From owner-chemistry@ccl.net Wed Feb 12 08:49:00 2014 From: "Varun Kundi chemvarun]![gmail.com" To: CCL Subject: CCL:G: Printing of two electron integrals. Message-Id: <-49673-140212061705-20365-7+eGSb5nZGMLSTxgrPCiXA|server.ccl.net> X-Original-From: "Varun Kundi" Date: Wed, 12 Feb 2014 06:17:04 -0500 Sent to CCL by: "Varun Kundi" [chemvarun(~)gmail.com] Hello, I have printed 2e integrals using #N HF/STO-3G SCF=Conventional IOP(3/33=6) EXTRALINKS=l316 NORAFF in Gaussian. How to read 2e integral values? From owner-chemistry@ccl.net Wed Feb 12 10:41:00 2014 From: "Steven Bachrach sbachrac_._trinity.edu" To: CCL Subject: CCL:G: How to consider charge on particular atom -lithium Message-Id: <-49674-140212100924-14208-PDFeIg+A/R0l5puyy2AnmA/a\server.ccl.net> X-Original-From: Steven Bachrach Content-Type: multipart/alternative; boundary=089e013cc070cdb76c04f236f4e0 Date: Wed, 12 Feb 2014 09:09:15 -0600 MIME-Version: 1.0 Sent to CCL by: Steven Bachrach [sbachrac^_^trinity.edu] --089e013cc070cdb76c04f236f4e0 Content-Type: text/plain; charset=ISO-8859-1 Chris is of course correct in his assessment of Bader charges. But I would like to add just a few observations. First off, there is no QM operator that corresponds to atomic charge - meaning there is no non-arbitrary method for obtaining the charge on an atom within a molecule. So it is pointless to argue about which method is "best". Rather, one should ask which is most appropriate for the question at hand. Bader charges are great if what you want to do is match up with the experimental electron distribution one gets from an x-ray diffraction experiment. Here one gets a 3-D map and one can partition the density according to the zero-flux surface and directly compare the experimental with computed "charge". If your desire to to mimic the electrostatic potential, then Chris is correct that the Bader charges are large and are compensated by higher order moments. So if you want to mimic the electrostatic potential with just monopoles, then Bader charges are a poor choice. But, if you allow for dipoles, quadrupoles, and higher moments, then the values one gets from AIM will be just fine. Keep in mind that ultimately one is making an arbitrary selection here, and both positive and negative characteristics pervade each choice. Choose your poison judiciously. Steven On Tue, Feb 11, 2014 at 5:31 PM, Christopher Cramer cramer%x%umn.edu < owner-chemistry]^[ccl.net> wrote: > > Sent to CCL by: Christopher Cramer [cramer**umn.edu] > Brian, > > Sure, I can take a run at your final question. The "problem" with Bader > charges is not that there is any lack of elegance in the definition of the > atomic basins, nor that they have a complete basis set limit (surely a good > thing!) The problem is that we usually want partial atomic charges to do > the best job possible of representing the molecular charge distribution > with a set of atom-centered monopoles, but there may be rather large HIGHER > moments of the charge within the Bader atomic basins (i.e., beyond the > difference in total number of electrons integrated over the Bader atom > subtracted from the nuclear charge). When that happens, collapsing all > those electrons entirely on the nuclear position, about which they are not > symmetrically disposed, can lead to rather bizarre charges. > > A good example that I recall from many years ago was work of Rainer > Glaser at Missouri. He showed that diazonium cations (i.e., RN2+ > structures) would show enormous variation in the two different N charges > with seemingly tiny variations in R (e.g., going from methyl to ethyl). The > problem is that LOTS of the electronic charge is built up about midway > between the two N atoms, but the zero-flux surface is VERY sensitive to the > substituent. So, a small movement of the zero-flux surface to one side or > the other suddenly moves a sizable fraction of an electron from one N > "atom" to the other. Of course, the ACTUAL charge distribution is actually > only very slightly perturbed, and if one were to consider the atomic dipole > moments within the basins, one would see that their variations largely > cancel the changes in the monopoles, but we aren't usually interested in > using higher atomic moments when we're looking for partial atomic charges -- > we're hoping for a simpler representation. > > So much is written on partial atomic charges that I won't indulge my > otherwise dangerous proclivity to lecture, but I thought I'd at least > address your question of why Bader charges tend to find little use in > communities not interested in considering higher moments of the atomic > basin charge distributions as well. > > Best, > > Chris > > On Feb 11, 2014, at 14:49, Salter-Duke, Brian James brian.james.duke## > gmail.com wrote: > > > > > Sent to CCL by: "Salter-Duke, Brian James " [ > brian.james.duke-x-gmail.com] > > I do not want to address how the various methods work in practice, as I > > do not have the experience. I do however want to make a general point > > and then ask a question. > > > > Mulliken charges are not basis set independent as they depend on the > > basis functions we happen to use on each atom. As another poster > > commented, if we use a one centre complete basis the method puts all the > > charge on that atom. Long ago there was a set of one-centre expansion > > calculations on simple systems such as methane. There are no basis > > functions on the hydrogen atoms. The Mulliken charges are C(4-) and > > H(+). If we used basis functions centered only on the H atoms we would > > get C(4+) and H(-). Mulliken charges do not have a basis set limit. Too > > often we use methods to interpret wave functions that do not have a > > basis set limit. I suggest we stop doing that and use methods that do, > > just as we have energies that do. We often extrapolate to that energy > > limit and we should extrapolate to limits for other properties, even > > when they are not observable properties such as charges. > > > > As others have shown many other methods can reduce the sensitivity of > > NPA charges to basis-set, but they still do not properly have a basis > > set limit. > > > > Bader charges depend on the basis set only in the sense that the density > > depends on the basis set. They do have a proper basis set limit. They > > are obtained by defining the boundaries of each atom and then > > integrating over the atoms. The AIM method of getting those boundaries > > seems soundly based. Nobody, I think, has come up with a better method. > > So why are Bader charges thought to be so bad and unacceptable. Has that > > question been properly considered and analysed? > > > > Brian Duke. > > > > On Tue, Feb 11, 2014 at 10:09:02AM -0500, Tian Lu sobereva-x-sina.comwrote: > >> > >> Sent to CCL by: "Tian Lu" [sobereva/./sina.com] > >> Hi, > >> > >> AFAIK, there is only one public program can realize Hirshfeld-I, namely > HiPart (http://molmod.ugent.be/software/). > > > >> Another modified Hirshfeld-based method alternative to Hirshfeld-I is > >> atomic dipole moment corrected Hirshfeld (ADCH) population method, > >> which is the one I highly recommend, see J. Theor. Comp. Chem., 2012, > >> 11: 163-183. ADCH charges have much better reproducibility for > >> electrostatic potential than Hirshfeld charges, and the molecular > >> dipole moment can be even exactly reproduced. ADCH has been > >> implemented in Multiwfn program (http://Multiwfn.codeplex.com, see > >> Section 4.7.2 of its manual for example). > > > >> NPA charges (also known as NBO charges) are also nice choice. In fact > >> they are not explicitly dependent on but only indirectly dependent on > >> the basis-set, because the original basis functions will be first > >> transformed to natural atomic orbitals before performing natural > >> population analysis, this step conspicuously reduces the sensitivity > >> of NPA charges to basis-set. According to my experiences, the relative > >> sensitivity to basis-set is Mulliken>=Lowdin>>NPAAIMcharges by fitting > >> ESP (MK,CHELPG,etc.) >= HirshfeldADCH. > > > >> Personally I don't recommend using AIM charges, since calculating AIM > >> charges is usually time-consuming, and their reproducibility for > >> observable quantities are quite poor. > > > >> A comprehensive comparison of atomic charges can be found in Acta > >> Phys.-Chim. Sinica, 2011, 28: 1-18 > >> (http://www.whxb.pku.edu.cn/EN/abstract/abstract27818.shtml) > > > >> > >> Tian Lu > >> > >> > >> > >> > >> ----- Original Message ----- > >>> From: "Susi Lehtola susi.lehtola]![alumni.helsinki.fi" > > >> To: "Lu, Tian " > >> Subject: CCL:G: How to consider charge on particular atom -lithium > >> Date: 2014-02-11 17:12 > >> > >> > >> > >> Sent to CCL by: Susi Lehtola [susi.lehtola-.-alumni.helsinki.fi] > >> On Mon, 10 Feb 2014 18:37:44 -0500 > >> "Jim Kress ccl_nospam_._kressworks.com" > wrote: > >>> Sent to CCL by: "Jim Kress" [ccl_nospam,kressworks.com] > >>> AIM and NBO charges would be the best choice. Mulliken and Lowdin are > far too basis set dependent. > >> NBO charges are explicitly dependent on the basis set, like Mulliken > >> and Lwdin. > >> AIM charges are not, but then again according to them e.g. H2O and HCN > >> are ionic... > >> Probably the best scheme is something like the recently proposed > >> iterative Hirshfeld schemes, but I'm not aware of any program that > >> implements these. > >> Another possibility are electrostatic potential charges, which are > >> available in e.g. Gaussian. > >> -- > >> --------------------------------------------------------------- > >> Mr. Susi Lehtola, PhD Research Associate > >> susi.lehtola%a%alumni.helsinki.fi Department of Applied Physics > >> http://www.helsinki.fi/~jzlehtol Aalto University > >> Finland > >> --------------------------------------------------------------- > >> Susi Lehtola, FT Tutkijatohtori > >> susi.lehtola%a%alumni.helsinki.fi Fysiikan laitos > >> http://www.helsinki.fi/~jzlehtol Aalto-yliopisto> > > > > -- > > 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> > > > > -- > Christopher J. Cramer > Elmore H. Northey Professor and > Associate Dean for Academic Affairs > University of Minnesota > Department of Chemistry and > College of Science & Engineering > Minneapolis, MN 55455-0431 > Phone: (612) 624-0859 (Chemistry) > Phone: (612) 624-9371 (CSE) > -------------------------- > Mobile: (952) 297-2575 > Email: cramer##umn.edu > Twitter: ##ChemProfCramer > Website: http://pollux.chem.umn.edu> > > -- Steven Bachrach Semmes Distinguished Professor and Assistant VP for Special Projects Department of Chemistry Trinity University Phone: 210-999-7379 1 Trinity Place Fax: 210-999-7569 San Antonio, TX 78212 email: sbachrach]^[trinity.edu --089e013cc070cdb76c04f236f4e0 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable
Chris is of course correct i= n his assessment of Bader charges.

But I would like to add jus= t a few observations. First off, there is no QM operator that corresponds t= o atomic charge - meaning there is no non-arbitrary method for obtaining th= e charge on an atom within a molecule. So it is pointless to argue about wh= ich method is "best". Rather, one should ask which is most approp= riate for the question at hand.

Bader charges are great if what you want to do is match up with t= he experimental electron distribution one gets from an x-ray diffraction ex= periment. Here one gets a 3-D map and one can partition the density accordi= ng to the zero-flux surface and directly compare the experimental with comp= uted "charge".

If your desire to to mimic the electrostatic potential, then Chri= s is correct that the Bader charges are large and are compensated by higher= order moments. So if you want to mimic the electrostatic potential with ju= st monopoles, then Bader charges are a poor choice. But, if you allow for d= ipoles, quadrupoles, and higher moments, then the values one gets from AIM = will be just fine.

Keep in mind that ultimately one is making an arbitrary selection= here, and both positive and negative characteristics pervade each choice. =

Choose your poison judiciously.

Steven


On Tue, Feb 1= 1, 2014 at 5:31 PM, Christopher Cramer cramer%x%= umn.edu <owner-chemistry]^[ccl.net> wrote:

Sent to CCL by: Christopher Cramer [cramer**umn.edu]
Brian,

   Sure, I can take a run at your final question. The “prob= lem” with Bader charges is not that there is any lack of elegance in = the definition of the atomic basins, nor that they have a complete basis se= t limit (surely a good thing!) The problem is that we usually want partial = atomic charges to do the best job possible of representing the molecular ch= arge distribution with a set of atom-centered monopoles, but there may be r= ather large HIGHER moments of the charge within the Bader atomic basins (i.= e., beyond the difference in total number of electrons integrated over the = Bader atom subtracted from the nuclear charge). When that happens, collapsi= ng all those electrons entirely on the nuclear position, about which they a= re not symmetrically disposed, can lead to rather bizarre charges.

   A good example that I recall from many years ago was work of R= ainer Glaser at Missouri. He showed that diazonium cations (i.e., RN2+ stru= ctures) would show enormous variation in the two different N charges with s= eemingly tiny variations in R (e.g., going from methyl to ethyl). The probl= em is that LOTS of the electronic charge is built up about midway between t= he two N atoms, but the zero-flux surface is VERY sensitive to the substitu= ent. So, a small movement of the zero-flux surface to one side or the other= suddenly moves a sizable fraction of an electron from one N “atom&rd= quo; to the other. Of course, the ACTUAL charge distribution is actually on= ly very slightly perturbed, and if one were to consider the atomic dipole m= oments within the basins, one would see that their variations largely cance= l the changes in the monopoles, but we aren’t usually interested in u= sing higher atomic moments when we’re looking for partial atomic char= ges — we’re hoping for a simpler representation.

   So much is written on partial atomic charges that I won’= t indulge my otherwise dangerous proclivity to lecture, but I thought I&rsq= uo;d at least address your question of why Bader charges tend to find littl= e use in communities not interested in considering higher moments of the at= omic basin charge distributions as well.

Best,

Chris

On Feb 11, 2014, at 14:49, Salter-Duke, Brian James  brian.james.duke#= #gmail.com <owner-che= mistry##ccl.net> wrote:=

>
> Sent to CCL by: "Salter-Duke, Brian James " [brian.james.duke-x-gmail= .com]
> I do not want to address how the various method= s work in practice, as I
> do not have the experience. I do however want to make a general point<= br> > and then ask a question.
>
> Mulliken charges are not basis set independent as they depend on the > basis functions we happen to use on each atom. As another poster
> commented, if we use a one centre complete basis the method puts all t= he
> charge on that atom. Long ago there was a set of one-centre expansion<= br> > calculations on simple systems such as methane. There are no basis
> functions on the hydrogen atoms. The Mulliken charges are C(4-) and > H(+). If we used basis functions centered only on the H atoms we would=
> get C(4+) and H(-). Mulliken charges do not have a basis set limit. To= o
> often we use methods to interpret wave functions that do not have a > basis set limit. I suggest we stop doing that and use methods that do,=
> just as we have energies that do. We often extrapolate to that energy<= br> > limit and we should extrapolate to limits for other properties, even > when they are not observable properties such as charges.
>
> As others have shown many other methods can reduce the sensitivity of<= br> > NPA charges to basis-set, but they still do not properly have a basis<= br> > set limit.
>
> Bader charges depend on the basis set only in the sense that the densi= ty
> depends on the basis set. They do have a proper basis set limit. They<= br> > are obtained by defining the boundaries of each atom and then
> integrating over the atoms. The AIM method of getting those boundaries=
> seems soundly based. Nobody, I think, has come up with a better method= .
> So why are Bader charges thought to be so bad and unacceptable. Has th= at
> question been properly considered and analysed?
>
> Brian Duke.
>
> On Tue, Feb 11, 2014 at 10:09:02AM -0500, Tian Lu sobereva-x-sina.com wrote:
>>
>> Sent to CCL by: "Tian  Lu" [sobereva/./sina.com]
>> Hi,
>>
>> AFAIK, there is only one public program can realize Hirshfeld-I, n= amely HiPart (http://molmod.ugent.be/software/).
>
>> Another modified Hirshfeld-based method alternative to Hirshfeld-I= is
>> atomic dipole moment corrected Hirshfeld (ADCH) population method,=
>> which is the one I highly recommend, see J. Theor. Comp. Chem., 20= 12,
>> 11: 163-183. ADCH charges have much better reproducibility for
>> electrostatic potential than Hirshfeld charges, and the molecular<= br> >> dipole moment can be even exactly reproduced. ADCH has been
>> implemented in Multiwfn program (http://Multiwfn.codeplex.com, see
>> Section 4.7.2 of its manual for example).
>
>> NPA charges (also known as NBO charges) are also nice choice. In f= act
>> they are not explicitly dependent on but only indirectly dependent= on
>> the basis-set, because the original basis functions will be first<= br> >> transformed to natural atomic orbitals before performing natural >> population analysis, this step conspicuously reduces the sensitivi= ty
>> of NPA charges to basis-set. According to my experiences, the rela= tive
>> sensitivity to basis-set is Mulliken>=3DLowdin>>NPAAIMcha= rges by fitting
>> ESP (MK,CHELPG,etc.) >=3D HirshfeldADCH.
>
>> Personally I don't recommend using AIM charges, since calculat= ing AIM
>> charges is usually time-consuming, and their reproducibility for >> observable quantities are quite poor.
>
>> A comprehensive comparison of atomic charges can be found in Acta<= br> >> Phys.-Chim. Sinica, 2011, 28: 1-18
>> (http://www.whxb.pku.edu.cn/EN/abstract/abstract2781= 8.shtml)
>
>>
>> Tian Lu
>>
>>
>>
>>
>> ----- Original Message -----
>>> From: "Susi Lehtola susi.lehtola]![alumni.helsinki.fi" <owner-ch= emistry+/-ccl.net>
>> To: "Lu, Tian " <sobereva+/-sina.com>
>> Subject: CCL:G: How to consider charge on particular atom -lithium=
>> Date: 2014-02-11 17:12
>>
>>
>>
>> Sent to CCL by: Susi Lehtola [susi.lehtola-.-alumni.helsinki.fi]
>> On Mon, 10 Feb 2014 18:37:44 -0500
>> "Jim Kress ccl_nospam_._kressworks.com" <owner-chemistry%a%ccl.net> wrote:
>>> Sent to CCL by: "Jim Kress" [ccl_nospam,kressworks.com]
>>> AIM and NBO charges would be the best choice. Mulliken and Low= din are far too basis set dependent.
>> NBO charges are explicitly dependent on the basis set, like Mullik= en
>> and Lwdin.
>> AIM charges are not, but then again according to them e.g. H2O and= HCN
>> are ionic...
>> Probably the best scheme is something like the recently proposed >> iterative Hirshfeld schemes, but I'm not aware of any program = that
>> implements these.
>> Another possibility are electrostatic potential charges, which are=
>> available in e.g. Gaussian.
>> --
>> --------------------------------------------------------------- >> Mr. Susi Lehtola, PhD Research Associate
>> susi.lehtola%a%alumni.helsinki.fi Department of Applied Physics
>> htt= p://www.helsinki.fi/~jzlehtol Aalto University
>> Finland
>> --------------------------------------------------------------- >> Susi Lehtola, FT Tutkijatohtori
>> susi.lehtola%a%alumni.helsinki.fi Fysiikan laitos
>> htt= p://www.helsinki.fi/~jzlehtol Aalto-yliopisto>
>
> --
>   Brian Salter-Duke (Brian Duke)   Brian.Salter-Duke|monash.edu
>                    A= djunct Associate Professor
>            Monash Institute of Pharmaceu= tical Sciences
>      Monash University Parkville Campus, VIC 3052, Aust= ralia>
>

--
Christopher J. Cramer
Elmore H. Northey Professor and
  Associate Dean for Academic Affairs
University of Minnesota
Department of Chemistry and
  College of Science & Engineering
Minneapolis, MN 55455-0431
Phone:  (6= 12) 624-0859 (Chemistry)
Phone:  (6= 12) 624-9371 (CSE)
--------------------------
Mobile: (952) 2= 97-2575
Email:  cramer##umn= .edu
Twitter:  ##ChemProfCramer
Website:  htt= p://pollux.chem.umn.edu



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

E-mail to administrators: CHEM= ISTRY-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.n= et/jobs
Conferences: http://server.ccl.net/chemistry/announcements/co= nferences/

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

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





--
Steven Bachrach
Semmes Distingui= shed Professor and Assistant VP for Special Projects
Department of Chemistry
Trinity University                 =          Phone: 210-999-7379
1 Trinity Place                 &nb= sp;            Fax: 210-999-7569
San Antonio, TX 78212               &nbs= p; email: sbachr= ach]^[trinity.edu
--089e013cc070cdb76c04f236f4e0-- From owner-chemistry@ccl.net Wed Feb 12 13:16:01 2014 From: "Cory Pye cpye,ap.smu.ca" To: CCL Subject: CCL: Converges using HF and hybrid DFT, but fails to converge using pure DFT Message-Id: <-49675-140212104013-16217-9AT4PsOJrUiMZ+GN6gms0A\a/server.ccl.net> X-Original-From: Cory Pye Content-Type: TEXT/PLAIN; charset=US-ASCII Date: Wed, 12 Feb 2014 11:40:06 -0400 (AST) MIME-Version: 1.0 Sent to CCL by: Cory Pye [cpye**ap.smu.ca] Hello, Another thing to try is turning off the incremental Fock matrix construction. SCF=(XQC,NoIncrFock) It appears that it is restarting the incremental Fock matrix construction several times before not converging. Also make sure that the accuracy of your grid matches the accuracy of your SCF. -Cory On Tue, 11 Feb 2014, Dr. Vitaly Chaban vvchaban::gmail.com wrote: > > Sent to CCL by: "Dr. Vitaly Chaban" [vvchaban/./gmail.com] > On Tue, Feb 11, 2014 at 6:57 PM, Sebastian Kozuch > seb.kozuch!A!gmail.com wrote: > > > > Sent to CCL by: Sebastian Kozuch [seb.kozuch(a)gmail.com] > > That happens sometimes. Pure GGA have some extra convergence difficulties. > > You may try to take the converged orbitals in the checkpoint file from a > > hybrid DFT and use them as a guess for the GGA, and try with scf(qc) or > > scf(xqc). More expensive, but life is always more expensive than our > > expectations. > > > The bad thing is that even SCF=QC with Guess=Read does not work > adequately. Below follows the logfile after a couple of hours of > iterations. The guess came from "B3LYP/CEP-31G* SCF=Conver=9", and the > failed convergence example used "BLYP/CEP-31G*" > > Unless there is no other magic way, I will have to decrease the SCF > criterion down to what BLYP can afford. Maybe indeed a bit exotic > system... > > >>>>>>>>>> Convergence criterion not met. > SCF Done: E(RB-LYP) = -477.317424055 A.U. after 65 cycles > NFock= 64 Conv=0.47D-05 -V/T= 2.4474 > ************* ! Dr. Cory C. Pye ***************** ! Associate Professor *** ** ** ** ! Theoretical and Computational Chemistry ** * **** ! Department of Chemistry, Saint Mary's University ** * * ! 923 Robie Street, Halifax, NS B3H 3C3 ** * * ! cpye::crux.stmarys.ca http://apwww.stmarys.ca/~cpye *** * * ** ! Ph: (902)-420-5654 FAX:(902)-496-8104 ***************** ! ************* ! Les Hartree-Focks (Apologies to Montreal Canadien Fans) From owner-chemistry@ccl.net Wed Feb 12 16:15:00 2014 From: "Frank Jensen frj**chem.au.dk" To: CCL Subject: CCL:G: How to consider charge on particular atom -lithium Message-Id: <-49676-140212150229-25652-6Mj04FZzRYxIeYJsYOQxXw{:}server.ccl.net> X-Original-From: Frank Jensen Content-Language: en-US Content-Type: multipart/alternative; boundary="_000_D5BC00C0FB9AC34D9C8F4DC965C51689479F048FSRVUNIMBX07unia_" Date: Wed, 12 Feb 2014 20:01:42 +0000 MIME-Version: 1.0 Sent to CCL by: Frank Jensen [frj..chem.au.dk] --_000_D5BC00C0FB9AC34D9C8F4DC965C51689479F048FSRVUNIMBX07unia_ Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable Can't resist to add my 0.02$ As Steven points out, there is no QM operator for the atomic charge, and th= erefore no unique choice. The problem is not the QM operator for charge, th= at is just fine, it is the 'atomic' that causes the problem. In a molecule,= where does one 'atom' stop and the next begin? The Bader (QTAIM) definition in terms of zero-flux surfaces is appealing, b= ut beside the sensitivity towards small changes in the system that Chris me= ntions, there are also the troublesome systems where the zero-flux partitio= ning of the 3D space leaves a volume of electron density without a nucleus.= And the reproduction of the molecular moment to order N requires that one = includes all the atomic multipole moments to order N. The experimental dipo= le moment of CO, for example is 0.12 D, which in the QTAIM partitioning is = reproduced by a large monopole (atomic charges) component of -6.54 D and an= atomic dipole component of +6.66 D. NPA charges are defined from the basis set, but since they effectively are = projected onto a minimum basis set, any sensible choice of basis sets will = display a smooth convergence towards a basis set limiting value. But they c= annot be derived from experimental electron densities, like QTAIM. Mulliken and Lowdin are just so terrible sensitive to the present of even s= lightly diffuse basis functions that they never should be used (but will ne= vertheless be used in 90+% of the publications appearing....) Hirshfeld charges are also appealing, but the original method depends on th= e definition of the reference atoms. The iterative version alleviates this = problem, but for strongly ionic system requires interpolation between singl= y and doubly negative charged atoms, and no doubly negative ions are stable= in the gas phase..... The best representation of the total electrostatic potential by atomic char= ges are obtained by explicit fitting, but as many studies have shown, this = is a mathematically ill-defined problem, as the parameter space defined by = the atomic charges is near-redundant for all by the simplest systems. This = leads to conformational and system-sensitive atomic charges, which only to = a certain degree can be alleviated by introducing constraints. Other less used atomic charge definitions have similar pro's and con's. Wit= h Stevens words: Choose your poison. Frank Frank Jensen Assoc. Prof. Dept. of Chemistry Aarhus University http://old.chem.au.dk/~frj > From: owner-chemistry+frj=3D=3Dchem.au.dk(!)ccl.net [mailto:owner-chemistry+f= rj=3D=3Dchem.au.dk(!)ccl.net] On Behalf Of Steven Bachrach sbachrac_._trinity= .edu Sent: 12. februar 2014 16:09 To: Frank Jensen Subject: CCL:G: How to consider charge on particular atom -lithium Chris is of course correct in his assessment of Bader charges. But I would like to add just a few observations. First off, there is no QM = operator that corresponds to atomic charge - meaning there is no non-arbitr= ary method for obtaining the charge on an atom within a molecule. So it is = pointless to argue about which method is "best". Rather, one should ask whi= ch is most appropriate for the question at hand. Bader charges are great if what you want to do is match up with the experim= ental electron distribution one gets from an x-ray diffraction experiment. = Here one gets a 3-D map and one can partition the density according to the = zero-flux surface and directly compare the experimental with computed "char= ge". If your desire to to mimic the electrostatic potential, then Chris is corre= ct that the Bader charges are large and are compensated by higher order mom= ents. So if you want to mimic the electrostatic potential with just monopol= es, then Bader charges are a poor choice. But, if you allow for dipoles, qu= adrupoles, and higher moments, then the values one gets from AIM will be ju= st fine. Keep in mind that ultimately one is making an arbitrary selection here, and= both positive and negative characteristics pervade each choice. Choose your poison judiciously. Steven On Tue, Feb 11, 2014 at 5:31 PM, Christopher Cramer cramer%x%umn.edu > wrote= : Sent to CCL by: Christopher Cramer [cramer**umn.edu] Brian, Sure, I can take a run at your final question. The "problem" with Bader = charges is not that there is any lack of elegance in the definition of the = atomic basins, nor that they have a complete basis set limit (surely a good= thing!) The problem is that we usually want partial atomic charges to do t= he best job possible of representing the molecular charge distribution with= a set of atom-centered monopoles, but there may be rather large HIGHER mom= ents of the charge within the Bader atomic basins (i.e., beyond the differe= nce in total number of electrons integrated over the Bader atom subtracted = > from the nuclear charge). When that happens, collapsing all those electrons= entirely on the nuclear position, about which they are not symmetrically d= isposed, can lead to rather bizarre charges. A good example that I recall from many years ago was work of Rainer Glas= er at Missouri. He showed that diazonium cations (i.e., RN2+ structures) wo= uld show enormous variation in the two different N charges with seemingly t= iny variations in R (e.g., going from methyl to ethyl). The problem is that= LOTS of the electronic charge is built up about midway between the two N a= toms, but the zero-flux surface is VERY sensitive to the substituent. So, a= small movement of the zero-flux surface to one side or the other suddenly = moves a sizable fraction of an electron from one N "atom" to the other. Of = course, the ACTUAL charge distribution is actually only very slightly pertu= rbed, and if one were to consider the atomic dipole moments within the basi= ns, one would see that their variations largely cancel the changes in the m= onopoles, but we aren't usually interested in using higher atomic moments w= hen we're looking for partial atomic charges - we're hoping for a simpler r= epresentation. So much is written on partial atomic charges that I won't indulge my oth= erwise dangerous proclivity to lecture, but I thought I'd at least address = your question of why Bader charges tend to find little use in communities n= ot interested in considering higher moments of the atomic basin charge dist= ributions as well. Best, Chris On Feb 11, 2014, at 14:49, Salter-Duke, Brian James brian.james.duke##gmai= l.com > wrote: > > Sent to CCL by: "Salter-Duke, Brian James " [brian.james.duke-x-gmail.com= ] > I do not want to address how the various methods work in practice, as I > do not have the experience. I do however want to make a general point > and then ask a question. > > Mulliken charges are not basis set independent as they depend on the > basis functions we happen to use on each atom. As another poster > commented, if we use a one centre complete basis the method puts all the > charge on that atom. Long ago there was a set of one-centre expansion > calculations on simple systems such as methane. There are no basis > functions on the hydrogen atoms. The Mulliken charges are C(4-) and > H(+). If we used basis functions centered only on the H atoms we would > get C(4+) and H(-). Mulliken charges do not have a basis set limit. Too > often we use methods to interpret wave functions that do not have a > basis set limit. I suggest we stop doing that and use methods that do, > just as we have energies that do. We often extrapolate to that energy > limit and we should extrapolate to limits for other properties, even > when they are not observable properties such as charges. > > As others have shown many other methods can reduce the sensitivity of > NPA charges to basis-set, but they still do not properly have a basis > set limit. > > Bader charges depend on the basis set only in the sense that the density > depends on the basis set. They do have a proper basis set limit. They > are obtained by defining the boundaries of each atom and then > integrating over the atoms. The AIM method of getting those boundaries > seems soundly based. Nobody, I think, has come up with a better method. > So why are Bader charges thought to be so bad and unacceptable. Has that > question been properly considered and analysed? > > Brian Duke. > > On Tue, Feb 11, 2014 at 10:09:02AM -0500, Tian Lu sobereva-x-sina.com wrote: >> >> Sent to CCL by: "Tian Lu" [sobereva/./sina.com] >> Hi, >> >> AFAIK, there is only one public program can realize Hirshfeld-I, namely = HiPart (http://molmod.ugent.be/software/). > >> Another modified Hirshfeld-based method alternative to Hirshfeld-I is >> atomic dipole moment corrected Hirshfeld (ADCH) population method, >> which is the one I highly recommend, see J. Theor. Comp. Chem., 2012, >> 11: 163-183. ADCH charges have much better reproducibility for >> electrostatic potential than Hirshfeld charges, and the molecular >> dipole moment can be even exactly reproduced. ADCH has been >> implemented in Multiwfn program (http://Multiwfn.codeplex.com, see >> Section 4.7.2 of its manual for example). > >> NPA charges (also known as NBO charges) are also nice choice. In fact >> they are not explicitly dependent on but only indirectly dependent on >> the basis-set, because the original basis functions will be first >> transformed to natural atomic orbitals before performing natural >> population analysis, this step conspicuously reduces the sensitivity >> of NPA charges to basis-set. According to my experiences, the relative >> sensitivity to basis-set is Mulliken>=3DLowdin>>NPAAIMcharges by fitting >> ESP (MK,CHELPG,etc.) >=3D HirshfeldADCH. > >> Personally I don't recommend using AIM charges, since calculating AIM >> charges is usually time-consuming, and their reproducibility for >> observable quantities are quite poor. > >> A comprehensive comparison of atomic charges can be found in Acta >> Phys.-Chim. Sinica, 2011, 28: 1-18 >> (http://www.whxb.pku.edu.cn/EN/abstract/abstract27818.shtml) > >> >> Tian Lu >> >> >> >> >> ----- Original Message ----- >>> From: "Susi Lehtola susi.lehtola]![alumni.helsinki.fi" > >> To: "Lu, Tian " > >> Subject: CCL:G: How to consider charge on particular atom -lithium >> Date: 2014-02-11 17:12 >> >> >> >> Sent to CCL by: Susi Lehtola [susi.lehtola-.-alumni.helsinki.fi] >> On Mon, 10 Feb 2014 18:37:44 -0500 >> "Jim Kress ccl_nospam_._kressworks.com" > wrote: >>> Sent to CCL by: "Jim Kress" [ccl_nospam,kressworks.com] >>> AIM and NBO charges would be the best choice. Mulliken and Lowdin are f= ar too basis set dependent. >> NBO charges are explicitly dependent on the basis set, like Mulliken >> and Lwdin. >> AIM charges are not, but then again according to them e.g. H2O and HCN >> are ionic... >> Probably the best scheme is something like the recently proposed >> iterative Hirshfeld schemes, but I'm not aware of any program that >> implements these. >> Another possibility are electrostatic potential charges, which are >> available in e.g. Gaussian. >> -- >> --------------------------------------------------------------- >> Mr. Susi Lehtola, PhD Research Associate >> susi.lehtola%a%alumni.helsinki.fi Department = of Applied Physics >> http://www.helsinki.fi/~jzlehtol Aalto University >> Finland >> --------------------------------------------------------------- >> Susi Lehtola, FT Tutkijatohtori >> susi.lehtola%a%alumni.helsinki.fi Fysiikan la= itos >> http://www.helsinki.fi/~jzlehtol Aalto-yliopisto> > > -- > 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> > -- Christopher J. Cramer Elmore H. Northey Professor and Associate Dean for Academic Affairs University of Minnesota Department of Chemistry and College of Science & Engineering Minneapolis, MN 55455-0431 Phone: (612) 624-0859 (Chemistry) Phone: (612) 624-9371 (CSE) -------------------------- Mobile: (952) 297-2575 Email: cramer##umn.edu Twitter: ##ChemProfCramer Website: http://pollux.chem.umn.edu -=3D This is automatically added to each message by the mailing script =3D-=
or use= :E-mail to administrators: CHEMISTRY-REQUEST::ccl.net or usehttp://www.ccl.net/chemistry/sub_unsub.shtml --_000_D5BC00C0FB9AC34D9C8F4DC965C51689479F048FSRVUNIMBX07unia_ Content-Type: text/html; charset="us-ascii" Content-Transfer-Encoding: quoted-printable

Can’t resist to add= my 0.02$

 <= /p>

As Steven points out, the= re is no QM operator for the atomic charge, and therefore no unique choice.= The problem is not the QM operator for charge, that is just fine, it is the ‘atomic’ that causes the problem. In a mo= lecule, where does one ‘atom’ stop and the next begin?

 <= /p>

The Bader (QTAIM) definit= ion in terms of zero-flux surfaces is appealing, but beside the sensitivity= towards small changes in the system that Chris mentions, there are also the troublesome systems where the zero-flux partitioning of= the 3D space leaves a volume of electron density without a nucleus. And th= e reproduction of the molecular moment to order N requires that one include= s all the atomic multipole moments to order N. The experimental dipole moment of CO, for example is 0.12 D, w= hich in the QTAIM partitioning is reproduced by a large monopole (atomic ch= arges) component of -6.54 D and an atomic dipole component of +6.66 D.<= o:p>

 <= /p>

NPA charges are defined f= rom the basis set, but since they effectively are projected onto a minimum = basis set, any sensible choice of basis sets will display a smooth convergence towards a basis set limiting value. But they cannot b= e derived from experimental electron densities, like QTAIM.

 <= /p>

Mulliken and Lowdin are j= ust so terrible sensitive to the present of even slightly diffuse basis fun= ctions that they never should be used (but will nevertheless be used in 90+% of the publications appearing….)

 <= /p>

Hirshfeld charges are als= o appealing, but the original method depends on the definition of the refer= ence atoms. The iterative version alleviates this problem, but for strongly ionic system requires interpolation between singly and do= ubly negative charged atoms, and no doubly negative ions are stable in the = gas phase…..

 <= /p>

The best representation o= f the total electrostatic potential by atomic charges are obtained by expli= cit fitting, but as many studies have shown, this is a mathematically ill-defined problem, as the parameter space defined by the atomic charges = is near-redundant for all by the simplest systems. This leads to conformati= onal and system-sensitive atomic charges, which only to a certain degree ca= n be alleviated by introducing constraints.

 <= /p>

Other less used atomic ch= arge definitions have similar pro’s and con’s. With Stevens wor= ds: Choose your poison.

 <= /p>

Frank

 <= /p>

Frank Jensen

Assoc. Prof.

Dept. of Chemistry

Aarhus Univer= sity

http://old.chem.au.dk/~frj

 

From: owner-ch= emistry+frj=3D=3Dchem.au.dk(!)ccl.net [mailto:owner-chemistry+frj=3D= =3Dchem.au.dk(!)ccl.net] On Behalf Of Steven Bachrach sbachrac_._trinity.edu
Sent: 12. februar 2014 16:09
To: Frank Jensen
Subject: CCL:G: How to consider charge on particular atom -lithium

 

Chris is of course co= rrect in his assessment of Bader charges.

But I would like to a= dd just a few observations. First off, there is no QM operator that corresp= onds to atomic charge - meaning there is no non-arbitrary method for obtain= ing the charge on an atom within a molecule. So it is pointless to argue about which method is "best". Rather= , one should ask which is most appropriate for the question at hand.

Bader charges are gre= at if what you want to do is match up with the experimental electron distri= bution one gets from an x-ray diffraction experiment. Here one gets a 3-D m= ap and one can partition the density according to the zero-flux surface and directly compare the experimental w= ith computed "charge".

If your desire to to = mimic the electrostatic potential, then Chris is correct that the Bader cha= rges are large and are compensated by higher order moments. So if you want = to mimic the electrostatic potential with just monopoles, then Bader charges are a poor choice. But, if you all= ow for dipoles, quadrupoles, and higher moments, then the values one gets f= rom AIM will be just fine.

Keep in mind that ult= imately one is making an arbitrary selection here, and both positive and ne= gative characteristics pervade each choice.

Choose your poison ju= diciously.

Steven

 

On Tue, Feb 11, 2014 at 5:31 PM, Christopher Cramer = cramer%x%umn.edu <owner-chemistry::ccl.net> wro= te:


Sent to CCL by: Christopher Cramer [cramer**umn.edu]
Brian,

   Sure, I can take a run at your final question. The “prob= lem” with Bader charges is not that there is any lack of elegance in = the definition of the atomic basins, nor that they have a complete basis se= t limit (surely a good thing!) The problem is that we usually want partial atomic charges to do the best job possible of represe= nting the molecular charge distribution with a set of atom-centered monopol= es, but there may be rather large HIGHER moments of the charge within the B= ader atomic basins (i.e., beyond the difference in total number of electrons integrated over the Bader atom= subtracted from the nuclear charge). When that happens, collapsing all tho= se electrons entirely on the nuclear position, about which they are not sym= metrically disposed, can lead to rather bizarre charges.

   A good example that I recall from many years ago was work of R= ainer Glaser at Missouri. He showed that diazonium cations (i.e., RN2+ = structures) would show enormous variation in the two different N charges wi= th seemingly tiny variations in R (e.g., going from methyl to ethyl). The problem is that LOTS of the electronic charge i= s built up about midway between the two N atoms, but the zero-flux surface = is VERY sensitive to the substituent. So, a small movement of the zero-flux= surface to one side or the other suddenly moves a sizable fraction of an electron from one N “atom= 221; to the other. Of course, the ACTUAL charge distribution is actually on= ly very slightly perturbed, and if one were to consider the atomic dipole m= oments within the basins, one would see that their variations largely cancel the changes in the monopoles, but we aren&= #8217;t usually interested in using higher atomic moments when we’re = looking for partial atomic charges — we’re hoping for a simpler= representation.

   So much is written on partial atomic charges that I won’= t indulge my otherwise dangerous proclivity to lecture, but I thought IR= 17;d at least address your question of why Bader charges tend to find littl= e use in communities not interested in considering higher moments of the atomic basin charge distributions as well.

Best,

Chris

On Feb 11, 2014, at 14:49, Salter-Duke, Brian James  brian.james.duke#= #gmail.com <owner-che= mistry##ccl.net> wrote:=

>
> Sent to CCL by: "Salter-Duke, Brian James " [brian.james.duke-x-gmail= .com]

> I do not want to= address how the various methods work in practice, as I
> do not have the experience. I do however want to make a general point<= br> > and then ask a question.
>
> Mulliken charges are not basis set independent as they depend on the > basis functions we happen to use on each atom. As another poster
> commented, if we use a one centre complete basis the method puts all t= he
> charge on that atom. Long ago there was a set of one-centre expansion<= br> > calculations on simple systems such as methane. There are no basis
> functions on the hydrogen atoms. The Mulliken charges are C(4-) and > H(+). If we used basis functions centered only on the H atoms we w= ould
> get C(4+) and H(-). Mulliken charges do not have a basis set limit= . Too
> often we use methods to interpret wave functions that do not have a > basis set limit. I suggest we stop doing that and use methods that do,=
> just as we have energies that do. We often extrapolate to that energy<= br> > limit and we should extrapolate to limits for other properties, even > when they are not observable properties such as charges.
>
> As others have shown many other methods can reduce the sensitivity of<= br> > NPA charges to basis-set, but they still do not properly have a basis<= br> > set limit.
>
> Bader charges depend on the basis set only in the sense that the densi= ty
> depends on the basis set. They do have a proper basis set limit. They<= br> > are obtained by defining the boundaries of each atom and then
> integrating over the atoms. The AIM method of getting those boundaries=
> seems soundly based. Nobody, I think, has come up with a better method= .
> So why are Bader charges thought to be so bad and unacceptable. Has th= at
> question been properly considered and analysed?
>
> Brian Duke.
>
> On Tue, Feb 11, 2014 at 10:09:02AM -0500, Tian Lu sobereva-x-sina.com wrote:
>>
>> Sent to CCL by: "Tian  Lu" [sobereva/./sina.com]
>> Hi,
>>
>> AFAIK, there is only one public program can realize Hirshfeld-I, n= amely HiPart (http://molmod.ugent.be/software/).
>
>> Another modified Hirshfeld-based method alternative to Hirshfeld-I= is
>> atomic dipole moment corrected Hirshfeld (ADCH) population method,=
>> which is the one I highly recommend, see J. Theor. Comp. Chem., 20= 12,
>> 11: 163-183. ADCH charges have much better reproducibility for
>> electrostatic potential than Hirshfeld charges, and the molecular<= br> >> dipole moment can be even exactly reproduced. ADCH has been
>> implemented in Multiwfn program (http://Multiwfn.codeplex.com, see
>> Section 4.7.2 of its manual for example).
>
>> NPA charges (also known as NBO charges) are also nice choice. In f= act
>> they are not explicitly dependent on but only indirectly dependent= on
>> the basis-set, because the original basis functions will be first<= br> >> transformed to natural atomic orbitals before performing natural >> population analysis, this step conspicuously reduces the sensitivi= ty
>> of NPA charges to basis-set. According to my experiences, the rela= tive
>> sensitivity to basis-set is Mulliken>=3DLowdin>>NPAAIMcha= rges by fitting
>> ESP (MK,CHELPG,etc.) >=3D HirshfeldADCH.
>
>> Personally I don't recommend using AIM charges, since calculating = AIM
>> charges is usually time-consuming, and their reproducibility for >> observable quantities are quite poor.
>
>> A comprehensive comparison of atomic charges can be found in Acta<= br> >> Phys.-Chim. Sinica, 2011, 28: 1-18
>> (http://www.whxb.pku.edu.cn/EN/abstract/abstract2781= 8.shtml)
>
>>
>> Tian Lu
>>
>>
>>
>>
>> ----- Original Message -----
>>> From: "Susi Lehtola susi.lehtola]![alumni.helsinki.fi" <owner-ch= emistry+/-ccl.net><= br> >> To: "Lu, Tian " <sobereva+/-sina.com>
>> Subject: CCL:G: How to consider charge on particular atom -lithium=
>> Date: 2014-02-11 17:12
>>
>>
>>
>> Sent to CCL by: Susi Lehtola [susi.lehtola-.-alumni.helsinki.fi]
>> On Mon, 10 Feb 2014 18:37:44 -0500
>> "Jim Kress ccl_nospam_._kressworks.com" <owner-chemistry%a%ccl.net> wrote:
>>> Sent to CCL by: "Jim Kress" [ccl_nospam,kressworks.com]
>>> AIM and NBO charges would be the best choice. Mulliken and Low= din are far too basis set dependent.
>> NBO charges are explicitly dependent on the basis set, like Mullik= en
>> and Lwdin.
>> AIM charges are not, but then again according to them e.g. H2O and= HCN
>> are ionic...
>> Probably the best scheme is something like the recently proposed >> iterative Hirshfeld schemes, but I'm not aware of any program that=
>> implements these.
>> Another possibility are electrostatic potential charges, which are=
>> available in e.g. Gaussian.
>> --
>> --------------------------------------------------------------- >> Mr. Susi Lehtola, PhD Research Associate
>> susi.lehtola%a%alumni.helsinki.fi Department of Applied Physics
>> htt= p://www.helsinki.fi/~jzlehtol Aalto University
>> Finland
>> --------------------------------------------------------------- >> Susi Lehtola, FT Tutkijatohtori
>> susi.lehtola%a%alumni.helsinki.fi Fysiikan laitos
>> htt= p://www.helsinki.fi/~jzlehtol Aalto-yliopisto>
>
> --
>   Brian Salter-Duke (Brian Duke)   Brian.Salter-Duke|monash.edu
>                    A= djunct Associate Professor
>            Monash Institute of Pharmaceu= tical Sciences
>      Monash University Parkville Campus, VIC 3052, Aust= ralia>
>

--
Christopher J. Cramer
Elmore H. Northey Professor and
  Associate Dean for Academic Affairs
University of Minnesota
Department of Chemistry and
  College of Science & Engineering
Minneapolis, MN 55455-0431
Phone:  (612) 624-0859 (Chemi= stry)
Phone:  (612) 624-9371 (CSE)<= br> --------------------------
Mobile: (952) 297-2575<= /p>

Email:  cramer##umn.edu
Twitter:  ##ChemProfCramer
Website:  htt= p://pollux.chem.umn.edu




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


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

E-mail to administrators: CHE= MISTRY-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.n= et/jobs
Conferences: http://server.ccl.net/chemistry/announcements/conferences/

Search Messages: http://www.ccl.net/chemistry/searchccl/index.shtml
<br
      http://www.ccl.net/spammers.txt

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




--
Steven Bachrach
Semmes Distinguished Professor and Assistant VP for Special Projects
Department of Chemistry
Trinity University                 =          Phone: 210-999-7379
1 Trinity Place                 &nb= sp;            Fax: 210-999-7569
San Antonio, TX 78212               &nbs= p; email: sbachrach::trinity.edu

--_000_D5BC00C0FB9AC34D9C8F4DC965C51689479F048FSRVUNIMBX07unia_-- From owner-chemistry@ccl.net Wed Feb 12 17:05:00 2014 From: "N. Sukumar nagams::rpi.edu" To: CCL Subject: CCL:G: How to consider charge on particular atom -lithium Message-Id: <-49677-140212163352-5391-gG2tyRiavE0YdQF3M/CbWw^^server.ccl.net> X-Original-From: "N. Sukumar" Content-Disposition: inline Content-Transfer-Encoding: binary Content-Type: text/plain Date: Wed, 12 Feb 2014 16:34:19 -0500 MIME-Version: 1.0 Sent to CCL by: "N. Sukumar" [nagams]=[rpi.edu] "why are Bader charges thought to be so bad and unacceptable?" Bader charges are thought to be so "bad and unacceptable" because people are so used to thinking in terms of point charges or classical ball-and-stick type spherical "atoms". Bader's atoms-in-molecules are, in general, very far from spherical. Thus the electron population in an AIM or the atomic monopole moment (nuclear charge minus the electron population) will not, on its own, reproduce the electrostatic potential or any other physical property. However with inclusion of higher multipole moments the Bader atoms DO reproduce the electrostatic potentials and other physical effects. Whitehead, et al “Transferable Atom Equivalent Multi-Centered Multipole Expansion Method” J. Comp. Chem. 24, 512-529 (2003) N. Sukumar Professor of Chemistry Shiv Nadar University, India ---------------------------- "Pursue something so important that even if you fail, the world is better off with you having tried." -- Tim O'Reilly http://as.wiley.com/WileyCDA/WileyTitle/productCd-0470769009.html ==============Original message text=============== On Tue, 11 Feb 2014 15:49:29 EST "Salter-Duke, Brian James brian.james.duke##gmail.com" wrote: Sent to CCL by: "Salter-Duke, Brian James " [brian.james.duke-x-gmail.com] I do not want to address how the various methods work in practice, as I do not have the experience. I do however want to make a general point and then ask a question. Mulliken charges are not basis set independent as they depend on the basis functions we happen to use on each atom. As another poster commented, if we use a one centre complete basis the method puts all the charge on that atom. Long ago there was a set of one-centre expansion calculations on simple systems such as methane. There are no basis functions on the hydrogen atoms. The Mulliken charges are C(4-) and H(+). If we used basis functions centered only on the H atoms we would get C(4+) and H(-). Mulliken charges do not have a basis set limit. Too often we use methods to interpret wave functions that do not have a basis set limit. I suggest we stop doing that and use methods that do, just as we have energies that do. We often extrapolate to that energy limit and we should extrapolate to limits for other properties, even when they are not observable properties such as charges. As others have shown many other methods can reduce the sensitivity of NPA charges to basis-set, but they still do not properly have a basis set limit. Bader charges depend on the basis set only in the sense that the density depends on the basis set. They do have a proper basis set limit. They are obtained by defining the boundaries of each atom and then integrating over the atoms. The AIM method of getting those boundaries seems soundly based. Nobody, I think, has come up with a better method. So why are Bader charges thought to be so bad and unacceptable. Has that question been properly considered and analysed? Brian Duke. On Tue, Feb 11, 2014 at 10:09:02AM -0500, Tian Lu sobereva-x-sina.com wrote: > > Sent to CCL by: "Tian Lu" [sobereva/./sina.com] > Hi, > > AFAIK, there is only one public program can realize Hirshfeld-I, namely HiPart (http://molmod.ugent.be/software/). > Another modified Hirshfeld-based method alternative to Hirshfeld-I is > atomic dipole moment corrected Hirshfeld (ADCH) population method, > which is the one I highly recommend, see J. Theor. Comp. Chem., 2012, > 11: 163-183. ADCH charges have much better reproducibility for > electrostatic potential than Hirshfeld charges, and the molecular > dipole moment can be even exactly reproduced. ADCH has been > implemented in Multiwfn program (http://Multiwfn.codeplex.com, see> Section 4.7.2 of its manual for example). > NPA charges (also known as NBO charges) are also nice choice. In fact > they are not explicitly dependent on but only indirectly dependent on > the basis-set, because the original basis functions will be first > transformed to natural atomic orbitals before performing natural > population analysis, this step conspicuously reduces the sensitivity > of NPA charges to basis-set. According to my experiences, the relative > sensitivity to basis-set is Mulliken>=Lowdin>>NPAAIMcharges by fitting > ESP (MK,CHELPG,etc.) >= HirshfeldADCH. > Personally I don't recommend using AIM charges, since calculating AIM > charges is usually time-consuming, and their reproducibility for > observable quantities are quite poor. > A comprehensive comparison of atomic charges can be found in Acta > Phys.-Chim. Sinica, 2011, 28: 1-18 > (http://www.whxb.pku.edu.cn/EN/abstract/abstract27818.shtml) > > Tian Lu > > > > > ----- Original Message ----- > > From: "Susi Lehtola susi.lehtola]![alumni.helsinki.fi" > To: "Lu, Tian " > Subject: CCL:G: How to consider charge on particular atom -lithium > Date: 2014-02-11 17:12 > > > > Sent to CCL by: Susi Lehtola [susi.lehtola-.-alumni.helsinki.fi] > On Mon, 10 Feb 2014 18:37:44 -0500 > "Jim Kress ccl_nospam_._kressworks.com" wrote: > > Sent to CCL by: "Jim Kress" [ccl_nospam,kressworks.com] > > AIM and NBO charges would be the best choice. Mulliken and Lowdin are far too basis set dependent. > NBO charges are explicitly dependent on the basis set, like Mulliken > and Lwdin. > AIM charges are not, but then again according to them e.g. H2O and HCN > are ionic... > Probably the best scheme is something like the recently proposed > iterative Hirshfeld schemes, but I'm not aware of any program that > implements these. > Another possibility are electrostatic potential charges, which are > available in e.g. Gaussian. > -- > --------------------------------------------------------------- > Mr. Susi Lehtola, PhD Research Associate > susi.lehtola%a%alumni.helsinki.fi Department of Applied Physics > http://www.helsinki.fi/~jzlehtol Aalto University> Finland > --------------------------------------------------------------- > Susi Lehtola, FT Tutkijatohtori > susi.lehtola%a%alumni.helsinki.fi Fysiikan laitos > http://www.helsinki.fi/~jzlehtol Aalto-yliopisto> -- Brian Salter-Duke (Brian Duke) Brian.Salter-Duke|monash.edu Adjunct Associate Professor Monash Institute of Pharmaceutical Sciences Monash University Parkville Campus, VIC 3052, Australiahttp://www.ccl.net/cgi-bin/ccl/send_ccl_messagehttp://www.ccl.net/chemistry/sub_unsub.shtmlhttp://www.ccl.net/spammers.txt===========End of original message text=========== From owner-chemistry@ccl.net Wed Feb 12 18:01:00 2014 From: "Salter-Duke, Brian James - brian.james.duke/./gmail.com" To: CCL Subject: CCL:G: Printing of two electron integrals. Message-Id: <-49678-140212135550-6177-DwUmSDH/Sqdl0aA8CEWdxg*|*server.ccl.net> X-Original-From: "Salter-Duke, Brian James -" Content-Disposition: inline Content-Type: text/plain; charset=us-ascii Date: Thu, 13 Feb 2014 05:55:38 +1100 MIME-Version: 1.0 Sent to CCL by: "Salter-Duke, Brian James -" [brian.james.duke#gmail.com] On Wed, Feb 12, 2014 at 06:17:04AM -0500, Varun Kundi chemvarun]![gmail.com wrote: > > Sent to CCL by: "Varun Kundi" [chemvarun(~)gmail.com] > Hello, > I have printed 2e integrals using #N HF/STO-3G SCF=Conventional > IOP(3/33=6) EXTRALINKS=l316 NORAFF in Gaussian. How to read 2e integral > values? What is the problem? Can you not look at the output, understand the format of the integrals and then write code to read them? Maybe I misundertand what you are doing. However, if you have Gaussian09 Version D1 or later, then there is a facility to output an integral file. That, together with the *chk file transformed to *.fchk, can be used to get all information from Gaussian for another program to use. This is fairly recent however. Brian.> -- 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 Feb 12 20:39:01 2014 From: "Salter-Duke, Brian James - brian.james.duke^^gmail.com" To: CCL Subject: CCL:G: How to consider charge on particular atom -lithium Message-Id: <-49679-140212174909-9448-76LhEpaB9XqRzp7YFwl1Gw||server.ccl.net> X-Original-From: "Salter-Duke, Brian James -" Content-Disposition: inline Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset=windows-1252 Date: Thu, 13 Feb 2014 09:48:57 +1100 MIME-Version: 1.0 Sent to CCL by: "Salter-Duke, Brian James -" [brian.james.duke . gmail.com] On Tue, Feb 11, 2014 at 05:31:22PM -0600, Christopher Cramer cramer%x%umn.edu wrote: > > Sent to CCL by: Christopher Cramer [cramer**umn.edu] > Brian, > Sure, I can take a run at your final question. The “problem” with > Bader charges is not that there is any lack of elegance in the > definition of the atomic basins, nor that they have a complete > basis set limit (surely a good thing!) The problem is that we > usually want partial atomic charges to do the best job possible of > representing the molecular charge distribution with a set of > atom-centered monopoles, but there may be rather large HIGHER > moments of the charge within the Bader atomic basins (i.e., beyond > the difference in total number of electrons integrated over the > Bader atom subtracted from the nuclear charge). When that happens, > collapsing all those electrons entirely on the nuclear position, > about which they are not symmetrically disposed, can lead to rather > bizarre charges. Thanks, Chris, for a masterfull explanation. So the Bader charges are OK, but the problems arise when we use them. The first point of using them is collapsing the charge to the nuclear position. I have been getting insight about valence bond orbitals by calculating their centroid of charge which tells me much about how an "atomic" VB orbitals moves away from the nucleus (newish code in VB2000 which is now fully incorporated into GAMESS(US) - end of ad.). The centroid of charge of the Bader atom could be evaluated by a messy integration over the atomic basis using the dipole moment integrals. We would then have the electron "charges" on one set of points and the nuclear "charges" on another set of points. Would this work? Would it improve matters? has it already been tried? Cheers, Brian. > A good example that I recall from many years ago was work of Rainer > Glaser at Missouri. He showed that diazonium cations (i.e., RN2+ > structures) would show enormous variation in the two different N > charges with seemingly tiny variations in R (e.g., going from > methyl to ethyl). The problem is that LOTS of the electronic charge > is built up about midway between the two N atoms, but the zero-flux > surface is VERY sensitive to the substituent. So, a small movement > of the zero-flux surface to one side or the other suddenly moves a > sizable fraction of an electron from one N “atom” to the other. Of > course, the ACTUAL charge distribution is actually only very > slightly perturbed, and if one were to consider the atomic dipole > moments within the basins, one would see that their variations > largely cancel the changes in the monopoles, but we aren’t usually > interested in using higher atomic moments when we’re looking for > partial atomic charges — we’re hoping for a simpler representation. > So much is written on partial atomic charges that I won’t indulge > my otherwise dangerous proclivity to lecture, but I thought I’d at > least address your question of why Bader charges tend to find > little use in communities not interested in considering higher > moments of the atomic basin charge distributions as well. > Best, > > Chris > > On Feb 11, 2014, at 14:49, Salter-Duke, Brian James brian.james.duke##gmail.com wrote: > > > > > Sent to CCL by: "Salter-Duke, Brian James " [brian.james.duke-x-gmail.com] > > I do not want to address how the various methods work in practice, as I > > do not have the experience. I do however want to make a general point > > and then ask a question. > > > > Mulliken charges are not basis set independent as they depend on the > > basis functions we happen to use on each atom. As another poster > > commented, if we use a one centre complete basis the method puts all the > > charge on that atom. Long ago there was a set of one-centre expansion > > calculations on simple systems such as methane. There are no basis > > functions on the hydrogen atoms. The Mulliken charges are C(4-) and > > H(+). If we used basis functions centered only on the H atoms we would > > get C(4+) and H(-). Mulliken charges do not have a basis set limit. Too > > often we use methods to interpret wave functions that do not have a > > basis set limit. I suggest we stop doing that and use methods that do, > > just as we have energies that do. We often extrapolate to that energy > > limit and we should extrapolate to limits for other properties, even > > when they are not observable properties such as charges. > > > > As others have shown many other methods can reduce the sensitivity of > > NPA charges to basis-set, but they still do not properly have a basis > > set limit. > > > > Bader charges depend on the basis set only in the sense that the density > > depends on the basis set. They do have a proper basis set limit. They > > are obtained by defining the boundaries of each atom and then > > integrating over the atoms. The AIM method of getting those boundaries > > seems soundly based. Nobody, I think, has come up with a better method. > > So why are Bader charges thought to be so bad and unacceptable. Has that > > question been properly considered and analysed? > > > > Brian Duke. > > > > On Tue, Feb 11, 2014 at 10:09:02AM -0500, Tian Lu sobereva-x-sina.com wrote: > >> > >> Sent to CCL by: "Tian Lu" [sobereva/./sina.com] > >> Hi, > >> > >> AFAIK, there is only one public program can realize Hirshfeld-I, namely HiPart (http://molmod.ugent.be/software/). > > > >> Another modified Hirshfeld-based method alternative to Hirshfeld-I is > >> atomic dipole moment corrected Hirshfeld (ADCH) population method, > >> which is the one I highly recommend, see J. Theor. Comp. Chem., 2012, > >> 11: 163-183. ADCH charges have much better reproducibility for > >> electrostatic potential than Hirshfeld charges, and the molecular > >> dipole moment can be even exactly reproduced. ADCH has been > >> implemented in Multiwfn program (http://Multiwfn.codeplex.com, see > >> Section 4.7.2 of its manual for example). > > > >> NPA charges (also known as NBO charges) are also nice choice. In fact > >> they are not explicitly dependent on but only indirectly dependent on > >> the basis-set, because the original basis functions will be first > >> transformed to natural atomic orbitals before performing natural > >> population analysis, this step conspicuously reduces the sensitivity > >> of NPA charges to basis-set. According to my experiences, the relative > >> sensitivity to basis-set is Mulliken>=Lowdin>>NPAAIMcharges by fitting > >> ESP (MK,CHELPG,etc.) >= HirshfeldADCH. > > > >> Personally I don't recommend using AIM charges, since calculating AIM > >> charges is usually time-consuming, and their reproducibility for > >> observable quantities are quite poor. > > > >> A comprehensive comparison of atomic charges can be found in Acta > >> Phys.-Chim. Sinica, 2011, 28: 1-18 > >> (http://www.whxb.pku.edu.cn/EN/abstract/abstract27818.shtml) > > > >> > >> Tian Lu > >> > >> > >> > >> > >> ----- Original Message ----- > >>> From: "Susi Lehtola susi.lehtola]![alumni.helsinki.fi" > >> To: "Lu, Tian " > >> Subject: CCL:G: How to consider charge on particular atom -lithium > >> Date: 2014-02-11 17:12 > >> > >> > >> > >> Sent to CCL by: Susi Lehtola [susi.lehtola-.-alumni.helsinki.fi] > >> On Mon, 10 Feb 2014 18:37:44 -0500 > >> "Jim Kress ccl_nospam_._kressworks.com" wrote: > >>> Sent to CCL by: "Jim Kress" [ccl_nospam,kressworks.com] > >>> AIM and NBO charges would be the best choice. Mulliken and Lowdin are far too basis set dependent. > >> NBO charges are explicitly dependent on the basis set, like Mulliken > >> and Lwdin. > >> AIM charges are not, but then again according to them e.g. H2O and HCN > >> are ionic... > >> Probably the best scheme is something like the recently proposed > >> iterative Hirshfeld schemes, but I'm not aware of any program that > >> implements these. > >> Another possibility are electrostatic potential charges, which are > >> available in e.g. Gaussian. > >> -- > >> --------------------------------------------------------------- > >> Mr. Susi Lehtola, PhD Research Associate > >> susi.lehtola%a%alumni.helsinki.fi Department of Applied Physics > >> http://www.helsinki.fi/~jzlehtol Aalto University > >> Finland > >> --------------------------------------------------------------- > >> Susi Lehtola, FT Tutkijatohtori > >> susi.lehtola%a%alumni.helsinki.fi Fysiikan laitos > >> http://www.helsinki.fi/~jzlehtol Aalto-yliopisto> > > > > -- > > 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> > > > > -- > Christopher J. Cramer > Elmore H. Northey Professor and > Associate Dean for Academic Affairs > University of Minnesota > Department of Chemistry and > College of Science & Engineering > Minneapolis, MN 55455-0431 > Phone: (612) 624-0859 (Chemistry) > Phone: (612) 624-9371 (CSE) > -------------------------- > Mobile: (952) 297-2575 > Email: cramer##umn.edu > Twitter: ##ChemProfCramer > Website: http://pollux.chem.umn.edu> -- Brian Salter-Duke (Brian Duke) Brian.Salter-Duke|a|monash.edu Adjunct Associate Professor Monash Institute of Pharmaceutical Sciences Monash University Parkville Campus, VIC 3052, Australia