|McWeeny was born in 1924 in Yorkshire, England. He studied at Leeds and Oxford where he finished his thesis in 1948. He then became professor at Newcastle upon Thyne, Keele, Sheffield (all England) and Pisa (Italy). One should consult his highly enjoyable and very readable autobiography in the Intern. J. of Quantum Chemistry, 60, 3-19 (1996).|
Pisa, University, Thursday, September 9, 1993, 16:00 - 17:00
Anders: Professor McWeeny, would you think it worthwhile to write something historical, non-technical about the very early numerical approaches and methods of the Hückel theory and the following other semiempirical methods?
McWeeny: I think that could be very interesting because so many people just don't know about that period and in some sense Hückel theory is still relevant nowadays. For if you want a first approximation to molecular orbitals, for anything, you make first make very basic approximations to see if things look reasonable.
A: It is surprising that Hückel himself never followed up those things after his first papers. Do you happen to know any reasons why Hückel went out of this field?
McW: I really don't know why.
A: Do you have any suggestions where to look for an answer for this question?
McW: After the papers of the early thirties one heard little more of him.
A: Longuet-Higgins was another person who left the field ...
McW: That was much later. He went out of quantum chemistry altogether and became interested in artificial intelligence and that kind of thing. Which was surprising, because he hated computations. But this aspect of the use of computers, things like pattern recognition and so on, appeared to attract him.
A: An interesting point, to be sure. In the Hückel theory...
McW: In those times we were interested in the problems that could be solved without using computers and there were many problems which again became fashionable like one dimensional crystals, polymer chains and so on. I did calculations of the energy bands for polyphenyle and polyacines and so on in the early ´fifties. With the usual Hückel assumptions, you could get everything out in analytical form - energy levels, AO coefficients and so on ...
A: ... and that was the general scientific feeling then ...
McW: Yes, yes. I also got interested in the magnetic properties at the same time because you remember, Fritz London had some papers on diamagnetic susceptibilities of small ring systems and I was working at that time in a department doing research on cokes and other carboniferous materials and they were interested in their magnetic properties. I started doing calculations on large condensed ring systems. And with London's approximation - the simple extension of Hückel's theory for the magnetic field - you could get nice expressions for the magnetic susceptibilities (the diamagnetic susceptibility that is) and so on.
A: That was in the fifties?
McW: Yes, '52 - '55. They appeared in the Proceedings of the Physical Society so the chemists never looked into it. There is this sort of compartmentalization which is very bad for the development of any field.
A: What about the famous book of Streitwieser?
McW: Well, I think Streiwieser was more one of the "consumers" rather than the innovators in this field. People fall into those two categories. Not to say that both categories don't do valuable work! Streitwieser did a good job in selling molecular orbital theory to organic chemists - which was quite an uphill struggle in those days!
A: Actually, what was the separation between these two categories. Pullman for instance, who applied Hückel theory widely and wrote two books about it, seemed really to believe in it.
McW: Well, it's the only thing you can use for really big systems. Even nowadays, if you want to do some gigantic biologically interesting molecule ...
A: ... and other things like Buckminsterfullerens, for instance ...
McW: Of course, when you have simplifications due to symmetry you can always boil it down. But if you have no symmetry and thousands of atoms, thousands and thousands of orbitals, it's very difficult even nowadays.
A: What about the fact that oscillating calculations were often terminated at the most suitable point to the author's ideas?
McW: There is one anecdote here that might be of interest. At one of the first meetings I ever went to, I think it was in '57, both Pauling and Mulliken were present. It was a meeting arranged by Löwdin in Sweden. And at that meeting the phrase was coined :"The Pauling Point". That is the point at which - when you get oscillating results - you stop. Because Pauling always right was, he stopped at the right point!
A: Mulliken characterized Pauling to be a good salesman and a good showman ...
McW: Oh, splendid, yes, yes - quite the opposite of Mulliken.
A: Who else would you consider to be of interest in the context of these early theories?
McW: I think I would put Dewar in the same category at that time, as Streitwieser. All his early work, like Molecular Orbital Theory for Organic Chemists, was based on simple Hückel ideas. But I think he went further than Streitwieser because he introduced some rather nice things here, such as molecular orbital perturbation theory, and he would use it for dealing with quite large systems. I think he was one of the more imaginative people.
A: Wasn't there a problem as to who calculated which molecule first?
McW: Yes, I recall there once was a bit of a quarrel between Coulson and Streitwieser. Because Streitwieser was bringing out a book on the results of MO calculations. There were many collaborators; it was a big hefty book. Previous to that, Coulson had compiled a very nice but much more limited "dictionary" of results on a similar range of hydrocarbons and so on. But it was not published commercially. Streitwieser I think made no acknowledgement of that, of these compilations. But I think they solved it amicably in the end by bringing out two simultaneous volumes.
A: Would you think it possible to read between the lines in these early papers and books concerning conflicts of ideas, priorities and this sort of thing?
McW: Well, I think it's usually done by omission. There are many sins of omission. Sometimes of course the literature was not properly checked.
A: As it was in your case with the density matrices?
McW: (Laughing) Yes. In fact Löwdin and I brought out our density matrix papers at the same time, more or less, in 1955. But he published an enormous paper in Physical Review. I had published my results first as a technical report from the group at MIT, together with reports by Slater, Koster and others. And it had happened that Löwdin and I were at the MIT at the same time. So we talked about these things. (Laughing): I know he learned some things from me that he didn't know about, for instance the famous McDonald theorem, and the fact that it was Husimi who invented reduced density matrices.
A: Were there any other occurrences of this type in your career?
McW: There was another case that I resented very much at the time. Of course I was only a very young researcher in those days. I had just taken my first teaching job and I had been to Oxford, invited by Coulson who was working there. So I got talking to Mulliken, who was on leave of absence in St. John's College, and we had a long chat together. I didn't realize that he was soaking up everthing. I had published a small paper in 1951 in which I introduced what I called atom-and-bond charges. They could be obviously generalized, as I indicated, by breaking the charge density down into the atomic parts and the bonding parts. Anyway, these atom-and-bond charges became "populations" out of which population analysis developed in 1954-55. So it's always called nowadays "Mulliken's population analysis". In his first paper there was a footnote saying : "Dr. R. McWeeny had a similar idea" (or something of this kind) - and that was the end of the quote. But Mulliken was a very famous person and I was a nobody. But that sort of thing just happens. You are not going to print all this, I hope!?
A: There is a book by Coulson on "Hückel Theory for Organic Chemists".
McW: That was posthumous in fact, wasn't it? It was based on the lecture notes Coulson had been giving for many years to organic chemists.
A: Apparently he liked giving those lectures.
McW: Yes, he enjoyed them enormously.
A: How was the reception among organic chemists?
McW: Well, I think it was reasonably warm. Probably because he was giving the lectures to the young organic chemists. He wasn't giving them to the old boys who didn't want to look beyond their familiar horizons.
A: Another aspect. Was there at any time a competition between the so called free electron model and the Hückel theory?
McW: I don't think they were in competition. Although in fact there were notes published pointing out the connection between the free electron model and the Hückel theory. The free electron model was used extensively by Platt. For the case of one-dimensional chains, and even for two- and three-dimensional systems you can show that if you use an LCAO-type approximation then the coefficients follow a standing wave pattern. So there is a 1:1 correspondence between the free electron model and the Hückel orbitals.
A: Especially organic chemists working with dyestuffs seemed to like these early ideas.
McW: What people were looking for were the simplest possible mathematical models of the situation; and if you know that some electrons are mobile you can try to describe them first of all as free electrons and then put in the atoms and use a variational orbital approximation.
A: In those early days one can see that some very optimistic decimal places were used. Did the "consumers" really believe in this sort of accuracy?
McW: Well, I think we all believed in our approximations if they seemed to be physically reasonable. Even though we were not getting so many significant figures or anything quantitatively meaningfull we seemed to believe to get the right general trends in many issues.
A: The Pullmans, in France, and Sandorfy, in Canada, were also very active in using Hückel-type approximations.
McW: Yes, and Sandorfy was the pioneer in the development from pi-electron systems to more general saturated ones ...
A: ... which nobody really seemed to use ...
McW: ...but which was taken up in the period of 1951-53 by Pople and Santry. Pople, along with Pariser and Parr, made the zero differential overlap approximations which seemed to be absolutely indefensible because they were throwing away overlaps of 0.5, 0.7 and so on. Many of us tried to justify this type of ZDO by designing the basis functions so that they were in fact orthogonal. And when you do that you find that there is some basis for a ZDO approximation. Well, I mean, even when it's not numerically excellent it makes the theory so much more realistic if the approximations are not too far away from the facts. One feels more confident when the approximations are basically more realistic.
A: At some stage Sandorfy extended the Hückel model to sigma-electrons.
McW: Yes, this is what I had in mind a few moments ago. Unfortunately, these developments, in which the sigma-electrons, were included as well as the pi-electrons came just a few years before the big boom in computer development. And then people started neglecting the simple models altogether and trying to do something more or less ab initio.
A: So you would say the computer more or less ousted the early methods?
McW: It did, I suppose. People became more ambitious as soon as they had computing power.
A: PPP inlcuded.
McW: PPP was the next step after Hückel.
A: And then CNDO, INDO, MNDO, MINDO/3...
McW: No very great conceptual advances I would say in that period of the INDO - MINDO era. But by fitting the parameters or increasing the number of parameters it seems they could do remarkably well. And then of course Pople did a lot in extending the field of application. Because that's another valid way of proceeding, concentrating on expanding the applications, doing more and more new applications (e.g. to NMR and ESR), even if you are not changing the fundamentals.
A: And then Pople left the semiempirical field altogether!?
McW: Pople was another person who made a very dramatic break with his past activities. At one time, like Longuet-Higgins, he thought that anything you couldn't do on the back of an envelope wasn't worth doing. (I think that remark is attributed to Longuet-Higgins himself!). Pople was in the same school at the time. He slipped into computing, and then went really head-long into big computing. The "Gaussian News" I receive every month, I seem to be on the mailing list...And there is another funny story: I was the first person to use Gaussians in chemistry!
A: Not Boys?
McW: No, I used them in '46! But they were never put out until '51, and '53-'54. And the story is I have them in my thesis which I did while working with Coulson. As I was a physicist by training he gave me a physics-type thesis: he gave me a copy of Svartholm´s thesis on the binding energies of the lightest atomic nuclei and said : Well, wouldn't it be possible to do something like that for molecules? Svartholm had used iterative techniques. Anyway, in the course of this work I noticed that Svartholm had used Gaussians in the nuclear problem and they were very well adapted to nuclear wavefunctions because of the different force law. So you could get extremely good approximations with Gaussian functions. But I used them for atoms and simple molecules! And then I also contracted Gaussians! That was in '53. It began in connection with work on X-ray scattering; but it was published in Acta Crystallographica, so nobody knew about it. And then of course Boys was developing the ab initio approach for many-electron wavefunctions. He used Gaussian wavefunctions in a very general and systematic way.
A: When was that?
McW: That was about 1950. He and I were working with Gaussians at the same time, And then I think it was in 1954 at a meeting in Sweden we discovered that we had the same formulas for the matrix elements of the Hamiltonian between different spin-coupled functions. He got them one way, I got them another way. So usually many things are going on at the same time and nobody knows about what everybody else is doing.
A: Allow me to come back to another question concerning Hückel theory. Hückel threw out all these integrals and terms. Was that - in the view of those days - acceptable?
McW: Yes, I think it was a very general feeling. It was impossible to confront this enormous problem. You had to simplify by looking for a mathematical model that included the most relevant features of the system and which one could solve. I remember very clearly Coulson expounding on this philosophy in saying that this is the way we should do things, we cannot compete with nature in getting the exact energies, say, of the atoms and molecules. But we can look for the main features and for the trends.
A: Something else which just came into my mind. Pullman in his first book hardly gave credit to Hückel while he was using nothing else but Hückel's method. Could that have been due to the German - French encounter in the last war?
McW: I don't know. I think in Germany most of this work was done in departments of physics and in France in chemistry departments.
A: Staying with the French: What about the immediate fate of Daudel's "loge", which went by rather unnoticed?
McW: It's true, yes. It was largely ignored and partly replaced, I think, by an attempt to look for localized wavefunctions without introducing Daudel's ideas about "loge". He did it one way - other people did it in terms of localizing their orbitals looking for a prescription for localizing molecular orbitals so that they could get the wavefunction expressed in terms of the localized orbitals instead of delocalized ones. Which is closer to the ways chemists want to see it. The neglect of the "loge" was not a question of antagonism, but largely a question of ignorance because Daudel published his papers in French.
For historical material, are you familiar with the after dinner speech by Coulson which he made in - I think 1955? It was published in the Reviews of Modern Physics. That was a very nice paper, frequently quoted, in which he expanded these ideas about the role of quantum chemistry. He has a nice phrase: "primitive patterns of understanding" which he used in describing this kind of simple model, - which did indeed give us exactly that.
A: Ruedenberg did a lot of work in the 70's...
McW: Yes, I think he lost interest to a large extent in the semiempirical models. He did a lot of work in the 50's and 60's developing very beautiful mathematical models and so on. But then he didn't use them very much and went off into ab initio calculations. Of course he did all his early work on the 1,2,3 and 4-center integrals - a gigantic piece of work. And then the development of the ab initio methods. But it was really the semiempirical developments that interested him most - and, of course, the "understanding" of the chemical bond.
A: What other people of these days can you think of right now?
McW: Well, of course Matsen used to do a lot in semiempirical methods. But then he fell in love with group theory and never went back to semiempirical chemistry. And then Del Re also made a lot of contributions in the semiempirical field. He worked very close to the French. He was interested in the localized orbitals and the description of the chemical bond. But with the development of computers he felt there was nothing more to do and became more interested in other fields, such as epistomology and the philosophy of science.
A: Well, Professor McWeeny, I don't want to use too much of your time ...
McW: Well, I'm sure there are a lot of people we havn't put into our conversation. Think of Ballhausen, and Fischer-Hjalmars with the question of how realistic you could be with the ZDO approximation. She did a lot of work trying to justify that kind of semiempirical chemistry. And all the others ...
A: Well, Professor McWeeny, thank you very much for your kindness to talk about this period of quantum chemistry.