[CP2K:2030] Re: derivation method for second_derivative of XC functionals

Juerg Hutter hut... at pci.uzh.ch
Wed Apr 22 06:59:44 UTC 2009


Hi

sorry, I misinterpreted the error message. The problem was not the
2nd derivative of the XC functional but the 2nd derivative of
the electron density needed for the XC functional kernel.
This derivative is only available through plane wave derivatives
and not with spline derivatives.

The implemented TDDFT method is equivalent to Casida's method.
However, most implementations work in the MO basis, whereas
CP2K does the calculation in the AO basis.

regards

Juerg Hutter

----------------------------------------------------------
Juerg Hutter                   Phone : ++41 44 635 4491
Physical Chemistry Institute   FAX   : ++41 44 635 6838
University of Zurich           E-mail: hut... at pci.uzh.ch
Winterthurerstrasse 190
CH-8057 Zurich, Switzerland
----------------------------------------------------------


On Tue, 21 Apr 2009, Hanning Chen wrote:

> Juerg,
>  No, I did not make any changes besides the xc_derivative. I read the code,
> and it seems that only a few derivative methods are actually supported (the
> default one is unfortunately not).
>
>  Regarding the missing oscillator strength calculation, I may be able to
> fix it if the CP2K TDDFT module is based on Casida's SOS formula. Is it?
>
>  Thanks.
>
> Hanning
>
>
>
> On Tue, Apr 21, 2009 at 2:14 AM, Juerg Hutter <hut... at pci.uzh.ch> wrote:
>
>>
>>
>> Hi
>>
>> your first error message was related to a missing 2nd derivative
>> of a density functional. Your apparent fix however, is related
>> to the way the derivative of the density is computed.
>> I'm surprised that this had an impact. Maybe you did other changes?
>>
>> The TDDFT code has not been worked on for three years.
>> The performance is not very good and the oscillator strength
>> calculation is missing.
>>
>> regards
>>
>> Juerg Hutter
>>
>> ----------------------------------------------------------
>> Juerg Hutter                   Phone : ++41 44 635 4491
>> Physical Chemistry Institute   FAX   : ++41 44 635 6838
>> University of Zurich           E-mail: hut... at pci.uzh.ch
>> Winterthurerstrasse 190
>> CH-8057 Zurich, Switzerland
>> ----------------------------------------------------------
>>
>>
>> On Sat, 18 Apr 2009, Hanning Chen wrote:
>>
>>> Axel,
>>>
>>>  Yes. You are right. By putting
>>>
>>>
>>>    &XC_GRID
>>>          XC_DERIV SPLINE3_SMOOTH
>>>    &END XC_GRID
>>>
>>>  the error message went away. It seems that only a few derivative methods
>>> are supported by the XC functionals that I used.
>>>
>>>  Thanks.
>>>
>>> Hanning
>>>
>>>
>>>
>>> On Sat, Apr 18, 2009 at 3:49 PM, Axel <akoh... at gmail.com> wrote:
>>>
>>>>
>>>>
>>>>
>>>> On Apr 18, 2:22 pm, Hanning Chen <chenh... at gmail.com> wrote:
>>>>> Dear CP2K users,
>>>>>
>>>>>   I ran a TDDFPT calculation for excitation energies using PBE as my XC
>>>>> functionals. I encountered an error "Please provide a derivation method
>>>>> 12:50:04 ERRORL2 in pw_poisson_methods:pw_poisson_solve err=-300
>> 12:50:06
>>>>> ERRORL2 in xc:xc_calc_2nd_deriv err=-300  condition FAILED". I guess
>> that
>>>>> was caused by a missing parameter for the second derivative of the XC
>>>> terms.
>>>>> Please help me to resolve this issue.
>>>>
>>>> a little grepping through the sources reveals that you should have a
>>>> look at:
>>>>
>>>>
>> http://cp2k.berlios.de/manual/CP2K_INPUT/FORCE_EVAL/DFT/XC/XC_GRID.html#desc_XC_DERIV
>>>>
>>>> i don't much about TDDFT, but my guess is that there is no analytical
>>>> 2nd derivative
>>>> for the functional that you are using, so you'd have to use a
>>>> numerical method.
>>>>
>>>> have you compared your input to existing test cases?
>>>>
>>>> axel.
>>>>>
>>>>>  Thanks.
>>>>>
>>>>> Hanning Chen
>>>>>
>>>>> Department of Chemistry
>>>>> Northwestern University
>>>>> Evanston, IL 60208
>>>>>
>>>>
>>>
>>>>
>>>
>>
>>>
>>
>
> >
>



More information about the CP2K-user mailing list