Problem with DETCI and 7Z basis set

1

I am facing a problem running the DETCI module for the 7-zeta basis set. The input file is:
#! C
memory 40 gb
molecule{
0 3
C 0. 0. 0.
}
set{
reference rohf
docc [2,0,0,0,0,0,0,0]
socc [0,0,0,0,0,0,1,1]
basis aug-cc-pv7z
guess gwh
df_scf_guess false
frozen_docc [1,0,0,0,0,0,0,0]
e_convergence 9
r_convergence 4
maxiter 20
}
energy(‘cisd’)

The DETCI module crashes at the “==> Setting up CI strings <==” step with the message (in fact, some 10^6 messages):
(subgr_lex_addr): Impossible walk!

and then:
There are 104941 alpha and 267 beta strings
The CI space requires 67188 (6.72E+04) determinants and 16 blocks

The calculation with the 6-zeta basis sets yields fine results. Also, the 7-zeta HF energy looks fine.
I would greatly appreciate any advice.

2

@sherrill has identified the problem that detci requires less than 256 active orbitals. Solution not imminent, I regret.

3

But if the limitation is just the datatype, shouldn’t it be rather straightforward to fix it…?

Also, the run segfaults on my computer.

4

Yes, a check for >256 active orbitals will be added to the driver so that it fails gracefully.

It could definitely be rewritten, but detci is really not optimized for cisd, cisdt, etc., and the limitation doesn’t impede detci’s strengths. For purpose-written cisd in psi4, there is the fnocc module (default for RHF), but it’s RHF-only.

5

Lori, thanks! So, what will be first on the list, detci for >256 active orbitals or fnocc for rohf? Or none?

6

Neither, I’m afraid. I think storage is too valuable in detci to double the size for a nice but untargeted method, cisd, and fnocc is Eugene DePrince’s code, and his research has moved on. For open-shell at the same gross level, all I can offer is lccd and ccsd.