[CP2K-user] Negative Hartree energy in SIC (self-interaction corrected) calculation

[Patrick Gono]

Dear CP2K experts,

Running a charged calculation with the self-interaction correction by Mauri
et al. (as defined in CP2K_INPUT / FORCE_EVAL / DFT / SIC) leads to a large
*negative* Hartree energy contribution -- something that doesn't look
physically possible:

==========================================================
  *** SCF run converged in    37 steps ***


  Electronic density on regular grids:      -1054.9999988859
 0.0000011141
  Core density on regular grids:             1055.9999999431
-0.0000000569
  Total charge density on r-space grids:        1.0000010572
  Total charge density g-space grids:           1.0000010572

  Overlap energy of the core charge distribution:
0.00001597916987
  Self energy of the core charge distribution:
 -7659.99497268730556
  Core Hamiltonian energy:
3104.15741917188643
*  Hartree energy:
 -12743.39184120392383*
  Exchange-correlation energy:
-820.98172868636971

  Total energy:
 -18120.21110742654128
==========================================================

After this section of the output file, a warning is printed ( *WARNING :
did not converge in ot_eigensolver* ) and the HOMO-LUMO gap for this spin
channel is nonsense (64 eV). Finally, a *CPASSERT* on *line 528* of the
file *qs_vxc.F* aborts the calculation (see the .out file). I assume this
is the consequence (or at least related to) the large negative Hartree
energy in the first place.

We are running this using CP2K version 4.1, but the same happens with
version 6.1. The input file, initial structure, and the full output file
are attached to this email. The initial structure is sensible and the
system behaves well under regular PBE exchange & correlation (no SIC, all
other settings kept identical), as well as under the hybrid PBE0 functional.

Are we doing something wrong? I would appreciate any and all input.

Yours sincerely,
Patrick Gono
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20190923/1c066dc1/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Ga2O3.inp
Type: application/octet-stream
Size: 2584 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20190923/1c066dc1/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Ga2O3.xyz
Type: chemical/x-xyz
Size: 7855 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20190923/1c066dc1/attachment.xyz>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Ga2O3.out
Type: application/octet-stream
Size: 216173 bytes
Desc: not available
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20190923/1c066dc1/attachment-0001.obj>