Molcas Forum

Nuno

Member

Registered: 2016-01-15

Posts: 14

Error in MP2 optimisation

While running the 8.0 freely distributed binaries I came across a strange error message while performing an MP2 optimisation run.
Does anyone on the dev team know what is going wrong here?

----------------------------------------------------------------------------------------------------
 DGA/MPI-2 Parallel Environment:    4 Molcas's processes are running on   1 node(s) x  4 cores each
----------------------------------------------------------------------------------------------------
 ###############################################################################
 ###############################################################################
 ###                                                                         ###
 ###                                                                         ###
 ###    OldFCM: iInter.ne.nInter .or. iInter.le.0                            ###
 ###                                                                         ###
 ###                                                                         ###
 ###############################################################################
 ###############################################################################
 iInter,nInter=                    4                    3
MOLCAS error: Terminating!, code = 128
--------------------------------------------------------------------------
MPI_ABORT was invoked on rank 2 in communicator MPI_COMM_WORLD 
with errorcode 128.

NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes.
You may or may not see output from other processes, depending on
exactly when Open MPI kills them.
--------------------------------------------------------------------------
MOLCAS error: Terminating!, code = 128
--------------------------------------------------------------------------

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,012

Re: Error in MP2 optimisation

Is it with numerical gradients? Can you show your input? Which program gives the error (slapaf, mbpt2, numerical_gradient...)?

Nuno

Member

Registered: 2016-01-15

Posts: 14

Re: Error in MP2 optimisation

Actually it is with the Numerical_Gradient module. I did not explicitly ask for it in &Slapaf as I had thought that analytic gradients existed for MP2 but apparently it is automatically chosen for me. I could send the input file to you in private should you wish to see it.

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,012

Re: Error in MP2 optimisation

The relevant part of the input (next time please post it in the forum, if possible) is:

&GATEWAY
...

>> Do   While

 &SEWARD
CHOLESKY
End of input

&SCF
 charge 
 2
End of input

&MBPT2 
End of input

&SLAPAF
 MAXSTEP
 0.01
 CARTESIAN
End of input

>> EndDo

You get numerical gradients because you used "CHOLESKY" in &seward. You may have better luck (analytical gradients) if you use "RICD" in &gateway instead (recommended), but that won't work in parallel.

The error you are getting is, I believe, fixed for later versions, but for 8.0 you may be able to work around it by distorting your initial geometry so that it's not exactly linear.

Nuno

Member

Registered: 2016-01-15

Posts: 14

Re: Error in MP2 optimisation

Thanks. Does the 15-06-18 revision still have this bug ?
From what you say I understand that not using the Cholesky integral approximation will make the program adopt an analytic gradient optimisation is that correct?

Last edited by Nuno (2016-01-15 15:52:00)

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,012

Re: Error in MP2 optimisation

Thanks. Does the 15-06-18 revision still have this bug ?

Yes, as far as I can see.

From what you say I understand that not using the Cholesky integral approximation will make the program adopt an analytic gradient optimisation is that correct?

Right. Conventional integrals (serial or parallel) or RICD integrals (serial) should give you correct analytical gradients for MP2 without symmetry.