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".

Pages: 1

#1 2016-09-28 13:03:14

Nuno
Member
Registered: 2016-01-15
Posts: 16

CASPT2 molden file

Dear all,

Is there any undocumented keyword/procedure that would allow the direct extraction of CASPT2 natural orbitals onto a molden file ?

I'm aware that a &RASSCF run with CIONLY can produce it from a fed in PT2ORB file but that molden file will not contain the natural occupations in the same way as it can come out from an &mbpt2 run. I was planning to use such a molden file by a third party population analysis program which of course needs the correct occupations.

Best regards,
Nuno

Offline

#2 2016-09-28 13:21:44

Ignacio
Administrator
From: Uppsala
Registered: 2015-11-03
Posts: 1,085

Re: CASPT2 molden file

I'm not sure there's any automated way in 8.0, generating a caspt2.molden file was added in the development version about a year ago. But the occupation numbers should be written in the Pt2Orb file, you may be able to transfer the numbers to the molden file. You could also try molpy (https://github.com/steabert/molpy), but I'm not sure it can do that either (the Pt2Orb does not contain all the information needed for the molden file). In any case, transferring the orbitals from Pt2Orb to molden is mostly a matter of rearranging the coefficients, I may have some old script that could do the job...

Offline

#3 2016-09-28 14:33:14

valera
Administrator
Registered: 2015-11-03
Posts: 129

Re: CASPT2 molden file

Nuno,
there is one 'undocumented' feature: run &SCF with ITER=0. It will do nothing with the orbitals, but compute all properties, e.g. charges, and produce a molden file

Offline

#4 2016-09-28 14:49:53

Nuno
Member
Registered: 2016-01-15
Posts: 16

Re: CASPT2 molden file

Thanks!

Molpy looks promising but it mentions an hdf file which I have no idea what it is. RunFile perhaps?

Valera's approach to cheat the program seems to be the most pragmatic, until the next version comes out that is.

Offline

#5 2016-10-04 13:00:46

Nuno
Member
Registered: 2016-01-15
Posts: 16

Re: CASPT2 molden file

Actually I've tried this with MOLCAS 8.0 and what I got was a memory error and no molden file. sad 


++    Optimization specifications:
      ----------------------------

      SCF Algorithm: LK-Cholesky
      The actual AO density is used
      Number of density matrices in core                1
      Maximum number of NDDO SCF iterations             0
      Maximum number of HF  SCF iterations              0
      Threshold for SCF energy change            0.10E-08
      Threshold for density matrix               0.10E-03
      Threshold for Fock matrix                  0.15E-03
      Threshold for linear dependence            0.10E-08
      Threshold at which DIIS is turned on       0.15E+00
      Threshold at which QNR/C2DIIS is turned on 0.15E+00
      Threshold for Norm(delta) (QNR/C2DIIS)     0.20E-04
      DIIS extrapolation of the SCF procedure
      All orbitals punched on: SCFORB
--

      Input vectors read from INPORB
      Orbital file label: * CASPT2


---------------------------------------------------------------------------------------------
  Nr.	 Label		Type		Offset		Length	   Atime	  Address
---------------------------------------------------------------------------------------------
  1	IP_DUM      	REAL	       5129180	           8           2	[0x3865020]
  2	IP_SDUM     	SNGL	      10258344	           4           3	[0x3864fe0]
  3	IP_IDUM     	INTE	       5129176	           8           4	[0x3865000]
  4	IGATIM      	REAL	       5126722	         160           5	[0x3860350]
  5	IGASTAT     	REAL	       5126744	          56           6	[0x3860400]
  6	ONEHAM      	REAL	       5134554	       84680          12	[0x386f810]
  7	OVLP        	REAL	       5145140	       84712          13	[0x38842e0]
  8	LOWDIN      	REAL	17482276017114	      168200          14	[0x7f3347e4d010]
  9	TRMAT       	REAL	       5155730	      168200          18	[0x3898dd0]
 10	CMO         	REAL	       5176756	      168200          19	[0x38c1ee0]
 11	FOCK        	REAL	       5197782	       84712          20	[0x38eaff0]
 12	OCCNO       	REAL	       5208372	        1160          21	[0x38ffae0]
 13	EORB        	REAL	       5208518	        1160          22	[0x38fff70]
 14	DENS        	REAL	       5208664	      169360          23	[0x3900400]
 15	TWOHAM      	REAL	       5229836	      169360          24	[0x39299a0]
 16	AW          	REAL	       5253768	       84680          57	[0x3958580]
 17	DNSS        	REAL	       5264354	      168200          58	[0x396d050]
 18	LWFSQ       	REAL	       5302158	      168200          59	[0x39b6db0]
 19	TEMPFLT     	REAL	       5285380	       84680          60	[0x3996160]
 20	NVEC_A+     	INTE	       5126778	          16          61	[0x3860510]
 21	CHOMOS      	REAL	       5323184	      168200          62	[0x39dfec0]
 22	DIAGI       	REAL	       5344210	       84664          64	[0x3a08fd0]
 23	DIAHI       	REAL	       5354794	      168200          65	[0x3a1da90]
 24	ABSC        	REAL	       5251928	        1160          66	[0x3954c00]
 25	YC          	REAL	       5375820	      168200          67	[0x3a46ba0]
 26	MLK         	REAL	       5295966	       12760          68	[0x39aac30]
 27	SKSH        	REAL	       5297562	       12760          69	[0x39ade10]
 28	INDX        	INTE	       5299158	       13920          70	[0x39b0ff0]
 29	IP_LAB      	INTE	       5252074	          88          71	[0x3955090]
 30	IP_KOFF     	INTE	       5252086	          88          72	[0x39550f0]
 31	IP_NNBF     	INTE	       5252098	         528          73	[0x3955150]
 32	IP_ISHP     	INTE	       5252166	         528          74	[0x3955370]
 33	IP_SVSH     	REAL	       5252234	        1056          75	[0x3955590]
---------------------------------------------------------------------------------------------
Maximal available memory for Molcas = 3143490092
MEMORY ERROR: Memory is exhausted!
MEMORY ERROR: Available memory = 3143490092 ( 2997 Mb ) !
MEMORY ERROR: Requested memory = 3145728088 ( 3000 Mb ) !
MEMORY ERROR: The suggested MOLCASMEM=5997 !
C_GetMem Calling parameters: ('F(K)SS ','ALLO   ','REAL   ',59114736,393216011)

++ Convergence information
                                             No optimization is performed
                                  Results refer to input orbitals read from INPORB

       Single iteration finished!

--

MMA failed to allocate a memory block.
--- Stop Module:  scf at Tue Oct  4 13:01:38 2016 /rc= _MEMORY_ERROR_ ---
     
     Segmentation dump
     
     ...................................................................................................
     ...................................................................................................
     ....Let me put it this way, Mr. Amor............................................................... 
     ....The 9000 series is the most reliable computer ever made........................................
     ....No 9000 computer has ever made a mistake or distorted information..............................
     ....We are all, by any practical definition of the words, foolproof and incapable of error.........
     ...................................................................................................

Offline

#6 2016-10-04 13:48:23

Nuno
Member
Registered: 2016-01-15
Posts: 16

Re: CASPT2 molden file

I found out what's causing the error. The &SCF module won't finish the job if CHOLESKY is chosen in &Seward. This could take some time for big molecules.

Last edited by Nuno (2016-10-04 13:48:38)

Offline

#7 2016-10-04 16:56:46

Ignacio
Administrator
From: Uppsala
Registered: 2015-11-03
Posts: 1,085

Re: CASPT2 molden file

I'd suggest you use RICD in GATEWAY instead of Cholesky in SEWARD...

Offline

#8 2016-10-08 01:26:20

josejc
Member
Registered: 2016-09-18
Posts: 9

Re: CASPT2 molden file

Perform a CASSCF calculation with zero iterations, with only the optimization of the CI vector.
The calculation ends quickly without converge, and MOLCAS prints the CASPT2 Molden file.

&GATEWAY
coord=test.xyz
basis=ANO-S-VTZP
Group=Nosym
&SEWARD
&RASSCF
FileOrb=$DIR/test.Pt2Orb
symmetry=1
spin=1
INACtive=23
nactel=12 0 0
Ras2=11
CIONly
CIMX=0
LumOrb

If the SCF calculation is performed the occupation numbers are assigned incorrectly.

Last edited by josejc (2016-10-08 01:27:31)

Offline

#9 2016-10-14 11:01:49

Nuno
Member
Registered: 2016-01-15
Posts: 16

Re: CASPT2 molden file

Dear josejc

I do not want to optimize the CI vector. That has already been done by the CASPT2 module. In fact I don't want MOLCAS to touch the MOs at all.

In fact I'm not even sure the MO order in the PT2Orb file is the same as in a RasOrb file since post-perturbation natural occupations can change slightly especially if you have moderately significant intruder states or you're using constraints as with supsym. Since the logical move would be to order the MOs by growing natural occupation the Pt2Orb file may well have a different orbital ordering from that of RasOrb as a consequence.

Valera's approach does work but only without Cholesky in &Seward. I have no idea why. I haven't tried it with RICD but I will do.

Last edited by Nuno (2016-10-14 11:02:38)

Offline

#10 2016-10-14 16:41:51

josejc
Member
Registered: 2016-09-18
Posts: 9

Re: CASPT2 molden file

The RASSCF module is called in this case only for read the INPORB file and print the Molden file. The starting molecular orbitals (CASPT2) are included in the RASSCF module as "FileOrb=$DIR/test.Pt2Orb", and the the name of the caspt2 molden file will be $input.rasscf.molden as they were generated with the rasscf module.

The calculation does not optimize, does nothing (CIONly;CIMX=0); begins, read the starting MO, does nothing, finish and print the "$ input.rasscf.molden" file. In this case, the rasscf output (ci vector, etc.) makes no sense at all, but the Molden file is generated correctly.

In that way I get the caspt2 molden files with molcas78 (without Cholesky decomposition). Some time ago, I wrote a perl script to get the castp2 molden file directly from the Pt2Orb file (with other information from the Molcas output) and I got the same result as perform the fake rasscf calculation.

If you do the SCF calculation with zero iterations (the Valera's approach) the caspt2 molden file is also generated correctly, but in this case the occupation numbers are set to 0 and 2.

Offline

#11 2016-10-14 17:01:22

Nuno
Member
Registered: 2016-01-15
Posts: 16

Re: CASPT2 molden file

josejc wrote:

If you do the SCF calculation with zero iterations (the Valera's approach) the caspt2 molden file is also generated correctly, but in this case the occupation numbers are set to 0 and 2.

This is not true. The occupation numbers are copied verbatim from INPORB. The only thing different is that the orbitals have one electron energies assigned to them in the molden file. This of course on molcas 8sp1.

Offline

Pages: 1

Links: OpenMolcas | Molcas | Manual | Facebook

Board footer

Powered by FluxBB 1.5.11

Last refresh: Today 13:51:42