[Wien] Force convergence within SCF cycle

[Oleg Rubel]

Dear WIEN2k users,

would you please comment on convergence of forces during the SCF-run. I
perform the energy minimization with respect to atomic positions for
GaAs:N [64-atom zinc-blend supercell with one As atom replaced by N
(nitrogen)]. After the first SCF cycle I have:

~> grep :FOR002 *scf
GaAsN64_1.scf::FOR002:   2.ATOM        129.498          0.000          0.000        129.498
GaAsN64_1.scf::FOR002:   2.ATOM        162.260          0.000          0.000        162.260
GaAsN64_1.scf::FOR002:   2.ATOM         60.859          0.000          0.000         60.859
GaAsN64_1.scf::FOR002:   2.ATOM         36.923          0.000          0.000         36.923
GaAsN64_1.scf::FOR002:   2.ATOM         19.557          0.000          0.000        -19.557
GaAsN64_1.scf::FOR002:   2.ATOM         37.737          0.000          0.000         37.737
GaAsN64_1.scf::FOR002:   2.ATOM         27.782          0.000          0.000        -27.782
GaAsN64_1.scf::FOR002:   2.ATOM          3.583          0.000          0.000         -3.583
GaAsN64_1.scf::FOR002:   2.ATOM          2.685          0.000          0.000         -2.685
GaAsN64_1.scf::FOR002:   2.ATOM          3.882          0.000          0.000          3.882
GaAsN64_1.scf::FOR002:   2.ATOM          6.754          0.000          0.000          6.754
GaAsN64_1.scf::FOR002:   2.ATOM          7.020          0.000          0.000          7.020
GaAsN64_1.scf::FOR002:   2.ATOM          7.479          0.000          0.000          7.479
GaAsN64_1.scf::FOR002:   2.ATOM         81.647          0.000          0.000         81.647

QUESTION: Do forces converge within the SCF-cycle? The last step
(transition from 7.479 to 81.647) looks suspicious. How does it correspond
to the -fc 1.0 option?

ADDITIONAL INFORMATION: Run command ~>min -i 100 -s 10 -j 'run_lapw -I -i
40 -fc 1.0 '; method PORT; LDA; Spin polarization - NO; Rmt(Ga)=1.92;
Rmt(As)=2.0;  Rmt(N)=1.75; Kmax*Rmt_min = 7; 1 k-point (shifted) (2x2x2);
Lmax = 12 (in *in1c); for the rest I use default values. Such parameters
produce reliable results for binaries (GaAs and GaN) and ensure absence of
ghost-states.

Thank you in advance,
Oleg Rubel