SMEAR vs. LSD+RELAX_MULTIPLICITY/MULTIPLICITY

[Ivan Gladich]

Dear CP2K users,

 

I would like some clarifications about SMEAR vs. LSD +RELAX_MULTIPLICTY in cp2k. I am trying to simulate a metallic nanoparticle (few tens of atoms) in water

 

I have two (maybe simple) questions:

 

a) From this post 

https://groups.google.com/forum/#!topic/cp2k/GEKLwRQat_M

 

my understanding is that the SMEAR option populates the alpha and beta channel according to a Fermi-Dirac 

 

In LSD + MULTIPLICITY (or LSD + RELAX MULTIPLICITY)

https://groups.google.com/forum/#!searchin/cp2k/smearing$20spin%7Csort:date/cp2k/ugkJLz4PSSI/C-ProBR6BwAJ

 

my understating is that after energy levels are calculated, electrons are placed according to Aufbau principle, from the lowest to the highest in an unrestricted scheme. If MULTIPLICITY is set, the multiplicity is fixed, with RELAX_MULTIPLICITY no

 

Please correct me if my understanding is mistaken.

 

Question: are the two approaches (SMEAR vs. LSD +RELAX_MULTIPLICTY) equivalent? One approach is more appropriate than the other for particular systems? Is the computational cost the same?

 

b) If I use LSD +MULTIPLICTY in the output I get the

 

“Integrated absolute spin density” and  “Ideal and single determinant S**2”

 

That are equations (2.271) and (2.272) in “Modern Quantum Chemistry: Introduction to Advanced Electronic Structure Theory,Attila Bangha Szabo, Neil S. Ostlund “

 

https://groups.google.com/forum/#!topic/cp2k/flEEaseAd7c

 

 

If I see the output of my calculation with SMEAR I do not see such or any other information about the spin state of my system. 

 

Question: Is this expected? How I can easily get information about the spin state of my system when using SMEAR?

 

Thank you very much for any possible help

 

Best regards

Ivan

[Krack Matthias (PSI)]

Dear Ivan

 

a)       SMEAR allows for fractional orbital occupations whereas LSD with RELAX_MULTIPLICITY is meant for systems with integer occupations, e.g. to allow for the spin flip needed to switch automatically from a singlet to a possibly lower lying triplet state. Thus, these schemes are not equivalent.

b)      The expectation value of S**2 is only defined for calculations with integer orbital occupations Therefore this properties is not calculated/printed when SMEAR is active.

 

HTH

 

Matthias

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/80e157d7-5559-4066-bbf5-bca66ce52addo%40googlegroups.com.

[Marcella Iannuzzi]

Dear Ivan, 

in addition to what Matthias already wrote, by default the multiplicity is not fixed with SMEAR + LSD. 

Working with fractional orbital occupations requires the use of diagonalisation schemes, which are in general  more demanding in terms of computational costs if compared with the orbital transformation (OT) method. 

Kind regards

Marcella

[Ivan Gladich]

Dear Matthias and Marcella

thank you very much for your reply. 

Now is much more clear.

However, I still have one doubt.

While in LSD we know the spin state of the system (i.e., S**2 and integrated spin mentioned above) at any step,

it seems to me that in SMEAR we lost such information: as Matthias said, .S**2 and integrated spin cannot be calculated for fractional occupation in SMEAR.

Using SMEAR, is it possible to define some average quantities defining the spin state of the system?

I do not know, for example,  the average occupation of the alpha and beta channels?

Or something else?

Thank you very very much for your replies and help

Best

Ivan 

[Marcella Iannuzzi]

Dear Ivan,

The projected density of states, calculated individually for the alpha and beta channels, give the information on how the electrons are distributed. 

Also the population analysis can provide information, or at least trends, on the fluctuations on the atomic charges and spin.

Kind regards

Marcella

[Ivan Gladich]

Dear Marcella,

This answers my question.

Thank you very much

Grazie

Best regards

Ivan