SA-CASSCF S1/S0 CONICAL INTERSECTION GEOM OPT. - &ALASKA STOPS SUDDEN

tuancpacheco · 2026-04-11 18:28:17

Hello there!

I'm using OpenMolcas v. 26.02 and I'm now trying to optimize conical intersection geometries for a 24 atoms system with SA-CASSCF method, but a sudden stop in module ALASKA occurred.

Input orbitals are from a previously converged SA-CASSCF(12,11) single-point for the starting geometry which is a candidate for S1/S0 crossing geometry.

Here's the input

 
&GATEWAY

	coord=v_oriri.xyz
	basis=def2-tzvp
	group=nosym
	Constraints
	 a=Ediff 1 2
	Value
	 a=0.000
	End of Constraints



>> DO WHILE <<


&SEWARD

	DoAnalytical

	CHOInput

		1-CEnter

	EndCHOInput

&RASSCF

	FILEOrb=cas_oriri.RasOrb
	spin=1
	nactel=12
	inactive=54
	ras2=11
	CIROOT=2 2 1
	RLXRoot=1


&SLAPAF

	ITERATIONS=300

>> ENDDO <<

and I got this non-zero code exit at the output after convergence was signaled after 18 cycles

--- Stop Module: mclr at Fri Apr 10 17:42:50 2026 /rc=_RC_ALL_IS_WELL_ ---
*** files: xmldump
    saved to directory $SCRIPT_SUBMIT_DIR
--- Module mclr spent 23 hours 1 minute 41 seconds ---

*** symbolic link created: INPORB -> opt_oriri.RasOrb
--- Start Module: alaska at Fri Apr 10 17:43:26 2026 ---
 
()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()

                                              &ALASKA

             launched 16 MPI processes, running in PARALLEL mode (work-sharing enabled)
                        available to each process: 10 GB of memory, 1 thread
                                        master pid: 1026832
()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()()
 
 
                     Threshold for contributions to the gradient: 1.000E-07
 
 
                    ********************************************
                    * Symmetry Adapted Cartesian Displacements *
                    ********************************************
 
 
           Irreducible representation : a  
           Basis function(s) of irrep: x, y, xy, Rz, z, xz, Ry, yz, Rx, I                                              
 
 Basis Label        Type   Center Phase
   1   N1           x         1     1
   2   N1           y         1     1
   3   N1           z         1     1
   4   C2           x         2     1
   5   C2           y         2     1
   6   C2           z         2     1
   7   C3           x         3     1
   8   C3           y         3     1
   9   C3           z         3     1
  10   N4           x         4     1
  11   N4           y         4     1
  12   N4           z         4     1
  13   C5           x         5     1
  14   C5           y         5     1
  15   C5           z         5     1
  16   C6           x         6     1
  17   C6           y         6     1
  18   C6           z         6     1
  19   N7           x         7     1
  20   N7           y         7     1
  21   N7           z         7     1
  22   C8           x         8     1
  23   C8           y         8     1
  24   C8           z         8     1
  25   N9           x         9     1
  26   N9           y         9     1
  27   N9           z         9     1
  28   C10          x        10     1
  29   C10          y        10     1
  30   C10          z        10     1
  31   N11          x        11     1
  32   N11          y        11     1
  33   N11          z        11     1
  34   C12          x        12     1
  35   C12          y        12     1
  36   C12          z        12     1
  37   C13          x        13     1
  38   C13          y        13     1
  39   C13          z        13     1
  40   S14          x        14     1
  41   S14          y        14     1
  42   S14          z        14     1
  43   N15          x        15     1
  44   N15          y        15     1
  45   N15          z        15     1
  46   C16          x        16     1
  47   C16          y        16     1
  48   C16          z        16     1
  49   H17          x        17     1
  50   H17          y        17     1
  51   H17          z        17     1
  52   H18          x        18     1
  53   H18          y        18     1
  54   H18          z        18     1
  55   H19          x        19     1
  56   H19          y        19     1
  57   H19          z        19     1
  58   H20          x        20     1
  59   H20          y        20     1
  60   H20          z        20     1
  61   H21          x        21     1
  62   H21          y        21     1
  63   H21          z        21     1
  64   H22          x        22     1
  65   H22          y        22     1
  66   H22          z        22     1
  67   H23          x        23     1
  68   H23          y        23     1
  69   H23          z        23     1
  70   H24          x        24     1
  71   H24          y        24     1
  72   H24          z        24     1
 
                     No automatic utilization of translational and rotational invariance of the energy is employed.
 
 Cholesky-ERI gradients!
--- Stop Module: alaska at Fri Apr 10 17:45:24 2026 /rc=_RC_EXTERNAL_TERM_ ---
*** files: xmldump
    saved to directory $SCRIPT_SUBMIT_DIR
--- Module alaska spent 1 minute 58 seconds ---

.########################.
.# Non-zero return code #.
.########################.

    Timing: Wall=84046.14 User=37008.39 System=6989.19

I looked for similar problems in forum and it didn't helped me to figure out what is wrong. Thank you.

Last edited by tuancpacheco (2026-04-13 11:17:46)

tuancpacheco · 2026-04-13 11:17:19

I thought it could be some bug or memory issue from the cluster because of the return code _EXTERNAL_TERM_, then I launched 32 processors instead of 16 for the same single job with 10000 MB for each and the same crash occurred. What else could it be?

The SLURM output ends up with

--------------------------------------------------------------------------
Primary job  terminated normally, but 1 process returned
a non-zero exit code. Per user-direction, the job has been aborted.
--------------------------------------------------------------------------
--------------------------------------------------------------------------
mpiexec noticed that process rank 25 with PID 1066623 on node n02 exited on signal 9 (Killed).
--------------------------------------------------------------------------
slurmstepd-n02: error: Detected 8 oom_kill events in StepId=550.batch. Some of the step tasks have been OOM Killed.

Last edited by tuancpacheco (2026-04-13 11:25:41)

Ignacio · 2026-04-14 15:42:02

oom_kill means "out of memory" (I believe). Make sure *each* of the processes has access to the requested 10GB (non-overlapping). More processes will only make it more likely to crash.

tuancpacheco · 2026-04-16 02:14:13

Thanks Ignacio. I launched 2 nodes (16 processors each) and reduced mem to 8000 MB. Now the return code is not external nor oom_kill anymore. It crashes at the same step but shows

                     No automatic utilization of translational and rotational invariance of the energy is employed.
 
 Cholesky-ERI gradients!
 Not Implemented for Cholesky yet!
--- Stop Module: alaska at Tue Apr 14 11:37:19 2026 /rc=41 ---
*** files: xmldump
    saved to directory $SCRIPT_SUBMIT_DIR
--- Module alaska spent 10 minutes 31 seconds ---

.########################.
.# Non-zero return code #.
.########################.

    Timing: Wall=4391.09 User=24712.47 System=15124.19

I thus launched with no Cholesky keyword and it is too slow, apparently spending more than 13h in first Seward module neutral

Last edited by tuancpacheco (2026-04-16 02:31:57)

Ignacio · 2026-04-17 09:50:15

Remove ChoInput and add RICD in GATEWAY.

tuancpacheco · 2026-04-27 11:09:52

It worked!!! Crystal Clear convergence!

Thanks a lot, Ignacio! My investigation can proceed hereafter.

Molcas Forum

Announcement

#1 2026-04-11 18:28:17

SA-CASSCF S1/S0 CONICAL INTERSECTION GEOM OPT. - &ALASKA STOPS SUDDEN

#2 2026-04-13 11:17:19

Re: SA-CASSCF S1/S0 CONICAL INTERSECTION GEOM OPT. - &ALASKA STOPS SUDDEN

#3 2026-04-14 15:42:02

Re: SA-CASSCF S1/S0 CONICAL INTERSECTION GEOM OPT. - &ALASKA STOPS SUDDEN

#4 2026-04-16 02:14:13

Re: SA-CASSCF S1/S0 CONICAL INTERSECTION GEOM OPT. - &ALASKA STOPS SUDDEN

#5 2026-04-17 09:50:15

Re: SA-CASSCF S1/S0 CONICAL INTERSECTION GEOM OPT. - &ALASKA STOPS SUDDEN

#6 2026-04-27 11:09:52

Re: SA-CASSCF S1/S0 CONICAL INTERSECTION GEOM OPT. - &ALASKA STOPS SUDDEN

Board footer