Problem with the "grid-it" script

cymantren · 2019-01-28 15:54:03

Hello everybody,

I made a kind of an update- installing the new version of OPENMOLCAS- however if I use "&Grid-It All" I take several hours but it does'nt finish.
So for example I calculated B3:
---------------------------------------------

 &GATEWAY
Title= energy of B3 dubl: MO's
coord=B3_unopt.xyz
basis=6-31G**
Group=Full

 &Seward
 &SCF
UHF
ZSpin=1
End of input

 &Grid_It
All

---------------------------------------------

but ps shows me:

18657 pts/1    00:00:00 python3
18671 pts/1    02:30:35 grid_it.exe

At the end a file "B3_engy....lus" is produced, but it has 0 Bytes !

What is wrong with this script ?! I have no idea.
Until now, this script always works fine, but now it fails all the time.

It would be very kind,if someone has an idea what goes wrong.

Kindly regards

Alfred Güthler

Ignacio · 2019-01-29 14:42:23

For the record, I can reproduce it with the "external" grid_it (https://sourceforge.net/projects/gridit/), with the xyz:

3

B 0.0 0.0 0.0
B 1.0 0.0 0.0
B 0.0 1.0 0.0

But it works fine with the code in molcas-extra.

It seems this is a bug in the External/grid_it/pickorb.f file, where at the end the ik_ab variable takes some crazy value, like 4987793981803345217, resulting in an almost infinite loop.

If you need to view orbitals or generate volumetric data, you can maybe use Pegamoid instead.

cymantren · 2019-02-01 12:07:27

Dear Ignacio,

thank you for your reply !

As I said before I use Openmolcas for some time and had never problems with the "grid_it"-script,but after installing a new version of OpenMolcas, this problem arise.
You said,that it is to be a problem with the "grid_it" script and that I can use the external script. This is an alternative, however there is no documentation, about how to compile the scirpt. There are Fortran -files and to compile these programs, I just know this syntax: "gfortran -o file file.f" But it does'nt work mainly because I don't know which is the main-file.
How does one compile this external "grid_it"script ?
It would be very kind, to get some hints.
Regarding to the "Pegamoid" - script I can tell you, that it needs newer libraries of VTK,python ... Libraries, that are hard to get, if one uses "debian-stretch", which supports only e.g. vtk 6.
It seems to be mostly a problem with the libraries of the OS and not directly with these scripts

However the problem remains and for a e.g. CASSCF, RASSCF-calculation, it is sensible to have some programs like "luscus,pegamoid" to get the active space and continue the calculation. It should be also possible to get the necessary information by looking at the MO's ?! So for example, if the occupation of the natural MO's are less than 2 or 1, then these orbitals belong to the active space. This is what I once have learned and it is reliable, although not so easy to recognize. Moreover, one has to know something about the molecular system, e.g. the NO's ,occupation number, to start the calculation. It is a kind of a vicious circle. One can only use the one's chemical intuition with "trial-and error" to get an approximate answer.
These programs "only" support one's understanding, but are also reliable and are - in this sense - necessary.

Thank you in advance for your help !

best wishes

Alfred Güthler

Ignacio · 2019-02-01 13:02:33

Grid_it is not a script, it is a program. With OpenMolcas you can only use the "external" one, meaning that the source code is not included in OpenMolcas, but is downloaded from a different repository. What I say is that this external code has a bug. If it worked before upgrading OpenMolcas it may have been just by pure luck.

Regarding Pegamoid, you may be able to install the requirements with "pip", "anaconda", or other python package manager, aside from the OS libraries, but there may be OS limitations too.

For the last part of your message, I don't quite understand. For CASSCF you always need to select an active space, and this needs some user input, since the "correct" orbitals depend on what you want to study. Sure, some viewer program is necessary, unless you can guess the orbitals from the MO coefficients. If you can't get Pegamoid to work, I'd recommend using Molden.

cymantren · 2019-02-05 12:03:14

Dear Ignacio,

thank you for your suggestions !

I have already installed the new version on OPENMOLCAS and tried the "grid-it" program with a simple H2O.input- and it worked ! However trying to do almost the same thing with the B3 input- it did'nt worked !! So I compared both inputs but I found nothing suspicious and the fact remains: The program "grid_it" works for a simple H2O-input, but not for B3 - why ?

Input for H2O:

  &GATEWAY
 coord=water.xyz
 basis=sto-3g
&SEWARD
 &SCF

 &GRID_IT
ALL

water.xyz

3
Angstrom
O  0.000000  0.000000  0.000000
H  0.758602  0.000000  0.504284
H  0.758602  0.000000 -0.504284

--------------

B3_engy_dubl_input:

&GATEWAY
Title= energy of B3 dubl: MO's
coord=B3_unopt.xyz
basis=6-31G**
Group=Full
 &SEWARD
 &SCF
UHF
ZSpin=1
End of input

 &GRID_IT
ALL

B3_unopt.xyz:

3
	Energy:     280.2443669
B         -2.42319        5.44721        0.00000
B         -1.27041        4.25832        0.00000
B         -2.87641        3.85443        0.00000

Can you figure out what is still going wrong ?

Thank you in advance for your helps !

kindly regards

Alfred

Ignacio · 2019-02-05 13:37:48

Grid_it is broken for UHF.

cymantren · 2019-02-14 18:01:27

Dear Mr. Ignacio,

thank you for your clear statement saying that the "grid_it" program is broken for UHF wavefunctions.

But what's about this: ScB2 optimization at RASSCF.
Does'nt the program calculates a former UHF-energy and then the latter RASSCF-energy ? It seems this is not the case, because a - file.lus - was produced and you mentioned, that UHF is broken. The manual says that it is not possible to calculate ROHF (p.92), however it is possible using RASSCF.
------------------------------

**     MOLCAS.input generated by molcas_input.py Version 2.0

&GATEWAY
COORD
3

Sc1      0.000000000      0.000000000     -0.667723000
B2      0.783281000      0.000000000      1.363305000
B3     -0.783281000      0.000000000      1.363305000
GROUP = nosym
TITLE = Molcas-Opt
BASIS = cc-pVDZ


**     ================ Optimization ================

>> COPY FORCE QuartetOrbitals.RasOrb INPORB
>>> DO WHILE

&SEWARD

&RASSCF
SPIN   = 4
NACTEL = 5,0,0
INACT  = 13
RAS2   = 5
CIROOT = 3,3,1
LUMORB
RLXROOT = 1

&SLAPAF
>>> ENDDO

**     ================ Frequencies ================

&MCKINLEY
&Grid_it
ALL

* Infos:
* ~/home at somestring
* Date: 2018-11-14 09:05:19.462538
* Current directory: ~/home/sharc-2.0/ScB2_opt

------------------------------

However the fact remains, that the "grid_it" program is still broken for UHF.
Maybe the problem is solved in later releases ?!

Best regards

Alfred Güthler

Ignacio · 2019-02-14 18:27:39

cymantren wrote:

Does'nt the program calculates a former UHF-energy and then the latter RASSCF-energy ? It seems this is not the case, because a - file.lus - was produced and you mentioned, that UHF is broken. The manual says that it is not possible to calculate ROHF (p.92), however it is possible using RASSCF.

I don't know what you mean, but RASSCF does not calculate UHF or ROHF. These are single-determinant methods, RASSCF computes multi-configurational wavefunctions.

valera · 2019-02-15 14:42:37

Ignacio wrote:

Grid_it is broken for UHF.

Supported vesrion of Grid_it (in Molcas) works for UHF. Unless there is a strong demand, I consider as a very low priority fixing the bugs in nonsupported versions.

cymantren · 2019-02-21 11:49:16

Dear Ignacio,Valera

I know that RASSCF calculates multideterminant wavefunctions, but I meant BEFORE it calculates the RASSCF-energy it calculates, as a starting point, the SCF-UHF-energy ? Anyway, as you and Mr. Veryazov already said,"Grid-it" is broken for UHF. I just wondered why the "Grid_it" works for RASSCF, BUT now IT IS CLEAR !
However, usually one uses UHF as a starting point and switch later to advanced methods. Of course it is sensible to look at the UHF-orbitals and then try to decide from the output-file, which orbitals are active. But this is not the correct topic for here.

Thank you both of you, for your answer !

kind greetings

Alfred

valera · 2019-02-21 19:03:45

cymantren
I don't agree with your statement that UHF orbitals give more inspiration, comparing to RHF orbitals, for future CAS calculations. UHF orbitals break symmetry and RASSCF should recover from that..
Sometimes I have heard a statement, well - I have no other choice since I have odd number of electrons. But it is also a faulty reason for choosing UHF: make the number even...

Ignacio · 2019-02-21 19:58:44

You can use the "UHF natural orbitals" ($Project.UnaOrb), I think grid_it will work with that.

Molcas Forum

Announcement

#1 2019-01-28 15:54:03

Problem with the "grid-it" script

#2 2019-01-29 14:42:23

Re: Problem with the "grid-it" script

#3 2019-02-01 12:07:27

Re: Problem with the "grid-it" script

#4 2019-02-01 13:02:33

Re: Problem with the "grid-it" script

#5 2019-02-05 12:03:14

Re: Problem with the "grid-it" script

#6 2019-02-05 13:37:48

Re: Problem with the "grid-it" script

#7 2019-02-14 18:01:27

Re: Problem with the "grid-it" script

#8 2019-02-14 18:27:39

Re: Problem with the "grid-it" script

#9 2019-02-15 14:42:37

Re: Problem with the "grid-it" script

#10 2019-02-21 11:49:16

Re: Problem with the "grid-it" script

#11 2019-02-21 19:03:45

Re: Problem with the "grid-it" script

#12 2019-02-21 19:58:44

Re: Problem with the "grid-it" script

Board footer