[Pw_forum] Phonon frequencies different for equivalent q points

stewart at cnf.cornell.edu stewart at cnf.cornell.edu
Tue Apr 11 20:45:42 CEST 2006


Hi everyone, 

   I have been running some phonon calculations for an hcp Pd system and I 
am running into a problem. 

I am looking at the high symmetry lines, Gamma-K-M along the [x,x,0] 
direction going from [0,0,0] to [1,1,0]*(2pi/a) and the Gamma-M line along 
the [x,0,0] going from [0,0,0] to [1/sqrt(3),0,0]*(2pi/a). 

The phonon frequencies predicted at the M point for each calculation 

[1,1,0]*(2pi/a) and [1/sqrt(3),0,0]*(2pi/a) should be equivalent. 

However, they are fairly different.
[1,1,0] 

    omega( 1) =       2.805385 [THz] =      93.578178 [cm-1]
    omega( 2) =       3.004742 [THz] =     100.228087 [cm-1]
    omega( 3) =       4.199464 [THz] =     140.079956 [cm-1]
    omega( 4) =       5.667289 [THz] =     189.041683 [cm-1]
    omega( 5) =       5.820376 [THz] =     194.148127 [cm-1]
    omega( 6) =       6.358473 [THz] =     212.097247 [cm-1] 


[1/sqrt(3),0,0] 

    omega( 1) =       3.747138 [THz] =     124.991899 [cm-1]
    omega( 2) =       4.763040 [THz] =     158.878972 [cm-1]
    omega( 3) =       4.797494 [THz] =     160.028243 [cm-1]
    omega( 4) =       5.155421 [THz] =     171.967476 [cm-1]
    omega( 5) =       5.867315 [THz] =     195.713877 [cm-1]
    omega( 6) =       6.220546 [THz] =     207.496457 [cm-1] 

Now hcp Pd only exists as small films on special substrates.  Normally, Pd 
is fcc.  However, the rest of the phonon dispersion curve looks fine (no 
imaginary data points).  I am using ecutwfc=44.0 and ecutrho=448 for my scf 
calculations with a 20x20x16 k-point grid.  The total stress on my relaxed 
system is 0.09 kbar so I don't think that is the issue. 

For the phonon calculations, my input is as follows 

phonons of pd at 1.00
&inputph
   tr2_ph=1.0d-14,
   alpha_mix(1)=0.7,
   prefix='pd',
   amass(1)=106.42,
   fildyn='pd.dyn1.00',
/
1.00 1.00 0.00 

I am using espresso-3.0.  Are there any changes in the CVS version that 
would affect this calculation? 

Any suggestions or thoughts would be greatly appreciated! 

Thanks, 

Derek 


################################
Derek Stewart, Ph. D.
Scientific Computation Associate
250 Duffield Hall
Cornell Nanoscale Facility (CNF)
Ithaca, NY 14853
stewart (at) cnf.cornell.edu
(607) 255-2856 




More information about the users mailing list