[CP2K-user] [CP2K:21468] EPS_DEFAULT Meaning
Marcella Iannuzzi
marci.akira at gmail.com
Wed May 14 14:45:48 UTC 2025
Dear ?
As you might have noticed, the EPS_DEFAULT is in a section of the input,
the QS section, where several other EPS_* are present.
Some of these can be controlled all at once by setting EPS_DEFAULT. Setting
them individually overwrites the eps_default derived value.
Some more details on the meaning of each of them can be revealed by their
description.
If you have still doubts on what they refer to, you might have a look at
the original QS papers.
Using default values for all of them is in general a good choice, if you
do not have a reason for doing otherwise.
EPS_SCF, as other "accuracy control parameters", is instead not
controlled by EPS_DEFAULT
The exact meaning of EPS_SCF depends on the optimisation method in use, but
it is not the energy convergence,
With OT it can be considered as the maximin gradient.
Regards
Marcella
On Wednesday, May 14, 2025 at 4:36:20 PM UTC+2 rwmo... at gmail.com wrote:
> I believe you are invoking that "epsilon" is often used as a *symbol* for
> a tolerance in numerical work. That is not my question. I am asking for the
> mathematical definition being invoked here; it is a control parameter *for
> what*? I am guessing that this is a control parameter for the SCF energy?
>
> EPS_SCF appears in a different CP2K manual and has the definition "Target
> accuracy for the SCF convergence" This is a somewhat unclear definition,
> but given that it has a default threshold of 10**-5, I will assume it is
> stating that the largest matrix element difference between iterations must
> be no bigger than the threshold, since that is the usual criterion (as
> opposed to converged energy or converged eigenvalues).
>
> Is EPS_DEFAULT overriding EPS_SCF values to govern convergence? Assuming
> that "epsilon" here is a numerical tolerance for an SCF, what specific
> element of the SCF process is it managing?
>
> On Wednesday, May 14, 2025 at 8:51:12 AM UTC-4 Krack Matthias wrote:
>
>> “eps” is a short for “epsilon” and refers to a threshold value which
>> controls the numerical accuracy.
>>
>>
>>
>> *From: *cp... at googlegroups.com <cp... at googlegroups.com> on behalf of
>> Robert Molt <rwmo... at gmail.com>
>> *Date: *Wednesday, 14 May 2025 at 14:29
>> *To: *cp2k <cp... at googlegroups.com>
>> *Subject: *Re: [CP2K:21464] EPS_DEFAULT Meaning
>>
>> I already cited the manual in my question. I do not know what "eps" is.
>> The manual does not seem to say it, as far as I can see. I want to guess
>> that this is referring to the eigenvalues of the Fock matrix, but this is
>> not directly stated. Can this be defined relative to somehting in SCF
>> theory? I want to guess that this puts a tolerance on eigenvalues, but then
>> this would seemingly be redundant to having an EPS_SCF keyword? If this is
>> different than EPS_SCF, can the definition be given?
>>
>> On Wednesday, May 14, 2025 at 8:00:32 AM UTC-4 Krack Matthias wrote:
>>
>> Hi
>>
>>
>>
>> Open the CP2K input reference
>>
>>
>> https://manual.cp2k.org/trunk/CP2K_INPUT/FORCE_EVAL/DFT/QS.html#CP2K_INPUT.FORCE_EVAL.DFT.QS.EPS_DEFAULT
>>
>> in your browser and search/find all occurrences of “EPS_DEFAULT”.
>> EPS_DEFAULT defines the default values for all EPS_xxx keywords referring
>> to EPS_DEFAULT in their description.
>>
>>
>>
>> Alternatively, you can consult the source code:
>>
>> https://github.com/cp2k/cp2k/blob/master/src/cp_control_utils.F -
>> L837-L858
>> <https://github.com/cp2k/cp2k/blob/master/src/cp_control_utils.F#L837-L858>
>>
>>
>>
>> *From: *cp... at googlegroups.com <cp... at googlegroups.com> on behalf of
>> Robert Molt <rwmo... at gmail.com>
>> *Date: *Wednesday, 14 May 2025 at 13:32
>> *To: *cp2k <cp... at googlegroups.com>
>> *Subject: *[CP2K:21462] EPS_DEFAULT Meaning
>>
>> Good morning,
>>
>>
>>
>> Can someone clarify the EPS_DEFAULT keyword? The manual is unclear. It
>> states:
>>
>> "Try setting all EPS_xxx to values leading to an energy correct up to
>> EPS_DEFAULT"
>>
>> --
>> You received this message because you are subscribed to the Google Groups
>> "cp2k" group.
>> To unsubscribe from this group and stop receiving emails from it, send an
>> email to cp2k+uns... at googlegroups.com.
>> To view this discussion visit
>> https://groups.google.com/d/msgid/cp2k/b109503b-03b8-4545-b957-6ccdc1f6c329n%40googlegroups.com
>> <https://groups.google.com/d/msgid/cp2k/b109503b-03b8-4545-b957-6ccdc1f6c329n%40googlegroups.com?utm_medium=email&utm_source=footer>
>> .
>>
>> --
>> You received this message because you are subscribed to the Google Groups
>> "cp2k" group.
>> To unsubscribe from this group and stop receiving emails from it, send an
>> email to cp2k+uns... at googlegroups.com.
>>
>> To view this discussion visit
>> https://groups.google.com/d/msgid/cp2k/3f301911-b73c-4845-b422-b00126d81b2dn%40googlegroups.com
>> <https://groups.google.com/d/msgid/cp2k/3f301911-b73c-4845-b422-b00126d81b2dn%40googlegroups.com?utm_medium=email&utm_source=footer>
>> .
>>
>
--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+unsubscribe at googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/cp2k/737ba5a2-27e0-4530-8873-c6764d67fe1bn%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20250514/1b563239/attachment-0001.htm>
More information about the CP2K-user
mailing list