Good day people
I'm trying to optimise a biphenyl with substituents on using DFT rB3LYP aug-cc-pVDZ.
I optimised biphenyl on its own with no problems, but now I added substituents on it and now optimisation fails due to convergence failure.
.log files ends in the following
" >>>>>>>>>> Convergence criterion not met.
SCF Done: E(RB3LYP) = -620.626597935 A.U. after 513 cycles
NFock=512 Conv=0.17D-05 -V/T= 2.0061
KE= 6.168849716931D+02 PE=-3.423301472620D+03 EE= 1.192804461235D+03
Convergence failure -- run terminated.
Error termination via Lnk1e in /gaussian09/g09/l502.exe at Fri Mar 8 00:12:47 2019.
Job cpu time: 1 days 21 hours 3 minutes 44.5 seconds.
File lengths (MBytes): RWF= 147 Int= 0 D2E= 0 Chk= 5 Scr= 1
"
By default it had 128 SCF cycles in which I then increased it to 512 but it got that far without converging
Other than increasing SCF cycles what could I try?
@Jared thank you. I just tried scf=(qc,maxcycle=512) which didn't work so now trying xqc
@Dariusz thank you for your reply. When you preoptimise what do you mean exactly? Because surely if DFT isn't working then HF won't be able to optimise either?
First, reason of nonconvergence has to be located.
As I understood SCF cycles converges and optimisation don't. Then changing of electronic converger will have no effect.
Does final geometry looks reasonable? If YES:
Is energy oscillating? Did oscillation starts after 102-108 optimisation steps? Does restart from final geometry or from lowest energy geometry helps? One can try to make calculation in Cartesian coordinates instead of default internals. Is it calculation with continuum solvent? Is dispersion correction used? If answer to all questions of this paragraph NO - than try to optimise with nonaugumented set.
" Because surely if DFT isn't working then HF won't be able to optimise either? " - No, its absolutely independent, HF may or may not converges better than DFT. But it is unreasonable to preoptimise with HF, while difference in price is to small. Traditional organics can be nicely preoptimised with forcefields. Sometimes semiempirics might be used. DFT with very small basis set.
- Badges
- Science topic
- Similar topics
- Filing