
 LARGEST GRAD (eV/A)  : 4.684e-01
 LARGEST GRAD (eV/A)  : 3.124e-01
 LARGEST GRAD (eV/A)  : 2.034e-01
 LARGEST GRAD (eV/A)  : 1.742e-01
 LARGEST GRAD (eV/A)  : 1.420e-01
 LARGEST GRAD (eV/A)  : 9.183e-02
 LARGEST GRAD (eV/A)  : 2.199e-01
 LARGEST GRAD (eV/A)  : 1.050e-01
 LARGEST GRAD (eV/A)  : 4.196e-01
 LARGEST GRAD (eV/A)  : 3.955e-01
 LARGEST GRAD (eV/A)  : 3.732e-01
 LARGEST GRAD (eV/A)  : 2.661e-01
 LARGEST GRAD (eV/A)  : 1.190e-01
 LARGEST GRAD (eV/A)  : 1.170e-01
 LARGEST GRAD (eV/A)  : 8.109e-02
 LARGEST GRAD (eV/A)  : 8.653e-02
 LARGEST GRAD (eV/A)  : 7.796e-02
 LARGEST GRAD (eV/A)  : 8.355e-02
 LARGEST GRAD (eV/A)  : 7.857e-02
 LARGEST GRAD (eV/A)  : 8.052e-02
 LARGEST GRAD (eV/A)  : 8.736e-02
 LARGEST GRAD (eV/A)  : 8.405e-02
 LARGEST GRAD (eV/A)  : 1.382e-01
 LARGEST GRAD (eV/A)  : 1.908e-01
 LARGEST GRAD (eV/A)  : 5.281e-01
 LARGEST GRAD (eV/A)  : 1.326e-01
 LARGEST GRAD (eV/A)  : 1.707e-01
 LARGEST GRAD (eV/A)  : 1.672e-01
 LARGEST GRAD (eV/A)  : 1.543e-01
 LARGEST GRAD (eV/A)  : 1.231e-01
 LARGEST GRAD (eV/A)  : 3.850e-01
 LARGEST GRAD (eV/A)  : 3.736e-01
 LARGEST GRAD (eV/A)  : 3.912e-01
 LARGEST GRAD (eV/A)  : 3.302e-01
 LARGEST GRAD (eV/A)  : 1.751e-01
 LARGEST GRAD (eV/A)  : 1.740e-01
 LARGEST GRAD (eV/A)  : 1.266e-01
 LARGEST GRAD (eV/A)  : 1.124e-01
 LARGEST GRAD (eV/A)  : 9.376e-02
 LARGEST GRAD (eV/A)  : 9.862e-02
 LARGEST GRAD (eV/A)  : 6.588e-02
 LARGEST GRAD (eV/A)  : 7.130e-02
 LARGEST GRAD (eV/A)  : 6.827e-02
 LARGEST GRAD (eV/A)  : 7.439e-02
 LARGEST GRAD (eV/A)  : 6.740e-02
 LARGEST GRAD (eV/A)  : 6.753e-02
 LARGEST GRAD (eV/A)  : 7.536e-02
 LARGEST GRAD (eV/A)  : 7.064e-02
 LARGEST GRAD (eV/A)  : 9.137e-02
 LARGEST GRAD (eV/A)  : 5.862e-02
 LARGEST GRAD (eV/A)  : 1.492e-01
 LARGEST GRAD (eV/A)  : 1.357e-01
 LARGEST GRAD (eV/A)  : 1.784e-01
 LARGEST GRAD (eV/A)  : 1.781e-01
 LARGEST GRAD (eV/A)  : 1.644e-01
 LARGEST GRAD (eV/A)  : 6.229e-02
 LARGEST GRAD (eV/A)  : 9.653e-02
 LARGEST GRAD (eV/A)  : 9.853e-02
 LARGEST GRAD (eV/A)  : 1.096e-01
 LARGEST GRAD (eV/A)  : 7.053e-02
 LARGEST GRAD (eV/A)  : 1.634e-01
 LARGEST GRAD (eV/A)  : 8.365e-02
 LARGEST GRAD (eV/A)  : 1.259e-01
 LARGEST GRAD (eV/A)  : 1.098e-01
 LARGEST GRAD (eV/A)  : 7.137e-02
 LARGEST GRAD (eV/A)  : 4.693e-02

Which is also not normal

smear_method=gau, sigma=0.001

kirk0830 commented 9 months ago

link #3002

QuantumMisaka commented 9 months ago

NOTICE: ASE have the different max force definition from ABACUS. in ABACUS, the largest gradient is defined as the max number in all force result along x or y or z direction, while ASE calculating max force by first calculating the vector sum of force than find max force. so the largest gradient of ASE-ABACUS and ABACUS itself will have difference even the force of ASE is read from ABACUS

QuantumMisaka commented 9 months ago

This Pt111 system will have magnetic moment, I tried running y-vac Pt111 model with symmetry 0 and initial magnetic moment set to 1.0 for all Pt atoms, the calculating result showed that the magnetic moment is exists:

----------------------------------------------------------------
TOTAL-STRESS (KBAR)                                           
----------------------------------------------------------------
      -98.3623345965         0.0001050145         0.0000608537
        0.0001050145       -50.0163785364        -0.0001330044
        0.0000608537        -0.0001330044       -98.3620829441
----------------------------------------------------------------
 ETOT DIFF (eV)       : -3.355e-05
 LARGEST GRAD (eV/A)  : 3.910e-01
 BFGS TRUST (Bohr)    : 8.140e-05
 -------------------------------------------
 STEP OF ION RELAXATION : 132
 -------------------------------------------
 DONE(1.843e+04  SEC) : INIT SCF
 ITER   TMAG      AMAG      ETOT(eV)       EDIFF(eV)      DRHO       TIME(s)    
 GE1    2.21e+00  2.32e+00  -2.112467e+05  0.000000e+00   1.909e-05  1.233e+01  
 GE2    2.21e+00  2.33e+00  -2.112467e+05  -3.714867e-06  1.004e-05  9.590e+00  
 GE3    2.21e+00  2.33e+00  -2.112467e+05  -3.824899e-07  8.737e-06  9.387e+00  
 GE4    2.21e+00  2.33e+00  -2.112467e+05  -1.685381e-07  5.369e-06  9.158e+00  
 GE5    2.21e+00  2.33e+00  -2.112467e+05  -3.158666e-07  2.766e-06  9.125e+00  
 GE6    2.21e+00  2.33e+00  -2.112467e+05  -2.291227e-07  2.190e-06  8.377e+00  
 GE7    2.21e+00  2.33e+00  -2.112467e+05  -1.142545e-06  1.910e-06  8.586e+00  
 GE8    2.21e+00  2.33e+00  -2.112467e+05  2.156917e-06   8.710e-07  9.238e+00  
----------------------------------------------------------------
TOTAL-STRESS (KBAR)                                           
----------------------------------------------------------------
      -98.3881315119         0.0001048862         0.0000611617
        0.0001048862       -50.0403201533        -0.0001327824
        0.0000611617        -0.0001327824       -98.3878795368
----------------------------------------------------------------
 ETOT DIFF (eV)       : -2.095e-05
 LARGEST GRAD (eV/A)  : 4.103e-01
 BFGS TRUST (Bohr)    : 8.954e-05
 -------------------------------------------
 STEP OF ION RELAXATION : 133
 -------------------------------------------
 DONE(1.853e+04  SEC) : INIT SCF
 ITER   TMAG      AMAG      ETOT(eV)       EDIFF(eV)      DRHO       TIME(s)    
 GE1    2.22e+00  2.33e+00  -2.112467e+05  0.000000e+00   2.291e-05  1.142e+01  
 GE2    2.21e+00  2.32e+00  -2.112467e+05  -3.432881e-06  1.725e-05  8.615e+00  
 GE3    2.21e+00  2.33e+00  -2.112467e+05  -2.289817e-06  1.420e-05  8.909e+00  
 GE4    2.21e+00  2.33e+00  -2.112467e+05  1.416042e-06   6.434e-06  9.230e+00  
 GE5    2.21e+00  2.33e+00  -2.112467e+05  4.042687e-07   3.154e-06  9.228e+00

But the relax is still slow, which still stuck to some point:

(base) [liuzq@amd41 nosymm+yvac+mag]$ grep GRAD abacus.log 
 LARGEST GRAD (eV/A)  : 6.743e-01
 LARGEST GRAD (eV/A)  : 4.041e-01
 LARGEST GRAD (eV/A)  : 2.031e+00
 LARGEST GRAD (eV/A)  : 4.058e-01
 LARGEST GRAD (eV/A)  : 4.047e-01
 LARGEST GRAD (eV/A)  : 4.043e-01
 LARGEST GRAD (eV/A)  : 4.039e-01
 LARGEST GRAD (eV/A)  : 4.059e-01
 LARGEST GRAD (eV/A)  : 4.038e-01
 LARGEST GRAD (eV/A)  : 4.076e-01
 LARGEST GRAD (eV/A)  : 4.090e-01
 LARGEST GRAD (eV/A)  : 4.068e-01
 LARGEST GRAD (eV/A)  : 4.032e-01
 LARGEST GRAD (eV/A)  : 4.002e-01
 LARGEST GRAD (eV/A)  : 3.994e-01
 LARGEST GRAD (eV/A)  : 4.032e-01
 LARGEST GRAD (eV/A)  : 4.104e-01
 LARGEST GRAD (eV/A)  : 4.037e-01
 LARGEST GRAD (eV/A)  : 4.009e-01
 LARGEST GRAD (eV/A)  : 3.985e-01
 LARGEST GRAD (eV/A)  : 3.996e-01
 LARGEST GRAD (eV/A)  : 3.976e-01
 LARGEST GRAD (eV/A)  : 3.996e-01
 LARGEST GRAD (eV/A)  : 3.970e-01
 LARGEST GRAD (eV/A)  : 3.995e-01
 LARGEST GRAD (eV/A)  : 3.964e-01
 LARGEST GRAD (eV/A)  : 4.006e-01
 LARGEST GRAD (eV/A)  : 3.996e-01
 LARGEST GRAD (eV/A)  : 4.036e-01
 LARGEST GRAD (eV/A)  : 3.958e-01
 LARGEST GRAD (eV/A)  : 4.038e-01
 LARGEST GRAD (eV/A)  : 3.968e-01
 LARGEST GRAD (eV/A)  : 4.100e-01
 LARGEST GRAD (eV/A)  : 3.951e-01
 LARGEST GRAD (eV/A)  : 4.177e-01
 LARGEST GRAD (eV/A)  : 3.954e-01
 LARGEST GRAD (eV/A)  : 4.191e-01
 LARGEST GRAD (eV/A)  : 3.943e-01
 LARGEST GRAD (eV/A)  : 4.200e-01
 LARGEST GRAD (eV/A)  : 3.916e-01
 LARGEST GRAD (eV/A)  : 4.160e-01
 LARGEST GRAD (eV/A)  : 3.909e-01
 LARGEST GRAD (eV/A)  : 4.148e-01
 LARGEST GRAD (eV/A)  : 3.902e-01
 LARGEST GRAD (eV/A)  : 4.149e-01
 LARGEST GRAD (eV/A)  : 3.899e-01
 LARGEST GRAD (eV/A)  : 4.206e-01
 LARGEST GRAD (eV/A)  : 3.921e-01
 LARGEST GRAD (eV/A)  : 4.187e-01
 LARGEST GRAD (eV/A)  : 3.891e-01
 LARGEST GRAD (eV/A)  : 4.168e-01
 LARGEST GRAD (eV/A)  : 3.925e-01
 LARGEST GRAD (eV/A)  : 4.200e-01
 LARGEST GRAD (eV/A)  : 3.903e-01
 LARGEST GRAD (eV/A)  : 4.196e-01
 LARGEST GRAD (eV/A)  : 3.905e-01
 LARGEST GRAD (eV/A)  : 4.127e-01
 LARGEST GRAD (eV/A)  : 4.007e-01
 LARGEST GRAD (eV/A)  : 4.118e-01
 LARGEST GRAD (eV/A)  : 4.003e-01
 LARGEST GRAD (eV/A)  : 4.111e-01
 LARGEST GRAD (eV/A)  : 4.021e-01
 LARGEST GRAD (eV/A)  : 4.095e-01
 LARGEST GRAD (eV/A)  : 4.055e-01
 LARGEST GRAD (eV/A)  : 4.077e-01
 LARGEST GRAD (eV/A)  : 4.061e-01
 LARGEST GRAD (eV/A)  : 4.060e-01
 LARGEST GRAD (eV/A)  : 4.059e-01
 LARGEST GRAD (eV/A)  : 4.064e-01
 LARGEST GRAD (eV/A)  : 4.050e-01
 LARGEST GRAD (eV/A)  : 4.078e-01
 LARGEST GRAD (eV/A)  : 4.055e-01
 LARGEST GRAD (eV/A)  : 4.156e-01
 LARGEST GRAD (eV/A)  : 4.084e-01
 LARGEST GRAD (eV/A)  : 4.098e-01
 LARGEST GRAD (eV/A)  : 4.048e-01
 LARGEST GRAD (eV/A)  : 4.068e-01
 LARGEST GRAD (eV/A)  : 4.044e-01
 LARGEST GRAD (eV/A)  : 4.076e-01
 LARGEST GRAD (eV/A)  : 4.065e-01
 LARGEST GRAD (eV/A)  : 4.068e-01
 LARGEST GRAD (eV/A)  : 4.058e-01
 LARGEST GRAD (eV/A)  : 4.042e-01
 LARGEST GRAD (eV/A)  : 4.092e-01
 LARGEST GRAD (eV/A)  : 4.049e-01
 LARGEST GRAD (eV/A)  : 4.056e-01
 LARGEST GRAD (eV/A)  : 4.039e-01
 LARGEST GRAD (eV/A)  : 4.030e-01
 LARGEST GRAD (eV/A)  : 4.043e-01
 LARGEST GRAD (eV/A)  : 4.042e-01
 LARGEST GRAD (eV/A)  : 3.967e-01
 LARGEST GRAD (eV/A)  : 4.064e-01
 LARGEST GRAD (eV/A)  : 4.033e-01
 LARGEST GRAD (eV/A)  : 4.035e-01
 LARGEST GRAD (eV/A)  : 3.990e-01
 LARGEST GRAD (eV/A)  : 4.002e-01
 LARGEST GRAD (eV/A)  : 3.979e-01
 LARGEST GRAD (eV/A)  : 4.012e-01
 LARGEST GRAD (eV/A)  : 3.990e-01
 LARGEST GRAD (eV/A)  : 3.978e-01
 LARGEST GRAD (eV/A)  : 3.981e-01
 LARGEST GRAD (eV/A)  : 3.977e-01
 LARGEST GRAD (eV/A)  : 3.993e-01
 LARGEST GRAD (eV/A)  : 3.964e-01
 LARGEST GRAD (eV/A)  : 3.955e-01
 LARGEST GRAD (eV/A)  : 3.998e-01
 LARGEST GRAD (eV/A)  : 3.997e-01
 LARGEST GRAD (eV/A)  : 4.015e-01
 LARGEST GRAD (eV/A)  : 4.022e-01
 LARGEST GRAD (eV/A)  : 4.033e-01
 LARGEST GRAD (eV/A)  : 4.034e-01
 LARGEST GRAD (eV/A)  : 4.039e-01
 LARGEST GRAD (eV/A)  : 4.028e-01
 LARGEST GRAD (eV/A)  : 4.023e-01
 LARGEST GRAD (eV/A)  : 4.020e-01
 LARGEST GRAD (eV/A)  : 4.034e-01
 LARGEST GRAD (eV/A)  : 4.022e-01
 LARGEST GRAD (eV/A)  : 4.027e-01
 LARGEST GRAD (eV/A)  : 3.984e-01
 LARGEST GRAD (eV/A)  : 3.989e-01
 LARGEST GRAD (eV/A)  : 3.889e-01
 LARGEST GRAD (eV/A)  : 3.988e-01
 LARGEST GRAD (eV/A)  : 4.018e-01
 LARGEST GRAD (eV/A)  : 4.052e-01
 LARGEST GRAD (eV/A)  : 4.015e-01
 LARGEST GRAD (eV/A)  : 4.048e-01
 LARGEST GRAD (eV/A)  : 3.997e-01
 LARGEST GRAD (eV/A)  : 4.036e-01
 LARGEST GRAD (eV/A)  : 3.929e-01
 LARGEST GRAD (eV/A)  : 4.087e-01
 LARGEST GRAD (eV/A)  : 3.910e-01
 LARGEST GRAD (eV/A)  : 4.103e-01

The algorism should have problem.

QuantumMisaka commented 3 months ago

For abacus 3.6.0, this BFGS problem is solved

 LARGEST GRAD (eV/A)  : 0.465027
 LARGEST GRAD (eV/A)  : 0.350196
 LARGEST GRAD (eV/A)  : 0.426523
 LARGEST GRAD (eV/A)  : 0.310645
 LARGEST GRAD (eV/A)  : 0.438280
 LARGEST GRAD (eV/A)  : 0.293222
 LARGEST GRAD (eV/A)  : 0.064077
 LARGEST GRAD (eV/A)  : 0.066276
 LARGEST GRAD (eV/A)  : 0.023082

deepmodeling / abacus-develop