The ABINIT Users Mailing List ( CLOSED )

["Eric J. Walter" <ewalter@particle.physics.wm.edu>]

Dear Abinit forum,

It seems to me that it is not possible to enforce the acoustic sum
rule in anaddb when doing a metal and skipping the DDK step (since it
shouldn't be necessary, right?). After some investigation, I have
found that in anaddb.F90 the electric field perturbation variable is
always set to 2 before the read of the gamma q-point DDB info:

------------------------------------------------------------
! **********************************************************************

!Dielectric Tensor and Effective Charges

!Look after the Gamma Blok in the DDB
 qphon(:,1)=0.0d0
 qphnrm(1)=0.0d0
 rfphon(1:2)=1
 rfelfd(1:2)=2

 rfstrs(1:2)=0
 rftyp=anaddb_dtset%rfmeth

 call gtblk9(blkflg,blknrm,blkqpt,blktyp,iblok,mpert,msize,natom,nblok,&
& qphon,qphnrm,qtol,rfphon,rfelfd,rfstrs,rftyp)

------------------------------------------------------------

As far as I can see, there is no way to tell anaddb not to expect a
d/dk part of the DDB when dealing with gamma.  This ends up causing
many of the following COMMENTS in STDOUT:

------------------------------------------------------------
gtblk9 : COMMENT -
  The blok     1 does not match the requirement
  because it lacks the element with
  idir1,ipert1,idir2,ipert2,index=    1    1    1    3  127
 gtblk9 : COMMENT -
  The blok     1 does not match the requirement
  because it lacks the element with
  idir1,ipert1,idir2,ipert2,index=    1    1    2    3  148
 gtblk9 : COMMENT -
  The blok     1 does not match the requirement
  because it lacks the element with
  idir1,ipert1,idir2,ipert2,index=    1    1    3    3  169
------------------------------------------------------------

Here, ipert2 means the electric field perturbation since this is a 1
atom unit cell (natom+2 = electric field pert).

Finally, this leads to the ASR never being applied and anaddb claiming
that it exited successfully (!).

To temporarily remedy this situation, I have added some code to
anaddb.F90 (here is a diff):

------------------------------------------------------------

--- anaddb.F90_orig     Sat Mar 19 13:01:36 2005
+++ anaddb.F90_ejw      Tue Jun 14 10:11:11 2005
@@ -218,9 +218,22 @@
  qphnrm(1)=0.0d0
  rfphon(1:2)=1
  rfelfd(1:2)=2
+
  rfstrs(1:2)=0
  rftyp=anaddb_dtset%rfmeth
 
+ if (occopt>=3.and.occopt<=7) then
+  rfelfd(1:2)=0                      
+  write(message, '(a)') "COMMENT:  rfelfd set to 0 since due to metal 
occupations"
+  call wrtout(6,message,'COLL')
+  call wrtout(iout,message,'COLL')
+
+  write(message, '(a)') "COMMENT:  this change made by EJW, NOT PART OF THE 
STANDARD ABINIT RELEASE!!!"
+  call wrtout(6,message,'COLL')
+  call wrtout(iout,message,'COLL')
+
+ endif
+

------------------------------------------------------------

This results in rfelfd set to 0 if metallic occupations are present.
This allows the d/dk part of the DDB to be absent and yet still apply
the ASR to metallic calculations.

I have tested this approach by comparing the results of phonon
frequencies calculated at q=0,0,0 and q=0.05, 0.05, 0 with the ASR
applied in two different ways:

1) With an unmodified 4.4.4 release, I actually included the d/dk step
as if the calculation were an insulator:

  Phonon wavevector (reduced coordinates) :  0.00000  0.00000  0.00000
 Phonon energies in Hartree :
   0.000000E+00  0.000000E+00  0.000000E+00
 Phonon energies in meV     :
-  0.000000E+00  0.000000E+00  0.000000E+00
 Phonon frequencies in cm-1    :
-  0.000000E+00  0.000000E+00  0.000000E+00
 Phonon frequencies in Thz     :
-  0.000000E+00  0.000000E+00  0.000000E+00
 Phonon energies in Kelvin  :
-  0.000000E+00  0.000000E+00  0.000000E+00
.
.
.
  Phonon wavevector (reduced coordinates) :  0.05000  0.05000  0.00000
 Phonon energies in Hartree :
   5.816119E-04  7.040367E-04  3.426734E-03
 Phonon energies in meV     :
-  1.582647E+01  1.915781E+01  9.324618E+01
 Phonon frequencies in cm-1    :
-  1.276491E+02  1.545182E+02  7.520812E+02
 Phonon frequencies in Thz     :
-  3.826822E+00  4.632339E+00  2.254683E+01
 Phonon energies in Kelvin  :
-  1.836583E+02  2.223169E+02  1.082076E+03


2) With the d/dk step skipped using my modified anaddb code:

  Phonon wavevector (reduced coordinates) :  0.00000  0.00000  0.00000
 Phonon energies in Hartree :
   0.000000E+00  0.000000E+00  0.000000E+00
 Phonon energies in meV     :
-  0.000000E+00  0.000000E+00  0.000000E+00
 Phonon frequencies in cm-1    :
-  0.000000E+00  0.000000E+00  0.000000E+00
 Phonon frequencies in Thz     :
-  0.000000E+00  0.000000E+00  0.000000E+00
 Phonon energies in Kelvin  :
-  0.000000E+00  0.000000E+00  0.000000E+00

  Phonon wavevector (reduced coordinates) :  0.05000  0.05000  0.00000
 Phonon energies in Hartree :
   5.816119E-04  7.040367E-04  3.426734E-03
 Phonon energies in meV     :
-  1.582647E+01  1.915781E+01  9.324618E+01
 Phonon frequencies in cm-1    :
-  1.276491E+02  1.545182E+02  7.520812E+02
 Phonon frequencies in Thz     :
-  3.826822E+00  4.632339E+00  2.254683E+01
 Phonon energies in Kelvin  :
-  1.836583E+02  2.223169E+02  1.082076E+03



I have posted the input and output files on http://www.cpd.wm.edu/ejw_abinit

Is this a bug or am I missing something? Shouldn't it be ok to skip
the d/dk step in a metal?

I would be happy to supply more information if necessary.

Thanks,

Eric J. Walter