[Pw_forum] about phonon frequencies

[sun jason]

Dear all,

I found my calculated phonon frequencies are not accurate enough. for 
instance, when I vc-relax the cubic BN with LDA (B.pz-vbc.UPF and 
N.pz-vbc.UPF) the result is
-------------
     Final estimate of lattice vectors (input alat units)
  -0.491348786   0.000000000   0.491348786
   0.000000000   0.491348786   0.491348786
  -0.491348786   0.491348786   0.000000000
  final unit-cell volume =     75.6348 (a.u.)^3
  input alat =       6.8314 (a.u.)
--------------
so I use 3.615*(0.491348786/0.5)=3.552 to calcalte phonon frequency. the 
result is
0.0000   0.0000   0.0000   1089.4410(TO)   1089.4410(TO)   1318.1950(LO)
---------they are too high

if I vc-relax the lattice use GGA_PBE (B.pbe-n-van_ak.UPF and 
N.pbe-n-van_ak.UPF)
the result is 3.618, a little large than experimental value. and the phonon 
frequencies are 
0.0001   0.0001   0.0000   1029.5267   1029.5267   1270.6955
---------they are too low

so how can I get a more accurate phonon frequencies?
any comment would be appreciate.

best

Jian SUN