mix=.t. works for both HF and DFT but there is no warranty that
you will indeed obtain a broken-symmetry solution. You can try
several times using different settings for SCF convergers.
Alternatively, you can try to use better guess for broken-symmetry
solutions using e.g. AOMIX program by S. Gorelsky.
Note, with mix=.t. you need to switch symmetry off if HOMO and
LUMO are of different symmetry.
On Sat Dec 21 '13 7:42am, Pavlo Solntsev wrote
>Dear FF users.
>Could you share your experience with broken-symmetry calculation using HF and DFT? I have an example below. If i use DFT i have Sz=0, S^2 = 0 for broken-symmetry singlet. If i use HF i have Sz=0 S^2 = 1.8. Which is reasonable. Does it mean, mix=.t. works only for HF? I know, we can remove switch symmetry completely off (nosym=1), and in this case i got S^2 = 0.8 for DFT which is also acceptable. I confused why we can use full orbital symmetry within HF, but not for DFT.
> $SYSTEM MWORDS=50 $END
> $CONTRL RUNTYP=energy dfttyp=bp86 $END
> $CONTRL SCFTYP=UHF $END
> $CONTRL MAXIT=100 $END
> $CONTRL ICHARG=-2 MULT=1 $END
> $BASIS GBASIS=n31 ngauss=6 $END
> $BASIS ndfunc=1 $END
> $guess mix=.t. $end
>[Cu2F6]^-2 BS singlet
> CU 29.0 1.5298305586 0.0000000000 0.0000000000
> F 9.0 2.7561841734 1.3744839332 0.0000000000
> O 8.0 0.0000000000 1.1934572146 0.0000000000
> H 1.0 0.0000000000 2.1349088145 0.0000000000