Re: [Wien] regarding k-mesh and tolf

2018-01-27 Thread Xavier Rocquefelte 
Could you give the chemical formula? It will really help answering your 
question from my side.


Cheers

Xavier


Le 27/01/2018 à 14:34, chin Sabsu a écrit :

Dear Respected Xavaier Sir
Thank you very much for detailed explanation.
My system is not having any d-d interaction (no TM element) so I do 
not think I should go for  DFT+U, Onsite Hybrid functional and Full 
hybrid calculations!


With TB-mBJ it is showing a band gap. The VBM is widely dispersed, 
flat for a wide range, near the VBM. So this is a semiconductor.


As we can not relax the structure using mBJ so there is no way to use 
it for the phonopy purpose (phonon bands).


Due to lack of good computational facilities, I want to avoid hybrid 
calculation which will never finish if I go for Phonopy.


What I see it, I use the combination of tolf and -fc such that it 
gives minimum forces. But I am sure the forces will not be less than 
2mRy/Bohr. Again this process I see a contradiction with Peter Sir's 
statement: forces should be less than 1mRy/bohr for phonon 
calculations and which I will not be having!!
How much the ~2-3mRy/bohr forces on the atom affects the phonon band 
structure?
can I proceed for phonon bands with the structure having forces ~2-3 
mRy/Bohr?


Thank you very much!

Chin S.




On Saturday 27 January 2018, 6:31:35 PM IST, Xavier Rocquefelte 
 wrote:



Dear Chin S.

Your system is metallic from the point of view of DFT which is not a 
surprise.


In your last calculation, which is the more accurate you simply catch 
this fact ... In other words in your previous calculations you had 0.1 
eV gap due to less kpoints and/or less accurate calculations. To 
really understand what is happening you should draw the DOS and 
bandstructure to see if you have band crossing at the Fermi level 
which could lead to occupancy of an expected empty band. If yes, it 
will have an impact on the chemical bond description and also on the 
phonons.


Depending on the system you will have to consider methods going beyond 
DFT, which in WIEN2k could be based on:


- DFT+U

- Onsite Hybrid functional

- Full hybrid

- TB-mBJ potential

If your band gap is based on charge transfer transitions you cannot 
used DFT+U and onsite-hybrid.
Thus the less expensive approach will be TB-mBJ potential which will 
give the band gap but not the forces.


Thus only one option remains which is full hybrid.

If your system is based on d-d transitions, or highly correlated 
states, then you can used DFT+U or Onsite-hybrid.


Hope it will help you.

Cheers

Xavier


Le 27/01/2018 à 12:28, chin Sabsu a écrit :

Dear Stefaan Sir

Below are my updates:

A. For mesh size
I got your point and in all different distorted structure (with 
different inequivalent atoms) I am having the same number of 
electrons while the number of inequivalent atoms are different. NE 
varies from 154 to 616(1x1x1 cell) to 4928 (2x2x2 cell).
So I will divide xxx numk by 4 and 32 in 1x1x1 and 2x2x2 supercell 
structure. Thanks for this excellent recommendation!!!
For 2x2x2 supercell in phonopy calculations, we need to go for mpi 
jobs for certain cases!!


B. Now I have done some tests for my cubic perovskite and below are 
my observations:


B1. For tolf 5 and -fc1 I for FOR003 = ~4 mRy / bohr   with band gap 
~0.1 eV
B2. For tolf 2 and - fc 0.1 forces are almost same. and the band gap 
is still ~0.1ev


B3. Now to reduce the forces less then 0.1mRy/ bohr , I reduced the 
tolf to 0.5 and - fc 0.1 - 0.01. I got forces at the desired accurecy 
but the band gap vanished and the system becomes metallic. The 
experimentally reported band gap is 1.2eV

I used RKMAX 8, numk 2000, vxc 19.

What is your suggestion for point number B, why I am getting the 
system metallic?

Can I use this relaxed structure for Phonopy?

Kind regards

Chin S.


For t olf



___
Wien mailing list
Wien@zeus.theochem.tuwien.ac.at 
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
SEARCH the MAILING-LIST 
at:http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html


___
Wien mailing list
Wien@zeus.theochem.tuwien.ac.at 
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
SEARCH the MAILING-LIST at: 
http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html



___
Wien mailing list
Wien@zeus.theochem.tuwien.ac.at
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
SEARCH the MAILING-LIST at:  
http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html


___
Wien mailing list
Wien@zeus.theochem.tuwien.ac.at
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
SEARCH the MAILING-LIST at:  
http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html

Re: [Wien] regarding k-mesh and tolf

2018-01-27 Thread Xavier Rocquefelte 

Dear Chin S.

Your system is metallic from the point of view of DFT which is not a 
surprise.


In your last calculation, which is the more accurate you simply catch 
this fact ... In other words in your previous calculations you had 0.1 
eV gap due to less kpoints and/or less accurate calculations. To really 
understand what is happening you should draw the DOS and bandstructure 
to see if you have band crossing at the Fermi level which could lead to 
occupancy of an expected empty band. If yes, it will have an impact on 
the chemical bond description and also on the phonons.


Depending on the system you will have to consider methods going beyond 
DFT, which in WIEN2k could be based on:


- DFT+U

- Onsite Hybrid functional

- Full hybrid

- TB-mBJ potential

If your band gap is based on charge transfer transitions you cannot used 
DFT+U and onsite-hybrid.
Thus the less expensive approach will be TB-mBJ potential which will 
give the band gap but not the forces.


Thus only one option remains which is full hybrid.

If your system is based on d-d transitions, or highly correlated states, 
then you can used DFT+U or Onsite-hybrid.


Hope it will help you.

Cheers

Xavier


Le 27/01/2018 à 12:28, chin Sabsu a écrit :

Dear Stefaan Sir

Below are my updates:

A. For mesh size
I got your point and in all different distorted structure (with 
different inequivalent atoms) I am having the same number of electrons 
while the number of inequivalent atoms are different. NE varies from 
154 to 616(1x1x1 cell) to 4928 (2x2x2 cell).
So I will divide xxx numk by 4 and 32 in 1x1x1 and 2x2x2 supercell 
structure. Thanks for this excellent recommendation!!!
For 2x2x2 supercell in phonopy calculations, we need to go for mpi 
jobs for certain cases!!


B. Now I have done some tests for my cubic perovskite and below are my 
observations:


B1. For tolf 5 and -fc1 I for FOR003 = ~4 mRy/bohr  with band gap ~0.1 eV
B2. For tolf 2 and -fc 0.1 forces are almost same. and the band gap is 
still ~0.1ev


B3. Now to reduce the forces less then 0.1mRy/bohr, I reduced the tolf 
to 0.5 and -fc 0.1 - 0.01. I got forces at the desired accurecy but 
the band gap vanished and the system becomes metallic. The 
experimentally reported band gap is 1.2eV

I used RKMAX 8, numk 2000, vxc 19.

What is your suggestion for point number B, why I am getting the 
system metallic?

Can I use this relaxed structure for Phonopy?

Kind regards

Chin S.


For tolf



___
Wien mailing list
Wien@zeus.theochem.tuwien.ac.at
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
SEARCH the MAILING-LIST at:  
http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html


___
Wien mailing list
Wien@zeus.theochem.tuwien.ac.at
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
SEARCH the MAILING-LIST at:  
http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html

[Wien] regarding k-mesh and tolf

2018-01-27 Thread chin Sabsu 
Dear Stefaan Sir
Below are my updates:
A. For mesh sizeI got your point and in all different distorted structure (with 
different inequivalent atoms) I am having the same number of electrons while 
the number of inequivalent atoms are different. NE varies from 154 to 616(1x1x1 
cell) to 4928 (2x2x2 cell).So I will divide xxx numk by 4 and 32 in 1x1x1 and 
2x2x2 supercell structure. Thanks for this excellent recommendation!!!For 2x2x2 
supercell in phonopy calculations, we need to go for mpi jobs for certain 
cases!!

B. Now I have done some tests for my cubic perovskite and below are my 
observations:
B1. For tolf 5 and -fc1 I for FOR003 = ~4 mRy/bohr  with band gap ~0.1 eV
B2. For tolf 2 and -fc 0.1 forces are almost same. and the band gap is still 
~0.1ev

B3. Now to reduce the forces less then 0.1mRy/bohr, I reduced the tolf to 0.5 
and -fc 0.1 - 0.01. I got forces at the desired accurecy but the band gap 
vanished and the system becomes metallic. The experimentally reported band gap 
is 1.2eV
I used RKMAX 8, numk 2000, vxc 19.
What is your suggestion for point number B, why I am getting the system 
metallic?Can I use this relaxed structure for Phonopy?
Kind regards
Chin S.
 
For tolf

 ___
Wien mailing list
Wien@zeus.theochem.tuwien.ac.at
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
SEARCH the MAILING-LIST at:  
http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html