Wikipedia:Scientific peer review/Computational chemistry
Review of Computational chemistry.
This article is being reviewed as part of the development process for Scientific peer review. Please help in reviewing the article and commenting on the process at Wikipedia Talk:Scientific peer review.
The intro is short but OK. But from than on nobody searching for a a good explenation for computational chemistry only gets the methodes which can also go into a small list at the end of the article.
Examples what computational chemistry is doing, are missing. There are a few words with no explenation in the intro giving no clue how and why you get IR-spectra or something else from a computer. For every major use there should be one section or a exapmle.
TS and IM for reactions or searching for the most stable conformer are exaples which can be shown in a small picture and everybody will get a picture what this stuff is about.
Then ther should be a description of why you do it at all. Most of the data you get from CC you can get by measurments. What are the benefits and the problems with CC. And beeing onest with problems is OK.
But for SPR this article is far more easy to comment than the aricle about science.
--Stone 07:57, 3 April 2006 (UTC)
The two most important questions to ask of an article are "is it scientifically correct?" and "is it complete?", so I start with these.
Is it correct? There are a number of problems, some arising from an attempt to simplify difficult concepts.
- The second paragraph under "Introduction" talks of one very accurate method. This is unclear, but the error is in stating that the cost rises factorially. Since this section is about an "in principle" method, it is talking about an exact solution to the Schrodinger equation. I have seen a reference that suggests that the exact solution scales as (N!)(N!). This is more than factorially. The question of scaling should be dealt with under different methods and this section here deleted. Computational chemistry is not even close to an exact solution so only a brief mention of the "in principle" method should be made.
- Under "electronic structure", this is described as the set of molecular orbitals as determined by solving the time-independent Schrödinger equation. This is false. The electronic structure is the total wave function, not the molecular orbitals and the latter are obtained from solving the Hartree-Fock equations.
- The Hartree-Fock method scales as N4 only in principle as that is how the number of 2-electron integrals scales. In practice it scales closer to N3, as all codes recognize zero or extremely small integrals and do not calculate them. This is minor point, but it is a common statement while all HF calculations scale rather better.
- The new image is excellent but unfortunately it includes an error. Full CI is inclusion of all configurations arising from a fixed finite basis set and thus is not the exact solution of the nonrelativistic equation. I think the term that is used is 'complete CI' which is the limit of 'full CI' as the basis set is increased, in the same way that the Hartree-Fock limit arises from a limit of HF energies as the basis set is increased.
Is it complete? Of course no article can be totally complete, but this article is missing a very important point.
- A vast majority of the computer time used by computational chemists is used to calculate derivatives of the total energy with respect to atomic coordinates and to use this to determine stationary points on the energy hypersurface, or to calculate the related second derivatives (and hence frequencies) to determine the nature of the stationary points. This applies to ab initio, semi-empirical and molecular mechanics. This process is usually called "geometry optimization". A superficial reading of the article suggests that this is entirely absent but it is hinted at. The prediction of molecular geometries for a series of isomers and the prediction of transition state structures for chemical reactions and the conversion between isomers is massively important and should be dealt with in detail and by links to other articles.
In his review, User:Stone, now copied above from the article's talk page, makes a number of useful suggestions. He also says "Most of the data you get from CC you can get by measurements". This comment is entirely understandable for an experimental chemist to deduce from this article, but it is false. Most of the data you get from CC is not available from experiment. For example, the predictions of structures for unstable isomers and transition states. This is why the article needs a major rewrite in this respect. Stone also argues for examples and he is absolutely right. I propose adding this example once the structure of the article has been modified.
Example.
A series of ab initio studies of Si2H2 shows clearly the power of computational chemistry. They go back over 20 years. Initially the question was whether Si2H2 had the same structure as ethyne (acetylene), C2H2. Slowly (because this started before geometry optimization was widespread), it became clear that linear Si2H2 was a transition structure between two equivalent trans-bent structures and that it was rather high in energy. The ground state was predicted to be a four-membered ring bent into a 'butterfly' structure with hydrogen atoms bridged between the two silicon atoms. Interest then moved to look at whether structures equivalent to vinylidene - Si=SiH2 - existed. This structure is predicted to be a local minimum, i. e. an isomer of Si2H2, lying higher in energy than the ground state but below the energy of the trans-bent isomer. Then surprisingly a new isomer was predicted by Brenda Colgrove in Henry F. Schaefer, III's group. This prediction was so surprising that it needed extensive calculations to confirm it. It requires post Hartree-Fock methods to obtain a local minimum for this structure. It does not exist on the Hartree-Fock energy hypersurface. The new isomer is a planar structure with one bridging hydrogen atom and one terminal hydrogen atom, cis to the bridging atom. Its energy is above the ground state but below that of the other isomers. Similar results were later obtained for Ge2H2 and SiGeH2. More interestingly, similar results were obtained for Al2H2 (and then Ga2H2 and AlGaH2) which have two electrons less than the Group 14 molecules. The only difference is that the four-membered ring ground state is planar and not bent. The cis-mono-bridged and vinylidene-like isomers are present. Experimental work on these molecules is not easy, but matrix isolation spectroscopy of the products of the reaction of hydrogen atoms and silicon and aluminium surfaces has found the ground state ring structures and the cis-mono-bridged structures for Si2H2 and Al2H2. Theoretical predictions of the vibrational frequencies were crucial in understanding the experimental observations of the spectra of a mixture of compounds. This may appear to be an obscure area of chemistry, but the differences between carbon and silicon chemistry is always a lively question, as are the differences between group 13 and group 14 (mainly the B and C differences). The silicon and germanium compounds were the subject of a Journal of Chemical Education article (B. J. DeLeeuw, R. S. Grev and H. F. Schaefer III, Vol 69, Page 441, 1992).
Other problems.
The other problems with this article are its organization and the absence in some cases of flow links from one paragraph to the next. It needs rewriting. I suggest that this rewrite removes some material which can go into new articles. Each main section should start with the line "For more information see name of new article". This certainly should apply to the ab initio and semi-empirical sections. I note in passing that the words "semi-empirical methods" are linked on many articles to the section of this article and that this is perhaps an indication that a separate article is needed. The "Linear scaling methods" paragraph needs adding to and putting into a separate article, as does the paragraph on "solid state physics".
The table of software packages has some structural problems that have raised recent issues about criteria for inclusion. There is a separate link to valence bond programs and the DFT article lists DFT software. The intention I believe was to list programs that include many different methods. I think this will be clarified if the column headed "periodic" is removed and these programs listed separately perhaps in an article that comes out of the solid state paragraph. Then only programs that show two or more "Y"s should be retained. Programs that are purely semi-empirical should be listed on the page that is created for semi-empirical. Programs that are purely DFT are already listed on the DFT page and should be removed from this list. The other lists should be linked as the VB and DFT lists are already above the table.
The section on chemical dynamics stands out as unrelated to the rest. I am not an expert on this area, but I suggest that the section should start with the use of ab initio methods to calculate the potential energy hypersurface in detail and provide a surface for classical dynamics. The work of Michael Collins in Canberra could feature here,
A few small points:
- I do not see Computational Chemistry as a branch of theoretical chemistry. I suggest the article starts with something like "Computational chemistry is a branch of chemistry that uses the results of theoretical chemistry incorporated into efficient computer programs to calculate the structures and properties of molecules".
- The "See also", "References" and "External links" sections need some thought. Do we, for example, list more journals other than J Comp Chem or list no journals? Are there books we should add? Is it right to list Schaefer's Group web site but not other groups?
I may add more later, but that for now uses the time I can sensibly give to this project.
--Bduke 03:01, 7 April 2006 (UTC)
A further point. The section on the Bader Atoms in Molecules approach needs broading to include many other approaches to interpreting or understanding the results of ab initio calculations. It needs to include something on Mulliken charges, localised orbitals, NBO method, etc. --Bduke 18:52, 9 April 2006 (UTC)
Comments on above reviews
[edit]Date of instable molecules are in fact hard to get, but most of the time computational chemists like me predict these numbers by computational methodes, knowing exactly that nobody will ever prove them wrong!
The Example you give is good and could be also a section in the article.
The numbers you get from computers should have some use, so state it somewhere in the article. Examples are the best methode for complexe things like computational chemistry. A good predicted mechanism for a reaction later verified by experimentalist whould show the effectivity of computational chemistry. --Stone 07:05, 7 April 2006 (UTC)
energy and interaction energy some of them you get from experimental data charges, dipoles and higher multipole moments are also possible to get frim experimental data vibrational frequencies or other spectroscopic quantitities use spectroscopy to get them reactivity do some kinetics cross sections for collision with other particles yes yes!
The point is that state clearly what are the benefits of CC and why is it better to use CC for predicting reactivity than to do 1000 experiments. The instable molecules are the best example for the benefits! When its instable you can not do experimental work and you should do CC. Transition states and exited states are also better researched with CC, because you get better results in less time.--Stone 07:13, 7 April 2006 (UTC)
Well I think much material has been recently added to the article. This material is usually of high value and should be kept but I regret the logical structure has been lost on the way. "Molecular structure" is a floppy concept which should be distinguished into "electronic structure" and "dynamics". The electronic structure is what is described by the electronic Hamiltonian and the dynamics correspond to the molecular Hamiltonian. I have more comment left and shall come back soon. Vb 19:57, 21 April 2006 (UTC)
In restructuring the article I tried to give emphasis and structure to the methods that have lead in the last decade or so to a massive increase in the references to computational chemistry in the chemistry literature. That was lost in the earlier version, which was very unclear for someone wanting to know what computational chemistry is all about. I was however, well aware that I had not done justice to dynamics. That section needs to be developed and I think it needs to start by linking into what is already there by talking about methods of dynamics that start from an ab initio potential energy surface. I am thinking in particular about the method of Michael Collins from Canberra. The emphasis should be on chemistry and we need to clear up the connection between this article and several more physics articles on molecular physics, Atomic, molecular, and optical physics and others in the categories of those names. There is a lot of confusion here. We also need to sort out what is to be left to the article on Molecular dynamics and what stays in the main article. I was trying to make what appears in the main article unstandable to a fairly wide audience leaving the more advanced detail to the linked sub-articles, but I doubt I succeeded. --Bduke 23:01, 21 April 2006 (UTC)