Doubt about M062X functional with LIBXC and proper Pseudopotentials

[Lucas Lodeiro]

Hello all,

I want to run some calculations with the M062X functional, but I am not acquainted with functionals from LIBXC. Particularly I am confused about the exchange part, which is called:

&XC

  &XC_FUNCTIONAL

&LIBXC

  FUNCTIONAL HYB_MGGA_X_M06_2X 

  &END LIBXC

  &END

&END

but, M062X functional uses just 46% of mGGA exchange, then I think it is necessary to scale it:

&XC

  &XC_FUNCTIONAL

&LIBXC

  FUNCTIONAL HYB_MGGA_X_M06_2X

      SCALE 0.46

  &END LIBXC

  &END

&END

Is the latter correct? or is the LIBXC exchange already "correct" for M062X?

About Pseudopotentials, I want to use the MOLOPT basis sets, then the GTH_pp is mandatory, and M062X based pp are not present. Which PP is the best for M062X? PBE, PBE0?

Regards - Lucas Lodeiro

[Krack Matthias (PSI)]

Hi Lucas

 

I think rather the first version without the explicit SCALE parameter is right, because LIBXC already includes the correct MGGA exchange scaling parameter 0.46 as _A0. Instead, you should add the &HF section with FRACTION 0.54 for the matching EXX contribution. I am not aware of any M06-2X optimized PPs for CP2K. In case, you have only light elements, you may consider all-electron runs for which you can find validation data in the literature.

 

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/CAOFT4P%2BsfDL5yrH0-HGsf0W4RuwxzM7T4sBzjebrfEKtXy53CQ%40mail.gmail.com.

[Lucas Lodeiro]

Hi Matthias,

Thanks for your quick response. Sadly, I have an Iodine atom in the system, so it is mandatory to use PPs. I read from the seminal paper of M062X, the exchange functional is based on the PBE exchange form... Then I guess PBE and mainly PBE0 PPs will work for M062X. What do you think? 

Somebody has used M062X + pseudopotentials?

Regards - Lucas Lodeiro

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/801deac568d34123804008e89f1d8d6f%40psi.ch.

[Filipe Menezes]

Hi Lucas,

if you don't mind my answering with a question, why do you need M06-2X, is there any particular system you are interested in, for which the functional is particularly adequate?

Though I never used the functional with PPs, I suppose that the "correctness" of the calculation will depend on the amount of exchange in the functional. This means, though you use PBE exchange, you still mix it with HF, which should correspondingly change the properties and characteristics of the PP. My suggestion is to try to find some similar model system, for which you can benchmark if possible. You can potentially use thast to tune your calculation.

Best

Filipe

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CAOFT4PKooBxx%2BRO5OiHbK4ndO6gwr_H69FewZ_iFpz1PCN8Adg%40mail.gmail.com.

[Lucas Lodeiro]

Hi Filipe,

yes, it is not a problem. I need to use M062X because the system has a complex electronic structure. Other authors showed the system is highly multirefence, and they could match their MR calculations using M062X... I try a couple of different functionals and they crash finding the electronic structure or the desired geometries... this is why I need M062X.

Probably I will do some benchmarks, but even if they works fine, it is not fully clear if it will work consistently. That is why I am asking some advice based on experience.

About the correctness related to the EXX amount, yes, it is probably true, but it is not so clear where the line of no return be... based on that idea, I guess PBE0 PPs would be better than PBE ones.

regards - Lucas

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CAE9PG0rU%3Dwy5vCFBwxJT0%3DXXMzrqVk%2BvATVpuRZkzadph6UYZg%40mail.gmail.com.

[Filipe Menezes]

Hi Lucas.

if you search for multi-reference in this pdf 

https://www.tandfonline.com/doi/pdf/10.1080/00268976.2017.1333644?needAccess=true (this is the doi if you want to google instead 10.1080/00268976.2017.1333644) 

you will find that M06-2X might give large errors for multireference cases. I worked recently on some borderline MR cases and I was told by a former colleague to try M06-2X (borderline means T1 diagnostics of 0.02). He used it for some reaction energies involving ozone and he said it matched perfectly. In my case M06-2X failed relatively badly, to the point I could not really use it. And it makes sense that it "fails", since the amount of HF is relatively large. Maybe you could try some of the other functionals with less HF exchange? Why not PBE0 itself? In my case it worked qualitatively well. Sometimes even quantitative!

MR is a real pain in the neck =).

Best

Filipe

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CAOFT4PKxjQA_HEDEJ_B7bqOFLAar44UtvdXKq5FhtAj9oeWjsA%40mail.gmail.com.

[Lucas Lodeiro]

Thanks for the paper,

I see MR is complex for high HF exchange functionals, but in this particular case M062X works very well and quantitatively... instead GGA functionals just qualitatively. I will explore more to decide which of them I will use.

Regards

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CAE9PG0pdwccaKuqu1TMTCKA_C46n%2BNnHu_Wu-1ub8GksNbsVYA%40mail.gmail.com.

[Filipe Menezes]

You welcome, I hope I was of some help.

Best

Filipe

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CAOFT4PJE5nhgKzbnta3YypbZe2zJey4q21uo-G478EkYz%3D_6Uw%40mail.gmail.com.