From owner-chemistry@ccl.net Thu Sep 10 06:11:01 2015 From: "Martin Korth martin.korth\a/gmail.com" To: CCL Subject: CCL: Case Studies of QM Computational Chemistry in Reactivity Message-Id: <-51688-150910060905-8043-7JRjLReO2vw3qDG3t5Oaqw*o*server.ccl.net> X-Original-From: Martin Korth Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset=windows-1252 Date: Thu, 10 Sep 2015 12:08:57 +0200 MIME-Version: 1.0 Sent to CCL by: Martin Korth [martin.korth_-_gmail.com] I completey agree; not only 'level nerds', but also 'impact geeks' (like me nowadays) should try to be informed about the state of the art - there is simply no good argument for being lazy about methods ... Am 10.09.2015 um 03:31 schrieb Susi Lehtola susi.lehtola ~~ alumni.helsinki.fi: > > Sent to CCL by: Susi Lehtola [susi.lehtola%a%alumni.helsinki.fi] > On 09/09/2015 10:28 AM, Tom Albright talbright1234]=[gmail.com wrote: >> I wholeheartedly agree with you Victor. Too often I see "how high can >> I go" rather than a coherent explanation of why I got the results >> that I did and can I extrapolate this "understanding" to other >> examples. And no this is not the '90s. > > That's simply not true. If you use a crappy method (like 6-31G*/B3LYP > really is), then you are in no position to make any claims on why you > get the results you get, because you're relying on fortuitous error > cancellation between the method and the basis set. You're so far from > the basis set limit that you can't make real quantitative judgements > whatever the method is you're using. > > Yes, it's not the 90s, but too many people act like if it were. > Computational methods have progressed a whole lot since those days, > but still way too many people stick to old habits without so much as > an afterthought. From owner-chemistry@ccl.net Thu Sep 10 07:07:00 2015 From: "Peter Jarowski peterjarowski===gmail.com" To: CCL Subject: CCL: Case Studies of QM Computational Chemistry in Reactivity Message-Id: <-51689-150910022646-8634-vIWXtALq11HKvK9kGOehSg]_[server.ccl.net> X-Original-From: Peter Jarowski Content-Type: multipart/alternative; boundary=001a1141f37ea8ba6f051f5eae29 Date: Thu, 10 Sep 2015 07:26:41 +0100 MIME-Version: 1.0 Sent to CCL by: Peter Jarowski [peterjarowski-,-gmail.com] --001a1141f37ea8ba6f051f5eae29 Content-Type: text/plain; charset=UTF-8 Dear All, Typically I see this discussion changing direction to a topic that is familiar and comfortable for theorist. However, the original question was very uncomfortable for many. Are we as theorist to avoid the most basic question that needs to be answered in order to justify our work? I hear all the time about the large numbers of graduates from comp chem and their problems getting academic and industrial jobs. Isn't this partly explained by our unwillingness to justify our work with hard metrics vis-a-via experiment. I ask a simple question and further ask everyone to pretend they are at an interview outside of an academic environment where you will find willing and interested parties who enjoy the alphabet soup of DFT. what example and what statistics would you present to justify your work. Clearly, you are talking to experimentalists, as theory does not exist without them. How does theory drive experiment? How much money does it save? How much revenue can it produce? How reliable is it? How essential is it and what can it do that experiment can not? If not, please go back to an academic hole where you will find people who will pat your back for adding an extra letter to some EC functional. I am an academic, I have used lots of functionals and methods. I am a friend of theory, but I am also pragmatic and I want to be able to make hard arguments that are easily translatable to others for the value of theory. Best to all, Peter On Thursday, September 10, 2015, Susi Lehtola susi.lehtola ~~ alumni.helsinki.fi wrote: > > Sent to CCL by: Susi Lehtola [susi.lehtola%a%alumni.helsinki.fi] > On 09/09/2015 10:28 AM, Tom Albright talbright1234]=[gmail.com wrote: > >> I wholeheartedly agree with you Victor. Too often I see "how high can I >> go" rather than a coherent explanation of why I got the results that I did >> and can I extrapolate this "understanding" to other examples. And no this >> is not the '90s. >> > > That's simply not true. If you use a crappy method (like 6-31G*/B3LYP > really is), then you are in no position to make any claims on why you get > the results you get, because you're relying on fortuitous error > cancellation between the method and the basis set. You're so far from the > basis set limit that you can't make real quantitative judgements whatever > the method is you're using. > > Yes, it's not the 90s, but too many people act like if it were. > Computational methods have progressed a whole lot since those days, but > still way too many people stick to old habits without so much as an > afterthought. > -- > ----------------------------------------------------------------------- > Mr. Susi Lehtola, PhD Chemist Postdoctoral Fellow > susi.lehtola]![alumni.helsinki.fi Lawrence Berkeley National Laboratory > http://www.helsinki.fi/~jzlehtol USA > -----------------------------------------------------------------------http://www.ccl.net/chemistry/sub_unsub.shtmlConferences: > http://server.ccl.net/chemistry/announcements/conferences/> > > --001a1141f37ea8ba6f051f5eae29 Content-Type: text/html; charset=UTF-8 Content-Transfer-Encoding: quoted-printable Dear All,

Typically I see this discussion changing direc= tion to a topic that is familiar and comfortable for theorist. However, the= original question was very uncomfortable for many. Are we as theorist to a= void the most basic question that needs to be answered in order to justify = our work? I hear all the time about the large numbers of graduates from com= p chem and their problems getting academic and industrial jobs. Isn't t= his partly explained by our unwillingness to justify our work with hard met= rics vis-a-via=C2=A0experiment.

I ask a simple que= stion and further ask everyone to pretend they are at an interview outside = of an academic environment where you will find willing and interested parti= es who enjoy the alphabet soup of DFT. what example and what statistics wou= ld you present to justify your work. Clearly, you are talking to experiment= alists, as theory does not exist without them.=C2=A0

How does theory drive experiment?
How much money does it save?=
How much revenue can it produce?
How reliable is it?
How essential is it and what can it do that experiment can not?

If not, please go back to an academic hole where you= will find people who will pat your back for adding an extra letter to some= EC functional.=C2=A0

I am an academic, I have use= d lots of functionals and methods. I am a friend of theory, but I am also p= ragmatic and I want to be able to make hard arguments that are easily trans= latable to others for the value of theory.

Best to= all,

Peter

On Thursday, Septe= mber 10, 2015, Susi Lehtola susi.lehtola ~~ alumni.helsinki.fi <owner-chemistry*_*ccl.net> wrote:
=
Sent to CCL by: Susi Lehtola [susi.lehtola%a%alumni.helsinki.fi]
On 09/09/2015 10:28 AM, Tom Albright talbright1234]=3D[gmail.com wrote:
I wholeheartedly agree with you Victor. Too often I see "how high can = I go" rather than a coherent explanation of why I got the results that= I did and can I extrapolate this "understanding" to other exampl= es. And no this is not the '90s.

That's simply not true. If you use a crappy method (like 6-31G*/B3LYP r= eally is), then you are in no position to make any claims on why you get th= e results you get, because you're relying on fortuitous error cancellat= ion between the method and the basis set. You're so far from the basis = set limit that you can't make real quantitative judgements whatever the= method is you're using.

Yes, it's not the 90s, but too many people act like if it were. Computa= tional methods have progressed a whole lot since those days, but still way = too many people stick to old habits without so much as an afterthought.
--
-----------------------------------------------------------------------
Mr. Susi Lehtola, PhD=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0Chemis= t Postdoctoral Fellow
susi.lehtola]![alum= ni.helsinki.fi=C2=A0 =C2=A0Lawrence Berkeley National Laboratory
http://www.h= elsinki.fi/~jzlehtol=C2=A0 USA
-----------------------------------------------------------------------



-=3D This is automatically added to each message by the mailing script =3D-=
E-mail to subscribers: CHEMISTRY*_*ccl.net or use:
=C2=A0 =C2=A0 =C2=A0http://www.ccl.net/cgi-bin/ccl/send_ccl_message<= br>
E-mail to administrators: CHEMISTRY-REQUEST*_*ccl.net or use
=C2=A0 =C2=A0 =C2=A0http://www.ccl.net/cgi-bin/ccl/send_ccl_message<= br>
Subscribe/Unsubscribe:=C2=A0 =C2=A0 =C2=A0 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/anno= uncements/conferences/

Search Messages: http://www.ccl.net/chemistry/searchccl/index.shtml
=C2=A0 =C2=A0 =C2=A0http://www.ccl.net/spammers.txt

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


--001a1141f37ea8ba6f051f5eae29-- From owner-chemistry@ccl.net Thu Sep 10 07:42:00 2015 From: "Peter Jarowski peterjarowski|*|gmail.com" To: CCL Subject: CCL: Case Studies of QM Computational Chemistry in Reactivity Message-Id: <-51690-150910031216-10633-4s151BQXxUrvrjgJG9iAzw]^[server.ccl.net> X-Original-From: Peter Jarowski Content-Type: multipart/alternative; boundary=001a1141f37e502c9a051f5f51e9 Date: Thu, 10 Sep 2015 09:12:10 +0200 MIME-Version: 1.0 Sent to CCL by: Peter Jarowski [peterjarowski+*+gmail.com] --001a1141f37e502c9a051f5f51e9 Content-Type: text/plain; charset=UTF-8 Hi Lars: There is no need for defense. We all agree that method development is important. Personally, I am a bit defensive of my own work which was published in 2009 in a non-theoretical journal and consist of 80% experiment. I provided that example to show one where theory (whatever it is) worked well to the point of prediction. Again, this is what I am looking for in this thread. Unfortunately, theory can not exist on its own and the finances flow from experiment downward. You will find only blank looks with the arguments you have made outside of a room full of theorists and that is the reality. On the other hand, often we are in front of chemists, at least, and having verified chemical examples of the importance of theory in necessary. These should be chemically understandable and should be dealing with issues that are important to the audience. They should be transferable. That means using standard methods with the right statistics and history that can be computed in a reasonable time and executed, potentially, by non-experts. So, while I love MRCI, for example, even the best of us will have trouble getting it right on relevant molecules. Best, Peter On Wed, Sep 9, 2015 at 4:06 PM, Lars Goerigk lars.goerigk-,-unimelb.edu.au < owner-chemistry**ccl.net> wrote: > > Sent to CCL by: "Lars Goerigk" [lars.goerigk : unimelb.edu.au] > Hi Peter, > > in defense of us people that believe in obtaining the right result for the > right reason: even if industry is interested in quick-and-dirty results, > would it not be highly embarassing to make recommendations based on a > method that relies on unforeseeable error compensation, only to then see > that the experimentalists cannot reproduce your predictions? Susi's comment > was therefore valid, albeit I would have worded it in a more diplomatic way. > > I find it very important to apply a level of theory that comes with a low > risk of unsystematic error compensation. For example, a > dispersion-corrected (double-)hybrid density functional with at least a > triple-zeta basis set is overall much more reliable than the popular > B3LYP/6-31G* level; the reasons for that are clear and they have been > discussed extensively in the literature and in this forum. If you cannot > afford a large basis set, but you want to quickly obtain results for bigger > systems with an acceptable accuracy, there are other promising methods out > there (e.g. HF-3c or PBEh-3c). Moreover, it is also important to use > reliable solvation models (e.g. COSMO-RS) and reliable enthalpy and entropy > corrections. > > As to your comment on "level nerds" being a barrier for successful > interaction between theory and experiment, quantum chemists do not > investigate and develop new methdods just because they do not know what > else to do with their time. I would rather say that this discussion > highlights a lack of communication between method developers and the users > of QM methods and as always the fault for that lies probably on both sides; > one consequence of that is that B3LYP/6-31G* is still so popular in the > year 2015. > > Giving a general answer to your question is difficult, as it is not quite > sure what property you want to calculate, which is exactly why the various > answers to your question differ so much. However, there are numerous > benchmark studies out there that provide some guidance on which methods to > use and which to avoid; I have given some hints on reliable methods above. > > Of course, there is never 100% certainty that your predictions are right > because in the end you try to describe a highly complex system with a > theoretical model, which by definition is only a simplification of the real > world. That complexity also means that there is not one sole, easy solution > to your question. Therefore, it is probably wise to run calculations at two > different levels of theory to distinguish between general trends and > methodological artifacts. By carefully choosing a statistically reliable > level of theory, you can at least minimise the possibilibity of making > wrong predictions, and that is something that also industry people will > accept if properly explained. I hope some of these hints were helpful. > > Finally, I hope you forgive me for saying that having efficient and ever > improving QM methods is something that would not be possible without all > those dedicated level nerds out there, which is why we should regard that > term as a compliment. > > Cheers, > Lars > > --- > Dr. Lars Goerigk > ARC DECRA Fellow > School of Chemistry > The University of Melbourne > VIC 3010 > Australia > > Research profile: > http://www.chemistry.unimelb.edu.au/dr-lars-goerigk > List of my publications: > http://www.researcherid.com/rid/D-3717-2009 > Follow me on Twitter: https://twitter.com/lgoer_compchem> > > --001a1141f37e502c9a051f5f51e9 Content-Type: text/html; charset=UTF-8 Content-Transfer-Encoding: quoted-printable
Hi Lars:

There is no need for = defense. We all agree that method development is important.

Persona= lly, I am a bit defensive of my own work which was published in 2009 in a n= on-theoretical journal and consist of 80% experiment. I provided that examp= le to show one where theory (whatever it is) worked well to the point of pr= ediction. Again, this is what I am looking for in this thread.

Unfo= rtunately, theory can not exist on its own and the finances flow from exper= iment downward. You will find only blank looks with the arguments you have = made outside of a room full of theorists and that is the reality. On the ot= her hand, often we are in front of chemists, at least, and having verified = chemical examples of the importance of theory in necessary. These should be= chemically understandable and should be dealing with issues that are impor= tant to the audience. They should be transferable. That means using standar= d methods with the right statistics and history that can be computed in a r= easonable time and executed, potentially, by non-experts. So, while I love = MRCI, for example, even the best of us will have trouble getting it right o= n relevant molecules.

=C2=A0Best,

Peter
=

On Wed, Sep 9, 20= 15 at 4:06 PM, Lars Goerigk lars.goerigk-,-unimelb.edu.au <owner-chemistry**ccl.net> wrote:<= br>

Sent to CCL by: "Lars=C2=A0 Goerigk" [lars.goerigk : unimelb.edu.au]
Hi Peter,

in defense of us people that believe in obtaining the right result for the = right reason: even if industry is interested in quick-and-dirty results, wo= uld it not be highly embarassing to make recommendations based on a method = that relies on unforeseeable error compensation, only to then see that the = experimentalists cannot reproduce your predictions? Susi's comment was = therefore valid, albeit I would have worded it in a more diplomatic way.
I find it very important to apply a level of theory that comes with a low r= isk of unsystematic error compensation. For example, a dispersion-corrected= (double-)hybrid density functional with at least a triple-zeta basis set i= s overall much more reliable than the popular B3LYP/6-31G* level; the reaso= ns for that are clear and they have been discussed extensively in the liter= ature and in this forum. If you cannot afford a large basis set, but you wa= nt to quickly obtain results for bigger systems with an acceptable accuracy= , there are other promising methods out there (e.g. HF-3c or PBEh-3c). More= over, it is also important to use reliable solvation models (e.g. COSMO-RS)= and reliable enthalpy and entropy corrections.

As to your comment on "level nerds" being a barrier for successfu= l interaction between theory and experiment, quantum chemists do not invest= igate and develop new methdods just because they do not know what else to d= o with their time. I would rather say that this discussion highlights a lac= k of communication between method developers and the users of QM methods an= d as always the fault for that lies probably on both sides; one consequence= of that is that B3LYP/6-31G* is still so popular in the year 2015.

Giving a general answer to your question is difficult, as it is not quite s= ure what property you want to calculate, which is exactly why the various a= nswers to your question differ so much. However, there are numerous benchma= rk studies out there that provide some guidance on which methods to use and= which to avoid; I have given some hints on reliable methods above.

Of course, there is never 100% certainty that your predictions are right be= cause in the end you try to describe a highly complex system with a theoret= ical model, which by definition is only a simplification of the real world.= That complexity also means that there is not one sole, easy solution to yo= ur question. Therefore, it is probably wise to run calculations at two diff= erent levels of theory to distinguish between general trends and methodolog= ical artifacts. By carefully choosing a statistically reliable level of the= ory, you can at least minimise the possibilibity of making wrong prediction= s, and that is something that also industry people will accept if properly = explained. I hope some of these hints were helpful.

Finally, I hope you forgive me for saying that having efficient and ever im= proving QM methods is something that would not be possible without all thos= e dedicated level nerds out there, which is why we should regard that term = as a compliment.

Cheers,
Lars

---
Dr. Lars Goerigk
ARC DECRA Fellow
School of Chemistry
The University of Melbourne
VIC 3010
Australia

Research profile:
http://www.chemistry.unimelb.edu.au/dr-lars-goeri= gk
List of my publications:
http://www.researcherid.com/rid/D-3717-2009
Follow me on Twitter: https://twitter.com/lgoer_compchem



-=3D This is automatically added to each message by the mailing script =3D-=
E-mail to subscribers: CHEMISTRY**ccl.n= et or use:
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/cgi-bin/ccl/s= end_ccl_message

E-mail to administrators: CHEM= ISTRY-REQUEST**ccl.net or use
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/cgi-bin/ccl/s= end_ccl_message

Subscribe/Unsubscribe:
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/chemistry/sub_un= sub.shtml

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

Job: http://www.ccl.net/jobs
Conferences: http://server.ccl.net/chemist= ry/announcements/conferences/

Search Messages: http://www.ccl.net/chemistry/sear= chccl/index.shtml
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/spammers.txt

RTFI: http://www.ccl.net/chemistry/aboutccl/ins= tructions/



--001a1141f37e502c9a051f5f51e9-- From owner-chemistry@ccl.net Thu Sep 10 09:23:01 2015 From: "Joseph Leonard jleonard42__gmail.com" To: CCL Subject: CCL: Case Studies of QM Computational Chemistry in Reactivity Message-Id: <-51691-150910090044-21270-4HyiBXQ1PrwhYvvtq2TXoQ^_^server.ccl.net> X-Original-From: Joseph Leonard Content-Type: text/plain; charset=UTF-8 Date: Thu, 10 Sep 2015 09:00:37 -0400 MIME-Version: 1.0 Sent to CCL by: Joseph Leonard [jleonard42_+_gmail.com] This makes a good point. If computational methods are to be used they must be useful to those who have "project questions": - Can you explain conformational preference? When hypervalent atoms are involved? - Can you estimate pKa/protonation states? - Can you estimate tautomeric forms, most-common topology? - Can you predict experimental values/spectra? - Does your salvation model properly correct for gas-phase artifacts? Predict bulk-solvent properties? - Can you answer to molecular similarity questions (scaffold hopping, etc)? - Can you estimate protein-ligand or protein-protein interaction energies? - Can you (semi-)quantitatively model chemical reactions? In the cleft? - Can you generate consistent missing forcefield parameters? There are other things that theory is useful for, but these came to mind quickly as tasks that would directly help project teams. And, of course, the calculations have to be faster than the corresponding experiment or accurate enough to make synthesizing the compound optional. It also has to be sufficiently better than empirical methods to even be considered as an option given the staggering performance difference. Joe On Thu, Sep 10, 2015 at 2:26 AM, Peter Jarowski peterjarowski===gmail.com wrote: > Dear All, > > Typically I see this discussion changing direction to a topic that is > familiar and comfortable for theorist. However, the original question was > very uncomfortable for many. Are we as theorist to avoid the most basic > question that needs to be answered in order to justify our work? I hear all > the time about the large numbers of graduates from comp chem and their > problems getting academic and industrial jobs. Isn't this partly explained > by our unwillingness to justify our work with hard metrics vis-a-via > experiment. > > I ask a simple question and further ask everyone to pretend they are at an > interview outside of an academic environment where you will find willing and > interested parties who enjoy the alphabet soup of DFT. what example and what > statistics would you present to justify your work. Clearly, you are talking > to experimentalists, as theory does not exist without them. > > How does theory drive experiment? > How much money does it save? > How much revenue can it produce? > How reliable is it? > How essential is it and what can it do that experiment can not? > > If not, please go back to an academic hole where you will find people who > will pat your back for adding an extra letter to some EC functional. > > I am an academic, I have used lots of functionals and methods. I am a friend > of theory, but I am also pragmatic and I want to be able to make hard > arguments that are easily translatable to others for the value of theory. > > Best to all, > > Peter > > On Thursday, September 10, 2015, Susi Lehtola susi.lehtola ~~ > alumni.helsinki.fi wrote: >> >> >> Sent to CCL by: Susi Lehtola [susi.lehtola%a%alumni.helsinki.fi] >> On 09/09/2015 10:28 AM, Tom Albright talbright1234]=[gmail.com wrote: >>> >>> I wholeheartedly agree with you Victor. Too often I see "how high can I >>> go" rather than a coherent explanation of why I got the results that I did >>> and can I extrapolate this "understanding" to other examples. And no this is >>> not the '90s. >> >> >> That's simply not true. If you use a crappy method (like 6-31G*/B3LYP >> really is), then you are in no position to make any claims on why you get >> the results you get, because you're relying on fortuitous error cancellation >> between the method and the basis set. You're so far from the basis set limit >> that you can't make real quantitative judgements whatever the method is >> you're using. >> >> Yes, it's not the 90s, but too many people act like if it were. >> Computational methods have progressed a whole lot since those days, but >> still way too many people stick to old habits without so much as an >> afterthought. >> -- >> ----------------------------------------------------------------------- >> Mr. Susi Lehtola, PhD Chemist Postdoctoral Fellow >> susi.lehtola]![alumni.helsinki.fi Lawrence Berkeley National Laboratory >> http://www.helsinki.fi/~jzlehtol USA >> ----------------------------------------------------------------------->> E-mail to subscribers: CHEMISTRY[]ccl.net or use:>> >> E-mail to administrators: CHEMISTRY-REQUEST[]ccl.net or usehttp://www.ccl.net/chemistry/sub_unsub.shtmlConferences: >> http://server.ccl.net/chemistry/announcements/conferences/>> >> > -- You cannot "freeze dry" alcohol because alcohol is a pure liquid at room temperature and to make it solid you would need a temperature of -78C which is a little on the cold side for anyone not Canadian. From owner-chemistry@ccl.net Thu Sep 10 12:19:01 2015 From: "Andreas Bender, PhD ab454-.-cam.ac.uk" To: CCL Subject: CCL: "Data In Drug Discovery - Time To Get Honest!" - Programme Now Available, 23 Sep 2015, Cambridge/UK Message-Id: <-51692-150910101050-12689-+gEolFXpKDC4Q7iroNOiTg]-[server.ccl.net> X-Original-From: "Andreas Bender, PhD" Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset=UTF-8; format=flowed Date: Thu, 10 Sep 2015 15:10:44 +0100 MIME-Version: 1.0 Sent to CCL by: "Andreas Bender, PhD" [ab454++cam.ac.uk] Dear All, We have now finalized the programme for the "Data In Drug Discovery" symposium to be held on Wednesday, 23 September 2015 in Cambridge/UK, and we are very happy to present speakers from all of the different areas we were planning to cover: From using genomics, proteomics and image-based readouts for characterizing and understanding biology to modelling compound properties, as well as the utilization of those (and other) types of data in discovery in a pharmaceutical setting. Please find the full programme for the day below for your information, which can also be accessed in PDF format here: http://www.c-inf.net/20150923_Programme_DataInDrugDiscovery.pdf . If you are interested in attending the meeting you are very welcome to register (for free) by email to this address - and you are also welcome to forward the invitation to anyone else who might be interested in joining us. Looking forward to seeing you soon! Best wishes, Andreas "Data In Drug Discovery – It’s Time To Get Honest!" 10am – 5.30pm, 23 September 2015 Unilever Lecture Theatre, Centre for Molecular Informatics (CMI), Department of Chemistry, Lensfield Road, University of Cambridge, Cambridge CB2 1EW, United Kingdom Contact and registration: Andreas Bender, ab454 _ cam.ac.uk, 079 – 061 585 33 Setting the Scene 10am Andreas Bender, Department of Chemistry, University of Cambridge, Cambridge / UK "So why are we here? From observations and experience to understanding and insight" 10.15am Matt Segall, Optibrium, Cambridge / UK "The challenges of making decisions using uncertain data" Understanding Gene Expression, Genome-Proteome Relationships, and Cellular Phenotypes 10.45am Paul Schofield, Department of Physiology, University of Cambridge, Cambridge / UK "Exploiting gene expression signatures and the phenotypes of mutant mice to predict drug action" 11.15am Tea (CMI Foyer) 11.45am Christoph Schlaffner, Sanger Institute and Department of Chemistry, Cambridge / UK "Quantitative proteogenomics application to personalised proteomics" 12.15pm Bram Herpers, OcellO, Leiden / NL "Compound screening and profiling in 3D cultured human tissues: How morphometric data can be used to determine compound efficacy and discriminate mode of action" 12.45pm "The limits of data: Digiti-mediated circumflexion of Phaseolus vulgaris and related chemical matter increases ion channel and GPCR-mediated signalling in the oral cavity" (aka "tacos for lunch"; top floor of the CMI, room U202) Genetic and Network Information for Cell Sensitivity Prediction and Target Discovery 2pm Mathew Garnett, Sanger Institute, Hinxton / UK "High throughput chemical and genetic screening data: separating the wheat from the chaff" 2.30pm Tamas Korcsmaros, The Genome Analysis Centre (TGAC) and Institute of Food Research (IFR), Norwich / UK "Challenges and opportunities of network biology approaches to promote drug discovery" Data in Drug Discovery – Opportunities and Challenges 3pm Tim James, Evotec, Abingdon / UK "Annotation bias in phenotypic deconvolution" 3.30pm Coffee (CMI Foyer) 4pm Matt Segall, Optibrium, Cambridge / UK "How confident can we be in ADME predictions?" 4.30pm Aidan MacNamara, GlaxoSmithKline (GSK), Stevenage / UK "Data to therapy at GSK: examples of current data-intensive projects for drug discovery" 5pm William Spooner, Eagle Genomics, Cambridge / UK "The increasing problem of silos in data-driven drug discovery; what can be done?" 5.30pm Symposium close Visit to the Alma (Russell Court, CB2 1HW) From owner-chemistry@ccl.net Thu Sep 10 12:54:01 2015 From: "Joseph T Golab joseph.golab%ineos.com" To: CCL Subject: CCL: Case Studies of QM Computational Chemistry in Reactivity Message-Id: <-51693-150910103036-21200-5c75bLucz3YvDPJO1fT9qA-.-server.ccl.net> X-Original-From: "Joseph T Golab" Date: Thu, 10 Sep 2015 10:30:34 -0400 Sent to CCL by: "Joseph T Golab" [joseph.golab-*-ineos.com] A theorist (or Theoretically a Chemist, as my colleagues refer to me ;) in an industrial environment must strive to present relevant technical information to the decision makers. This is exactly what the experimentalists do. Working with a good experimental person is also essential because no one believes (really) model predictions without at least comparative agreement with "accepted" experiments. Accepted here meaning what the overall group you are working with traditionally believes. As in any area of science, the key to contributing to a group goal with modeling results is knowing what question to have the application answer so that reliable data is produced in a timely manner. You, as the modeler/ theorist, are responsible for the reliability part of the prediction. Correlation with experimental data is crucial as a gauge. Often times, correlating data, especially if done with an experimental colleague, leads to conjectures that can be pursued in a laboratory. With enough give and take between prediction and experiment, the important details of a chemical process are teased out and better understood. I often find that experimentalists are wonderfully creative people with lots of ideas on what to study with their equipment. They welcome the results of modeling to help them pursue those ideas that are more likely to succeed. This allows us to get to the goal line quicker and with less expense. Modeling is an enabling technology; it cannot automate processes nor will it will replace or eliminate (all) experimentation. It is definitely not a stand alone protocol for problem resolution (at least at the chemistry industrial scale!). However, in conjunction with experimentation, it can help prioritize experiments, improve experimental design, and predict experimental participants & results, sometimes ones that are not expected (by chemical engineers -- but that's a whole other story!). Finally, something that shouldn't be discounted (and has already been mentioned in this thread), modeling constantly improves; the questions that "can't" be answered today most assuredly will be answerable within a few years. And more cheaply too. As for industrial modeling examples, several computational chemistry software companies collect these and post 1-2 slide summaries on their web site(s). Obviously, these examples use their software so they are somewhat of an advert for them; however, that should not discount the hard work summarized as an application of "theory" by an industrial chemist to a problem of importance to industry. Joe _____________________________ Joseph T Golab, PhD Modeling & Simulation Scientist INEOS Technologies | www.ineostechnologies.com Naperville | +1 (630) 420-5063 (o) | +1 (630) 336-0063 (m) From owner-chemistry@ccl.net Thu Sep 10 13:29:01 2015 From: "Thomas Manz thomasamanz_-_gmail.com" To: CCL Subject: CCL: Case Studies of QM Computational Chemistry in Reactivity Message-Id: <-51694-150910115912-28218-4Tp7HZ5ilwgcBnFeeMA/3g{}server.ccl.net> X-Original-From: Thomas Manz Content-Type: multipart/alternative; boundary=001a1139644ec9c0a1051f66ad5e Date: Thu, 10 Sep 2015 09:59:06 -0600 MIME-Version: 1.0 Sent to CCL by: Thomas Manz [thomasamanz[#]gmail.com] --001a1139644ec9c0a1051f66ad5e Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: quoted-printable Hi Peeter, There is a fundamental distinction between the current conversation focused on exchange-correlation theories and basis sets and the earlier discussion focused on atomic properties. If one increases the basis set size, exchange-correlation functionals such as B3LYP, M06, or whatever one you care to use will approach a well-defined mathematical limit. We can then discuss what the relative accuracy of that mathematical limit is in comparison to experimental properties and also discuss how close we are to that mathematical limit with a particular basis set. Thus, it is meaningful to discuss how adequate an exchange-correlation theory or basis set are for a particular research problem. Of course, the goal is to choose an adequate level that is not too computationally expensive for the particular research question being studied. In contrast, Mulliken and Lowdin population analysis schemes do not have any defined mathematical limits. As the basis set is increased and the energy and electron density approach the complete basis set limit, the Mulliken and Lowdin populations behave erratically and blow up. This is how we know for sure that Mulliken and Lowdin population analysis schemes are utter nonsense and should never be used for publication results. As pointed out by one person, their only purpose is for debugging calculations to see if the symmetry or other basic features of the input geometry are malformed= . It is not the earlier discussion on atomic charges that is "nonsense" but rather the Mulliken and Lowdin populations that are nonsense, because they have no defined mathematical limits. This has nothing to do with atomic charges, per se. The Mulliken and Lowdin populations do not measure anything physical. They do not measure atomic charges. Probably the confusion has been propagated by calling Mulliken and Lowdin populations as types of "atomic charges", but really the Mulliken and Lowdin populations cannot be atomic charges, because they have no defined mathematical limits. In the future, I shall try to avoid referring to Mulliken and Lowdin populations as types of atomic charges, because I think this error is responsible for the confusion surrounding the definition of atomic charges. While we may not be able to measure atomic charges as precisely as energies in experiments, it is not true to say atomic charges are not experimentally observable. They can be observed and measured through spectroscopy experiments, albeit with much less precision than we are able to measure energies. I could go into more extensive details and examples if you are interested. While Mulliken and Lowdin populations do not quantify atomic charges (because Mulliken and Lowdin populations have no mathematical limits as described above), there are methods that have been developed to quantify net atomic charges. In contrast to the Mulliken and Lowdin populations, valid methods for quantifying net atomic charges approach a well-defined mathematical limit as the basis set size is increased. It is difficult for me to comprehend how this extremely simple and basic element eludes the grasp of so many people. As I have said before, this has been explained in numerous journal articles over the past 50 years. Could you please explain to me why a significant fraction of the computational chemistry research community is so slow to comprehend this basic fact? Sincerely, Tom >In that sense also the earlier discussion on which atomic charges are better was a bit nonsense to me as the atomis charges can not be physically measurable. And if I recall right Eugen Schwarz has allready in 1994 shown that "Only a single principal component of ionicity has been found, which is common to all the various operational charge definitions. " J. Phys.Chem. 98, 1994, p. 8245. On Wed, Sep 9, 2015 at 11:48 AM, Peeter Burk peeter.burk|a|ut.ee < owner-chemistry{=}ccl.net> wrote: > > Sent to CCL by: Peeter Burk [peeter.burk-$-ut.ee] > Sorry, could not hold me back! > > Any theory that explains or gives "insight on understanding what > experiments produce and what influences the experimental properties" is > theoretically useless. The only useful theory PREDICTS something that has > not yet been measured and can thus been verified or falsified. If the > HF/STO-3G can predict, let's say dipole moments (or HF/6-31G* the harmoni= c > frequencies) with given/known accuracy, then it is a valid and useful > theory. Chasing for newest (and thus usually more expensive) and often > uncalibrated (unverified against experiment) theories seems to be a commo= n > mistake... > > In that sense also the earlier discussion on which atomic charges are > better was a bit nonsense to me as the atomis charges can not be physical= ly > measurable. And if I recall right Eugen Schwarz has allready in 1994 show= n > that "Only a single principal component of ionicity has been found, which > is common to all the various operational charge definitions. " J. > Phys.Chem. 98, 1994, p. 8245. > > Best regards to everybody > Peeter Burk > University of Tartu, Estonia > > On 09/09/2015 04:28 PM, V=C3=ADctor Lua=C3=B1a Cabal victora/fluor.quimic= a.uniovi.es > wrote: > >> >> Sent to CCL by: =3D?iso-8859-1?Q?V=3DEDctor_Lua=3DF1a?=3D Cabal [victor:= +: >> fluor.quimica.uniovi.es] >> On Wed, Sep 09, 2015 at 07:19:58AM +0100, Peter Jarowski peterjarowski##= # >> gmail.com wrote: >> >>> Thanks for the snarky comment it highlights an important point. I am >>> really >>> not interested in methodology for this present work and neither is >>> industry. They care about prediction and having it fast. I see level >>> nerds >>> as being a barrier to a better interaction between theory and experimen= t. >>> So, people, if you have HF/sto-3g calcs that do the job, I want to know= . >>> >> >> Peter, >> >> Your remark points to a important quastion about the role of theory >> in science, and particularly in scientific simulation. >> >> This is a debatable subject, certainly, but my opinion is that the >> objective of theory is not reproducing experiments. Nothing surpasses >> experimental measurement on that. The role of theory is insight on >> understanding what experiments produce and what influences the >> experimental properties. >> >> You point to HF/sto-3g calcs as a good example of demod=C3=A9 techniques= that >> can be useful for a purpose. I point to the H=C3=BCckel and extended-H= =C3=BCckel >> techniques that were used by a Nobel prize school: Roald Hoffmann. >> >> Good luck in your chase for heroic examples of theoretical chemistry >> examples, >> V=C3=ADctor >> -- >> . . "In science a person can be convinced by a good argument. >> / `' \ That is almost impossible in politics or religion" >> /(o)(o)\ (Adapted from Carl Sagan) >> /`. \/ .'\ "Lo mediocre es peor que lo bueno, pero tambi=C3=A9n es p= eor >> / '`'` \ que lo malo, porque la mediocridad no es un grado, es una >> | \'`'`/ | actitud" -- Jorge Wasenberg, 2015 >> | |'`'`| | (Mediocre is worse than good, but it is also worse than >> \/`'`'`'\/ bad, because mediocrity is not a grade, it is an attitude= ) >> =3D=3D=3D(((=3D=3D)))=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D+=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D >> ! Dr.V=C3=ADctor Lua=C3=B1a, in silico chemist & prof. !"I have two kind= s of >> problems, >> ! Departamento de Qu=C3=ADmica F=C3=ADsica y Anal=C3=ADtica ! the urgent= and the >> important. >> ! Universidad de Oviedo, 33006-Oviedo, Spain ! The urgent are not >> important, >> ! e-mail: victor###fluor.quimica.uniovi.es ! and the important are >> never >> ! phone: +34-985-103491 fax: +34-985-103125 ! urgent. >> +--------------------------------------------+ (Dwight D. >> Eisenhower) >> GroupPage : http://azufre.quimica.uniovi.es/> >> > > > > -=3D This is automatically added to each message by the mailing script = =3D-http://www.ccl.net/chemistry/sub_unsub.shtmlConferences: > http://server.ccl.net/chemistry/announcements/conferences/> > > --001a1139644ec9c0a1051f66ad5e Content-Type: text/html; charset=UTF-8 Content-Transfer-Encoding: quoted-printable
Hi Peeter,

There is a fundamental distin= ction between the current conversation focused on exchange-correlation theo= ries and basis sets and the earlier discussion focused on atomic properties= . If one increases the basis set size, exchange-correlation functionals suc= h as B3LYP, M06, or whatever one you care to use will approach a well-defin= ed mathematical limit. We can then discuss what the relative accuracy of th= at mathematical limit is in comparison to experimental properties and also = discuss how close we are to that mathematical limit with a particular basis= set. Thus, it is meaningful to discuss how adequate an exchange-correlatio= n theory or basis set are for a particular research problem. Of course, the= goal is to choose an adequate level that is not too computationally expens= ive for the particular research question being studied.

In contrast, Mulliken and Lowdin populat= ion analysis schemes do not have any defined mathematical limits. As the ba= sis set is increased and the energy and electron density approach the compl= ete basis set limit, the Mulliken and Lowdin populations behave erratically= and blow up.=C2=A0This= is how we know for sure that Mulliken and Lowdin population analysis schem= es are utter nonsense and should never be used for publication results. As = pointed out by one person, their only purpose is for debugging calculations= to see if the symmetry or other basic features of the input geometry are m= alformed.

It is not the earlier discussion on atomic charges that is &quo= t;nonsense" but rather the Mulliken and Lowdin populations that are no= nsense, because they have no defined mathematical limits. This has nothing = to do with atomic charges, per se. The Mulliken and Lowdin populations do n= ot measure anything physical. They do not measure atomic charges. Probably = the confusion has been propagated by calling Mulliken and Lowdin population= s as types of "atomic charges", but really the Mulliken and Lowdi= n populations cannot be atomic charges, because they have no defined mathem= atical limits. In the future, I shall try to avoid referring to Mulliken an= d Lowdin populations as types of atomic charges, because I think this error= is responsible for the confusion surrounding the definition of atomic char= ges. While we may not be able to measure atomic charges as precisely as ene= rgies in experiments, it is not true to say atomic charges are not experime= ntally observable. They can be observed and measured through spectroscopy e= xperiments, albeit with much less precision than we are able to measure ene= rgies. I could go into more extensive details and examples if you are inter= ested.

While Mulliken= and Lowdin populations do not quantify atomic charges (because Mulliken an= d Lowdin populations have no mathematical limits as described above), there= are methods that have been developed to quantify net atomic charges. In co= ntrast to the Mulliken and Lowdin populations, valid methods for quantifyin= g net atomic charges approach a well-defined mathematical limit as the basi= s set size is increased. It is difficult for me to comprehend how this extr= emely simple and basic element eludes the grasp of so many people. As I hav= e said before, this has been explained in numerous journal articles over th= e past 50 years. Could you please explain to me why a significant fraction = of the computational chemistry research community is so slow to comprehend = this basic fact? =C2=A0

Sincerely,

Tom

>In that sense also t= he earlier discussion on which atomic charges are better was a bit nonsense= to me as the atomis charges can not be physically measurable. And if I rec= all right Eugen Schwarz has allready in 1994 shown that "Only a single= principal component of ionicity has been found, which is common to all the= various operational charge definitions. " J. Phys.Chem. 98, 1994, p. = 8245.

On Wed, Sep 9, 2015 at 11:48 AM, Peeter Burk peeter.burk|a|ut.ee <owner-chemistry{=}ccl.net> wrote:

Sent to CCL by: Peeter Burk [peeter.burk-$-ut.ee]
Sorry, could not hold me back!

Any theory that explains or gives "insight on understanding what exper= iments produce and what influences the experimental properties" is the= oretically useless. The only useful theory PREDICTS something that has not = yet been measured and can thus been verified or falsified. If the HF/STO-3G= can predict, let's say dipole moments (or HF/6-31G* the harmonic frequ= encies) with given/known accuracy, then it is a valid and useful theory. Ch= asing for newest (and thus usually more expensive) and often uncalibrated (= unverified against experiment) theories seems to be a common mistake...

In that sense also the earlier discussion on which atomic charges are bette= r was a bit nonsense to me as the atomis charges can not be physically meas= urable. And if I recall right Eugen Schwarz has allready in 1994 shown that= "Only a single principal component of ionicity has been found, which = is common to all the various operational charge definitions. " J. Phys= .Chem. 98, 1994, p. 8245.

Best regards to everybody
Peeter Burk
University of Tartu, Estonia

On 09/09/2015 04:28 PM, V=C3=ADctor Lua=C3=B1a Cabal victora/fluor.qui= mica.uniovi.es wrote:

Sent to CCL by: =3D?iso-8859-1?Q?V=3DEDctor_Lua=3DF1a?=3D Cabal [victor:+:<= a href=3D"http://fluor.quimica.uniovi.es" rel=3D"noreferrer" target=3D"_bla= nk">fluor.quimica.uniovi.es]
On Wed, Sep 09, 2015 at 07:19:58AM +0100, Peter Jarowski peterjarowski###gmail.com wrote:
Thanks for the snarky comment it highlights an important point. I am really=
not interested in methodology for this present work and neither is
industry. They care about prediction and having it fast. I see level nerds<= br> as being a barrier to a better interaction between theory and experiment. So, people, if you have HF/sto-3g calcs that do the job, I want to know.

Peter,

Your remark points to a important quastion about the role of theory
in science, and particularly in scientific simulation.

This is a debatable subject, certainly, but my opinion is that the
objective of theory is not reproducing experiments. Nothing surpasses
experimental measurement on that. The role of theory is insight on
understanding what experiments produce and what influences the
experimental properties.

You point to HF/sto-3g calcs as a good example of demod=C3=A9 techniques th= at
can be useful for a purpose. I point to the H=C3=BCckel and extended-H=C3= =BCckel
techniques that were used by a Nobel prize school: Roald Hoffmann.

Good luck in your chase for heroic examples of theoretical chemistry
examples,
=C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 V=C3=ADctor
--
=C2=A0 =C2=A0 =C2=A0 .=C2=A0 .=C2=A0 =C2=A0 "In science a person can b= e convinced by a good argument.
=C2=A0 =C2=A0 =C2=A0/ `' \=C2=A0 =C2=A0That is almost impossible in pol= itics or religion"
=C2=A0 =C2=A0 /(o)(o)\=C2=A0 (Adapted from Carl Sagan)
=C2=A0 =C2=A0/`. \/ .'\=C2=A0 "Lo mediocre es peor que lo bueno, p= ero tambi=C3=A9n es peor
=C2=A0 /=C2=A0 =C2=A0'`'`=C2=A0 =C2=A0\ que lo malo, porque la medi= ocridad no es un grado, es una
=C2=A0 |=C2=A0 \'`'`/=C2=A0 | actitud" -- Jorge Wasenberg, 201= 5
=C2=A0 |=C2=A0 |'`'`|=C2=A0 | (Mediocre is worse than good, but it = is also worse than
=C2=A0 =C2=A0\/`'`'`'\/=C2=A0 bad, because mediocrity is not a = grade, it is an attitude)
=3D=3D=3D(((=3D=3D)))=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D+=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
! Dr.V=C3=ADctor Lua=C3=B1a, in silico chemist & prof. !"I have tw= o kinds of problems,
! Departamento de Qu=C3=ADmica F=C3=ADsica y Anal=C3=ADtica ! the urgent an= d the important.
! Universidad de Oviedo, 33006-Oviedo, Spain ! The urgent are not important= ,
! e-mail:=C2=A0 =C2=A0victor###fluor.quimica.uniovi.es=C2=A0 =C2= =A0! and the important are never
! phone: +34-985-103491=C2=A0 fax: +34-985-103125 ! urgent.
+--------------------------------------------+=C2=A0 =C2=A0 =C2=A0 =C2=A0 (= Dwight D. Eisenhower)
=C2=A0 GroupPage : http://azufre.quimica.uniovi.es/>



-=3D This is automatically added to each message by the mailing script =3D-=
E-mail to subscribers: CHEMISTRY{=}ccl.net or use:
=C2=A0 =C2=A0 =C2=A0http://www.ccl.net/cgi-bin/ccl/se= nd_ccl_message

E-mail to administrators: CHEMISTRY-REQUEST{=}ccl.net or use
=C2=A0 =C2=A0 =C2=A0http://www.ccl.net/cgi-bin/ccl/se= nd_ccl_message

Subscribe/Unsubscribe:=C2=A0 =C2=A0 =C2=A0 http://www.cc= l.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/conferences/

Search Messages: http://www.ccl.net/chemistry/sear= chccl/index.shtml

=C2=A0 =C2=A0 =C2=A0http://www.ccl.net/spammers.txt

RTFI: http://www.ccl.net/chemistry/aboutccl/ins= tructions/



--001a1139644ec9c0a1051f66ad5e-- From owner-chemistry@ccl.net Thu Sep 10 15:17:01 2015 From: "Jan Goetze jpg9 .. st-andrews.ac.uk" To: CCL Subject: CCL: Case Studies of QM Computational Chemistry in Reactivity Message-Id: <-51695-150910113800-26578-NNcuKTMm5Hn6Q/U/s6rWsg||server.ccl.net> X-Original-From: Jan Goetze Content-Language: en-US Content-Type: multipart/alternative; boundary="_000_HE1PR06MB1484ADACAA6DEF9BD68C93F6F0510HE1PR06MB1484eurp_" Date: Thu, 10 Sep 2015 15:37:27 +0000 MIME-Version: 1.0 Sent to CCL by: Jan Goetze [jpg9]~[st-andrews.ac.uk] --_000_HE1PR06MB1484ADACAA6DEF9BD68C93F6F0510HE1PR06MB1484eurp_ Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Dear Peter, are you looking for qualitative or quantitative predictions? Are the system= s large or small? The latter ones are usually what the "pure" theoretical chemists are concer= ned about. For the large systems, the method is just one degree of freedom = among many others (choice of solvent, removing apparently irrelevant parts = of the system, conformation...) This is the domain of the "pure" computatio= nal chemist. People tending towards one side will often disregard caveats from the other= in favour of higher publication output. Hence, more theoretically oriented= chemists will often stick to simple models, showing that their methods rep= roduce, while more computationally oriented chemists often employ old, esta= blished methods on elaborate, realistic models. A combined approach is desi= rable, but hurts the publication output... Cheers from Scotland, Jan Jan Philipp G=F6tze School of Chemistry University of St Andrews North Haugh St Andrews, Fife KY16 9ST, Scotland, United Kingdom Tel.: +44 (0)1334-464748 (ext. 14141) > Hi Lars: > > There is no need for defense. We all agree that method development is > important. > > Personally, I am a bit defensive of my own work which was published in > 2009 > in a non-theoretical journal and consist of 80% experiment. I provided > that > example to show one where theory (whatever it is) worked well to the poin= t > of prediction. Again, this is what I am looking for in this thread. > > Unfortunately, theory can not exist on its own and the finances flow from > experiment downward. You will find only blank looks with the arguments yo= u > have made outside of a room full of theorists and that is the reality. On > the other hand, often we are in front of chemists, at least, and having > verified chemical examples of the importance of theory in necessary. Thes= e > should be chemically understandable and should be dealing with issues tha= t > are important to the audience. They should be transferable. That means > using standard methods with the right statistics and history that can be > computed in a reasonable time and executed, potentially, by non-experts. > So, while I love MRCI, for example, even the best of us will have trouble > getting it right on relevant molecules. > > Best, > > Peter > > On Wed, Sep 9, 2015 at 4:06 PM, Lars Goerigk lars.goerigk-,-unimelb.edu.a= u > < > owner-chemistry-#-ccl.net> wrote: > >> >> Sent to CCL by: "Lars Goerigk" [lars.goerigk : unimelb.edu.au] >> Hi Peter, >> >> in defense of us people that believe in obtaining the right result for >> the >> right reason: even if industry is interested in quick-and-dirty results, >> would it not be highly embarassing to make recommendations based on a >> method that relies on unforeseeable error compensation, only to then see >> that the experimentalists cannot reproduce your predictions? Susi's >> comment >> was therefore valid, albeit I would have worded it in a more diplomatic >> way. >> >> I find it very important to apply a level of theory that comes with a >> low >> risk of unsystematic error compensation. For example, a >> dispersion-corrected (double-)hybrid density functional with at least a >> triple-zeta basis set is overall much more reliable than the popular >> B3LYP/6-31G* level; the reasons for that are clear and they have been >> discussed extensively in the literature and in this forum. If you cannot >> afford a large basis set, but you want to quickly obtain results for >> bigger >> systems with an acceptable accuracy, there are other promising methods >> out >> there (e.g. HF-3c or PBEh-3c). Moreover, it is also important to use >> reliable solvation models (e.g. COSMO-RS) and reliable enthalpy and >> entropy >> corrections. >> >> As to your comment on "level nerds" being a barrier for successful >> interaction between theory and experiment, quantum chemists do not >> investigate and develop new methdods just because they do not know what >> else to do with their time. I would rather say that this discussion >> highlights a lack of communication between method developers and the >> users >> of QM methods and as always the fault for that lies probably on both >> sides; >> one consequence of that is that B3LYP/6-31G* is still so popular in the >> year 2015. >> >> Giving a general answer to your question is difficult, as it is not >> quite >> sure what property you want to calculate, which is exactly why the >> various >> answers to your question differ so much. However, there are numerous >> benchmark studies out there that provide some guidance on which methods >> to >> use and which to avoid; I have given some hints on reliable methods >> above. >> >> Of course, there is never 100% certainty that your predictions are right >> because in the end you try to describe a highly complex system with a >> theoretical model, which by definition is only a simplification of the >> real >> world. That complexity also means that there is not one sole, easy >> solution >> to your question. Therefore, it is probably wise to run calculations at >> two >> different levels of theory to distinguish between general trends and >> methodological artifacts. By carefully choosing a statistically reliable >> level of theory, you can at least minimise the possibilibity of making >> wrong predictions, and that is something that also industry people will >> accept if properly explained. I hope some of these hints were helpful. >> >> Finally, I hope you forgive me for saying that having efficient and ever >> improving QM methods is something that would not be possible without all >> those dedicated level nerds out there, which is why we should regard >> that >> term as a compliment. >> >> Cheers, >> Lars >> >> --- >> Dr. Lars Goerigk >> ARC DECRA Fellow >> School of Chemistry >> The University of Melbourne >> VIC 3010 >> Australia >> >> Research profile: >> http://www.chemistry.unimelb.edu.au/dr-lars-goerigk >> List of my publications: >> http://www.researcherid.com/rid/D-3717-2009 >> Follow me on Twitter: https://twitter.com/lgoer_compchem> >> >> > --_000_HE1PR06MB1484ADACAA6DEF9BD68C93F6F0510HE1PR06MB1484eurp_ Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable

Dear Peter,


are you looking for qualitative or quantitative predictions? Are the sys= tems large or small? 


The latter ones are usually what the "pure" theoretical chemis= ts are concerned about. For the large systems, the method is just one degre= e of freedom among many others (choice of solvent, removing apparently irre= levant parts of the system, conformation...) This is the domain of the "pure" computational chemist.


People tending towards one side will often disregard caveats from the ot= her in favour of higher publication output. Hence, more theoretically orien= ted chemists will often stick to simple models, showing that their methods = reproduce, while more computationally oriented chemists often employ old, established methods on elaborate, real= istic models. A combined approach is desirable, but hurts the publication o= utput...


Cheers from Scotland,

Jan


Jan Philipp G=F6tze
School of Chemistry
University of St Andrews
North Haugh
St Andrews, Fife
KY16 9ST, Scotland, United Kingdom

Tel.: +44 (0)1334-464748 (ext. 14141)


> Hi Lars:
>
> There is no need for defense. We all agree that method development is<= br> > important.
>
> Personally, I am a bit defensive of my own work which was published in=
> 2009
> in a non-theoretical journal and consist of 80% experiment. I provided=
> that
> example to show one where theory (whatever it is) worked well to the p= oint
> of prediction. Again, this is what I am looking for in this thread. >
> Unfortunately, theory can not exist on its own and the finances flow f= rom
> experiment downward. You will find only blank looks with the arguments= you
> have made outside of a room full of theorists and that is the reality.= On
> the other hand, often we are in front of chemists, at least, and havin= g
> verified chemical examples of the importance of theory in necessary. T= hese
> should be chemically understandable and should be dealing with issues = that
> are important to the audience. They should be transferable. That means=
> using standard methods with the right statistics and history that can = be
> computed in a reasonable time and executed, potentially, by non-expert= s.
> So, while I love MRCI, for example, even the best of us will have trou= ble
> getting it right on relevant molecules.
>
>  Best,
>
> Peter
>
> On Wed, Sep 9, 2015 at 4:06 PM, Lars Goerigk lars.goerigk-,-unimelb.ed= u.au
> <
> owner-chemistry-#-ccl.net> wrote:
>
>>
>> Sent to CCL by: "Lars  Goerigk" [lars.goerigk : uni= melb.edu.au]
>> Hi Peter,
>>
>> in defense of us people that believe in obtaining the right result= for
>> the
>> right reason: even if industry is interested in quick-and-dirty re= sults,
>> would it not be highly embarassing to make recommendations based o= n a
>> method that relies on unforeseeable error compensation, only to th= en see
>> that the experimentalists cannot reproduce your predictions? Susi'= s
>> comment
>> was therefore valid, albeit I would have worded it in a more diplo= matic
>> way.
>>
>> I find it very important to apply a level of theory that comes wit= h a
>> low
>> risk of unsystematic error compensation. For example, a
>> dispersion-corrected (double-)hybrid density functional with at le= ast a
>> triple-zeta basis set is overall much more reliable than the popul= ar
>> B3LYP/6-31G* level; the reasons for that are clear and they have b= een
>> discussed extensively in the literature and in this forum. If you = cannot
>> afford a large basis set, but you want to quickly obtain results f= or
>> bigger
>> systems with an acceptable accuracy, there are other promising met= hods
>> out
>> there (e.g. HF-3c or PBEh-3c). Moreover, it is also important to u= se
>> reliable solvation models (e.g. COSMO-RS) and reliable enthalpy an= d
>> entropy
>> corrections.
>>
>> As to your comment on "level nerds" being a barrier for = successful
>> interaction between theory and experiment, quantum chemists do not=
>> investigate and develop new methdods just because they do not know= what
>> else to do with their time. I would rather say that this discussio= n
>> highlights a lack of communication between method developers and t= he
>> users
>> of QM methods and as always the fault for that lies probably on bo= th
>> sides;
>> one consequence of that is that B3LYP/6-31G* is still so popular i= n the
>> year 2015.
>>
>> Giving a general answer to your question is difficult, as it is no= t
>> quite
>> sure what property you want to calculate, which is exactly why the=
>> various
>> answers to your question differ so much. However, there are numero= us
>> benchmark studies out there that provide some guidance on which me= thods
>> to
>> use and which to avoid; I have given some hints on reliable method= s
>> above.
>>
>> Of course, there is never 100% certainty that your predictions are= right
>> because in the end you try to describe a highly complex system wit= h a
>> theoretical model, which by definition is only a simplification of= the
>> real
>> world. That complexity also means that there is not one sole, easy=
>> solution
>> to your question. Therefore, it is probably wise to run calculatio= ns at
>> two
>> different levels of theory to distinguish between general trends a= nd
>> methodological artifacts. By carefully choosing a statistically re= liable
>> level of theory, you can at least minimise the possibilibity of ma= king
>> wrong predictions, and that is something that also industry people= will
>> accept if properly explained. I hope some of these hints were help= ful.
>>
>> Finally, I hope you forgive me for saying that having efficient an= d ever
>> improving QM methods is something that would not be possible witho= ut all
>> those dedicated level nerds out there, which is why we should rega= rd
>> that
>> term as a compliment.
>>
>> Cheers,
>> Lars
>>
>> ---
>> Dr. Lars Goerigk
>> ARC DECRA Fellow
>> School of Chemistry
>> The University of Melbourne
>> VIC 3010
>> Australia
>>
>> Research profile:
>> http://www.chemistry.unimelb.edu.au/dr-lars-goerigk
>> List of my publications:
>> http://www.researcherid.com/rid/D-3717-2009
>> Follow me on Twitter: https://twitter.com/lgoer_compchem>
>>
>>
>

--_000_HE1PR06MB1484ADACAA6DEF9BD68C93F6F0510HE1PR06MB1484eurp_-- From owner-chemistry@ccl.net Thu Sep 10 15:52:01 2015 From: "Tom Albright talbright1234-$-gmail.com" To: CCL Subject: CCL: Case Studies of QM Computational Chemistry in Reactivity Message-Id: <-51696-150910125843-31179-FWjdbGifqUiGdJ3XTbRuVA[]server.ccl.net> X-Original-From: Tom Albright Content-Type: multipart/alternative; boundary=Apple-Mail-12--259353049 Date: Thu, 10 Sep 2015 11:58:34 -0500 Mime-Version: 1.0 (Apple Message framework v1085) Sent to CCL by: Tom Albright [talbright1234[#]gmail.com] --Apple-Mail-12--259353049 Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset=us-ascii Dear Susi - you have completely missed my point. So I do a calculation = of the transition state for some reaction at the aug-cc-pCVTZ level with = CCSD(T) or some double functional with dispersion and get a barrier of = 40.35 kcal/mol. Great, so the fact that this reaction does not = experimentally occur is in perfect agreement with my theoretical = results. So this is the end of the story, right? Wrong! The point is why = is the barrier so large - the conclusion should not be that one is = reasonably assured that the numerology matches what we know is true (ie = experiment). An explanation is useful, for example, if one can make a = prediction of how one can modify the reactants to get a lower barrier. = In another context one can show that the orientation of the Cr(CO)3 = group is determined by the substituents on the arene in arene-Cr(CO)3 = complexes. Furthermore the orientation of the Cr(CO)3 group directs = attack by nucleophiles on the arene. All this at the (almost, but I am = sure in your mind, totally) unreliable EHT level. The energy differences = are small but one can formulate a coherent explanation of the = regioselectivity. This is useful. Experimentalists can and have used = this to design the synthesis of compounds. I understand your position - = if one is just reporting numbers then it is important to use a reliable = technique - but is this really the sole role of a theoretician? =46rom = the old "Dragnet" TV show - "Just give me the facts, ma'am". I see a = spectrum of questions in this forum that range from a range from the = relatively sophisticated concerning a computational approach to the = incredibly naive where even a basic knowledge of chemistry is absent. = Unfortunately with the "black boxes" available, the latter can easily = figure out what level to use and, for example, compute every diatomic in = the known universe at a reliable level (at that time). Yeah - that = really pushes chemistry forward. Victor was correct... On Sep 9, 2015, at 8:31 PM, Susi Lehtola susi.lehtola ~~ = alumni.helsinki.fi wrote: >=20 > Sent to CCL by: Susi Lehtola [susi.lehtola%a%alumni.helsinki.fi] > On 09/09/2015 10:28 AM, Tom Albright talbright1234]=3D[gmail.com = wrote: >> I wholeheartedly agree with you Victor. Too often I see "how high can = I go" rather than a coherent explanation of why I got the results that I = did and can I extrapolate this "understanding" to other examples. And no = this is not the '90s. >=20 > That's simply not true. If you use a crappy method (like 6-31G*/B3LYP = really is), then you are in no position to make any claims on why you = get the results you get, because you're relying on fortuitous error = cancellation between the method and the basis set. You're so far from = the basis set limit that you can't make real quantitative judgements = whatever the method is you're using. >=20 > Yes, it's not the 90s, but too many people act like if it were. = Computational methods have progressed a whole lot since those days, but = still way too many people stick to old habits without so much as an = afterthought. > --=20 > = ----------------------------------------------------------------------- > Mr. Susi Lehtola, PhD Chemist Postdoctoral Fellow > susi.lehtola]![alumni.helsinki.fi Lawrence Berkeley National = Laboratory > http://www.helsinki.fi/~jzlehtol USA > = ----------------------------------------------------------------------- >=20 >=20 >=20 > -=3D This is automatically added to each message by the mailing script = =3D- > To recover the email address of the author of the message, please = change>=20>=20>=20=>=20>=20Conferences: = http://server.ccl.net/chemistry/announcements/conferences/ >=20>=20>=20>=20 >=20 With Best Regards Tom Albright --Apple-Mail-12--259353049 Content-Transfer-Encoding: quoted-printable Content-Type: text/html; charset=us-ascii Dear = Susi - you have completely missed my point. So I do a calculation of the = transition state for some reaction at the aug-cc-pCVTZ level with = CCSD(T) or some double functional with dispersion and get a barrier of = 40.35 kcal/mol. Great, so the fact that this reaction does not = experimentally occur is in perfect agreement with my theoretical = results. So this is the end of the story, right? Wrong! The point is why = is the barrier so large - the conclusion should not be that one is = reasonably assured that the  numerology matches what we know is = true (ie experiment). An explanation is useful, for example, if one can = make a prediction of how one can modify the reactants to get a lower = barrier. In another context one can show that the orientation of the = Cr(CO)3 group is determined by the substituents on the arene in = arene-Cr(CO)3 complexes. Furthermore the orientation of the Cr(CO)3 = group directs attack by nucleophiles on the arene. All this at the = (almost, but I am sure in your mind, totally) unreliable EHT level. The = energy differences are small but one can formulate a coherent = explanation of the regioselectivity. This is useful. Experimentalists = can and have used this to design the synthesis of compounds. I = understand your position - if one is just reporting numbers then it is = important to use a reliable technique - but is this really the sole role = of a theoretician? =46rom the old "Dragnet" TV show - "Just give me the = facts, ma'am". I see a spectrum of questions in this forum that range = > from a range from the relatively sophisticated concerning a = computational approach to the incredibly naive where even a basic = knowledge of chemistry is absent. Unfortunately with the "black boxes" = available, the latter can easily figure out what level to use and, for = example, compute every diatomic in the known universe at a reliable = level (at that time). Yeah - that really pushes chemistry forward. = Victor was correct...
On Sep 9, 2015, at 8:31 PM, Susi = Lehtola susi.lehtola ~~ alumni.helsinki.fi wrote:


Sent to CCL by: Susi Lehtola = [susi.lehtola%a%alumni.helsinki.fi]
On 09/09/2015 10:28 AM, Tom = Albright talbright1234]=3D[gmail.com wrote:
I= wholeheartedly agree with you Victor. Too often I see "how high can I = go" rather than a coherent explanation of why I got the results that I = did and can I extrapolate this "understanding" to other examples. And no = this is not the '90s.

That's simply not true. If you = use a crappy method (like 6-31G*/B3LYP really is), then you are in no = position to make any claims on why you get the results you get, because = you're relying on fortuitous error cancellation between the method and = the basis set. You're so far from the basis set limit that you can't = make real quantitative judgements whatever the method is you're = using.

Yes, it's not the 90s, but too many people act like if it = were. Computational methods have progressed a whole lot since those = days, but still way too many people stick to old habits without so much = as an afterthought.
-- =
----------------------------------------------------------------------= -
Mr. Susi Lehtola, PhD =             Ch= emist Postdoctoral Fellow
susi.lehtola]![alumni.helsinki.fi =   Lawrence Berkeley National Laboratory
http://www.helsinki.fi/~jzlehtol= =  USA
-------------------------------------------------------------= ----------



-=3D This is automatically added to each = message by the mailing script =3D-
To recover the email address of = the author of the message, please change
the strange characters on = the top line to the ** sign. You can also
look up the X-Original-From: = line in the mail header.

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

E-mail to administrators: CHEMISTRY-REQUEST**ccl.net = or use
    http://www.ccl.ne= t/cgi-bin/ccl/send_ccl_message
http://www.ccl.net/c= hemistry/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/conferences/

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

If your mail bounces = > from CCL with 5.7.1 error, check:
    http://www.ccl.net/spammers.txt

RTFI: http://www.cc= l.net/chemistry/aboutccl/instructions/


<= /div>
With = Best Regards
Tom Albright



= --Apple-Mail-12--259353049-- From owner-chemistry@ccl.net Thu Sep 10 16:27:01 2015 From: "N. Sukumar nagams[a]rpi.edu" To: CCL Subject: CCL: Case Studies of QM Computational Chemistry in Reactivity Message-Id: <-51697-150910131035-6062-HQyK4GEczfKa1k8lnUtAlQ- -server.ccl.net> X-Original-From: "N. Sukumar" Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset=UTF-8; format=flowed Date: Thu, 10 Sep 2015 22:40:28 +0530 MIME-Version: 1.0 Sent to CCL by: "N. Sukumar" [nagams**rpi.edu] > theory can do something experiment can not: it can provide "error-less" > results. That reminds me of something I heard recently: in theory (although I prefer to use the term "computation" to refer to the results of various DFT functionals) the errors are front-loaded into the parameters that make up the model. Computations (of the non-statistical kind) can produce scatter-free results, bot not error-free results. -- N. SUKUMAR Professor & Head, Department of Chemistry Director, Center for Informatics Shiv Nadar University, India On 2015-09-09 23:59, Peter Jarowski peterjarowski%gmail.com wrote: > Hi Victor: > > Very much agreed. I remind readers that I was looking for cases where > theory predicted something that had not been experimentally performed > and then was later verified by experiment. Reproducing experiment with > theory is useful for verification, but doesn't go much beyond that. > > I would argue though that theory can do something experiment can not: > it can provide "error-less" results. More specifically, experiments > present so many confounds, known and unknown, whereas theory can model > the heart of the matter. Heats of formation, for example, may be more > accurate and more readily comparable by very robust theory versus > trolling the literature for these data derived under varied > conditions, at various times, and using different approaches. > > So lets keep them coming and later on I will summarize what was sent > to me here and on linkedin/researchgate and see what we all think. > > Best Regards, > > Peter > > On Wed, Sep 9, 2015 at 3:28 PM, Víctor Luaña Cabal > victora/fluor.quimica.uniovi.es [1] wrote: > >> Sent to CCL by: =?iso-8859-1?Q?V=EDctor_Lua=F1a?= Cabal >> [victor:+:fluor.quimica.uniovi.es [1]] >> On Wed, Sep 09, 2015 at 07:19:58AM +0100, Peter Jarowski >> peterjarowski###gmail.com [2] wrote: >>> Thanks for the snarky comment it highlights an important point. I >> am really >>> not interested in methodology for this present work and neither >> is >>> industry. They care about prediction and having it fast. I see >> level nerds >>> as being a barrier to a better interaction between theory and >> experiment. >>> So, people, if you have HF/sto-3g calcs that do the job, I want >> to know. >> >> Peter, >> >> Your remark points to a important quastion about the role of theory >> in science, and particularly in scientific simulation. >> >> This is a debatable subject, certainly, but my opinion is that the >> objective of theory is not reproducing experiments. Nothing >> surpasses >> experimental measurement on that. The role of theory is insight on >> understanding what experiments produce and what influences the >> experimental properties. >> >> You point to HF/sto-3g calcs as a good example of demodé >> techniques that >> can be useful for a purpose. I point to the Hückel and >> extended-Hückel >> techniques that were used by a Nobel prize school: Roald Hoffmann. >> >> Good luck in your chase for heroic examples of theoretical >> chemistry >> examples, >>            Víctor >> -- >>      .  .    "In science a person can be convinced by a good >> argument. >>     / `'    That is almost impossible in politics or religion" >>    /(o)(o)  (Adapted from Carl Sagan) >>   /`. / .'  "Lo mediocre es peor que lo bueno, pero también es >> peor >>  /   '`'`    que lo malo, porque la mediocridad no es un >> grado, es una >>  |  '`'`/  | actitud" -- Jorge Wasenberg, 2015 >>  |  |'`'`|  | (Mediocre is worse than good, but it is also worse >> than >>   /`'`'`'/  bad, because mediocrity is not a grade, it is an >> attitude) >> > ===(((==)))==================================+========================= >> ! Dr.Víctor Luaña, in silico chemist & prof. !"I have two kinds >> of problems, >> ! Departamento de Química Física y Analítica ! the urgent and >> the important. >> ! Universidad de Oviedo, 33006-Oviedo, Spain ! The urgent are not >> important, >> ! e-mail:   victor###fluor.quimica.uniovi.es [1]   ! and the >> important are never >> ! phone: +34-985-103491 [3]  fax: +34-985-103125 [4] ! urgent. >> +--------------------------------------------+        (Dwight >> D. Eisenhower) >>  GroupPage : http://azufre.quimica.uniovi.es/ [5] >> >> -= This is automatically added to each message by the mailing >> script =- >> >> E-mail to subscribers: CHEMISTRY*ccl.net or use: >>      [6] >> >> E-mail to administrators: CHEMISTRY-REQUEST*ccl.net or use >>      [6]>>      [7][8][9] >> Conferences: >> http://server.ccl.net/chemistry/announcements/conferences/ [10]>> [11] >> >>      [12][13] > > > > Links: > ------ > [1] http://fluor.quimica.uniovi.es > [2] http://gmail.com > [3] tel:%2B34-985-103491 > [4] tel:%2B34-985-103125 > [5] http://azufre.quimica.uniovi.es/ > [6]> [7]> [8] http://www.ccl.net > [9] http://www.ccl.net/jobs > [10] http://server.ccl.net/chemistry/announcements/conferences/ > [11] http://www.ccl.net/chemistry/searchccl/index.shtml > [12]> [13] http://www.ccl.net/chemistry/aboutccl/instructions/ From owner-chemistry@ccl.net Thu Sep 10 17:01:00 2015 From: "Jerome Kieffer Jerome.Kieffer.+/-.terre-adelie.org" To: CCL Subject: CCL: Case Studies of QM Computational Chemistry in Reactivity Message-Id: <-51698-150910095048-5330-D4wf5evuzSHrKN9jYUzeHQ+/-server.ccl.net> X-Original-From: Jerome Kieffer Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset=UTF-8 Date: Thu, 10 Sep 2015 15:50:40 +0200 MIME-Version: 1.0 Sent to CCL by: Jerome Kieffer [Jerome.Kieffer*o*terre-adelie.org] On Thu, 10 Sep 2015 07:26:41 +0100 "Peter Jarowski peterjarowski===gmail.com" wrote: > I ask a simple question and further ask everyone to pretend they are at an > interview outside of an academic environment where you will find willing > and interested parties who enjoy the alphabet soup of DFT. what example and > what statistics would you present to justify your work. Clearly, you are > talking to experimentalists, as theory does not exist without them. > > How does theory drive experiment? > How much money does it save? > How much revenue can it produce? > How reliable is it? > How essential is it and what can it do that experiment can not? Hi all, I had the opportunity to do computational chemistry as part of a (large) industrial group, a decade ago. Such questions were often asked and it was not easy to be honest. When speaking of reactivity, everything is about the energy barrier between two (or more) path which leads to different products. Some critical steps may be mono, the other bi-molecular, so it is harder to compare them... Then comes the enthalpic and entropic contribution, not speaking about solvatation for spices that do not exist per-se as they are transition states. I left this job a while ago and I have to admit the "correctness" of most result I obtained at that time were mainly "lucky error cancelation": Concidering : A --> B A --> C in kinetic condition, let AB and AC be the transition states: [C]/[B] = exp(-(G[AC*]-G[AB*])/RT) The error in the exponential is 2x the error in calculating the free energy of one TS, one ends with a huge error bars which is anything but conclusive at that time (one error was about 6kcal/mol ... while with ony 1kcal/mol error, no conclusion was possible) . Cheers, -- Jérôme Kieffer From owner-chemistry@ccl.net Thu Sep 10 17:37:01 2015 From: "Stefan Grimme grimme,,thch.uni-bonn.de" To: CCL Subject: CCL: Case Studies of QM Computational Chemistry in Reactivity Message-Id: <-51699-150910160445-8248-sg9SvisPBZb/aP3qxoK4lg[A]server.ccl.net> X-Original-From: "Stefan Grimme" Date: Thu, 10 Sep 2015 16:04:44 -0400 Sent to CCL by: "Stefan Grimme" [grimme|*|thch.uni-bonn.de] Dear Tom, I followed this discussion quietly for some time but now can't resist to comment on this too extreme viewpoint: 1. Methods can be useful and reasonable without a definite mathematical limit. A Mulliken or Loewdin population analysis gives a definite result for a given well-defined AO basis set. If the set is small (minimal) the derived atomic charges are chemically reasonable and correlate well with those from other methods for well understood reasons. I don't want to defend orbital based partitionings (I prefer observables) but making the mathematical limit to the encompassing requirement seems nonsense to me. There are other useful and widely used QC methods like Moeller-Plesset perturbation theory which are often divergent (or at least convergence is unlcear) in large one-particle basis sets and hence also do not have a definite mathematical limit. Is this a good reason to abandon all MP2 calculations? 2. The word "observe" in our context can only mean "observable" in a QM sense. Hence, because there is no operator for "atomic charge" an observable atomic charge does not exist in a strict sense. You probably mean correlations of spectroscopic signatures with atomic charges when writing "They can be observed and measured through spectroscopy experiments". If you have another opinion on that I would like to know more details on how to measure atomic charges. Best wishes Stefan >Hi Peeter, >There is a fundamental distinction between the current conversation focused on exchange-correlation theories and basis sets and the earlier discussion focused on atomic properties. If one increases the basis set size, exchange-correlation functionals such as B3LYP, M06, or whatever one you care to use will approach a well-defined mathematical limit. We can then discuss what the relative accuracy of that mathematical limit is in comparison to experimental properties and also discuss how close we are to that mathematical limit with a particular basis set. Thus, it is meaningful to discuss how adequate an exchange-correlation theory or basis set are for a particular research problem. Of course, the goal is to choose an adequate level that is not too computationally expensive for the particular research question being studied. >In contrast, Mulliken and Lowdin population analysis schemes do not have any defined mathematical limits. As the basis set is increased and the energy and electron density approach the complete basis set limit, the Mulliken and Lowdin populations behave erratically and blow up. This is how we know for sure that Mulliken and Lowdin population analysis schemes are utter nonsense and should never be used for publication results. As pointed out by one person, their only purpose is for debugging calculations to see if the symmetry or other basic features of the input geometry are malformed. >It is not the earlier discussion on atomic charges that is "nonsense" but rather the Mulliken and Lowdin populations that are nonsense, because they have no defined mathematical limits. This has nothing to do with atomic charges, per se. The Mulliken and Lowdin populations do not measure anything physical. They do not measure atomic charges. Probably the confusion has been propagated by calling Mulliken and Lowdin populations as types of "atomic charges", but really the Mulliken and Lowdin populations cannot be atomic charges, because they have no defined mathematical limits. In the future, I shall try to avoid referring to Mulliken and Lowdin populations as types of atomic charges, because I think this error is responsible for the confusion surrounding the definition of atomic charges. While we may not be able to measure atomic charges as precisely as energies in experiments, it is not true to say atomic charges are not experimentally observable. They can be observed and measured through spectroscopy experiments, albeit with much less precision than we are able to measure energies. I could go into more extensive details and examples if you are interested. From owner-chemistry@ccl.net Thu Sep 10 18:12:01 2015 From: "Jens Spanget-Larsen spanget~!~ruc.dk" To: CCL Subject: CCL: SV: Case Studies of QM Computational Chemistry in Reactivity Message-Id: <-51700-150910174806-27345-ZY/7HEncFmbVLSTqt46L5Q^-^server.ccl.net> X-Original-From: Jens Spanget-Larsen Content-Language: da-DK Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset="us-ascii" Date: Thu, 10 Sep 2015 21:47:35 +0000 MIME-Version: 1.0 Sent to CCL by: Jens Spanget-Larsen [spanget#%#ruc.dk] "Quantum chemistry is faced with two fundamental problems: How to reduce chemistry, in a practical way, to numerics; And how to avoid doing so and still benefit from theory" K. Wittel, S.P. McGlynn, Chem. Rev. 77, 745 (1977) ------------------------------------------------------ JENS SPANGET-LARSEN Office: +45 4674 2710 Dept. of Science Fax: +45 4674 3011 Roskilde University Mobile: +45 2320 6246 P.O.Box 260 E-Mail: spanget-*-ruc.dk DK-4000 Roskilde, Denmark http://www.ruc.dk/~spanget ------------------------------------------------------ From owner-chemistry@ccl.net Thu Sep 10 19:35:01 2015 From: "Susi Lehtola susi.lehtola+*+alumni.helsinki.fi" To: CCL Subject: CCL: Case Studies of QM Computational Chemistry in Reactivity Message-Id: <-51701-150910190719-26293-PBo8K5v5rn26TNxe2lYGug!^!server.ccl.net> X-Original-From: Susi Lehtola Content-Transfer-Encoding: 7bit Content-Type: text/plain; charset=windows-1252; format=flowed Date: Thu, 10 Sep 2015 16:07:06 -0700 MIME-Version: 1.0 Sent to CCL by: Susi Lehtola [susi.lehtola*alumni.helsinki.fi] On 09/10/2015 09:58 AM, Tom Albright talbright1234-$-gmail.com wrote: > Dear Susi - you have completely missed my point. So I do a > calculation of the transition state for some reaction at the > aug-cc-pCVTZ level with CCSD(T) or some double functional with > dispersion and get a barrier of 40.35 kcal/mol. Great, so the fact > that this reaction does not experimentally occur is in perfect > agreement with my theoretical results. So this is the end of the > story, right? Wrong! The point is why is the barrier so large - the > conclusion should not be that one is reasonably assured that the > numerology matches what we know is true (ie experiment). And so have you. Sure, I'd trust the value 40.35 kcal/mol given by a CCSD(T)/aug-cc-pCVTZ calculation to be sufficiently close to the true result (assuming the system is close to single reference). But, this does not say at all whether B3LYP/6-31G* would also give you a reasonably accurate answer; it might be anything from -60 to 100 kcal/mol [exaggeration for sake of argument]. As has already been said, a lot of the success of the B3LYP/6-31G* model relies on fortunate cancellations of errors between the small basis set and the B3LYP functional, but this may not be the case in all systems, or especially all properties. Now, if you know that B3LYP/6-31G* is good enough to yield accurate results for your system and property (compared to higher level ab initio data[*]), then all is fine. However, the past successes of B3LYP/6-31G* have made a good part of the applications community blind to all of the advances made in the past 20+ years. Depending on the problem, one might get much more accurate results with only a trivial amount of more resources. [*] Straight comparison of a low-level chemical model to experiment may lead you astray, since the experimental data might not have been analyzed correctly, or the theoretical values ignore some relevant chemistry (say, zero point or solvation effects). -- ----------------------------------------------------------------------- Mr. Susi Lehtola, PhD Chemist Postdoctoral Fellow susi.lehtola!A!alumni.helsinki.fi Lawrence Berkeley National Laboratory http://www.helsinki.fi/~jzlehtol USA ----------------------------------------------------------------------- From owner-chemistry@ccl.net Thu Sep 10 20:09:01 2015 From: "Thomas Manz thomasamanz-*-gmail.com" To: CCL Subject: CCL:G: Case Studies of QM Computational Chemistry in Reactivity Message-Id: <-51702-150910200217-25569-XEGze+UI5r58Ul8PhtoYgA++server.ccl.net> X-Original-From: Thomas Manz Content-Type: multipart/alternative; boundary=001a11400d6a68dd4f051f6d6d57 Date: Thu, 10 Sep 2015 18:02:10 -0600 MIME-Version: 1.0 Sent to CCL by: Thomas Manz [thomasamanz|gmail.com] --001a11400d6a68dd4f051f6d6d57 Content-Type: text/plain; charset=UTF-8 Hi Stefan, In regards to your questions about MP2, one has to be extremely careful with such an approach, because the denominator is of the form (energy_1 - energy_2) which causes the denominator to become zero when energy_1 = energy_2. This can cause perturbation methods to blow up. For this reason, I generally prefer non-perturbative methods such as CCSD, when a higher-level calculation result is needed. I would recommend CCSD as opposed to MP2, simply because CCSD has a more well-defined mathematical limit on the results of the calculation. Personally, I don't use MP2 calculations for this reason, but this doesn't necessarily mean others can't. However, I wouldn't go so far as to say that MP2 calculations don't have a well-defined basis set limit. I believe that for most systems the complete basis set limit would be well-defined for MP2 calculations. In this sense, the MP2 calculations are much more well-defined than Mulliken or Lowdin populations, which definitely do not have a basis set limit. > If the set is small (minimal) the derived atomic charges are chemically reasonable and correlate well with those from other methods for well understood reasons. The populations of the density matrix projected onto a smaller basis set is usually referred to by a different name. At least in Gaussian programs, it is called Pop=MBS. In Gaussian programs, this is a different algorithm than Pop=Regular which performs Mulliken analysis in the current basis set. In my experience, the Pop=MBS method is not very useful and tends to crash a large percentage of the time. It seems to crash especially often for heavier atoms and for those with pseudopotentials. Also, people have tested the idea to project plane-wave basis sets onto minimal localized atomic orbital basis sets, but this results in charge leakage where the density matrix in the smaller basis set does not accurately represent the true density matrix. In general, the small basis sets do not represent the density matrix with high accuracy. Therefore, in general, I cannot recommend the approach you mentioned. There are certainly much better approaches if the goal is to compute net atomic charges. Best, Tom On Thu, Sep 10, 2015 at 2:04 PM, Stefan Grimme grimme,,thch.uni-bonn.de < owner-chemistry/a\ccl.net> wrote: > > Sent to CCL by: "Stefan Grimme" [grimme|*|thch.uni-bonn.de] > Dear Tom, > I followed this discussion quietly for some time but now can't resist to > comment on this too extreme viewpoint: > > 1. Methods can be useful and reasonable without a definite mathematical > limit. A Mulliken or Loewdin population analysis gives a definite result > for a given well-defined AO basis set. If the set is small (minimal) the > derived atomic charges are chemically reasonable and correlate well with > those from other methods for well understood reasons. I don't want to > defend orbital based partitionings (I prefer observables) but making the > mathematical limit > to the encompassing requirement seems nonsense to me. > There are other useful and widely used QC methods like Moeller-Plesset > perturbation theory which are often divergent (or at least convergence is > unlcear) in large one-particle basis sets and hence also do not have a > definite mathematical limit. Is this a good reason to abandon all MP2 > calculations? > > 2. The word "observe" in our context can only mean "observable" in a QM > sense. Hence, because there is no operator for "atomic charge" an > observable atomic charge does not exist in a strict sense. You probably > mean > correlations of spectroscopic signatures with atomic charges when writing > "They can be observed and measured through spectroscopy experiments". > If you have another opinion on that I would like to know more details on > how to measure atomic charges. > > > Best wishes > Stefan > > >Hi Peeter, > > >There is a fundamental distinction between the current conversation > focused on exchange-correlation theories and basis sets and the earlier > discussion focused on atomic properties. If one increases the basis set > size, exchange-correlation functionals such as B3LYP, M06, or whatever one > you care to use will approach a well-defined mathematical limit. We can > then discuss what the relative accuracy of that mathematical limit is in > comparison to experimental properties and also discuss how close we are to > that mathematical limit with a particular basis set. Thus, it is meaningful > to discuss how adequate an exchange-correlation theory or basis set are for > a particular research problem. Of course, the goal is to choose an adequate > level that is not too computationally expensive for the particular research > question being studied. > > >In contrast, Mulliken and Lowdin population analysis schemes do not have > any defined mathematical limits. As the basis set is increased and the > energy and electron density approach the complete basis set limit, the > Mulliken and Lowdin populations behave erratically and blow up. This is how > we know for sure that Mulliken and Lowdin population analysis schemes are > utter nonsense and should never be used for publication results. As pointed > out by one person, their only purpose is for debugging calculations to see > if the symmetry or other basic features of the input geometry are malformed. > > >It is not the earlier discussion on atomic charges that is "nonsense" but > rather the Mulliken and Lowdin populations that are nonsense, because they > have no defined mathematical limits. This has nothing to do with atomic > charges, per se. The Mulliken and Lowdin populations do not measure > anything physical. They do not measure atomic charges. Probably the > confusion has been propagated by calling Mulliken and Lowdin populations as > types of "atomic charges", but really the Mulliken and Lowdin populations > cannot be atomic charges, because they have no defined mathematical limits. > In the future, I shall try to avoid referring to Mulliken and Lowdin > populations as types of atomic charges, because I think this error is > responsible for the confusion surrounding the definition of atomic charges. > While we may not be able to measure atomic charges as precisely as energies > in experiments, it is not true to say atomic charges are not experimentally > observable. They can be observed and m! > easured through spectroscopy experiments, albeit with much less precision > than we are able to measure energies. I could go into more extensive > details and examples if you are interested.> > > --001a11400d6a68dd4f051f6d6d57 Content-Type: text/html; charset=UTF-8 Content-Transfer-Encoding: quoted-printable
Hi Stefan,

In regards to your questions= about MP2, one has to be extremely careful with such an approach, because = the denominator is of the form (energy_1 - energy_2) which causes the denom= inator to become zero when energy_1 =3D energy_2. This can cause perturbati= on methods to blow up. For this reason, I generally prefer non-perturbative= methods such as CCSD, when a higher-level calculation result is needed. I = would recommend CCSD as opposed to MP2, simply because CCSD has a more well= -defined mathematical limit on the results of the calculation. Personally, = I don't use MP2 calculations for this reason, but this doesn't nece= ssarily mean others can't. However, I wouldn't go so far as to say = that MP2 calculations don't have a well-defined basis set limit. I beli= eve that for most systems the complete basis set limit would be well-define= d for MP2 calculations. In this sense, the MP2 calculations are much more w= ell-defined than Mulliken or Lowdin populations, which definitely do not ha= ve a basis set limit.

> If the set is small (minimal) the derived atomic char= ges are chemically reasonable and correlate well with those from other meth= ods for well understood reasons.

The populations of the density matrix projected onto a smal= ler basis set is usually referred to by a different name. At least in Gauss= ian programs, it is called Pop=3DMBS. In Gaussian programs, this is a diffe= rent algorithm than Pop=3DRegular which performs Mulliken analysis in the c= urrent basis set. In my experience, the Pop=3DMBS method is not very useful= and tends to crash a large percentage of the time. It seems to crash espec= ially often for heavier atoms and for those with pseudopotentials. Also, pe= ople have tested the idea to project plane-wave basis sets onto minimal loc= alized atomic orbital basis sets, but this results in charge leakage where = the density matrix in the smaller basis set does not accurately represent t= he true density matrix. In general, the small basis sets do not represent t= he density matrix with high accuracy. Therefore, in general, I cannot recom= mend the approach you mentioned. There are certainly much better approaches= if the goal is to compute net atomic charges.

Best,

Tom



= On Thu, Sep 10, 2015 at 2:04 PM, Stefan Grimme grimme,,thch.uni-bonn.de <owner-chemistry/a\ccl.net>= ; wrote:

Sent to CCL by: "Stefan=C2=A0 Grimme" [grimme|*|thch.uni-bonn.de]
Dear Tom,
I followed this discussion quietly for some time but now can't resist t= o
comment on this too extreme viewpoint:

1. Methods can be useful and reasonable without a definite mathematical lim= it. A Mulliken or Loewdin population analysis gives a definite result for a= given well-defined AO basis set. If the set is small (minimal) the derived= atomic charges are chemically reasonable and correlate well with those fro= m other methods for well understood reasons. I don't want to defend orb= ital based partitionings (I prefer observables) but making the mathematical= limit
to the encompassing requirement seems nonsense to me.
There are other useful and widely used QC methods like Moeller-Plesset
perturbation theory which are often divergent (or at least convergence is unlcear) in large one-particle basis sets and hence also do not have a
definite mathematical limit. Is this a good reason to abandon all MP2
calculations?

2. The word "observe" in our context can only mean "observab= le" in a QM
sense. Hence, because there is no operator for "atomic charge" an=
observable atomic charge does not exist in a strict sense. You probably mea= n
correlations of spectroscopic signatures with atomic charges when writing "They can be observed and measured through spectroscopy experiments&qu= ot;.
If you have another opinion on that I would like to know more details on how to measure atomic charges.


Best wishes
Stefan

>Hi Peeter,

>There is a fundamental distinction between the current conversation foc= used on exchange-correlation theories and basis sets and the earlier discus= sion focused on atomic properties. If one increases the basis set size, exc= hange-correlation functionals such as B3LYP, M06, or whatever one you care = to use will approach a well-defined mathematical limit. We can then discuss= what the relative accuracy of that mathematical limit is in comparison to = experimental properties and also discuss how close we are to that mathemati= cal limit with a particular basis set. Thus, it is meaningful to discuss ho= w adequate an exchange-correlation theory or basis set are for a particular= research problem. Of course, the goal is to choose an adequate level that = is not too computationally expensive for the particular research question b= eing studied.

>In contrast, Mulliken and Lowdin population analysis schemes do not hav= e any defined mathematical limits. As the basis set is increased and the en= ergy and electron density approach the complete basis set limit, the Mullik= en and Lowdin populations behave erratically and blow up. This is how we kn= ow for sure that Mulliken and Lowdin population analysis schemes are utter = nonsense and should never be used for publication results. As pointed out b= y one person, their only purpose is for debugging calculations to see if th= e symmetry or other basic features of the input geometry are malformed.

>It is not the earlier discussion on atomic charges that is "= ;nonsense" but rather the Mulliken and Lowdin populations that are non= sense, because they have no defined mathematical limits. This has nothing t= o do with atomic charges, per se. The Mulliken and Lowdin populations do no= t measure anything physical. They do not measure atomic charges. Probably t= he confusion has been propagated by calling Mulliken and Lowdin populations= as types of "atomic charges", but really the Mulliken and Lowdin= populations cannot be atomic charges, because they have no defined mathema= tical limits. In the future, I shall try to avoid referring to Mulliken and= Lowdin populations as types of atomic charges, because I think this error = is responsible for the confusion surrounding the definition of atomic charg= es. While we may not be able to measure atomic charges as precisely as ener= gies in experiments, it is not true to say atomic charges are not experimen= tally observable. They can be observed and m!
=C2=A0easured through spectroscopy experiments, albeit wit= h much less precision than we are able to measure energies. I could go into= more extensive details and examples if you are interested.



-=3D This is automatically added to each message by= the mailing script =3D-
E-mail to subscribers: CHEMISTR= Y/a\ccl.net or use:
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/cgi-bin/ccl/s= end_ccl_message

E-mail to administrators: CHEM= ISTRY-REQUEST/a\ccl.net or use
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/cgi-bin/ccl/s= end_ccl_message

Subscribe/Unsubscribe:
=C2=A0 =C2=A0 =C2=A0 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/chemist= ry/announcements/conferences/

Search Messages: http://www.ccl.net/chemistry/sear= chccl/index.shtml
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/spammers.txt

RTFI: http://www.ccl.net/chemistry/aboutccl/ins= tructions/



--001a11400d6a68dd4f051f6d6d57-- From owner-chemistry@ccl.net Thu Sep 10 22:48:00 2015 From: "Thomas Manz thomasamanz:_:gmail.com" To: CCL Subject: CCL:G: Case Studies of QM Computational Chemistry in Reactivity Message-Id: <-51703-150910224035-7011-Q6RCHEoK+Y0aoa52cA7QHQ|*|server.ccl.net> X-Original-From: Thomas Manz Content-Type: multipart/alternative; boundary=001a1136e03a70fe58051f6fa2aa Date: Thu, 10 Sep 2015 20:40:10 -0600 MIME-Version: 1.0 Sent to CCL by: Thomas Manz [thomasamanz]*[gmail.com] --001a1136e03a70fe58051f6fa2aa Content-Type: text/plain; charset=UTF-8 Stephen, >The word "observe" in our context can only mean "observable" in a QM sense. Hence, because there is no operator for "atomic charge" an observable atomic charge does not exist in a strict sense. You probably mean correlations of spectroscopic signatures with atomic charges when writing "They can be observed and measured through spectroscopy experiments". If you have another opinion on that I would like to know more details on how to measure atomic charges. Yes, the experimental information we have about the net charges of individual atoms in materials comes from spectroscopic signatures and other related experimental data. The N**C60 endohedral complex is an exemplary case. For this complex, the spectroscopy of the N atom is a quartet that closely resembles the quartet state of an isolated N atom, with some tiny shifts in the peak positions. The N**C60 spectroscopic signature shows that the N atom resides in the middle of the C60 cage and that it has an electron configuration almost identical to the isolated N atom. [Weiden, N.; Kass, H.; Dinse, K. P., Pulse electron paramagnetic resonance (EPR) and electron-nuclear double resonance (ENDOR) investigation of N**C-60 in polycrystalline C-60. *J. Phys. Chem. B * *1999,* *103*, 9826-9830.] Therefore, we know that charge transfer between the N atom and the C60 cage is minimal. This implies the net atomic charge of N in N**C60 is approximately zero. Using this information, it is possible to apply the scientific method to different charge assignment schemes. If a charge assignment scheme gives a net atomic charge of +3.567 for this N atom, this result is clearly wrong. On the other hand, if a charge assignment scheme gives a net atomic charge of +0.07 electrons for this N atom, this results is reasonably consistent with experiment. The results of one system alone is not sufficient to make conclusions about the overall accuracy of a method for assigning net atomic charges. But, if we repeat the above kinds of experiments and analysis for many different types of materials, then we can make meaningful assessments about which methods for assigning net atomic charges are accurate and which are not (compared to experiments). Rather than going through an extensive set of additional experimental examples here, I will update the CCL when a journal article on this topic that I'm co-authoring becomes published. This article will discuss numerous experimental comparisons of net atomic charges in depth and should be of immense interest to CCL readers. Sincerely, Tom On Thu, Sep 10, 2015 at 6:02 PM, Thomas Manz thomasamanz-*-gmail.com < owner-chemistry**ccl.net> wrote: > Hi Stefan, > > In regards to your questions about MP2, one has to be extremely careful > with such an approach, because the denominator is of the form (energy_1 - > energy_2) which causes the denominator to become zero when energy_1 = > energy_2. This can cause perturbation methods to blow up. For this reason, > I generally prefer non-perturbative methods such as CCSD, when a > higher-level calculation result is needed. I would recommend CCSD as > opposed to MP2, simply because CCSD has a more well-defined mathematical > limit on the results of the calculation. Personally, I don't use MP2 > calculations for this reason, but this doesn't necessarily mean others > can't. However, I wouldn't go so far as to say that MP2 calculations don't > have a well-defined basis set limit. I believe that for most systems the > complete basis set limit would be well-defined for MP2 calculations. In > this sense, the MP2 calculations are much more well-defined than Mulliken > or Lowdin populations, which definitely do not have a basis set limit. > > > If the set is small (minimal) the derived atomic charges are chemically > reasonable and correlate well with those from other methods for well > understood reasons. > > The populations of the density matrix projected onto a smaller basis set > is usually referred to by a different name. At least in Gaussian programs, > it is called Pop=MBS. In Gaussian programs, this is a different algorithm > than Pop=Regular which performs Mulliken analysis in the current basis set. > In my experience, the Pop=MBS method is not very useful and tends to crash > a large percentage of the time. It seems to crash especially often for > heavier atoms and for those with pseudopotentials. Also, people have tested > the idea to project plane-wave basis sets onto minimal localized atomic > orbital basis sets, but this results in charge leakage where the density > matrix in the smaller basis set does not accurately represent the true > density matrix. In general, the small basis sets do not represent the > density matrix with high accuracy. Therefore, in general, I cannot > recommend the approach you mentioned. There are certainly much better > approaches if the goal is to compute net atomic charges. > > Best, > > Tom > > > > On Thu, Sep 10, 2015 at 2:04 PM, Stefan Grimme grimme,,thch.uni-bonn.de < > owner-chemistry]~[ccl.net> wrote: > >> >> Sent to CCL by: "Stefan Grimme" [grimme|*|thch.uni-bonn.de] >> Dear Tom, >> I followed this discussion quietly for some time but now can't resist to >> comment on this too extreme viewpoint: >> >> 1. Methods can be useful and reasonable without a definite mathematical >> limit. A Mulliken or Loewdin population analysis gives a definite result >> for a given well-defined AO basis set. If the set is small (minimal) the >> derived atomic charges are chemically reasonable and correlate well with >> those from other methods for well understood reasons. I don't want to >> defend orbital based partitionings (I prefer observables) but making the >> mathematical limit >> to the encompassing requirement seems nonsense to me. >> There are other useful and widely used QC methods like Moeller-Plesset >> perturbation theory which are often divergent (or at least convergence is >> unlcear) in large one-particle basis sets and hence also do not have a >> definite mathematical limit. Is this a good reason to abandon all MP2 >> calculations? >> >> 2. The word "observe" in our context can only mean "observable" in a QM >> sense. Hence, because there is no operator for "atomic charge" an >> observable atomic charge does not exist in a strict sense. You probably >> mean >> correlations of spectroscopic signatures with atomic charges when writing >> "They can be observed and measured through spectroscopy experiments". >> If you have another opinion on that I would like to know more details on >> how to measure atomic charges. >> >> >> Best wishes >> Stefan >> >> >Hi Peeter, >> >> >There is a fundamental distinction between the current conversation >> focused on exchange-correlation theories and basis sets and the earlier >> discussion focused on atomic properties. If one increases the basis set >> size, exchange-correlation functionals such as B3LYP, M06, or whatever one >> you care to use will approach a well-defined mathematical limit. We can >> then discuss what the relative accuracy of that mathematical limit is in >> comparison to experimental properties and also discuss how close we are to >> that mathematical limit with a particular basis set. Thus, it is meaningful >> to discuss how adequate an exchange-correlation theory or basis set are for >> a particular research problem. Of course, the goal is to choose an adequate >> level that is not too computationally expensive for the particular research >> question being studied. >> >> >In contrast, Mulliken and Lowdin population analysis schemes do not have >> any defined mathematical limits. As the basis set is increased and the >> energy and electron density approach the complete basis set limit, the >> Mulliken and Lowdin populations behave erratically and blow up. This is how >> we know for sure that Mulliken and Lowdin population analysis schemes are >> utter nonsense and should never be used for publication results. As pointed >> out by one person, their only purpose is for debugging calculations to see >> if the symmetry or other basic features of the input geometry are malformed. >> >> >It is not the earlier discussion on atomic charges that is "nonsense" >> but rather the Mulliken and Lowdin populations that are nonsense, because >> they have no defined mathematical limits. This has nothing to do with >> atomic charges, per se. The Mulliken and Lowdin populations do not measure >> anything physical. They do not measure atomic charges. Probably the >> confusion has been propagated by calling Mulliken and Lowdin populations as >> types of "atomic charges", but really the Mulliken and Lowdin populations >> cannot be atomic charges, because they have no defined mathematical limits. >> In the future, I shall try to avoid referring to Mulliken and Lowdin >> populations as types of atomic charges, because I think this error is >> responsible for the confusion surrounding the definition of atomic charges. >> While we may not be able to measure atomic charges as precisely as energies >> in experiments, it is not true to say atomic charges are not experimentally >> observable. They can be observed and m! >> easured through spectroscopy experiments, albeit with much less >> precision than we are able to measure energies. I could go into more >> extensive details and examples if you are interested.>> E-mail to subscribers: CHEMISTRY]~[ccl.net or use:>> >> E-mail to administrators: CHEMISTRY-REQUEST]~[ccl.net or use>> >> >> > --001a1136e03a70fe58051f6fa2aa Content-Type: text/html; charset=UTF-8 Content-Transfer-Encoding: quoted-printable
Stephen,

>The word "observe" in our context can only mea= n "observable" in a QM=C2=A0sense. Hence, because there is no operator for "atomic = charge" an
observable atomic charge does not exist = in a strict sense. You probably mean=C2=A0correlations of spectroscopic signatures with atomic cha= rges when writing
"They can be observed and measure= d through spectroscopy experiments".=C2=A0If you have another opinion on that I would like to= know more details on
how to measure atomic charges.

<= /div>
Yes, the experimenta= l information we have about the net charges of individual atoms in material= s comes from spectroscopic signatures=C2=A0
and other related experimental data.=C2=A0The N**C60 endohedral comp= lex is an exemplary case. For this complex, the spectroscopy of the=C2=A0
N atom is a qu= artet that closely resembles the=C2=A0quartet state of an isolated N atom, with some tiny shifts i= n the peak positions.=C2=A0
The=C2= =A0N**C60=C2=A0spectroscopic signature shows th= at the N atom=C2=A0resi= des in the middle of the C60 cage and that it has an electron=C2=A0<= /div>
configuration almost= identical to the isolated N atom.=C2=A0
[Weiden, N.; Kass, H.; Dinse, K. P., Pul= se electron paramagnetic resonance (EPR) and electron-nuclear double resonance=C2=A0
(ENDOR) investigation of N**C-60 in polycrystalline C-60. J. Phys. Chem. B 1999, 103, 9826-9830.]

Therefore, we=C2=A0know that charge transfer between the= N atom and the C60 cage is minimal.=C2=A0
This implies the net atomic charge of N in N**= C60 is approximately zero.=C2=A0Using this information, it is possible to apply the scientific met= hod to=C2=A0
= different charge assignment schemes. If a charge assignment scheme gives=C2= =A0a net atomic charge = of +3.567 for this N atom, this result is clearly wrong.=C2=A0
=
On the other hand, if a c= harge assignment scheme gives a=C2=A0net atomic charge of +0.07 electrons for this N atom,<= /div>
=C2=A0this results i= s reasonably consistent with experiment.

The results of one system alone is not sufficient to ma= ke conclusions about the overall accuracy of a method for=C2=A0
assigning net atomic cha= rges. But, if we repeat the above kinds of experiments and analysis for man= y different types of materials,=C2=A0
then we can make meaningful assessments about whic= h methods for assigning net atomic charges are accurate and=C2=A0
which are not (compare= d to experiments).

Ra= ther than going through an extensive set of additional experimental example= s here, I will update the CCL when a journal article=C2=A0on this topic
that I'm co-authoring=C2=A0<= span style=3D"font-size:12.8000001907349px">becomes published. This article= will discuss numerous experimental comparisons of net atomic charges
in depth and shoul= d be of=C2=A0immense in= terest to CCL readers.

Sincerely,

Tom

On Thu, Sep 10, 2015 at 6:02 PM, Thomas Manz thomasamanz-*-gmail.com <owner-chemistry**ccl.net> wrote:
Hi Stefan,
In regards to your questions about MP2, one has to be extremely= careful with such an approach, because the denominator is of the form (ene= rgy_1 - energy_2) which causes the denominator to become zero when energy_1= =3D energy_2. This can cause perturbation methods to blow up. For this rea= son, I generally prefer non-perturbative methods such as CCSD, when a highe= r-level calculation result is needed. I would recommend CCSD as opposed to = MP2, simply because CCSD has a more well-defined mathematical limit on the = results of the calculation. Personally, I don't use MP2 calculations fo= r this reason, but this doesn't necessarily mean others can't. Howe= ver, I wouldn't go so far as to say that MP2 calculations don't hav= e a well-defined basis set limit. I believe that for most systems the compl= ete basis set limit would be well-defined for MP2 calculations. In this sen= se, the MP2 calculations are much more well-defined than Mulliken or Lowdin= populations, which definitely do not have a basis set limit.
> If the set is= small (minimal) the derived atomic charges are chemically reasonable and c= orrelate well with those from other methods for well understood reasons.

=
The populations of = the density matrix projected onto a smaller basis set is usually referred t= o by a different name. At least in Gaussian programs, it is called Pop=3DMB= S. In Gaussian programs, this is a different algorithm than Pop=3DRegular w= hich performs Mulliken analysis in the current basis set. In my experience,= the Pop=3DMBS method is not very useful and tends to crash a large percent= age of the time. It seems to crash especially often for heavier atoms and f= or those with pseudopotentials. Also, people have tested the idea to projec= t plane-wave basis sets onto minimal localized atomic orbital basis sets, b= ut this results in charge leakage where the density matrix in the smaller b= asis set does not accurately represent the true density matrix. In general,= the small basis sets do not represent the density matrix with high accurac= y. Therefore, in general, I cannot recommend the approach you mentioned. Th= ere are certainly much better approaches if the goal is to compute net atom= ic charges.
<= br>
Best,

Tom



On Thu, Sep 10, 2015 at 2:04 PM, St= efan Grimme grimme,,t= hch.uni-bonn.de <owner-chemistry]~[ccl.net> wrot= e:

Sent to CCL by: "Stefan=C2=A0 Grimme" [grimme|*|thch.uni-bonn.de]
Dear Tom,
I followed this discussion quietly for some time but now can't resist t= o
comment on this too extreme viewpoint:

1. Methods can be useful and reasonable without a definite mathematical lim= it. A Mulliken or Loewdin population analysis gives a definite result for a= given well-defined AO basis set. If the set is small (minimal) the derived= atomic charges are chemically reasonable and correlate well with those fro= m other methods for well understood reasons. I don't want to defend orb= ital based partitionings (I prefer observables) but making the mathematical= limit
to the encompassing requirement seems nonsense to me.
There are other useful and widely used QC methods like Moeller-Plesset
perturbation theory which are often divergent (or at least convergence is unlcear) in large one-particle basis sets and hence also do not have a
definite mathematical limit. Is this a good reason to abandon all MP2
calculations?

2. The word "observe" in our context can only mean "observab= le" in a QM
sense. Hence, because there is no operator for "atomic charge" an=
observable atomic charge does not exist in a strict sense. You probably mea= n
correlations of spectroscopic signatures with atomic charges when writing "They can be observed and measured through spectroscopy experiments&qu= ot;.
If you have another opinion on that I would like to know more details on how to measure atomic charges.


Best wishes
Stefan

>Hi Peeter,

>There is a fundamental distinction between the current conversation foc= used on exchange-correlation theories and basis sets and the earlier discus= sion focused on atomic properties. If one increases the basis set size, exc= hange-correlation functionals such as B3LYP, M06, or whatever one you care = to use will approach a well-defined mathematical limit. We can then discuss= what the relative accuracy of that mathematical limit is in comparison to = experimental properties and also discuss how close we are to that mathemati= cal limit with a particular basis set. Thus, it is meaningful to discuss ho= w adequate an exchange-correlation theory or basis set are for a particular= research problem. Of course, the goal is to choose an adequate level that = is not too computationally expensive for the particular research question b= eing studied.

>In contrast, Mulliken and Lowdin population analysis schemes do not hav= e any defined mathematical limits. As the basis set is increased and the en= ergy and electron density approach the complete basis set limit, the Mullik= en and Lowdin populations behave erratically and blow up. This is how we kn= ow for sure that Mulliken and Lowdin population analysis schemes are utter = nonsense and should never be used for publication results. As pointed out b= y one person, their only purpose is for debugging calculations to see if th= e symmetry or other basic features of the input geometry are malformed.

>It is not the earlier discussion on atomic charges that is "= ;nonsense" but rather the Mulliken and Lowdin populations that are non= sense, because they have no defined mathematical limits. This has nothing t= o do with atomic charges, per se. The Mulliken and Lowdin populations do no= t measure anything physical. They do not measure atomic charges. Probably t= he confusion has been propagated by calling Mulliken and Lowdin populations= as types of "atomic charges", but really the Mulliken and Lowdin= populations cannot be atomic charges, because they have no defined mathema= tical limits. In the future, I shall try to avoid referring to Mulliken and= Lowdin populations as types of atomic charges, because I think this error = is responsible for the confusion surrounding the definition of atomic charg= es. While we may not be able to measure atomic charges as precisely as ener= gies in experiments, it is not true to say atomic charges are not experimen= tally observable. They can be observed and m!
=C2=A0easured through spectroscopy experiments, albeit with much less= precision than we are able to measure energies. I could go into more exten= sive details and examples if you are interested.



-=3D This is automatically added to each message by the mailin= g script =3D-
E-mail to subscribers: CHEMISTRY]~[ccl.net or use:
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/cgi-bin/ccl/s= end_ccl_message

E-mail to administrators: CHEMISTRY-REQUEST]~[ccl.net or use
=C2=A0 =C2=A0 =C2=A0 http://www.ccl.net/cgi-bin/ccl/s= end_ccl_message

Subscribe/Unsubscribe:


--001a1136e03a70fe58051f6fa2aa-- From owner-chemistry@ccl.net Thu Sep 10 23:23:01 2015 From: "Andreas Klamt klamt*|*cosmologic.de" To: CCL Subject: CCL: Case Studies of QM Computational Chemistry in Reactivity Message-Id: <-51704-150910214648-32720-VoP6GyFL90BNUTMsL2JyWQ[a]server.ccl.net> X-Original-From: Andreas Klamt Content-Transfer-Encoding: 8bit Content-Type: text/plain; charset=utf-8 Date: Fri, 11 Sep 2015 03:46:43 +0200 MIME-Version: 1.0 Sent to CCL by: Andreas Klamt [klamt,cosmologic.de] Hi all, now let me put my 5 cents into the discussion, having spent 12 years as computational chemist at Bayer, finally as head of the group, an being in close contact with many comp. chem. group in industry all over the world since more than 25 years now: I consider it as one of the greatest successes and values of computational chemistry in industry to rule out the impossible and leave a smaller set of potentially doable alternatives. We can rarely decide whether a reaction will really be feasible, because our error bars are much too large for that. But we can often decide that a reaction will be impossible in the proposed way, because the barrier is much to high, even taking into account the computational uncertainties. And such a decision can save a lot of time and money in industry. But such success will not be published or communicated as big successes, because the colleague who came to you with the question, will not be enthusiastic if you post the fact, that he came with an idea to you that turned out to be completely unrealistic. Therefore industrial success cases often are of the kind: Within a set of alternatives, find those which we should try in the lab and rule out those which are clearly impossible, e.g. suggest a ranked list of solvents for a certain reaction or separation, so that the experimentalist can focus on the top 20 or so. If you are lucky, there may be an unexpected candidate under the top 20, and the experimentalist then test it and finds it as very good. That would be considered as a success case. But most often you are just narrowing the choices for the experimentalist. If the colleague is fair he will admit that this was helpful. Or he may say that he new those candidates upfront. Andreas Am 10.09.2015 um 15:50 schrieb Jerome Kieffer Jerome.Kieffer./a\.terre-adelie.org: > > Sent to CCL by: Jerome Kieffer [Jerome.Kieffer*o*terre-adelie.org] > On Thu, 10 Sep 2015 07:26:41 +0100 > "Peter Jarowski peterjarowski===gmail.com" wrote: > >> I ask a simple question and further ask everyone to pretend they are at an >> interview outside of an academic environment where you will find willing >> and interested parties who enjoy the alphabet soup of DFT. what example and >> what statistics would you present to justify your work. Clearly, you are >> talking to experimentalists, as theory does not exist without them. >> >> How does theory drive experiment? >> How much money does it save? >> How much revenue can it produce? >> How reliable is it? >> How essential is it and what can it do that experiment can not? > > Hi all, > > I had the opportunity to do computational chemistry as part of a > (large) industrial group, a decade ago. > Such questions were often asked and it was not easy to be honest. > > When speaking of reactivity, everything is about the energy barrier > between two (or more) path which leads to different products. > Some critical steps may be mono, the other bi-molecular, so it is > harder to compare them... > Then comes the enthalpic and entropic contribution, not speaking about > solvatation for spices that do not exist per-se as they are transition > states. > > I left this job a while ago and I have to admit the "correctness" of > most result I obtained at that time were mainly "lucky error cancelation": > Concidering : > A --> B > A --> C > in kinetic condition, let AB and AC be the transition states: > [C]/[B] = exp(-(G[AC*]-G[AB*])/RT) > > The error in the exponential is 2x the error in calculating the free > energy of one TS, one ends with a huge error bars which is anything but > conclusive at that time (one error was about 6kcal/mol ... while with > ony 1kcal/mol error, no conclusion was possible) . > > Cheers, > -- -------------------------------------------------- Prof. Dr. Andreas Klamt CEO / Geschäftsführer COSMOlogic GmbH & Co. KG Imbacher Weg 46 D-51379 Leverkusen, Germany phone +49-2171-731681 fax +49-2171-731689 e-mail klamt/a\cosmologic.de web www.cosmologic.de [University address: Inst. of Physical and Theoretical Chemistry, University of Regensburg] HRA 20653 Amtsgericht Koeln, GF: Prof. Dr. Andreas Klamt Komplementaer: COSMOlogic Verwaltungs GmbH HRB 49501 Amtsgericht Koeln, GF: Prof. Dr. Andreas Klamt