I am trying to run NAMD in Q-Chem to study the surface hopping from the S1 state of retinal to its ground state. To do so I have defined the $rem section as follows:

$rem

JOBTYPE aimd

EXCHANGE B3LYP

BASIS 6-31g

MEM_TOTAL 10000

CIS_N_ROOTS 1

SYMMETRY off

SYM_IGNORE true

CIS_SINGLETS true

CIS_TRIPLETS false

PROJ_TRANSROT true

FSSH_LOWESTSURFACE 0

FSSH_NSURFACES 2 ! hop between S0 and S1

FSSH_INITIALSURFACE 1 ! start on S1

AFSSH 1 ! decoherence

CALC_NAC true

AIMD_STEPS 1000 ! Typically more would be used

TIME_STEP 14

AIMD_SHORT_TIME_STEP 2

AIMD_TIME_STEP_CONVERSION 1 ! Do not alter time_step duration

AIMD_PRINT 1

AIMD_INIT_VELOC thermal

AIMD_TEMP 300

AIMD_INTEGRATION vverlet

FOCK_EXTRAP_ORDER 6

FOCK_EXTRAP_POINTS 12

$end

but I end up with a Bad Overlap error after 1 or 2 steps:

overlap matrix:

1.8957e-05 1.4558e-02 -6.8652e-03 -1.5981e-04

-4.0988e-04 4.8832e-04 -7.0997e-04 6.3056e-04

3.2727e-03 -6.1944e-04 -3.6709e-03 -5.0990e-03

-1.6554e-03 -7.6313e-03 -2.0576e-02 7.0119e-05

Q-Chem fatal error occurred in module aimdman/finite_difference_dc.C, line 754:

Ending Q-Chem run because of bad overlap.

I have tried already to increase the number of roots/fssh_nsurfaces (up to 4) and to reduce the time step ranging from 40 to 6. Despite this I always get the same error.

Interestingly, I didn’t get any errors when not including the ground state (FSSH_LOWESTSURFACE = 1) in the calculations, i.e. NAMD with surface hopping from S1 to higher states.

Is there any other suggestion about how to resolve this?