[gpaw-users] Problem with the Davidson eigensolver

[Jussi Enkovaara]

On 2013-10-31 13:35, Jens Jørgen Mortensen wrote:
> Den 31-10-2013 09:17, Ulrik Grønbjerg Vej-Hansen skrev:
>> Dear list
>>
>> I have just begun to use the Davidson eigensolver, in an attempt to
>> handle semi-core p electrons without resorting to cg. However, I have
>> a encountered an error which I do not know how to solve, and I hope
>> you can help me out. I have attached the script and the output file
>> from gpaw, and the wrror message is found below.
>
> Jussi:  What would you suggest that the Davidson algorithm should do
> when this happens?

Hi,
I think that the reason for the error is that there is some linear 
dependency in the basis formed by wavefunctions and preconditioned 
residuals. The linear dependency comes mostly like from some residuals 
being very small. I try to investigate this particular case in more 
detail, but one could try to make check that if preconditioned residual 
is very small it won't be included in Davidson subspace.

Best regards,
Jussi


>
> Jens Jørgen
>
>>
>> Best wishes
>> Ulrik
>>
>> Traceback (most recent call last):
>>   File "relax-structure.py", line 72, in <module>
>>     qn.run(fmax=0.05)
>>   File "/home/camp/ugan/ase/ase/optimize/optimize.py", line 114, in run
>>     f = self.atoms.get_forces()
>>   File "/home/camp/ugan/ase/ase/atoms.py", line 685, in get_forces
>>     forces = self._calc.get_forces(self)
>>   File "/home/camp/ugan/gpaw/gpaw/aseinterface.py", line 69, in
>> get_forces
>>     force_call_to_set_positions=force_call_to_set_positions)
>>   File "/home/camp/ugan/gpaw/gpaw/paw.py", line 270, in calculate
>>     self.occupations):
>>   File "/home/camp/ugan/gpaw/gpaw/scf.py", line 46, in run
>>     wfs.eigensolver.iterate(hamiltonian, wfs)
>>   File "/home/camp/ugan/gpaw/gpaw/eigensolvers/eigensolver.py", line
>> 64, in iterate
>>     e, psit_nG = self.iterate_one_k_point(hamiltonian, wfs, kpt)
>>   File "/home/camp/ugan/gpaw/gpaw/eigensolvers/davidson.py", line 154,
>> in iterate_one_k_point
>>     general_diagonalize(H_2n2n, eps_2n, S_2n2n)
>>   File "/home/camp/ugan/gpaw/gpaw/utilities/lapack.py", line 84, in
>> general_diagonalize
>>     raise RuntimeError('general_diagonalize error: %d' % info)
>> RuntimeError: general_diagonalize error: 0
>> GPAW CLEANUP (node 0): <type 'exceptions.RuntimeError'> occurred.
>> Calling MPI_Abort!
>>
>> Ulrik Grønbjerg Vej-Hansen
>> PhD student
>> DTU Physics
>>
>> Technical University of Denmark
>>
>>
>> Department of Physics
>> Street
>> Building 307
>> DK - 2800 Kgs. Lyngby
>> Denmark
>> Direct +45 45253223
>> ulrik.groenbjerg at fysik.dtu.dk <mailto:ugan at fysik.dtu.dk>
>> www.fysik.dtu.dk <http://www.fysik.dtu.dk/>
>>
>
>
>