Talk:Ab initio quantum chemistry methods

Unbiased

This page seems to be an advertizement for Fritz Schaefer and his research. What happened to un-biased, reasoned, educational discussion (referencing papers other than your own)? C'mon guys. This is a cheap way to get your name out!—The preceding unsigned comment was added by 130.202.94.248 (talk • contribs).

I have removed the section from the article and pasted it below. It appears indeed that this needs to be rewritten. --Dirk Beetstra ^{T C} 16:22, 20 August 2007 (UTC)

I have done a bit of a search. That all references in the example are from one group is unrelated to the editor who added the example originally. Though the rest of the article is unreferenced, the example is indeed the only part that is referenced (I deleted the pasted paragraph) --Dirk Beetstra ^{T C} 16:33, 20 August 2007 (UTC)

I moved the example from Computational chemistry where it had been for a long time. This was because nobody added examples to other sections, if they had done so it would have made the article too long, and the GA folks criticized the example when they delisted that article as a GA. Schaefer has made something of a thing about making predictions from theory that are later confirmed by experiment. The first was of course the prediction that methylene was bent and not linear which forced Hersberg to go back and check his experiment and to find that he was originally wrong. I wanted an example for a larger molecule and Si₂H₂ seems a classic case. I added his photo because the GA article folks suggested more photographs on Computational chemistry and it is not easy to get free photographs of people. As to my involvement with this, I did learn about these calculation in 1990 when I spent some time in the Schaefer group. Later I collaborated on calculations on SiGe₂ and AlGa₂ but they are not mentioned in this article. Fritz Schaefer has certainly not suggested that the example is on this article. I am in touch with him and the last time was to pursuade him to not edit his own article and to allow that to be NPOV. If people can come up with another example that really shows ab initio theory at its very best, we can either add it or use it to replace this example. As for references, I keep intending to add more. I'll try to do that today. --Bduke 22:30, 20 August 2007 (UTC)

As someone who uses these methods almost daily, I was highly disturbed by the example used in this paper. First of all, this is not a prominent or historically important group in the field, secondly the example didn't justify the inclusion of a picture of the group leader (why was that?), third of all the group leader in question is notable mostly for his opinions on creationism, not on ab initio methods. It is very very clear that this example doesn't belong here. There are many textbook examples available that can be used without having to cite someone's own paper. I don't know who put it here and why, but please refrain from advertizing your work or the work of your favorite creationist scientist in an article that is supposed to enlighten people on ab-initio quantum chemistry methods. Mipmip (talk) 18:49, 10 February 2008 (UTC)

Please take your personal attacks of Schaefer elsewhere. There is no way to verify your claim that you "use these methods almost daily," but your claim that Schaefer and his research group are not prominent or historically important in the field of ab initio quantum chemistry is demonstrably false. Logically, that means that you are either deliberately lying or don't know what you are talking about. Schaefer is one of the most highly cited chemists in the world. He has received numerous prestigious awards for his role in the development of ab initio methods and their application. In 1979, he received theAmerican Chemical Society Award in Pure Chemistry "for the development of computational quantum chemistry into a reliable quantitative field of chemistry and for prolific exemplary calculations of broad chemical interest." He received the 2003 American Chemical Society Award in Theoretical Chemistry "for his development of novel and powerful computational methods of electronic structure theory, and their innovative use to solve a host of important chemical problems"). In 2003 he also received the annual American Chemical Society Ira Remsen Award "for work that resulted in more than one hundred distinct, critical theoretical predictions that were subsequently confirmed by experiment and for work that provided a watershed in the field of quantum chemistry, not by reproducing experiment, but using state-of-the-art theory to make new chemical discoveries and, when necessary, to challenge experiment." The quotes are taken from the citations that went along with the awards. Anyone who cares to do so can look them up in the appropriate back issues of Chemical and Engineering News (the offical publication of the ACS).Mosquito bait (talk) 04:49, 9 August 2008 (UTC)

Oh, dear. Where do I start. First, Schaefer is most certainly prominent in the field of ab initio quantum chemistry. He has well over 1000 publications in the field with many in the Journal of the American Chemical Society and Journal of Chemical Physics, with the rest almost entirely in high impact journals. He has been awarded many Prizes. He is a member of the highly regarded International Academy of Quantum Molecular Science and has been President of the World Association of Theoretical and Computational Chemists. Please do not insult one of the leaders of our field because you do not like creation science. Your statement "this is not a prominent or historically important group in the field" is just false. In science, he is notable for his work in ab intio theory and practice. Most scientists do not even know he is a supporter of ID. Second, I added the photograph when this example was in Computational Chemistry before I moved the example here as others had not taken the bait and added examples for other parts of that article. I added it because the article was criticized for having no photographs by the good article reviewers. Third, Schaefer is sort of a creation scientist (he does not believe in a short age for the earth, but is doubtful about evolution), but I and very many other scientists respect him for his science while hating his views on evolution. His group is not mine, although I did spend 5 months in his group on sabbatical in 1990. Schaefer is well know for doing calculations in his group that force experimentalists to look at their evidence again. Most famously, he told Herzberg that he was wrong that triplet methylene was linear and predicted it had an angle of about 140^o. Schaefer was right. This example is similar, but for a larger molecule which is more appropriate for today.. A surprising result forced the experimentalist to look again at the data and they found the unusual monobridged structure. The work is not all by the Schaefer group, but that is not clear in the article. The early discovery of the butterfly ring structure was by other groups. That should be mentioned. It is not "very very clear that this example doesn't belong here". It has been on WP for quite a long time with only minor discussion on it as above. However, I wait for the views of others. Perhaps in the meantime, you could suggest an example from a textbook which does as least as well as this example in showing the power of ab initio methods to predict new chemistry, not just get high accuracy for properties. The real issue is whether we have examples at all. Choosing one, means you do not chose others and this could be seen as not NPOV. On the other hand, examples can help to show people what the method is capable of and where it helps the general chemist. I am not sure how to resolve this, as it is not simple. --Bduke (talk) 01:01, 11 February 2008 (UTC)

Look, I'm not the first person to object to this, so I don't think I'm unreasonable here. At none of the other chemistry articles you will find a picture of someone who didn't invent the technique or won Nobel Prize for it (Ideally, these should be the same), And, if this article needs a picture, why not add the output of an ab-initio calculation instead of a staff member of a group you happened to visit. Like this, for example: http://nobelprize.org/nobel_prizes/chemistry/laureates/1998/press.html You did take an effort writing down this example, maybe a bit too detailed, but it's not a trivial task so I suggest we leave that until someone comes up with a clearer example. I and others would however be highly content if you would remove the picture of this group leader, as it creates a visual link between this guy and ab-initio quantum chemistry methods, which is very unfair towards ALL other ab-initio groups around the earth. Do you agree with me that this is indeed unfair to other science groups? And could you therefore please remove the picture, since you were the person who put it there and reverting won't bring any of us any further? Thanks! Mipmip (talk) 23:32, 12 February 2008 (UTC)

Again, Mipmip is either deliberately lying or simply doesn't know what he is talking about. Please take your personal attacks of Schaefer elsewhere. You may not like his views on ID, but you cannot expect to post false information about his work in the field of ab inito quantum chemistry and get away with it. In fact, the Schaefer Group has published landmark papers on the development and efficient implementation of most of the ab initio methods listed in this article (HF, ROHF, CI, CC, MRSCF, and MRCI). That's one reason why Schaefer is one of the most highly cited chemists in the world (see link above). Schaefer's publication list is a matter of public record, and anyone can find these landmark papers if they search the list. For example, see publications 250, 251, 254, 255, 363, 381, etc. That doesn't mean that Schaefer's picture should be included here, however. It simply isn't possible to point to a single individual as the "inventor" of any of these methods. Mosquito bait (talk) 04:57, 9 August 2008 (UTC)

OK, I do not want to upset people and have an edit war. I have therefore removed the image. I note however that you do not recognize my other arguments and that you were quite wrong in describing the Schaefer Group as not prominent and him as my favorite creationist. I am not clear that it it is unfair to other ab initio groups. I have not seen that argument before and I think many ab initio groups would think it is quite appropriate. The Good Article Review people like photographs. It would be nice to have a photograph of Pople and I may in time be able to get a free image of him. I note however, that while he won the Nobel Prize, he did not invent ab initio quantum chemistry. It would be nice to have images of several people on this article and related articles. At that point, the image of Schaefer might return. There is no doubt that he did invent some of the most useful methods for doing configuration interaction and has done some of the most significant calculations. He did not win the Nobel Prize and it is irritating that the Discovery Institute goes on about him being nominated when such information is supposed to be private. I have however no doubt that he was close to joining the pair who did win the 1998 Prize. Can we let it rest now? --Bduke (talk) 00:44, 13 February 2008 (UTC)

Also, with respect to Schaefer's notability, here's a member bio on him from the International Academy of Quantum Molecular Science that was pointed out on Talk:Henry F. Schaefer, III that will provides a brief summary of his career if interested.--Roswell native (talk) 02:47, 13 February 2008 (UTC)

Example: Is Si2H2 like acetylene (C2H2)?

Hope I don't start an edit war, but I restored the Si2H2 example section. First, it's a good example of the power of ab initio computations and helps illustrate as such in the article. If you don't like the example, feel free to write up a better one and suggest it as a replacement here on the talk page rather than just delete the one that has existed for quite some time. Second, whether you like Fritz Schaefer or not, folks in the field will undeniably agree that his group is prolific. My $.02. --Roswell native (talk) 01:41, 11 February 2008 (UTC)

What about Density Functional Theory?

Yes, I am aware of the typical chemist's objections that DFT is not ab initio because the commonly used functionals are parametrized. However, as we all (should) know, DFT as a theory is accurate (unlike HF without MRCI), assumes nothing, and needs no parameters. In my opinion, HF without as such is no more ab initio than LDA, because the correlation effects are neglected, and thus one starts from an approximate Hamiltonian. In my opinion, the description of HF et al is much too detailed, and should be accompanied by theories obtained from other variational principles, too (i.e. DFT, and of course Green's function methods). Accum ulate (talk) 05:44, 10 June 2009 (UTC)

I agree that there should be a section on Density functional theory linking to the larger article on that topic. The description of HF is needed because all the other methods build on HF. --Bduke (Discussion) 07:00, 10 June 2009 (UTC)

As a follow-up. the method that is used in practice, overwhelming more than any other method is B3LYP. That does have parameters. B3LYP is about 80% of all published DFT papers. --Bduke (Discussion) 07:03, 10 June 2009 (UTC)

For now I have added DFT to the See also section so that it is at least mentioned in this article. I agree however that a brief section linking to DFT would be preferable, if someone will make the effort of writing an appropriate text. Dirac66 (talk) 12:56, 10 June 2009 (UTC)

Definition of ab initio

I have remembered why DFT is excluded from ab initio. Levine's Quantum Chemistry (4th edn 1991, p.455) introduces an ab initio calculation as one which uses the correct Hamiltonian and does not use experimental data. The variational theorem shows that the expectation value of the correct Hamiltonian over an approximate wavefunction is an upper bound to the true energy.

I disagree with Accumulate's statement above that "HF without as such is no more ab initio than LDA, because the correlation effects are neglected, and thus one starts from an approximate Hamiltonian." The neglect of correlation means that the wavefunction is approximate, but the energy expectation value of the approximate wavefunction is calculated with the correct Hamiltonian and so the variational theorem applies and the calculation is ab initio.

For DFT on the other hand, even if there are no parameters the Hamiltonian is approximate, so the calculation is not variational and not classed as ab initio. Of course this is considered less important than it used to be, because of modern improvements in DFT accuracy. I suggest that we mention DFT as a method which is of comparable (and sometimes better) accuracy to ab initio methods. Dirac66 (talk) 01:40, 11 June 2009 (UTC)

I'm a little confused. Are you saying that a method being variational is what makes it ab initio? Or are you saying using an approximate Hamiltonian is? Jal173 (talk) 21:44, 19 February 2012 (UTC)

We do have to be careful about talking about the Hamiltonian. For example, Hartree-Fock is an exact solution to an approximate Hamiltonian (the sum of Fock operators), but is an approximate solution to the exact Hamiltonian. I think DFT is similar. This is why some people count DFT as ab initio and others do not. Hybrid DFT methods like B3LYP where terms are empirically fitted to a set of experimental values such as the G2 set, are clearly not ab initio, but semi-empirical. However it is clear that these days the boundary between ab initio and semi-empirical is fuzzy. Use of the variational theorem ensures the the energy is an upper bound to the exact energy. However, if we used this to define ab initio, we would exclude perturbation methods like MP2 and, I think, coupled cluster. --Bduke (Discussion) 07:25, 20 February 2012 (UTC)

Full CI

The following statement has been questioned today with a Citation Needed tag. "In particular, configuration interaction, where all possible configurations are included (called "Full CI") tends to the exact non-relativistic solution of the Schrödinger equation."

It should be noted that the articles on Configuration interaction and on Full configuration interaction both refer more cautiously to an exact solution within the space spanned by the one-electron basis set. Possibly the statement in this article needs the same qualification to be valid. I am not at all certain however, since I don't really know what is meant by an exact solution to a differential equation within a given linear space. Could someone clarify these points? Dirac66 (talk) 01:18, 27 October 2009 (UTC)

This does need some clarification. First, it can only be exact within the confines of the Born-Oppenheimer approximation. Second, then my understanding is that it is exact at the limit of a complete basis set. That of course has an infinite number of basis functions, so an infinite number of molecular orbitals and an infinite number of configurations in the CI. In practice is is not that bad. The use of the cc-pV6Z basis set for H2 is close to the exact result for full CI, but full CI here is just CISD as only single and double excitations are possible. So "tends" is correct if that means "as we go to the complete basis set limit". Note too that not all infinite basis sets are complete. --Bduke (Discussion) 02:02, 27 October 2009 (UTC)

Thank you. That is about what I remembered, though with more detail and more precision. Thank you. I have now revised this article to specify a complete basis set as well as the B-O approximation.

Unfortunately "complete set" does not seem to be defined on Wikipedia at the moment. There is an article on Completeness in a number of contexts, but it does not seem too helpful. Perhaps one of the mathematical definitions is what we need here? Dirac66 (talk) 01:05, 1 November 2009 (UTC)

The lede

I think that the problem with the lede can be simply solved by deleting the second paragraph. It adds nothing useful and is complex. What it is talking about might go elsewhere. --Bduke (Discussion) 21:27, 7 March 2015 (UTC)