Issues with Slippage in Head-to-Head stacked composites during CDFT optimization

Questions
#1

Hi developers and users,

I am working on the optimization of head-to-head stacked composites using the cdft method with the B3LYP functional and the 6-31+G(d,p) basis set. After achieving convergence, I observed that the stacked composite slips significantly from its initial geometry.
The input as follows,
$molecule
read opt.txt
$end

$rem
jobtype opt
METHOD B3LYP
BASIS 6-31+G(d,p)
XC_GRID 000099000590
DFT_D D3_BJ
max_scf_cycles 200
GEOM_OPT_MAX_CYCLES 200
SCF_ALGORITHM DIIS
SCF_GUESS SAD
SCF_CONVERGENCE 8
THRESH 12
SCF_FINAL_PRINT 1
mem_static 4000
MEM_TOTAL 32000
SYMMETRY FALSE
SYM_IGNORE TRUE
CDFT TRUE
CDFT_THRESH 7
CDFT_BECKE_POP TRUE
BECKE_SHIFT UNSHIFTED
$end

$cdft
2
-1 11 24
+1 1 10
$end

I also tried with the 6-31G(d) and 6-31+G(d) basis sets, but the slippage issue still persists.
Has anyone else encountered this problem? Could this be due to the choice of functional/basis set, or should I try an alternative approach (eg. dispersion-corrected functionals or constrained during optimization)?

#2

Hard to say without your geometry, but is there a reason to think this result is wrong? You could try wB97X-D/6-31+G(d), which is sort of a modern replacement for B3LYP+D3, although it’s often qualitatively similar. wB97M-V is probably the best all-around functional for noncovalent interactions right now. Literature suggests that it requires larger basis sets to converge (e.g., def2-ma-TZVP) although for a geometry optimization you could probably get away with def2-ma-SVP. The “ma” are minimally augmented versions of the Karlsruhe basis sets.