CIS(D) calculation memory problem

Questions
#1

Dear all,
My CIS(D) calculation failed due to some memory issue. I try to do a simple sp calculation:

$rem
BASIS cc-pvtz
JOBTYPE sp
METHOD CIS(D)
EE_SINGLETS 32
SYMMETRY false
SYM_IGNORE true
CCMAN2 false
MEM_STATIC 2000
MEM_TOTAL 4000
$end

For a small molecule (C8H10), it finished successfully. However, when I tried a larger molecule (C16H18), it failed with the following error:

In TrnAlg: available memory = 155198092
needed mmory = 439204000

Q-Chem fatal error occurred in module 0, line 283:

Insufficient memory for semidirect MP2

It seems that the “available memory” is determined by “MEM_STATIC”, but the maximal value is 2000mb in qchem. Can anyone point me to possible ways to solve this problem? Thank you very much.

Theta

#2

There is a known issue that prevents Q-Chem from routing correctly via METHOD=CIS(D), so the correct input is

$rem
BASIS cc-pvtz
JOBTYPE sp
METHOD MP2
CIS_SINGLETS 1
CIS_N_ROOTS 32
SYMMETRY false
SYM_IGNORE true
MEM_STATIC 2000
MEM_TOTAL 4000
AO2MO_DISK 100000
$end

Unfortunately this will still die with insufficient MEM_STATIC. Using RI might help:

METHOD RI-MP2
AUX_BASIS rimp2-cc-pvtz
1 Like
Effective Fragment Potential Method
#3

I try to use a very big memory, 234000MB. Why did I get this strange error?

Setting up for CIS(D)
SETman timing summary (seconds)
CPU time 16892.57s
System time 0.01s
Wall time 794.71s
Algorithm is semi-direct
Memory given = -401 MB Disk given = 1000000 MB
not enough memory
not enough disk

#4

Both conventional and RI perturbative doubles for CIS use MEM_STATIC to allocate memory, which is limited to about 8GB. Increasing MEM_TOTAL will have no effect. If you are running out of memory with conventional (D), try using RI as shown in the comment above. If your job runs out of memory with RI, I recommend contacting support@q-chem.com with your input files for the issue to be resolved in a future version of Q-Chem.