[Pw_forum] Error in routine bfgs (1): dE0s is positive which should never happen

Paolo Giannozzi

2014-09-22 19:52:05 UTC

This kind of errors invariably happens when you are very close
to the minimum and you have some numerical noise on forces. It
is useless in my opinion to insist: your system is sufficiently
relaxed

Paolo

Post by amitharani
Hi
I am getting an error { Error in routine bfgs (1): dE0s is positive
which should never happen }. The output file is attached for your kind
perusal.This is the 2nd time I am getting the same error consecutively.
I followed what Nick had suggested on the pwforum to take the relaxed
co-ordinates, make a new input file and resubmit, but it didn't
work.Thanks.
_______________________________________________
Pw_forum mailing list
Pw_forum at pwscf.org
http://pwscf.org/mailman/listinfo/pw_forum

--
Paolo Giannozzi, Dept. Chemistry&Physics&Environment,
Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
Phone +39-0432-558216, fax +39-0432-558222

Lorenzo Paulatto

2014-09-22 20:46:58 UTC

In this specific case, as dispersion correction is a relatively new
addition to the code, it /may/ also be caused by some minor bug causing
energy and forces to not be perfectly consistent. I can tell from
experience that even a difference as small as 10^-5 Ry can cause BFGS to
fail when close to the minimum.

It would take some very careful finite difference calculation to,
possibly performed by someone who knows the dispersion correction code,
to be 100% sure. However I would say that for most purposes you can
proceed with no worries, and eventually use damped dynamics for the last
part of the geometry optimization.

regards

Post by Paolo Giannozzi
This kind of errors invariably happens when you are very close
to the minimum and you have some numerical noise on forces. It
is useless in my opinion to insist: your system is sufficiently
relaxed
Paolo

--
Lorenzo Paulatto - Paris 18

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://pwscf.org/pipermail/pw_forum/attachments/20140922/23552dad/attachment.html

Alaska Subedi

2016-09-22 12:56:26 UTC

Dear Paolo,

Sorry to resurrect this after a long time.

This problem is innocuous but annoying nonetheless. I am using QE to
relax structures generated by USPEX. For around five percent of the
structures, I get this error. When this error happens, pw.x exits by
only printing atomic positions but not the cell parameters. This gives
problems to USPEX.

I tried to find which call to output_tau is responsible for this, to
no avail. I'd greatly appreciate if someone can point me to a
modification that will make the code output both the cell_parameters
and atomic_positions during exit after this error.

Thanks,
Alaska

On Mon, Sep 22, 2014 at 9:52 PM, Paolo Giannozzi

--
Paolo Giannozzi, Dept. Chemistry&Physics&Environment,
Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
Phone +39-0432-558216, fax +39-0432-558222
_______________________________________________
Pw_forum mailing list
http://pwscf.org/mailman/listinfo/pw_forum

Paolo Giannozzi

2016-09-22 16:03:06 UTC

Once the error is detected, the code stops immediately and prints nothing
useful. Where do you see that there are no cell parameters? And when:
during a variable-cell calculation, or also (or only) in a fixed-cell one?
If I remember correctly, cell parameters are reprinted on putput at each
step in the former case, never in the latter.

Unfortunately the code performing the various MD and structural
optimizations is rather obscure and it is not straightforward to figure out
how to modify it.

Paolo

Post by Alaska Subedi
Dear Paolo,
Sorry to resurrect this after a long time.
This problem is innocuous but annoying nonetheless. I am using QE to
relax structures generated by USPEX. For around five percent of the
structures, I get this error. When this error happens, pw.x exits by
only printing atomic positions but not the cell parameters. This gives
problems to USPEX.
I tried to find which call to output_tau is responsible for this, to
no avail. I'd greatly appreciate if someone can point me to a
modification that will make the code output both the cell_parameters
and atomic_positions during exit after this error.
Thanks,
Alaska
On Mon, Sep 22, 2014 at 9:52 PM, Paolo Giannozzi

_______________________________________________
Pw_forum mailing list
http://pwscf.org/mailman/listinfo/pw_forum

--
Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
Phone +39-0432-558216, fax +39-0432-558222

Alaska Subedi

2016-09-23 09:31:34 UTC

Post by Paolo Giannozzi
Once the error is detected, the code stops immediately and prints nothing
useful. Where do you see that there are no cell parameters? And when: during
a variable-cell calculation, or also (or only) in a fixed-cell one? If I
remember correctly, cell parameters are reprinted on putput at each step in
the former case, never in the latter.

In a vc-relax calculation, the code prints atomic_positions but not
cell_parameters after this error. Generally, vc-relax prints both
cell_parameters and atomic_positions. But this does not happen when
the code stops with this error. It would be nice if it also printed
cell_parameters before exiting.

Thanks,
Alaska

Post by Paolo Giannozzi
Unfortunately the code performing the various MD and structural
optimizations is rather obscure and it is not straightforward to figure out
how to modify it.
Paolo

_______________________________________________
Pw_forum mailing list
http://pwscf.org/mailman/listinfo/pw_forum

--
Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
Phone +39-0432-558216, fax +39-0432-558222
_______________________________________________
Pw_forum mailing list
http://pwscf.org/mailman/listinfo/pw_forum

Paolo Giannozzi

2016-09-23 11:58:12 UTC

Please send me an output showing the problem you mention. Paolo

Post by Paolo Giannozzi
Once the error is detected, the code stops immediately and prints nothing

during

Post by Paolo Giannozzi
a variable-cell calculation, or also (or only) in a fixed-cell one? If I
remember correctly, cell parameters are reprinted on putput at each step

Post by Paolo Giannozzi
the former case, never in the latter.

Post by Paolo Giannozzi
Unfortunately the code performing the various MD and structural
optimizations is rather obscure and it is not straightforward to figure

out

Post by Paolo Giannozzi
how to modify it.
Paolo

Post by amitharani
Hi
I am getting an error { Error in routine bfgs (1): dE0s is positive
which should never happen }. The output file is attached for your

kind

Post by amitharani
perusal.This is the 2nd time I am getting the same error

consecutively.

Post by amitharani
I followed what Nick had suggested on the pwforum to take the relaxed
co-ordinates, make a new input file and resubmit, but it didn't
work.Thanks.
_______________________________________________
Pw_forum mailing list
http://pwscf.org/mailman/listinfo/pw_forum

_______________________________________________
Pw_forum mailing list
http://pwscf.org/mailman/listinfo/pw_forum

--
Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
Phone +39-0432-558216, fax +39-0432-558222

Alaska Subedi

2016-09-23 16:13:33 UTC

Dear Paolo,

I tried to reproduce this problem at different nodes in the cluster,
and the problem seems to be node dependent. Let me verify that it is
not the problem of the environment before I take more of your time.

Best,
Alaska

Post by Paolo Giannozzi
Please send me an output showing the problem you mention. Paolo