Dear Alberto,
Thank you for your file. I looked carefully at this file and performed
several tests. What I found out may be useful.
Here's what I did.
I compared the optical calculation in SIESTA 2.0 and SIESTA 3.0 versions
playing with TYPE and SIZE of the basis set.
It turns out that in SIESTA 2.0 there is no problem with this calculation
no matter which basis size/type you use (I checked and found that
up to DZP all types of basis are ok). In contrast, in SIESTA 3.0 SZ basis
size should not cause any problems no matter which basis type you want
(split, splitgauss, node, nonodes), DZ is ok only with nodes basis type
whereas for SZP and DZP there is no available basis type compatible with
optical calculations. I think that in SIESTA 3.0 the inclusion of
polarization orbitals is somehow in conflict with optical calculations
which leads to the
"linear dependence of basis vectors". Perhaps, there is a way how to fix it
in the MINVEC.f program,I don't know how.
In SIESTA 3.1 I did not experience any problem with different combinations
of basis type vs. basis size, DZP is available.
Hope, this helps.
Best,
Artem Baskin
Chemistry Department
University of Illinois at Chicago
On Mon, Jan 30, 2012 at 3:41 AM, Alberto <aeljar...@el.ub.es> wrote:
> Oops, I forgot to add the .fdf file.
>
> Fixed!
>
>
> 2012/1/30 Alberto <aeljar...@el.ub.es>
>
>> Dear Artem,
>>
>> Thank you very much for your reply. It seems to me too that we are
>> dealing with a bug of the program, as you point out in you mail.
>>
>> Together with this mail you will fing a "minimal working example" of my
>> calculation, I performed the same calculation as the one in this .fdf
>>
>
> file that I send now, but only in the Gamma point, and the error message
>> persisted. Anyway, I will keep trying to sort out this problem by any
>> means.
>>
>> Thank you again,
>>
>> Alberto Eljarrat
>> Departamet d'Electrònica,
>> Universitat de Barcelona.
>>
>> 2012/1/27 Artem Baskin <abas...@gmail.com>
>>
>>> Dear Alberto,
>>>
>>> I have also experienced the same problem with siesta-3.0 version and
>>> could not not find any solution. However, what I have found may relevant
>>> and pertinent to your embarrassment.
>>>
>>> I was trying to calculate optical properties of one atomic layer of
>>> graphene (as a test system for siesta 3.0) and I got the same error
>>> ('MINVEC:
>>> BASIS VECTORS ARE LINEARLY DEPENDENT').
>>> My lattice vectors were absolutely ok, moreover, this error appears only
>>> when "Optical.calculation .true." , like in your case. Then I looked
>>> throughout siesta archive and found many users experiencing
>>> the same problem. Marcos Verissimo proposed that "by basis vectors
>>> siesta does not mean the lattice vectors, it's related to the basis
>>> functions instead."
>>> As other siesta user suggested, I deceased the size of one of my lattice
>>> vectors in such a way that instead of planar structure I have the bulky
>>> material (graphite, or so). When I did that the error message
>>> no longer appears but SystemLabel.EPSIGM file was not created and
>>> calculations stopped right before the evaluation of dipole moment. This
>>> looks weird.
>>>
>>> Then I contacted Dr. Daniel Sanchez and he assumed that "something went
>>> corrupted with this bit of the code during the migration to latest
>>> version". Probably this is a bug. I had to go back to siesta 2.0.
>>> It's even more surprising that you've managed to get your optical
>>> properties with siesta 3.1.
>>> Up to now, that's all I know about this issue.
>>>
>>> I would be very obliged if you could send me your .fdf file to check it.
>>>
>>> Best regards,
>>>
>>> Artem Baskin
>>> Chemistry Department
>>> University of Illinois at Chicago
>>>
>>>
>>>
>>>
>>>
>>> On Thu, Jan 26, 2012 at 4:10 AM, Alberto <aeljar...@el.ub.es> wrote:
>>>
>>>> Dear Siesta Users,
>>>>
>>>> I'm trying to use siesta-ldau to improve an optical calculation of
>>>> Hexagonal GaN. Former calculations in siesta-3.1 with GGAs and LDAs where
>>>> working just nice, but now (using siesta-ldau) i'm getting an error message
>>>> without even changing the calculation: just normal LDA or GGA.
>>>>
>>>> The error message appears only when "Optical.calculation .true." and
>>>> goes like this:
>>>>
>>>> 'MINVEC: BASIS VECTORS ARE LINEARLY DEPENDENT'
>>>>
>>>> Repeated once for each of the processor if running a parallel
>>>> calculation.
>>>>
>>>> If one searches for the routine minvec.f in the siesta-ldau/Src folder,
>>>> something like this is written in the comments:
>>>>
>>>> C *******************************************************************
>>>> C FINDS THE LATTICE BASIS OF MINIMUM LENGTH, I.E. SUCH TAHT ANY
>>>> C OTHER BASIS (NOT EQUIVALENT BY SYMMETRY) HAS ONE VECTOR LONGER.
>>>> C WRITTEN BY J.MORENO AND J.SOLER. AUGUST 1989 AND OCTOBER 1997.
>>>> C ********* INPUT ***************************************************
>>>> C REAL*8 B0 : Cell vectors B0(xyj,vector)
>>>> C ********* OUTPUT **************************************************
>>>> C REAL*8 BMIN : Minimum cell vectors B0(xyj,vector)
>>>> C *******************************************************************
>>>>
>>>> I guess this error has something to do with the lattice vectors. I run
>>>> the calculation with a fixed lattice like this:
>>>>
>>>> outcell: Cell vector modules (Ang) : 3.208006 3.208006
>>>> 5.232989
>>>> outcell: Cell angles (23,13,12) (deg): 90.0000 90.0000
>>>> 120.0000
>>>> outcell: Cell volume (Ang**3) : 46.6392
>>>>
>>>> (I'm pasting the result of "grep outcell output.fdf")
>>>>
>>>> A quick investigation of the mailing-list showed that the misterious
>>>> force behind this error message has not been yet revealed, although some
>>>> people have been complaining about getting the same error message. Everyone
>>>> of them had been trying to execute optical calculations (which I have been
>>>> perfectly able of doing in the my former siesta experiences).
>>>>
>>>> Does anyone have any kind of insight or clue for this problem? If so,
>>>> help would be greatly appreciated.
>>>>
>>>> Thank you in advance,
>>>>
>>>> Alberto Eljarrat Ascunce
>>>> Departamet d'Electrònica,
>>>> Universitat de Barcelona.
>>>>
>>>>
>>>
>>
>
