From ogitsu at llnl.gov  Fri Sep  1 21:19:41 2006
From: ogitsu at llnl.gov (Tadashi Ogitsu)
Date: Fri, 1 Sep 2006 12:19:41 -0700
Subject: [Wannier] parameter to control convergence
In-Reply-To: <Pine.LNX.4.63.0608251039450.8161@localhost.localdomain>
References: <6C4C3303-C2C4-4035-B3EA-B6BCFF0E9B4E@llnl.gov> <44EC5E1C.2070608@mit.edu> <E95100A3-5C49-4579-9DCD-60B81B9B7D23@llnl.gov> <Pine.LNX.4.63.0608231145520.6741@localhost.localdomain> <4872F20A-6DA1-4F89-B4BA-67DFA0E467E3@llnl.gov> <Pine.LNX.4.63.0608251039450.8161@localhost.localdomain>
Message-ID: <D1B6D63F-37D7-45AC-9A01-8E77F6261C4F@llnl.gov>

>

Dear Jonathan,

Thanks for the new pw2wannier90 programs to save some memory size.  
Finally, I got it run without error/warning or crash. Now, let me  
come back to the original topic which is the assessment of the  
quality of MLWF.


> 2- That the MLWF are essentially real (the plotting routine tells  
> you this)
>

Smaller Im/Re ratio is better, right? Generally speaking, how small  
one should expect?

I've checked the routine. It calculate the phase at the position that  
the MLWF's amplitude gets max, and subtract the phase from the MLWF  
at each grid point. Then, calculate absolute value of Im/Re ratio  
where the MLWF's amplitude get large than 0.01, and printout the  
maximum number for each MLWF. Do I understand correctly?

With my beta-boron, MLWFs are very well localized (around 1.5A) but  
Im/Re ratio are still large, the largest one is 340(!). I'm wondering  
what is the implication (or wrong)... Is it mixing unconnected bands?  
Any comments/suggestion will be appreciated.

Best,
Tadashi



From ogitsu at llnl.gov  Fri Sep  1 21:44:08 2006
From: ogitsu at llnl.gov (Tadashi Ogitsu)
Date: Fri, 1 Sep 2006 12:44:08 -0700
Subject: [Wannier] interface with xcrysden
Message-ID: <FA492936-B1DC-4554-AA7D-7C76E1D499BA@llnl.gov>

Dear Wannier developers,

I seem to have gotten a problem in plotting the MLWF using xcrysden.  
Basically, xcrysden can not read a xsf file the wannier code produced.

I contacted Tone Kokalj, the developer of xcrysden, and he suggested  
me to notify the wannier developers as well.

I'll put the error log from xcrysden and the input file I've used in  
tar + gzip format. Let me know if you need more info. Thanks.

Tadashi
-------------- next part --------------
A non-text attachment was scrubbed...
Name: xcrysden.tgz
Type: application/octet-stream
Size: 668654 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/wannier/attachments/20060901/828d5513/attachment.obj>

From ogitsu at llnl.gov  Tue Sep  5 20:04:26 2006
From: ogitsu at llnl.gov (Tadashi Ogitsu)
Date: Tue, 5 Sep 2006 11:04:26 -0700
Subject: [Wannier] Fwd: XCrysDen question
References: <46676.130.192.118.84.1157478201.squirrel@nabiralnik.ijs.si>
Message-ID: <CFF791DD-E9B4-429D-BA4B-842C85039689@llnl.gov>

Dear Wannier90 developers,

I've reported that XCrysden got an error in reading the output (xsf  
file) from wannier90 last week. I've gotten the feedback from Tone,  
the developer of the crysden as below. So, it was not designed to  
deal with more than 100 atoms... I understand.

May I suggest not to use a fixed format as it often causes not only  
the format length problem but the precision problem. (i.e., you reuse  
the output data for further processing, where you don't have the  
double precision anymore) I know it is mostly for peace of mind, but  
people do crazy (unexpected use) things (like me)! Let me know if it  
sounds reasonable proposal.

Tadashi

Begin forwarded message:

> From: "Tone Kokalj" <tone.kokalj at ijs.si>
> Date: September 5, 2006 10:43:21 AM PDT
> To: "Tadashi Ogitsu" <ogitsu at llnl.gov>
> Subject: Re: XCrysDen question
> Reply-To: tone.kokalj at ijs.si
>
>> Hi Tone,
>>
>> Thanks for your help. I'm sending the error log file and the input
>> file. (Slightly different from the previous report.)
>
> The problem is the following (see the content of you XSF file):
>
> PRIMCOORD
> **  1
> B       0.0000006  -1.1451458  12.9473163
> ...
>
> The wannier developers assumed only two digits for number-of-atoms.
> Apparently you have more than 100 atoms, hence the stars.
>
> Correct the "**" to the actal number of atoms.
>
> Please, inform the wannier developers about this problem.
>
> Regards, Tone
>



From jryates at lbl.gov  Tue Sep  5 20:36:12 2006
From: jryates at lbl.gov (Jonathan Yates)
Date: Tue, 5 Sep 2006 11:36:12 -0700 (PDT)
Subject: [Wannier] parameter to control convergence
In-Reply-To: <D1B6D63F-37D7-45AC-9A01-8E77F6261C4F@llnl.gov>
References: <6C4C3303-C2C4-4035-B3EA-B6BCFF0E9B4E@llnl.gov> <44EC5E1C.2070608@mit.edu>
 <E95100A3-5C49-4579-9DCD-60B81B9B7D23@llnl.gov>
 <Pine.LNX.4.63.0608231145520.6741@localhost.localdomain>
 <4872F20A-6DA1-4F89-B4BA-67DFA0E467E3@llnl.gov>
 <Pine.LNX.4.63.0608251039450.8161@localhost.localdomain>
 <D1B6D63F-37D7-45AC-9A01-8E77F6261C4F@llnl.gov>
Message-ID: <Pine.LNX.4.63.0609051119500.23282@localhost.localdomain>

On Fri, 1 Sep 2006, Tadashi Ogitsu wrote:

> Smaller Im/Re ratio is better, right? Generally speaking, how small one 
> should expect?

  Yes. In all the systems I've looked at it has been possible to find MLWF 
with a Im/Re ratio < 0.001. However, my experience is with relatively 
small systems (<40WF).
  With the disentanglement procedure it is possible to find WF that are 
localised, but not real. I generally take this as an indication that I 
should extract a different number of WF, or change the energy windows.

> I've checked the routine. It calculate the phase at the position that the 
> MLWF's amplitude gets max, and subtract the phase from the MLWF at each grid 
> point. Then, calculate absolute value of Im/Re ratio where the MLWF's 
> amplitude get large than 0.01, and printout the maximum number for each MLWF. 
> Do I understand correctly?

  You do.


  Thanks for pointing out the bug in the plotting routine when the number 
of atoms is >100. Tone is right, just replace the ** with 105.

  I should add that the xsf file you sent has another problem which may 
prevent it from working; namely the file contains uninitialised data - it's 
not clear to me how this happened. But I'll contact you directly about it.

  Yours
   Jonathan


-- 
Dr Jonathan Yates:  email: jryates at lbl.gov   phone: 510-642-7302
Material Science Division, Lawrence Berkeley National Laboratory  and
Department of Physics, University of California,
366 Le Conte Hall #7300, Berkeley, CA 94720-7300, USA