Facing large number of imaginary frequencies in LRC-wPBE functional

Questions
#1

Hi, I recently performed geometry optimization followed by frequency analysis using the LRC-ωPBE functional, but I am obtaining multiple imaginary frequencies (around 20-25). However, when I carried out the same procedure using CAM-B3LYP, the optimization converged successfully with no imaginary frequencies. I’m wondering if this issue is specific to LRC-ωPBE or if there are particular settings or strategies I should adjust? I am sharing my input file here.
$molecule
0 1
H 3.16448000 2.15377000 -0.15950000
C 3.67121000 1.19825000 -0.08353000
C 2.90948000 0.00000000 0.00000000
C 3.67121000 -1.19825000 0.08353000
H 3.16448000 -2.15377000 0.15950000
C 5.05577000 -1.19349000 0.08271000
H 5.58532000 -2.14066000 0.15117000
C 5.77640000 0.00000000 0.00000000
H 6.86140000 0.00000000 0.00000000
C 5.05577000 1.19349000 -0.08271000
H 5.58532000 2.14066000 -0.15117000
C 1.46255000 0.00000000 0.00000000
C 0.68551000 -1.20011000 0.00096000
C -0.68551000 -1.20011000 -0.00096000
C -1.46255000 -0.00000000 0.00000000
C -0.68551000 1.20011000 0.00096000
C 0.68551000 1.20011000 -0.00096000
C -2.90948000 -0.00000000 0.00000000
C -3.67121000 -1.19825000 -0.08353000
C -5.05577000 -1.19349000 -0.08271000
C -5.77640000 -0.00000000 0.00000000
C -5.05577000 1.19349000 0.08271000
C -3.67121000 1.19825000 0.08353000
H 1.18512000 -2.16375000 -0.00451000
H -1.18512000 -2.16375000 0.00451000
H -1.18512000 2.16375000 -0.00451000
H 1.18512000 2.16375000 0.00451000
H -3.16448000 -2.15377000 -0.15950000
H -5.58532000 -2.14066000 -0.15117000
H -6.86140000 -0.00000000 0.00000000
H -5.58532000 2.14066000 0.15117000
H -3.16448000 2.15377000 0.15950000
N -0.00171900 -0.00297200 2.93517500
C -1.43021400 -0.25815800 2.92412900
C -1.90959700 -0.91639900 4.22182800
C 0.49049800 1.36208300 2.92114100
C 0.15721100 2.10942700 4.21641500
C 0.93406700 -1.11227700 2.92837900
C 1.75956700 -1.18236500 4.21646500
H -1.61805800 -0.92629300 2.07592200
H -1.94150900 0.68806900 2.74915400
H -1.42562600 -1.88045700 4.38799700
H -2.98462200 -1.08351300 4.13331900
H -1.72474500 -0.26849100 5.08010500
H 0.00614000 1.85631300 2.07122100
H 1.56560500 1.33243200 2.74642500
H -0.91977800 2.15959500 4.38530600
H 0.53812600 3.12772800 4.12122800
H 0.63388200 1.63495700 5.07562700
H 1.59611300 -0.95039000 2.06986100
H 0.36835600 -2.02991900 2.77043500
H 2.35388000 -0.27909700 4.36473500
H 2.44078800 -2.03046500 4.12840700
H 1.11708400 -1.33672600 5.08481000
$end

$rem
JOB_TYPE Optimization
METHOD LRC-wPBE
OMEGA 300
BASIS 6-31G(d,p)
SCF_CONVERGENCE 8
SCF_MAX_CYCLES 200
GEOM_OPT_MAX_CYCLES 200
UNRESTRICTED TRUE
SYMMETRY FALSE
SYM_IGNORE TRUE
GUI 2
CDFT TRUE
CDFT_PRINT TRUE
CDFT_BECKE_POP TRUE
CDFT_THRESH 8
BECKE_SHIFT UNSHIFTED
SOLVENT_METHOD PCM
MEM_TOTAL 248000
MEM_STATIC 4000
$end

$pcm
THEORY CPCM
method swig
radii bondi
solver inversion
$end

$cdft
2
-1 1 32
1 33 54
$end

@@@

$molecule
read
$end

$rem
JOB_TYPE Freq
METHOD LRC-wPBE
OMEGA 300
BASIS 6-31G(d,p)
SCF_CONVERGENCE 8
UNRESTRICTED TRUE
SYMMETRY FALSE
SYM_IGNORE TRUE
SOLVENT_METHOD PCM
CDFT TRUE
CDFT_PRINT TRUE
CDFT_BECKE_POP TRUE
CDFT_THRESH 8
BECKE_SHIFT UNSHIFTED
MEM_TOTAL 248000
MEM_STATIC 4000
$end

$pcm
THEORY CPCM
method swig
radii bondi
solver inversion
$end

$cdft
2
-1 1 32
1 33 54
$end

#2

This is constrained DFT, so it’s not guaranteed that solution is a minimum, is it? Are you obtaining what looks to be the same SCF solution with both methods?

#3

I checked the charge distribution for both methods, and in both cases the CDFT constraints are satisfied. However, the final SCF energies differ significantly. With CAM-B3LYP, the energy is −986.17821808 Hartree and no imaginary frequencies are observed, whereas with LRC-ωPBE, the energy is −985.86775647 Hartree and the structure shows 20–25 imaginary frequencies.

#4

You can’t rely on those numbers because different functionals give different energies. For example, that molecule in the gas phase, normal SCF, gives the following:

    8    -986.1753858447      2.44e-06  Convergence criterion met (CAM-B3LYP)
    7    -985.8605171264      7.65e-06  Convergence criterion met (LRC-wPBE)

So when I say “did both reach the same SCF solution”, you need to examine that in a more qualitative way (orbitals, energy levels, etc.) Or just try to perturb the LRC-wPBE calculation away from the saddle point.