1. Introduction
2. The Dawn of NMR in Chemistry
3. The "Do it yourself" era
4. High resolution NMR
5. Two key doctoral theses
5.1 James T. Arnold
5.2 Weston A. Anderson
6. Electromagnets
7. Permanent magnets
8. Commercial NMR Spectrometers
9. Sensitivity
9.1 Multichannel excitation
9.2 Fourier transform NMR
When Chemists First "Discovered" NMR
This review charts the early history of NMR from the time that it first began to be of interest in chemistry. It begins with the efforts of a few far-sighted chemists to build their own spectrometers, exemplified by one such laboratory in Oxford in the early 1950s. The key breakthrough into high-resolution NMR by scientists at Stanford is highlighted by citing the classic papers written by Jim Arnold and Wes Anderson, based on their doctoral research. The evolution of electromagnet and permanent magnet technology is described, culminating in the introduction of commercial NMR spectrometers which opened the field to a broad community of organic chemists. The quest for higher sensitivity led to the introduction of the time averaging technique and the concept of multichannel excitation, triggering the Fourier transform revolution and the advent of practical carbon-13 spectroscopy.
1. Introduction
When the phenomenon of nuclear magnetic resonance1,2 was first discovered, it was exclusively the province of physicists. Almost all the known elements possess one or more isotopes with magnetic properties, and there was a concerted effort to search for these unknown resonance responses, as there was no way to predict the frequency at which they would appear in a given magnetic field. Once the nuclear resonance frequency of a particular nuclear species has been measured, it defines the magnetic moment. Although frequencies can be measured with very high precision - magnetic field strengths are best derived by way of magnetic resonance; consequently the new magnetic moments were usually related to that of the proton. At the time, the magnetic moment was believed to represent an important fundamental property of the nucleus.
This new property aroused intense interest among physicists because it seemed to offer new insight into nuclear structure. Then (horror of horrors) evidence began to emerge that these fundamental "constants" were not really constant at all, but appeared to depend on the nature of the sample used for the measurement. This disturbing observation is vividly illustrated by the case of ammonium nitrate, investigated by Proctor and Yu
3 in an attempt to derive an accurate value for the magnetic moment of 14N. To their consternation they observed two resonance lines. Owing to an understandable reluctance to accept that this might be a mere chemical effect, every effort was made to seek a less disturbing explanation. To this end they placed an order for a sample of 15N-enriched ammonium nitrate in the (vain) hope that the unwanted second resonance line was actually from the 15N isotope. In fact the two resonances represented two different 14N chemical sites in the same molecule. About the same time, Dickinson4 found similar "discrepancies" in the 19F resonance frequencies of several fluorine compounds, Lindstrom5 showed that the ratio of 2H and 1H resonance frequencies differed in water and paraffin oil, and Thomas6 observed further instances where the proton resonance frequencies depended on the sample used. Something was seriously wrong. How could all the excitement about a fundamental, high-precision nuclear property of such intrinsic importance be so cruelly dashed by a mere chemical shift? Most physicists gave up in disgust.
2. The Dawn of NMR in Chemistry
Many setbacks in science turn out to have a silver lining. A few far-sighted scientists began to wonder if this new magnetic resonance discovery could be applied to chemistry, by concentrating attention on the chemical effects so despised by the physicists. Rex Richards at Oxford believed that NMR would be a promising new avenue for chemistry research, but the physicists he consulted were uniformly discouraging. Fortunately the world-famous chemist Linus Pauling was visiting Oxford at the time and he told Rex "Don't believe anything the physicists tell you", so Rex went ahead. The justification for such optimism was rather thin at that time, but at least NMR was an open field, and Rex is a courageous researcher. However in that era, a typical chemistry laboratory would be mainly devoted to "wet chemistry" and it was really no place for two-ton magnets. To introduce any large physical machine into that environment was a real battle, involving banishment to some dreary basement room that no one else wanted to use. On the positive side for an NMR pioneer there was a good supply of war-surplus radar equipment that could be usefully cannibalized, and it was always possible to ask those now disillusioned physicists for technical advice.
It was the work of George Pake
7 that indicated a possible line of attack for chemists. Pake chose to investigate a single crystal of gypsum (CaSO4.2H2O) where the two protons of the water molecule are the only significantly magnetic species. He showed that the NMR signal was a doublet of resonances, representing the fact that each proton in the water molecule could "see" the two possible orientations (up or down) of its partner. The splitting of this doublet depended on the orientation of the crystal with respect to the magnetic field and gave a measure of the inter-proton distance. An important justification for studying solid-state NMR was that it filled a serious gap in the coverage of X-ray crystallography, where the scattering was mainly from the heavier atoms and it was quite difficult to pin down the location of protons. To render the method more generally applicable, Pake extended the theory to powder samples where the spectrum is made up of the superposition of subspectra from a very large number of microcrystals with all possible orientations in space. The result was a characteristic broad line shape - the "Pake doublet". A later collaboration with two chemists, Herbert Gutowsky and George Kistiakowsky8, confirmed that this was indeed a valid approach for determining molecular structure. Chemistry was beginning to gain a foothold in magnetic resonance.
However it was not a simple task to identify suitable chemical applications of Pake's work. In Rex Richards' fledgling NMR laboratory in the 1950's, all the "easy" experiments seemed to have been exhausted. The remaining chemical samples on offer were either alarmingly toxic or explosively unstable, and any junior research student took his life into his own hands. In fact, once an NMR measurement was complete, we carefully buried our samples (after a brief ceremony) in a grassy slope adjacent to the laboratory; that lawn still retains some strange colorations. To make matters worse, the NMR sample had to be cooled to minimize motional effects, and the only available refrigerants at that time were liquid oxygen and liquid hydrogen, an unholy alliance. Our small 12.2 MHz electromagnet was powered by a bank of war-surplus submarine batteries that had to be recharged overnight (just as in the original submarines). There was an inevitable slow drift of the magnet field as the voltage fell off with time. This drift was manually corrected by a research student monitoring a sensitive mirror galvanometer, and compensating the downward trend in current by winding a resistive graphite rod down into a pool of mercury. Just how smoothly the mercury/carbon interface behaved proved to be critical, and we discovered that Eagle Turquoise 6H pencil leads were the only reliable candidates for the graphite rod (the Oxford stationers were bewildered by this unexplained popularity of one particular pencil). We like to think of this tedious scheme as a crude fore-runner of the very successful Varian "superstabilizer" where magnet drift was compensated electronically, using two photocells to replace the research student, and a current feedback system.
My own contribution
10 to these endeavours was to measure the Pake doublet in a powder sample of potassium amide (another unstable compound that had to be prepared and kept under dry nitrogen). Since the spectrometer recorded the first derivative of the NMR absorption-mode signal, integrations had to be calculated, using a planimeter that eventually turned out to be faulty and had to be replaced. In polycrystalline powders the effect of the dipolar interactions with more-distant protons caused a broadening that could be represented by a Gaussian function. Consequently the fitting of the theoretical and experimental lineshapes required some time-consuming calculations, and the available computers of that era were essentially hand-cranked adding machines. This expenditure of blood, sweat and tears was aggravated when a referee (quite reasonably) demanded that the calculations be repeated with slightly different combinations of the inter-proton distance and the Gaussian broadening function in order to obtain an estimate of the accuracy of the result. After allowance for small torsional and vibration motions of the amide group, the H-H distance turned out to be 1.63 ± 0.03 Angstrom units, not exactly a world-shaking discovery. It was this sobering experience that convinced me that solid-state NMR would always be slow and unrewarding work. It was not until many years later that innovations by John Waugh11, and Peter Mansfield12 opened up this field enormously, and solid-state NMR became popular again. Could it be that Linus Pauling's suggestion about NMR and chemistry had been somewhat over-optimistic?
3. The "Do it yourself" era
Initially the most daunting barrier to chemists hoping to try NMR was the realization that a large magnet and complex electronics would have to be assembled from scratch; there were no available commercial NMR spectrometers. Nor was there appreciable funding for purchase of components, since most chemistry laboratories operated on a shoestring budget geared to the purchase of glassware and a few chemicals. In the Physical Chemistry Laboratory at Oxford, Rex Richard did most of the construction work with his own hands. He acquired an old London taxi, the kind with a reinforced platform next to the driver to carry large items of luggage, and used this to transport several large chunks of soft iron, one at a time, to the local car factory at Cowley for machining. He wound the energizing coils of the electromagnet by hand, totalling nine miles of copper wire. All the electronics was from stripped-down war surplus radar equipment, indeed even our nuts and bolts had been carefully rescued from these radar sets. (It is interesting to note that the early pioneers of NMR had all been participants in wartime radar research, and it was this expertise that encouraged them to search for magnetic resonance in bulk matter.) It was clear that the magnet pole caps were critical components as they determined the eventual uniformity of the field, but the existing prescriptions for the best geometry and metallurgy somehow smacked of witchcraft. There are several accounts of endless, patient hand-polishing of pole faces to achieve a flat surface, and even one apocryphal story about a field contour map that was misinterpreted, so that regions making lower contributions to the magnetic field were ground down even further. The direct current power supply also proved to be a tricky item, for this is what determines the stability of the magnetic field. It required a set of large vacuum tubes for current regulation operating in parallel. There were concerns (later discovered to be largely unfounded) that switching off the magnet might induce an unacceptably large surge in current. Eventually our own monster magnet was installed; basic physics had at last insinuated itself into the Oxford chemistry laboratory. Some wag attached a notice that read "Magnet, not to be removed from room 16".
This new apparatus was used to investigate "other nuclei" (anything except the proton) in the liquid state. Surely, with so many different magnetic isotopes in the periodic table, something of chemical significance would come to light? Rather like those early explorers, we set out with the expectation that we would inevitably stumble on some new land to call our own. Our trusted vessel was the "Pound spectrometer" an oscillating detector
9 that allowed one to search the largely empty radiofrequency spectrum in the hope of finding the NMR response. Of course we knew roughly where to look, for the values for these magnetic moments had already been measured by zealous (but misguided) physicists. It was nevertheless a real challenge to locate the tiny blip of an NMR signal in a vast sea of noise, using a homemade spectrometer where the actual strength of the magnetic field was only known very approximately. The Pound box (sometimes known as a regenerative oscillator, autodyne detector, or marginal oscillator) proved well suited to this task, making it easy to sweep the radiofrequency by driving a tuning capacitor with a clock motor and reduction gears. The regenerative feature of the Pound spectrometer served to amplify weak NMR signals to the extent that no further radiofrequency amplification was required. All that remained was the straightforward task of relating the NMR frequency to a standard 5 MHz signal emitted by a radio transmitter located at Daventry.
Investigations were initiated on several of the "other" nuclear species -
7Li, 59Co, 69Ga, 71Ga, 113In, 115In, 203Th, and 205Th. Of these, the cobalt project proved to be the most rewarding. Norman Ramsey13 had just developed an important theory of chemical shifts. Even at the time of the early molecular beam magnetic resonance experiments it was understood that the magnetic field at the nucleus was slightly different from the external applied field owing to the circulation of the atomic electrons (the diamagnetic shielding). In molecules there is also a paramagnetic term that reflects the fact that the electron distribution around an atom that forms part of a molecule no longer has spherical symmetry. Ramsey's calculation of this chemical shielding effect involves a summation over all the excited electronic state wavefunctions. He pointed out that the resulting chemical shift would be particularly large (and temperature dependent) if there were a low-lying electronic state, and suggested that this might occur in cobalt (III) complexes. In this special case one term dominates all the others, and the chemical shift should be inversely proportional to the electronic energy gap Δ(sometimes called the ligand field splitting) It is this particular electronic transition that gives rise to the colour of these cobalt salts, so it can be measured in a standard ultraviolet-visible spectrometer. We measured a series of symmetrical cobalt complexes14 and found that the NMR frequencies (spanning the enormous range of 13,000 ppm) were directly proportional to the corresponding ultraviolet-visible wavelengths, thus corroborating Norman Ramsey's theory. A physical chemist just loves a straight-line graph:
To our chagrin, what seemed at the time to be a rewarding study on 59Co NMR, was soon spectacularly upstaged by developments some 5,000 miles away to the West. Scientists at Stanford University had just invented a new form of spectroscopy that would change chemistry forever.
4. High resolution NMR
In one of those fortunate coincidences that sometimes change the entire direction of science, an Indian postdoctoral chemist Shrinivas Dharmatti, happened to be working in the physics laboratory of Stanford University. Having read about the chemical shifts observed by Warren Proctor and others, he pointed out that almost any organic molecule should show a proton spectrum consisting of many chemically shifted lines, an idea that seems not to have occurred to the physicists at that time for they had focussed on simple materials like water, and had not paid much attention to seeking high resolving power. Using a very small sample and painstakingly searching for a location in the magnet gap where the field was most uniform, Martin Packard15 discovered the famous three-line spectrum of ethanol. Chemistry entered a new golden age. Ask a freshman chemist, and he will happily write down CH3CH2OH for the structure of ethanol. Ask him how we know this to be the case, and his justification becomes quite complicated, relying on a rich history of interconnected reactions investigated by "wet chemistry", including the oxidation of ethanol to acetaldehyde and acetic acid. In contrast, here for the first time was direct evidence for three distinct chemical sites, with the intensity ratio 3:2:1. Martin Packard (a physicist) always asserts that he never believed the structural formulae written by chemists until he saw for himself the three lines of ethanol.
5. Two key doctoral theses
Although the three-line ethanol spectrum is widely accepted as the trigger for the NMR revolution that changed structural chemistry forever, it was the doctoral theses of two young Stanford students, Jim Arnold and Wes Anderson, that really demonstrated the incredibly broad scope of this new form of spectroscopy. When their supervisor Felix Bloch was appointed Director of the European Centre for Nuclear Research (CERN) in Geneva, he took these research students with him, along with their NMR equipment. There they wrote up their thesis work in two classic papers19,20 that show for the first time the incredible potential of high-resolution NMR. All aspiring NMR spectroscopists should read these two seminal publications.
5.1 James T. Arnold
An essential breakthrough was a remarkable permanent magnet designed and constructed by Jim Arnold19. He set out to achieve a degree of uniformity of the magnetic field that no one had ever dreamed was possible (approaching one part in 108). Since the magnetic material (Alnico) had to be magnetized while the magnet gap was closed, one pole of the magnet was mounted on roller bearings, and once the magnet had been energized, the gap was opened hydraulically (against a five-ton magnetic force of attraction) by means of war surplus Boeing B-17 landing-gear struts. The initial charging stage used a current of almost 400 amperes at a power level of 2 megawatts, and was carried out in the middle of the night to avoid any serious disruption of the main Stanford electrical supply. The pole caps, of very pure Armco ingot iron, were ground flat and parallel by an outside contractor; everything else was done in the Stanford physics laboratory. A permanent magnet is quite susceptible to field distortions by nearby iron or steel objects such as tools, key rings, belt buckles or pocket knives. In an amusing overreaction to this challenge, Jim Arnold would strip to his underwear before operating the spectrometer. Resolution was significantly improved by stirring the sample, an idea that came to Felix Bloch21 one day while he was stirring his tea, surmising that this motion would cause the nuclei to experience a magnetic field that was averaged over a circular path. At first, stirring was achieved by means of a rotating glass paddle inside the sample tube, but it was soon found that spinning the cylindrical sample tube about its long axis was simpler and just as effective. This soon became the accepted practice in all high-resolution spectrometers. Great care was exercised in choosing the materials of the radiofrequency probe since the presence of any paramagnetic particles would be disastrous, and even diamagnetic materials induce significant distortions of the field at the sample. All this meticulous work culminated in a magnet with previously unheard-of stability and uniformity, demonstrating for the very first time that an NMR line could be recorded that was only 0.5 Hz wide. Even today this would represent a perfectly acceptable resolving power.
For the very first time the rich complexity of a high-resolution NMR spectrum became evident for all to see. In his new magnet Jim Arnold showed that the three chemically shifted lines of ethanol carried additional fine structure, attributed to a new kind of interaction between adjacent nuclei. This "spin-spin coupling" indicated the number of statistical combinations of "up" or "down" states of neighbouring groups of protons. For example, the methyl resonance was split into a 1:2:1 pattern by the two protons in the adjacent methylene group. The methylene resonance acquired a 1:3:3:1 pattern defined by the four possible combinations of the spin orientations within the adjacent the methyl group. Once spectroscopists had absorbed these new rules for spin-spin multiplet structure, it came as a real disappointment to discover further complicating effects attributable to "strong coupling". These intensity distortions and new splittings could be calculated by second-order perturbation theory
19. In today's spectrometers, operating at far higher magnetic fields, strong coupling effects normally disappear.
Jim Arnold can also take credit for discovering the phenomenon of chemical exchange. In a very pure sample of ethanol the rate of exchange was slow enough that the coupling between CH
2 and OH came into play - the hydroxyl resonance was a 1:2:1 triplet, and the methylene response acquired an additional doubling. In contrast, when very small concentrations of hydrogen ions were added, the exchange rate was increased to the point that the OH resonance reverted to a singlet and the corresponding splittings on the CH2 resonance disappeared. Fast exchange is said to "wash out" the spin-spin splittings. Thus the first organic chemical chosen as an NMR sample brought to light three new phenomena - proton spin-spin coupling, strong coupling effects, and chemical exchange. The extraordinary potential of NMR in chemistry was there for all to see.
5.2 Weston A. Anderson
Wes Anderson's thesis explored the high-resolution spectra of several other simple organic compounds20. In retrospect, his most important discovery concerned the first proton-proton double resonance experiment. Although Robert Pound22 had earlier shown that the double resonance technique could be useful in NMR, Virginia Royden23 had decoupled the spin-spin interaction between 13C and 1H in an enriched sample of methyl iodide, and Arnold Bloom and Jim Shoolery24 had decoupled 19F from 31P, and 31P from 19F, these were relatively straightforward heteronuclear experiments. Wes Anderson's decoupling of one group of protons from another group of protons presented a far more serious experimental challenge. His double irradiation experiment20 on 2,3-dibromopropene (Figure 2) removed the overlying triplet structure on the trans proton (bottom right) by selectively decoupling the resonance of the CH2Br group (arrowed).
In its simplest form homonuclear decoupling identifies groups that are coupled by the spin-spin interaction; today we would say that these chemical sites are correlated.
It is the philosophy behind these double resonance experiments that has had the most important ramifications. At that period, spectroscopy in general was straightforward - there was a direct one-to-one relationship between sample and spectrum, nothing could be changed in the spectrometer itself. Wes Anderson demonstrated for the first time how proton NMR spectra could be manipulated - altered by deliberate intervention of the spectroscopist to extract further useful information. The manipulations soon became more intricate. As one illustrative example, consider the three-spin "AMX" proton spectrum
25. of 2,3-dibromopropionic acid (Figure 3).
We know that the M nuclei consist of almost exactly equal numbers of "up" and "down" spins. Suspend disbelief for a moment, and assume that we can somehow physically separate the sample into two flasks, one containing only the "up" M spins and the other the "down" M spins. (We might usefully invoke James Clerk Maxwell's "demon" which tests the second law of thermodynamics by identifying fast-moving gas molecules and allowing them to pass through a trap door into a second chamber, thereby warming the gas in that chamber.) In our proposed gedanken NMR experiment, a sample taken from the flask with only "up" spins could be examined in the NMR spectrometer. Disregard for the moment the M response; imagine M as merely a spectator. A sample taken from the first flask would now show the A and X responses reduced to the "red" doublets, whereas a sample from the second flask would show only the "blue" doublets. Our Maxwell demon has succeeded in separating two subspectra. Now in making this assignment we have tacitly assumed that the two coupling constants JAM and JMX have like signs. If these signs are opposite, then the colour coding of the upper trace should be changed to that shown in the lower trace. Of course we cannot really call on the services of a Maxwell demon, but in the real world the equivalent of this hypothetical experiment can indeed be performed by double resonance. A selective irradiation experiment can be set up to distinguish the upper spectrum of from the lower spectrum. By suitably choosing the level of irradiation, it is feasible to irradiate the "red" A doublet (red arrow) without appreciably affecting the "blue" A doublet, thus decoupling the "red" X doublet without affecting the "blue" X doublet. The experimental double resonance response (Figure 3, bottom right) corresponds unequivocally to the case of opposite signs. Consequently JAM (a geminal coupling) is opposite in sign to JMX (a vicinal coupling). This result was of some significance at the time because theoretical calculations26 had predicted that geminal and vicinal couplings should all be positive.
This one example serves to demonstrate that manipulation of NMR spectra by double resonance can reveal quite subtle properties of the molecule. Extension of this technique by Forsén and Hoffman
27 led to studies of slow chemical exchange, where irradiated spins were physically transferred to a second chemical site by the exchange process. A second powerful application, now widely employed for structural studies, is the evaluation of proton-proton distances by the nuclear Overhauser effect28. In a similar manner, broadband irradiation of protons is now used routinely to simplify 13C spectra and to obtain a roughly three-fold signal enhancement by the nuclear Overhauser effect29. Much later, in 1971, Jean Jeener30 showed that most double resonance experiments could be more efficiently implemented by two-dimensional NMR, and this soon became the standard "manipulation" procedure. Seldom is an NMR spectrum recorded today in its pristine, unmodified, one-dimensional form. We could argue that Wes Anderson's experiments provided the first impetus to the development of the art of "spin choreography" that is fundamental to modern multidimensional NMR. We have all become manipulators.
6. Electromagnets
The high-resolution spectrum of ethanol highlighted the need for high stability in any magnet designed for chemical applications of proton magnetic resonance. There is no point in looking for fine detail on a picture that is moving and shaking. Harry Weaver in the Stanford group designed and built an electromagnet that operated at a proton frequency of 30 MHz, chosen because the readily available war surplus radar equipment employed an intermediate frequency amplifier (IF strip) at this very frequency. The first step towards improving stability was to regulate the current by means of eight parallel 304TL vacuum tubes. (These later became an alarming feature of the early Varian spectrometers; the electrodes become red-hot.) The next step was the idea of "field/frequency" regulation, based on a proton "error signal" from a second probe in the magnet gap. One such scheme is described in an early patent by Russell Varian31 and the first practical implementation was demonstrated by Baker and Burd32. Another ingenious scheme was invented by Robert Pound15 who employed a pulse-modulated super-regenerative oscillator where the phase of the radiofrequency pulses is triggered by the NMR signal itself, so that the oscillator frequency automatically follows any deviations of the magnetic field. Mention has already been made of the "research student" drift compensation method employed in Rex Richards' laboratory. In a similar spirit, Varian built a very effective "super-stabilizer" that worked by sensing small field changes with a pick-up coil feeding a sensitive galvanometer that shone a light beam onto a pair of photocells. The field variations were compensated by a correction current in coils attached to the pole faces. Wes Anderson16 devised a regulation scheme based on a "nuclear sideband oscillator" that permitted the new Varian A60 spectrometer to use precalibrated charts. It worked so well that Jim Shoolery tells the story that he was amazed to be able to record an NMR spectrum twice, with the recorder pen exactly retracing the first spectrum - proof that the regulation scheme was very effective indeed. A final perfection was suggested by Hans Primas17 who realized that an NMR signal from an internal reference (the twelve equivalent protons of tetramethylsilane) would provide the best possible regulation because it avoids the possibility of field changes between two geometrically separated probes. This idea proved particularly useful for performing frequency-sweep double resonance experiments18. Stabilization via an internal reference signal soon became the accepted practice; later the proton signal from tetramethylsilane was replaced by a deuterium signal from a deuterated solvent, often CDCl3.
Spatial uniformity (homogeneity) of the magnetic field is the second essential requirement. Not surprisingly, this represented an enormous challenge to magnet technology. Expressed in terms of an astronomy analogy, the resolution today achievable in NMR (approaching 1 part in 10
9) would correspond to a telescope capable of resolving the images of two cats sitting side by side on the surface of the moon. Before the work at Stanford, no one had foreseen that this degree of field uniformity could be achieved, or even felt that there would be any need for it. Considerable thought was given to the design of pole caps, both with respect to metallurgy and geometry, since these are critical in determining the uniformity of the field. Electromagnets at that time had to cycle the initial current according to a prescription that ensured a slightly "dished" configuration of the magnetic field. A later initiative was to shape the edges of the pole caps according to a logarithmic function. Field distortions due to magnetic susceptibilities of the copper wire used for the radiofrequency coil were eventually corrected by devising a sandwich of two conducting materials with opposite signs of their diamagnetic susceptibilities.
However it soon became clear that the most general solution to the resolution problem was to accept that there would always be some residual field inhomogeneities, but to compensate them empirically with a set of correction coils. Based on calculations in terms of a spherical harmonic expansion, Marcel Golay and Ed Jaynes independently designed correction coils where the adjustments of individual gradients were approximately independent. Golay
33 made his first correction coils by embroidering them on calico; later printed circuits were used. Wes Anderson34 designed a set of coils that were generally adopted for all Varian high-resolution magnets. There was much talk of orthogonality of these "shim coils" although in practice the adjustment of the different field gradients always involves some interaction, and considerable patience and skill is still required to set up the most uniform magnetic field.
7. Permanent magnets
As mentioned above, Jim Arnold opted for a permanent magnet, thus avoiding many of the stability problems inherent in electromagnets. This seemed like the preferred approach at that time, despite the fact that the maximum achievable magnetic field was rather less than that of a well-designed electromagnet. One early pioneer of high-resolution NMR, Herbert Gutowsky35 adopted a permanent magnet for his home-built spectrometer. Characteristically for a chemist, Gutowsky's turbines for spinning the sample were carefully bored corks driven by an air jet.
At Oxford, Rex Richards
36 also decided to assemble a high-resolution spectrometer based on a prototype permanent magnet constructed by the Mullard Company for a sum of £2,000; it operated at a proton frequency of 30 MHz. Since a permanent magnet is quite sensitive to temperature, it was placed in an insulated enclosure with thermostatic control. The detection of the NMR signal was achieved by means of a commercial communications receiver. The very first high-resolution spectrum was rather disappointing - it seemed to be a band of pure noise - but closer examination showed some narrow gaps where the noise had disappeared. The fact that the receiver incorporated an automatic gain control had been overlooked, and once this circuitry was disabled a beautiful high-resolution spectrum was revealed.
8. Commercial NMR Spectrometers
The full impact of NMR in chemistry had to await the advent of working spectrometers that could be purchased from commercial companies, since few chemists were prepared to make the big investment in time and effort to build their own spectrometers from scratch. These instrument companies included Varian Associates in California, Japan Electron Optics Ltd. in Japan, Associated Electrical Industries in England, Perkin Elmer Corporation in England, and Trüb Täuber (later Bruker Spectrospin) in Switzerland. The complexity of these machines and the limited market meant that this was not really a profitable business venture. Varian NMR prospered by being part of a larger company that made profitable microwave devices; indeed the Varian brothers (Russell and Sigurd) had deliberately embarked on magnetic resonance in the full knowledge that it would not be a viable commercial proposition. As protection against competition, Varian Associates had acquired the Bloch-Hansen patent that covered the NMR phenomenon itself, and Bloch's spinning sample patent.
These early high-resolution spectrometers employed large electromagnets and were not easy to operate or maintain. The best results required careful regulation of the temperature of the water used to cool the energizing coils, and often the magnet was housed in an insulating enclosure to avoid the effects of changes in ambient air temperature. Stability was greatly improved by the introduction of the "superstabilizer" that monitored field changes and compensated them with a current feedback scheme (see above). The definitive breakthrough came with the introduction of the Varian A60 in 1961. This spectrometer was designed specifically for the chemist who had no particular desire to become an expert in radiofrequency electronics. It was a simplified, slimmed-down version of the earlier Varian machines, with a much smaller magnet, power supply and cooling system, all in a single enclosure. For the first time, a minimalist design was adopted for the operating controls, making this a truly user-friendly NMR spectrometer (some critical controls were hidden behind a trapdoor to discourage incorrigible knob-twiddlers). The console was dominated by a wide flat-bed recorder that boasted a large paper chart precalibrated in ppm, reproducibility being guaranteed by a regulation scheme based on Wes Anderson's nuclear sideband oscillator. By defining a standard layout for the NMR spectrum on a calibrated chart, this encouraged chemists to acquire a "feel" for interpretation of spectra, and to adopt an empirical approach to spectral assignment.
The A60 opened the way for a much wider acceptance of high-resolution NMR among organic chemists. This was deliberately encouraged by the Varian personnel, who toured the chemistry departments of universities and industries to explain the basic principles of this alien new technology, and to demonstrate how beautifully it could solve structural problems. They were led by a great enthusiast, Jim Shoolery, one of the first chemists to fully appreciate the enormous potential of the technique. Every effort was made to keep the talks pictorial and non-mathematical, as a profound knowledge of the fundamental physics was not considered essential for the acquisition of NMR spectra. This educational initiative was reinforced by the introduction of a series of NMR "workshops" where owners and prospective owners could acquire "hands on" experience of the new machines. Although limited to proton studies, this can be regarded as the golden age of high resolution NMR. Almost any organic compound taken down from the shelf would yield an interesting proton spectrum.
9. Sensitivity
Even during the euphoria of the early 1960's some scientists began to ponder the question of poor NMR sensitivity, limited by the low Boltzmann factor of nuclear spin populations at room temperature. The first attack on this problem came from Klein and Barton37 who converted a digital pulse-height analyser into a device that added together successive NMR traces, thereby increasing the signal-to-noise ratio by the square root of the number of accumulations. At an American Chemical Society "Meeting in Miniature" in December 1962, the audience was amazed by Melvin Klein's practical demonstration of an NMR trace that improved in quality before our very eyes. This seemed like magic. About the same time Jardetsky et al38 employed a commercially available Mnemotron "computer of average transients" to enhance the NMR signals of some biochemical samples. Very quickly the time averaging method was generally adopted. The only drawback was the law of diminishing returns imposed by the square root dependence; high signal enhancement factors inevitably meant unacceptably long time-averaging durations. The practical limit appeared to be signal accumulation over an entire weekend, provided that the spectrometer was not required for more important work.
9.1 Multichannel excitation
It was Wes Anderson who realized that the conventional field sweep (or frequency sweep) procedure was inherently inefficient with regard to sensitivity. At any given instant, the spectrometer receiver detects the intensity at only one point on a particular resonance line in a spectrum of many lines, indeed at other instants it wastes time gathering only baseline noise. Wes suggested that a multichannel irradiation device that excited all frequency regions simultaneously would be far more efficient. With N independent excitation channels the signal-to-noise ratio should improve as √N, up to the approximate limit of one channel per line width. Wes loves mechanical gadgets, so he set out to design a device to generate a set of N regularly spaced audio frequencies. A rotating hollow Lucite cylinder had a strip light running down the central axis (Figure 4).
The outer circumference was machined horizontally in 20 vertical steps with linearly increasing radii. Strips of photographic film with alternating dark and light bands were glued around the different levels so that as the cylinder rotated, the light that fell on a vertical array of photoelectron detectors was chopped at a set of gradually incremented frequencies. These were then used to generate a regular "comb" of modulation sidebands that covered the entire proton NMR spectrum. The same cylinder could be used in reverse to demodulate the received NMR signal. The device became known as the "Prayer Wheel", being reminiscent of those used by Buddhist monks to enhance their prayers by constant repetition.
The prayer wheel was never used in anger because a more promising method of broadband excitation emerged, and Wes reluctantly allowed his brainchild to be taken away and installed in the Smithsonian Institution in Washington. This may stand as a lesson to all, that certain seminal ideas can turn out to be more important than their actual practical implementation. Much later it became possible to vindicate the concept of multichannel excitation by using modern waveform generator technology to excite part of the 500 MHz proton NMR spectrum of strychnine (Figure 5) with 2048 separate irradiation frequencies spaced 1Hz apart, using Hadamard encoding to differentiate the resulting 2048 responses
39. There is no perceptible difference between this multiplex spectrum and that recorded by the conventional Fourier transform method.
9.2 Fourier transform NMR
The approach that superseded the prayer wheel was based on an idea for broadband excitation of the entire NMR spectrum by an intense radiofrequency pulse, followed by Fourier transformation of the resulting transient free induction decay. This concept was first formulated by Russell Varian in a patent filed on 29 August 1956. This revolutionary patent even contained suggestions for an analogue Fourier transformation scheme based on repetitive reading of the transient free induction signal stored on a loop of magnetic tape, picking out the individual frequency components by means of a variable-frequency lock-in detector. In his laboratory notebook (dated 3 June 1964) Wes Anderson sets out a practical block diagram (Figure 6) for pulse excitation of all the different NMR frequencies in a proton spectrum, and a scheme for sorting the different frequency responses by Fourier transformation.
The method relies implicitly on two important features - time averaging of many successive free induction decays, and the availability of a suitable computer for Fourier transformation.
Practical implementation of the concept was greatly hampered by the complications involved in digitising the transient NMR signal, and the difficulty in finding a computer to perform the Fourier transforms. Small laboratory computers had yet to be invented. One very cumbersome approach - the spectrum analyser, repetitively examines the time-domain signal and slowly picks out the component frequencies one by one. It is interesting to note that the first digital computer programs for Fourier transformation employed an analogous stepwise examination of the spectral frequencies, entailing a delay of 20 minutes or so for a typical transformation. Completion of the transform in a more reasonable time (seconds) had to await the Cooley-Tukey fast Fourier transform algorithm
At Varian at that time it happened that Richard Ernst had been working on methods to improve the sensitivity of the conventional NMR sweep protocol by optimizing the various operating parameters, so he was naturally well primed to attack the proposed new Fourier transform scheme. It is greatly to the credit of Ernst and Anderson
41 that they persevered with this daunting project even though it appeared rather impractical at the time. Digitization and temporary storage of the accumulated transient NMR signal involved a tedious conversion, first to paper tape and then to IBM punched cards - a whole box of cards was used to carry a single free induction decay to a central computer located at the far end of Palo Alto. The large IBM 7090 computer at the local Service Bureau relegated such "trivial" calculations to overnight runs. Typically, the first spectrum would need readjustment of the phase (and the phase gradient) to give pure absorption-mode signals, requiring another round trip to the service bureau. Not many chemists were prepared to wait a matter of days for a properly phased NMR spectrum. Nevertheless the final results were impressive - an order of magnitude improvement in proton sensitivity compared with the conventional sweep method executed in the same total time. It is a sad reflection on our peer-review system that referees twice contrived to find reasons to reject Ernst and Anderson's manuscript submitted to a respected physical chemistry journal; the story eventually appeared in press in 1966. NMR spectroscopy was changed forever.
The most significant outcome of this new Fourier transform spectroscopy was a gift to the chemists that they had always desired, but had been afraid to request -
13C spectra from samples of natural isotopic abundance (1.08%). Carbon defines the basic framework of any organic molecule. Because 13C spectra are simpler and the information is more widely dispersed than in proton spectra, they offer a more reliable fingerprint for identification and structural determination. The Fourier transform made this possible, aided by an important spectral enhancement technique - broadband irradiation of the protons, which greatly simplifies 13C spectra by removing the 13C-H fine structure and inducing a roughly threefold signal enhancement through the nuclear Overhauser effect. The first flowering of NMR in chemistry was now virtually complete.
The Encyclopedia of Magnetic Resonance, Volume 1: Historical Perspectives (John Wiley & Sons, 1996; Editors in Chief: David M. Grant and Robin K. Harris) proved very useful for checking chronology, references, and some half-remembered facts about the early days of NMR in chemistry. I thank the NMR pioneers (Rex Richards, Anatole Abragam, Ionel Solomon, Bob Pound, John Pople, David Whiffen, Warren Proctor, Martin Packard, Jim Arnold, Wes Anderson, Jim Shoolery, Richard Ernst, Jean Jeener) all of whom I was fortunate to know personally, and I ask for their forbearance if my memory of some details has not been as accurate as they might wish.
1.  E. M. Purcell, H. C. Torrey and R. V. Pound, Phys. Rev. 69, 37 (1946).
2.  F. Bloch, W. W. Hansen and M. E. Packard, Phys. Rev. 69, 127 (1946).
3.  W. G. Proctor and F. C. Yu, Phys. Rev. 77, 717 (1950).
4.  W. C. Dickinson, Phys. Rev. 77, 736 (1950).
5.  G. Lindström, Phys. Rev. 78, 817 (1950).
6.  H. A. Thomas, Phys. Rev. 80, 901 (1950).
7.  G. E. Pake, J. Chem. Phys. 16, 327 (1948).
8.  H. S. Gutowsky, G. B. Kistiakowsky, G. E. Pake and E. M. Purcell, J. Chem. Phys. 17, 972 (1948).
9.  R. V. Pound, Phys. Rev. 72, 527 (1947).
10. R. Freeman and R. E. Richards, Trans. Faraday Soc. 52, 802 (1956).
11. J. S. Waugh, L. M. Huber and U. Haeberlen, Phys. Rev. Lett. 20, 180 (1968).
12. P. Mansfield, Phys. Lett. 32A, 485 (1970).
13. N. F. Ramsay, Phys. Rev. 78, 699 (1950).
14. R. Freeman, G. R. Murray and R. E. Richards, Proc. Roy. Soc.(London) A. 242, 455 (1957).
15. R. V. Pound and R. Freeman, Rev. Sci. Instr. 31, 96 (1960).
16. W. A. Anderson, Rev. Sci. Instr. 33, 1160 (1962).
17. H. Primas, Fifth European Congress on Molecular Spectroscopy, Amsterdam (1961).
18. W. A. Anderson and R. Freeman, J. Chem. Phys. 37, 85 (1962).
19. J. T. Arnold, Phys. Rev. 102, 136 (1956).
20. W. A. Anderson, Phys. Rev. 102, 151 (1956).
21. F. Bloch, Phys. Rev. 94, 496 (1954).
22. R. V. Pound, Phys. Rev. 79, 685 (1950).
23. V. Royden, Phys. Rev. 96, 543 (1954).
24. A. L. Bloom and J. N. Shoolery, Phys. Rev. 97, 1261 (1955).
25. R. Freeman and D. H. Whiffen, Mol. Phys. 4, 321 (1961).
26. M. Karplus, J. Am. Chem. Soc. 84, 2458 (1962).
27. S. Forsén and R. A. Hoffman, Acta Chem. Scand. 17, 1787 (1963).
28. F. A. L. Anet and A. J. R. Bourn, J. Am. Chem. Soc. 87, 5250 (1965).
29. K. F. Kuhlmann and D. M. Grant, J. Am. Chem. Soc. 90, 7355 (1968).
30. J. Jeener, Ampere International Summer School, Basko Polje, Yugoslavia (1971).
31. R. H. Varian, U.S. Patent 3,109,732, Filed 29 August 1956.
32. E. B. Baker and L. W. Burd, Rev. Sci. Instr. 28, 313 (1957).
33. M. J. E. Golay, Rev. Sci. Instr. 29, 313 (1958).
34. W. A. Anderson, Rev. Sci. Instr. 32, 241 (1961).
35. H. S. Gutowsky C. J. Hoffmann and R. E. McClure, Phys. Rev. 81, 305 (1951).
36. J. B. Leane, R. E. Richards and T. P. Schaefer, J. Sci. Instr. 36, 230 (1959).
37. M. P. Klein and G. W. Barton, Rev. Sci. Instr. 34, 754 (1963).
38. O. Jardetsky, N. G. Wade, and J. J. Fischer, Nature (London) 197, 183 (1963).
39. E. Kupce and R. Freeman, J. Magn. Reson. 162, 158 (2003).
40. J. W. Cooley and J. W. Tukey, Math. Comput. 19, 297 (1965).
41. R. R. Ernst and W. A. Anderson, Rev. Sci. Instr. 37, 93 (1966).
Ray Freeman FRS
Ray Freeman
© 2012 Ray Freeman Web site by Quincunx