Excited state optimisation and frequency

Questions
#1

I first optimsed the ground state for the water molecule and checked for the imaginary frequencies and there were none and so I moved to excited state calculations. The issue is that I am getting one imaginary frequency for excited state calculation. I have also attached the input

The molecule coordinates which I am using is optimised GS geometry

$molecule
0 1
O -0.0000000000 -0.0000000000 0.1068100006
H -0.7643723408 -0.0000000000 -0.4777331049
H 0.7643723408 0.0000000000 -0.4777331049
$end

$rem
JOBTYPE opt ! Single-point energy calculation
EXCHANGE B3LYP ! Range-separated hybrid functional (good for excited states)
BASIS def2-TZVP ! Triple-zeta basis set for better accuracy
CIS_N_ROOTS 5 ! Compute the first 5 excited states
CIS_SINGLETS TRUE
CIS_TRIPLETS FALSE
CIS_STATE_DERIV 1! Compute singlet excited staites
SCF_CONVERGENCE 8
MEM_TOTAL 220000
$end

@@@

$molecule
read
$end

$rem
JOBTYPE freq ! Single-point energy calculation
EXCHANGE B3LYP ! Range-separated hybrid functional (good for excited states)
BASIS def2-TZVP ! Triple-zeta basis set for better accuracy
CIS_N_ROOTS 5 ! Compute the first 5 excited states
CIS_SINGLETS TRUE
CIS_TRIPLETS FALSE
CIS_STATE_DERIV 1 ! Compute singlet excited states
SCF_CONVERGENCE 8
MEM_TOTAL 220000
$end

#2

Take a look at the single imaginary mode - it is the asymmetric stretch. If you displace about that mode (by providing unequal O-H bond lengths) with, say,

0 1
O
H  1 1.05
H  1 1.15     2 113.622080

as the guess geometry you’ll nudge the optimizer out of the saddle point onto the true minima.

0 1
O
H  1 0.976734
H  1 1.700162 2 101.026610

Providing it a guess geometry with equal bond lengths was steering the optimizer towards the saddle point.