<HTML><HEAD><TITLE>Mail</TITLE>
<META name=GENERATOR
content="KsDHTMLEDLib.ocx, FreeWare HTML Editor 1.164.2, ?Kurt Senfer">
<META content="text/html; charset=GB2312" http-equiv=Content-Type></HEAD>
<BODY style="FONT-FAMILY: ÐÂËÎÌå; FONT-SIZE: 9pt" leftMargin=5 topMargin=5 #ffffff>
<DIV>Dear all, </DIV>
<DIV> </DIV>
<DIV>I was planning to use d3 to do Gamma point linewidth calculation. I
start with silicon, which is an example provided in the phonon package. I
successively run pw, ph and d3 and then I can get the d3 dynamical matrix. The
input files and the final output from d3 are shown below: (the version I use is
5.0.3)</DIV>
<DIV> </DIV>
<DIV><STRONG>pw.x:</STRONG></DIV>
<DIV> </DIV>
<DIV>&control<BR> calculation= 'scf'<BR>
wf_collect=.true.<BR> restart_mode= 'from_scratch'<BR>
pseudo_dir= '/home/jiawei/pseudo/'<BR> outdir=
'/scratch/jiawei/silicon/'<BR> prefix=
'Si-ph-k7-e20-a10.21-pp:pz-vbc'<BR> tprnfor= .true.<BR>
tstress= .true.<BR>/<BR>&system<BR> ibrav= 2, celldm(1)= 10.21,
nat= 2, ntyp= 1<BR>
ecutwfc=20<BR>/<BR>&electrons<BR> conv_thr=
1.0d-12<BR> diagonalization= 'david'<BR> mixing_beta=
0.7<BR>/<BR>ATOMIC_SPECIES<BR> Si 28.0855
Si.pz-vbc.UPF<BR>ATOMIC_POSITIONS<BR> Si 0.00 0.00 0.00<BR> Si 0.25
0.25 0.25<BR>K_POINTS {automatic}<BR> 7 7 7 0 0 0<BR></DIV>
<DIV><STRONG>ph.x:</STRONG></DIV>
<DIV> </DIV>
<DIV>phonons of Si at Gamma<BR>&inputph<BR>
recover=.false.<BR> tr2_ph=1.0d-14<BR> prefix=
'Si-ph-k7-e20-a10.21-pp:pz-vbc'<BR> epsil=.false.<BR>
ldisp=.false.<BR> amass(1)=28.0855<BR>
outdir='/scratch/jiawei/silicon/'<BR>
fildyn='Si.dyn'<BR> fildrho='Si.drhoG'<BR>/<BR>0.0 0.0 0.0<BR></DIV>
<DIV><STRONG>d3.x:</STRONG></DIV>
<DIV> </DIV>
<DIV>anharm at BZ<BR>&inputph<BR> prefix=
'Si-ph-k7-e20-a10.21-pp:pz-vbc'<BR> amass(1)=28.0855<BR>
outdir='/scratch/jiawei/silicon/'<BR>
fildyn='Si.anhG'<BR> fildrho='Si.drhoG'<BR>
fild0rho='Si.drhoG'<BR> wraux=.true.<BR>/<BR> 0.000 0.000
0.000<BR></DIV>
<DIV><STRONG>Si.anhG:</STRONG></DIV>
<DIV> </DIV>
<DIV>Derivative of the force constants</DIV>
<DIV> 1 2 2 10.2100000 0.0000000
0.0000000 0.0000000 0.0000000
0.0000000<BR>
1 'Si ' 28.085500000000000<BR>
1 1
0.0000000 0.0000000
0.0000000<BR> 2
1 0.2500000
0.2500000 0.2500000</DIV>
<DIV> Third derivative in cartesian axes</DIV>
<DIV> q = ( 0.000000000
0.000000000 0.000000000 )</DIV>
<DIV><BR>
modo: 1</DIV>
<DIV> 1 1<BR>
-0.416333634234E-16
0.000000000000E+00
-0.192592994439E-33<BR>
0.000000000000E+00
0.192592994439E-33
0.000000000000E+00<BR>
-0.416333634234E-16
0.000000000000E+00
0.192592994439E-33<BR>
0.000000000000E+00
0.385847337244E+00
0.000000000000E+00<BR>
0.277555756156E-16
0.000000000000E+00
0.385847337244E+00<BR>
0.000000000000E+00
0.192592994439E-33
0.000000000000E+00<BR>......</DIV>
<DIV>(here I just show the first several numbers)</DIV>
<DIV> </DIV>
<DIV>Up to now there is no problem. But I also wanted to know the
wavefunction on the whole Brillouin zone. Because pw.x will automatically
find the IBZ and only calculate wavefunction on that reduced k-mesh, I want to
turn off the symmetry and see what I can get. There are two switches I
found that I can use: nosym_evc and noinv. So I set them both to be true:
(inside the &system control flag)</DIV>
<DIV> </DIV>
<DIV><STRONG>nosym_evc=.true.<BR>noinv=.true.</STRONG></DIV>
<DIV><STRONG></STRONG> </DIV>
<DIV>So the pw.x will not use any symmetry and do the calculation in the FBZ.
The result from pw.x clearly shows no symmetry found and 343(7*7*7) k-points are
used. The results from phonon code are also the same.(dynamical matrix at Gamma
point as well as the frequency). However, when it comes to d3, the results are
different. It runs without any problem, but the final d3.anhG file I get is
below:</DIV>
<DIV> </DIV>
<DIV><STRONG>d3.anhG</STRONG> (symmetry turnned off)</DIV>
<DIV> </DIV>
<DIV>
modo: 1</DIV>
<DIV> 1 1<BR>
0.000000000000E+00
0.000000000000E+00
0.832667268469E-16<BR>
0.000000000000E+00
0.832667268469E-16
0.000000000000E+00<BR>
0.277555756156E-16
0.000000000000E+00
-0.277555756156E-16<BR>
0.000000000000E+00
-0.747798812221E+00
0.000000000000E+00<BR>
0.111022302463E-15
0.000000000000E+00
-0.747798812221E+00<BR>
0.000000000000E+00
0.832667268469E-16 0.000000000000E+00<BR>
1 2<BR>
0.000000000000E+00
0.000000000000E+00
0.111022302463E-15<BR>
0.000000000000E+00
-0.111022302463E-15
0.000000000000E+00<BR>
0.111022302463E-15
0.000000000000E+00
0.111022302463E-15<BR>
0.000000000000E+00
-0.177369122767E+01
0.000000000000E+00<BR>
-0.111022302463E-15
0.000000000000E+00
-0.177369122767E+01<BR>
0.000000000000E+00
-0.111022302463E-15
0.000000000000E+00<BR>.......</DIV>
<DIV> </DIV>
<DIV>The results are different from the previous one. I wonder why
this happens? Does the symmetry helps to constrain the d3 dynamical matrix
elements, so that it cannot be simply neglected like that is done in pw and
ph?</DIV>
<DIV> </DIV>
<DIV>The other question goes into more details. I tried to see what effects the
symmetry will have on wavefunctions. Originally I thought as long as we get the
ground-state wavefunction for different bands on a k-mesh from pw.x, we can know
any of its value on any k, including k+q.(these are used by phonon I think). At
gamma point, it should be no different. I noticed that phq_init subroutine read
wavefunctions (from files to evc and evq, each corresponding to point k and k+q
if q is not 0), so I output them to an extra file to see how they look like. For
this I use 4.2.1 version (because I cannot change the already installed QE on
the cluster). When symmetry is active, I get:</DIV>
<DIV> </DIV>
<DIV> q = ( 0.000000000000000E+000 , 0.000000000000000E+000
,<BR> 0.000000000000000E+000 )<BR> nksq
=
20<BR> ************************************************<BR> npol
= 1<BR> npwx
= 411<BR> nbnd
= 4<BR> ik #
= 1
-----------------------------------<BR>
0.69783903837624816635E+00
-0.64649301435265482674E+00
0.69783903837624816635E+00
-0.64649301435265482674E+00<BR>
-0.19216558783363027322E-16
0.87586028340881804830E-16
-0.19216558783363027322E-16
0.87586028340881804830E-16<BR>
0.55275566596732210915E-16
0.32076646606791784316E-17
0.55275566596732210915E-16
0.32076646606791784316E-17<BR>
0.23060842986941295573E-17
-0.42753486774548910442E-16
0.23060842986941295573E-17
-0.42753486774548910442E-16<BR>
0.99051076923549141728E-01
0.37832014530605620820E-02
0.99051076923549141728E-01
0.37832014530605620820E-02<BR>......(these shows the first several numbers for G
at k=0,q=0,ibnd=1,ig=1..ngm)</DIV>
<DIV> </DIV>
<DIV>When symmetry is turnned off, I get:</DIV>
<DIV> </DIV>
<DIV> q = ( 0.000000000000000E+000 , 0.000000000000000E+000
,<BR> 0.000000000000000E+000 )<BR> nksq
=
343<BR> ************************************************<BR> npol
= 1<BR> npwx
= 411<BR> nbnd
= 4<BR> ik #
= 1
-----------------------------------<BR>
-0.49664369043450279362E+00
0.81134307529730043118E+00
-0.49664369043450279362E+00
0.81134307529730043118E+00<BR>
-0.27818613173485040276E-10
0.54661997423613342026E-10
-0.27818613173485040276E-10
0.54661997423613342026E-10<BR>
0.43853898353156132494E-10
-0.42976651079895819979E-10
0.43853898353156132494E-10
-0.42976651079895819979E-10<BR>
0.63231205525296367916E-11
-0.10330070651313073490E-10
0.63231205525296367916E-11
-0.10330070651313073490E-10<BR>
-0.96373137459236146718E-01
0.23187212493507927680E-01
-0.96373137459236146718E-01
0.23187212493507927680E-01<BR>.......(at k=0,q=0,ibnd=1,ig=1..ngm)</DIV>
<DIV> </DIV>
<DIV>Which confuses me is that they look quite different, because I orginally
thought they should be the same as long as we are looking at the same point. I
thought maybe this has something to do with the d3 code. But I have no
idea. Besides, although phonon code output different wavefunctions at
the same k-point, they give the same final results (dynamical matrix and
frequencies). So I am wondering how phonon do the sum over k-mesh and
why the wavefunctions would be different for the same point and band.</DIV>
<DIV> </DIV>
<DIV>I will appreciate a lot for any of your suggestions. Thanks!</DIV>
<DIV> </DIV>
<DIV>Regards,</DIV>
<DIV> </DIV>
<DIV>Jiawei</DIV>
<DIV>Massachusetts Institute of Technology</DIV>
<DIV> </DIV>
<DIV> </DIV>
<DIV>---------------------------------</DIV>
<DIV></DIV>
<DIV>Jiawei Zhou, <A
href="mailto:zhoujw20@gmail.com">zhoujw20@gmail.com</A></DIV>
<DIV>2014/4/22 </DIV></BODY></HTML>