[molpro-user] unrestricted DF-LUCCSD(T)-F12

Martin Diefenbach diefenbach at chemie.uni-frankfurt.de
Thu Feb 18 20:04:39 CET 2016


Dear Joachim,

Thanks again for the details! I will experiment with

df-luccsd(t)-f12,ansatz=3*A(Fix,NoX)

for both closed- and open-shell cases and wait for the publication of the 
reference and the documentation in the meantime.

Will the keyword for treatment of close pairs (how_treatclswk) also be 
available in the unrestricted df-luccsd(t) and df-luccsd(t)-f12 versions?

Best wishes,
Martin
-- 
Dr. Martin Diefenbach
Institut für Anorganische und Analytische Chemie
der Johann Wolfgang Goethe-Universität Frankfurt
Max-von-Laue-Straße 7, D-60438 Frankfurt am Main
Campus Riedberg, N160/207, Tel: +49-69-798-29445
E-Mail: diefenbach at chemie.uni-frankfurt.de

On Wed, 17 Feb 2016, Hans-Joachim Werner wrote:

> Dear Martin,
> I think what you write is incorrect. The following should be equivalent:
>
> df-lccsd(t)-f12,ansatz=3*A(Fix,NoX),projf=1
> df-luccsd(t)-f12,ansatz=3*A(Fix,NoX)
>
> The default in closed-shell df-lccsd(t)-f12 is projf=2 (see manual). This yields slightly different energies but should be more efficient.
> The option projf=2 is not implemented for open-shell. I would recommend to use df-luccsd(t)-f12,ansatz=3*A(Fix,NoX) for both open-
> and closed shell cases, if you need to compute energy differences between open- and closed-shell systems. Even though the above give the same
> energy for small cases, there may be small differences for large cases, when screening and weak pair approximations etc. come into play.
> Filipe Agapito is checking this currently in more detail.
>
> I am sorry that all this is still a bit confusing, but as I wrote earlier the df-luccsd(t)-f12 is still an undocumented feature
> and not published yet.
>
> Best wishes
> Joachim
>
> ---
> Prof. Dr. Hans-Joachim Werner
> Institute for Theoretical Chemistry
> University of Stuttgart
> Pfaffenwaldring 55
> D-70569 Stuttgart
> Tel: +49 711 / 685 64400
> Fax: +49 711 / 685 64442
> email: werner at theochem.uni-stuttgart.de
>
>
>> Am 12.02.2016 um 19:53 schrieb Martin Diefenbach <diefenbach at chemie.uni-frankfurt.de>:
>>
>> Dear Joachim and Grant,
>>
>> thank you for the details! Am I correct in assuming that the default ansatz for df-lccsd(t)-f12 "3*A(Loc,Fix)" is equivalent to "3*A(Fix,NoX),singles=0,projf=2", and that for technical reasons the former 3*A(Loc,Fix) is unavailable in the unrestricted version?
>>
>> Suppose I wanted to compute singlet-triplet energy splittings --- would it make sense to compare a closed-shell calculation using
>>
>> {df-hf}
>> {df-lccsd(t)-f12}
>>
>> with an open-shell one with
>>
>> {df-rhf}
>> {df-luccsd(t)-f12,3*A(Fix,NoX),singles=0,projf=2}
>>
>> Best regards,
>> Martin
>> --
>> Dr. Martin Diefenbach
>> Institut für Anorganische und Analytische Chemie
>> der Johann Wolfgang Goethe-Universität Frankfurt
>> Max-von-Laue-Straße 7, D-60438 Frankfurt am Main
>> Campus Riedberg, N160/207, Tel: +49-69-798-29445
>> E-Mail: diefenbach at chemie.uni-frankfurt.de
>>
>> On Thu, 11 Feb 2016, Hans-Joachim Werner wrote:
>>
>>> Dear Grant and Martin,
>>> to the best of my knowledge, this should work correctly, but it is not yet published. The reference
>>> is F. Agapito, Y. Liu and H.-J. Werner, to be published.
>>> Best wishes
>>> Joachim
>>> ---
>>> Prof. Dr. Hans-Joachim Werner
>>> Institute for Theoretical Chemistry
>>> University of Stuttgart
>>> Pfaffenwaldring 55
>>> D-70569 Stuttgart
>>> Tel: +49 711 / 685 64400
>>> Fax: +49 711 / 685 64442
>>> email: werner at theochem.uni-stuttgart.de
>>>
>>>
>>>> Am 11.02.2016 um 17:52 schrieb Grant Hill <grant.hill at sheffield.ac.uk>:
>>>>
>>>> Dear Martin,
>>>>
>>>> I’ve been able to run DF-LUCCSD(T)-F12 calculations using:
>>>>
>>>> explicit,ansatz=3*A(Fix,NoX)
>>>>
>>>> However, I have no idea how “safe” this choice of F12 Ansatz is.
>>>>
>>>> Best wishes,
>>>> Grant
>>>>
>>>>
>>>>> On 11 Feb 2016, at 13:50, Martin Diefenbach <diefenbach at chemie.uni-frankfurt.de> wrote:
>>>>>
>>>>> Dear all,
>>>>>
>>>>> is the unrestricted DF-LUCCSD(T)-F12 variant implemented in the new 2015 version of Molpro already? Using the current 2015.1.2 binary release, the "df-luccsd(t)-f12" keyword is parsed without complaint/without the "unknown directive" message.
>>>>>
>>>>> However, a simple open-shell calculation such as:
>>>>>
>>>>> ***,df-luccsd(t)-f12
>>>>> memory,500,m
>>>>> basis={default=vdz-f12}
>>>>> nosym;
>>>>> noorient;
>>>>> ang
>>>>> geometry={
>>>>> O
>>>>> O  1  r1
>>>>> }
>>>>> r1=1.21
>>>>>
>>>>> {df-rhf
>>>>> occ,   9
>>>>> closed,7
>>>>> wf,16,1,2
>>>>> }
>>>>> {df-luccsd(t)-f12}
>>>>> ---
>>>>>
>>>>> aborts with the following message right after convergence of the RHF-SCF deck
>>>>>
>>>>> Input parameters for LOCAL:
>>>>> LOCAL  =   -1.000000
>>>>> THRLOC =   -1.000000
>>>>> Local=1 not possible for ansatz 3*A(LOC,FIX)
>>>>> ? Error
>>>>> ? Illegal option
>>>>> ? The problem occurs in check_ansatz
>>>>> GLOBAL ERROR fehler on processor   0
>>>>>
>>>>> Can anyone reproduce this or point towards what I am missing here?
>>>>>
>>>>> Best regards,
>>>>> Martin
>>>>> --
>>>>> Dr. Martin Diefenbach
>>>>> Institut für Anorganische und Analytische Chemie
>>>>> der Johann Wolfgang Goethe-Universität Frankfurt
>>>>> Max-von-Laue-Straße 7, D-60438 Frankfurt am Main
>>>>> Campus Riedberg, N160/207, Tel: +49-69-798-29445
>>>>> E-Mail: diefenbach at chemie.uni-frankfurt.de_______________________________________________
>>>>> Molpro-user mailing list
>>>>> Molpro-user at molpro.net
>>>>> http://www.molpro.net/mailman/listinfo/molpro-user
>>>>
>>>>
>>>> _______________________________________________
>>>> Molpro-user mailing list
>>>> Molpro-user at molpro.net
>>>> http://www.molpro.net/mailman/listinfo/molpro-user
>>>
>
>


More information about the Molpro-user mailing list