Divergence of Broyden mixing method without OT

[Mohammad Shakiba]

Hello everyone,

I wanted to print the MOs and I used the Broyden mixing method instead of DIIS while running MD for an optimized perovskite system with OT method. The DIIS mixing diverges quickly and the Broyden mixing does not change significantly and the convergence does not go further than 1.0E-2. My system is a MAPbI3 perovskite consisting of Pb and Iiodine atoms. I use the DZVP-MOLOPT basis set and GTH pseudopotential.

I tried reducing EPS_DEFAULT and change other parameters but did not work. When I use it with the OT method by reducing the EPS_DEFAULT I obtain the convergence but not able to print the MOs containing LUMOs. I even used the wfn file obtained from the OT calculation with Broyden but again it does not work. Also, I used SZV basis set and different types of solvers for POISSON. None of them worked and I am completely confused. Can anyone help me to obtain the convergence? Or any hint so I can print the MOs with LUMOs (since with OT I can print MOs but they do not contain the LUMOs)?

Thanks in advance.

Here is my input:

[hut...@chem.uzh.ch]

Hi

this should work for you (as far as I get from your information).

1) Optimize wfn with OT
2) Restart using Diagonalization and ADDED_MOS
This should be converged in 1) so no SCF iterations
Calculate (print) MOs

regards

Juerg Hutter

--------------------------------------------------------------
Juerg Hutter Phone : ++41 44 635 4491
Institut für Chemie C FAX : ++41 44 635 6838
Universität Zürich E-mail: hut...@chem.uzh.ch
Winterthurerstrasse 190
CH-8057 Zürich, Switzerland
---------------------------------------------------------------

-----cp...@googlegroups.com wrote: -----
To: "cp2k" <cp...@googlegroups.com>
From: "Mohammad Shakiba"
Sent by: cp...@googlegroups.com
Date: 10/31/2019 11:19AM
Subject: [CP2K:12437] Divergence of Broyden mixing method without OT

--
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/273264c3-b84c-4dbd-8736-74df084aa9c4%40googlegroups.com.


[attachment "B.inp" removed by Jürg Hutter/at/UZH]

[Mohammad Shakiba]

Dear Juerg Hutter,

Thanks for you reply. I tried this but it does not converge. I used this method for other systems (for example for CdSe or PbSe quantum dots) but when I use it for perovskite, it does not work. I thought it probably is because of iodine, so changed the iodine to Br or Cl but did not change anything. Now I wonder that it is because of my molecular structure since I have optimized the structure using OT method. Also, I have run MD at 300K and the MD results are good. Do you think that the structure is my problem and I have to change it? The same structure image is shown below:

Thanks and Best.

 To unsubscribe from this group and stop receiving emails from it, send an email to cp...@googlegroups.com.

[hut...@chem.uzh.ch]

Hi

if the restart with Diagonalization is not converged, this means
that your OT has converged to an excited state (assuming no other trivial
mistakes of handling files etc.).

You can check this with this procedure:

1) converge using OT
2) restart using Diagonalization, do only 1 SCF step
3) restart using OT, you should now get a lower energy.

This usually is possible for systems with a very small gap.
Try Diagonalization using smearing in that case.

regards

Juerg Hutter
--------------------------------------------------------------
Juerg Hutter Phone : ++41 44 635 4491
Institut für Chemie C FAX : ++41 44 635 6838
Universität Zürich E-mail: hut...@chem.uzh.ch
Winterthurerstrasse 190
CH-8057 Zürich, Switzerland
---------------------------------------------------------------

-----cp...@googlegroups.com wrote: -----
To: "cp2k" <cp...@googlegroups.com>
From: "Mohammad Shakiba"
Sent by: cp...@googlegroups.com

Date: 10/31/2019 12:27PM
Subject: Re: [CP2K:12439] Divergence of Broyden mixing method without OT

Dear Juerg Hutter,

Thanks for you reply. I tried this but it does not converge. I used this method for other systems (for example for CdSe or PbSe quantum dots) but when I use it for perovskite, it does not work. I thought it probably is because of iodine, so changed the iodine to Br or Cl but did not change anything. Now I wonder that it is because of my molecular structure since I have optimized the structure using OT method. Also, I have run MD at 300K and the MD results are good. Do you think that the structure is my problem and I have to change it? The same structure image is shown below:

Thanks and Best.

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/d0a88369-6d46-4d04-980e-6e6cb0759b39%40googlegroups.com.

[Mohammad Shakiba]

Dear Juerg Hutter ,

Thanks for your useful information. You're right. I checked it and it seems it has converged to an excited state. I think the problem is because of my molecular structure since smearing also does not converge.

Thanks again and Best.

 To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/d0a88369-6d46-4d04-980e-6e6cb0759b39%40googlegroups.com.
 

[Mohammad Shakiba]

Dear Juerg Hutter,

I used Fermi-Dirac smearing and the calculation converged successfully with the Broyden mixing method. Also, the MOs were printed with LUMOs.

Thanks again and Best.

 To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/d0a88369-6d46-4d04-980e-6e6cb0759b39%40googlegroups.com.