From seung43210 at yahoo.com  Mon Nov  2 22:02:16 2009
From: seung43210 at yahoo.com (seunghwan lee)
Date: Mon, 2 Nov 2009 13:02:16 -0800 (PST)
Subject: [Wannier] Initial choice for projection
Message-ID: <90987.98350.qm@web50508.mail.re2.yahoo.com>

Dear Young-Su and Nick,

I looked at the initial orbitals I set in the .win file and they are 
actually consistent with what you suggested. I tried 2 unit cell as well
and got the same error message. (I attached .win file and scf input file 
for 2 unit cell)

   In dis_proj_froz... done
     1    95    95  1.001183578513  0.000000000000
 The trial orbitals for disentanglement are not orthonormal
 Exiting.......
 Error in dis_main: orthonormal error 1

It looks like the problem is in NORMALIZATION of orbital #95 (s-orbital, 
if W90 keep the ordering of initial orbitals), not orthogonality.
But it is difficult to understand that same orbitals defined in W90 are 
normalized at one location and not normalized at another location.

There is a couple of even more strange behavior.

[1] If I change the number of significant figures for the location of
initial orbital, the nomalization changes. For example, if I use

c= 3.3780, -0.7128, -0.6157 :pz :z = 3.3780, -0.7128, 0.0000 :x=0,0,1

the error message was

--------------------------------------------------------------
 1    95    95  1.006000381851  0.000000000000
 The trial orbitals for disentanglement are not orthonormal
 Exiting.......
-------------------------------------------------------------

But, if I use

c= 3.378034, -0.712837, -0.615700 :pz :z = 3.378034, -0.712837, 0.000000 :x=0,0,1

the error message was

--------------------------------------------------------------
     1    95    95  1.001183578513  0.000000000000
 The trial orbitals for disentanglement are not orthonormal
 Exiting.......
-------------------------------------------------------------

Why normalization is anything to do with the location of orbitals?

[2] If I change the order of initial orbitals in the input file,
I am still getting the error for orbital #95. So, it is not really
a problem of orbital #95 in the previous input file.

It is possible that I made some stupid error in the input file 
(see my attached input), but it looks like there is something strange 
going on with normalization for a large carbon nanotube. But, PRL paper
(Lee, Nardelli and Marzari) used even larger nanotube and, so it has to 
work. Any idea what might cause this kind of behavior? I am wondering
if you used different initial projections for 5 unit cell calculations 
reported in PRL paper.

Thank you. I really appreciate your help.

Seunghwan Lee




--- On Thu, 10/29/09, Young-Su Lee <lee0su at kist.re.kr> wrote:

> From: Young-Su Lee <lee0su at kist.re.kr>
> Subject: RE: [Wannier] Initial choice for projection
> To: "'seunghwan lee'" <seung43210 at yahoo.com>
> Cc: Wannier at quantum-espresso.org
> Date: Thursday, October 29, 2009, 9:20 PM
> 
> >I thought "random" means pz orbital on each carbon and
> s orbitals on random 
> >places (according to the user manual). But, still, it
> can't be a good
> >choice as you said. Where should I look to get
> H-matrix?
> 
> Yes, you are right. I got confused.
> To get H-matrix, turn on hr_plot option:
> 
> hr_plot = .true.
> 
> then you'll have seedname_hr.dat
> If you run transport part, then the code outputs "Maximum
> imaginary part of the real-space Hamiltonian".
> 
> >In nanotube case, I have at least an example to
> compare. But, if I have a
> >system that has never been investigated, how do I know
> that the results of
> >W90 calculation are correct if the results depend so
> much on the initial choice of projection?
> 
> But, if you start from reasonably good initial guesses,
> they'll give you the same WFs after dis+loc in most (not
> all) cases.
>   
> You may check things like these
>  1) Do WFs match well with our chemical intuition of
> (anti-)bonding?
>  2) Are all the elements of H matix real?
>  3) Do the interpolated band structure look okay?
> 
> Young-Su
> 
> 
>

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: cnt55.scf.in
Type: application/octet-stream
Size: 2304 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/wannier/attachments/20091102/e3996b97/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: cnt55.win
Type: application/octet-stream
Size: 9347 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/wannier/attachments/20091102/e3996b97/attachment-0001.obj>

From lee0su at kist.re.kr  Wed Nov  4 03:44:46 2009
From: lee0su at kist.re.kr (Young-Su Lee)
Date: Wed, 4 Nov 2009 11:44:46 +0900
Subject: [Wannier] Initial choice for projection
In-Reply-To: <90987.98350.qm@web50508.mail.re2.yahoo.com>
References: <90987.98350.qm@web50508.mail.re2.yahoo.com>
Message-ID: <000a01ca5cf8$c7367480$55a35d80$@re.kr>

Dear Seunghwan,

I ran espresso/wannier90 with your input file to check whether there is a
bug in dis_proj_froz routine.
But, I couldn't reproduce the error you mentioned. Wannier90 finishes
without an error. 
Just one thing is that your frozen window is a bit high and seems to include
some states that should not be frozen.
Reducing it to 1.5 gives a good result. You can compare how they are
different - convergence, WF spread, imaginary part of H matrix.
Not sure what went wrong in  your calculation...

Young-Su





-----Original Message-----
From: seunghwan lee [mailto:seung43210 at yahoo.com] 
Sent: Tuesday, November 03, 2009 6:02 AM
To: Young-Su Lee
Cc: Wannier at quantum-espresso.org
Subject: RE: [Wannier] Initial choice for projection

Dear Young-Su and Nick,

I looked at the initial orbitals I set in the .win file and they are
actually consistent with what you suggested. I tried 2 unit cell as well and
got the same error message. (I attached .win file and scf input file for 2
unit cell)

   In dis_proj_froz... done
     1    95    95  1.001183578513  0.000000000000
 The trial orbitals for disentanglement are not orthonormal  Exiting.......
 Error in dis_main: orthonormal error 1

It looks like the problem is in NORMALIZATION of orbital #95 (s-orbital, if
W90 keep the ordering of initial orbitals), not orthogonality.
But it is difficult to understand that same orbitals defined in W90 are
normalized at one location and not normalized at another location.

There is a couple of even more strange behavior.

[1] If I change the number of significant figures for the location of
initial orbital, the nomalization changes. For example, if I use

c= 3.3780, -0.7128, -0.6157 :pz :z = 3.3780, -0.7128, 0.0000 :x=0,0,1

the error message was

--------------------------------------------------------------
 1    95    95  1.006000381851  0.000000000000
 The trial orbitals for disentanglement are not orthonormal  Exiting.......
-------------------------------------------------------------

But, if I use

c= 3.378034, -0.712837, -0.615700 :pz :z = 3.378034, -0.712837, 0.000000
:x=0,0,1

the error message was

--------------------------------------------------------------
     1    95    95  1.001183578513  0.000000000000
 The trial orbitals for disentanglement are not orthonormal  Exiting.......
-------------------------------------------------------------

Why normalization is anything to do with the location of orbitals?

[2] If I change the order of initial orbitals in the input file, I am still
getting the error for orbital #95. So, it is not really a problem of orbital
#95 in the previous input file.

It is possible that I made some stupid error in the input file (see my
attached input), but it looks like there is something strange going on with
normalization for a large carbon nanotube. But, PRL paper (Lee, Nardelli and
Marzari) used even larger nanotube and, so it has to work. Any idea what
might cause this kind of behavior? I am wondering if you used different
initial projections for 5 unit cell calculations reported in PRL paper.

Thank you. I really appreciate your help.

Seunghwan Lee

-------------- next part --------------
A non-text attachment was scrubbed...
Name: cnt55.wout_1.8
Type: application/octet-stream
Size: 564321 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/wannier/attachments/20091104/c6ed6571/attachment.obj>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: cnt55.wout_1.5
Type: application/octet-stream
Size: 492125 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/wannier/attachments/20091104/c6ed6571/attachment-0001.obj>

From a.mostofi at imperial.ac.uk  Sat Nov 21 21:24:16 2009
From: a.mostofi at imperial.ac.uk (Arash Mostofi)
Date: Sat, 21 Nov 2009 20:24:16 +0000
Subject: [Wannier] Multiple calls in library mode
In-Reply-To: <200909251634.35245.wieferink@uni-muenster.de>
References: <200909251634.35245.wieferink@uni-muenster.de>
Message-ID: <4B084C70.6080008@imperial.ac.uk>

Dear Juergen,

Thanks for your email and for pointing out this "feature". It arises 
because the value of the logical variable first_pass, defined in 
wann_omega, is preserved between successive calls to wannier_run(). 
We'll be sure to fix this in the next release.

Best wishes and thanks once again,

Arash

:------------------------------------------------------------:
: Dr. Arash A. Mostofi           :: a.mostofi at imperial.ac.uk :
: Lecturer and RCUK Fellow       ::                          :
: Depts. of Materials & Physics  ::                          :
: Imperial College London        :: T  +44 (0)207 594 8154   :
: London SW7 2AZ, United Kingdom :: F  +44 (0)207 594 6757   :
:------- http://www.cmth.ph.ic.ac.uk/people/a.mostofi -------:

Juergen Wieferink wrote:
> Hi list,
> 
> recently, I have implemented spin-polarization within my DFT code.
> The simplest way to adjust the interface to wannier90 seemed to be
> to call wannier_setup() and wannier_run() separatly for each spin
> channel.  But these subroutines are not designed to be called more
> than once within the run of a program.
> 
> After applying the attached patch, it seems to work, however.  Are
> there any other known problems with calling wannier_setup() and
> wannier_run() more than once?
> 
> Best regards,
> Juergen
> 
> 
> --
> Juergen Wieferink
> AG Pollmann
> Institute of Solid State Physics
> University of Muenster
> 
> 


From a.mostofi at imperial.ac.uk  Sat Nov 21 21:25:43 2009
From: a.mostofi at imperial.ac.uk (Arash Mostofi)
Date: Sat, 21 Nov 2009 20:25:43 +0000
Subject: [Wannier] Convergence problem of wannierisation
In-Reply-To: <EECF718003702B4F89104E3DBAA0AD03077C989F71@MBX-CLUSTER01.ad.fz-juelich.de>
References: <EECF718003702B4F89104E3DBAA0AD03077C989F71@MBX-CLUSTER01.ad.fz-juelich.de>
Message-ID: <4B084CC7.90605@imperial.ac.uk>

Dear Hongbin,

Others who have studied this particular system may be able to give more
specific advice.

The first things I would check are that you are including a sufficient
number of conduction bands for the disentanglement, using an appropriate
inner (or "frozen") wiindow and appropriate initial guesses
("projections") for your Wannier functions.

Best wishes,

Arash

:------------------------------------------------------------:
: Dr. Arash A. Mostofi           :: a.mostofi at imperial.ac.uk :
: Lecturer and RCUK Fellow       ::                          :
: Depts. of Materials & Physics  ::                          :
: Imperial College London        :: T  +44 (0)207 594 8154   :
: London SW7 2AZ, United Kingdom :: F  +44 (0)207 594 6757   :
:------- http://www.cmth.ph.ic.ac.uk/people/a.mostofi -------:

Zhang, Hongbin wrote:
> Dear all,
> 
> At present I am trying to play with Wannier 90 to calculate some properties like anomalous Hall effect. However, I found that sometimes it is hard to get the convergence of minimazation of both the Omega_I and Omega, that is, disentanglement and wannierization. This is true especially for the latter. I have tried on hcp Co, following the paper by E. Roman (Phys. Rev. Lett. 103, 097203). However, after 5000 iterations, the wannierisation converges to only 1E-7 (enough already? )  like below (Disentanglement has been achieved to 1E-10):
> 
>    4984    -0.148E-06     0.0003041075      106.4767765619   90254.25  <-- CONV
>    4985     0.875E-05     0.0001849761      106.4767853125   90269.00  <-- CONV
>    4986    -0.736E-05     0.0023025636      106.4767779520   90288.01  <-- CONV
>    4987    -0.155E-05     0.0010750558      106.4767764030   90301.06  <-- CONV
>    4988    -0.660E-06     0.0005507459      106.4767757427   90312.06  <-- CONV
>    4989    -0.856E-07     0.0002900457      106.4767756571   90329.45  <-- CONV
>    4990    -0.695E-07     0.0001142533      106.4767755876   90345.29  <-- CONV
>    4991     0.000E+00     0.0001822559      106.4767755865   90350.65  <-- CONV
>    4992    -0.436E-07     0.0001817570      106.4767755429   90369.36  <-- CONV
>    4993    -0.488E-07     0.0001090447      106.4767754941   90383.58  <-- CONV
>    4994     0.000E+00     0.0001206043      106.4767754934   90387.75  <-- CONV
>    4995    -0.697E-07     0.0001203373      106.4767754237   90403.39  <-- CONV
>    4996    -0.312E-06     0.0005823860      106.4767751115   90416.78  <-- CONV
>    4997    -0.471E-07     0.0001267525      106.4767750644   90433.15  <-- CONV
>    4998    -0.410E-07     0.0001274398      106.4767750233   90451.88  <-- CONV
>    4999     0.000E+00     0.0001121615      106.4767750226   90459.01  <-- CONV
>    5000    -0.797E-07     0.0001117099      106.4767749430   90477.35  <-- CONV
> 
> I have tried also different initial projections, but it does not help so much. So could you please tell me how to speed up the convergence process? And how to achieve better convergence?
> 
> I do not see a lot parameters to play with from the input file. Any suggestion is appreciated.
> 
> Best wishes,
> 
> Hongbin Zhang
> 
> ------------------------------------------------------------------------------------------------
> ------------------------------------------------------------------------------------------------
> Forschungszentrum Juelich GmbH
> 52425 Juelich
> Sitz der Gesellschaft: Juelich
> Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498
> Vorsitzende des Aufsichtsrats: MinDir'in Baerbel Brumme-Bothe
> Geschaeftsfuehrung: Prof. Dr. Achim Bachem (Vorsitzender),
> Dr. Ulrich Krafft (stellv. Vorsitzender), Prof. Dr.-Ing. Harald Bolt,
> Prof. Dr. Sebastian M. Schmidt
> ------------------------------------------------------------------------------------------------
> ------------------------------------------------------------------------------------------------
> _______________________________________________
> Wannier mailing list
> Wannier at quantum-espresso.org
> http://www.democritos.it/mailman/listinfo/wannier


From a.mostofi at imperial.ac.uk  Sat Nov 21 21:18:59 2009
From: a.mostofi at imperial.ac.uk (Arash Mostofi)
Date: Sat, 21 Nov 2009 20:18:59 +0000
Subject: [Wannier] Convergence problem of wannierisation
In-Reply-To: <EECF718003702B4F89104E3DBAA0AD03077C989F71@MBX-CLUSTER01.ad.fz-juelich.de>
References: <EECF718003702B4F89104E3DBAA0AD03077C989F71@MBX-CLUSTER01.ad.fz-juelich.de>
Message-ID: <4B084B33.9000900@imperial.ac.uk>

Dear Hongbin,

Others who have studied this particular system may be able to give more 
specific advice.

The first things I would check are that you are including a sufficient 
number of conduction bands for the disentanglement, using an appropriate 
inner (or "frozen") wiindow and appropriate initial guesses 
("projections") for your Wannier functions.

Best wishes,

Arash


:------------------------------------------------------------:
: Dr. Arash A. Mostofi           :: a.mostofi at imperial.ac.uk :
: Lecturer and RCUK Fellow       ::                          :
: Depts. of Materials & Physics  ::                          :
: Imperial College London        :: T  +44 (0)207 594 8154   :
: London SW7 2AZ, United Kingdom :: F  +44 (0)207 594 6757   :
:------- http://www.cmth.ph.ic.ac.uk/people/a.mostofi -------:

Zhang, Hongbin wrote:
> Dear all,
> 
> At present I am trying to play with Wannier 90 to calculate some properties like anomalous Hall effect. However, I found that sometimes it is hard to get the convergence of minimazation of both the Omega_I and Omega, that is, disentanglement and wannierization. This is true especially for the latter. I have tried on hcp Co, following the paper by E. Roman (Phys. Rev. Lett. 103, 097203). However, after 5000 iterations, the wannierisation converges to only 1E-7 (enough already? )  like below (Disentanglement has been achieved to 1E-10):
> 
>    4984    -0.148E-06     0.0003041075      106.4767765619   90254.25  <-- CONV
>    4985     0.875E-05     0.0001849761      106.4767853125   90269.00  <-- CONV
>    4986    -0.736E-05     0.0023025636      106.4767779520   90288.01  <-- CONV
>    4987    -0.155E-05     0.0010750558      106.4767764030   90301.06  <-- CONV
>    4988    -0.660E-06     0.0005507459      106.4767757427   90312.06  <-- CONV
>    4989    -0.856E-07     0.0002900457      106.4767756571   90329.45  <-- CONV
>    4990    -0.695E-07     0.0001142533      106.4767755876   90345.29  <-- CONV
>    4991     0.000E+00     0.0001822559      106.4767755865   90350.65  <-- CONV
>    4992    -0.436E-07     0.0001817570      106.4767755429   90369.36  <-- CONV
>    4993    -0.488E-07     0.0001090447      106.4767754941   90383.58  <-- CONV
>    4994     0.000E+00     0.0001206043      106.4767754934   90387.75  <-- CONV
>    4995    -0.697E-07     0.0001203373      106.4767754237   90403.39  <-- CONV
>    4996    -0.312E-06     0.0005823860      106.4767751115   90416.78  <-- CONV
>    4997    -0.471E-07     0.0001267525      106.4767750644   90433.15  <-- CONV
>    4998    -0.410E-07     0.0001274398      106.4767750233   90451.88  <-- CONV
>    4999     0.000E+00     0.0001121615      106.4767750226   90459.01  <-- CONV
>    5000    -0.797E-07     0.0001117099      106.4767749430   90477.35  <-- CONV
> 
> I have tried also different initial projections, but it does not help so much. So could you please tell me how to speed up the convergence process? And how to achieve better convergence?
> 
> I do not see a lot parameters to play with from the input file. Any suggestion is appreciated.
> 
> Best wishes,
> 
> Hongbin Zhang
> 
> ------------------------------------------------------------------------------------------------
> ------------------------------------------------------------------------------------------------
> Forschungszentrum Juelich GmbH
> 52425 Juelich
> Sitz der Gesellschaft: Juelich
> Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498
> Vorsitzende des Aufsichtsrats: MinDir'in Baerbel Brumme-Bothe
> Geschaeftsfuehrung: Prof. Dr. Achim Bachem (Vorsitzender),
> Dr. Ulrich Krafft (stellv. Vorsitzender), Prof. Dr.-Ing. Harald Bolt,
> Prof. Dr. Sebastian M. Schmidt
> ------------------------------------------------------------------------------------------------
> ------------------------------------------------------------------------------------------------
> _______________________________________________
> Wannier mailing list
> Wannier at quantum-espresso.org
> http://www.democritos.it/mailman/listinfo/wannier