TDDFT Optimization using M06-2X

Questions
#1

Hello. I’m trying to proceed the geometry optimizations using TD-M06-2X.
However, the calculations look strange compared to the results of other functionals. My rem section is like below.

$rem
JOBTYPE = OPTIMIZATION
EXCHANGE = M062X
BASIS = 6-31G*
GEOM_OPT_MAX_CYCLES = 999
GEOM_OPT_DMAX = 300
GEOM_OPT_TOL_GRADIENT = 150
GEOM_OPT_TOL_DISPLACEMENT = 600
GEOM_OPT_TOL_ENERGY = 50
SCF_CONVERGENCE = 8
SCF_MAX_CYCLES = 1204
SCF_ALGORITHM = DIIS
DFT_D = EMPIRICAL_GRIMME
MEM_TOTAL = 48000
MEM_STATIC = 8000
THRESH = 14
CIS_STATE_DERIV = 1
CIS_N_ROOTS = 3
CIS_SINGLETS = TRUE
CIS_TRIPLETS = FALSE
$end

$empirical_dispersion
s6 1.00
$end

1st, the warning messages occur before starting the calculations.

NAlpha2: 526
NElect   526
Mult     1


!!********************************************.
!!******************Warning*******************.
!!Desired Analytical derivatives not available.
!!Finite difference job might take a long time.
!!********************************************.
!!********************************************.


 1-order derivative to be evaluated numerically with energy values

Checking the input file for inconsistencies... 	...done.
Using s6 = 1.000000

2nd. The evaluations of gradient, displacement, and energy change are omitted. Also, the geometries are seldom updated for each steps.

Before, the optimizations of S0 state using M06-2X-D containing dispersion option showed strange results. So I’m not sure the performance of M06-2X functional. What do you think of?

#2

Firstly, as I mentioned in a previous post,

I think it’s M06-2X with the empirical dispersion correction that’s the problem, not M06-2X in general, as documented here:

I had suggested that you turn off empirical dispersion in the ground state and that suggestion still holds for the excited state.

Secondly, the message you’re seeing is because analytic excited-state gradient is not implemented for this functional. It is available in current version of Q-Chem, what version are you running?

#3

Ok, I understand the combination of M06-2X and dispersion correction may cause the problems.
Q-Chem vesion that installed in our server is 5.2.

#4

Looks like TDDFT analytic gradients for meta-GGA functionals were added in Q-Chem 5.3.

#5

I understand. Thank you for your answers. :slight_smile: