QuantumATK Forum

QuantumATK => General Questions and Answers => Topic started by: Manish95 on January 31, 2019, 07:23

Title: Geometry optimization not converging
Post by: Manish95 on January 31, 2019, 07:23

I have a 5x5 supercell monolayer MoS2, on top of it I have placed 2 Cr atoms (cleaved along 100) after converting them into molecule. I want to study the effect of Cr on DOS of MoS2. But the geometry optimization didn't converge. Any inputs?

Title: Re: Geometry optimization not converging
Post by: Petr Khomyakov on January 31, 2019, 15:48

What do you mean "did not converge"? Please attach your script and log file.

Title: Re: Geometry optimization not converging
Post by: Manish95 on February 1, 2019, 10:26

The calculation was running for more than 12 hours even though I ran the simulation on 4 cores. So I had to kill the process. I am unable to attach both log and script file due to size limitation here.

Title: Re: Geometry optimization not converging
Post by: Ulrik G. Vej-Hansen on February 1, 2019, 12:46

Sounds like you should give it more time - 12 hours on 4 cores is probably not enough. Hard to say more without the log though.

I would also recommend that you use the PseudoDojo pseudopotentials instead. The OMX ones can be quite hard as described here: https://docs.quantumwise.com/manual/technicalnotes/atk_pseudo_basis/atk_pseudo_basis.html

Title: Re: Geometry optimization not converging
Post by: Petr Khomyakov on February 1, 2019, 13:31

In addition, I think you do not really have to relax volume of your supercell, just ions . It would make sense to relax MoS2 lattice parameter before introducing any foreign atoms.

Also, when placing these 2 Cr atoms, one should have a decent initial guess for their positions on the MoS2 surface, e.g., put one atom on a hollow site of MoS2 hexagonal lattice, and the second Cr atom on some other hollow site. Or attach them to Mo or S atoms, or think of any other arrangement that would make sense from the chemistry point of view.

I guess you should try various options to find the most stable, low-energy configuration(s). 

Title: Re: Geometry optimization not converging
Post by: Manish95 on February 1, 2019, 13:57

Quote from: Ulrik G. Vej-Hansen on February  1, 2019, 12:46

Sounds like you should give it more time - 12 hours on 4 cores is probably not enough. Hard to say more without the log though.

I would also recommend that you use the PseudoDojo pseudopotentials instead. The OMX ones can be quite hard as described here: https://docs.quantumwise.com/manual/technicalnotes/atk_pseudo_basis/atk_pseudo_basis.html

I am using ATK version 2017.2 and it does not have PseudoDojo Pseudopotential. My band structure calculations for MoS2 were done with OMX PP for Mo and S atoms. So I was using OMX for Cr also. Can I use a different set of PP for Mo, S atoms and some other for Cr atom?

Title: Re: Geometry optimization not converging
Post by: Petr Khomyakov on February 1, 2019, 16:25

You may also use SG15 pseudopotentials in 2017.2, which will perhaps be as good as PseudoDojo for your system.

Title: Re: Geometry optimization not converging
Post by: Manish95 on February 2, 2019, 06:30

I got band gap of 1.81 eV for monolayer MoS2 using OMX PP for both Mo and S atoms.  Shall I change to SG15 PP for all the atoms or only for Cr (and keep OMX PP for Mo and S atoms)?

Title: Re: Geometry optimization not converging
Post by: Petr Khomyakov on February 3, 2019, 21:16

It would make sense to use the same pseudopotential for all the elements.