[Pw_forum] Re: How to determine the thickness in surface phonons calculation?

[英利 牛]

Prof Nicola Marzari,
  Thanks very much for your help!
  
  I am Yingli Niu. My mailbox didn't deliver the mail
to the forum. So I used Lijun Zhang's mailbox. Now I
change a mailbox.
  
  I have test the vacuum thickness and other
parameters, for 14 layers C(2x1) and 2 layers H:
the vacuum thickness is 10 Ang.
    nat       = 32,
    ntyp      = 2,
    ecutwfc   = 50,
    ecutrho   = 600,
    nbnd      = 90
 
    diagonalization='david'
    conv_thr    = 1.0d-8

    C  12.0107 C_GGA.af
    H  1.00794 H.pbe-van_bm.UPF

    K_POINTS {automatic}
      4 2 1 0 0 0
    EOF

    tr2_ph=1.0d-14,

And I get the frequences:
     omega( 1) =      -1.271939 [THz] =     -42.427596
[cm-1]
     omega( 2) =       0.605728 [THz] =      20.205038
[cm-1]
     omega( 3) =       0.675694 [THz] =      22.538880
[cm-1]
     ......
     omega(93) =      89.296054 [THz] =    2978.615556
[cm-1]
     omega(94) =      89.322861 [THz] =    2979.509737
[cm-1]
     omega(95) =      89.993780 [THz] =    3001.889321
[cm-1]
     omega(96) =      90.062173 [THz] =    3004.170679
[cm-1]

   It's clearly that two sides of the slab interact
each other. So the highest optical frequences become
so high when then vacuum thickness is equal to 8 Ang.
Now I change it to 10 Ang, then the frequences reduce.
Surely according to Prof Nicola Marzari's advice, I
should do painful, slow, and *necessary* convergence
test.
    
   I have read the article of Prof  Claudia Bungaro,
Surface lattice dynamics of Mg(0001), PRB
62,17012,(2000).
A periodically repeated slab geometry with
Mg(0001)slabs of nine atomic layers separated by a
vacuum region of 13 Å (equivalent to five atomic
layers) was utilized to study the surface. Then he
calculated the phonons of the slab. 
******
    What's the nature of the phonons of the slab?
Couldn't we separate the surface modes from the bulk
modes in the thin slab model?
******    
    Prof Nicola Marzari built a thicker slab in order
to separate the surface modes from the bulk modes. He
said:

    The dynamical matrix was built up of a much
thicker slab by inserting a number of bulk layers in
the middle of the ninelayer slab. In particular, the
force constants for the interactions of atoms in the
first five surface layers are those calculated for the
nine-layer slab; all the others are described by the
bulk force constants. In this way we can calculate the
phonon dispersion and the corresponding atomic
vibrations for a slab of Mg(0001) as thick as
desired.To describe the lattice dynamics of the
surface, we have used a slab of 63 Mg layers that is
thick enough to decouple completely the surface
vibrations localized at the two opposite surfaces of
the slab.

    So,he built a "64 bulk layers" + "9 surface
layers", then insert the constant into the case.fc. 
******
    Can anyone tell me the details about insertion
skills? Should the 64 bulk layers be in the middle of
the 9 surface layers(where is the middle of 9 surface
layers)? I think I should build up a bulk primitive
unit cell first,then calculate the bulk force constant
with large grids as I want, e.g. 32*32*1(or 32*32*32 ?
),and I can get the bulk force constants. The surface
force constants will be the ones which was calculated
in the 9-layer slab. Then how can I get the force
constant between the bulk and the surface?  If I get
all the force constants of the slab, I can calculate
the phonon frequences at any qpoint in the q-space.
******


		
___________________________________________________________ 
抢注雅虎免费邮箱-3.5G容量，20M附件！ 
http://cn.mail.yahoo.com