## Calculating spin-orbit matrix elements

(It is not necessary to upload input files, they can be found in the output file.)

JennyP
Posts: 7
Joined: 08 Apr 2016, 16:59
First name(s): Jenny
Last name(s): Pirillo
Affiliation: University
Country: Italy

### Calculating spin-orbit matrix elements

I would like to calculate spin-orbit matrix elements between excited states (S1 -> Tn), by the quadratic-response TDDFT approach, on systems containing heavy atoms, therefore I used a pseudopotential.

I tried to use both AMFI and effective-core one-electron spin-orbit operator methods.

The obtained results are substantially different (of orders of magnitude). This seems to be wrong.

Has anyone had the same problem?

Thanks

taylor
Posts: 595
Joined: 15 Oct 2013, 05:37
First name(s): Peter
Middle name(s): Robert
Last name(s): Taylor
Affiliation: Tianjin University
Country: China

### Re: Calculating spin-orbit matrix elements

I think posting a couple of outputs would help us better understand!

Best regards
Pete

JennyP
Posts: 7
Joined: 08 Apr 2016, 16:59
First name(s): Jenny
Last name(s): Pirillo
Affiliation: University
Country: Italy

### Re: Calculating spin-orbit matrix elements

below inputs and outputs for MNF and Effective-core one-electron SO, respectively 1 and 2

1. INPUT MNF-SO

**DALTON INPUT

.RUN RESPONS

.DIRECT

**INTEGRALS

.MNF-SO

**WAVE FUNCTIONS

.DFT

B3LYP

**RESPONS

.DOUBLE RESIDUE

.ISPABC

1 0 1

.PROPRT

X1MNF-SO

.PROPRT

Y1MNF-SO

.PROPRT

Z1MNF-SO

.ROOTS

3

**END OF INPUT

OUTPUT

@ Transition moment <B | A - <A> | C> in a.u. for

@ A operator label, symmetry, spin: X1MNF-SO 1 1

@ B excited state no., symmetry, spin: 1 1 0

@ C excited state no., symmetry, spin: 1 1 1

@ B and C excitation energies, moment: 0.09898754 0.08346398 -0.00007303

@ Transition moment <B | A - <A> | C> in a.u. for

@ A operator label, symmetry, spin: Y1MNF-SO 1 1

@ B excited state no., symmetry, spin: 1 1 0

@ C excited state no., symmetry, spin: 1 1 1

@ B and C excitation energies, moment: 0.09898754 0.08346398 0.00001433

@ Transition moment <B | A - <A> | C> in a.u. for

@ A operator label, symmetry, spin: Z1MNF-SO 1 1

@ B excited state no., symmetry, spin: 1 1 0

@ C excited state no., symmetry, spin: 1 1 1

@ B and C excitation energies, moment: 0.09898754 0.08346398 -0.00014859

2. INPUT Effective-core one-electron SO

**DALTON INPUT

.RUN RESPONS

.DIRECT

**INTEGRALS

**WAVE FUNCTIONS

.DFT

B3LYP

**RESPONS

.DOUBLE RESIDUE

.ISPABC

1 0 1

.PROPRT

X1SPNSCA

.PROPRT

Y1SPNSCA

.PROPRT

Z1SPNSCA

.ROOTS

3

**END OF INPUT

OUTPUT

@ Transition moment <B | A - <A> | C> in a.u. for

@ A operator label, symmetry, spin: X1SPNSCA 1 1

@ B excited state no., symmetry, spin: 1 1 0

@ C excited state no., symmetry, spin: 1 1 1

@ B and C excitation energies, moment: 0.08084941 0.06840855 0.00995203

@ Transition moment <B | A - <A> | C> in a.u. for

@ A operator label, symmetry, spin: Y1SPNSCA 1 1

@ B excited state no., symmetry, spin: 1 1 0

@ C excited state no., symmetry, spin: 1 1 1

@ B and C excitation energies, moment: 0.08084941 0.06840855 -0.00191314

@ Transition moment <B | A - <A> | C> in a.u. for

@ A operator label, symmetry, spin: Z1SPNSCA 1 1

@ B excited state no., symmetry, spin: 1 1 0

@ C excited state no., symmetry, spin: 1 1 1

@ B and C excitation energies, moment: 0.08084941 0.06840855 0.02035584

Bernd S.
Posts: 17
Joined: 19 Feb 2015, 13:31
First name(s): Bernd
Last name(s): Schimmelpfennig
Affiliation: KIT
Country: Germany

### Re: Calculating spin-orbit matrix elements

What system is it and what are the basis sets? Such cut out pieces are pretty useless. There are good reasons to allow arrachments of the full information.

JennyP
Posts: 7
Joined: 08 Apr 2016, 16:59
First name(s): Jenny
Last name(s): Pirillo
Affiliation: University
Country: Italy

### Re: Calculating spin-orbit matrix elements

the system is Te-Guanosine.
Below the file.mol:

ATOMBASIS

Atomtypes=5 Nosymmetry
Charge=52.0 Atoms=1 Basis=cc_pvdz_pp ECP=cc_pvdz_pp
Te01 -11.350719 2.791915 -18.465085
Charge=8.0 Atoms=3 Basis=cc-pVDZ
O001 -16.206008 1.460138 -3.952406
O002 -11.114993 -0.556411 -4.814234
O003 -8.976364 2.397996 -0.192905
Charge=7.0 Atoms=5 Basis=cc-pVDZ
N001 -10.708834 2.008199 -8.409843
N002 -13.218831 2.335322 -11.845633
N003 -6.965334 2.361906 -14.865610
N004 -3.002484 2.159150 -12.919893
N005 -6.591620 2.002310 -10.414884
Charge=6.0 Atoms=10 Basis=cc-pVDZ
C001 -15.087685 -0.497353 -2.455516
C002 -12.233798 -0.137839 -2.353949
C003 -9.983661 1.696403 -5.763806
C004 -13.146738 2.147349 -9.379240
C005 -10.704125 2.318097 -12.592918
C006 -9.582081 2.463035 -14.983162
C007 -5.552099 2.141904 -12.691556
C008 -9.121542 2.099667 -10.452077
C009 -11.336709 2.512909 -1.518799
C010 -10.860262 3.851842 -4.026970
Charge=1.0 Atoms=13 Basis=cc-pVDZ
H001 -18.024259 1.225612 -3.965539
H002 -15.501321 -2.361067 -3.267713
H003 -15.826402 -0.447138 -0.515190
H004 -11.472824 -1.583585 -1.079740
H005 -7.928284 1.492751 -5.774293
H006 -14.762302 2.062984 -8.126763
H007 -6.039765 2.479447 -16.545491
H008 -2.162457 1.802540 -14.599111
H009 -2.000829 1.684440 -11.361351
H010 -12.757874 3.477535 -0.364362
H011 -12.621727 4.669239 -4.726472
H012 -9.443226 5.344215 -3.859125
H013 -9.244402 1.595193 1.436455

taylor
Posts: 595
Joined: 15 Oct 2013, 05:37
First name(s): Peter
Middle name(s): Robert
Last name(s): Taylor
Affiliation: Tianjin University
Country: China

### Re: Calculating spin-orbit matrix elements

Could you please post your output files! Bernd was, I think, trying to stress that an expert who wishes to help may get useful information from things like convergence behaviour of the response equations and other parts of the calculation. And please upload files to the website, do not paste information inline. On the web page posting inline in the message leads to huge gaps between lines at least when I view it, and the copy auto-Emailed to me by the website is full of unhelpful linebreaks.

Best regards
Pete

JennyP
Posts: 7
Joined: 08 Apr 2016, 16:59
First name(s): Jenny
Last name(s): Pirillo
Affiliation: University
Country: Italy

### Re: Calculating spin-orbit matrix elements

attached files required

Attachments
MNF-SO.out
MNO-SO.dal
effectiveCoreOneElectron.out
effectiveCoreOneElectron.dal
TeG_S0.mol

taylor
Posts: 595
Joined: 15 Oct 2013, 05:37
First name(s): Peter
Middle name(s): Robert
Last name(s): Taylor
Affiliation: Tianjin University
Country: China

### Re: Calculating spin-orbit matrix elements

There is a line in your MNF-SO calculation output that worries me a bit (but I am not an expert here): it lists the calculation of the various spin-orbit integrals before going on to do the DFT step and the line in questions reads

Code: Select all

``````SO-integrals are calculated for Cr: [Ar]4s^2 3d^4
``````
Now when you calculate Te with an ECP there are 28 electrons that are replaced by the ECP, which then leaves 24 "explicit electrons" . I can accept all that, but the output line seems to imply that it is calculating the MNF-SO integrals for an element with atomic number 24 (Cr). Again, I'm not an expert, if that's what the code is doing there will be a problem, because the integral values will be for a different atom and mean-field occupation.

By the way, your Dalton version (2013.0) is getting very out-of-date, and I recommend whoever looks after it updates to the current version (2016.1). It is difficult to support all older versions of the code, and it can mean (and this might be the case here) that the issue has already been identified and fixed.

Best regards
Pete

Bernd S.
Posts: 17
Joined: 19 Feb 2015, 13:31
First name(s): Bernd
Last name(s): Schimmelpfennig
Affiliation: KIT
Country: Germany

### Re: Calculating spin-orbit matrix elements

Pete correctly spotted the trouble maker. NEVER combine PP with MNF-SO or a more general SO like Breit-Pauli as the nodal structure of the orbitals is just wrong and the screening core orbitals are missing.

taylor
Posts: 595
Joined: 15 Oct 2013, 05:37
First name(s): Peter
Middle name(s): Robert
Last name(s): Taylor
Affiliation: Tianjin University
Country: China

### Re: Calculating spin-orbit matrix elements

Bernd, I'd ask you to check me on what I'm posting here: I don't want to mislead the original poster.

Jenny, if you want to do the spin-orbit this way it appears that neither MNF nor Breit-Pauli is going to be workable, in which case if you want to keep going with it you will have to do it all-electron. This (I would have thought) isn't that big a deal --- a few more basis functions and another 14 doubly occupied orbitals if my maths is right.

However, I looked at the EMSL Basis Set Exchange and all-electron basis sets are pretty thin on the ground for Te. There are one or two smaller TZP-type sets, and sets ADZP, ATZP, AQZP, that seem to be a pretty good-sized, but I have no familiarity with them at all. There is of course the ANO-RCC set from the late Bjoern Roos, but here a word of caution is necessary. This basis set was constructed using the Douglas-Kroll-Hess model for scalar relativistics. It is my limited experience that such sets do not work well (especially for ANOs the contraction coefficients will be wrong) if you run without DKH. But, Bernd, if she runs with DKH, does that impact the subsequent MNF or Breit-Pauli treatment?

Best regards
Pete

Bernd S.
Posts: 17
Joined: 19 Feb 2015, 13:31
First name(s): Bernd
Last name(s): Schimmelpfennig
Affiliation: KIT
Country: Germany

### Re: Calculating spin-orbit matrix elements

In combination with DKH Dalton should switch to the according MNF-SO using Foldy-Wuithuysen instead of Breit-Pauli, so Björn's basis sets should be a fair choice to get some reasonable treatment of the mainly Te-centred SO effects.

JennyP
Posts: 7
Joined: 08 Apr 2016, 16:59
First name(s): Jenny
Last name(s): Pirillo
Affiliation: University
Country: Italy

### Re: Calculating spin-orbit matrix elements

Thank you for your precious suggestions. But, I'd like to know if it is possible to obtain reasonable results regarding the SO effects using an ECP, as I have the same problem for a system in which ruthenium is the heavy atom.

Bernd S.
Posts: 17
Joined: 19 Feb 2015, 13:31
First name(s): Bernd
Last name(s): Schimmelpfennig
Affiliation: KIT
Country: Germany

### Re: Calculating spin-orbit matrix elements

I honestly haven't followed the implementation of PP-SO operators over the years, so I cannot give any advice. From approximately twenty years working on and with SO-MNF I can say it is only completely off for very pathological cases.

JennyP
Posts: 7
Joined: 08 Apr 2016, 16:59
First name(s): Jenny
Last name(s): Pirillo
Affiliation: University
Country: Italy

### Re: Calculating spin-orbit matrix elements

What about using effective core one electron method with PP for obtain SOC? (attached files: EffectiveCoreOneElectron.dal EffectiveCoreOneElectron.out)

taylor
Posts: 595
Joined: 15 Oct 2013, 05:37
First name(s): Peter
Middle name(s): Robert
Last name(s): Taylor
Affiliation: Tianjin University
Country: China

### Re: Calculating spin-orbit matrix elements

I'm not sure what your worry is about ruthenium? It is lighter than Te. Is it that the valence-shell electronic structure is more complicated than Te? That is most likely true, but none of those valence shell problems would solved by an ECP, except for those derived from relativistic effects in the core and that DKH should take care of. And Bjoern's ANO-RCC sets are available for the whole periodic table (but you may need my little aces2dalton utility for basis set conversion if you are using an older version of Dalton).

Best regards
Pete

xiongyan21
Posts: 281
Joined: 24 Sep 2014, 08:36
First name(s): yan
Last name(s): xiong
Affiliation: CENTRAL CHINA NORMAL UNIVERSITY
Country: China

### Re: Calculating spin-orbit matrix elements

Dear Drs. Schimmelpfenning and Taylor

Your opinions are insightful, because although in Dalton2016.1, the SCF iterations are much more efficient and more new trial vectors are used, generally, the obtained results are also substantially different, i.e., disparities possibly being of several orders of magnitude.

olav
Posts: 145
Joined: 28 Aug 2013, 06:20
First name(s): Olav
Last name(s): Vahtras

### Re: Calculating spin-orbit matrix elements

Dear Jenny,
The calculation you enclose is an ECP calculation with a one-electron spin-orbit with scaled charges. The scaled charges are from e.g. Koseki, J Phys Chem A 99, 12764 (1995) where they were parameterized to fit spin-orbit splittings when used with ECPs. You can see in the output that the
scaled charge for Te is 5720, which is what you find in this article. So in this sense your input is correct. But the method is very crude and if it gives any reasonable results for excited state transition moments is another matter. It may just work or it may not.

Regards,
Olav

JennyP
Posts: 7
Joined: 08 Apr 2016, 16:59
First name(s): Jenny
Last name(s): Pirillo
Affiliation: University
Country: Italy

### Re: Calculating spin-orbit matrix elements

Dear Olav, Pete and Bernd,

thanks for your suggestion. I will try to use all-electron basis sets.

Regards,
Jenny

taylor
Posts: 595
Joined: 15 Oct 2013, 05:37
First name(s): Peter
Middle name(s): Robert
Last name(s): Taylor
Affiliation: Tianjin University
Country: China

### Re: Calculating spin-orbit matrix elements

I have left this on this thread, just in case, but Yan, are you saying that the spin-orbit matrix elements "differ by orders of magnitude" in 2016.1 from previous versions? And if so why do you think improved SCF iterations has anything to do with that happening?

If you think there is a problem that is not with spin-orbit, it should be moved to another thread under a topic like bug reports.

Best regards
Pete
P.S. To the best of my knowledge there has been little or no change in the testing done of functionality like spin-orbit recently, which means if the tests passed several years ago, they must still pass or that would be flagged up in nightly testing.

xiongyan21
Posts: 281
Joined: 24 Sep 2014, 08:36
First name(s): yan
Last name(s): xiong
Affiliation: CENTRAL CHINA NORMAL UNIVERSITY
Country: China

### Re: Calculating spin-orbit matrix elements

Dear Dr. Taylor
I mean the disparities of the results obtained by the two methods in Dalton2016.1 also can be several orders of magnitude, the reasons of which
might be what you, Drs.Schimmelpfenning and Vahtras have proposed above.
For example, one method gets
@ Transition moment <B | A - <A> | C> in a.u. for
@ A operator label, symmetry, spin: Z1MNF-SO 1 1
@ B excited state no., symmetry, spin: 3 1 0
@ C excited state no., symmetry, spin: 3 1 1

@ B and C excitation energies, moment: 0.12114850 0.11470039 0.00004751
but the other gets
@ Transition moment <B | A - <A> | C> in a.u. for
@ A operator label, symmetry, spin: Z1SPNSCA 1 1
@ B excited state no., symmetry, spin: 3 1 0
@ C excited state no., symmetry, spin: 3 1 1

@ B and C excitation energies, moment: 0.12114850 0.11470039 0.01723192

Best Regards!
Last edited by xiongyan21 on 21 Apr 2016, 05:26, edited 2 times in total.

taylor
Posts: 595
Joined: 15 Oct 2013, 05:37
First name(s): Peter
Middle name(s): Robert
Last name(s): Taylor
Affiliation: Tianjin University
Country: China

### Re: Calculating spin-orbit matrix elements

You do not post outputs, but your two fragments are from two different calculations --- one is MNF and the other effective-core one-electron. So why should they be the same? (Olav has already commented on issues with the latter.)

Best regards
Pete

taylor
Posts: 595
Joined: 15 Oct 2013, 05:37
First name(s): Peter
Middle name(s): Robert
Last name(s): Taylor
Affiliation: Tianjin University
Country: China

### Re: Calculating spin-orbit matrix elements

Following up briefly, it is difficult to say more without knowing the system in question, but I doubt very much that the SCF convergence behaviour has anything to do with it. Certainly the MNF value is effectively noise.

Best regards
Pete

Bernd S.
Posts: 17
Joined: 19 Feb 2015, 13:31
First name(s): Bernd
Last name(s): Schimmelpfennig
Affiliation: KIT
Country: Germany

### Re: Calculating spin-orbit matrix elements

The MNF calculation was according to the attached output still using PP and the related basis set on Te. Please to the calculation once with DOUGLAS-KROLL and ANO-RCC basis sets. Even VDZP quality should give some pretty good idea about the SO-matrix elements.

qinghua
Posts: 10
Joined: 10 May 2016, 12:39
First name(s): Qinghua
Last name(s): Wu
Affiliation: Anhui Normal University
Country: China

### Re: Calculating spin-orbit matrix elements

Dear everyone,

I am just a novice in Dalton. I want to computing SOC between higher-lying singlet and triplet excited states. May i ask some questions?

1. When i was doing SOC calculation, should i use the ground state or excited state (S1 or T1)?

2. When the symmetry of the molecule is C2, should i use the input?
.ISPABC
2 0 1

Best regard!

Qinghua

olav
Posts: 145
Joined: 28 Aug 2013, 06:20
First name(s): Olav
Last name(s): Vahtras

### Re: Calculating spin-orbit matrix elements

Your reference calculation should be the ground singlet state (i.e. S0). The input option which is not so self-evident refers to A=spin of operator, B, C spin of excited states, i.e. for spin-orbit S-T transitions in excited states it is always .

Code: Select all

``````.ISPABC
1 0 1``````
The point group symmetry C2 has nothing to do with this option.

Regards,
Olav

### Who is online

Users browsing this forum: Bing [Bot] and 7 guests