Re: [cclib-devel] [cclib-users] orbital occupation numbers

[Karol L. <kar...@gm...>]

Hi Axel,

I took a closer look at the files. The formaldehyde logfile actually breaks
our parser, and I've created a ticket for that since it doesn't seem
trivial to figure out (https://github.com/cclib/cclib/issues/150).

The Psi3 files also broke the parser, but that was an easy fix. The other
files seems to parse OK and I'll use them for implementing nocoeffs and
nooccnos for the other programs, until I extend the unit tests we have.

Cheers,
Karol

On Fri, Nov 7, 2014 at 10:18 AM, Karol Langner <kar...@gm...>
wrote:

> Hi Axel,
>
> I see... I thought you might have attached the wrong file :)
>
> Anyway, I don't think it is a good idea to parse the occuptation of
> canonical MOs from RHF as nooccnos, because these are not calculated but
> simply defined as occupied or not. I would reserve the natural orbital
> attributes to things that are not equivalent to the mo* attributes. In any
> case, we don't parse the occuptaion of canonical MOs, simply because they
> are trivial and can be deduced from the HOMO.
>
> IMO, a better way to convey more information downstream is to parse some
> metadata about the calculation, namely if it's RHF/CASSCF/CI, whatever.
> There are also other reasons to do that, so this is a general thing that I
> think will be added at some point (although we have resisted it so far).
> Would it be useful for your purposes?
>
> Best,
> Karol
>
>
> On Fri, Nov 7, 2014 at 5:41 AM, Axel Schild <axe...@go...>
> wrote:
>
>> Hi Karol,
>>
>> in water_molpro.out and water_gaussian.inp.log we have a Hartree-Fock
>> calculation, for which the natural orbitals are the same as the canonical
>> orbitals (in the Molpro output they are just called "ELECTRON ORBITALS" but
>> in the Gaussian output they are given below "Natural Orbital Coefficients"
>> ).
>>
>> For a Hartree-Fock calculation they are trivial, but it would
>> nevertheless be useful to have nooccnos return these values so that we need
>> not to distinguish between different types of calculations.
>>
>> Cheers
>> Axel
>>
>> On Fri, Nov 7, 2014 at 6:08 AM, Karol Langner <kar...@gm...>
>> wrote:
>>
>>> Axel,
>>>
>>> Thanks! I will include some of these files as regressions tess. However,
>>> the Gaussian and Molpro output files for water do not contain natural
>>> orbitals... or am I missing something?
>>>
>>> Karol
>>>
>>> On Thu, Nov 6, 2014 at 11:57 PM, Karol Langner <kar...@gm...>
>>> wrote:
>>>
>>>> Just forwarding this to cclib-dev...
>>>>
>>>>
>>>> ---------- Forwarded message ----------
>>>> From: Axel Schild <axe...@go...>
>>>> Date: Thu, Nov 6, 2014 at 6:13 AM
>>>> Subject: Re: [cclib-users] orbital occupation numbers
>>>> To: "Karol M. Langner" <kar...@gm...>
>>>>
>>>>
>>>> Hi Karol,
>>>>
>>>> sorry for the delay, I first had to check with the other developers and
>>>> this took some time...
>>>>
>>>> Attached please find the files that you may use, i.e. that we release
>>>> to the Public Domain.
>>>>
>>>> Thanks!
>>>> Axel
>>>>
>>>> On Wed, Oct 29, 2014 at 3:50 PM, Karol M. Langner <
>>>> kar...@gm...> wrote:
>>>>
>>>>> Axel,
>>>>>
>>>>> Thanks for the files. Can I use them in our publicly available test
>>>>> suite? In other
>>>>> words, do you release these files into the Public Domain? Also, if you
>>>>> have the Molpro
>>>>> file with the long list of coefficients, that would be helpful, too.
>>>>>
>>>>> We are about to release cclib v1.3, and this certainly won't make it
>>>>> there. But I have
>>>>> already written some code which will go into the dev version:
>>>>> https://github.com/cclib/cclib/pull/148
>>>>>
>>>>> Thanks,
>>>>> Karol
>>>>>
>>>>> On Oct 27 2014, Axel Schild wrote:
>>>>> > Karol,
>>>>> >
>>>>> > I could give you more Gaussian or GamessUS files, but they don't give
>>>>> > anything new compared to the files in
>>>>> > https://github.com/cclib/cclib/tree/master/data
>>>>> >
>>>>> > Attached please find two Molpro logfiles and a psi3 logfile. We
>>>>> don't have
>>>>> > other file formats because we don't use the respective programs (but
>>>>> we
>>>>> > thought about using cclib to load these file formats for users that
>>>>> do use
>>>>> > them).
>>>>> >
>>>>> > Cheers
>>>>> > Axel
>>>>> >
>>>>> > PS: I removed the long list of CI coefficients of the Molpro output
>>>>> for
>>>>> > CH2O.
>>>>> >
>>>>> > On Fri, Oct 24, 2014 at 3:35 PM, Karol Langner <
>>>>> kar...@gm...>
>>>>> > wrote:
>>>>> >
>>>>> > > Axel,
>>>>> > >
>>>>> > > Would you be wlling to help by providing logfiles for programs
>>>>> that you
>>>>> > > would like parse this for?
>>>>> > >
>>>>> > > Karol
>>>>> > >
>>>>> > > On Fri, Oct 24, 2014 at 2:42 AM, Axel Schild <
>>>>> axe...@go...>
>>>>> > > wrote:
>>>>> > >
>>>>> > >> Hi Karol,
>>>>> > >>
>>>>> > >> thank you! I just compared to my test files and yes, those are the
>>>>> > >> numbers that we want.
>>>>> > >>
>>>>> > >> I will follow the progress on the development version to see when
>>>>> we can
>>>>> > >> use cclib for orbkit.
>>>>> > >>
>>>>> > >> Cheers
>>>>> > >> Axel
>>>>> > >>
>>>>> > >> On Thu, Oct 23, 2014 at 6:38 PM, Karol Langner <
>>>>> kar...@gm...>
>>>>> > >> wrote:
>>>>> > >>
>>>>> > >>> I created an issue for this:
>>>>> > >>> https://github.com/cclib/cclib/issues/143
>>>>> > >>>
>>>>> > >>>
>>>>> > >>> On Thu, Oct 23, 2014 at 12:31 PM, Karol Langner <
>>>>> kar...@gm...
>>>>> > >>> > wrote:
>>>>> > >>>
>>>>> > >>>> Axel,
>>>>> > >>>>
>>>>> > >>>> The terminology seems to be a little tricky. To be specific,
>>>>> the NO
>>>>> > >>>> coefficients that we currently parse (for Gaussian) are those
>>>>> that define
>>>>> > >>>> the natural orbital as an eigenvector of the diagonalized
>>>>> density matrix.
>>>>> > >>>>
>>>>> > >>>> Here is an example of the output from out unit tests:
>>>>> > >>>>
>>>>> > >>>>
>>>>> https://github.com/cclib/cclib/blob/master/data/Gaussian/basicGaussian09/water_cis.log#L578
>>>>> > >>>>
>>>>> > >>>> The eigenvalues printed there, I think, are the occupoation
>>>>> numbers you
>>>>> > >>>> need (the population in each orbital), although in that
>>>>> particular case
>>>>> > >>>> some are negative (maybe because the basis set is so small?).
>>>>> Anyway, we
>>>>> > >>>> don't parse them, but we could of course, if that will be
>>>>> useful for you.
>>>>> > >>>>
>>>>> > >>>> Cheers,
>>>>> > >>>> Karol
>>>>> > >>>>
>>>>> > >>>> P.S. A new release of cclib (1.3) is coming out soon. Which
>>>>> version are
>>>>> > >>>> you using? In any case, this will not make it into 1.3, so you
>>>>> will need to
>>>>> > >>>> use the dev version from github.
>>>>> > >>>>
>>>>> > >>>>
>>>>> > >>>> On Thu, Oct 23, 2014 at 12:07 PM, Axel Schild <
>>>>> > >>>> axe...@go...> wrote:
>>>>> > >>>>
>>>>> > >>>>> Hi Karol,
>>>>> > >>>>>
>>>>> > >>>>> I suppose that NO coefficients and orbital occupation numbers
>>>>> are the
>>>>> > >>>>> same, i.e. they should be the diagonals of the diagonalized
>>>>> density
>>>>> > >>>>> matrix.For single-determinant methods they are 2 (or 1) and for
>>>>> > >>>>> multi-determinant methods they have some value between 0 and 2
>>>>> (or 1).
>>>>> > >>>>>
>>>>> > >>>>> At least I cannot think of any other meaning of "NO
>>>>> coefficients", but
>>>>> > >>>>> I never heard that term so far.
>>>>> > >>>>>
>>>>> > >>>>> Thanks, I hope that this feature is available soon as it would
>>>>> be
>>>>> > >>>>> pretty nice to use cclib!
>>>>> > >>>>>
>>>>> > >>>>> Cheers
>>>>> > >>>>> Axel
>>>>> > >>>>>
>>>>> > >>>>> On Thu, Oct 23, 2014 at 2:18 PM, Karol Langner <
>>>>> > >>>>> kar...@gm...> wrote:
>>>>> > >>>>>
>>>>> > >>>>>> Hi Axel,
>>>>> > >>>>>>
>>>>> > >>>>>> The attribute nocoeffs parses NO coefficients, and only for
>>>>> Gaussian
>>>>> > >>>>>> at the moment, I believe. It could be easily extended to
>>>>> other programs,
>>>>> > >>>>>> assuming our test logfiles contain the data.
>>>>> > >>>>>>
>>>>> > >>>>>> If I understand correctly, occupation numbers can be
>>>>> calculated from
>>>>> > >>>>>> the NO coefficients. Is that correct?
>>>>> > >>>>>>
>>>>> > >>>>>> Cheers,
>>>>> > >>>>>> Karol
>>>>> > >>>>>>
>>>>> > >>>>>>
>>>>> > >>>>>> On Thu, Oct 23, 2014 at 3:40 AM, Axel Schild <
>>>>> > >>>>>> axe...@go...> wrote:
>>>>> > >>>>>>
>>>>> > >>>>>>> Hi,
>>>>> > >>>>>>>
>>>>> > >>>>>>> I'm one of the developers of orbkit (
>>>>> > >>>>>>> http://sourceforge.net/projects/orbkit). So far, we are
>>>>> using our
>>>>> > >>>>>>> own Molden, GAMESS US, Gaussian log-file/fchk-file, and
>>>>> WFN-file parser,
>>>>> > >>>>>>> but we thought about (also) using cclib to be able to read
>>>>> all the file
>>>>> > >>>>>>> formats that you support.
>>>>> > >>>>>>>
>>>>> > >>>>>>> However, we need orbital occupation numbers (like those of
>>>>> MCSCF
>>>>> > >>>>>>> orbitals or of natural orbitals).
>>>>> > >>>>>>>
>>>>> > >>>>>>> Is this the data that is stored in the variable <nocoeffs>
>>>>> in the
>>>>> > >>>>>>> development version? If so, will it be available for any
>>>>> other than the
>>>>> > >>>>>>> GAMESS and Gaussian parsers anytime soon? If not, is this
>>>>> feature planned
>>>>> > >>>>>>> to be included?
>>>>> > >>>>>>>
>>>>> > >>>>>>> Cheers
>>>>> > >>>>>>> Axel
>>>>> > >>>>>>>
>>>>> > >>>>>>>
>>>>> > >>>>>>>
>>>>> ------------------------------------------------------------------------------
>>>>> > >>>>>>>
>>>>> > >>>>>>> _______________________________________________
>>>>> > >>>>>>> cclib-users mailing list
>>>>> > >>>>>>> ccl...@li...
>>>>> > >>>>>>> https://lists.sourceforge.net/lists/listinfo/cclib-users
>>>>> > >>>>>>>
>>>>> > >>>>>>>
>>>>> > >>>>>>
>>>>> > >>>>>
>>>>> > >>>>
>>>>> > >>>
>>>>> > >>
>>>>> > >
>>>>>
>>>>> --
>>>>> written by Karol M. Langner
>>>>> Wed Oct 29 10:46:23 EDT 2014
>>>>>
>>>>
>>>>
>>>>
>>>
>>
>