From owner-chemistry@ccl.net Wed Mar 14 01:08:00 2012 From: "Nicholas Mayhall nmayhall(!)gmail.com" To: CCL Subject: CCL: CASSCF does not produce spin densities Message-Id: <-46487-120313234645-2869-MCMfOy9MFnnxximHOf+l7w.@.server.ccl.net> X-Original-From: Nicholas Mayhall Content-Type: multipart/alternative; boundary=f46d0442726c9c23ba04bb2bd2f1 Date: Tue, 13 Mar 2012 20:46:18 -0700 MIME-Version: 1.0 Sent to CCL by: Nicholas Mayhall [nmayhall+/-gmail.com] --f46d0442726c9c23ba04bb2bd2f1 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable "all excitations are NOT created equal!" : ) Seth, I agree that there is no "clean" manner in which one may classify a single component of a determinantal expansion as contributing to dynamical or static correlation energy. And for the FCI case, an attempt to categorize would be unnecessary. But the recognition that some determinants may be strongly correlated to the reference was the motivation of CASSCF-type methods, and still proves to be useful in developing approximate theories that closely approximate FCI without the factorial explosion. I'm happy to see this sacred cow still alive and kicking! Regards, Nick Mayhall, PhD nmayhall[-]berkeley.edu On Tue, Mar 13, 2012 at 4:26 PM, Seth Olsen seth.olsen-x-uq.edu.au < owner-chemistry[-]ccl.net> wrote: > > Sent to CCL by: Seth Olsen [seth.olsen=3D=3D=3Duq.edu.au] > > With all due respect, I think this conversation is getting foolish. The > concepts of "static" vs "dynamic" correlation do not have meaning in the > context of CASSF models. That is, they usually don't have well-defined > meaning in any case (I.e. everyone has their own) but any reasonable > meaning falls apart in a complete active space context - because the spac= e > is complete! In the asymptotic limit where all orbitals are included, it > becomes fill CI, and obviously treats both. I know of no sensible way to > cleanly distinguish the turnover where "static" correlation ends and wher= e > "dynamic" correlation begins, regardless of the length of the expansion. > > This field needs a revision of its core concepts, soon. Slaughter this > sacred cow before it rots on its own. > > Seth Olsen, Ph.D. > ARC Australian Research Fellow > School of Mathematics & Physics > The University of Queensland > Brisbane, QLD 4072 > Australia > +61 7 3365 2816 > > On 14/03/2012, at 4:17, "Neese, Frank frank.neese{:}mpi-mail.mpg.de" > wrote: > > > In response to Georg and Jim: > > > > > >> with all respect, but Frank Neese's statement is wrong. > > > > it looks like we are opening Pandora's box here :-) > > > > I nevertheless take the liberty to disagree with the statement brought > forward by Georg. (Not with the stated facts of course. I do understand > what CASSCF is and having programmed diverse multireference methods over > the years I do claim to have at least some appreciation of multireferenc= e > wavefunction theory; I know the Schlegl/Robb papers that were quoted by > Georg; they report a convergence enhancement algorithm for CASSCF wave > functions and an application to a reaction between two very small > molecules, i.e. H2CO --> H2 + CO. Very nice work! But IMHO these papers = do > not disprove anything I said). > > > > My arguments in favor of stating what I have stated in my previous post > are summarized below: > > > > As we all know CASSCF is limited to very few number of active orbitals. > The majority of reseachers who have worked intensely in the field of > multiconfigurational self consistent field theory agree that this type of > wavefunction is not meant to cover dynamic correlation which is what woul= d > be required to really go beyond Hartree-Fock level accuracy. Compared to, > say, a closed shell or ROHF CASSCF obviously does include a tiny amount o= f > dynamic correlation (unless your active space is trivially small and > reduces to the closed shell or ROHF determinant). However, that tiny litt= le > bit of dynamic correlation is essentially irrelevant. What CASSCF is > designed to do is to generalize the Hartree-Fock method to cases where t= he > single determinant Ansatz is qualitatively wrong. There are many papers b= y > eminent researchers like Peter Pulay or Jean Paul Malrieu who state that = it > would be an abuse of the CASSCF method to want to cover more than the > essential model space (which is the m! > inimal active space that is needed to cover the static correlation; that > again can mean many things. Sometimes it means ensuring that your > dissociation limit is correctly represented. Obviously here is where ones > chemical thinking comes in). I personally agree with this reasoning. It = is > then logical to state that CASSCF is not a high level method because its > accuracy will be similar to that of, say, RHF in the cases where a closed > shell determinant is reasonable. Thus, CASSCF is of broader applicability > (even much broader) and it is more consistent than R(O)HF, but it is not = of > strikingly higher accuracy - provided one compares apples with apples and > oranges with oranges. > > > > Of course if one defines anything that has more than one determinant as > a post Hartree-Fock method then, yes, there is no doubt that CASSCF is > post-HF. It is not, however, my favorite definition of what a post-HF > method is designed to be. I would tend to use this term if it is implied > that a substantial fraction of the (basis set) dynamic correlation energy > is recovered by the treatment (what is meant by "dynamic correlation" is, > again, a matter of the reference point. If you argue > from a > multireference perturbation theory or multireference configuration > interaction or multireference coupled cluster point of view then CASSCF i= s > the starting point and does not contain dynamic correlation). > > > > At the end of the day everybody is of course free to do whatever > calculation one wants and maybe one has good reasons for wanting to cover > some dynamic correlation already with a CASSCF wave function. A typical > example is the double d-shell effect for transition metal calculations in > conjunction with the CASSCF/CASPT2 method that in particular Kristin > Pierloot has elaborated on many times (in this case one covers a radial > correlation effect). My personal preference, however, is to do this with = a > post-CASSCF treatment (if one can afford it). > > > > We too have had many good results with CASSCF and transition metals. My > concerns were directed at the particular problem that the colleague was > mentioning. Having worked on both, multireference wavefunctions and iron > nitrosyl systems, I have simply expressed concerns to regard CASSCF as a > high level treatment that should be given preference over any other > theoretical method. To do so would, in my opinion, require solid backup > > from comparison of calculated to observed physical properties. > > > > The discussion that we have been starting here could very quickly turn > into a hair splitting dispute about terminology and I hope that this is n= ot > what will happen. I have simply tried here to explain what my take on th= e > subject is and I do, of course, respect any other approach or opinion tha= t > leads to creative and interesting science. > > > > Best regards > > Frank > > > > -----------------------------------------------------------------------= -- > > Prof. Dr. Frank Neese > > - Director - > > Max-Planck Institut f=FCr Bioanorganische Chemie > > Stiftstr. 34-36 > > D-45470 M=FClheim an der Ruhr > > Germany > > E-Mail: Frank.Neese]~[mpi-mail.mpg.de > > -----------------------------------------------------------------------= -- > > > > > > -=3D This is automatically added to each message by the mailing script = =3D-> > > --f46d0442726c9c23ba04bb2bd2f1 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable "all excitations are NOT created equal!" : )=A0

Seth, I ag= ree that there is no "clean" manner in which one may classify a s= ingle component of a determinantal expansion as contributing to dynamical o= r static correlation energy. And for the FCI case, an attempt to=A0categori= ze=A0would be unnecessary. But the recognition that some determinants may b= e strongly correlated to the reference was the motivation of CASSCF-type me= thods, and still proves to be useful in developing approximate theories tha= t closely approximate FCI without the factorial explosion.=A0

I'm happy to see this sacred cow still alive and kicking= !

Regards,

Nick Mayhall, = PhD
nmayhall[-]berkeley.ed= u


On Tue, Mar 13, 201= 2 at 4:26 PM, Seth Olsen seth.ols= en-x-uq.edu.au <owner-chemistry[-]ccl.net> wrote:

Sent to CCL by: Seth Olsen [seth.olsen=3D=3D=3Duq.edu.au]

With all due respect, I think this conversation is getting foolish. The con= cepts of "static" vs "dynamic" correlation do not have = meaning in the context of CASSF models. =A0That is, they usually don't = have well-defined meaning in any case (I.e. everyone has their own) but any= reasonable meaning falls apart in a complete active space context - becaus= e the space is complete! =A0In the asymptotic limit where all orbitals are = included, it becomes fill CI, and obviously treats both. I know of no sensi= ble way to cleanly distinguish the turnover where "static" correl= ation ends and where "dynamic" correlation begins, regardless of = the length of the expansion.

This field needs a revision of its core concepts, soon. =A0Slaughter this s= acred cow before it rots on its own.

Seth Olsen, Ph.D.
ARC Australian Research Fellow
School of Mathematics & Physics
The University of Queensland
Brisbane, QLD 4072
Australia
+61 7 3365 2= 816

On 14/03/2012, at 4:17, "Neese, Frank frank.neese{:}mpi-mail.mpg.de" <owner-chem= istry..ccl.net> wrote:<= br>
> In response to Georg and Jim:
>
>
>> with all respect, but Frank Neese's statement is wrong.
>
> it looks like we are opening Pandora's box here :-)
>
> I nevertheless take the liberty to disagree with the statement brought= forward by Georg. (Not with the stated facts of course. I do understand wh= at CASSCF is and having programmed diverse multireference methods over the = years I do claim to have at least some =A0appreciation of multireference wa= vefunction theory; I know the Schlegl/Robb papers that were quoted by Georg= ; they report a convergence enhancement algorithm for CASSCF wave functions= and an application to a reaction between two very small molecules, i.e. = =A0H2CO --> H2 + CO. Very nice work! But IMHO these papers do not dispro= ve anything I said).
>
> My arguments in favor of stating what I have stated in my previous pos= t are summarized below:
>
> As we all know CASSCF is limited to very few number of active or= bitals. The majority of reseachers who have worked intensely in the field o= f multiconfigurational self consistent field theory agree that this type of= wavefunction is not meant to cover dynamic correlation which is what would= be required to really go beyond Hartree-Fock level accuracy. Compared to, = say, a closed shell or ROHF CASSCF obviously does include a tiny amount of = dynamic correlation (unless your active space is trivially small and reduce= s to the closed shell or ROHF determinant). However, that tiny little bit o= f dynamic correlation is essentially irrelevant. What CASSCF is designed to= =A0do is to generalize the Hartree-Fock method to cases where the single d= eterminant Ansatz is qualitatively wrong. There are many papers by eminent = researchers like Peter Pulay or Jean Paul Malrieu who state that it would b= e an abuse of the CASSCF method to want to cover more than the essential mo= del space (which is the m!
=A0inimal active space that is needed to cover the= static correlation; that again can mean many things. Sometimes it means en= suring that your dissociation limit is correctly represented. Obviously her= e is where ones chemical thinking comes in). I personally =A0agree with thi= s reasoning. It is then logical to state that CASSCF is not a high level me= thod because its accuracy will be similar to that of, say, RHF in the cases= where a closed shell determinant is reasonable. Thus, CASSCF is of broader= applicability (even much broader) and it is more consistent than R(O)HF, b= ut it is not of strikingly higher accuracy - provided one compares apples w= ith apples and oranges with oranges.
>
> Of course if one defines anything that has more than one determinant a= s a post Hartree-Fock method then, yes, there is no doubt that CASSCF is po= st-HF. It is not, however, my favorite definition of what a post-HF method = is designed to be. I would tend to use this term if it is implied that a su= bstantial fraction of the (basis set) dynamic correlation energy is recover= ed by the treatment (what is meant by "dynamic correlation" is, a= gain, a matter of the reference point. If you argue > from a multirefere= nce perturbation theory or multireference configuration interaction or mult= ireference coupled cluster point of view then CASSCF is the starting point = and does not contain dynamic correlation).
>
> At the end of the day everybody is of course free to do whatever calcu= lation one wants and maybe one has good reasons for wanting to cover some d= ynamic correlation already with a CASSCF wave function. A typical example i= s the double d-shell effect for transition metal calculations in conjunctio= n with the CASSCF/CASPT2 method that in particular Kristin Pierloot has ela= borated on many times (in this case one covers a radial correlation effect)= . My personal preference, however, is to do this with a post-CASSCF treatme= nt (if one can afford it).
>
> We too have had many good results with CASSCF and transition metals. M= y concerns were directed at the particular problem that the colleague was m= entioning. Having worked on both, multireference wavefunctions and iron nit= rosyl systems, I have simply expressed concerns to regard CASSCF as a high = level treatment that should be given preference over any other theoretical = method. To do so would, in my opinion, require solid backup > from compa= rison of calculated to observed physical properties.
>
> The discussion that we have been starting here could very quickly turn= into a hair splitting dispute about terminology and I hope that this is no= t what will happen. I have simply tried here to explain what my take =A0on = the subject is and I do, of course, respect any other approach or opinion t= hat leads to creative and interesting science.
>
> Best regards
> Frank
>
> ----------------------------------------------------------------------= ---
> Prof. Dr. Frank Neese
> - Director -
> Max-Planck Institut f=FCr Bioanorganische Chemie
> Stiftstr. 34-36
> D-45470 M=FClheim an der Ruhr
> Germany
> E-Mail: Frank.Neese]~[mpi-mail.mpg.de
> ----------------------------------------------------------------------= ---
>



-=3D This is automatically ad= ded to each message by the mailing script =3D-
E-mail to subscribers: CHEMISTRY[-]ccl.n= et or use:
=A0 =A0 =A0http://www.ccl.net/cgi-bin/ccl/send_ccl_message

E-mail to administrators: CHEM= ISTRY-REQUEST[-]ccl.net or use
=A0 =A0 =A0http://www.ccl.net/cgi-bin/ccl/send_ccl_message

Subscribe/Unsubscribe:
=A0 =A0 =A0http://www.ccl.net/chemistry/sub_unsub.shtml

Before posting, check wait time at: http://www.ccl.net

Job: http://www.ccl.n= et/jobs
Conferences: http://server.ccl.net/chemistry/announcements/co= nferences/

Search Messages: http://www.ccl.net/chemistry/searchccl/index.shtml
=A0 =A0 =A0h= ttp://www.ccl.net/spammers.txt

RTFI: http://www.ccl.net/chemistry/aboutccl/instructions/



--f46d0442726c9c23ba04bb2bd2f1-- From owner-chemistry@ccl.net Wed Mar 14 04:31:01 2012 From: "Torsten Bruhn torsten.bruhn]|[mail.uni-oldenburg.de" To: CCL Subject: CCL:G: SpecDis 1.53 released Message-Id: <-46488-120314043018-25921-oL06UMTwNkbLmtknmmkdIA(_)server.ccl.net> X-Original-From: "Torsten Bruhn" Date: Wed, 14 Mar 2012 04:30:16 -0400 Sent to CCL by: "Torsten Bruhn" [torsten.bruhn::mail.uni-oldenburg.de] Hi all, we released a new version of SpecDis, a simple Windows tool to elucidate CD and UV calculations of g03, g09, Turbomole, and ORCA. SpecDis will calculate Gaussian or Lorentzian curves and one can directly compare the results with experimental CD or UV spectra. It is also possible to do a Boltzmann weighting or to apply an UV shift. In combination with gnuplot graphics can be plotted in publication quality. For details have a look at: http://www-organik.chemie.uni-wuerzburg.de/lehrstuehlearbeitskreise/bringmann/specdis/ If you have any questions or suggestions please send me an email, we are always looking for new ideas and will also add support for other software packages if there is the need for. Best regards, Torsten From owner-chemistry@ccl.net Wed Mar 14 12:25:00 2012 From: "Sebastian Kozuch kozuchs+/-yahoo.com" To: CCL Subject: CCL: CASSCF does not produce spin densities Message-Id: <-46489-120314034114-22349-0V6HxAoqYkc5lVQ12Cluyw()server.ccl.net> X-Original-From: Sebastian Kozuch Content-Type: multipart/alternative; boundary="-492277813-1910997932-1331710865=:92403" Date: Wed, 14 Mar 2012 00:41:05 -0700 (PDT) MIME-Version: 1.0 Sent to CCL by: Sebastian Kozuch [kozuchs{=}yahoo.com] ---492277813-1910997932-1331710865=:92403 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: quoted-printable If we are getting philosophical, there are lots of distinctions in our fiel= d that have a very blurred limit, but are still extremely useful (as long a= s you know that their limit is blurred). Forget about static and dynamic fo= r a second, and consider the full concept of "correlation". The most classi= cal (Lowdin's) definition is "all that Hartree-Fock lacks". However, the HF= exchange is a clear type of electronic correlation! And let's not start di= scussing about DFTcorrelation or the influence of the basis set. These defi= nitions are useful entities, not clear cut properties that can be represent= ed by a hermitian operator.=0A=0AIn the same way, traditional static and dy= namic correlation are distinguished by something like "all that CCSD(T) lac= ks" and "all that a minimal window CASSCF lacks", even though both methods = include a bit of the correlation that was "not intended" to be included. Of= course that if you have a full-CI, or a CCSDTQ567... or a CASSCF(inf,inf) = you will have everything, so static and dynamic correlation have a meaning = only in the "small" version of those methods.=0AI feel that the conceptual = definitions of those issues (i.e. needing more than one determinant and cor= relating the movement of electron pairs) came after it was comprehended wha= t each post-HF method can correct.=0A=0AAnyway, after we finish discussing = the correlation, we can start trying to define the formal charges, or the i= dea of atoms in a molecule, or if hybridization exists, or if Schrodinger's= cat is finally dead. Some concepts are to be used, not to be philosophized= .=0A=0A=A0=0ABest,=0ASebastian=0A=0A=0A=0A________________________________= =0A From: Nicholas Mayhall nmayhall(!)gmail.com = =0ATo: "Kozuch, Sebastian " =0ASent: Wednesday,= March 14, 2012 5:46 AM=0ASubject: CCL: CASSCF does not produce spin densit= ies=0A =0A=0A"all excitations are NOT created equal!" : )=A0=0A=0ASeth, I a= gree that there is no "clean" manner in which one may classify a single com= ponent of a determinantal expansion as contributing to dynamical or static = correlation energy. And for the FCI case, an attempt to=A0categorize=A0woul= d be unnecessary. But the recognition that some determinants may be strongl= y correlated to the reference was the motivation of CASSCF-type methods, an= d still proves to be useful in developing approximate theories that closely= approximate FCI without the factorial explosion.=A0=0A=0AI'm happy to see = this sacred cow still alive and kicking!=0A=0ARegards,=0A=0ANick Mayhall, P= hD=0Anmayhall(~)berkeley.edu=0A=0A=0A=0A=0AOn Tue, Mar 13, 2012 at 4:26 PM,= Seth Olsen seth.olsen-x-uq.edu.au wrote:=0A=0A= =0A>Sent to CCL by: Seth Olsen [seth.olsen=3D=3D=3Duq.edu.au]=0A>=0A>With a= ll due respect, I think this conversation is getting foolish. The concepts = of "static" vs "dynamic" correlation do not have meaning in the context of = CASSF models. =A0That is, they usually don't have well-defined meaning in a= ny case (I.e. everyone has their own) but any reasonable meaning falls apar= t in a complete active space context - because the space is complete! =A0In= the asymptotic limit where all orbitals are included, it becomes fill CI, = and obviously treats both. I know of no sensible way to cleanly distinguish= the turnover where "static" correlation ends and where "dynamic" correlati= on begins, regardless of the length of the expansion.=0A>=0A>This field nee= ds a revision of its core concepts, soon. =A0Slaughter this sacred cow befo= re it rots on its own.=0A>=0A>Seth Olsen, Ph.D.=0A>ARC Australian Research = Fellow=0A>School of Mathematics & Physics=0A>The University of Queensland= =0A>Brisbane, QLD 4072=0A>Australia=0A>+61 7 3365 2816=0A>=0A>=0A>On 14/03/= 2012, at 4:17, "Neese, Frank frank.neese{:}mpi-mail.mpg.de" wrote:=0A>=0A>> In response to Georg and Jim:=0A>>=0A>>=0A>>> w= ith all respect, but Frank Neese's statement is wrong.=0A>>=0A>> it looks l= ike we are opening Pandora's box here :-)=0A>>=0A>> I nevertheless take the= liberty to disagree with the statement brought forward by Georg. (Not with= the stated facts of course. I do understand what CASSCF is and having prog= rammed diverse multireference methods over the years I do claim to have at = least some =A0appreciation of multireference wavefunction theory; I know th= e Schlegl/Robb papers that were quoted by Georg; they report a convergence = enhancement algorithm for CASSCF wave functions and an application to a rea= ction between two very small molecules, i.e. =A0H2CO --> H2 + CO. Very nice= work! But IMHO these papers do not disprove anything I said).=0A>>=0A>> My= arguments in favor of stating what I have stated in my previous post are s= ummarized below:=0A>>=0A>> As we all know CASSCF is limited to very few num= ber of active orbitals. The majority of reseachers who have worked intensel= y in the field of multiconfigurational self consistent field theory agree t= hat this type of wavefunction is not meant to cover dynamic correlation whi= ch is what would be required to really go beyond Hartree-Fock level accurac= y. Compared to, say, a closed shell or ROHF CASSCF obviously does include a= tiny amount of dynamic correlation (unless your active space is trivially = small and reduces to the closed shell or ROHF determinant). However, that t= iny little bit of dynamic correlation is essentially irrelevant. What CASSC= F is designed to =A0do is to generalize the Hartree-Fock method to cases wh= ere the single determinant Ansatz is qualitatively wrong. There are many pa= pers by eminent researchers like Peter Pulay or Jean Paul Malrieu who state= that it would be an abuse of the CASSCF method to want to cover more than = the essential model space (which is the m!=0A>=0A>=A0inimal active space that is needed = to cover the static correlation; that again can mean many things. Sometimes= it means ensuring that your dissociation limit is correctly represented. O= bviously here is where ones chemical thinking comes in). I personally =A0ag= ree with this reasoning. It is then logical to state that CASSCF is not a h= igh level method because its accuracy will be similar to that of, say, RHF = in the cases where a closed shell determinant is reasonable. Thus, CASSCF i= s of broader applicability (even much broader) and it is more consistent th= an R(O)HF, but it is not of strikingly higher accuracy - provided one compa= res apples with apples and oranges with oranges.=0A>>=0A>> Of course if one= defines anything that has more than one determinant as a post Hartree-Fock= method then, yes, there is no doubt that CASSCF is post-HF. It is not, how= ever, my favorite definition of what a post-HF method is designed to be. I = would tend to use this term if it is implied that a substantial fraction of= the (basis set) dynamic correlation energy is recovered by the treatment (= what is meant by "dynamic correlation" is, again, a matter of the reference= point. If you argue > from a multireference perturbation theory or multire= ference configuration interaction or multireference coupled cluster point o= f view then CASSCF is the starting point and does not contain dynamic corre= lation).=0A>>=0A>> At the end of the day everybody is of course free to do = whatever calculation one wants and maybe one has good reasons for wanting t= o cover some dynamic correlation already with a CASSCF wave function. A typ= ical example is the double d-shell effect for transition metal calculations= in conjunction with the CASSCF/CASPT2 method that in particular Kristin Pi= erloot has elaborated on many times (in this case one covers a radial corre= lation effect). My personal preference, however, is to do this with a post-= CASSCF treatment (if one can afford it).=0A>>=0A>> We too have had many goo= d results with CASSCF and transition metals. My concerns were directed at t= he particular problem that the colleague was mentioning. Having worked on b= oth, multireference wavefunctions and iron nitrosyl systems, I have simply = expressed concerns to regard CASSCF as a high level treatment that should b= e given preference over any other theoretical method. To do so would, in my= opinion, require solid backup > from comparison of calculated to observed = physical properties.=0A>>=0A>> The discussion that we have been starting he= re could very quickly turn into a hair splitting dispute about terminology = and I hope that this is not what will happen. I have simply tried here to e= xplain what my take =A0on the subject is and I do, of course, respect any o= ther approach or opinion that leads to creative and interesting science.=0A= >>=0A>> Best regards=0A>> Frank=0A>>=0A>> ---------------------------------= ----------------------------------------=0A>> Prof. Dr. Frank Neese=0A>> - = Director -=0A>> Max-Planck Institut f=FCr Bioanorganische Chemie=0A>> Stift= str. 34-36=0A>> D-45470 M=FClheim an der Ruhr=0A>> Germany=0A>> E-Mail: Fra= nk.Neese]~[mpi-mail.mpg.de=0A>> -------------------------------------------= ------------------------------=0A>>=0A>=0A>=0A>=0A>=0A>-=3D This is automat= ically added to each message by the mailing script =3D-=0A>=0A>E-mail to su= bscribers: CHEMISTRY(~)ccl.net or use:=0A>=A0 =A0 =A0http://www.ccl.net/cgi= -bin/ccl/send_ccl_message=0A>=0A>E-mail to administrators: CHEMISTRY-REQUES= T(~)ccl.net or use=0A>=A0 =A0 =A0http://www.ccl.net/cgi-bin/ccl/send_ccl_me= ssage=0A>=0A>Subscribe/Unsubscribe:=0A>=A0 =A0 =A0http://www.ccl.net/chemis= try/sub_unsub.shtml=0A>=0A>Before posting, check wait time at: http://www.c= cl.net=0A>=0A>Job: http://www.ccl.net/jobs=0A>Conferences: http://server.cc= l.net/chemistry/announcements/conferences/=0A>=0A>Search Messages: http://w= ww.ccl.net/chemistry/searchccl/index.shtml=0A>=0A>=A0 =A0 =A0http://www.ccl= .net/spammers.txt=0A>=0A>RTFI: http://www.ccl.net/chemistry/aboutccl/instru= ctions/=0A>=0A>=0A> ---492277813-1910997932-1331710865=:92403 Content-Type: text/html; charset=iso-8859-1 Content-Transfer-Encoding: quoted-printable
If we are getting phi= losophical, there are lots of distinctions in our field that have a very bl= urred limit, but are still extremely useful (as long as you know that their= limit is blurred). Forget about static and dynamic for a second, and consi= der the full concept of "correlation". The most classical (Lowdin's) defini= tion is "all that Hartree-Fock lacks". However, the HF exchange is a clear = type of electronic correlation! And let's not start discussing about DFT correlation or the influence of the basis set. Thes= e definitions are useful entities, not clear cut properties that can be rep= resented by a hermitian operator.
In the same wa= y, traditional static and dynamic correlation are distinguished by somethin= g like "all that CCSD(T) lacks" and "all that a minimal window CASSCF lacks", even though both methods include a bit of the correlation that was= "not intended" to be included. Of course that if you have a full-CI, or a = CCSDTQ567... or a CASSCF(inf,inf) you will have everything, so static and d= ynamic correlation have a meaning only in the "small" version of those meth= ods.
I feel that the conceptual definitions of those= issues (i.e. needing more than one determinant and correlating the movemen= t of electron pairs) came after it was comprehended what each post-HF metho= d can correct.

Anyway, a= fter we finish discussing the correlation, we can start trying to define th= e formal charges, or the idea of atoms in a molecule, or if hybridization e= xists, or if Schrodinger's cat is finally dead. Some concepts are to be use= d, not to be philosophized.
 

Best,
S= ebastian

F= rom: Nicholas Mayhall nmayhall(!)gmail.com <owner-chemistry|a|c= cl.net>
To: "Kozuch= , Sebastian " <kozuchs|a|yahoo.com>
Sent: Wednesday, March 14, 2012 5:46 AM
<= span style=3D"font-weight: bold;">Subject: CCL: CASSCF does not = produce spin densities

=0A
"= all excitations are NOT created equal!" : ) 

Seth, I agree that= there is no "clean" manner in which one may classify a single component of= a determinantal expansion as contributing to dynamical or static correlati= on energy. And for the FCI case, an attempt to categorize would b= e unnecessary. But the recognition that some determinants may be strongly c= orrelated to the reference was the motivation of CASSCF-type methods, and s= till proves to be useful in developing approximate theories that closely ap= proximate FCI without the factorial explosion. 
=0A=0A
I'm happy to see this sacred cow still alive and kicking!

=
Regards,

Nick Mayhall, PhD
nmayhall(~)berkeley.edu
=0A=0A


On Tue, Mar 13, 2012 at 4:26 PM, Seth Olsen seth.olsen-x-uq.edu.au= <owner-chemistry(~)ccl.net> wrote:
=0A=0A
=0ASent to CCL by: Seth Olsen [seth.o= lsen=3D=3D=3Duq.edu.au]
=0A
=0AWith all due respect, I think this conversati= on is getting foolish. The concepts of "static" vs "dynamic" correlation do= not have meaning in the context of CASSF models.  That is, they usual= ly don't have well-defined meaning in any case (I.e. everyone has their own= ) but any reasonable meaning falls apart in a complete active space context= - because the space is complete!  In the asymptotic limit where all o= rbitals are included, it becomes fill CI, and obviously treats both. I know= of no sensible way to cleanly distinguish the turnover where "static" corr= elation ends and where "dynamic" correlation begins, regardless of the leng= th of the expansion.
=0A=0A=0A
=0AThis field needs a revision of its = core concepts, soon.  Slaughter this sacred cow before it rots on its = own.
=0A
=0ASeth Olsen, Ph.D.
=0AARC Australian Research Fellow=0ASchool of Mathematics & Physics
=0AThe University of Queensland<= br>=0ABrisbane, QLD 4072
=0AAustralia
=0A+61 7 33= 65 2816
=0A

=0AOn 14/03/2012, at 4:= 17, "Neese, Frank frank.neese{:}mpi-mail.mpg.de" <owner-chemistry..ccl.net> wr= ote:
=0A=0A=0A
=0A> In response to Georg and Jim:
=0A>
= =0A>
=0A>> with all respect, but Frank Neese's statement is wro= ng.
=0A>
=0A> it looks like we are opening Pandora's box here := -)
=0A>
=0A> I nevertheless take the liberty to disagree with t= he statement brought forward by Georg. (Not with the stated facts of course= . I do understand what CASSCF is and having programmed diverse multireferen= ce methods over the years I do claim to have at least some  appreciati= on of multireference wavefunction theory; I know the Schlegl/Robb papers th= at were quoted by Georg; they report a convergence enhancement algorithm fo= r CASSCF wave functions and an application to a reaction between two very s= mall molecules, i.e.  H2CO --> H2 + CO. Very nice work! But IMHO th= ese papers do not disprove anything I said).
=0A=0A=0A>
=0A> My= arguments in favor of stating what I have stated in my previous post are s= ummarized below:
=0A>
=0A
> As we all know CASSCF is limit= ed to very few number of active orbitals. The majority of reseachers who ha= ve worked intensely in the field of multiconfigurational self consistent fi= eld theory agree that this type of wavefunction is not meant to cover dynam= ic correlation which is what would be required to really go beyond Hartree-= Fock level accuracy. Compared to, say, a closed shell or ROHF CASSCF obviou= sly does include a tiny amount of dynamic correlation (unless your active s= pace is trivially small and reduces to the closed shell or ROHF determinant= ). However, that tiny little bit of dynamic correlation is essentially irre= levant. What CASSCF is designed to  do is to generalize the Hartree-Fo= ck method to cases where the single determinant Ansatz is qualitatively wro= ng. There are many papers by eminent researchers like Peter Pulay or Jean P= aul Malrieu who state that it would be an abuse of the CASSCF method to wan= t to cover more than the essential model space (which is the m!
=0A=0A=0A
 inimal active space that is needed to cover= the static correlation; that again can mean many things. Sometimes it mean= s ensuring that your dissociation limit is correctly represented. Obviously= here is where ones chemical thinking comes in). I personally  agree w= ith this reasoning. It is then logical to state that CASSCF is not a high l= evel method because its accuracy will be similar to that of, say, RHF in th= e cases where a closed shell determinant is reasonable. Thus, CASSCF is of = broader applicability (even much broader) and it is more consistent than R(= O)HF, but it is not of strikingly higher accuracy - provided one compares a= pples with apples and oranges with oranges.
=0A=0A=0A>
=0A> Of = course if one defines anything that has more than one determinant as a post= Hartree-Fock method then, yes, there is no doubt that CASSCF is post-HF. I= t is not, however, my favorite definition of what a post-HF method is desig= ned to be. I would tend to use this term if it is implied that a substantia= l fraction of the (basis set) dynamic correlation energy is recovered by th= e treatment (what is meant by "dynamic correlation" is, again, a matter of = the reference point. If you argue > from a multireference perturbation t= heory or multireference configuration interaction or multireference coupled= cluster point of view then CASSCF is the starting point and does not conta= in dynamic correlation).
=0A=0A=0A>
=0A> At the end of the day = everybody is of course free to do whatever calculation one wants and maybe = one has good reasons for wanting to cover some dynamic correlation already = with a CASSCF wave function. A typical example is the double d-shell effect= for transition metal calculations in conjunction with the CASSCF/CASPT2 me= thod that in particular Kristin Pierloot has elaborated on many times (in t= his case one covers a radial correlation effect). My personal preference, h= owever, is to do this with a post-CASSCF treatment (if one can afford it).<= br>=0A=0A=0A>
=0A> We too have had many good results with CASSCF a= nd transition metals. My concerns were directed at the particular problem t= hat the colleague was mentioning. Having worked on both, multireference wav= efunctions and iron nitrosyl systems, I have simply expressed concerns to r= egard CASSCF as a high level treatment that should be given preference over= any other theoretical method. To do so would, in my opinion, require solid= backup > from comparison of calculated to observed physical properties.=
=0A=0A=0A>
=0A> The discussion that we have been starting here= could very quickly turn into a hair splitting dispute about terminology an= d I hope that this is not what will happen. I have simply tried here to exp= lain what my take  on the subject is and I do, of course, respect any = other approach or opinion that leads to creative and interesting science.=0A=0A=0A>
=0A> Best regards
=0A> Frank
=0A>
=0A&= gt; -----------------------------------------------------------------------= --
=0A> Prof. Dr. Frank Neese
=0A> - Director -
=0A> Max-= Planck Institut f=FCr Bioanorganische Chemie
=0A> Stiftstr. 34-36
= =0A> D-45470 M=FClheim an der Ruhr
=0A> Germany
=0A> E-Mail:= Frank.Neese]~[mpi-mail.mpg.de
=0A> ----------------------------------= ---------------------------------------
=0A>
=0A
=0A
=0A
= =0A
-= =3D This is automatically added to each message by the mailing script =3D-<= br>=0A
=0AE-mail to subscribers: CHEMISTRY(~)ccl.net or use:
=0A      http://www.ccl.net/cgi-bin/ccl/send_ccl_message
=0A
=0AE= -mail to administrators: CHEMISTRY-REQUEST(~)ccl.net or use
=0A      = ;http://www.ccl.net/cgi-bin/ccl/send_ccl_message=0A
=0ASubscribe/Unsubscribe:
=0A      http://www.ccl.net/chemistry/sub_unsub.shtml
=0A
=0ABefore pos= ting, check wait time at: http://www.ccl.net
=0A
=0AJob: http://www.ccl.net/jo= bs
=0AConferences: http://server.ccl.= net/chemistry/announcements/conferences/
=0A
=0ASearch Messages: = http://www.ccl.net/chemistry/searchccl/index.shtml
=0A
=0A      http://www.ccl.net/spammers.txt
=0A
=0ARTFI: http://www.ccl.net/chemistry/a= boutccl/instructions/
=0A
=0A
=0A

=0A


---492277813-1910997932-1331710865=:92403-- From owner-chemistry@ccl.net Wed Mar 14 13:01:01 2012 From: "Georg Lefkidis lefkidis[]physik.uni-kl.de" To: CCL Subject: CCL: AW: CASSCF does not produce spin densities Message-Id: <-46490-120314044105-8008-klFS503TC2zzpfYVLbewHQ|server.ccl.net> X-Original-From: "Georg Lefkidis" Content-Language: de Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset="utf-8" Date: Wed, 14 Mar 2012 09:40:49 +0100 MIME-Version: 1.0 Sent to CCL by: "Georg Lefkidis" [lefkidis^^^physik.uni-kl.de] Dear all, perhaps I need to clarify that my answer's intention was not to advocate the merits of CASSCF. I merely wanted to point out that CASSCF is *not* at HF level. The discussion about static and/or dynamic correlations is a hot topic which can be both system- and property-specific. Even full CI is not equal full CI, since, mathematically speaking, full CI can be considered full only if the basis set spans the complete function-Hilbert space, etc. In my humble opinion, any multideterminant method which can reproduce (even at the lowest level), e.g., an open shell singlet (AF coupling) cannot be set equal to HF. This motivated my posting with respect to spin density. In theory an infinite window would make CASSCF perfect, but of course this is not what happens in real calculations. Still, I have been able in the past to reproduce to an excellent level experimental results (admittedly for very specific systems and properties). For other systems I found CASSCF to be inadequate. But then again there are systems where even EOMCC fails! I am thankful that not all issues are conclusively settled, so that I can still enjoy years of research! Best regards George -----Ursprüngliche Nachricht----- Von: owner-chemistry+lefkidis==physik.uni-kl.de- -ccl.net [mailto:owner-chemistry+lefkidis==physik.uni-kl.de- -ccl.net] Im Auftrag von Seth Olsen seth.olsen-x-uq.edu.au Gesendet: Mittwoch, 14. März 2012 00:26 An: Lefkidis, Georg Betreff: CCL: CASSCF does not produce spin densities Sent to CCL by: Seth Olsen [seth.olsen===uq.edu.au] With all due respect, I think this conversation is getting foolish. The concepts of "static" vs "dynamic" correlation do not have meaning in the context of CASSF models. That is, they usually don't have well-defined meaning in any case (I.e. everyone has their own) but any reasonable meaning falls apart in a complete active space context - because the space is complete! In the asymptotic limit where all orbitals are included, it becomes fill CI, and obviously treats both. I know of no sensible way to cleanly distinguish the turnover where "static" correlation ends and where "dynamic" correlation begins, regardless of the length of the expansion. This field needs a revision of its core concepts, soon. Slaughter this sacred cow before it rots on its own. Seth Olsen, Ph.D. ARC Australian Research Fellow School of Mathematics & Physics The University of Queensland Brisbane, QLD 4072 Australia +61 7 3365 2816 On 14/03/2012, at 4:17, "Neese, Frank frank.neese{:}mpi-mail.mpg.de" wrote: > In response to Georg and Jim: > > >> with all respect, but Frank Neese's statement is wrong. > > it looks like we are opening Pandora's box here :-) > > I nevertheless take the liberty to disagree with the statement brought forward by Georg. (Not with the stated facts of course. I do understand what CASSCF is and having programmed diverse multireference methods over the years I do claim to have at least some appreciation of multireference wavefunction theory; I know the Schlegl/Robb papers that were quoted by Georg; they report a convergence enhancement algorithm for CASSCF wave functions and an application to a reaction between two very small molecules, i.e. H2CO --> H2 + CO. Very nice work! But IMHO these papers do not disprove anything I said). > > My arguments in favor of stating what I have stated in my previous post are summarized below: > > As we all know CASSCF is limited to very few number of active orbitals. The majority of reseachers who have worked intensely in the field of multiconfigurational self consistent field theory agree that this type of wavefunction is not meant to cover dynamic correlation which is what would be required to really go beyond Hartree-Fock level accuracy. Compared to, say, a closed shell or ROHF CASSCF obviously does include a tiny amount of dynamic correlation (unless your active space is trivially small and reduces to the closed shell or ROHF determinant). However, that tiny little bit of dynamic correlation is essentially irrelevant. What CASSCF is designed to do is to generalize the Hartree-Fock method to cases where the single determinant Ansatz is qualitatively wrong. There are many papers by eminent researchers like Peter Pulay or Jean Paul Malrieu who state that it would be an abuse of the CASSCF method to want to cover more than the essential model space (which is the m! inimal active space that is needed to cover the static correlation; that again can mean many things. Sometimes it means ensuring that your dissociation limit is correctly represented. Obviously here is where ones chemical thinking comes in). I personally agree with this reasoning. It is then logical to state that CASSCF is not a high level method because its accuracy will be similar to that of, say, RHF in the cases where a closed shell determinant is reasonable. Thus, CASSCF is of broader applicability (even much broader) and it is more consistent than R(O)HF, but it is not of strikingly higher accuracy - provided one compares apples with apples and oranges with oranges. > > Of course if one defines anything that has more than one determinant as a post Hartree-Fock method then, yes, there is no doubt that CASSCF is post-HF. It is not, however, my favorite definition of what a post-HF method is designed to be. I would tend to use this term if it is implied that a substantial fraction of the (basis set) dynamic correlation energy is recovered by the treatment (what is meant by "dynamic correlation" is, again, a matter of the reference point. If you argue > from a multireference perturbation theory or multireference configuration interaction or multireference coupled cluster point of view then CASSCF is the starting point and does not contain dynamic correlation). > > At the end of the day everybody is of course free to do whatever calculation one wants and maybe one has good reasons for wanting to cover some dynamic correlation already with a CASSCF wave function. A typical example is the double d-shell effect for transition metal calculations in conjunction with the CASSCF/CASPT2 method that in particular Kristin Pierloot has elaborated on many times (in this case one covers a radial correlation effect). My personal preference, however, is to do this with a post-CASSCF treatment (if one can afford it). > > We too have had many good results with CASSCF and transition metals. My concerns were directed at the particular problem that the colleague was mentioning. Having worked on both, multireference wavefunctions and iron nitrosyl systems, I have simply expressed concerns to regard CASSCF as a high level treatment that should be given preference over any other theoretical method. To do so would, in my opinion, require solid backup > from comparison of calculated to observed physical properties. > > The discussion that we have been starting here could very quickly turn into a hair splitting dispute about terminology and I hope that this is not what will happen. I have simply tried here to explain what my take on the subject is and I do, of course, respect any other approach or opinion that leads to creative and interesting science. > > Best regards > Frank > > ---------------------------------------------------------------------- > --- > Prof. Dr. Frank Neese > - Director - > Max-Planck Institut f r Bioanorganische Chemie Stiftstr. 34-36 > D-45470 M lheim an der Ruhr > Germany > E-Mail: Frank.Neese]~[mpi-mail.mpg.de > ---------------------------------------------------------------------- > ---http://www.ccl.net/cgi-bin/ccl/send_ccl_messagehttp://www.ccl.net/chemistry/sub_unsub.shtmlhttp://www.ccl.net/spammers.txt From owner-chemistry@ccl.net Wed Mar 14 13:36:01 2012 From: "Anja Jatsch pipek..gmx.net" To: CCL Subject: CCL:G: PBC calculations Message-Id: <-46491-120314050642-19056-Mg8pfjGiQCzjNpv+Rkdt/g^^^server.ccl.net> X-Original-From: "Anja Jatsch" Date: Wed, 14 Mar 2012 05:06:41 -0400 Sent to CCL by: "Anja Jatsch" [pipek:+:gmx.net] Dear ccl users, I am trying to set up a PBC calculation on a measured crystal structure (organic), but I always end up with the same error in the MO guess: Dimensions of /B2/ exceeded in AtP2B2. Error termination via Lnk1e in c:\program files\G09W\l401.exe I am using #p pbepbe/3-21g/auto and I really dont have a clue whats the problem. There are almost 200 atoms in the unit cell, maybe its to large? But Gaussian doesnt complain about the memory Thanks a lot for any idears! Anja From owner-chemistry@ccl.net Wed Mar 14 14:11:00 2012 From: "=?iso-8859-1?Q?J=FCrgen_Gr=E4fenstein?= jurgen:+:chem.gu.se" To: CCL Subject: CCL: CASSCF does not produce spin densities Message-Id: <-46492-120314060859-15983-bNVeN6i+VjlersusO8QZuA=-=server.ccl.net> X-Original-From: =?iso-8859-1?Q?J=FCrgen_Gr=E4fenstein?= Content-Language: en-US Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset="iso-8859-1" Date: Wed, 14 Mar 2012 10:08:44 +0000 MIME-Version: 1.0 Sent to CCL by: =?iso-8859-1?Q?J=FCrgen_Gr=E4fenstein?= [jurgen]_[chem.gu.se] +++ Seth Olsen wrote: The concepts of "static" vs "dynamic" correlation do not have meaning in the context of CASSF models. That is, they usually don't have well-defined meaning in any case (I.e. everyone has their own) +++ You are right, there is no exact and generally accepted distinction between th two types of correlations. Notwithstanding, it is obvious that there are two essentially different correlation patterns, which require essentially different ways of treatment in quantum chemistry. but any reasonable meaning falls apart in a complete active space context - because the space is complete! +++ This is plainly incorrect. "Complete active space" means that for the orbitals chosen to be active, all possible configurations are incorporates (as opposed to RASSCF methods and the like). It does not mean that all orbitals are considered active. +++ The art with CASSCF is just to select just those orbitals as active that account for the non-dynamic correlations (and to do this in a consistent fashion for reactants, intermediates/TS's, and products). In the asymptotic limit where all orbitals are included, it becomes fill CI, and obviously treats both. +++ This is uncontroversial but of little practical relevance. Thus, we have to continue bothering about static correlation. I know of no sensible way to cleanly distinguish the turnover where "static" correlation ends and where "dynamic" correlation begins, regardless of the length of the expansion. +++ If one finds an explicit (physically motivated) separation between non-dynamic and dynamic correlation, probably any CASSCF wavefunction will contain both of them. With increasing size of the active space, the portion of dynamical correlation will increase. There will be no CASSCF WF that contains all of the static and none of the dynamic correlation. See e.g. Gräfenstein/Cremer, PCCP 2, 2091 (2000); Mol. Phys. 103, 279 (2005) for a discussion. This field needs a revision of its core concepts, soon. +++ A strict definition of non-dynamic vs dynamic correlation would be extremely valuable; I agree with this. Slaughter this sacred cow before it rots on its own. +++ It won't. If you doubt this, just try to describe dissociating H_2 with MP2/MP3/MP4. Best regards, Jürgen Gräfenstein University of Gothenburg, Department of Chemistry SE-412 96 GÖTEBORG, SWEDEN Phone: +46-31-786 9016 Fax: +46-31-772 3840 e-mail: Jurgen.Grafenstein^chem.gu.se From owner-chemistry@ccl.net Wed Mar 14 14:46:00 2012 From: "Seth Olsen seth.olsen!^!uq.edu.au" To: CCL Subject: CCL: CASSCF does not produce spin densities Message-Id: <-46493-120314084901-21640-13oPfKaJfIeO/iTHOycaNQ(_)server.ccl.net> X-Original-From: Seth Olsen Content-Type: multipart/alternative; boundary=Apple-Mail-2--912677580 Date: Wed, 14 Mar 2012 22:48:44 +1000 Mime-Version: 1.0 (Apple Message framework v1084) Sent to CCL by: Seth Olsen [seth.olsen##uq.edu.au] --Apple-Mail-2--912677580 Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset=iso-8859-1 It's not that there isn't a sensible way to categorize components in any = particular model that bothers me... its that the physical concepts = themselves are not defined. At best, they are usually defined relative = to some-method-or-other (for example, statements like "static = correlation is the correlation represented in a full valence CASSCF = technique"). It's not satisfactory to me to have no notion of a = physical effect that isn't tied implicitly to some specific = computational model. =20 It's also worth mentioning that I've only ever dealt with state-averaged = problems. It is possible that distinguishing between "static" and = "dynamic" correlations may be operationally useful for ground state = problems, where the idea of a well-defined reference makes somewhat more = sense (although still not very much, because of the orbital invariance = of the CAS expansion). =20 It's pretty clear that the consensus is that "static" correlations are = correlations required by the constraint that the state transform as a = particular irrep of the symmetric group, and that that "dynamic" = correlations are associated with the Coulomb hole. I get all that. = But, this tells you right away that it is not possible to build any = operator whose expectation will give you a measure of pure "static" or = "dynamic" correlations. This is because the requirement of = transformation as an irrep of S_n will entangle all degrees of freedom, = while the Coulomb operator generates pairwise entanglements. The = operators act on different Hilbert spaces. -Seth On 14/03/2012, at 1:46 PM, Nicholas Mayhall nmayhall(!)gmail.com wrote: > "all excitations are NOT created equal!" : )=20 >=20 > Seth, I agree that there is no "clean" manner in which one may = classify a single component of a determinantal expansion as contributing = to dynamical or static correlation energy. And for the FCI case, an = attempt to categorize would be unnecessary. But the recognition that = some determinants may be strongly correlated to the reference was the = motivation of CASSCF-type methods, and still proves to be useful in = developing approximate theories that closely approximate FCI without the = factorial explosion.=20 >=20 > I'm happy to see this sacred cow still alive and kicking! >=20 > Regards, >=20 > Nick Mayhall, PhD > nmayhall(~)berkeley.edu >=20 >=20 > On Tue, Mar 13, 2012 at 4:26 PM, Seth Olsen seth.olsen-x-uq.edu.au = wrote: >=20 > Sent to CCL by: Seth Olsen [seth.olsen=3D=3D=3Duq.edu.au] >=20 > With all due respect, I think this conversation is getting foolish. = The concepts of "static" vs "dynamic" correlation do not have meaning in = the context of CASSF models. That is, they usually don't have = well-defined meaning in any case (I.e. everyone has their own) but any = reasonable meaning falls apart in a complete active space context - = because the space is complete! In the asymptotic limit where all = orbitals are included, it becomes fill CI, and obviously treats both. I = know of no sensible way to cleanly distinguish the turnover where = "static" correlation ends and where "dynamic" correlation begins, = regardless of the length of the expansion. >=20 > This field needs a revision of its core concepts, soon. Slaughter = this sacred cow before it rots on its own. >=20 > Seth Olsen, Ph.D. > ARC Australian Research Fellow > School of Mathematics & Physics > The University of Queensland > Brisbane, QLD 4072 > Australia > +61 7 3365 2816 >=20 > On 14/03/2012, at 4:17, "Neese, Frank frank.neese{:}mpi-mail.mpg.de" = wrote: >=20 > > In response to Georg and Jim: > > > > > >> with all respect, but Frank Neese's statement is wrong. > > > > it looks like we are opening Pandora's box here :-) > > > > I nevertheless take the liberty to disagree with the statement = brought forward by Georg. (Not with the stated facts of course. I do = understand what CASSCF is and having programmed diverse multireference = methods over the years I do claim to have at least some appreciation of = multireference wavefunction theory; I know the Schlegl/Robb papers that = were quoted by Georg; they report a convergence enhancement algorithm = for CASSCF wave functions and an application to a reaction between two = very small molecules, i.e. H2CO --> H2 + CO. Very nice work! But IMHO = these papers do not disprove anything I said). > > > > My arguments in favor of stating what I have stated in my previous = post are summarized below: > > > > As we all know CASSCF is limited to very few number of active = orbitals. The majority of reseachers who have worked intensely in the = field of multiconfigurational self consistent field theory agree that = this type of wavefunction is not meant to cover dynamic correlation = which is what would be required to really go beyond Hartree-Fock level = accuracy. Compared to, say, a closed shell or ROHF CASSCF obviously does = include a tiny amount of dynamic correlation (unless your active space = is trivially small and reduces to the closed shell or ROHF determinant). = However, that tiny little bit of dynamic correlation is essentially = irrelevant. What CASSCF is designed to do is to generalize the = Hartree-Fock method to cases where the single determinant Ansatz is = qualitatively wrong. There are many papers by eminent researchers like = Peter Pulay or Jean Paul Malrieu who state that it would be an abuse of = the CASSCF method to want to cover more than the essential model space = (which is the m! > inimal active space that is needed to cover the static correlation; = that again can mean many things. Sometimes it means ensuring that your = dissociation limit is correctly represented. Obviously here is where = ones chemical thinking comes in). I personally agree with this = reasoning. It is then logical to state that CASSCF is not a high level = method because its accuracy will be similar to that of, say, RHF in the = cases where a closed shell determinant is reasonable. Thus, CASSCF is of = broader applicability (even much broader) and it is more consistent than = R(O)HF, but it is not of strikingly higher accuracy - provided one = compares apples with apples and oranges with oranges. > > > > Of course if one defines anything that has more than one determinant = as a post Hartree-Fock method then, yes, there is no doubt that CASSCF = is post-HF. It is not, however, my favorite definition of what a post-HF = method is designed to be. I would tend to use this term if it is implied = that a substantial fraction of the (basis set) dynamic correlation = energy is recovered by the treatment (what is meant by "dynamic = correlation" is, again, a matter of the reference point. If you argue > = > from a multireference perturbation theory or multireference = configuration interaction or multireference coupled cluster point of = view then CASSCF is the starting point and does not contain dynamic = correlation). > > > > At the end of the day everybody is of course free to do whatever = calculation one wants and maybe one has good reasons for wanting to = cover some dynamic correlation already with a CASSCF wave function. A = typical example is the double d-shell effect for transition metal = calculations in conjunction with the CASSCF/CASPT2 method that in = particular Kristin Pierloot has elaborated on many times (in this case = one covers a radial correlation effect). My personal preference, = however, is to do this with a post-CASSCF treatment (if one can afford = it). > > > > We too have had many good results with CASSCF and transition metals. = My concerns were directed at the particular problem that the colleague = was mentioning. Having worked on both, multireference wavefunctions and = iron nitrosyl systems, I have simply expressed concerns to regard CASSCF = as a high level treatment that should be given preference over any other = theoretical method. To do so would, in my opinion, require solid backup = > from comparison of calculated to observed physical properties. > > > > The discussion that we have been starting here could very quickly = turn into a hair splitting dispute about terminology and I hope that = this is not what will happen. I have simply tried here to explain what = my take on the subject is and I do, of course, respect any other = approach or opinion that leads to creative and interesting science. > > > > Best regards > > Frank > > > > = ------------------------------------------------------------------------- > > Prof. Dr. Frank Neese > > - Director - > > Max-Planck Institut f=FCr Bioanorganische Chemie > > Stiftstr. 34-36 > > D-45470 M=FClheim an der Ruhr > > Germany > > E-Mail: Frank.Neese]~[mpi-mail.mpg.de > > = ------------------------------------------------------------------------- > > >=20 >=20 >=20 > -=3D This is automatically added to each message by the mailing script = =3D- >=20 >=20 > E-mail to subscribers: CHEMISTRY(~)ccl.net or use:>=20 > E-mail to administrators: CHEMISTRY-REQUEST(~)ccl.net or use>=20>=20>=20> Conferences: = http://server.ccl.net/chemistry/announcements/conferences/ >=20>=20 >=20>=20>=20 >=20 >=20 --------------------------------------------------- Seth Olsen ARC Australian Research Fellow 6-431 Physics Annexe School of Mathematics and Physics The University of Queensland Brisbane QLD 4072 Australia seth.olsen,+,uq.edu.au +61 7 3365 2816 --------------------------------------------------- Unless stated otherwise, this e-mail represents only the views of the = Sender and not the views of The University of Queensland --Apple-Mail-2--912677580 Content-Transfer-Encoding: quoted-printable Content-Type: text/html; charset=iso-8859-1 It's = not that there isn't a sensible way to categorize components in any = particular model that bothers me... its that the physical concepts = themselves are not defined.  At best, they are usually defined = relative to some-method-or-other (for example, statements like "static = correlation is the correlation represented in a full valence CASSCF = technique").  It's not satisfactory to me to have no notion of a = physical effect that isn't tied implicitly to some specific = computational model.  

It's also worth = mentioning that I've only ever dealt with state-averaged problems. =  It is possible that distinguishing between "static" and "dynamic" = correlations may be operationally useful for ground state problems, = where the idea of a well-defined reference makes somewhat more sense = (although still not very much, because of the orbital invariance of the = CAS expansion).  

It's pretty clear that = the consensus is that "static" correlations are correlations required by = the constraint that the state transform as a particular irrep of the = symmetric group, and that that "dynamic" correlations are associated = with the Coulomb hole.  I get all that.  But, this tells you = right away that it is not possible to build any operator whose = expectation will give you a measure of pure "static" or "dynamic" = correlations.  This is because the requirement of transformation as = an irrep of S_n will entangle all degrees of freedom, while the Coulomb = operator generates pairwise entanglements.  The operators act on = different Hilbert spaces. -Seth


On = 14/03/2012, at 1:46 PM, Nicholas Mayhall nmayhall(!)gmail.com wrote:

"all = excitations are NOT created equal!" : ) 

Seth, I agree that = there is no "clean" manner in which one may classify a single component = of a determinantal expansion as contributing to dynamical or static = correlation energy. And for the FCI case, an attempt = to categorize would be unnecessary. But the recognition that = some determinants may be strongly correlated to the reference was the = motivation of CASSCF-type methods, and still proves to be useful in = developing approximate theories that closely approximate FCI without the = factorial explosion. 

I'm happy to see this sacred cow still alive and = kicking!

Regards,

Nick = Mayhall, PhD


On Tue, Mar 13, = 2012 at 4:26 PM, Seth Olsen seth.olsen-x-uq.edu.au <owner-chemistry(~)ccl.net>= ; wrote:

Sent to CCL by: Seth Olsen [seth.olsen=3D=3D=3Duq.edu.au]

With all due respect, I think this conversation is getting foolish. The = concepts of "static" vs "dynamic" correlation do not have meaning in the = context of CASSF models.  That is, they usually don't have = well-defined meaning in any case (I.e. everyone has their own) but any = reasonable meaning falls apart in a complete active space context - = because the space is complete!  In the asymptotic limit where all = orbitals are included, it becomes fill CI, and obviously treats both. I = know of no sensible way to cleanly distinguish the turnover where = "static" correlation ends and where "dynamic" correlation begins, = regardless of the length of the expansion.

This field needs a revision of its core concepts, soon.  Slaughter = this sacred cow before it rots on its own.

Seth Olsen, Ph.D.
ARC Australian Research Fellow
School of Mathematics & Physics
The University of Queensland
Brisbane, QLD 4072
Australia
+61 7 = 3365 2816

On 14/03/2012, at 4:17, "Neese, Frank frank.neese{:}mpi-mail.mpg.de" = <owner-chemistry..ccl.net> wrote:

> In response to Georg and Jim:
>
>
>> with all respect, but Frank Neese's statement is wrong.
>
> it looks like we are opening Pandora's box here :-)
>
> I nevertheless take the liberty to disagree with the statement = brought forward by Georg. (Not with the stated facts of course. I do = understand what CASSCF is and having programmed diverse multireference = methods over the years I do claim to have at least some =  appreciation of multireference wavefunction theory; I know the = Schlegl/Robb papers that were quoted by Georg; they report a convergence = enhancement algorithm for CASSCF wave functions and an application to a = reaction between two very small molecules, i.e.  H2CO --> H2 + = CO. Very nice work! But IMHO these papers do not disprove anything I = said).
>
> My arguments in favor of stating what I have stated in my previous = post are summarized below:
>
> As we all know CASSCF is limited to very few number of active = orbitals. The majority of reseachers who have worked intensely in the = field of multiconfigurational self consistent field theory agree that = this type of wavefunction is not meant to cover dynamic correlation = which is what would be required to really go beyond Hartree-Fock level = accuracy. Compared to, say, a closed shell or ROHF CASSCF obviously does = include a tiny amount of dynamic correlation (unless your active space = is trivially small and reduces to the closed shell or ROHF determinant). = However, that tiny little bit of dynamic correlation is essentially = irrelevant. What CASSCF is designed to  do is to generalize the = Hartree-Fock method to cases where the single determinant Ansatz is = qualitatively wrong. There are many papers by eminent researchers like = Peter Pulay or Jean Paul Malrieu who state that it would be an abuse of = the CASSCF method to want to cover more than the essential model space = (which is the m!
 inimal active space that is needed to = cover the static correlation; that again can mean many things. Sometimes = it means ensuring that your dissociation limit is correctly represented. = Obviously here is where ones chemical thinking comes in). I personally =  agree with this reasoning. It is then logical to state that CASSCF = is not a high level method because its accuracy will be similar to that = of, say, RHF in the cases where a closed shell determinant is = reasonable. Thus, CASSCF is of broader applicability (even much broader) = and it is more consistent than R(O)HF, but it is not of strikingly = higher accuracy - provided one compares apples with apples and oranges = with oranges.
>
> Of course if one defines anything that has more than one = determinant as a post Hartree-Fock method then, yes, there is no doubt = that CASSCF is post-HF. It is not, however, my favorite definition of = what a post-HF method is designed to be. I would tend to use this term = if it is implied that a substantial fraction of the (basis set) dynamic = correlation energy is recovered by the treatment (what is meant by = "dynamic correlation" is, again, a matter of the reference point. If you = argue > from a multireference perturbation theory or multireference = configuration interaction or multireference coupled cluster point of = view then CASSCF is the starting point and does not contain dynamic = correlation).
>
> At the end of the day everybody is of course free to do whatever = calculation one wants and maybe one has good reasons for wanting to = cover some dynamic correlation already with a CASSCF wave function. A = typical example is the double d-shell effect for transition metal = calculations in conjunction with the CASSCF/CASPT2 method that in = particular Kristin Pierloot has elaborated on many times (in this case = one covers a radial correlation effect). My personal preference, = however, is to do this with a post-CASSCF treatment (if one can afford = it).
>
> We too have had many good results with CASSCF and transition = metals. My concerns were directed at the particular problem that the = colleague was mentioning. Having worked on both, multireference = wavefunctions and iron nitrosyl systems, I have simply expressed = concerns to regard CASSCF as a high level treatment that should be given = preference over any other theoretical method. To do so would, in my = opinion, require solid backup > from comparison of calculated to = observed physical properties.
>
> The discussion that we have been starting here could very quickly = turn into a hair splitting dispute about terminology and I hope that = this is not what will happen. I have simply tried here to explain what = my take  on the subject is and I do, of course, respect any other = approach or opinion that leads to creative and interesting science.
>
> Best regards
> Frank
>
> = -------------------------------------------------------------------------<= br> > Prof. Dr. Frank Neese
> - Director -
> Max-Planck Institut f=FCr Bioanorganische Chemie
> Stiftstr. 34-36
> D-45470 M=FClheim an der Ruhr
> Germany
> E-Mail: Frank.Neese]~[mpi-mail.mpg.de
> = -------------------------------------------------------------------------<= br> >



-=3D This is automatically = added to each message by the mailing script =3D-


E-mail to subscribers: CHEMISTRY(~)ccl.net or use:
     http://www.ccl.net/cgi-bin/ccl/send_ccl_message

E-mail to administrators: CHEMISTRY-REQUEST(~)ccl.net or use
     http://www.ccl.net/cgi-bin/ccl/send_ccl_message

Subscribe/Unsubscribe:
     http://www.ccl.net/chemistry/sub_unsub.shtml

Before posting, check wait time at: http://www.ccl.net

Job: http://www.ccl.net/jobs
Conferences: http://server.ccl.net/chemistry/announcements/conference= s/

Search Messages: http://www.ccl.net/chemistry/searchccl/index.shtml

     http://www.ccl.net/spammers.txt

RTFI: http://www.ccl.net/chemistry/aboutccl/instructions/<= br>



seth.olsen,+,uq.edu.au
+61 7 3365 = 2816
---------------------------------------------------
Unless = stated otherwise, this e-mail represents only the views of the = Sender and not the views of The University of = Queensland

= --Apple-Mail-2--912677580-- From owner-chemistry@ccl.net Wed Mar 14 19:26:00 2012 From: "=?iso-8859-1?Q?J=FCrgen_Gr=E4fenstein?= jurgen-$-chem.gu.se" To: CCL Subject: CCL: CASSCF does not produce spin densities Message-Id: <-46494-120314192328-26044-Zro7T+WsdSNg5FPTMKL2VQ[-]server.ccl.net> X-Original-From: =?iso-8859-1?Q?J=FCrgen_Gr=E4fenstein?= Content-ID: <3B10CBBF996F0545A96802EB5D733A75[-]chalmers.se> Content-Language: en-US Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset="iso-8859-1" Date: Wed, 14 Mar 2012 23:23:15 +0000 MIME-Version: 1.0 Sent to CCL by: =?iso-8859-1?Q?J=FCrgen_Gr=E4fenstein?= [jurgen:+:chem.gu.se] On 14 Mar, 2012, at 13:48 , Seth Olsen seth.olsen!^!uq.edu.au wrote: > > It's also worth mentioning that I've only ever dealt with state-averaged problems. It is possible that distinguishing between "static" and "dynamic" correlations may be operationally useful for ground state problems, where the idea of a well-defined reference makes somewhat more sense (although still not very much, because of the orbital invariance of the CAS expansion). The latter is actually no issue. You can rotate the orbitals in a HF or CASSCF wave function but you will still keep its character, i.e. single-reference or multi-reference. > It's pretty clear that the consensus is that "static" correlations are correlations required by the constraint that the state transform as a particular irrep of the symmetric group, and that that "dynamic" correlations are associated with the Coulomb hole. If there was no Coulomb repulsion the electrons would not avoid each other and the ground state for H_2, however stretched, would be (1 \sigma_g)^2. Unfortunately, misleading statements of the kind "dynamic correlation is driven by Coulomb interaction, static correlation by the near-degeneracy of two or more configurations" are quite common in the literature. Both kind of correlations are driven by Coulomb repulsion; however, a set of quasi-degenerate configurations responds differently to electron-electron repulsion than a bunch of configurations higher up in energy. On the other hand, both static and dynamic correlation has to maintain the symmetry (more strictly, the IRREP) of the wave function. That is, in both cases only configurations from the right IRREP contribute to the CI expansion. Also, dynamic correlation is probably dominated by, but definitely not restricted to, two.electron interactions. > I get all that. But, this tells you right away that it is not possible to build any operator whose expectation will give you a measure of pure "static" or "dynamic" correlations. This is because the requirement of transformation as an irrep of S_n will entangle all degrees of freedom, while the Coulomb operator generates pairwise entanglements. The operators act on different Hilbert spaces. -Seth The proper definition of static and dynamic correlation becomes topical in connection with CAS-DFT methods, which a number of authors (including myself) have struggled and struggle with. In this context, a physically motivated, "waterproof" definition of the two correlation contributions would be of great value. Best regards, Jürgen From owner-chemistry@ccl.net Wed Mar 14 20:01:01 2012 From: "Brian Salter-Duke brian.james.duke*o*gmail.com" To: CCL Subject: CCL: CASSCF does not produce spin densities Message-Id: <-46495-120314141352-1261-4WezsA1tG/jCk9oyTNWsyg]=[server.ccl.net> X-Original-From: Brian Salter-Duke Content-Disposition: inline Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset=iso-8859-1 Date: Wed, 14 Mar 2012 18:13:43 +0000 MIME-Version: 1.0 Sent to CCL by: Brian Salter-Duke [brian.james.duke]~[gmail.com] On Wed, Mar 14, 2012 at 09:26:15AM +1000, Seth Olsen seth.olsen-x-uq.edu.au wrote: I agree with my Australian colleage, Seth, but would like to add a different point. First, it is of course well known that valence bond theory introduces static correlation. Second, it is also know from the work of David Cooper and others, that CASSCF functions can be strongly linked to spin coupled VB functions. The latter can be projected out of the CASSCF function and are a very good approximation to the CASSCF. Thus usefull CASSCF functions are ones that are linked in this way to VB functions. If we have a larger active space, yes, we are including dynamic correlation, but perhaps we would be better doing it a different way from CASSCF. Brian. > Sent to CCL by: Seth Olsen [seth.olsen===uq.edu.au] > With all due respect, I think this conversation is getting foolish. > The concepts of "static" vs "dynamic" correlation do not have meaning > in the context of CASSF models. That is, they usually don't have > well-defined meaning in any case (I.e. everyone has their own) but any > reasonable meaning falls apart in a complete active space context - > because the space is complete! In the asymptotic limit where all > orbitals are included, it becomes fill CI, and obviously treats both. > I know of no sensible way to cleanly distinguish the turnover where > "static" correlation ends and where "dynamic" correlation begins, > regardless of the length of the expansion. > This field needs a revision of its core concepts, soon. Slaughter > this sacred cow before it rots on its own. > Seth Olsen, Ph.D. > ARC Australian Research Fellow > School of Mathematics & Physics > The University of Queensland > Brisbane, QLD 4072 > Australia > +61 7 3365 2816 > > On 14/03/2012, at 4:17, "Neese, Frank frank.neese{:}mpi-mail.mpg.de" wrote: > > > In response to Georg and Jim: > > > > > >> with all respect, but Frank Neese's statement is wrong. > > > > it looks like we are opening Pandora's box here :-) > > > > I nevertheless take the liberty to disagree with the statement brought forward by Georg. (Not with the stated facts of course. I do understand what CASSCF is and having programmed diverse multireference methods over the years I do claim to have at least some appreciation of multireference wavefunction theory; I know the Schlegl/Robb papers that were quoted by Georg; they report a convergence enhancement algorithm for CASSCF wave functions and an application to a reaction between two very small molecules, i.e. H2CO --> H2 + CO. Very nice work! But IMHO these papers do not disprove anything I said). > > > > My arguments in favor of stating what I have stated in my previous post are summarized below: > > > > As we all know CASSCF is limited to very few number of active orbitals. The majority of reseachers who have worked intensely in the field of multiconfigurational self consistent field theory agree that this type of wavefunction is not meant to cover dynamic correlation which is what would be required to really go beyond Hartree-Fock level accuracy. Compared to, say, a closed shell or ROHF CASSCF obviously does include a tiny amount of dynamic correlation (unless your active space is trivially small and reduces to the closed shell or ROHF determinant). However, that tiny little bit of dynamic correlation is essentially irrelevant. What CASSCF is designed to do is to generalize the Hartree-Fock method to cases where the single determinant Ansatz is qualitatively wrong. There are many papers by eminent researchers like Peter Pulay or Jean Paul Malrieu who state that it would be an abuse of the CASSCF method to want to cover more than the essential model space (which is the m! > inimal active space that is needed to cover the static correlation; that again can mean many things. Sometimes it means ensuring that your dissociation limit is correctly represented. Obviously here is where ones chemical thinking comes in). I personally agree with this reasoning. It is then logical to state that CASSCF is not a high level method because its accuracy will be similar to that of, say, RHF in the cases where a closed shell determinant is reasonable. Thus, CASSCF is of broader applicability (even much broader) and it is more consistent than R(O)HF, but it is not of strikingly higher accuracy - provided one compares apples with apples and oranges with oranges. > > > > Of course if one defines anything that has more than one determinant as a post Hartree-Fock method then, yes, there is no doubt that CASSCF is post-HF. It is not, however, my favorite definition of what a post-HF method is designed to be. I would tend to use this term if it is implied that a substantial fraction of the (basis set) dynamic correlation energy is recovered by the treatment (what is meant by "dynamic correlation" is, again, a matter of the reference point. If you argue > from a multireference perturbation theory or multireference configuration interaction or multireference coupled cluster point of view then CASSCF is the starting point and does not contain dynamic correlation). > > > > At the end of the day everybody is of course free to do whatever calculation one wants and maybe one has good reasons for wanting to cover some dynamic correlation already with a CASSCF wave function. A typical example is the double d-shell effect for transition metal calculations in conjunction with the CASSCF/CASPT2 method that in particular Kristin Pierloot has elaborated on many times (in this case one covers a radial correlation effect). My personal preference, however, is to do this with a post-CASSCF treatment (if one can afford it). > > > > We too have had many good results with CASSCF and transition metals. My concerns were directed at the particular problem that the colleague was mentioning. Having worked on both, multireference wavefunctions and iron nitrosyl systems, I have simply expressed concerns to regard CASSCF as a high level treatment that should be given preference over any other theoretical method. To do so would, in my opinion, require solid backup > from comparison of calculated to observed physical properties. > > > > The discussion that we have been starting here could very quickly turn into a hair splitting dispute about terminology and I hope that this is not what will happen. I have simply tried here to explain what my take on the subject is and I do, of course, respect any other approach or opinion that leads to creative and interesting science. > > > > Best regards > > Frank > > > > ------------------------------------------------------------------------- > > Prof. Dr. Frank Neese > > - Director - > > Max-Planck Institut für Bioanorganische Chemie > > Stiftstr. 34-36 > > D-45470 Mülheim an der Ruhr > > Germany > > E-Mail: Frank.Neese]~[mpi-mail.mpg.de > > -------------------------------------------------------------------------> -- Brian Salter-Duke (Brian Duke) Brian.Salter-Duke-$-monash.edu Adjunct Associate Professor Monash Institute of Pharmaceutical Sciences Monash University Parkville Campus, VIC 3052, Australia From owner-chemistry@ccl.net Wed Mar 14 20:36:01 2012 From: "Ananth Kaushik pa234]![cornell.edu" To: CCL Subject: CCL: Freezing only x coordinates of atoms in g09 optimizations Message-Id: <-46496-120314161817-2446-Jn78GjSk6F6u7fxt8xgXig[]server.ccl.net> X-Original-From: "Ananth Kaushik" Date: Wed, 14 Mar 2012 16:18:15 -0400 Sent to CCL by: "Ananth Kaushik" [pa234-#-cornell.edu] Hi, I am trying to do a geometry optimization on a large system but wish to contrain most of the atoms and allow only a few atoms to move. I want to have all x,y,z coordinates of some of the atoms to be frozen. For this, I have used the opt=modredundant option and after the molecule specification I have: X 1 F X 2 F From owner-chemistry@ccl.net Wed Mar 14 21:11:00 2012 From: "Seth Olsen seth.olsen[a]uq.edu.au" To: CCL Subject: CCL: CASSCF does not produce spin densities Message-Id: <-46497-120314172638-23180-jPuzSoD7aJg1RZFjCuSU3Q:server.ccl.net> X-Original-From: Seth Olsen Content-Language: en-US Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset="utf-8" Date: Thu, 15 Mar 2012 07:26:20 +1000 MIME-Version: 1.0 Sent to CCL by: Seth Olsen [seth.olsen-x-uq.edu.au] Comment I settled below with >> Seth Olsen, Ph.D. ARC Australian Research Fellow School of Mathematics & Physics The University of Queensland Brisbane, QLD 4072 Australia +61 7 3365 2816 On 15/03/2012, at 4:54, "Jürgen Gräfenstein jurgen:+:chem.gu.se" wrote: > > Sent to CCL by: =?iso-8859-1?Q?J=FCrgen_Gr=E4fenstein?= [jurgen]_[chem.gu.se] > +++ Seth Olsen wrote: > > The concepts of "static" vs "dynamic" correlation do not have meaning in the context of CASSF models. That is, they usually don't have well-defined meaning in any case (I.e. everyone has their own) > > +++ You are right, there is no exact and generally accepted distinction between th two types of correlations. Notwithstanding, it is obvious that there are two essentially different correlation patterns, which require essentially different ways of treatment in quantum chemistry. > >> I'm sure it's obvious if you already believe it. > but any reasonable meaning falls apart in a complete active space context - because the space is complete! > > +++ This is plainly incorrect. "Complete active space" means that for the orbitals chosen to be active, all possible configurations are incorporates (as opposed to RASSCF methods and the like). It does not mean that all orbitals are considered active. > >> Yes, I do know this. I suppose I meant "complete within the active space". For an evenly weighted average of states, the identity is also resolved on the target many-electron subspace. The point is that all correlations are included in the spaces over which the identity is resolved, and there is no natural distinction on these. > +++ The art with CASSCF is just to select just those orbitals as active that account for the non-dynamic correlations (and to do this in a consistent fashion for reactants, intermediates/TS's, and products). > >>I appreciate that it is an art form. I love CASSCF dearly (particularly stare-averaged, where the extra layer of classical probability makes it more interesting). I use it all the time. Look me up if you doubt this. > In the asymptotic limit where all orbitals are included, it becomes fill CI, and obviously treats both. > > +++ This is uncontroversial but of little practical relevance. Thus, we have to continue bothering about static correlation. > >>you have to understand how to interpret the method. I'm not saying you don't. I'm saying that these terms have become dogma that everyone uses but no one understands, so they don't end up clarifying anything. > I know of no sensible way to cleanly distinguish the turnover where "static" correlation ends and where "dynamic" correlation begins, regardless of the length of the expansion. > > +++ If one finds an explicit (physically motivated) separation between non-dynamic and dynamic correlation, probably any CASSCF wavefunction will contain both of them. With increasing size of the active space, the portion of dynamical correlation will increase. There will be no CASSCF WF that contains all of the static and none of the dynamic correlation. See e.g. Gräfenstein/Cremer, PCCP 2, 2091 (2000); Mol. Phys. 103, 279 (2005) for a discussion. > >> I'll look this up and tell you what I think. I am keen to find discussions. Many thanks. > This field needs a revision of its core concepts, soon. > > +++ A strict definition of non-dynamic vs dynamic correlation would be extremely valuable; I agree with this. > > Slaughter this sacred cow before it rots on its own. > > +++ It won't. If you doubt this, just try to describe dissociating H_2 with MP2/MP3/MP4. > I know what happens, and it doesn't sway me. Saying that something is wrong is not the same as proposing a solution. > Best regards, > Jürgen Gräfenstein > University of Gothenburg, Department of Chemistry > SE-412 96 GÖTEBORG, SWEDEN > Phone: +46-31-786 9016 > Fax: +46-31-772 3840 > e-mail: Jurgen.Grafenstein^chem.gu.se> > From owner-chemistry@ccl.net Wed Mar 14 21:46:00 2012 From: "Guenter Grethe ggrethe---att.net" To: CCL Subject: CCL: Reminder to apply for CINF-FIZ Scholarships for Scientific Excellence Message-Id: <-46498-120314183434-7223-rQYIQSi8imYFWOIGX7I61g]^[server.ccl.net> X-Original-From: "Guenter Grethe" Date: Wed, 14 Mar 2012 18:34:22 -0400 Sent to CCL by: "Guenter Grethe" [ggrethe/a\att.net] Last Call --- Last Call --- Last Call 2012 CINF Scholarship for Scientific Excellence Sponsored by FIZ Chemie Berlin The scholarship program of the Division of Chemical Information (CINF) of the American Chemical Society (ACS) funded by FIZ Chemie Berlin is designed to reward graduate and postdoctoral students in chemical information and related sciences for scientific excellence and to foster their involvement in CINF. Up to three scholarships valued at $1,000 each will be presented at the 244th ACS National Meeting in Philadelphia, PA, August 19 August 23, 2012. Applicants must be enrolled at a certified college or university, and they will present a poster during the Welcoming Reception of the division on Sunday evening at the National Meeting. Additionally, they will have the option to also show their poster at the Sci-Mix session on Monday night. Abstracts for the poster must be submitted electronically through PACS, the abstract submission system of ACS. To apply, please inform the Chair of the selection committee, Guenter Grethe at ggrethe##att.net that you are applying for a scholarship. Submit your abstract to http://abstracts.acs.org using your ACS ID. If you do not have an ACS ID, follow the registration instructions and submit your abstract in the CINF program for the session CINF Scholarship for Scientific Excellence. PACS will be open for abstract submissions from January 23 to March 31, 2012. Additionally, please send a 2,000-word abstract describing the work to be presented in electronic form to the Chair of the selection committee by June 30, 2012. Any questions related to applying for one of the scholarships should be directed to the same e-mail address. Winners will be chosen based on contents, presentation and relevance of the poster and they will be announced during the reception. The contents shall reflect upon the students work and describe research in the field of cheminformatics and related sciences. Winning posters will be marked Winner of FIZ Chemie-CINF Scholarship for Scientific Excellence at the poster session. Guenter Grethe