Sm Pseudopotentials used for GGA and HSE06

Problems running VASP: crashes, internal errors, "wrong" results.

Moderators: Global Moderator, Moderator

Post Reply
Message
Author
marykate_
Newbie
Newbie
Posts: 4
Joined: Wed Dec 01, 2021 1:35 pm

Sm Pseudopotentials used for GGA and HSE06

#1 Post by marykate_ » Mon Dec 06, 2021 7:22 pm

I'm looking at an oxide with many rare-earth cations which include Ce, La, Pr, and Sm. I know the difficulties of simulating these f-element electrons in DFT, but I've been having problems with Sm specifically.

I've been using VASP 5.4.4 for GGA calculations as well as HSE06 calculations using PAW-PBE for both.

My question is about the compatibilities of the two available Sm pseudopotentials, [Sm_3] and [Sm] with GGA and HSE06. I'm aware of the differences in the valence electrons between these two pseudopotentials. [Sm_3] has been optimized for Sm3+ and has 1 f-electron in valence. The [Sm] pseudopotential has all 6 f-electrons in valence.
It seems that the [Sm_3] pseudopotential works with GGA but not HSE06, and the [Sm] pseudopotential might work with HSE06 but not GGA.

Ultimately, I wanted to know if my only option is to run my HSE06 from scratch since it isn't possible to start from a converged GGA run for Sm (and anything that contains Sm like my material) due to incompatible Sm pseudopotentials. Likewise, I can only run my GGA calculations with an [Sm_3] pseudopotential optimized for the 3+ state.

I first had an issue when I was trying to run a HSE06 calculation from a converged GGA run using the [Sm_3] pseudopotential which resulted in these errors:
" internal error in SET_CORE_WF: core electrons incorrect 51.0000000000000
62.0000000000000"
"Error EDDDAV: Call to ZHEGV failed. Returncode = 14 2 18"
And I came across this post on the forum: forum/viewtopic.php?t=11307
So, I began to assume that the [Sm_3] pseudopotential is not compatible with HSE06.

I'd prefer to run my HSE06 from a converged GGA, which I know won't work using [Sm_3].
When I tried to run GGA with the [Sm] potential, I got the following errors:
"WARNING: CNORMN: search vector ill defined"
"Error EDDDAV: Call to ZHEGV failed. Returncode = 16 2 18"

I thus found the following forum post: https://w.vasp.at/forum/viewtopic.php?p=19871
So, I thought there might be issues in the code for Sm similar to Nd.

I also am still having some trouble running an HSE06 calculation from scratch using the [Sm] pseudopotential for my material with all the RE elements. However, I have gotten HSE06 to run for a plain Sm oxide with the [Sm] pseudopotential, but not the [Sm_3] pseudopotential.

To summarize:
The [Sm_3] pseudopotential works with GGA but not HSE06, and the [Sm] pseudopotential might work with HSE06 but definitely not GGA. These issues do not occur for Pr which also has two different f-electron containing pseudopotentials, [Pr_3] and [Pr]. Both of these pseudopotentials work in GGA and HSE06, which is not the case for Sm. I'd appreciate understanding why this is occuring, and I'm curious why I seem to only be having this issue with Sm.

I have attached the relevant zip files (as many as I could attach to the post). Thank you very much!
You do not have the required permissions to view the files attached to this post.

henrique_miranda
Global Moderator
Global Moderator
Posts: 414
Joined: Mon Nov 04, 2019 12:41 pm
Contact:

Re: Sm Pseudopotentials used for GGA and HSE06

#2 Post by henrique_miranda » Thu Dec 09, 2021 8:18 pm

Thank you for the detailed error report.
I will be looking at it.

The first thing that stands out for me is your choice of pseudopotentials.
Perhaps you could try running this system using the pseudos indicated in boldface on this page:
wiki/index.php/Available_PAW_potentials

marykate_
Newbie
Newbie
Posts: 4
Joined: Wed Dec 01, 2021 1:35 pm

Re: Sm Pseudopotentials used for GGA and HSE06

#3 Post by marykate_ » Fri Dec 10, 2021 2:43 pm

Thank you for the reply!

I have looked at those pseudopotentials. The [Sm_3] works easily in GGA, but it won't work in HSE06. I'd like to run a GGA calculation using the [Sm] rather than the [Sm_3] pseudopotential made for an Sm3+ state. I was hoping to run my HSE06 from a converged GGA run which seems to be that HSE06 will only run with the [Sm] pseudopotential, and I haven't been able to run GGA with [Sm]. Samarium has been the only rare-earth element in my oxide material giving me these issues.

I've also attached additional .zip files where I didn't have any errors in case they may help in addition to my error files previously attached. Thank you very much for your help! It is greatly appreciated.

-Mary Kate
You do not have the required permissions to view the files attached to this post.

henrique_miranda
Global Moderator
Global Moderator
Posts: 414
Joined: Mon Nov 04, 2019 12:41 pm
Contact:

Re: Sm Pseudopotentials used for GGA and HSE06

#4 Post by henrique_miranda » Sat Dec 11, 2021 2:42 pm

What I meant was that you could try using only pseudopotentials from the recommended list.
That would mean changing the Pr to Pr_3 as well.
Maybe in that case, both the DFT and hybrid calculations would converge.

But if there is a particular reason you don't want to use these POTCARs then the calculations are going to be a bit more tricky because of the f electrons.

First, try to get the DFT calculation running. You might have to run it in a few steps:
1. Deactivate the ionic relaxation tags, reduce the cutoff (ENCUT=200 eV for example), set ICHARG=12 to obtain an initial WAVECAR file.
2. Use the WAVECAR from step 1 to restart a calculation with the ENCUT=500 eV you wanted.
3. Remove ICHARG=12, start from the previous WAVECAR and run the calculation with ALGO=All and TIME=0.1 or 0.05.

We collected some of these ideas on this page
wiki/index.php/Diffcult_to_converge_systems

Note that these are just general guidelines, you might have to adapt a bit the steps for your particular system.
It is a good idea to first try these steps with a gamma-only calculation because the calculations run faster.
Once you find a recipe that works you can try the same steps with the k-mesh you want.

Let me know if this works for you and good luck!

marykate_
Newbie
Newbie
Posts: 4
Joined: Wed Dec 01, 2021 1:35 pm

Re: Sm Pseudopotentials used for GGA and HSE06

#5 Post by marykate_ » Sat Feb 05, 2022 10:19 pm

Hello!

Here is an update with some questions.

I followed the steps you suggested and I managed to converge the systems that contain Sm in GGA using the [Sm] pseudopotential! The only aspect of the procedure that didn't work for me was step 2; I wasn't able to use the WAVECAR from step 1 to restart the calculation with ENCUT=500eV and it resulted in the same errors. However when I skip step 2 and use the WAVECAR from step 1 for step 3, the system will converge properly.

I am currently still having some issues. I continue to have problems with running HSE06 for Sm and anything that contains Sm with the [Sm] pseudopotential (as already discussed in my previous post the [Sm_3] potential is not compatible with HSE06). I used the pre-converged WAVECAR, CONTCAR, and CHGCAR to initialize the HSE06 run, and despite anything I tried I continued to get the following errors again:
"WARNING: CNORMN: search vector ill defined"
"LAPACK: Routine ZPOTRF failed!"

I tried applying the procedure similar to the one that worked for GGA, and I ran into the same errors. I also attempted reducing TIME and removing symmetry, but I still have those errors in HSE06.
Additionally, I've been attempting to run DOS for Sm, using the same procedure that worked for GGA, but I haven't been able to get that to run properly either.

The relevant folders are attached which includes the properly converged GGA, the error HSE06, and the error DOS. Any advice would be greatly appreciated. Thank you so much!
You do not have the required permissions to view the files attached to this post.

Post Reply