Geometry Optimization with SS_PCM

Hi, I am encountering failures during excited-state geometry optimization using SS-PCM in Q-Chem. With internally iterative SS-PCM (IISS-PCM), the job stops with an “Imaginary RPA root detected (ω² < 0)” error, suggesting a possible instability of the TDDFT response or SCF reference in the presence of state-specific solvent polarization. With externally iterative SS-PCM (EISS-PCM), the calculation instead fails with “Unable to diagonalize A–B” during the TDDFT step.
I previously did the optimization process with LR-PCM approach and it worked fine. Could anyone help me to solve this issue.
I am sharing both input files:
GEOM OPT with EISS_PCM:
$molecule
0 1
H -3.1280200 -2.1697000 -0.0874000
C -3.6353300 -1.2144000 -0.0450500
C -2.8758100 0.0000000 0.0000000
C -3.6353300 1.2144000 0.0450500
H -3.1280200 2.1697000 0.0874000
C -5.0131700 1.2046000 0.0453200
H -5.5509800 2.1471000 0.0828500
C -5.7240100 0.0000000 0.0000000
H -6.8090200 0.0000000 0.0000000
C -5.0131700 -1.2046000 -0.0453200
H -5.5509800 -2.1471000 -0.0828500
C -1.4540200 0.0000000 0.0000000
C -0.6791600 1.2186200 0.0012500
C 0.6791600 1.2186200 -0.0012500
C 1.4540200 0.0000000 0.0000000
C 0.6791600 -1.2186200 0.0012500
C -0.6791600 -1.2186200 -0.0012500
C 2.8758100 0.0000000 0.0000000
C 3.6353300 1.2144000 -0.0450500
C 5.0131700 1.2046000 -0.0453200
C 5.7240100 0.0000000 0.0000000
C 5.0131700 -1.2046000 0.0453200
C 3.6353300 -1.2144000 0.0450500
H -1.1852900 2.1753700 -0.0018800
H 1.1852900 2.1753700 0.0018800
H 1.1852900 -2.1753700 -0.0018800
H -1.1852900 -2.1753700 0.0018800
H 3.1280200 2.1697000 -0.0874000
H 5.5509800 2.1471000 -0.0828500
H 6.8090200 0.0000000 0.0000000
H 5.5509800 -2.1471000 0.0828500
H 3.1280200 -2.1697000 0.0874000
N -0.0004510 -0.0004180 3.0104710
C 0.8034210 1.1410550 2.5655350
C 0.6008560 2.3835730 3.4236770
C 0.5844940 -1.2696120 2.5705030
C 1.7695850 -1.7085140 3.4216130
C -1.3914640 0.1258320 2.5675710
C -2.3663980 -0.6721130 3.4241840
H 0.6001440 1.3747620 1.5048230
H 1.8570400 0.8560040 2.6171830
H -0.4339030 2.7359610 3.4018780
H 1.2345320 3.1999000 3.0652400
H 0.8638930 2.1749400 4.4643570
H 0.8780320 -1.2177920 1.5064370
H -0.1884470 -2.0393930 2.6361370
H 2.5973050 -0.9956660 3.3776900
H 2.1483220 -2.6743240 3.0749820
H 1.4686620 -1.8115980 4.4677900
H -1.4929900 -0.1645390 1.5060640
H -1.6702560 1.1811320 2.6218040
H -2.1589760 -1.7452520 3.3960210
H -3.3902280 -0.5253640 3.0685110
H -2.3127390 -0.3455950 4.4663230
$end

$rem
BASIS 6-31G(d,p)
METHOD CAM-B3LYP
JOB_TYPE sp
CIS_N_ROOTS 3
CIS_SINGLETS TRUE
CIS_TRIPLETS FALSE
RPA TRUE
CIS_RELAXED_DENSITY TRUE
CIS_MAX_CYCLES 200
SOLVENT_METHOD PCM
SCF_CONVERGENCE 8
SCF_MAX_CYCLES 200
MEM_TOTAL 248000
MEM_STATIC 4000
$end

$pcm
Theory CPCM
StateSpecific External
LinearResponse false
$end

$solvent
Dielectric 37.5
OpticalDielectric 1.8068
$end

@@@

$molecule
read
$end

$rem
BASIS 6-31G(d,p)
METHOD CAM-B3LYP
JOB_TYPE Optimization
GEOM_OPT_MAX_CYCLES 200
CIS_N_ROOTS 3
CIS_SINGLETS TRUE
CIS_TRIPLETS FALSE
RPA TRUE
CIS_STATE_DERIV 1
CIS_RELAXED_DENSITY TRUE
CIS_MAX_CYCLES 200
SOLVENT_METHOD PCM
SCF_CONVERGENCE 8
SCF_MAX_CYCLES 200
MEM_TOTAL 248000
MEM_STATIC 4000
$end

$pcm
Theory CPCM
StateSpecific External
EqSolv 15
EqState 1
EqState_Follow true
LinearResponse false
$end

$solvent
Dielectric 37.5
OpticalDielectric 1.8068
$end
GEOM OPT with IISS_PCM:
$molecule
0 1
H -3.1280200 -2.1697000 -0.0874000
C -3.6353300 -1.2144000 -0.0450500
C -2.8758100 0.0000000 0.0000000
C -3.6353300 1.2144000 0.0450500
H -3.1280200 2.1697000 0.0874000
C -5.0131700 1.2046000 0.0453200
H -5.5509800 2.1471000 0.0828500
C -5.7240100 0.0000000 0.0000000
H -6.8090200 0.0000000 0.0000000
C -5.0131700 -1.2046000 -0.0453200
H -5.5509800 -2.1471000 -0.0828500
C -1.4540200 0.0000000 0.0000000
C -0.6791600 1.2186200 0.0012500
C 0.6791600 1.2186200 -0.0012500
C 1.4540200 0.0000000 0.0000000
C 0.6791600 -1.2186200 0.0012500
C -0.6791600 -1.2186200 -0.0012500
C 2.8758100 0.0000000 0.0000000
C 3.6353300 1.2144000 -0.0450500
C 5.0131700 1.2046000 -0.0453200
C 5.7240100 0.0000000 0.0000000
C 5.0131700 -1.2046000 0.0453200
C 3.6353300 -1.2144000 0.0450500
H -1.1852900 2.1753700 -0.0018800
H 1.1852900 2.1753700 0.0018800
H 1.1852900 -2.1753700 -0.0018800
H -1.1852900 -2.1753700 0.0018800
H 3.1280200 2.1697000 -0.0874000
H 5.5509800 2.1471000 -0.0828500
H 6.8090200 0.0000000 0.0000000
H 5.5509800 -2.1471000 0.0828500
H 3.1280200 -2.1697000 0.0874000
N -0.0004510 -0.0004180 3.0104710
C 0.8034210 1.1410550 2.5655350
C 0.6008560 2.3835730 3.4236770
C 0.5844940 -1.2696120 2.5705030
C 1.7695850 -1.7085140 3.4216130
C -1.3914640 0.1258320 2.5675710
C -2.3663980 -0.6721130 3.4241840
H 0.6001440 1.3747620 1.5048230
H 1.8570400 0.8560040 2.6171830
H -0.4339030 2.7359610 3.4018780
H 1.2345320 3.1999000 3.0652400
H 0.8638930 2.1749400 4.4643570
H 0.8780320 -1.2177920 1.5064370
H -0.1884470 -2.0393930 2.6361370
H 2.5973050 -0.9956660 3.3776900
H 2.1483220 -2.6743240 3.0749820
H 1.4686620 -1.8115980 4.4677900
H -1.4929900 -0.1645390 1.5060640
H -1.6702560 1.1811320 2.6218040
H -2.1589760 -1.7452520 3.3960210
H -3.3902280 -0.5253640 3.0685110
H -2.3127390 -0.3455950 4.4663230
$end

$rem
BASIS 6-31G(d,p)
METHOD CAM-B3LYP
JOB_TYPE Optimization
CIS_N_ROOTS 3
CIS_SINGLETS TRUE
CIS_TRIPLETS FALSE
RPA TRUE
CIS_RELAXED_DENSITY TRUE
CIS_MAX_CYCLES 200
CIS_STATE_DERIV 1
SOLVENT_METHOD PCM
SCF_CONVERGENCE 8
SCF_MAX_CYCLES 200
MEM_TOTAL 248000
MEM_STATIC 4000
$end

$pcm
Theory CPCM
StateSpecific Internal
ChargeSeparation MARCUS
InternalIteration AEM(f)
EqSolv 15
EqState 1
EqState_Follow true
EqS_Conv 4
LinearResponse false
$end

$solvent
Dielectric 37.5
OpticalDielectric 1.8068
$end

Sometimes one hits instabilities upon moving away from the ground-state geometry on an excited-state potential surface, as discussed here:
https://doi.org/10.1016/B978-0-323-91738-4.00005-1
Often there’s nothing to be done, it’s inherent, and can be more of a problem for functionals with a large fraction of exact exchange, including RSH functionals like CAM-B3LYP. I suggest making the Tamm-Dancoff approximation (RPA=FALSE) to avoid this. There’s not much to be gained by using RPA=TRUE.