Ge - jump in laplacian of electronic density

Elk Users
2013-03-13
2013-06-11
  • Dear Elk users,

    I'm quite new user of Elk and I have a problem with germanium compound.

    I use Dgrid to obtain eli-d and laplacian of electronic density from the results of Elk calculation, but I always have a "jump" on a border of muffin-tin sphere.
    I understand that for the smooth behavior of the 2-nd derivative of electronic density very accurate calculation is needed, but I didn't have such a problem with some other elements which I tried.

    I tried
    1) to increase the number of wave functions (up to rgkmax=11, lmaxapw=11, lmaxvr=12, gmaxvr=22)
    2) to use LAPW+LO instead of APW+lo+LO (if I understand it correctly)
    3) to use autolinengy
    4) to use a bit bigger number of k-points (6x6x10 instead of 4x4x4), I have a hexagonal structure. Unfortunately I have 8 atoms per unit cell, so increasing this number is very time-concuming
    5) to increase nrmt in Ge.in (I used 1500 instead of 500)

    All this haven't decided my problem. Can you suggest me something for better smoothing on muffin-tin bounder?

    Here is my Ge.in file
    'Ge'                                       : spsymb
    'germanium'                                : spname
      -32.0000                                  : spzn
       132414.6195                              : spmass
      0.353553E-06    2.0000   37.4309   1500    : sprmin, rmt, sprmax, nrmt !changed nrmt (500=>2000)
      12                                        : spnst
       1   0   1   2.00000    T                 : spn, spl, spk, spocc, spcore
       2   0   1   2.00000    T
       2   1   1   2.00000    T
       2   1   2   4.00000    T
       3   0   1   2.00000    F
       3   1   1   2.00000    F
       3   1   2   4.00000    F
       3   2   2   4.00000    F
       3   2   3   6.00000    F
       4   0   1   2.00000    F
       4   1   1   1.00000    F
       4   1   2   1.00000    F
       2                                        : apword
      0.1500   0  F                             : apwe0, apwdm, apwve
      0.1500   1  F                             : apwe0, apwdm, apwve
       0                                        : nlx
       7                                        : nlorb
       0   3                                    : lorbl, lorbord
      0.1500   0  F                             : lorbe0, lorbdm, lorbve
      0.1500   1  F
      0.1500   2  F           : all orbitals - LO instead of lo
       1   3                                    : lorbl, lorbord
      0.1500   0  F                             : lorbe0, lorbdm, lorbve
      0.1500   1  F
      0.1500   2  F
       2   3                                    : lorbl, lorbord
      0.1500   0  F                             : lorbe0, lorbdm, lorbve
      0.1500   1  F
      0.1500   2  F
       2   3                                    : lorbl, lorbord
      0.1500   0  F                             : lorbe0, lorbdm, lorbve
      0.1500   1  F
    -1.3259   0  T
       0   3                                    : lorbl, lorbord
      0.1500   0  F                             : lorbe0, lorbdm, lorbve
      0.1500   1  F
    -0.6880   0  T
       0   3                                    : lorbl, lorbord
      0.1500   0  F                             : lorbe0, lorbdm, lorbve
      0.1500   1  F
    -5.9000   0  T                :new s-LO
       1   3                                    : lorbl, lorbord
      0.1500   0  F                             : lorbe0, lorbdm, lorbve
      0.1500   1  F
    -3.9500   0  T               :new p-LO

    Thank you for any suggestion!

    Best regards,
    Olga

     
  • I forgot to add my elk.in. May be it can help.

    tasks
      0            : ground state

    avec
       0.000000000        0.000000000       11.04340000   
      0.0                          10.70719168      -3.681133333   
       9.272700000      -5.353595841      -3.681133333   

    atoms
       1                                    : nspecies
    'Ge.in'                                 : spfname
       8                                    : natoms; atposl, bfcmt below
        0.28250000    0.00000000    0.00000000    0.00000000  0.00000000  0.00000000
        0.71750000    0.00000000    0.00000000    0.00000000  0.00000000  0.00000000
        0.46760000    0.19600000    0.49460000    0.00000000  0.00000000  0.00000000
        0.53240000    0.80400000    0.50540000    0.00000000  0.00000000  0.00000000
        0.27160000    0.29860000    0.80400000    0.00000000  0.00000000  0.00000000
        0.72840000    0.70140000    0.19600000    0.00000000  0.00000000  0.00000000
        0.97300000    0.50540000    0.70140000    0.00000000  0.00000000  0.00000000
        0.02700000    0.49460000    0.29860000    0.00000000  0.00000000  0.00000000

    ngridk            
       4 4 4
      
    xctype
      3 0 0         

    vkloff             
      0.5  0.5  0.5
     
    lradstp
    1                    

    nempty
    20

    deband     
    0.005                

    mixtype      
    1            

    swidth       
    0.01           

    beta0               
    0.05              

    betamax            
    0.5           

    rgkmax         
    11.0             

    lmaxvr          
    11                

    lmaxapw           
    12                

    gmaxvr           
    22.0 

     
  • Dear Olga,

    the jump in laplacian of the density, should come from the fact that you use APW where the derivative of the orbital is not forced to be continuous. In the same way LAPW's are not either forced to have a continuous second derivative.

    Hence to get a smooth laplacian of the density you need to use so-called SLAPW -3 as basis for which the second derivative is continuos. This is a special case of S(uper)LAPW which on one hand lead to slower convergence, see e.g. Singh PRB 43, 6388 (91), but on the other hand, luckily for you, to smoother densities …
    They are already implemented in elk.

    In practice:
    in your species file you change for the (SL)APW's to
      3                                        : apword
      0.1500   0  F                             : apwe0, apwdm, apwve
      0.1500   1  F                             : apwe0, apwdm, apwve
      0.1500   2  F                             : apwe0, apwdm, apwve
    and for LO's to e.g.
    2   4                                    : lorbl, lorbord
      0.1500   0  F                             : lorbe0, lorbdm, lorbve
      0.1500   1  F
    -1.3259   0  T
    -1.3259   1  T
    You can get rid of the first three lo's (do they not lead to linear dependence in your secular equation?) since the energy derivatives are already included in the SLAPW.

    I hope this improve your case. Good luck.
    Lars

     
  • Dear Olga,
    Another thing that might improve your case, which you might have though of already, is to use as large MT radius as possible.
    Hence if the nearest neighbour distance allow for it, you should increase it. If the nn distance is small i your case, that is probably the cause for your specific  problem with this material …
    /Lars

     
  • Dear Lars,

    thank you very much for your advise!

    I will try both your suggestions, hope it will help. Using SLAPW is an interesting possibility, I haven't knew about it.

    Also I tried to increase lmaxapw from 10 to 20 with rgkmax=9 and this improved the situation at list for smaller germanium system.

    Best regards,
    Olga