search

deepmodeling / abacus-develop

An electronic structure package based on either plane wave basis or numerical atomic orbitals.
http://abacus.ustc.edu.cn
GNU Lesser General Public License v3.0
169 stars 129 forks source link

Bug：PBE0 for Fe cannot get normal result #5036

Open QuantumMisaka opened 1 month ago

QuantumMisaka commented 1 month ago

Describe the bug

While do EXX SCF calculation in Fe2 bulk (target: ferromagnetic system), HSE can give normal results with normal energy and magnetic moments

 Updating EXX and rerun SCF     4.928e+02 (s)
 GE0      5.50e+00   5.93e+00  -6.43742199e+03  -1.53905484e-10   1.0239e-07 502.35
 ITER      TMAG       AMAG        ETOT/eV          EDIFF/eV         DRHO     TIME/s
 GE1      5.50e+00   5.93e+00  -6.43742199e+03   0.00000000e+00   3.2955e-06   5.69
 GE2      5.50e+00   5.93e+00  -6.43742199e+03   1.59473923e-09   2.2303e-06   8.50
 GE3      5.50e+00   5.93e+00  -6.43742199e+03  -3.34106377e-10   1.4013e-06   6.08
 GE4      5.50e+00   5.93e+00  -6.43742199e+03  -2.24284373e-10   1.3880e-06   6.57
EDIFF/eV (outer loop): 7.68599345e-08 
 Updating EXX and rerun SCF     4.925e+02 (s)
 GE0      5.50e+00   5.93e+00  -6.43742199e+03  -1.36968147e-09   4.0459e-08 499.92
 ITER      TMAG       AMAG        ETOT/eV          EDIFF/eV         DRHO     TIME/s
 GE1      5.50e+00   5.93e+00  -6.43742199e+03   0.00000000e+00   1.4978e-06  13.42
EDIFF/eV (outer loop): 1.53696667e-08 
 Updating EXX and rerun SCF     4.920e+02 (s)
 GE0      5.50e+00   5.93e+00  -6.43742199e+03   3.51894448e-10   9.1585e-07 498.27
EDIFF/eV (outer loop): 1.85537314e-09 
 ITER      TMAG       AMAG        ETOT/eV          EDIFF/eV         DRHO     TIME/s
 GE1      5.50e+00   5.93e+00  -6.43742199e+03   0.00000000e+00   8.0938e-07  12.73
 >> Leave SCF iteration.
 * * * * * *
 ETOT DIFF (eV)       : 0
 LARGEST GRAD (eV/A)  : 1.14e-05

And magnetic moments for each Fe atom is 2.7484

But PBE0 cannot give normal results, sometimes, PBE0 will lead to high magnetic moments such as 7.9986, which also lead to unconvergence of SCF

 GE298    1.60e+01   1.60e+01  -1.11290220e+05  -5.83943700e-07   5.4251e-05   6.53
 GE299    1.60e+01   1.60e+01  -1.11290220e+05   1.09401277e-07   5.3031e-05   6.61
 GE300    1.60e+01   1.60e+01  -1.11290220e+05   1.62709805e-07   4.7227e-05   6.37
 >> Leave SCF iteration.
 * * * * * *
 !! CONVERGENCE HAS NOT BEEN ACHIEVED !!

and in other situation (only the initial magnetic moments is different, but they are all positive), the TMAG will disappear and lead to a weird anti-ferromagnetic SCF , this SCF cannot converge as well

 GE36     8.54e-05   3.31e+00  -6.43996556e+03  -9.10980479e-07   1.2279e-05   4.11
 GE37     1.09e-04   3.31e+00  -6.43996556e+03  -3.64578116e-08   4.2449e-06   4.20
 GE38     2.60e-05   3.31e+00  -6.43996556e+03  -6.51198077e-10   1.1083e-06   4.13
 Entering 2nd SCF, where EXX is updated
 GE0     -2.11e-06   3.31e+00  -6.43996556e+03   7.81128334e-10   6.9807e-07   4.28
 ITER      TMAG       AMAG        ETOT/eV          EDIFF/eV         DRHO     TIME/s
 GE1      8.86e-05   4.43e+00  -6.43595817e+03   0.00000000e+00   8.3515e-02 287.37
 GE2      1.86e-04   4.34e-01  -8.04937942e+04  -7.40578361e+04   3.9177e-01 291.12
 GE3     -1.62e-03   2.09e+00  -2.28595946e+05  -1.48102151e+05   4.3919e-01 296.31
 GE4     -1.55e-02   2.13e+00  -2.37775006e+05  -9.17906053e+03   4.3337e-01 297.33
 GE5     -2.96e-02   2.56e+00  -2.60465350e+05  -2.26903444e+04   4.5255e-01 298.15
 GE6     -7.25e-02   2.40e+00  -2.46636423e+05   1.38289276e+04   4.4750e-01 297.75
 GE7     -2.00e-01   2.43e+00  -2.61180091e+05  -1.45436687e+04   4.5008e-01 297.16
 GE8     -4.75e-01   2.54e+00  -2.43117418e+05   1.80626735e+04   4.5364e-01 295.24
 GE9     -5.11e-01   2.55e+00  -2.50808002e+05  -7.69058423e+03   4.5426e-01 298.66

There is also another situation where AMAG and TMAG both -> 0

 ITER      TMAG       AMAG        ETOT/eV          EDIFF/eV         DRHO     TIME/s
 GE1      1.84e-01   2.73e-01  -3.03967808e+05   0.00000000e+00   1.0364e-02   7.52
 GE2      1.81e-01   2.74e-01  -3.03967814e+05  -6.37141113e-03   6.1097e-03  13.64
 GE3      1.83e-01   2.74e-01  -3.03967814e+05  -1.65154177e-04   2.5188e-03   7.62
 GE4      1.83e-01   2.74e-01  -3.03967814e+05  -3.56292525e-05   1.9311e-04   6.19
 GE5      1.83e-01   2.74e-01  -3.03967814e+05   9.15352477e-07   4.9791e-05   7.31
 GE6      1.83e-01   2.74e-01  -3.03967814e+05  -2.92033722e-08   1.6763e-05   7.41
 GE7      1.83e-01   2.74e-01  -3.03967814e+05   1.79180012e-08   3.7352e-06   6.64
 GE8      1.83e-01   2.74e-01  -3.03967814e+05   2.57385653e-09   2.1483e-06   8.34
EDIFF/eV (outer loop): 7.61560651e+03 
 Updating EXX and rerun SCF     7.105e+02 (s)
 GE0      1.83e-01   2.74e-01  -3.03967814e+05   8.41453097e-10   2.2225e-07 717.68
 ITER      TMAG       AMAG        ETOT/eV          EDIFF/eV         DRHO     TIME/s
 GE1      4.23e-01   4.26e-01  -2.96352197e+05   0.00000000e+00   1.0259e-02  11.24
 GE2      4.30e-01   4.33e-01  -2.96352204e+05  -6.55905947e-03   5.9487e-03   5.69
 GE3      4.28e-01   4.31e-01  -2.96352204e+05  -1.41964819e-04   2.5128e-03   8.75
 GE4      4.28e-01   4.30e-01  -2.96352204e+05  -3.14751965e-05   1.7930e-04   9.12
 GE5      4.28e-01   4.31e-01  -2.96352204e+05   8.45511870e-07   5.1951e-05   9.47
 GE6      4.28e-01   4.31e-01  -2.96352204e+05  -2.50555034e-07   1.1991e-05   6.36
 GE7      4.28e-01   4.31e-01  -2.96352204e+05  -1.45521888e-08   3.7525e-06   6.66
 GE8      4.28e-01   4.31e-01  -2.96352204e+05  -3.46480687e-09   1.4731e-06   9.07

Expected behavior

PBE0 for Fe systems can run normally

To Reproduce

Fe-PBE0-weird.tar.gz

Environment

ABACUS : 3.7.4 Commit: 93badfa87 (Wed Aug 28 10:18:22 2024 +0800) LibRI: 0.2.0 LibComm: 0.1.1 Hardware: Intel 8358 (WM2) Software: Intel Toolchain with OneAPI 2023.0 LibXC: 6.2.2 ELPA: 2024.03.001

Additional Context

No response

Task list for Issue attackers (only for developers)

QuantumMisaka commented 1 month ago

Added: SCAN0 have the likely problem