The ABINIT Users Mailing List ( CLOSED )

["徐" <lanqingxu@fjnu.edu.cn>]

Dear Matteo,
Thanks very much for your help. I've revised the code and re-compile them. 
Now it is OK.
Best regards,
shelleny

>From: "Matteo Giantomassi" <Matteo.Giantomassi@uclouvain.be>
>Reply-To: forum@abinit.org
>To: forum@abinit.org
>Subject: Re: {Spam?} Re: [abinit-forum] negative  Phonon frequency at 
>gammaand wrongPHDOS
>Date:Sun, 26 Jul 2009 21:19:32 +0200
>
>> Dear Matteo,
>> I simply plot the omega ,PHDOS ,PJDOS[1],PJDOS[2] columes and compare
>> results with A.H.Romero et al.(PRB2008,78,224302). Their x axis is phonon
>> frequency from 0 to 250cm-1, my x axis is omega. The overall agreement is
>> bad.Should I convert omega to phonon frequency? I'm not sure about the
>> formula. The output PHDOS file pls see the attached document.
>
> Omega is the phonon frequency. By default it is given in Hartree.
> The header of the PHDOS file indeed reports:
>
> # Phonon density of states and projected DOS generated by anaddb
> #
> # energy in Ha , DOS in states/Ha
> # Gaussian method with smearing =   2.27816763E-05Ha , nqibz = 256
>
> If you want to compare with the results reported in the paper, you
> have to convert from Hartree to cm-1 both the x- and the y-axis.
>
> Ha_cmm1=219474.6313705  ! 1 Hartree, in cm^-1
>
> I had a look at your file and the weird results you are getting are due
> to a problem with the output format used to write the results,
> in particular the first and the second column (omega and PHDOS,
>respectively).
> The two columns should be separated by a blank. In your case,
> some values of PHDOS (2nd column) reach the maximum width specified in
>the code (f11.5).
> As a consequence the first and the second column are joined together and
>the plotter
> doesn't read correctly the values.
>
> I'm attaching the modified PHDOS.txt in which the two columns have been
> separated by hand using vi. The results now seem reasonable.
>
> The output format (f11.5) is not safe and I've fixed it in 5.8.4
> In the case you want to patch your version of m_phdos, this is the diff
>file:
>
>=== modified file 'src/77_ddb/m_phdos.F90'
>--- src/77_ddb/m_phdos.F90     2009-05-04 20:03:40 +0000
>+++ src/77_ddb/m_phdos.F90     2009-07-26 18:05:29 +0000
>@@ -161,9 +161,9 @@
> ! call wrtout(unt,msg,'COLL')
>   write(msg,'(5a)')'# ',ch10,'# omega     PHDOS    IPHDOS   PJDOS[1] 
>IPJDOS[1] ...  ',ch10,'# '
>   call wrtout(unt,msg,'COLL')
>-  write(frmt,*)'(',phonon_dos%ntypat,'(2f11.5))'
>+  write(frmt,*)'(',phonon_dos%ntypat,'(2es17.8))'
>   do io=1,phonon_dos%nomega
>-  
>write(unt,'(3f11.5)',advance='NO')phonon_dos%omega(io)*cfact,phonon_dos%phdos(io)/cfact,phonon_dos%phdos_int(io)/cfact
>+  
>write(unt,'(3es17.8)',advance='NO')phonon_dos%omega(io)*cfact,phonon_dos%phdos(io)/cfact,phonon_dos%phdos_int(io)/cfact
>    do itype=1,phonon_dos%ntypat
>     
> write(unt,frmt,advance='NO')phonon_dos%pjdos_typ(io,itype)/cfact,phonon_dos%pjdos_typ_int(io,itype)/cfact
>    end do
>
>
> The non-zero contribution to the PHDOS at negative frequencies is not a
>problem.
> The values are small and due to the fact that you are using a Gaussian
> to approximate the delta function.
> You might try the tetrahedron method (prtdos 2) or a gaussian with a smaller
> broadening (and a denser interpolating q-mesh) in order to have a nicer
>PHDOS
> curve with a smaller contribution for omega

> Best Regards
> Matteo Giantomassi
>
>
>