Imaginary frequency

Questions
#1

Hi,
I am trying to calculate the vibrational frequencies of the ground state for 2-methylAnthracene and I couldn’t get rid of the imaginary frequency. I tried some commonly used trics such as tightening the convergence criteria, using stability analysis and tweaking the geometry a bit along with the complex normal mode but that didn’t help.
Here is my input file:
$molecule
0 1
C 0.8681005 0.5236808 0.0000397
C -0.2907264 1.3131490 0.0000945
H -0.1941351 2.4002659 0.0001735
C -1.5730942 0.7421661 0.0000498
C 2.1824337 1.0948542 0.0000893
H 2.2716681 2.1828136 0.0002187
C 3.3138750 0.3145431 0.0000260
C 3.1662560 -1.1133953 -0.0000489
H 4.0640240 -1.7336604 -0.0000539
C 1.9303747 -1.7046712 -0.0000956
H 1.8388077 -2.7914802 -0.0001248
C 0.7355414 -0.9147021 -0.0000675
C -0.5435393 -1.4871912 -0.0000997
H -0.6400596 -2.5742783 -0.0001685
C -1.7037944 -0.6975034 -0.0000465
C -2.7635610 1.5378912 0.0001039
H -2.6653469 2.6241624 0.0001757
C -4.0046880 0.9512522 0.0000726
H -4.9020261 1.5702941 0.0001186
C -4.1334131 -0.4708129 -0.0000197
H -5.1272312 -0.9187378 -0.0000419
C -3.0164716 -1.2687908 -0.0000772
H -3.1126252 -2.3552333 -0.0001475
C 4.6984263 0.9113910 -0.0001720
H 4.6419423 2.0353429 0.6390515
H 5.0499419 1.1293476 -1.1342601
H 5.4824402 0.175325 0.6039344
$end

$rem
BASIS = aug-cc-pVDZ
GEOM_OPT_MAX_STEP_SIZE = 999
GEOM_OPT_TOL_DISPLACEMENT = 120
GEOM_OPT_TOL_ENERGY = 1
GEOM_OPT_TOL_GRADIENT = 10
GUI = 2
INTERNAL_STABILITY = 1
JOB_TYPE = Optimization
MEM_STATIC = 2000
MEM_TOTAL = 80000
METHOD = B3LYP
MOLDEN_FORMAT = 1
PURECART = 2222
SCF_CONVERGENCE = 8
SYM_IGNORE = 1
THRESH = 14
$end

@@@

$comment
Freq
$end

$molecule
read
$end

$rem
BASIS = cc-pVDZ
GUI = 2
JOB_TYPE = Frequency
MEM_STATIC = 2000
MEM_TOTAL = 80000
METHOD = B3LYP
MOLDEN_FORMAT = 1
SCF_CONVERGENCE = 8
SYM_IGNORE = 1
$end

Any suggestion is very helpful. Thank you

#2

The INTERNAL_STABILITY=1 keyword will verify the stability of the SCF wavefunction, but not the stability of the structure on PES. If SCF convergence is well behaved, this keyword is unnecessary and will only slow down the computation.

It is normal that sometimes geometry optimization reaches a saddle point as it happens in this case. To continue geometry optimization from the saddle point and toward the true minimum, restart JOBTYPE=OPT with GEOM_OPT_HESSIAN=READ:

$rem
JOBTYPE OPT
...
$end

@@@

$rem
JOBTYPE = FREQ
...
$end

@@@

$rem
JOBTYPE = OPT
GEOM_OPT_HESSIAN = READ
...
$end

@@@

$rem
JOBTYPE = FREQ
...
$end
#3

I ran this job overnight. The imaginary mode corresponds to a methyl rocking that is trying to break Cs symmetry. Even though you have used SYM_IGNORE=TRUE, you’ve landed on a geometry with symmetry and it can happen that the gradient is exactly zero there. Suggest that you manually twist the methyl rotor away from symmetric geometry.

#4

Thank you both for your answer. I tried both suggestions (combined) but unfortunately the imaginary vibration is still there. I started with a twisted methyl but the optimisation put it back to the saddle point geometry.

1
1966×836 59.1 KB
2
21054×836 63.5 KB

#5

If the magnitude of your imaginary frequency is small, then the problem might be due to the quadrature grid. Q-Chem uses SG-1 for B3LYP and this can lead to noise in the derivatives, especially the second derivatives. You could try changing XC_GRID to 2 (for SG-2) or 3 (for SG-3) to improve the grid and see if the problem is fixed.

#6

I am able to reproduce the imaginary frequency problem. It is caused by using different basis sets for optimization and frequencies. After correcting to BASIS=aug-cc-pVDZ in the JOBTYPE=FREQ part of the input, the lowest frequencies look like this:

 Mode:                 1                      2                      3
 Frequency:        72.25                 110.27                 137.67
 Force Cnst:      0.0127                 0.0211                 0.0132
 Red. Mass:       4.1212                 2.9438                 1.1809
 IR Active:          YES                    YES                    YES
 IR Intens:        1.048                  0.252                  0.008
 Raman Active:       YES                    YES                    YES