Memory and RICIS(D) calculations

Questions
#1

I am trying to test a RICIS(D) calculation on a 38-atom molecule and am having issues understanding whether it is possible or not.

$molecule
  0   1
  S  -1.629783     1.486780     1.602787
  C  -3.172502     0.660737     1.707813
  C  -3.171184    -0.671275     1.715410
  S  -1.626373    -1.494026     1.616410
  C  -0.672927    -0.002912     1.566063
  C   0.668335    -0.002912     1.561352
  S   1.628459     1.484068     1.572165
  C   3.166985     0.659479     1.699600
  C   3.186469    -0.674950     1.725186
  S   1.624518    -1.488871     1.603089
  H  -4.067465     1.275986     1.756352
  H  -4.064939    -1.287750     1.770649
  H   4.056837     1.280344     1.774784
  C   4.396566    -1.534518     1.921570
  C  -0.674382     1.248928    -1.603512
  C   0.667355     1.254227    -1.601682
  C   1.423987     0.008538    -1.556222
  C   0.676999    -1.243114    -1.602489
  C  -0.664712    -1.248318    -1.602697
  C  -1.421474    -0.002388    -1.555263
  C  -2.783555    -0.008233    -1.471895
  C  -3.542008     1.209629    -1.409803
  C  -3.532327    -1.232067    -1.410597
  C   2.785871     0.014766    -1.472754
  C   3.537145     1.236832    -1.413692
  C   3.542932    -1.203270    -1.414911
  N  -4.141245     2.196106    -1.343501
  N  -4.124340    -2.222918    -1.344413
  N   4.135288     2.224096    -1.349107
  N   4.139045    -2.192161    -1.355305
  H  -1.227484     2.184990    -1.626392
  H  -1.210300    -2.188862    -1.625422
  H   1.213657     2.194190    -1.623971
  H   1.229804    -2.179290    -1.626254
  H   4.394358    -2.385786     1.233493
  H   5.295973    -0.949244     1.703792
  S   4.598251    -2.114493     3.642597
  H   3.499525    -2.895097     3.677215
$end

$rem
  METHOD          RICIS(D)
  BASIS           cc-pVDZ
  AUX_BASIS       RIMP2-cc-pVDZ
  MEM_TOTAL       120000
  AO2MO_DISK      50000
  MEM_STATIC      2000
  CIS_N_ROOTS     3
  CIS_TRIPLETS    FALSE
  SCF_CONVERGENCE 8
  THRESH          14
$end

First, the $rem variable

AO2MO_DISK

seems to be used throughout different examples in the manual but never really explained. Second, I am confused in regards to what MEM_STATIC and MEM_TOTAL should be set to. Section 7.9.7 RI-CIS(D), SOS-CIS(D), and SOS-CIS(D0 ): Job Control in the manual says:

At least 150(N^2+N)D of MEM_STATIC is required (N: number of basis functions, D: size of a double precision storage, usually 8). Because a number of matrices with N^2 size also need to be stored, 32–160 MB of additional MEM_STATIC is needed.

In the example above,

 Molecular Point Group                 C1    NOp =  1
 Largest Abelian Subgroup              C1    NOp =  1
 Nuclear Repulsion Energy =        3450.86728495 hartrees
 There are      116 alpha and      116 beta electrons
 Requested basis set is cc-pVDZ
 There are 208 shells and 462 basis functions

 Total QAlloc Memory Limit 120000 MB
 Mega-Array Size      1956 MB
 MEM_STATIC part      2000 MB

so N = 462, and according to the formula in the manual, MEM_STATIC should be set to 256687200 which seems very high to me. Example 7.64 in the manual

$molecule
   0 1
   C     0.667472     0.000000     0.000000
   C    -0.667472     0.000000     0.000000
   H     1.237553     0.922911     0.000000
   H     1.237553    -0.922911     0.000000
   H    -1.237553     0.922911     0.000000
   H    -1.237553    -0.922911     0.000000
$end

$rem
   METHOD          ricis(d)
   BASIS           aug-cc-pVDZ
   MEM_TOTAL       1000
   MEM_STATIC      100
   AO2MO_DISK      1000
   AUX_BASIS       rimp2-aug-cc-pVDZ
   PURECART        1111
   CIS_N_ROOTS     10
   CIS_SINGLETS    true
   CIS_TRIPLETS    false
$end

has 82 basis functions which, per that formula, requires 8167200 of MEM_STATIC when, clearly, 100 is enough. Is the formula given in the manual wrong or am I misunderstanding it? When I try to run my calculation, I get

 RI-CIS(D) job begins now...

Calculating maximum allowed batch size for RI-CIS(D)
 Memory available: 119999 MB
 Disk   available: 50000 MB
 Maximum batch size: 101 (Allowed from memory: 245/ disk: 101)
 Memory required:  -10185 MB
 Disk   required:  49648 MB

Form (P|Q)**(-1/2)...
Forming (pq|P)...


 Q-Chem warning in module libmdc/NewQAlloc.C, line 469:

 QAllocGeneral() returns a null pointer when allocating 18446744062933890160 Bytes memory


 Q-Chem fatal error occurred in module libmdc/newfileman.C, line 348:

 FileMan error: End of file reached prematurely reading (5004544) bytes in file FILE_VIAQ_INTS
 Path: /scratch/jearias/AO2MO.20260310152610/617.0


 Please submit a crash report at q-chem.com/reporter

18446744062933890160 bytes of memory is obviously outside of my limits but the line Memory required: -10185 MB makes me believe I am missing some memory-related keyword or not properly adjusting the values and perhaps the calculation is indeed possible.

Any help would be appreciated.

#2

Try setting AO2MO_DISK to the max size of your disk. This looks like the old semi-direct MP2 code (now seemily deprecated for MP2, sadly) where you could always compromise between memory and disk. AO2MO_DISK is the disk space allowed for the 4-index integral transformation.

#3

According to the nodes available to me,

~$ free -m
              total        used        free      shared  buff/cache   available
Mem:         192070        1520      189428           0        1121      189221
Swap:         31249          45       31204

I tried giving

$rem
  METHOD          RICIS(D)
  BASIS           cc-pVDZ
  AUX_BASIS       RIMP2-cc-pVDZ
  MEM_TOTAL       192000
  AO2MO_DISK      192000
  MEM_STATIC      2000
  CIS_N_ROOTS     3
  CIS_TRIPLETS    FALSE
  SCF_CONVERGENCE 8
  THRESH          14
$end

and still got

 RI-CIS(D) job begins now...

Calculating maximum allowed batch size for RI-CIS(D)
 Memory available: 191999 MB
 Disk   available: 192000 MB
 Maximum batch size: 116 (Allowed from memory: 313/ disk: 436)
 Memory required:  -3666 MB
 Disk   required:  56005 MB

Form (P|Q)**(-1/2)...
Forming (pq|P)...
 

 Q-Chem warning in module libmdc/NewQAlloc.C, line 469:

 QAllocGeneral() returns a null pointer when allocating 18446744069769369280 Bytes memory


 Q-Chem fatal error occurred in module libmdc/newfileman.C, line 348:

 FileMan error: End of file reached prematurely reading (5004544) bytes in file FILE_VIAQ_INTS
 Path: /scratch/jearias/AO2MO.20260310170116/617.0


 Please submit a crash report at q-chem.com/reporter

Interestingly, the non-RI version

$rem
  METHOD          CIS(D)  
  BASIS           cc-pVDZ
  MEM_TOTAL       120000
  MEM_STATIC      2000
  AO2MO_DISK      32000
  CIS_N_ROOTS     3
  CIS_TRIPLETS    FALSE
  SCF_CONVERGENCE 8
  THRESH          14
$end

seems to at the very least be running.

#4

Sorry, I also meant to say: It’s possible that RI-CIS(D) is still using MEM_STATIC rather than MEM_TOTAL. A lot of that code, which was written in the mid-2000s, is based on the old version of the MP2 (before the mid-2010s rewrite), and the old MP2 code used MEM_STATIC and AO2MO_DISK. I’m not sure about any of this without looking inside but based on the history, these are my guesses.