《天然药物化学》课程教学资源（文献资料）Ernst, Richard R. - The Success Story of Fourier Transformation in NMR.pdf

Ernst, Richard R.: The Success Story of Fourier Transformation in NMR Richard R. Ernst Eidgen¨ossische Technische Hochschule, Z¨urich, Switzerland This chapter could also be entitled: How to be successful by borrowing clever ideas. During my undergraduate studies at the Swiss Federal Institute of Technology in Zurich ¨ (ETH Zurich) from 1952–56, the term nuclear magnetic ¨ resonance was never even mentioned, and when I started my thesis with Professor Hans H. Gunthard in 1958 after ¨ some healthy military training I knew nothing about my beloved future companion NMR. It was, so to say, a marriage arranged by my thesis advisor. Fortunately, an extraordinarily gifted, truly ingenious scientist, Hans Primas, was just in the course of finishing the construction of the first Swiss highresolution NMR spectrometer. I had the enormous luck to work under his guidance on my Ph.D. thesis for the following 4 years. HANS PRIMAS AND HIS NMR SPECTROMETERS Professor Hans H. Gunthard, at that time Associate ¨ Professor in the Laboratorium fur Organische Chemie, later ¨ Head of the Laboratorium fur Physikalische Chemie, was an ¨ enormously enterprising and stimulating leader (Figure 1). He set out to introduce all powerful modern spectroscopic tools into the Swiss university chemistry. It should also be mentioned here that he was strongly encouraged by Professor Leopold Ruzicka who was heading the Laboratorium fur¨ Organische Chemie with much foresight. At the early date of 1953, Professor Gunthard asked Hans Primas to pioneer ¨ the construction of a high-resolution NMR spectrometer useful for chemical applications (Figure 2). During a brief stay at the University of Zurich in 1953 with Professor Hans ¨ Staub, who was a former co-worker of Felix Bloch at Stanford, and with his associate Ernst Brun, Hans Primas was exposed for the first time to the practical aspects of NMR, here in the hands of enterprising solid state physicists. The 25 MHz proton resonance spectrometer that Primas and Gunthard constructed used a 0.6 T permanent magnet (Figure ¨ 3), spherical sample containers to improve the magnetic field homogeneity,1 – 4 and a field flux stabilizer.5 It was an important event when in 1956, Gunthard and Primas convinced ¨ the Swiss company Trub-T ¨ auber, with its scientific director ¨ Dr. L. Wegmann, to manufacture and commercialize this spectrometer under the name KIS I.6 Several instruments of this type were sold in Europe during the following years. Hans Primas decided to continue his engineering ventures by designing a high-field instrument working at 75 MHz proton Figure 1 Professor Hs.H. Gunthard, the spinning nucleus of the ¨ research group in 1958 Figure 2 Professor Hans Primas working on his electronic design for the 25 MHz NMR spectrometer in 1958 resonance, and he developed a new concept of shaped pole caps for electromagnets to achieve high-field homogeneity by avoiding local saturation.7 This spectrometer was also commercialized as KIS II, operating at a proton resonance frequency of 90 MHz. The company Trub-T ¨ auber was dissolved in 1965, and ¨ its NMR spectroscopy division became the germ for the well-known Swiss company Spectrospin AG in Fallanden, ¨ which is a member of the Bruker group. In some sense, the instruments built by Hans Primas are the ancestors of the modern high-resolution Bruker-Spectrospin spectrometers. Besides numerous instrumental innovations, Hans Primas also contributed during this time to new theoretical concepts such eMagRes, Online © 2007 John Wiley & Sons, Ltd. This article is © 2007 John Wiley & Sons, Ltd. This article was previously published in the Encyclopedia of Magnetic Resonance in 2007 by John Wiley & Sons, Ltd. DOI: 10.1002/9780470034590.emrhp0051

2 RICHARD R. ERNST Figure 3 The 6 kG permanent magnet for the 25 MHz proton magnetic resonance spectrometer in 1958. The coils for the magnetic flux stabilizer are visible on the pole pieces. The probe assembly is directly attached to the preamplifier. The radiofrequency transmitter stands on top of the magnet as a generalized perturbation theory in operator form8 and superoperator concepts for calculating NMR spectra.9,10 FIRST VENTURE IN STOCHASTIC RESONANCE I never even started my first thesis subject posed by Professor Gunthard in 1958, i.e. a group theoretical analysis ¨ of NMR spectra. After extensive ventures into electronics and the construction of NMR probe assemblies, frequently burning my fingers with solder and high voltages, I finished my graduate studies in 1962 with a purely theoretical investigation of stochastic magnetic resonance. The pertinent ideas for this study came from Hans Primas, who at that time was working on the theory of quantum mechanical systems with a stochastic Hamiltonian,11 inspired by the work of Norbert Wiener12 and others. The application of stochastic excitation for obtaining spectral information with an inherent multiplex advantage was not considered at that time. Rather, stochastic excitation was intended for decoupling purposes. By a broadband excitation of all spins except for one within a homonuclear spin system, broadband spin decoupling of this selected spin was attempted. Sweeping a broadband frequency source with a narrowband hole in its spectrum through the entire spectral range should lead to a completely decoupled homonuclear spectrum, which could be observed with a weak coherent irradiation in the center of the hole. It was disappointing for us to recognize that the principle only functions for weakly coupled spin systems, a situation that needed, in our view, anyway no further simplication. Weak coupling was too simple to appeal to us, and we missed the possibility of inventing stochastic heteronuclear decoupling. No wonder that we never did a stochastic resonance experiment. Our aim was more a theoretical excursion to explore the nonlinear stochastic response of a quantum mechanical system by means of an example.13 THE SENSITIVITY DEFECT OF NMR Anyone who performed NMR experiments in the 1950s will remember the exceedingly low sensitivity of NMR at that time. A signal-to-noise ratio of 5:1 for the spectrum of a 1% ethylbenzene solution was considered to be excellent. Indeed, nobody could envision the 1000-fold increase that would happen in the next 30 years. We thought a lot at that time about sensitivity enhancement techniques. Special low-noise preamplifiers using ceramic tubes of the type 7077-G-E- 60-17, optimized probe assemblies, and electronic filtering techniques were considered.14 Noise calculations of probes and of complete spectroscopic experiments were a daily routine in the laboratory of Professor Gunthard. This was ¨ normally the last resort for completing a Ph.D. thesis if the constructed spectrometer did not perform properly. However, this taught us more about response theory, Fourier transforms, and noise figures than any working spectrometer would have done. This was the time of the computer of averaged transients, CAT, which helped to improve sensitivity by the coaddition of slow passage spectra.15 At that time we advanced the idea of using one ultraslow passage instead of repeated scanning through the same spectrum. In this way we missed some important facts that became clear to me only after my arrival in the fall of 1963 at Varian Associates in Palo Alto for my postdoctoral years. Under the guidance of Dr. Weston A. Anderson, I started extensive investigations of rapid scan performance conditions based on numerical solutions of the Bloch equations. They revealed the possibility of significantly enhancing the sensitivity per unit time by rapid and repetitive scanning through the spectrum,16 – 18 facts that were already well known to scientists such as Bitter19 and Jacobsohn and Wangsness.20 However, at that time nobody recognized the possibility of using computers to remove the distortions which the rapid scan introduced into the spectrum. FOURIER TRANSFORM NMR SPECTROSCOPY I remember in the spring of 1964, when the calculations of performance conditions were in full progress, Wes Anderson came to me and mentioned the possibility of applying eMagRes, Online © 2007 John Wiley & Sons, Ltd. This article is © 2007 John Wiley & Sons, Ltd. This article was previously published in the Encyclopedia of Magnetic Resonance in 2007 by John Wiley & Sons, Ltd. DOI: 10.1002/9780470034590.emrhp0051

RICHARD R. ERNST 3 Figure 4 Dr. Weston A. Anderson and his Little Devil Prayer Wheel which he constructed for multiple channel NMR spectroscopy in 1963 repetitive pulses for the simultaneous excitation of all nuclear resonances and using a computer to analyze the response. I was asked to sign, as a witness, the important notebook pages by Wes Anderson that contained the principles of pulse Fourier spectroscopy. I was not very aware at that time of Wes’s previous work on the Wheel of Fortune (Figure 4); in fact, I never saw it. I do not remember whether Wes showed me at the same time or slightly later the highly important patent application by Russell Varian21 which already contained the basic idea. There was no rf pulse equipment available at Varian at that time as everybody was fully devoted to continuous wave spectroscopy. However, as a result of masterly support by the electronic engineer William Siebert, within two months a proton-pulse spectrometer with 300 W pulse power using a fluorine field/frequency lock without any magnetic field modulation was assembled. The crucial first experiments were performed during the summer months of 1964 while Wes Anderson was on an extensive journey abroad (Figure 5). When he came home, I was able to hand him the first successful results that surprised me as much as him. However, there was no reason to be overenthusiastic and we certainly did not think in terms of a champagne party. When one considers the cumbersome treatment of the data acquired in a time averaging CAT 400 computer, into paper tape, being converted into punched cards at IBM San Jose, then converted to magnetic tape at the Service Bureau Corporation, Palo Alto, Fourier-transformed on an IBM 7090, and plotted on a Calcom plotter, nobody Figure 5 Early experiments performed on 16 September 1964. After a few experiments on benzene, a complicated spin system, 1-bromo-4- fluorobenzene, was also treated on the first day. The figure shows the FID and a CW spectrum. The Fourier transform of the FID has been lost Figure 6 The author in front of the modified DP-60 NMR spectrometer on which the first Fourier transform NMR experiments were performed at Varian Associates, Palo Alto, CA, in 1965. The author is acting as a human interface between the time averaging computer C1024 and an IBM card punch could have been convinced by us of a time-saving advantage! Later, we had our own card punch to shorten the data pathway (Figure 6). Nevertheless, we decided to publish the work in the Journal of Chemical Physics. However, we had no success; the paper was rejected twice, being considered too technical and not of sufficient originality. We then submitted the paper to Review of Scientific Instruments on 9 July 1965 where, after further revision, it was finally accepted on 16 September 1965.22 I also discussed the subject at the 6th Experimental NMR Conference in Pittsburgh, 25–27 February 1965. In eMagRes, Online © 2007 John Wiley & Sons, Ltd. This article is © 2007 John Wiley & Sons, Ltd. This article was previously published in the Encyclopedia of Magnetic Resonance in 2007 by John Wiley & Sons, Ltd. DOI: 10.1002/9780470034590.emrhp0051

4 RICHARD R. ERNST general the response was not overwhelming, except from a few people with foresight like Oleg Jardetzky. The Varian management was also not too excited. With little enthusiasm, an accessory was developed for the HR-100 spectrometer; however, that never performed properly. Even the newly developed XL-100 concept, which was worked out in the following years, was designed with a field-modulation system that prohibited the implementation of pulse Fourier transform spectroscopy. Ultimately, it was the competitive German company Bruker Analytische Messtechnik with Toni Keller and Professor Gunther Laukien which demonstrated and sold ¨ the first properly designed Fourier transform equipment in 1969. With much foresight, they brought the first Fouriertransform-only spectrometer HX 90 onto the market in 1971. HETERONUCLEAR NOISE DECOUPLING Being out of better ideas for the time being, it was tempting in May 1965 to try some of the ideas left over from my thesis,14 thus justifying post mortem my otherwise spoiled best years. Broadband spin decoupling indeed appeared to be a worthwhile goal to me, especially for heteronuclear spin systems. The first experiments concentrated on a flourine–proton system. The necessary binary random noise was taken initially from the even/odd character of the last digits in the Palo Alto telephone directory, modulated onto the radiofrequency carrier, and applied to the protons, observing the fluorine resonance at the same time.23 This led to the initial pseudonym phone-book resonance. Only later was I informed by R. Whitehorn (Varian Associates) that binary noise could be generated in a more convenient and predictable manner by shift register generators24 that were used from September 1965 onwards. Numerous further modulation schemes such as repetitive 180 ◦ and other pulses were used and it was found that primarily the spectral properties had a significant effect, but otherwise the type of sequence was immaterial. There was no indication whatsoever in my experiments for the dramatic improvement that was discovered much later by Malcolm Levitt and Ray Freeman25 using composite pulse sequences. COMPUTERIZED SPECTROSCOPY Until 1966, Varian Associates did not possess even a single computer, and all the Fourier NMR experiments I undertook at Varian used off-company data processing. Finally, in October 1966, the first PDP8 computer with 4096 words of 12-bit memory arrived, and I interfaced it with the help of William Siebert to the most widely used routine NMR spectrometer, the A-60. This prohibited the use of pulse Fourier transform experiments for which the A-60 is unsuited by design. Numerous other applications were programmed and tried in a rapid sequence in view of a presentation at the 8th ENC, 2–4 March 1967, in Pittsburgh. The experiments included signal averaging with automatic drift compensation by searching for the TMS line, resolution enhancement by convolution, spectra analysis by multiplet identification, coding of spectra for a subsequent automatic library search, automatic 2D shim mapping, and automatic shimming by use of the simplex search algorithm. Except for the automatic shimming procedure,26 this work was not published at that time as it was felt to be of little originality. A later account can be found in Ref. 27. RETURN TO THE HOMELAND My return to Switzerland in the spring of 1968 proved to be a scientific disadvantage. Unsuitable equipment and the lack of support at the Laboratorium fur Physikalische Chemie, ¨ ETH Zurich, made progress difficult. I had to supervise an ill- ¨ suited Varian 220 MHz spectrometer that was strictly limited to continuous wave operation without a field/frequency lock system. With a streetcar line in the neigborhood, this turned out to be a disaster. Despite a homemade lock system, the spectrometer never performed satisfactorily. Scientific output was low, and after two years of struggling I was affected by a nervous breakdown in 1970. Looking back, I am tempted to rate my years from 1968 to 1974 as the dark Middle Ages after the productive and inspiring classical years in Palo Alto. We tried numerous new approaches, but none with an epoch-making breakthrough. (Personally, I became interested in the mysterious and fascinating Tibetan art during this time period.) With my first graduate student, Thomas W. Baumann (thesis 1969–74), we ventured on our first exploratory excursion into solid state NMR, constructing a 1.5-T solid state spectrometer. We devoted much time on cross polarization in the laboratory frame (35Cl → 1H) and in the rotating frame (1H → 13C). Our time-consuming solid state activities became more exciting in 1973 with the work of Luciano Muller which will be described ¨ later. My second graduate student, Dieter Welti (1970–76), together with Max Linder, explored the use of paramagnetic shift and relaxation reagents in NMR, using borneol and 1-propanol as model compounds. The question of accurate structure determinations using the multifaceted shift and relaxation effects was our primary interest.28 Just to keep our obstinate HR-220 Varian spectrometer busy, we started a series of investigations on simple molecules dissolved in a nematic phase. After the study of a few four-ring compounds by Christian Oertli, we explored, together with Alexander Frey, the distortion of cubic molecules by liquid crystalline solvents.29 Again the determination of molecular structures was our main interest. I was still interested in the slowly emerging Fourier transform spectroscopy, and I studied some subtle effects of pulse excitation such as the equivalence of slow passage and pulse spectroscopy,30 saturation effects in Fourier spectroscopy, together with my first postdoc Robert E. Morgan,31 and the causality principle in Fourier spectroscopy, together with my third graduate student Enrico Bartholdi.32 Motivated by the notorious field instability in Zurich, we also proposed at that ¨ time the usage of difference-frequency measurements to avoid the need for a field/frequency lock system.33,34 Another series of studies was devoted to the application of pulse spectroscopy to chemically induced dynamic nuclear eMagRes, Online © 2007 John Wiley & Sons, Ltd. This article is © 2007 John Wiley & Sons, Ltd. This article was previously published in the Encyclopedia of Magnetic Resonance in 2007 by John Wiley & Sons, Ltd. DOI: 10.1002/9780470034590.emrhp0051

RICHARD R. ERNST 5 polarization (CIDNP). It appeared to us very natural to investigate time-dependent phenomena with time-resolved measurement techniques. This was the subject of the thesis of Stefan Schaublin (1971–76). For a short period he was supported ¨ by Alexander Wokaun.35,36 This work also led to the recognition of the important intensity problems when investigating nonequilibrium populations by pulse experiments.37 Later, these studies were extended to stopped flow experiments by Rene O. K ´ uhne and Thierry Schaffhauser. ¨ 38 STOCHASTIC RESONANCE Pursuing my old love and motivated by a correspondence with Wes Anderson, in June 1968 I started the first attempts to compute and measure NMR spectra via stochastic noise excitation of the sample and by Fourier transforming the cross-correlation function that relates input and output noise.39 The first experiments already used shift register sequences as sources of binary pseudorandom noise.24 At virtually the same time, Reinhold Kaiser also proposed stochastic resonance for recording spectra, but using Gaussian noise to excite the spin system.40 Subsequently, the theory of stochastic resonance was worked out in great detail in the thesis of Enrico Bartholdi (1971–75) and by Alexander Wokaun.41 Experiments have been performed by Kurt A. Meier and Marco Genoni. Although the properties of higher correlation functions were well known to us at that time, we did not recognize that they contain multidimensional information that could be used and displayed in the form of 2D spectra. This became obvious to us however, in the course of the early calculations of Enrico Bartholdi in 1972 in the context of two-pulse 2D Fourier spectroscopy.42 THE DAWN OF TWO-DIMENSIONAL SPECTROSCOPY The participation of Thomas W. Bauman at the AMPERE Summer School in Basko Polje, Yugoslavia, in September 1971 was an extremely fortunate event. Being a meticulous scientist, he brought home a careful script of the lectures, among them one by Jean Jeener that attracted my attention immediately: a simple two-pulse experiment that produced revealing 2D spectra by 2D Fourier transformation of a 2D set of response signals (Figure 7 and 8). This was exactly the technique I had been waiting for. I had been thinking for some time about systematic computercontrolled double resonance experiments, but appreciated the complexity of the resulting 2D spectra should they follow the shape of the famous Anderson–Freeman plots.43 The Jeener two-pulse experiment appeared not to have this disadvantage. Enrico Bartholdi, who had just started his graduate studies, was willing to perform some analytical calculations on two-pulse experiments, taking relaxation into account. He confirmed the principal usefulness of the experiment and studied its features in detail. We did not plan to perform Figure 7 Professor Jean Jeener, the inventor of two–dimensional Fourier transform NMR Figure 8 Thomas Baumann (right) and the author trying to understand the notes of the AMPERE Summer School on the two-pulse experiment proposed by Jean Jeener in 1971. In the background is the modified DP- 60 Varian NMR spectrometer on which the first two-dimensional NMR experiments were performed experiments for two reasons. First we considered 2D spectroscopy as Jeener’s property and waited for his results to be published. Second, we thought that our computing equipment, with 16 000 words of memory, was inadequate for two-dimensional data storage and processing. We had some correspondence with Jean Jeener on the experiment in 1973, and he showed us a rough draft of an unfinished paper on this subject which, unfortunately, was never published. In 1974, I attended the 15th Experimental NMR Conference (ENC) and heard on 30 April 1974 an exciting lecture by Paul Lauterbur entitled Zeugmatography–Spatial Resolution of NMR Signals. He discussed the attainment of images by the projection–reconstruction method using continuous wave experiments. I immediately recognized that time domain experiments with switched magnetic field gradients in complete analogy to 2D spectroscopy would be the method of choice. This led to the concept of Fourier zeugmatography or Fourier imaging with a patent filed orginally on March 18, 1975, and issued on January 24, 1978. As I was acting eMagRes, Online © 2007 John Wiley & Sons, Ltd. This article is © 2007 John Wiley & Sons, Ltd. This article was previously published in the Encyclopedia of Magnetic Resonance in 2007 by John Wiley & Sons, Ltd. DOI: 10.1002/9780470034590.emrhp0051