Molcas Forum

Support and discussions for Molcas and OpenMolcas users and developers

You are not logged in.

Announcement

Welcome to the Molcas forum.

Please note: The forum's URL has changed. The new URL is: https://molcasforum.univie.ac.at. Please update your bookmarks!

You can choose an avatar and change the default style by going to "Profile" → "Personality" or "Display".

#1 2022-06-21 15:46:55

timothe33
Member
Registered: 2022-04-25
Posts: 12

RHF energy difference

Dear OpenMolcas users,

I am running a RHF calculation using STO-3G for LiH with OpenMolcas, GAMESS and Molpro.

I noticed that both GAMESS and Molpro energy are the same: -7.8633823379 Hartree but OpenMolcas lead to the energy: -7.8637374491

Any reasons why OpenMolcas does not give the same energy?

Here is my input for openMolcas:

&GATEWAY
Basis set
* LITHIUM  (6s,3p) -> [2s,1p]
 Li    / inline
      3.00   1
* S-type functions
     6    2
               0.1611957475E+02
               0.2936200663E+01
               0.7946504870E+00
               0.6362897469E+00
               0.1478600533E+00
               0.4808867840E-01
      0.1543289673E+00       0.00000000
      0.5353281423E+00       0.00000000
      0.4446345422E+00       0.00000000
      0.00000000            -0.9996722919E-01
      0.00000000             0.3995128261E+00
      0.00000000             0.7001154689E+00
* P-type functions
     3    1
               0.6362897469E+00
               0.1478600533E+00
               0.4808867840E-01
      0.1559162750E+00
      0.6076837186E+00
      0.3919573931E+00
Li 0.0 0.0 0.0000 /Angstrom
End of basis set
Basis set
* HYDROGEN  (3s) -> [1s]
 H    / inline
      1.00   0
* S-type functions
     3    1
               0.3425250914E+01
               0.6239137298E+00
               0.1688554040E+00
      0.1543289673E+00
      0.5353281423E+00
      0.4446345422E+00
H 0.0 0.0 1.5109 /Angstrom
End of basis set
*AMFI
*Symmetry = xy y
&SEWARD
*AMFI
*Expert
*Cholesky
&SCF
spin=1

Best,

Offline

#2 2022-06-21 16:33:17

Ignacio
Administrator
From: Uppsala
Registered: 2015-11-03
Posts: 1,079

Re: RHF energy difference

I get -7.8633821388

Offline

#3 2022-06-22 17:21:01

timothe33
Member
Registered: 2022-04-25
Posts: 12

Re: RHF energy difference

With the same input?

Offline

#4 2022-06-24 15:31:27

Ignacio
Administrator
From: Uppsala
Registered: 2015-11-03
Posts: 1,079

Re: RHF energy difference

Yes, copy-pasted. Version 22.02, tag 324-gf8fe7195b

Offline

#5 2022-06-24 16:53:46

timothe33
Member
Registered: 2022-04-25
Posts: 12

Re: RHF energy difference

I tried with this version and it di not change my result.
Do you think that it could be due to how I compile OpenMolcas? I followed the instruction from the github page but did not change the compiler, etc...

Best,

Offline

#6 2022-06-24 19:29:26

Ignacio
Administrator
From: Uppsala
Registered: 2015-11-03
Posts: 1,079

Re: RHF energy difference

It could be. Have you run the verification ("pymolcas verify") in the build directory? Are you running it in parallel (MPI or multithreaded)?

Offline

#7 2022-06-27 17:15:16

timothe33
Member
Registered: 2022-04-25
Posts: 12

Re: RHF energy difference

When I run the "pymolcs verify", all tests passed besides the grayzone one. I run the calculation on one node, one processor so not in parallel.

I will try to recompile with changing the settings.

Best,

Offline

#8 2022-06-29 19:22:31

timothe33
Member
Registered: 2022-04-25
Posts: 12

Re: RHF energy difference

Hey Ignacio,

I found the issues: by default my calculation (with the input given) were using SCF: "SCF Algorithm: LK-RI/DF"
When I added "noCD" in my input in GATEWAY, I got: "SCF Algorithm: Conventional" and the energy:  -7.8633821388
which is the value you got. I am not sure why default OpenMolcas does not do the conventional SCF but at least it solved my issue.

Thanks for your help.

Best,

Offline

#9 2022-06-30 08:35:39

Ignacio
Administrator
From: Uppsala
Registered: 2015-11-03
Posts: 1,079

Re: RHF energy difference

Ah, so I guess you had set MOLCAS_NEW_DEFAULTS="YES" (see https://molcas.gitlab.io/OpenMolcas/sph … variables)

The reason to use RICD by default is that it is faster, and needs fewer resources (at least for larger systems, not for LiH with minimal basis), and the errors in relative energies are negligible, although absolute energies are indeed not comparable with conventional calculations.

Offline

Board footer

Powered by FluxBB 1.5.11

Last refresh: Today 22:09:26