Colin Windsor 1938 - : His Science story

Colin was born into a very happy, if entirely unscientific family. His birthday 28/6/1938, was just before the start of the Second World War. His unscientific life-story is told on separate pages. Family life in the war was a worry for his parents: they often had to sleep in their concrete bunker in the garden, but it was not really a worry for him. He remembers happily shooting at passing German planes with his "ack-ack" toy gun from the roof of the bunker. His father George Windsor (1900 - 1991) was a civil servant going up to London every morning from their home in West Wickham, South East London. His mother "Ray" Windsor, nee Rayment, (1908-2000) was a full-time housewife. They were both interested in tennis and table tennis and music and these interests rubbed off on Colin.

He looks back thankfully to those days of the 11plus exam (a morning's exam with English, Arithmetic and IQ papers) which enabled the lucky ones to go to superb schools. He went to Beckenham and Penge Grammar School, a boys only school which his Father had attended before him, although in another building. The schooling was disciplined, ordered and good, and he thrived. The Science facilities were superb, with Senior and Junior labs for Physics and Chemistry, stocked full of ancient but serviceable apparatus. It is to his physics teacher Mr Davies and his chemistry teacher Mr Ridgewell that he owes his scientific life. Always busy, they always had time for his questions. In his slightly solitary way, he would take his textbooks home, and after homework was done, read them far ahead in the quiet of his bedroom.

Colin's father had as part of his job in the Ministry of Health visited factories. One day he took Colin along to a lens factory and he came away with a stack of reject lenses, large and small. The drawing shows Colin's room with his desk and the two-bar fire which he would sit near reading. The drawing was made with a sort of "camera obscura" from a wooden box, a nice big lens from the collection, a mirror and a glass plate on which you put some tracing paper. Click for more details. One particular lens, with a positive focal length of a metre was ideal for a telescope. This was duly made and cold evenings spent in the garden searching the heavens. The picture shows his photo of the moon made in the garden with a piece of old roll film and home developing.

His mother was patient. When he wanted to set up a Bunsen burner for chemistry experiments in her kitchen, she arranged for a proper fixing point - not the rubber pipe on to a tube used a school. He would take his Father's ammonium nitrate garden fertiliser and heat it to get ammonia. Physics took an equal place to Chemistry. It was the heyday of Meccano, with its metal strips joined by nuts and bolts. His parents bought him several sets. He scorned the designs in the instructions manuals and preferred his own designs with plenty of gears and motors. The grammar school headmaster Mr White put the idea of Magdalen College, Oxford into his head. Another pupil had made it there the year before. It meant staying on in the third year sixth form and taking the special Oxford exam in January. Colin remembers well the visit there, staying in a Magdalen College room, the really hard exam papers, but the friendly interview. He got a "demyship" (Magdalen's own word for a scholarship - it means a half fellowship). How lucky he feels it was to have that opportunity, without saddling himself or his parents in debt.

There was no point in staying on at school and he left for a 6-month job at the Mullard Research Laboratories at Redhill. His energetic boss, Dr Hoselitz, treated him as a part of the research team. He first made large single crystals of common salt, and later used a vacuum evaporator to produce thin films of iron. The idea was for fast memory storage using the anisotropy of the salt to hold the magnetisation of the iron. It was a good idea, but it did not work in Colin's time. He remembers the happy daily walks up the road outside the lab after lunch.

So in October 1957 we went up to Magdalen College to the gracious room seen in the picture. The toilet may have been across the quad and the communal bathroom even further, but it seemed like heaven. Those loving parents were far away. He was on his own. He could do whatever he wanted. He wrote his thoughts of life in an exercise book he still has. New friends were so easy to make. Many of those first friends were fellow Magdalen physicists. It was not all work. Tennis excursions along the lovely Addison's Walk to the college playing fields were frequent.

Those were the days of the individual tutor. For him it worked extraordinarily well. His tutor, Mr Milford, would set him an essay title, and he would go off to the College or University library and surround himself with books. He would try to condense the subject first into his own notes, now in a string binder with him today, and secondly into the written essay. Lectures were optional in those days, and he selected those he would attend with some care, like those given in a large room in the exam schools by the historian A J P Taylor.

His main tutor was Dr Griffiths, a real character who drove a Rolls Royce, looked after the College wines, and had a barrel of beer outside his bachelor rooms in college. He was to be Magdalen's President 1968-1979. He had the experienced teacher's knack of letting his pupils follow their own path, even way outside the syllabus. Colin was forever grateful of this opportunity. He remembers learning about "Monte Carlo" methods back then in the fifties, that were to become a major tool in his later research.

The first long vacation was spent with General Electric at their factory in Erith, Kent. He joined the team designing the reactor of the Trawsfynydd MAGNOX power station. The team was about 30 people housed in a modest hut on the giant site. It had a firm boss who sat at the top facing everyone else. He was the only one with a telephone. Of course there were hand calculators, and a few motor-driven calculators to break the silence. There was the occasional visit to Harwell to run their Mercury computer. They were an experienced team and had designed the Italian Latina reactor being constructed at the time. Colin joined with Mr Bryan in some juicy algebra, which aimed to simplify reactivity computations. It lead to his first classified publication: "A New Parameter for the Calculation of the Fast Fission Factor in Complex Fuel Elements.

There was plenty of music at Magdalen. Colin was always at peace attending Bernard Rose's evensong at 6.00pm. There was the college orchestra and amateur choir, and the peaceful music rooms with their grand piano at Magdalen and Somerville. There was the "Heritage Club" for guitars, which met in people's rooms with plenty of wine, where Colin met Margaret, another physicist, who was to become his first wife.

After finals Dr Griffith casually said, "How would you like to do some research with me?". That meant electron spin resonance. There was a flourishing lab mostly presided over by John Owen, a clever scientist, with specialities in superhyperfine structure and exchange interaction measurement for coupled "pairs". Colin joined this active team and did work in both fields. His real supervisor was neither James Griffiths, his official supervisor, nor John Owen, but John Thornley, a research student two years ahead who answered all his questions.

What a good education a DPhil was in those days! Everything that could be made, was made. Colin's first job was to make a Nuclear Magnetic Resonance field measurement box. He was given the circuit by expert Neville Robinson, a transistorised version and rather modern in 1961. The rest was made in the Clarendon Lab students' workshop presided over by "Jenks". He would see that you held a hacksaw properly and drilled the holes with the work securely clamped on the drill stand. He does not remember any accidents except the odd electric shock trying to mend the temperamental Mullard magnet that supplied the magnetic fields for the experiments.

An early task was the reconstruction and computer fitting of the superhyperfine structure in ammonium chloroiridate (NH₄)₂Ir,PtCl₆, a classic example of covalent bonding. Colin happened to be cycling past the University's new computer laboratory in South Parks Road on 26/11/1959, the day the new Ferranti Mercury computer was being delivered. He devoured the "Autocode" manual, talked to the helpful Dr Rollet about least squares fitting, and punched up some resonance traces. He was soon delivering his first punched paper tapes to Mrs Annette Fluendy. She was the lady who actually ran the computer, feeding in the boxes of paper tapes, watching the cathode ray screen for signs of an infinite loop, and boxing up the output paper tapes. The work lead to his first published paper Electron transfer in (IrCl₆)^--[1], and to several others in this field during his thesis [2,3,5,6].

His main thesis work, the measurement of exchange interactions between Mn⁺⁺ ions in K₂MnO₃ was altogether more difficult. A dilute single crystal of the manganese salt had to be grown and aligned. The dilution is critical: too dilute and the pairs will be very weak in intensity, too concentrated and there will be many "triples" and other complicating clusters. The figure shows the electron spin resonance spectrum. The 5-fold split central Mn⁺⁺ is clearly seen in the middle. The next most intense line is the spurious "half field" line. The lowest field pair line is clearly seen just below this. However the other lines seen are weak and poorly resolved. The exchange interaction had to be deduced from the temperature variation of these lines.

The thesis production was quite a tousle. Colin bought a typewriter, and borrowed his sister's typing course notes. Corrections to all those pages of matrices usually meant a new set of 4 sheets and carbons. It was ready in June 1963. The thesis examiners were Dr Baker from the Clarendon and Dr Marshall from Harwell. As the latter said: "There is a time in one's life when everyone needs to be tested to the limit and this is it".

These were the years of the "brain drain". Dr Werner Wolf at the Clarendon was starting a new spin resonance lab at Yale, and invited Colin to join him along with Mike Hutchings, also a spin resonance PostDoc. So it was that He and Margaret, on 22/8/63 and just married, set of on the SS Maasdam for New Haven, Connecticut. Although their research started of with an empty room containing a crate from the Clarendon, it was a good team and they were soon doing experiments. The picture shows research student Dave Landau, Colin, chemist Margaret Lyall, Werner Wolf, Secretary Cori and technician Cliff. Colin really enjoyed his first American Physical Society Conference at Atlantic City on 13/11/63. Papers were soon being written [8,11].

If the brain drain was easy, then so was the reverse. Harwell would send an interview team to New York, and pay for your trip back. So Colin arrived at Harwell in September 1964. After his thesis viva Walter Marshall had said how much easier it was to measure exchange interactions from spin wave neutron scattering than from pairs. It was true. By 1966 he had published a paper on spin waves in K₂MnF₃[10].

But his first task at Harwell was a more difficult problem, to extract a single crystal from the cubic antiferromagnet RbMnF₃ so that he could measure its spin waves. Prof Stevenson at Aberdeen had sent down a polycrystalline boule. He spent happy hours on a neutron diffractometer, setting the counter to the correct diffraction angle, searching for a big peak by changing the boule orientation, and then taking a "Polaroid" neutron snap of the reflection.

It was difficult and measuring the scattering from a polycrystalline RbMnF₃ powder collected from his chippings was a much easier project. It soon lead to Colin's first paper from Harwell [12]: to determine the exchange interactions from the paramagnetic inelastic neutron scattering. He turned to the PLUTO "chopper" a time of flight rotor-driven inelastic instrument. The measurements were easy: a suitably "high" temperature for paramagnetism was room temperature. He was able to analyse the scattering using theory from de Gennnes.

It was this theory that set his mind on computer simulation of Heisenberg magnets. The theory contains an equation:

dS_R/dt = 4pJ/h S_nn S_RxS_R'. This shows the direction a given spin vector S_R will move given its present value and that of its nearest neighbours S_R. Computers can evaluate this equation given a set of initial spin values. The differential is changed to find the actual spin movements DS_R within a small but finite time interval Dt. Paramagnets were easy, as any random set of initial directions will go. He set up a 10x10x10 computer model and followed it (in time units of 1/4pJS) for 4 time units with steps of 0.01. Having determined the spin motions, and he later made a movie picture of these, it was easy to determine the spin correlation functions <S₀(0).S_R(t)>, giving the mean value of the dot product of any given spin at time zero with any other spin at position R at a time t later. The figure shows the self correlation and the first, second and third neighbour correlations for a simple paramagnet. The results fitted well with the measurements on paramagnetic RbMnF₃ [15]. The work was extended to general temperatures using Monte Carlo methods and lead to a paper at the IAEA Inelastic Scattering Conference in Copenhagen 1968 [23].

At last a 0.2cm³ single crystal was cut out from the big boule of RbMnF₃. It was a bit small for inelastic scattering but again using the PLUTO "chopper", he obtained the data for the figure shown, later to be reprinted as an example of spin waves in Kittel's Solid State Physics. The almost linear way the dispersion curve approaches the origin shows the negligible anisotropy in this compound, making it a model antiferromagnet. He was to work on this compound over many years.

Colin's boss in those years was Dr Ray Lowde, a character with many years of neutron scattering to his credit. He was able to apply his infinite concentration to the discussion in hand. If the telephone rang, he would carry on the conversation without a falter as he opened his desk draw and put the ringing telephone inside. He had a project to measure the itinerant magnetic scattering from nickel. The discussions included most of the UK's considerable talent in this field: Mick Lomer, John Hubbard and Stephen Lovesey from Harwell, Seb Doniach, David Edwards, and Peter Wohlfarth from Imperial College. There were many visitors, Guy Allan from Lille, France, Joe Cable from Oak Ridge, USA, from Japan, Jiyo Komura a PostDoc, and Takeo Izuyama the joint discoverer of iterant magnons. The first part of the project was to acquire a nickel 60 isotopic single crystal. This had the same magnetism as ordinary nickel but much less of the unwanted nuclear scattering background. Again the Pluto chopper was used for the experiments. They were not easy. There were 34 counters to be calibrated with a vanadium scatterer and combined together. There were corrections for background, for furnace scattering, for elastic scattering, for multiple scattering, for phonon scattering, for resolution. The list seemed endless. But a series of papers were written, including three Physical Review Letters: a fit to the Hartree-Foch free electron model [13], and a fit to Doniach's free-electron gas [15]. Guy Allan was experienced in band structure calculation. His visit prompted a tight-binding model including the five 3d electron bands and Izuyama's exchange enhancement [22].

The next step was to measure the scattering as a function of temperature, showing the smooth transition from the spin waves predicted by Izuyama at low temperature through the phase transition to the high temperature paramagnetic state. The figure shows the calculated intensity the full temperature dependence of the nickel scattering and compare with the Izuyama exchanged enhanced theory. The spin wave scattering, which had always slightly problematic to measure over the relevant energy transfer range emerged very clearly in this calculation, and was equally clear in the room temperature experiments. Similar results were performed at four temperatures spanning the transition temperature 0.5Tc,0.9Tc, Tc and 1.6Tc The work was written up as an Advances in Physics review [28].

In 1973 the UK joined the Institute Laue Langevin in Grenoble and neutron experiments of a new quality became available to Harwell staff. Colin sometimes regrets that he did not post the application form for a job there that he once filled in. However he performed many experiments there, some successful, some difficult and never written up. One success was the measurement of the inelastic scattering from supercooled gallium which he performed in collaboration with the group of Louis Bosio from Orsey, Paris. It was an enjoyable collaboration and Colin's French was never so good. Gallium readily supercools and the experiment was to measure the dynamics of the liquid state as it approaches its stability limit at around 150K. It was an exciting experiment. One never knew when the gallium would suddenly crystallize as it was being cooled. The experiments were readily analysed and the inelastic width at each new temperature could be plotted within minutes of the scan being completed. The team slowly watched the widths decrease with decreasing temperature, with the straightforward result that the width linearly approached zero as the temperature approached the stability limit [37,44,48].

In 1975 Colin was asked to join the Linac Review Group considering the science possible with pulsed neutron sources. It was to change the direction of his research. Roger Sinclair's Total Scattering Spectrometer had opened a new field in determining the structure of liquids and glasses. Colin put a nickel powder sample on the spectrometer sample changer. A weekend of counting revealed a diffraction pattern extending to scattering vectors, Q), way higher than possible on reactors. Starting from first principles, Colin showed that in the epithermal high neutron energy range the scattering intensity per resolution element was independent of neutron energy. Colin immediately set to and developed a time of flight version of the Reitvelt profile refinement code he had used for reactor diffraction patterns. It was a new task, the wavelength dependence of the resolution was quite different and new peak profile shapes had to be determined. It was reported at the Pettan Neutron Diffraction conference in 1975 [43].

In fact the diffraction patterns continued until the structure was attenuated by the thermal vibration amplitude of the atoms, the Debye Waller factor. Colin and Roger Sinclair developed the code showing that these thermal parameters could be determined in detail [45]. The profile refinement programs now universally used on pulsed sources are all based on the classic paper of non-Dreele, Jorgenson and Windsor [81]. The new Harwell Linac was approved in 1976, and Colin became Group Leader of Pulsed Neutron Scattering. Although a group leader since 1972, they now had a real team with a mission. There was Roger Sinclair, Les Bunce, Chris Tasker, DAG Johnson, and part-time secretary Hilda Whitmarsh, who would drive them to the Cherry Tree in Steventon every Friday lunchtime. They had their own space: a reconditioned "terrapin" hut.

They had just one year before the old linac was dismantled to prove the their methods. It was the hot summer of 1976. They ran around the "Condensed Matter Hut", shirts off, moving the many resin blocks needed for the new instruments. In that summer with help from Brian Boland, Zoe Bowden, Dave Mildner and others from the Rutherford Laboratory, they designed, build and later wrote up a series of new instruments. The Back Scattering Spectrometer [51] was first. It followed as far as feasible the improvement in resolution which comes as back scattering is approached. The figure shows how the new diffraction pattern carries on from where the Total Scattering Spectrometer disappeared. This is partly because the 77K temperature of the sample limits the Debye Waller attenuation shown on the Total Scattering Spectrometer measurements.

Next was the Inelastic Rotor spectrometer [59], build by Les Bunce, Dave Mildner and Brian Boland from an old reactor instrument rotor, old reactor counters, and constructed from old resin blocks! It worked well and measured a series of experiments. The figure shows the famous vibrational spectrum of liquid water with its peak at 210 eV.

There was one quite novel spectrometer invented by Colin. A highly important field of neutron scattering was the measurement of coherent inelastic scattering, like the phonons and magnons of crystals. Around 1952 at Canada's Chalk River Laboratory Brockhouse had invented the triple axis spectrometer. It was soon ubiquitous at research reactors. They could not work on pulsed sources, and Colin spent hours dreaming how it could be done. The answer came to him in the big bath of Avebury House and he remember getting out, drying himself and going to write it down on paper. The "Constant Q Spectrometer" [58] moved quickly, and soon there was a prototype spectrometer in the waste beam behind the Total Scattering Spectrometer. It was to become one of the first spectrometers on ISIS [114]. The related "High Symmetry Spectrometer" was soon built at KENS the Japanese pulsed source. Colin went over there for 6 months on a grant from the Japanese Society for the Promotion of Science.

In addition a liquid nitrogen moderator was installed [57] and some initial small angle scattering experiments made [63].

Then in 1977 the old linac shut down. Colin and Roger Sinclair had already worked out their ideas for the instruments on the new linac. They were defined much as they were to come into being in their paper "A Discussion Paper for the Exploitation of the Harwell LINAC for Condensed Matter Studies". There was not so much to do. Colin set to work writing "Pulsed Neutron Scattering" [71] the first, and still the only book on the subject. It was published by Taylor and Francis in 1981.

Even before these pulsed neutron experiments in the mid sixties a wind of change was sweeping through Harwell. Walter Marshall became Deputy Director of Harwell and later Director. Harold Wilson wanted the "white heat" of technology spread to industry. Walter embraced it, as did Colin, and he with Group Leader Peter Egelstaff performed the first commercial neutron scattering experiments. They were for English China Clay and were successfully written up. The first industrial neutron publication was in 1974 for the Cement and Concrete Association [42]. They wished to measure, not just the amount of water in concrete, but the amount that was chemically bound to the concrete compared to the amount of free water. The figure shows the measurements. The broad peak is from free water and shows the characteristic width in energy arising from the diffusion of water molecules. The narrow peak has merely the broadening characteristic of the experimental resolution and represents the bound water. It was a simple matter to measure the free to total water fraction. There was no sign of any intermediate bonding and the measurements gave a much smaller fraction than did conventional drying measurements.

Numerous applied neutron experiments followed over the coming years, including in 1980 a review "Neutron beams for industry" with Chris Wright for New Scientist. Colin had a long partnership with Gerry Slattery from the Risley Laboratory of UKAEA leading to several papers and reports on the phases present in stainless steels [84] and maraging steels [92]. It also lead to a review of the "Metallurgical applications of neutron beams" [83]. A second long-lasting partnership was with the Berkeley Laboratory of the then Central Electricity Generating Board. The first study was on the so-called gamma prime phases present in stainless steel with Vince Callen and Pat Rose [89]. It moved on to a long-term study of the precipitate structure in pressure vessel steels.

Yet another long term partnership was with Alan Bowen of the Royal Aircraft Establishment Farnborough, on the clusters formed under heat treatment in evaporated steels. These steel had excellent mechanical properties due to their absence of conventional crystallites with grain boundaries. Small angle scattering was a powerful tool to reveal the size and concentration of the tiny clusters of solute atoms that were formed with heat treatment. We describe later how Colin's computer models were to interpret the data.

It was on a trip in 1980 to the Welding Institute at Abington near Cambridge to try to obtain industrial contracts that Colin stumbled on the importance of residual stresses. They were expert at that institution on the measurement of internal stresses in metals by shaving of slices of metals and measuring the surface strain changes. It was a time-consuming and difficult task and they impressed on Colin that the most important application of the penetrating power of neutrons would be to measure internal stresses in samples non-destructively. The method is easy in principle. A simple powder diffraction peak gives the lattice spacing and strain is simply the relative change of lattice spacing. The problem is the practice! The magnitude of typical residual stress lattice changes is only a few times 10^-3 and this is very high resolution by the neutron diffraction standards of the 1970s. But Colin had recently been immersed in this problem, and its solution via back scattering that he had used on the linac. In a short time he had devised the Double Back Scattering Spectrometer and it was detailed and constructed by Peter Schofield, Mike Hutchings and Andrew Allen for installation on the PLUTO reactor as shown on the left. The instrument worked only with steel specimens, and using the same iron 110 reflection near back scattering for both monochromator and specimen. Fortunately from old iterinant magnetism days Colin had a large single crystal of pure iron in his cupboard which was ideal as the neutron monochromator. With the iron 110 reflection at the wavelengths used, both reflections could be caused to occur at just 6^o from back scattering. The scattered powder pattern could be recorded by a Position Sensitive Neutron Detector so placed to span the expected scattering angles.

The first experiments were not a success. Some iron powder was chosen as the specimen and the beam opened. There was nothing to be seen. In fact unannealed iron powder has large residual stresses and a random broadening destroys all the signal.

There was a second serious practical problem. No one quite knew what the residual stresses were in any particular specimen. What was needed was a realistic sample whose internal stresses could be readily changed and measured. Colin devised the "Vice" shown to the right. It was a rectangular bar of annealed mild steel, clamped firmly at the upper end. Its lower end could be pushed by turning the vice screw so that the bar bent into a smooth arc. The macroscopic strain could be measured by the displacement of the bar at its bass, or more precisely by its curvature using a planimeter. Of course all deflections had to be within the elastic limit so that the changes were reversible. It proved extremely useful and effective from the first.

Then came success. The figure shows the response of the Position Sensitive Detector of the Double Back Scattering Spectrometer for 1/4, 1/2 and 3/4 turns of the vice screw. The shift in peak position was well within the experimental error and of the expected magnitude.

Things moved quickly. It was soon realised that the shifts seen were within the capabilities of the high resolution D1A diffractometer at the ILL High Flux reactor in Grenoble. Instrument time was booked and Carla Andreani, Andrew Allen and Colin set off for the experiment. Mike Hutchings was already there for another experiment. Everything went well. The photo shows the trio holding the sample and the experimental book. The crucial results were of the lattice strain measured from the neutron diffraction plotted against the applied stress as measured by the macroscopic strain and the known elastic modulus. The two curves are with the scattering vector both along the strain direction and perpendicular to it.

In 1981 the Sinclair ZX81 computer came out and Colin bought one. It was to change his life for a while. Previously computing had been about submitting jobs to a mainframe and waiting for the answers. Now you pressed the "ENTER" key and the answers were there immediately, or at least were being calculated. The ZX81 had a lot going for it. For display you simply plugged in an old TV. For storing programs and for reading them in, you plugged in a standard cassette tape recorder. It came fully loaded with BASIC and when you powered up there is was. You could type: PRINT "Hello World", ENTER, and there is was on the screen. A child could learn it, and Colin's children, about 7 at the time, did. He has still the little book Elizabeth kept so neatly with its little 5-line programs and their printouts pasted in. Yes, you could buy a tiny printer with useful commands like "Print Screen" which were not to be commonplace until very much later.

With its tiny free storage of about 1 kbyte, writing serious code was another matter. But its tiny size gave it some immense advantages over our current impenetrable operating systems. Its storage locations were fixed. You knew exactly where everything was in the computer. You knew the locations that put colours on the screen. Like computers still, its heart was an "Accumulator" register, and a dozen or so other named "Registers" which you could access using some stored "Program" which could perform simple sequential operations on the registers. Inside the ZX81 was the Z80 microprocessor. It had just 32 machine language commands which were easy to learn. "ld b,xx" loaded the number at address location "xx" into the Register "b", "add b,a" added the number in Register "b" to the accumulator. It was not difficult! You could see exactly what was going on in your computer in a way which is quite impossible with Windows Vista.

The Monte Carlo modelling work described earlier was just right for this environment. Colin devised a gas/solid/liquid model that you can see illustrated in the book cover left. The "Model" was two-dimensional with 10x10 possible atomic positions and a variable number of atoms within the model. The current state of the model was actually stored on the locations driving the screen display. After that, not much more storage is required for the Monte Carlo method, just straight number crunching which the Z80 could do with ease. At low temperatures the black atoms shown in the left hand square, had condensed into a solid with the small black squares indicating vaporized atoms. As you warm the model up, more atoms "evaporate" and the black areas break up and become changing with time. We have a liquid. The curve in the centre is the "Specific Heat" evaluated from the change in energy of the model with temperature. It shows the clear phase transition present in even this tiny model. Three more models were written demonstrating a magnetic 2-dimension "Ising" model, the correlations in a one-dimensional linear chain, and percolation in two dimensions. He published "4 Computer Models" himself in 1983.

In May 1984 the 4 models reached the Royal Society Soiree. With clever vacation student Bruce Normand and applied to the commercial aluminium alloy experiments mentioned earlier, it was able to provide a detailed fit of the growth of precipitates with time[115]. Later with student Ruth Barron the modelling was able to reveal the detailed form of alloy clusters as they grew in size [121].

Another event to change Colin's life occurred at the 1987 Reading University Solid State Physics conference. These annual meetings had been going on throughout Colin's professional career, and he had taken to them. He liked to meet his colleagues and hear what they were all doing. He became Secretary of the Solid State Subcommittee of the Institute of Physics in 1980. For 8 years, he and the Chairman Denis Weaire, worked hard to reinvigorate them [103].

The speaker at the crucial talk at Reading was physicist David Wallace from Edinburgh. He talked about neural networks, the rather new field that Colin was to embrace. His subject was "Content Addressable Memories" - very simple type of network that had been invented in 1982 by another physicist Hopfield. Colin was entranced! As he wrote in an article to Physics Bulletin [120] the following year: "We can all see in ourselves some of the essential ourselves some of the essential characteristics of human memory:

(i) memories need to be learnt and are built up by rehearsal;
(ii) we recall from a clue, or fragment of content;
(iii) memories decay gracefully with time, rather than catastrophically."

The article included a 67-line BASIC program including data which demonstrated these three properties. The lower three images in the figure show the 5x5 "attractors" which were "learnt" by the Hopfield network. The upper images show the "seed" images, representing the "clues" or "fragments of knowledge" which was able to recover the correct images. In the third case the network had been "corrupted" by introducing random changes to the network connections but still gave nearly the correct image. The article was to have quite an influence. Chris Bishop, soon to be one of the leaders in the field, was drawn to it by the article.

Harwell then formed a small team lead by Andrew Chadwick and later John Mason to exploit Neural Networks. We won an ESPRIT II European project "ANNIE" - The Exploitation of Neural Networks for Industry in Europe. Colin was the chairman of the "Image Analysis" group, once described by a referee as the "jewel in ANNNIE's crown. A highlight was in 1991 when Colin was asked by ESPRIT II to display the group's "On-line weld defect detection" at the international Expert Systems Conference at Avignon.[140]

Always on the lookout for the big application, Colin began a project to predict the Stock Exchange index. Other more "chartist" methods had looked for predictors within a single index on a fine time scale. Colin developed a neural network model with many annual inputs, so that fluctuations were averaged out and true economic trends could be revealed. Another reason was that annual data could be typed in from the Economist publication Economic Statistics 1900-1983. 12 inputs available there included the UK Share Index, Interest Rate, Ml Money Supple, Gross National Product, Personal Disposable Income, Balance of payments, Conservative Party Majority, Savings Ratio, Birth Rate, Unemployment and the US Balance of Payments and Dow Jones Index. The results obtained are shown in the figure from the paper given with Tony Harker at the IEEE International Neural Network Conference, Paris 1990 [126]. Colin remembers an amazing presentation. The quiet room suddenly became completely full and there were many questions. The later poster session was a rugby scrum with not a moment of peace! "Rocket Science" then began although Harwell considered that it had no long term potential!

They were heady times. In 1988 Colin took a trip to CNRS Paris to visit one of the early pioneers in the field Elie Bienenstock. It was a lovely department in the old building containing Marie Curie's still radioactive laboratory. He had developed clever techniques for the elastic matching of images. Colin absorbed what he could and in the days following developed a much simpler one-dimensional analogue of what he had been doing in two dimensions. Suddenly the complicated code running on a Cray became simple code running on a home Sinclair. Chemist Andrew Chadwick saw that the code had immediate application for gas chromatography and very soon C code for a PC was written along with a patent [122]. Another similar application arose from this. Chris Bishop had now joined the group from Culham and one day on a long plane journey he and Colin devised a related application: "Dynamic Signature Recognition". A hand-written signature can be captured by a special tablet so that it stores the path of the signature as a function of time. The signature is two-dimensional, but the time-line of the signature is only one-dimensional. In a remarkably short time, C++ code was developed and a patent in place [132]. It became a quite a project with official funding and code control so that the working code could only be modified by two signatures who had checked the changes! Colin remembers the trials they held in Harwell's Admin block Building 329. Trials were so important as the best way to improve the algorithm was to study the now tiny numbers of authentic signatures rejected and undetected forgeries. They went well until a Secretary there suddenly forged the Director's. Interestingly she was an amateur artist with a "picture" memory. Her excellent forgeries were a great help to us is improving the many parameters of our method.

In 1992 the project was robust enough to make "Tomorrow's World". It was scary as the programme is broadcast live and the program had to work unattended by us. Howard Stableford, the presentor, had to make the signatures look different on TV screens at home without the system saying "Forgery!".

The project grew with trials at the Department of Social Security, and a public display at the Royal Society Soiree of 1995. Eventually it was sold to PenOp an English competitor. On Colin's retirement in 1998, he joined PenOp as a Senior Consultant. PenOp had a good team with a fine technical leader Nick Mettyear. The system's performance became ever better. Sadly in 2001 it was sold again to yet another competitor the US firm CIC. Many of the old staff were laid off and formed their own company Florentis, which still markets yet better technology including a static signature verifier, which Colin had developed but which had never been brought to a product.

in 1994 Colin made a move to Culham Laboratory to help with a quite different application of neural networks. The left hand figure shows the COMPASS-D tokamak surrounded by its magnetic diagnostic coils. Signals from these could be used to find the plasma position and shape, but only in seconds using the specialist code TOPEOL. In order to achieve on-line control the tokamak position and shape, a much faster millisecond speed was necessary. The neural net achieved this by running many TOPEOL codes covering all likely plasma positions and shapes, storing them, and then training the network. The project was well underway when Colin joined but a new milestone was coming up the following year for the Extension to two plasma variables. He formed a team joining the disparate peronalities, experts in the codes, the electronics, the diagnostics and the COMPASS-D operations. It was not easy as the hardware network was locked away behind the COMPASS-D security system. Even changing the network demanded a shut-down to break the security. The milestone was met and the right hand figure shows the plasma height (upper) following a demanded profile (right). For an account [141]

In 1998 the JET tokamak had a series of "Tritium" runs when it run the true fusion Deuterium-Tritium reactions producing many neutrons. Colin had administrative charge of the neutron diagnostic team. Their "cameras" consisted of 10 vertical and 10 horizontal collimators with detectors measuring the neutron intensity along the lines shown in the left hand figure. Colin's aim was to make a simple model that fitted these intensities. The resulting extended Gaussian model had about 9 parameters to fit the 20 inputs. The model could be displayed in many ways including a movie film showing the plasma neutron intensity thoughout the shot as shown on the right.

One of the problems of tokamaks is that, under the very conditions giving good neutron and power output, high electron density, high temperatures, high magnetic pressure, the plasma can become unstable. There is a "disruption" and the plasma may dump all its current into the walls. These must be carefully controlled or the tokamak may be damaged. It is an important problem to try to predict the appearance of disruptions enough time ahead

After Colin's retirement in 1998 he began a long collaboration with Lorenzo Capineri from Florence University to detect and classify anti-personnel land mines buried in the soil. Landmines present a dreadful legacy to humanity. There are around 100,000,000 line mines still in the ground and around 15,000 casualties each year - half die, but the other half live maimed lives. It cost about $10 to make a landmine but about $500 to remove it from the ground. Using present methods and funding, it will take hundreds of years to remove them.

What can science do? Metal detectors are most commonly used but the amount of metal in a modern mine is tiny. radar is one possiblity since it sees both plastic and metal. Colin and Lorenzo had long experience working together on ulrasonic measurements from welds [140, 142] and radar is closely analogous. radar can do more than just detect mines, it can image them, and potentially classify them. A team was formed with many students and a collaboration with IDS, a Florence company who manufactured radar equipment and performed demining contracts. From the many papers [148,149,150,154,159] we pick out one [157] on The classification of buried objects.

The data were publicly available from DeTeC - the Demining Technology Center at the Ecole Polytechnique at Lausanne, Switzerland. The area analysed in the figure shows the reflected signals from a typical plastic anti-personnel mine (type PMN), a bottle, a metal ring, all buried in a big sand box. The sizes of the circles shown represent positive and negative radar reflection amplitudes. It is seen that these form vertical "pendants" of alternating sign as the initial radar pulse is reflected into a broad wavelet. It is these intensities which contain information on the nature of the reflecting object. These sets of amplitudes from each object located form the inputs to a classifier trained on other data with only that object present at varying depths. Classification without error was possible with a simple neural net classifier. The standard K nearest neighbour classifier worked equally well.

A write up of recent work on holographic radar has been prepared for the Royal Society 2010 Soiree.

A demonstration of the holographic radar from an image taken at Florence University has also been prepared for the Royal Society Soiree.