Negative frequency in excited state TDDFT calculation

Questions
#1

Hello,
I’ve been trying to optimize the S1 structure of [2,2]-paracyclophane using TD-DFT. For some unknown reason I keep getting a a single negative frequency in the normal mode analysis (indicating a saddle point). Iv’e tried everything, starting from ignoring symmetry in calculations, distorting the molecule along this negative mode, and reloading the hessian from the previous calculation for further iterations, but nothing seems to work. I also tried to change the XC grid.

It is worth saying that when using a simple 6-31G(d) basis, the molecule optimizes into a local minimum, resulting with all positive frequencies. The problem begins, when I use a the 6-31+G(d) basis.

Here is my input file:

$molecule
0 1
C -1.40512 -1.30536 0.14389
C -0.82457 -1.45948 -1.14380
H -1.46326 -1.47782 -2.02917
C 0.57679 -1.43544 -1.27970
H 1.02762 -1.40777 -2.27299
C 1.40514 -1.30533 -0.14389
C 0.82460 -1.45947 1.14380
H 1.46329 -1.47779 2.02917
C -0.57676 -1.43546 1.27970
H -1.02759 -1.40780 2.27299
C -2.82381 -0.78677 0.26176
H -3.16214 -0.87195 1.37988
H -3.55773 -1.44342 -0.38083
C 2.82383 -0.78671 -0.26176
H 3.16217 -0.87187 -1.37988
H 3.55775 -1.44335 0.38083
C 2.82381 0.78677 0.26176
H 3.16214 0.87195 1.37987
H 3.55772 1.44342 -0.38083
C 1.40512 1.30536 0.14389
C 0.82457 1.45948 -1.14380
H 1.46326 1.47782 -2.02917
C -0.57679 1.43544 -1.27970
H -1.02762 1.40777 -2.27299
C -1.40514 1.30533 -0.14389
C -0.82460 1.45947 1.14380
H -1.46329 1.47779 2.02917
C 0.57676 1.43545 1.27970
H 1.02759 1.40780 2.27299
C -2.82383 0.78671 -0.26176
H -3.16217 0.87187 -1.37988
H -3.55775 1.44335 0.38083
$end

$rem
JOBTYPE opt
EXCHANGE CAMB3LYP
BASIS 6-31+G(d)
DFT_D D4
CIS_STATE_DERIV 1
CIS_N_ROOTS 5
CIS_SINGLETS true
CIS_TRIPLETS false
RPA false
SYMMETRY false
SYM_IGNORE true
GEOM_OPT_TOL_DISPLACEMENT 10
GEOM_OPT_TOL_ENERGY 10
GEOM_OPT_TOL_GRADIENT 10
THRESH 14
GEOM_OPT_MAX_CYCLES 100
XC_GRID = 000075000302
mem_total 12000
$end

@@@

$molecule
read
$end

$rem
JOBTYPE freq
EXCHANGE CAMB3LYP
BASIS 6-31+G(d)
DFT_D D4
CIS_STATE_DERIV 1
CIS_N_ROOTS 2
CIS_SINGLETS true
CIS_TRIPLETS false
RPA false
mem_total 12000
SYM_IGNORE true
SYMMETRY false
XC_GRID = 000075000302
$end

@@@

$molecule
read
$end

$rem
JOBTYPE opt
GEOM_OPT_HESSIAN read
EXCHANGE CAMB3LYP
BASIS 6-31+G(d)
DFT_D D4
CIS_STATE_DERIV 1
CIS_N_ROOTS 5
CIS_SINGLETS true
CIS_TRIPLETS false
RPA false
SYMMETRY false
SYM_IGNORE true
GEOM_OPT_TOL_DISPLACEMENT 10
GEOM_OPT_TOL_ENERGY 10
GEOM_OPT_TOL_GRADIENT 10
THRESH 14
GEOM_OPT_MAX_CYCLES 100
XC_GRID = 000075000302
mem_total 12000
$end

@@@

$molecule
read
$end

$rem
JOBTYPE freq
EXCHANGE CAMB3LYP
BASIS 6-31+G(d)
DFT_D D4
CIS_STATE_DERIV 1
CIS_N_ROOTS 2
CIS_SINGLETS true
CIS_TRIPLETS false
RPA false
mem_total 12000
SYM_IGNORE true
SYMMETRY false
XC_GRID = 000075000302
$end

Thanks!
Omer.

#2

How large is this imaginary frequency and does it disappear if you turn off D4? What about if you set THRESH=14?

#3

Hi John, thanks for your reply.
I haven’t tried turning off the D4, but why should it affect the optimization process?
I don’t understand what you mean about THRESH? I did set it to 14 in the optimization part. Should I use it also in the frequency analysis?

thanks again!

#4

You should definitely use a consistent value of THRESH for opt + freq. As for D4, I was merely trying to test by process of elimination what could be causing the problem, in case it is some kind of bug. D4 is the newest part of the calculation you’ve put together.

#5

I tried eliminating the D4 part, as well as playing with the thresh component… It all leads to the same result.
What I find interesting is that when I replace the diffuse basis set with the simple 6-31Gd, the ES optimizes to a minimum with no negative frequencies… I’m struggling to understand why :(.