atomic start

[Peterson, Kirk]

Dear Dirac experts (and non-experts):

 

I'm attempting to use the atomic start routines to get  a good initial guess for a difficult diatomic. Unfortunately all the examples I've found either involve 1^st row atoms where the open-shell occupations involve all the orbitals in ungerade symmetry or a fully closed-shell atom.  I can't figure out how one would use this even for something like sulfur.  I would like to have multiple orbital strings with their occupations, e.g., one string for closed shell orbitals and another for open shells, but this doesn't seem to work.  Can someone post an example?

 

regards,

 

-Kirk

[Andre Gomes]

hi kirk,

here is an example for i have for uo2cl4-2

As you can see, you have two lines defining the closed shells of the atom, and two for the open shells.

.ATOMST
AFUXXX 2
1..43
1.00
45..58
0.21
AFOXXX 2
1,2
1.00
3..5
0.66

AFCLXX 2
1..6
1.00
7,8,9
0.83

i hope this helps.

best,

andre

--
You received this message because you are subscribed to the Google Groups "dirac-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to dirac-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dirac-users/F7568000-6069-4FEE-B3AB-6C0D133D9488%40wsu.edu.

[Peterson, Kirk]

Hi Andre,

 

ah, from the manual (and its examples), I took the trailing "2" after the character string to mean inversion symmetry such that the two sets of strings corresponded to one for gerade and one for ungerade.

 

thanks!

 

regards,  -Kirk

 

From: <dirac...@googlegroups.com> on behalf of Andre Gomes <asp...@gmail.com>
Reply-To: "dirac...@googlegroups.com" <dirac...@googlegroups.com>
Date: Saturday, October 22, 2022 at 7:39 AM
To: "dirac...@googlegroups.com" <dirac...@googlegroups.com>
Subject: Re: [dirac-users] atomic start

 

[EXTERNAL EMAIL]

To view this discussion on the web visit https://groups.google.com/d/msgid/dirac-users/CADF8p%3D-ZBac8DV5u2TdvrHcx0yZBCEyPBqyHFtcHjLYYMwWJWQ%40mail.gmail.com.

[Peterson, Kirk]

thanks again Andre, this works as expected now.  A follow-up question though.  Is it typical for the energy to rise by nearly 100 Hartrees in iteration 2 before it starts moving down again? It makes me suspect that the benefits of using the atomic start is completely spoiled at that point.

 

regards, -Kirk

To view this discussion on the web visit https://groups.google.com/d/msgid/dirac-users/F1EA22BE-66CC-42C1-A4F9-CB98E10381E6%40wsu.edu.

[Andre Gomes]

Hi Kirk,

The experience I had with uo2cl4-2 was that without atomic start I just couldn’t converge the calculation, so the benefit was obvious.

Perhaps others will have more information.

Best,

Andre 

To view this discussion on the web visit https://groups.google.com/d/msgid/dirac-users/000B82F4-60A1-48F7-ADF3-D12F5668FDF9%40wsu.edu.

[Peterson, Kirk]

Hi Andre,

 

in my case it converges without it but the resulting spinor occupation is not what I would have expected, so I wanted to try atomic start just to make sure it was converging to the lowest energy solution.

 

best,

To view this discussion on the web visit https://groups.google.com/d/msgid/dirac-users/CADF8p%3D-d0ii5Qd0apP29bFLZhsw8bLbj%2BOi6EtWXvXEdTmPj4Q%40mail.gmail.com.

[Ayaki Sunaga]

Dear Kirk,

>in my case it converges without it but the resulting spinor occupation is not what I would have expected, so I wanted to try atomic start just to make sure it was converging to the lowest energy solution.

If your target is a diatomic molecule, you can specify the occupation number per spinor using .MJSELE option.
http://www.diracprogram.org/doc/master/manual/wave_function/scf.html#mjsele

Best regards,
Ayaki

2022年10月24日(月) 2:54 'Peterson, Kirk' via dirac-users <dirac...@googlegroups.com>:

To view this discussion on the web visit https://groups.google.com/d/msgid/dirac-users/7F02B231-30B7-410E-BE37-212116F3942D%40wsu.edu.

[Peterson, Kirk]

Dear Ayaki,

 

thanks, I might try that next (I had forgotten I could do that). It's mostly the character of the spinors I'm worried about, but maybe setting Omega will help too.

 

best regards,

 

-Kirk

To view this discussion on the web visit https://groups.google.com/d/msgid/dirac-users/CAHv9aY6DdB%3DibBiS2q%3DB7OyNaikKJUTK%3DS0bh%2B2gtjAGXD3HOg%40mail.gmail.com.

[Peterson, Kirk]

Is it not possible to have 5 values of .MJSELE?   Four values work just fine.

Specifically I get a crash for:

 

.OPEN SHELL

1

4/8

.MJSELEC

5

 1  3  5  7  9

72 38 16  4  0

 2  2  2  0  2

 

output:

 

 * Vector selection based on m_j symmetry

   - inactive shell:

      72  38  16   4

   - active shell(s):

       2   2   2   0

 

INPUT ERROR! This shell should have  4 orbitals!

 

Perhaps I've misinterpreted how the input for this works?  I was using .KPSELEC as a template.

To view this discussion on the web visit https://groups.google.com/d/msgid/dirac-users/76C83631-C43E-4333-8C02-7040B2E4CEBA%40wsu.edu.

[Ayaki Sunaga]

Dear Kirk,

I tried a simpler molecule with a similar electronic occupation, and I also got the same error message.
I think it is a bug in the input reader. I will check the source and reply back to you.

Best regards,
Ayaki

2022年10月25日(火) 7:10 'Peterson, Kirk' via dirac-users <dirac...@googlegroups.com>:

To view this discussion on the web visit https://groups.google.com/d/msgid/dirac-users/73CF8273-BFB1-4E6B-BC87-79FB0DB7C289%40wsu.edu.

[Peterson, Kirk]

Thank you Ayaki.

To view this discussion on the web visit https://groups.google.com/d/msgid/dirac-users/CAHv9aY7C%3DoZKxudPxoMma72A-AmoeaAXO46YMfhaFtYvmrQ6%3DQ%40mail.gmail.com.

[Ayaki Sunaga]

Dear Kirk,

I fixed the bug and made a merge request to the public master.
The merge request will be checked and accepted by a developer of DIRAC.
If you want to try the fixed version soon, you can modify the source as follows.

line 1715-1717 of Dirac/src/dirac/dirrdn.F
Before:
               DOBOSSEL = .TRUE.
               IBOSSEL  = -1
               READ(LUCMD,*,IOSTAT=IOS) NMJ

After:
               DOBOSSEL = .TRUE.
               IBOSSEL  = -1
               IF(TWOCOMP)THEN !2c
                 NSUBBL = NHTYP
               ELSE !4c
                 NSUBBL = NHTYP-1
               ENDIF                
               READ(LUCMD,*,IOSTAT=IOS) NMJ
 
I attached a simple example and the corresponding basis set file.

Best regards,
Ayaki

2022年10月25日(火) 22:55 'Peterson, Kirk' via dirac-users <dirac...@googlegroups.com>:

To view this discussion on the web visit https://groups.google.com/d/msgid/dirac-users/7613B67F-4C04-4BF4-A053-C47B30098782%40wsu.edu.

[Peterson, Kirk]

fantastic, thank you!

 

best,

 

-Kirk

To view this discussion on the web visit https://groups.google.com/d/msgid/dirac-users/CAHv9aY7_bM5yTPS%3D3vxohE0zoi5eOVmruyvLn89Oovu89KGsyA%40mail.gmail.com.