[Pw_forum] vc-relax with PW

[Konstantin Kudin]

 Hi all,

 This is a public announcement coupled with a request for qualified
help, so if you are NOT interested, please disregard.

 I added some constraints on cell coordinates for vc-relax to the CVS,
so that one can optimize only a subset of those.

 Here is an example of cell_dynamics='damp-w' :

input lattice vectors (alat, hexagonal cell, vacuum in Z)
                 1.000000  0.000000  0.000000
                -0.500000  0.866025  0.000000
                 0.000000  0.000000  1.250000

first step stress:
          total   stress  (Ry/bohr**3)                   (kbar)     P= 
-15.58
  0.00066135   0.00012844   0.00000009         97.29     18.89     
0.01
  0.00012844  -0.00097768  -0.00000005         18.89   -143.82    
-0.01
  0.00000009  -0.00000005  -0.00000129          0.01     -0.01    
-0.19

 The old code would update to:
   1.000003703   0.000000719   0.000000001
  -0.500001229   0.866019900   0.000000000
   0.000000001   0.000000000   1.249999991

 The new code with constraints can do this:
   1.000003703   0.000000000   0.000000000
  -0.500000614   0.866019900   0.000000000
   0.000000000   0.000000000   1.250000000

 So only x1, x2, and y2 are updated.

 What is interesting is that it looks like the fully unconstrained code
is somewhat rotating the lattice vectors (note the +delta_y1 and
-delta_x2).

 If somebody understands what the VC code actually does, please take a
look.

 Kostya



 
____________________________________________________________________________________
Do you Yahoo!?
Everyone is raving about the all-new Yahoo! Mail beta.
http://new.mail.yahoo.com