Convergence and Bandstructure with GW correction

[rupesh]

Dear Dr. Daniele,

I am sorry to disurb you again but I am confused. I am getting this error

Code: Select all

 <---> [01] Files & I/O Directories
 <---> [02] CORE Variables Setup
 <---> [02.01] Unit cells
 <---> [02.02] Symmetries
 <---> [02.03] RL shells
 <---> [02.04] K-grid lattice
 <---> [02.05] Energies [ev] & Occupations
 <---> [03] Transferred momenta grid
 <---> [04] Bare local and non-local Exchange-Correlation[ubuntu:03283] *** Process received signal ***
[ubuntu:03283] Signal: Segmentation fault (11)
[ubuntu:03283] Signal code: Address not mapped (1)
[ubuntu:03283] Failing at address: (nil)
[ubuntu:03283] [ 0] /lib/x86_64-linux-gnu/libpthread.so.0(+0xfcb0) [0x7fc4036efcb0]
[ubuntu:03283] [ 1] yambo(xco_hartree_fock_+0x480) [0x4673b0]
[ubuntu:03283] [ 2] yambo(xco_driver_+0x351) [0x466bf1]
[ubuntu:03283] [ 3] yambo(yambo_driver_+0x661) [0x4253d1]
[ubuntu:03283] [ 4] yambo(main+0x5fe) [0x4232de]
[ubuntu:03283] [ 5] /lib/x86_64-linux-gnu/libc.so.6(__libc_start_main+0xed) [0x7fc40334276d]
[ubuntu:03283] [ 6] yambo() [0x423a49]
[ubuntu:03283] *** End of error message ***
Segmentation fault (core dumped)

I have checked setup file yes there is X-grid no uniform but I have made the changes you suggested to similar problems in blog (i.e. force_symmetric = .true.), I am doing calculation for Black Phosphorus k-grid is 8X6X8, and I have attached both scf and nscf files ,setup file is also there .why it is not getting X-grid uniform ? or there is any other error in input files?

Thank you !

Regards,
Rupesh

[Daniele Varsano]

Dear Rupesh,
I do not have a clear idea on what is going on.
Please check your qe input files. Note that

Code: Select all

CELL_PARAMETERS  {cubic}

"cubic" is not a correct attribute for cell_parameters.
Try to check your input crystal structure, at first glance I'm not sure it is correct. Pay attention on the units, atom per cell, positions.
Moreover, I also suggest you, before going through GW and so on, to compare your DFT results, in this case LDA, with the literature.
GW it is a perturbation theory, so if the ground state it is not accurate, it does no make any sense.
Finally, the keyword is non_symmorphic (instead of non_symmetric) which has a totally different meaning, anyway in the input file it is correctly spelled.

Best,
Daniele

[rupesh]

Dear Daniele,

I would like to ask you few questions about r-setup file :

Code: Select all

  [02.04] K-grid lattice
  ======================

  Compatible Grid is 3D
  B1 [rlu]=-0.125000  0.000000  0.000000
  B2      = 0.000000 -0.125000  0.166667
  B3      = 0.000000 -0.166667  0.125000
  [b]Grid dimensions               :  8   6   6[/b]
  K lattice UC volume       [au]:0.6941E-3

  [02.05] Energies [ev] & Occupations
  ===================================

  Fermi Level        [ev]: 6.753270
  Electronic Temp. [ev K]:  0.00      0.00    
  Bosonic    Temp. [ev K]:  0.00      0.00    
  El. density      [cm-3]: 0.249E+24
  States summary         : Full        Metallic    Empty
                           0001-0010               0011-0050
  Indirect Gaps      [ev]: 0.127296  2.694010
  Direct Gaps        [ev]: 0.255504  5.080506
  [b]X BZ K-points :  736[/b]

basic questions :

• Why is that grid dimentions coming 8 6 6 though I have given k-grid 8 6 8 ? and again k-points 736 why ?
  Shouldn't that be 8X6X8 = 384 ? but in my pwscf also I am not getting 384 ,its 372 at there? Is there upper limit for it ?
  Does Ex functional used in pwscf calculation is causing me trouble like error in the last post of mine ?

All input files are same as previous post.


Thank you !
Regards,
Rupesh

[Daniele Varsano]

Dear Rupesh,
nk1,nk2,nk3 are not the number of k-point grid, but the Monkhrost Pack index, I invite you to study what does it means.
736 --> Number of K-points in the entire Brillouin zone.
372--> Number of k-points in the Irriducible Brillouin zone.
Yambo make use of symmetries, so the k-point in the IBZ are expanded accordingly with the symmetries operations reported (in your case the inversion).
The functional should be not a problem, it is an LDA right. Anyway the pseudo you are using is an hgh, that is rather hard and you need a big cutoff to get convergences.
The error anyway comes from the fact you do have a non-uniform grid. My guess is there something wrong in your structure,
please check your cell/atomic_position/units lattice parameters etc. in your qe input.
Have you checked your ground state structure make sense?
According to this paper (tight binding):
H Asahina and A Morita 1984 J. Phys. C: Solid State Phys. 17 1839. doi:10.1088/0022-3719/17/11/006
and this more recent one (DFT hybrid functional):
http://www.nature.com/nnano/journal/v9/ ... 35_F1.html
it looks that black phospshorus has a direct band gap at Z point (while you find an indirect gap). In the first reference you can also find the crystal structure.
Try to be sure that all make sense at DFT stage before passing to GW.

Best,
Daniele

[rupesh]

Hi,

I think my DFT calculations are correct because I have used the same atomic positions and parameters mentioned in H Asahina earlier 1981 paper. And I got the band structure for Black Phosporus like this one there is small error may be due to band direction that I had given not correct.

[rupesh]

Dear Daniele,

I have verified the structure using XCrySDen , even if you get short time you can see from the scf file in second last post with XCrySDen . You are always attentive to my doubts and I am very very thank you for it . And really sure about atomic positions. Please reply if you get something .

Thank you !

Regards,
Rupesh

[Daniele Varsano]

Dear Rupesh,
unfortunately I can't install xcrysden in my laptop. If you are sure about your crystal structure try to play with the k-point grids, inspect it and try to figure out why your grid is not uniform. You can for instance try to add a shift in your kpoint grid.

Code: Select all

K_POINTS {automatic}
8 6 8 1 1 1
or 
8 6 8 1 0 0 

It looks me that in your kpoint grid (you can get it from the qe output or by typing:

Code: Select all

grep iku r_setup

you have in your sampling several equivalent k-points, this should not happen and it depends on your definition of the unit cell.

Best,
Daniele

[rupesh]

Dear Daniele,

I tried different cominations of 8X6X8 with shifts 1 1 1, 0 1 0, 1 0 0, 1 1 0, etc. but any case the error wasn't solve. Then I tried with 6X6X6 and 8X8X8 grid there was no error uniformity. So I have a simple question suppose I used these grid and made bandstructure calculation for Black Phosphorus and Phosphorene (which supposed to be 14X10X4 and 14X10X1 respectively) willI get GW correction to my DFT structure along same directions as used in DFT ? There will definitely be convergence issues with these new grid . Do you have some solution?

Thank you !!

Regards,
Rupesh

[Daniele Varsano]

Dear Rupesh,
1. Sure, GW calculation are done on the same point you calculated the DFT structure (as you need the KS structure to perform perturbation on top). Next you can interpolate the Band Structure for the desired path.
2. Convergence issues is surely something to be checked.

Two comments:
1) Phosphorene is a 2D structure, surely you do not need to sample the BZ in the direction orthogonal to it.
2) I'm not expert on your system anyway looking very quickly to the literature I can see that black phosphorus crystallize as base centered orthorombic lattice, and looking at this paper (e.g. here http://arxiv.org/pdf/1211.3512.pdf or http://isites.harvard.edu/fs/docs/icb.t ... apter2.pdf) you can have it considering an orthorombic lattice with 8 P atoms per cell. You can try to consider to build such input. You will have more atoms, but you should need a smaller BZ sampling and would be much more easy to look at the Brillouin zone shape and avoid to get non uniform grids.
You can also try to consider to build up your input with ibrav=9 but it should not affect, or try to ask to quantum espresso forum.

Best,
Daniele

[rupesh]

Dear Daniele,

I am now performing GW calculation for GaAs and l kept Phosphorene aside for time .I perform the calculations but the band gap I got in DFT calculation ~ 0.4 eV not differ much from GoWo calculation ~0.69 eV . I was expecting its above 1.0 eV by GoWo. How sshould I get it much close to experimental value or atleast different significantly from DFT calculations? I have attached zip folder containing scf and nscf inputs, yambo setup, ypp.in files. Please help me .


Thank you !!!

Regards,
Rupesh