The information here comes from a meeting in honour of Vic Saunders, see: http://www.ccp1.ac.uk/newsletters/Current/Saunders/MSSQC.html. The meeting was held in 2003 after Vic had retired from Daresbury.
Although we in the UK tend to think of Vic as the expert on gaussian integral evaluation, we do not suppose it his only contribution to the field. Peter Knowles, who was one of the students involved in that summer school, is actually speaking at this meeting and I rather think that he was greatly influenced by another of Vic's contributions and perhaps he will speak of it.
Vic was born in Birkenhead in April 1943. After getting his Grad. R.I.C. in 1965, he moved to Sussex for his post-graduate work and in 1969 he moved on to Manchester as a Research Fellow in Ian Hillier's group, leaving in 1971 to become an SSO at Atlas Lab. There he remained until SRC policy changes about the location of computational resources moved him and some of his colleagues to Daresbury Lab where he has remained.
Atlas Lab had been set up, under the direction of Jack Howlett, a former student of Douglas Hartree, to run a powerful UK designed and manufactured machine, Atlas, for the benefit of UK science. To understand what was involved it has to be remembered, no doubt to the astonishment of some younger members of the audience, that there was a time when computers were hideously expensive, quite enormous engines. (As an aside, Atlas as installed in 1964, spread over two full floors of air-conditioned space, one floor devoted to holding the CPU and the other to holding the peripherals, largely Ampex one-inch tape drives and an enormous drum-store. It took 14 truck-loads to get the machine to the lab and it took three months to set-up and test. It cost £2.5M at 1964 prices (so probably about £50M at today's prices) and it developed .25 MIPS.
They were prodigal consumers of power, requiring the utmost cosseting in terms of air-conditioning and of ministering staff. If they were satisfied with their treatment then they would run reliably for up to 12 hours or so and while running, could be approached by suitably skilled acolytes to be fed with programs written in a suitably abstruse language, which they would often run and would sometimes yield results on a great line printer operating noisily at 300 lines per minute. To get anything at all out of a machine in a reasonable time required the greatest ingenuity and programming skills. Such ingenuity and programming skills Vic had shown in a paper that he and Ian Hillier produced (Proc. Roy. Soc. A 320, 161, 1970) that introduced the idea of level shifters to ensure convergence in SCF calculations and I have no doubt that this achievement was what helped Jack Howlett to realise that Vic was just the man for the job that he had in mind.
What he had in mind was the foundation of what he called "meeting houses" that could consider the computational requirements of such disciplines as theoretical chemistry, chemical crystallography and the like and, by collaborative effort, develop program suites that were generally useful to the UK science community. To this end he began to appoint permanent staff to the lab to provide the nuclei around which the meeting houses would form. I think that Vic must have been the first appointment here but he was rapidly followed by Jed Brown, who died so tragically young late last year, and then by Martyn Guest, who is happily present here. The crystallographic and data-base work formed around Mike Elder and Pella Machin a little later. Both of them are still remembered with warmth by the community, after their untimely deaths in a climbing accident late in 1987. The molecular side of things begun by Vic and Martyn, developed with the addition of Bill Rodwell and Mike Chiu, neither of whom transferred to Daresbury, and by Alastair Rendall, who was at Daresbury for a time before heading for Australia.
To introduce a personal note here, relevant to Vic's contributions (though I shall stop myself from speaking of Vic's extramural activities such as chess and botanical horticulture). He and Jed Brown arranged a meeting in St Catherine's College, Oxford in April 1974 which was called Quantum Chemistry: The State of the Art, the proceedings of which still make interesting reading if you can run them down and spot some who were to become real stars in the field making their early contributions. I had been much concerned with developing, together with a student, Don Garton, a direct minimization approach to the solution of the SCF equations based upon the ideas of Fletcher and Powell on conjugate gradients and finally Don and I had realised that actually the level-shifter method proposed by Vic and Ian was actually just a neat way of writing the conjugate gradient method. When I broached this with Vic at the meeting and asked him if he knew it, he replied that he did and when I then asked him why he hadn't mentioned it in the paper, he said because he had thought that it was obvious! I retired suitably chastened. My point here is that it is an example of that which was to be a frequent occurrence, of Vic beginning something that was later to become "big business". Now we all use direct minimization techniques: Fletcher-Powell, Newton-Raphson, Marquardt-Levenberg or whatever, to converge our solutions or to optimize our geometries, but perhaps Peter Pulay will dwell a little on this topic.
Vic was to do something like this again a little more than ten years later in an amazing paper with Joop van Lenthe (Molec. Phys. 48, 923, 1983) called The direct CI method: a detailed analysis. It was again, a first, and it is difficult to over-estimate the effect that it had on the field. Perhaps Joop will enlarge a little on it in his talk. But I once more get a bit ahead of myself so let me go back a bit to put this paper in context.
By 1974 one part of the "working group" had developed into a "meeting house for molecular electronic structure calculations". This meeting house was to become Collaborative Computational Project 1 (CCP1) when the CCPs were founded by Phil Burke at Daresbury in 1977. The aim of this effort was to produce electronic structure computer programs that would be available to all UK users and it was as part of this endeavour that Vic and his colleagues created the program suite ATMOL. This suite had two enormous advantages, it was tremendously efficient and it was easy to use. I know that in these days of graphical input interfaces and readily available electronic structure suites, it must be difficult for anyone to remember what damnable nuisance the rigid Fortran format statement was in getting any input data correctly into the machine and how foul IBM job control language was, ruining I always thought, at least one submission in three, usually because you'd forgotten something like that you'd used the disk before and so needed "disp=old" not "disp=new" or you'd forgotten to specify the cylinders to be used. But the input to ATMOL was in free format and all the JCL was taken care of by default and never caused any trouble. One could also easily understand the output. I know this sounds corny, but to have things so easy, was at the time, just amazing.
The achievements of ATMOL are not perhaps as well-known as they might be, perhaps because it was not much used in the US. But when I look at the codes of some of the currently popular packages, I think that I can see a good deal of ATMOL in them. Of course one should, perhaps be careful here. It may well be that clever programmers arrive at similar solutions to a given problem.
When in 1979 the CRAY 1 arrived at Daresbury, I was a visitor to the Lab, spending four days a week there and being really foxed about how best to use this potentially powerful vector processor. Not so Vic! He announced that he and Martyn Guest would port the ATMOL codes to the CRAY in just one month and that they could subsequently be vectorised. There was a meeting at DL in November 1979 at which Vic read a paper entitled "The use of Vector Processors in Quantum Chemistry" and it still makes good reading. But he showed that he and Martyn had managed to make the ATMOL code run 16.2 times faster than the best code on the fastest scalar processor in the UK (the CDC 7600). When I heard the talk I was astounded and delighted, but I didn't know that to achieve these figures the CRAY basic linear algebra routines (BLAS) had been replaced by ones that Vic and Martyn had written and which were much faster than the CRAY supplied ones. But the delight at these achievements was general in the UK electronic structure community.
Of course, this being the UK, about to begin its great leap backwards, all was about to be spoiled by official intervention that shifted the CRAY from Daresbury to London. But it turned out not to be as bad as we all feared because of the growth of networking that made it less and less relevant precisely where users were located in relation to the machine that they were using.
In any case the whole structure of scientific computing was beginning to change, first with the advent of powerful mini-computers and later with the arrival of powerful work-stations and PCs. These in turn lead to the growth of parallel and distributed computing. The effect of these technological changes together with high speed networking, was, over the decade from '85 to '95 to change the nature of the endeavour of all kinds of computational chemist, including those involved in electronic structure calculations. They changed from being a large group enterprise centred about a central machine: if you will, like factory production; towards small group enterprise: if you will, like hand-loom weavers, a cottage industry.
But I stray too far from my brief. In the middle 80s Vic's interests began to move from the isolated molecule towards repeated molecular structures and working with Italian colleagues he begins to apply his molecular expertise in the study of polymers and becomes a participant in the development of CRYSTAL. I look forward to hearing what Professors Dovesi and Pisani have to say of this work.
When Nic Harrison asked me to do the introduction to this meeting for Vic, I felt honoured and delighted for I have, as have so many of us present here, been very much influenced for the better in my professional life by his example. My only problem was to limit myself to just a couple of examples of his influence. But I am sure that those many aspects of his work which I have neglected will be well supplied by those who follow me.
