ABINIT Discussion Forums

Posted: **Sun Oct 02, 2011 5:19 pm**

Dear Researchers,

I'm trying to calculate the optical properties of Barium Titanate in it's P4mm phase. Following the optics tutorial, I calculated the linear response with 4 * 4 * 4 and 8 * 8 * 8 K-grid. Obviously, they were not very accurate. For accuracy, I decided to use 24 * 24 * 24 kpoint-grid. To make sure my computer (Intel core i5, 4 GB RAM, Win 7 32 bit) can handel the memory need, I used mffmem 0 mkmem 500 mkqmem 500 mk1mem 500. After 21 hours, I got the WFK files necessary to run the next optics run.

But the problem is that the wfk files are 6~7 GB in size and optic is crashing with this error:
"Operating system error: Can not allocate memory Out of memory."

At this point I'm lost. Is it at all possible to handle a wfk file that big on a single PC (keeping my computer's specification in mind)?

Here is the input file to abinit:

Code: Select all

#Prepare the computation of linear and non-linear optic properties
#of GaAs crystal : ground-state with few bands,
#then non-SCF with a larger number of bands, then ddk for different directions
#Note that the k point sampling shoud be finer for significant results. The cut-off energy is also too low.

 ndtset 6

#First dataset : SC run with kpoints in the IBZ
   iscf1  3
  nband1  40     
  nstep1 1000   
 kptopt1 1
 nbdbuf1 0
 prtden1 1   getden1 0   getwfk1 0    ! Usual file handling data

#Second dataset : NSC run with large number of bands, and points in the IBZ
    iscf2 -2  
   nband2 40  ! This number of bands might be too low for non-linear optics and real part of linear optics
   nstep2 1000
  kptopt2  1
  getwfk2  1   getden2 1   ! Usual file handling data

#Third dataset : NSC run with large number of bands, and points in the the full BZ
    iscf3 -2  
   nband3 40  ! This number of bands might be too low for non-linear optics and real part of linear optics 
   nstep3 1000
  kptopt3  3
  getwfk3  2   getden3 1   ! Usual file handling data


#Fourth dataset : ddk response function along axis 1
   iscf4 -3
  nband4 40   ! This number of bands might be too low for non-linear optics and real part of linear optics
  nstep4  1  nline4  0
 kptopt4  3

   nqpt4  1  qpt4  0.0d0 0.0d0 0.0d0 
  rfdir4  1 0 0
 rfelfd4  2
 getwfk4  3

#Fifth dataset : ddk response function along axis 2
   iscf5 -3
  nband5 40   ! This number of bands might be too low for non-linear optics and real part of linear optics
  nstep5  1  nline5  0
 kptopt5  3

   nqpt5  1  qpt5  0.0d0 0.0d0 0.0d0
  rfdir5  0 1 0
 rfelfd5  2
 getwfk5  3

#Sixth dataset : ddk response function along axis 3
   iscf6 -3
  nband6 40   ! This number of bands might be too low for non-linear optics and real part of linear optics
  nstep6  1  nline6  0
 kptopt6  3

   nqpt6  1  qpt6  0.0d0 0.0d0 0.0d0
  rfdir6  0 0 1
 rfelfd6  2
 getwfk6  3


#Data common to all datasets
 nshiftk 1
 shiftk  0.5 0.5 0.5
         
 ngkpt  24 24 24       ! This is much too low : should be at least 24x24x24
mffmem 0
mkmem 500
mkqmem 500
mk1mem 500
 acell   7.5124752684E+00  7.5124752684E+00  7.6906441898E+00
 amu 1.37327000E+02  4.78800000E+01  1.59994000E+01
 diemac 4.00000000E+03
 ecut 10.00             ! This is also too low
 iscf 3
 ixc 11
 natom  5  
 ntypat  3
 rprim   1 0 0  0 1 0  0 0 1   
 xred 0.0000000000E+00  0.0000000000E+00  0.0000000000E+00
                     5.0000000000E-01  5.0000000000E-01  5.1430000000E-01
                     5.0000000000E-01  5.0000000000E-01 -3.0760000000E-02
                     5.0000000000E-01  0.0000000000E+00  4.8481703472E-01
                     0.0000000000E+00  5.0000000000E-01  4.8481703472E-01
                                               
nbdbuf 10
# tnons 72*0.0
 typat  1 2 3 3 3 tolwfr  1.e-20
 znucl 56 22 08

Here is the input file to optic

Code: Select all

toptic_1_o_DS4_1WF16
toptic_1_o_DS5_1WF17
toptic_1_o_DS6_1WF18
toptic_1_o_DS3_WFK
0.002         ! Value of the smearing factor, in Hartree
0.0003  0.3   ! Difference between frequency values (in Hartree), and maximum frequency ( 1 Ha is about 27.211 eV)
0.000         ! Scissor shift if needed, in Hartree
0.002         ! Tolerance on closeness of singularities (in Hartree)
3             ! Number of components of linear optic tensor to be computed
11 22 33      ! Linear coefficients to be computed (x=1, y=2, z=3)
2             ! Number of components of nonlinear optic tensor to be computed
123 222       ! Non-linear coefficients to be computed

Any help or comment would be greatly appreciated.

Posted: **Mon Apr 02, 2012 10:18 pm**

Hello,

1) you need a larger nstep for the non scf parts (= number of iterations for convergence).

2) In principle you should be able to do the calculation - what you use in optic is just the matrix elements of the p operator, which are less than 10% of the total 1WF and WFK files... You don't need to load all 7 Gb, but it depends how much memory you have. The big arrays are nkpt * nband**2, but you need several of them (3 directions + spin polarization eventually).

Matthieu

Posted: **Wed May 02, 2012 1:37 pm**

Thanks Sir. If I knew this beforehand, I would not have upgraded to 64 bit and installed the extra 4 GB RAM, but thanks nevertheless.

ABINIT Discussion Forums

Can optic handle VERY big wfk files?

Can optic handle VERY big wfk files?

Re: Can optic handle VERY big wfk files?

Re: Can optic handle VERY big wfk files?