Ceilidh ... (SOLVED, UPDATED) imaginary frequency along H-bond, how to get rid of it?

Firefly and PC GAMESS-related discussion club

Learn how to ask questions correctly

We are NATO-free zone

(SOLVED, UPDATED) imaginary frequency along H-bond, how to get rid of it?

FINAL UPDATE: The problem (described below after "Dear colleagues,") is solved.
Frankly speaking, it was not very wise (at least) of me to expect that hessian calculation without explicit additional settings (that is, with inplicit nvib=1 by default) would be of adequate quality to give reliable low-energy frequencies in my case. Actually, analytical hessian (which is of adequate accuracy) is realized for SCF-calculations (excluding UHF) only, and not for DFT (and for other procedures which recover electron correlation) - page 5 of FF8.1.0 full manual. This means that hessian for not-SCF calculations is obtained numerically and is of lower quality (with default nvib=1) unless higher accuracy (comparable with analytical calculation, at the cost of doubling calculation time) is requested with nvib=2 keyword in the $force group (see page 190 of FF8.1.0 full manual).

Initially, I didn't take into account (due to lack of knowledge) this difference in the way of hessian calculation and thus its quality, so I used default settings and caught imaginary values which are spurious and disappears after turning on higher hessian accuracy by "$force nvib=2 $end".

In fact, first six frequencies for hydrated nitromethane radical-anion (CH3NO2-H2O.-) were -32.5; 22.33; 18.78; 11.17; 7.4 and 11.77 when were calculated with defaults, and after recalculation of hessian with nvib=2 they are 3.79; 0.72; 0.42; 0.27; 0.33 and 1.61. Similarly, for hydrated trifluoronitromethane radical-anion (CF3NO2-H2O.-) set of six lowest frequensies (-41.55; 14.36; 10.34; 8; 6.82; 9.45) becomes (2.78; 2.12; 0.85; 0.24; 0.24; 0.53).

I hope that this confession excuses posting a problem which could be solved with very diligent and minute manual reading. : -)
And since questions regarding imaginary frequensies are more or less popular, I hope that my experience will be useful. : -)

Dear collegues,

During geometry optimization of a complex of trifluoronitromethane anion-radical and water molecule, I've got a TS-like structure of approx. Cs-symmetry (as it looks). I can't get rid of this im. freq. and ask your kind advice.

Even use of very tight convergence criteria (OPTTOL=0.000001, nconv=8) still left it. Similarly, small random displacement of oxygen and hydrogen atoms does not solve the problem, as well as displacement along imaginary oscillation (by catching most diturbed structure while visualizing this oscillation in Chemcraft, neither with 1-x displacement scale nor with 0.05-x displacement scale). Calculating of starting hessian also doesn't help.

(By the way: if the starting geometry is an optimum and input contains HESS=CALC and HSSEND = T, then THE SAME Hessian is computed twice without actual geometry change. Perphaps, it could be fixed, say, skip the second calculation if no atom displacement is made and display corresponding warning? For HESS=READ when the starting geometry occurs an optimum final hessian calculation is unnecessary too)

The starting geometry in the below input is optimum for U-B3LYP (and UHF too, if I'm not confusing my attempts) at conventional convergence criteria (OPTTOL=0.00001, nconv by default), tightening leads to very little energy gain without discernible geometry changes and still lefts -41.8 cm^-1

Thanks in advance

UPDATE: the same problem holds for less computer-demanding non-fluorinated analogue, which input follows initially given input of fluorinated complex

typical input for CF3NO2-H2O(.-) is as follows:

$BASIS GBASIS=N31 NGAUSS=6 NDFUNC=1 NPFUNC=1 DIFFSP=.T. DIFFS=.T. $END
$CONTRL
SCFTYP=UHF RUNTYP=OPTIMIZE maxit=3000 ICHARG=-1 MULT=2 DFTTYP=B3LYP
$END
$CONTRL INTTYP=HONDO, ICUT=11, ITOL=30 $END
$SCF FDIFF=.FALSE. DIIS = .T. SOSCF = .F. fshift=0.1 nconv=8 $END
$STATPT OPTTOL=0.000001 NSTEP=100 HSSEND=.TRUE. $END

$DATA
CF3NO2 radical-anion + AQUA U-B3LYP (from TS-like U-B3LYP-optmm)
C1
C 6.0 -5.6272534607 3.1026618478 0.3181518574
N 7.0 -4.3567243168 2.5364655946 -0.0536497344
O 8.0 -4.0473454257 2.6944770883 -1.3163317809
O 8.0 -3.4764618687 2.5084867643 0.9153946635
F 9.0 -6.0467149717 2.6341897226 1.5149610997
F 9.0 -6.5858757825 2.8046753514 -0.5874339103
F 9.0 -5.6111571750 4.4692610548 0.4249064004
O 8.0 -1.2053156062 2.4142654430 -0.8731233731
H 1.0 -1.6425223181 2.4008540086 -0.0007881277
H 1.0 -2.0097941315 2.5185954046 -1.4158506646
$END

input for U-B3LYP of CH3NO2-H2O(.-)
(NOTE!!! $HESS GROUP IS OMITTED; I CANT UPLOAD FILE AS ATTACHMENT, SO I OMIT IT TO SAVE SPACE)

$BASIS GBASIS=N31 NGAUSS=6 NDFUNC=1 NPFUNC=1 DIFFSP=.T. DIFFS=.T. $END
$CONTRL
SCFTYP=UHF RUNTYP=OPTIMIZE maxit=3000 ICHARG=-1 MULT=2 DFTTYP=B3LYP
$END
$CONTRL INTTYP=HONDO, ICUT=11, ITOL=30 $END
$SCF FDIFF=.FALSE. DIIS = .T. SOSCF = .F. fshift=0.1 $END
$STATPT HESS=READ OPTTOL=0.00001 NSTEP=100 HSSEND=.TRUE. $END

$DATA
nitromethane radical-anion - AQUA U-B3LYP (from UHF opt)
C1
CARBON 6.0 -5.690238467 3.123159861 0.342974456
NITROGEN 7.0 -4.408793571 2.571382467 -0.033886602
OXYGEN 8.0 -4.092995051 2.708372270 -1.274641507
OXYGEN 8.0 -3.524399211 2.542580274 0.901523600
HYDROGEN 1.0 -5.972164682 2.729477283 1.311039893
HYDROGEN 1.0 -6.421906901 2.860638826 -0.410260474
HYDROGEN 1.0 -5.611620885 4.207395957 0.405052132
OXYGEN 8.0 -1.232516766 2.415027424 -0.875638547
HYDROGEN 1.0 -1.650408000 2.409399347 -0.020040167
HYDROGEN 1.0 -2.013679661 2.515332453 -1.410831048
$END

[ This message was edited on Mon Aug 24 '15 at 3:21pm by the author ]

Mon Aug 24 '15 3:21pm

This message read 831 times