Input file for GaAs geometry optimization


  • Guest -
11:19:50 PM PDT - Sun, May 29th 2011
Hi all,
can you suggest an input file to do DFT geometry optimization for GaAs? I can't get it working (something about the choice of basis set, functional). Thank you!

Forum Vet
4:53:24 PM PDT - Wed, Jun 1st 2011
Gaussian DFT or plane waves? Could you copy the input file in a post so we have some information we can use to help you?

Thanks,

Bert


Quote: May 30th 6:19 am
Hi all,
can you suggest an input file to do DFT geometry optimization for GaAs? I can't get it working (something about the choice of basis set, functional). Thank you!

  • Guest -
2:19:27 AM PDT - Thu, Jun 2nd 2011
GaAs
Hi Bert,

My trial input looks like:

Jun 2nd 9:19 am Local TimeJun 2nd 9:19 am Local TimeJun 2nd 9:19 am Local TimeJun 2nd 9:19 am Local TimeJun 2nd 9:19 am Local TimeJun 2nd 9:19 am Local Time~
title "GaAs"

geometry units angstroms
Ga 0.0 0.0 -2.5
As 0.0 0.0 2.5
end

basis
Ga library sto-3g
As library sto-3g
end

dft; direct; end

task dft optimize


The goal is to optimize accurately the GaAs structure (using quality DFT functional). Either PW or Gaussian (any would do,
or as required by accuracy goal).

Forum Vet
3:55:33 PM PDT - Fri, Jun 3rd 2011
Well, a couple of things here. Having a Ga and As separated by 5 angstroms is creating some of your problems, try a more realistic bond length like 2.0, i.e.

geometry units angstroms
Ga 0.0 0.0 -1.0
As 0.0 0.0 1.0
end

You will need to increase the number of iterations:

dft; iterations 300; direct; end

Also, I hope you're going to use a reasonable basis set and not the minimal sto-3g.

You hinted at the calculation not being able to find a basis set. If this is another issue, you should set NWCHEM_BASIS_LIBRARY to point to the location where the NWChem basis sets are on your machine (if you build from source it will be in <nwchem-path>/src/basis/libraries)

  setenv NWCHEM_BASIS_LIBRARY <nwchem-path>/src/basis/libraries/

Hope this helps,

Bert

Quote: Jun 2nd 9:19 am
Hi Bert,

My trial input looks like:

Jun 2nd 9:19 am Local TimeJun 2nd 9:19 am Local TimeJun 2nd 9:19 am Local TimeJun 2nd 9:19 am Local TimeJun 2nd 9:19 am Local TimeJun 2nd 9:19 am Local Time~
title "GaAs"

geometry units angstroms
Ga 0.0 0.0 -2.5
As 0.0 0.0 2.5
end

basis
Ga library sto-3g
As library sto-3g
end

dft; direct; end

task dft optimize


The goal is to optimize accurately the GaAs structure (using quality DFT functional). Either PW or Gaussian (any would do,
or as required by accuracy goal).

Forum Regular
8:14:33 PM PDT - Sat, Jun 4th 2011
Hi,

Are you trying to do perform a periodic or molecular calculation ?

Our Gaussian basis DFT module is for molecular systems while the planewave module can handle periodic systems. Your input specifications above will invoke the Gaussian DFT module. In other words, the calculation will only be performed on 2 atoms.

Best,
-Niri
niri.govind@pnnl.gov

  • Guest -
11:14:44 PM PDT - Sun, Jun 5th 2011
Progress on GaAs
Hi Bert and Niri,

thanks for your quick replays. I reply to each below in detail. Additionally, do you know any way in NWChem to access 1) the band occupation numbers (ie, to change them at will) and 2) to calculate dipole transition moments (ie, <km|x,y,z|k'n> or momentum operator transitions)? Even to have the wave functions numerically could help by doing the dipole elements with quadrature (though not so accurately).

For Bert:
I went to 2 Angstr for the bond length. I tried some aug-cc-pvdz. Not much success. Could you suggest a suitable basis set for GaAs? At this point, I would be happy to see a converged calculation (ground state energy) at a given bond length (say 2 Angstrom).

For Niri:
A molecular calculation is wanted. As you can see above, I am missing some good basis set for GaAs. Basically, i would like to see a converged energy calculation at fixed bond length.

Thank you!

Forum Vet
3:47:30 PM PDT - Thu, Jun 9th 2011
The input deck below works fine for me:

title "GaAs"

geometry units angstroms
Ga 0.0 0.0 -1.5
As 0.0 0.0 1.5
end

basis
Ga library aug-cc-pvdz
As library aug-cc-pvdz
end


dft
iterations 300
end

task dft optimize


The first DFT converges in 29 steps. The geometry optimization converges in 7 steps to a bond length of 2.15429 Angstrom.

If you cannot reproduce this I wonder if something is up with your binary.

Bert




Quote: Jun 6th 6:14 am
Hi Bert and Niri,

thanks for your quick replays. I reply to each below in detail. Additionally, do you know any way in NWChem to access 1) the band occupation numbers (ie, to change them at will) and 2) to calculate dipole transition moments (ie, <km|x,y,z|k'n> or momentum operator transitions)? Even to have the wave functions numerically could help by doing the dipole elements with quadrature (though not so accurately).

For Bert:
I went to 2 Angstr for the bond length. I tried some aug-cc-pvdz. Not much success. Could you suggest a suitable basis set for GaAs? At this point, I would be happy to see a converged calculation (ground state energy) at a given bond length (say 2 Angstrom).

For Niri:
A molecular calculation is wanted. As you can see above, I am missing some good basis set for GaAs. Basically, i would like to see a converged energy calculation at fixed bond length.

Thank you!

  • Guest -
11:00:35 PM PDT - Thu, Jun 9th 2011
Working well so far...
Hi Bert,


Once optimized, is it now difficult to do a band structure plot? Or is that not implemented in NWChem? I saw that one can give the individual k point when asking for the energy levels (even then, I am not sure i know where to read for the energy levels at that given k -point). Here is what I used (and working):

title "GaAs"

geometry units angstroms
Ga 0.0 0.0 0
As 0.0 0.0 2.15429
end

basis
Ga library aug-cc-pvdz
As library aug-cc-pvdz
end


dft

iterations 300

end

task dft energy


NWPW

BRILLOUIN_ZONE
  ZONE_NAME zone_1

  KVECTOR 0.0 0.0 0.0 1.0
END

END

Forum Regular
5:53:28 AM PDT - Fri, Jun 10th 2011
Hi,

You performed a Gaussian basis set DFT calculation. Our Gaussian basis set module is purely for molecular or finite systems. If you want a band structure, you need to run the plane wave module for this system

Check the following tutorial in the Plane Wave DFT documentation page. It shows you how to performa band structure calculation on diamond.

http://nwchemgit.github.io/index.php/NWPW#Using_BAND_to_Calculate_the_Band_Structures_of_Dia...

Best,
-Niri

  • Guest -
12:41:51 AM PDT - Mon, Jun 13th 2011
Thank you. For the PW approach, I encounter some difficulties with running the script. I put below the file for the band structure (inspired from the diamond example). At some points, it crashes

"0:0:cpsi_data_alloc: out of heap memory:: 0 ..."

title "GaAs"

permanent_dir ./

start GaAs

memory 3000 mb

geometry center noautosym noautoz print 
 system crystal 

   lat_a 5.4d0

   lat_b 5.4d0 

   lat_c 5.4d0 

   alpha 90.0d0 

   beta  90.0d0 

   gamma 90.0d0 

 end

Ga  0.00000d0  0.00000d0  0.00000d0

As  0.25000d0  0.25000d0  0.25000d0
end

nwpw
 ewald_rcut 3.0

 ewald_ncut 8    #The default value of 1 needs to be increased

 lmbfgs

 xc pbe96



 monkhorst-pack 9 9 9
end

task band energy

nwpw
  virtual 16

  brillouin_zone

    zone_name fccpath

    path fcc l gamma x w k gamma

  end

  zone_structure_name fccpath
end

task band structure

Forum Vet
2:28:33 PM PDT - Mon, Jul 18th 2011
Try "memory heap 1400 stack 1500 global 100 mb"

Bert



Quote: Jun 13th 7:41 am
Thank you. For the PW approach, I encounter some difficulties with running the script. I put below the file for the band structure (inspired from the diamond example). At some points, it crashes

"0:0:cpsi_data_alloc: out of heap memory:: 0 ..."

title "GaAs"

permanent_dir ./

start GaAs

memory 3000 mb

geometry center noautosym noautoz print 
 system crystal 

   lat_a 5.4d0

   lat_b 5.4d0 

   lat_c 5.4d0 

   alpha 90.0d0 

   beta  90.0d0 

   gamma 90.0d0 

 end

Ga  0.00000d0  0.00000d0  0.00000d0

As  0.25000d0  0.25000d0  0.25000d0
end

nwpw
 ewald_rcut 3.0

 ewald_ncut 8    #The default value of 1 needs to be increased

 lmbfgs

 xc pbe96



 monkhorst-pack 9 9 9
end

task band energy

nwpw
  virtual 16

  brillouin_zone

    zone_name fccpath

    path fcc l gamma x w k gamma

  end

  zone_structure_name fccpath
end

task band structure


Forum >> NWChem's corner >> Running NWChem