Hi,

I have the same problem with my own vdw-df pseudopotentials. Everything
works fine with older version (< trunk 433). Starting from trunk 444, I
encountered ghost states. Using the fdf file generated from 433 to run with
444 does not solve the problem. A hidden new keyword, or a bug? I suggest
to use older version (< trunk 433).

Best regards,
Quan


On Tue, Jan 28, 2014 at 7:13 PM, Gomes Rocha <[email protected]> wrote:

>
> Maybe it can help change the compiler. Recently I have used cray compilers
> (ftn) and have been working really well. The optimization flags we used
> were -mavx -O3. Just a few more data in the calculation I ran: I still did
> not use the more recent siesta-trunk. I still have installed the 433. And
> since you have carbon atoms in the system, I also send the rc data that
> came out from one of my log files:
>
> Ground state valence configuration:   2s02  2p02
> Reading pseudopotential information in formatted form from C.psf
>
> Valence configuration for pseudopotential generation:
> 2s( 2.00) rc: 1.54
> 2p( 2.00) rc: 1.54
> 3d( 0.00) rc: 1.54
> 4f( 0.00) rc: 1.54
> Ground state valence configuration:   1s01
> Reading pseudopotential information in formatted form from H.psf
>
> Valence configuration for pseudopotential generation:
> 1s( 1.00) rc: 1.31
> 2p( 0.00) rc: 1.31
> 3d( 0.00) rc: 0.37
> 4f( 0.00) rc: 1.31
> Ground state valence configuration:   3s02  3p01
>
> I also used:
>
> %block PS.lmax
>    C   3
>    H   2
> %endblock PS.lmax
>
> and
>
> XC.functional     VDW
> XC.authors        DRSLL
>
> C. G. Rocha
>
>
>
> On 28 January 2014 19:49, Bailey, Steven <[email protected]> wrote:
>
>>  Thanks Gomes
>>
>> Yes this is how I created my VDW pseudopotentials. I have just tried your
>> H.psf and still have ghost states at L=1. Could this be an issue with intel
>> compilers? I have even reduced the flags to -O0 but again with no luck.
>>
>>
>>
>> Steve
>>
>>
>>
>> Dr S Bailey
>>
>> Department of Physics
>>
>> Lancaster University UK
>>
>> Email: [email protected]
>>
>> Tel: +441524592844
>>
>>
>>
>> *From:* [email protected] [mailto:[email protected]] *On
>> Behalf Of *Gomes Rocha
>> *Sent:* 28 January 2014 17:39
>> *To:* [email protected]
>> *Subject:* Re: [SIESTA-L] vdW
>>
>>
>>
>> I also had problems in this matter. I do not know if it will help but
>> what I did was to take LDA inputs for atom generator and I just changed the
>> flavour of the potential to "vw" and generated a new psf file. At least, I
>> got reasonable simulations with it. I attached the pseudopotential I used
>> for the hydrogen.
>>
>> best
>>
>> C. G. Rocha
>>
>>
>>
>> On 28 January 2014 15:34, Bailey, Steven <[email protected]>
>> wrote:
>>
>> Siesta users
>>
>> I have compiled siesta-trunk-453 on intel/13.0. This is of course the
>> serial version which I want to test out before implementing
>> openmpi/1.6.5-intel.
>>
>> I generate DRSLL VDW pseudopotentials in atom and try to run a system of
>> carbon with attached hydrogens. Unfortunately the Hvdw.psf always throws up
>> an error claiming the detection of ghost states for L=1.
>>
>> I have tried many variations of rc in the pseudopotential generation with
>> no success. I have tried imposing %block PS.KBprojectors projector limits
>> with no success.
>>
>> Eliminating the hydrogen from the system allows the calculation to run
>> and gives reasonable binding energies between carbon sheets.
>>
>> I have tried using the LMKLL functional and interestingly if I accidently
>> leave the XC.authors as DRSLL the ghost state error does not appear but the
>> program then crashes with references to allocation errors.
>>
>> Way back I compiled siesta-trunk-3 something using a pgi compiler. Using
>> the same pseudopotentials ( generated in the same way in atom) the
>> calculations finished and the binding energies were very good. I could well
>> go back to this version to try out using the intel/13.0 compiler if I could
>> find it!
>>
>> Any advice to resolve this problem would be very useful.
>>
>>
>>
>> Dr S Bailey
>>
>> Department of Physics
>>
>> Lancaster University UK
>>
>> Email: [email protected]
>>
>> Tel: +441524592844
>>
>>
>>
>>
>>
>
>

Responder a