Progress in Nuclear Magnetic Spectroscopy 28 (1995) 37-52

The development of experimental and analytical high resolution

NMR
James N. Shoolery’
Varian Associates. 611 Hansen Way, Palo Alto, CA 94303, USA

Received 25 February 1995

1. The beginnings about 5 gauss, easily observable in the magnets

In 1952, when I began working with high resolu-
tion nuclear magnetic resonance spectroscopy, a
suggestion that the technique might some day
determine the complete structure of a thirty-eight
carbon polycyclic alkaloid from a few milligrams of
sample would have seemed an impractical dream.
Today NMR is used regularly to solve such pro-
blems. In the five decades since its theoretical ori-
gins, nuclear magnetic resonance spectroscopy has
matured with unprecedented speed and revolutio-
nized the manner in which chemists conduct struc-
tural analysis. This paper reviews some of the
major milestones in that development.
1.2. From physics to chemistry
The first demonstrations of nuclear magnetic
resonance came about because physicists were
trying to measure the nuclear magnetic moment
more accurately. At Harvard, Purcell tackled the
measurement using paraffin, while Bloch at
Stanford chose to use water. Both materials
contain a high concentration of hydrogen, and
the proton dipole-dipole interactions in both
were predicted to broaden the NMR signals by
’ Retired.
then available. Purcell and Block shared the first
Nobel Prize associated with nuclear magnetic
resonance for their 1945 work. But their choices
of samples proved to be a first turning point. The
line widths in the water sample that Bloch used
proved to be much narrower than those in paraffin.
The overlooked phenomenon of isotropic mole-
cular tumbling in liquids had effectively averaged
out the static dipole-dipole effects. That opened
the way for Bloch to measure and correct for the
small but irritating influences of the chemical
surroundings and get down to real physics.
In 1951, two of Bloch’s students, Arnold and
Packard, constructed an improved instrument for
a world record attempt on another decimal place.
Their apparatus had a far more homogeneous elec-
tromagnet, a much smaller sample, and a very slow
sweep rate. A post-doctoral organic chemist at
Stanford, S. Dharmatti, provided the second
turning point. Curious about how an organic com-
pound would behave in the apparatus, he prepared
several samples, one of which was ethyl alcohol [ 11,
and the spectrum, shown in Fig. 1, revealed the
potential of NMR for determining the structures
of molecules.
The limitations of the early instruments and the
incomplete development of theory had focused
much of the earliest nuclear magnetic resonance

0079-6565/95/$29.00 0 1995 Elsevier Science B.V. All rights reserved

SSDZ 0079-6565(95)01019-X
38 J.N. ShoolerylProgress in Nuclear Magnetic Spectroscopy
Fig. 1. First high resolution NMR spectrum showing proton
chemical shifts in the ethyl alcohol molecule at 30 MHz.
work on studies of internuclear distances in solids,
a rather limited class of applications. Professor
H.S. Gutowsky at the University of Illinois,
Urbana, was among the first to appreciate the
possibilities for chemistry. Beginning in 1951, he
and his students used a permanent magnet high
resolution NMR apparatus to produce data on a
number of compounds. His pioneering work
brought high resolution NMR to the attention of
theoreticians and organic chemists in a most timely
way.
1.2. From academia to industry
As I described in an earlier paper, an unusual
confluence of events surrounds the beginnings of
high resolution NMR [2]. The inventor of the
klystron tube, Russell Varian (see Fig. 2), returned
to Stanford as a research associate after World
Warr II, with a dream of founding a company to
develop applications of new ideas in physics. When
he saw the results of the magnetic resonance experi-
ments, he immediately appreciated the potential
applications in chemistry. In 1946, he prepared a
patent application and persuaded Bloch and
Hansen to submit it. In return, he asked for and
received an exclusive license for his yet-to-be-
formed company to exploit the new technique.
Varian Associates was founded in 1948, and the
business philosophy of the fledgling company
proved well-suited to the rapid development of
the new analytical technique. Profits from the
klystron tube made it possible to invest in the initial
NMR development. As further profits were made,
28 (1995) 37-52
Fig. 2. Russell Varian, inventor of the klystron tube and one of
the founders of Varian Associates in 1948.
a large portion was reinvested in further research
and development. That positive feedback loop
powered an exponential growth in instrumental
capabilities and in the acceptance of NMR by
chemists and biochemists.
The development of nuclear magnetic resonance
spectrometers, in which Varian Associates played a
primary, almost solitary, role for two decades, pro-
vides a notable example of successful transfer of
academic insights to industry. As the technology
matured, other companies joined Varian in a
search for improvements. The dynamic industry
that resulted has served the scientific community
well.
One of Varian’s first products was an electro-
magnet of the type built by Arnold and Packard.
In early 1952, seeing a specification sheet for it that
included the ethyl alcohol spectrum, I wrote to
Varian Associates, suggesting that they needed an
applications laboratory, which I would be glad to
run. Fortunately for me, they agreed. By the time I
arrived a few months later, Varian Associates had
hired two of Bloch’s students, Martin Packard and
 J.N. Shoolery/ProRress in Nuclear Magnetic Specrroscopv 28 (1995) 37-52
Emery Rogers, and sold its first spectrometer. The
creative freedom and encouragement of individual
initiative promoted by the Varian brothers, Russell
and Sigurd, made working at Varian uniquely
exciting. Over the next few years, more of Bloch’s
students joined the group: Weston Anderson,
James Arnold, Warren Proctor, and Harry Weaver
all became part of the effort.
Varian’s NMR applications laboratory came
into existence even before the first NMR spectro-
meter was delivered. To my enthusiastic eye, the
prototype Varian spectrometer turned the nucleus
of the hydrogen atom into the equivalent of a cen-
trally located spy, willing to disclose its interactions
with the electrons in nearby chemical bonds. The
potential for extracting information about the
chemical structure was apparent. One had only to
correlate the observed spectral parameters with
known molecular properties and develop the
theory.
The prototype I worked on, the HR-30 based on
a Stanford-type electromagnet and operating at
30 MHz (7050 gauss), delivered state-of-the-art
performance, but, clearly, there was much room
for improvement (see Fig. 3). The first problem to
tackle was the magnet; it had to be stronger, more
homogeneous, and much more stable.
1.3. The heart of the instrument
As early work by Gutowsky and Hoffmann [3] at
Illinois and Arnold [4] at Stanford had illustrated,
permanent magnets bypass the difficulty of field
stability encountered with electromagnets. How-
ever, they sacrifice some chemical shift dispersion
and sensitivity, both of which increase with field
strength. I believed that proton chemical shifts
are just too small to resolve much structural
information at frequencies below 60 MHz. Accept-
ing this view, Varian bet on the electromagnet.
A Varian research program carried out by
Forrest Nelson showed that the important factors
in electromagnet design were concentrated in the
pole caps, adjacent to the air gap. Proper metal-
lurgical treatment was required to maintain a
fine-grained magnetic domain structure in the
special steel alloys found to perform best. Surfaces
were finished to a flatness of one wavelength of
39
light, and the caps were machined to a complex
mathematical shape to compensate for the varying
saturation of the iron. Methods for adjusting the
parallelism of the caps were also refined.
Little by little, those improvements were incor-
porated in the products. In 1955, the HR-40 was
announced, followed in 1958 by the HR-60, which
became a standard general purpose high resolution
instrument. In 1959, the final step in electromagnet
improvements was taken with the release of the
HR-100, capable of achieving a field uniformity
of 1 part in 10’ over the volume of a 5mm sample
tube at 100 MHz (23 490 gauss).
1.4. The spinning sample
Other improvements and innovations were
developed concurrently. The isotropic molecular
tumbling, or motional averaging, found in liquid
samples had been recognized in 1945, but reso-
lution was still poor because of the unresolved
problem of how to compensate for the X-, y-, and
z-gradients over a sample placed in the magnetic
field. Stirring a cup of tea one day in 1954, Felix
Bloch came up with the idea of rapidly stirring the
sample [S]. When that proved too difficult, he
settled on using a small air turbine to rotate the
sample tube [6]. The result was a dramatic improve-
ment in fine structure resolution. The gradients
transverse to the spinning axis were averaged to
zero, leaving only the gradient parallel to the axis
needing adjustment. The work was quickly pub-
lished and patented. It proved extremely simple
to add this capability to the HR-40 in production.
That invention, obvious in hindsight, resulted in
high quality data being generated years before the
equally effective compensatory technology appeared.
Once spin-spin coupling patterns could be resolved
routinely and reproducibly, NMR was rapidly
accepted. The fact that forty years later, the most
powerful and sophisticated NMR spectrometers
still use this simple device ranks its discovery as
one of the significant developments in experimental
technique.
1.5. Current shims
Early attempts to achieve optimum resolution
40 J.N. ShoolerylProgress in Nuclear Magnetic Spectroscopy 28 (1995) 37-52
Fig. 3. The author with Dr. Virginia Royden, operating
involved a random search for the best spot in the mag-
netic field, which was not necessarily in the center.
Golay [7] developed a current shim based on a
spherical harmonic representation of the field that
was introduced by Perkin-Elmer, Ltd. on the R-
12, a 60MHz permanent magnet spectrometer.
Once the magnet of the R-12 was energized and
stabilized thermally (which required temperature
control to a millidegree), only minor adjustments
of the shim currents were required from one sample
to the next. That feature was greatly appreciated by
users with routine applications.
But the problems of shimming an electromagnet
were less tractable. Arnold fastened a concentric
array of coils on the pole faces of the magnet at
the prototype HR-30 in the Varian NMR Applications Laboratory in 1953.
Stanford to correct the field of the main magnet.
When a set of those coils was added to my HR-30
in the Application Laboratory, I discovered that
the coils were highly interactive: a change in one
current required readjusting all the others. The
mechanical search had been converted into an
equally frustrating electrical one.
After several years of work, Wes Anderson
developed a set of non-interactive current shims
based on orthogonal gradients [8]. With the
development of field plotting apparatus, the shim
coils could be attached to the pole faces symmetri-
cally about the center and, with spinning, the
adjustment of the current for optimal resolution
was reduced to a single axis.
 J.N. ShoolerylProgress in Nuclear Magnetic Spectroscopy
Hysteresis in the iron pole caps caused problems
that were overcome by a procedure we developed
for energizing the magnet. By running it up to a
higher field strength and then backing it down to
the operating value, we were able to obtain a
reproducible field contour. This made the electro-
magnet-based instruments easier to adjust and less
baffling to the user.
Current shims optimized the quality of data and
allowed standards of performance to be set.
Schools of thought on shimming procedures
sprang up, and, to this day, some people regard
magnet shimming as more of an art than a science.
In the 1990s that view is uninformed. Non-spinning
line width and line shape specifications now rival
those attainable with spinning in suitably equipped
modern instruments. That standard was not
reached through an inspirational stroke of genius
but through the patient work of physicists and
engineers over a period of forty years.
1.6. Flux stabilization
NMR is built on a tripod of magnetic field
strength, resolution, and stability. The field
strength had been doubled, the spinning sample
and gradient correction coils had allowed unprece-
dented narrowing of the spectral lines, but the
spectra still looked terrible. Short-term fluctuations
in either field strength or frequency produced tran-
sient distortions that interfered with interpreting
the data. In addition, long-term drift precluded
the use of a linear, precalibrated scale for chemical
shifts. Without a convenient and reproducible pre-
sentation of data, NMR failed to win the hearts of
many chemists. Electronic regulation could only
stabilize the current to about one part in lo’,
three orders of magnitude less than was needed.
The choice of the electromagnet was beginning to
look questionable.
Then, in 1956, our colleagues at the Shell Devel-
opment Laboratories [9] in Houston, Texas, told us
of a physics experiment at Princeton [lo] that used
a degenerative feedback loop to stabilize a magnet.
The idea provided inspiration. Pick-up coils on the
magnet poles sensed any change in the magnetic
field as an induced current, which deflected a
galvanometer. A light beam reflected from a mirror
28 (1995) 37-52 41
on the suspension struck a split photocell, giving
an enormously amplified response to either an
increase or decrease in the field strength. An opera-
tional amplifier converted this response to a current
in a set of correction coils and stability was ours.
The wide frequency response from d.c. to hundreds
of hertz assured that both jitter and drift were
virtually eliminated.
The completed unit was called a Super-Stabilizer
and could be retrofitted to the earlier HR-30s and
4Os, bringing them up to the current standard. The
effect on chemist-users was almost as dramatic as
the improvement in the spectra. Like the double-
beam principle in infrared spectroscopy, flux-
stabilization brought reproducibility to high
resolution NMR. Chemists now could trust their
results.
1.7. Spin decoupling
Scalar coupling between nuclear spins is trans-
mitted through intervening bonds. The resulting
splittings can be interpreted to confirm the exist-
ence of those bonding pathways or to determine the
number and type of interacting nuclei. The inter-
pretation becomes complicated when too many
nuclei are involved, and in 1954, experiments at
Varian [l l] and Stanford [12] showed that the
spin-spin coupling can be averaged to zero by
the application of a strong r.f. field at the resonance
frequency of one of the spins. That technique was
called “spin decoupling,” and it was very useful for
simplifying complex spectra.
A more general theory was developed 1131that
showed that spin decoupling is a special case of
applying a perturbing r.f. field to one nucleus
and observing changes in the frequencies and
populations of the energy levels of nuclei coupled
to it. Recognizing that ability to transfer inform-
ation between spins was a turning point that
has led to important experimental improvements
and to .families of powerful and informative
experiments.
1.8. The A-60: NA4R enters the mainstream
Each high resolution instrument produced was
42 J.N. ShoolerylProgress in Nuclear Magnetic Spectroscopy 28 (1995) 37-52
better than the preceding one, and company policy
was that older instruments could be modernized by
adding a new module. The eventual result was a
spectrometer made up of racks of electronic units
that looked better suited to a broadcasting station
than a chemistry laboratory. By 1957, Varian had
the required technology to tackle building an NMR
spectrometer for general laboratory use by organic
chemists, and it set in motion a four-year-long
program to do so.
The spectrometer had to be a reliable, highly
sensitive, proton-only instrument that would gen-
erate reproducible spectra on a linear precalibrated
chart. In addition, it would have to be easy to use,
conveniently sized, and affordable. When the
trade-off between cost and performance led to
using a 1200 pound 60MHz electromagnet, the
spectrometer acquired a name, the A(for analyti-
Cal)-60.
The sample probe and r.f. circuitry were modi-
fied to minimize noise and optimize the signals
from the nuclei. A compact console with conveni-
ently grouped controls provided the means for the
operator to adjust, calibrate, acquire, and plot
data. In addition, Wes Anderson invented an
ingenious device that locked the frequency to the
precession of protons in a control sample of water
[ 14,151. If the field strength changed, the frequency
shifted to compensate for the change. This per-
mitted the introduction of precalibrated charts,
and NMR was ready to join the ranks of other
established spectroscopic techniques. Nearly all
high resolution NMR spectrometers now incorpor-
ate some form of field-frequency locking.
During the six years that the A-60 was in
production, NMR became firmly established as
the primary source of structural and analytical
information for organic chemists. Then, the A-60
with its electromagnet and vacuum tubes was
replaced by permanent magnet instruments with
solid-state electronics. The T-60, introduced in
1967 and followed by the EM-360 in 1973, was
joined during this period by similar instruments
of Japanese manufacture. Those descendants of
the A-60 continue today to serve the analytical
needs of organic chemists in industrial, govern-
ment, and academic laboratories throughout the
world.
2. The 1960s
2.1. Technological synergy
Most of the work published in organic chemistry
during the 1960s was carried out with commercially
available NMR instruments operating at 60 or
100 MHz. Knowledge about spin-spin couplings,
chemical shift effects, ring current effects, solvent
shifts, second-order spin-coupling effects, depend-
ence of coupling on stereochemistry, and the effects
of chemical exchange was applied to a wide variety
of chemical systems. With the explosive growth of
applications, even the skeptics became believers.
The HR- 100 demonstrated that results obtained
at 23 490 gauss were vastly superior for studies for
more complex organic compounds. Pharmaceuti-
cal research laboratories and even biochemists
were showing interest in higher field NMR. At
Varian, the climate was favorable for further
investment in innovative ideas.
2.2. Superconducting magnets
NMR differs from other spectroscopic tech-
niques in that the energy levels involved are not a
property of the spin system but depend upon
imposing an external field on the nuceli. That
makes it possible to improve the sensitivity and
chemical shift dispersion indefinitely. The only
limits are those imposed by the technology at hand.
Trying to generate more than 20000 gauss with
an iron electromagnet leads to rapidly diminishing
returns. The weight, energy consumption, and
cooling requirements become prohibitive for a
laboratory instrument. Further progress in field
strength required a new approach, and the
commercial availability of superconducting
niobium-zirconium wire offered a possible oppor-
tunity. The new material, unlike earlier supercon-
ductors, could remain superconducting in a strong
magnetic field. Test samples of the wire suggested
that it might be possible to reach fields as high as
71 kgauss (300 MHz for protons).
Harry Weaver was given the task of designing a
high resolution solenoid. The prototype unit [16]
was compensated with extra turns at the ends to
maintain a uniform field along the axis. In 1962 it
 J.N. ShoolerylProgress in Nuclear Magnetic Spectroscopy 28 (199.5) 37-52 43

was ready to be tested. After immersion in liquid

helium, the magnet was brought to operating field
strength by slowly increasing the energizing current
through a resistive load. At that point, a supercon-
ducting return path was established by turning off a
heater surrounding a length of superconducting
wire across the solenoid terminals, and the power
supply was removed, leaving the magnet in the
persistent state. The flow of electrons without
resistance came as close to perpetual motion as
anything we could imagine.
The prototype magnet [ 161reached 46 980 gauss
(200MHz for protons) and appeared capable of
maintaining that field indefinitely without signifi-
cant decay, given a continued supply of liquid
helium. If protected from local magnetic noise, it
appeared to be completely stable. The magnet
served as its own flux stabilizer! Homogeneity
was optimized with redesigned current shims
wound on cylindrical forms, and reached 0.2 Hz
line widths in test samples. Sensitivity was 2.7
times better than at 100MHz. With superior
stability, resolution, sensitivity, and chemical shift
dispersion, the superconducting magnet was a
Fig. 4. The Varian HR-220 Superconducting NMR Spectro-
winner on all counts. The days of the electromagnet meter.
were numbered.
A console of the HR-type was adapted to the developed, but suffered from the same drawbacks.
higher frequency. Cylindrical probes were designed The full benefit of superconducting magnets would
with the sample spinning about the longitudinal not be realized until the next decade, but the break-
axis of the solenoid, which was vertical. Since the through had been made.
field was parallel to the spinning axis, the coil,
which must be perpendicular to the field, had to 2.3. Computer-assisted NMR
be redesigned in a Helmholz pair configuration to
allow the sample tube to pass through it. The In 1964, Leland Allen, then at Yale University,
resulting 200 MHz prototype instrument was proposed a method of improving the sensitivity of
exhibited at the Pittsburgh Conference in 1962, the NMR spectrometer, based on a computer of
serving notice that NMR was ready to take on average transients (CAT). The CAT was being
tougher chemical problems. used at that time to detect weak electrical transients
In 1964, 51680 gauss (220 MHz) was reached. in biological systems, After the application of an
The first installation of an HR-220 (Fig. 4) was initiating stimulus, the transient response voltage
made in 1965 at the DuPont Research Laboratories was sampled at a high repetition rate, and the con-
in Wilmington, Delaware. The instrument per- secutive digitized values were stored in up to 400
formed well and confirmed the value of operating memory channels. The process was then repeated n
at a high field strength. However, the high con- times. The device was hard-wired to add each new
sumption of liquid helium, along with its high set of digital values to the sum already accumu-
cost and uncertain availability, offset the technical lated. The signal increased by the factor n, while
advantage of the HR-220, and the market was the random noise increased only as the square root
therefore small. In 1968 the HR-300 was ofn.
44 J.N. ShoolerylProgress in Nuclear Magnetic Spectroscopy 28 (1995) 37-52
Allen proposed triggering the CAT at the begin-
ning of the spectral scan in an NMR spectrometer,
using a reference peak in the spectrum. The experi-
ment worked, giving the expected improvement in
weak signals. Varian commissioned the Nicolet
Company to develop a 1024 channel accessory
for NMR, the C-1024. Hundreds of them were
purchased by chemists struggling to overcome the
problems of analyzing small or dilute samples.
Installed in the CW spectrometers, the computer
improved the sensitivity by about a factor of 10,
a very useful gain. The significance of the C-1024
was, however, far greater than this modest
improvement suggests. It ushered in a profound
change in the direction of development of NMR
instrumentation by demonstrating the usefulness
of digitizing the NMR signals and following with
digital processing.
The C- 1024 was soon replaced by programmable
computers, which could acquire, store, process,
and display the NMR data, and also control the
instrument with a speed and precision far beyond
the capability of a human operator. Freed from
those limitations, the development of NMR as a
structural and analytical tool soon entered an
exciting new period.
2.4. Pulsed Fourier transform
The third and probably most important break-
through of the 1960s was the realization of an idea
that had its roots in the early 1950s. In 1965,
Richard Ernst performed an experiment at Varian
Associates under the direction of Wes Anderson.
He applied a short, intense pulse of r.f., exciting the
nuclei over the entire spectral range. This wide-
band excitation was followed by the appearance
in the receiver coil of a complex waveform in the
time domain (the free induction decay-FID) that
contained all of the spectral information. Fourier
transformation was carried out on the digitized
FID by a digital computer and yielded the familiar
frequency spectrum. Sensitivity could be enhanced
by between 30- and lOO-fold.
The seeds for this discovery were planted by
Felix Bloch, who pointed out in his first detailed
paper about NMR in 1946, that a strong r.f. pulse
could reorient the nuclear moments, which would
then precess in the absence of a perturbing field.
Only Hahn seems to have taken that comment
seriously, and by 1950, he had adapted the use of
pulses to the measurement of relaxation times,
using his ingenious spin-echo pulse sequence [ 171.
By 1952, Hahn and Maxwell [18] had reported the
phenomenon of J-modulation by which the spin-
spin coupling frequencies are encoded into the echo
decay.
Russell Varian filed a patent application in 1948
for a magnetic field measuring device that
depended upon free precession [ 191. In 1956, he
filed another patent that described a system in
which a wide band of excitation is achieved by
pulses or white noise, and the detected signals are
separated into the component spectral frequencies
by complex optical or electronic means [20]. By the
time of his death in 1959, he had not succeeded in
persuading anyone to do the experiments. In retro-
spect, it is clear that this visionary man had a clear
understanding of the basic concepts of Fourier
Transform (FT) NMR and of its importance. For-
tunately, his ideas had made a lasting impression
on Wes Anderson.
In 1962, Wes began construction of a clever
device employing a rotating drum to generate a
band of frequencies and to decode the signals,
but by 1965, he had realized that a train of r.f.
pulses would be a better way to achieve the wide-
band excitation that Russell Varian had advocated.
The signals could be stored using a CAT and then
Fourier-transformed using a digital computer.
Richard Ernst assembled the necessary equipment
and achieved the first demonstration [21,22]. The
idea worked: his first FT spectrum showed more
than a lo-fold gain in sensitivity compared to a
CW experiment using the same data acquisition
time.
But in 1965, the new method was actually much
slower because digital computers were in their
infancy. The contents of the memory channels of
the CAT had to be transferred to punched paper
tape or cards, transported to a computing center,
and when time was available, the data could be
printed. The main frame computer in Varian’s
Accounting and Management Information
System’s group provided overnight service for
Ernst when computer time was available.
 J. N. ShoolerJ,/Progress it1Nuclear Magnetic Spectroscopy 28 (1995) 37-52
A truly useful FT-NMR instrument had to await
the development of minicomputers which could be
dedicated to the spectrometer. In 1968, Varian
offered an FT accessory to the HA-100. This
instrument, with a field-frequency lock, had
enough stability to demonstrate the enormous
gain in sensitivity that can be achieved. Its product-
ivity was limited, however, by a minicomputer with
only 4000 bytes of memory. That meant that the
transforms took minutes rather than seconds, and
the spectral plots were primitive.
In 1969, Bruker GmbH, a German company
which had recently entered the NMR business,
introduced an instrument that set the direction
for the 1970s. Combining the three great advances
of the 1960s in a single instrument, the WH-270
offered chemists more problem-solving capability
than ever before. Combining improvements in
dewar design with a state-of-the-art minicomputer,
this FT-NMR instrument offered practical solu-
tions to problems that had been holding up pro-
gress. Its immediate success alerted chemists all
over the world to the power of NMR to solve struc-
tural problems quickly and elegantly and launched
NMR into the frenzied development and amazing
accomplishments of the next twenty-five years.
3. The 1970s
3.1. Evolution of’ modern NMR experimental
capabilities
As the 1970s began, Bruker demonstrated that
FT-NMR spectrometers could also bring chemists
the benefits of observing carbon-13. Limited in
sensitivity by its low magnetic moment and 1%
abundance, the C-13 nucleus is also frequently
coupled to protons through one to three bonds.
The resulting multiplet patterns are split into weak,
overlapping peaks, further reducing sensitivity.
Richard Ernst had developed a technique for
broad-band decoupling of the protons from the
carbons, using noise modulation of the proton
irradiation frequency [23], which collapsed all of
the C-13 peak intensity into a single sharp line
and also led serendipitously to a nuclear Over-
hauser enhancement (NOE) of the C-13 signals of
45
up to a factor of three. Combined with pulsed FT
operation, larger samples, and longer acquisition
times, the technique was shown to make the acqui-
sition of C-13 data practical for the chemist.
Once again, Varian embarked on a five-year-
long project to design a workhorse instrument,
this time to be able to produce high quality
carbon-l 3 spectra in a laboratory environment
and in an affordable and straightforward way.
Using a small electromagnet operating at
18 792 gauss, the CFT-20 met those objectives in
1972 and established the potential of FT-NMR to
study other nuclei besides the proton.
Once chemists recognized the value of having
information from both the protons and the carbons
of a molecule, ingenious experiments to take
advantage of the C-H couplings proliferated.
Nearly twenty-five years later, some organic
chemists still remain to be converted, but the
trend is unmistakably toward this growing family
of powerful multichannel experiments.
3.2. The sqftware revolution: NMR’s second Nobel
Prize
In 1971, Jean Jeener, a Belgian, proposed an
ingenious experiment for selectively extracting
information from the nuclei. In ordinary one-
dimensional NMR, chemical shifts and spin-spin
couplings are spread out along a single line, often
badly overlapped. Jeener suggested applying two
separated pulses in a series of experimental incre-
ments, increasing the pulse spacing for each new
increment. After Fourier transformation, the digi-
tized spectra would be arranged as the rows of a
matrix, and a second Fourier transformation
would be performed on the columns. The two
types of information in the resulting spectra
would determine the peak positions in a two-
dimensional array.
In 1971, spectrometers lacked enough data
storage and computing power to perform that
experiment. By 1974, however, improved compu-
ters allowed Richard Ernst to put the proposal into
practice [24]. His success started a new direction of
development in experimental and analytical NMR,
which could be called the software revolution.
Jeener had hit upon a most important fact. If the
46 J.N. ShooleryjProgress in Nuclear Mugncric Spectroscopy 28 (1995) 37-52
data are not acquired immediately after a pulse
excites the nuclei, the interactions within the system
can evolve and be reconstructed later, allowing
pulse sequences to be tailored to select various
types of spectral information.
Of special value are the correlation experiments
that display peaks at the chemical shift coordinates
of coupled nuclei, thus elucidating bonding path-
ways within the molecular structures. Those
include COSY, HETCOR, TOCSY, and variations
of them. Chemists and biochemists can also estab-
lish molecular conformations using the map of
interproton distances provided by the NOESY
and ROESY experiments. Richard Ernst was
awarded the Nobel Prize in Chemistry in 1991 for
his role in developing FT-NMR and 2D-NMR and
for systematizing entire families of 2-D experi-
ments and demonstrating their usefulness.
Pulse sequence development flourished and
attracted other outstanding innovators, such as
Freeman, Bax, and Mullet-, to single out a few.
Techniques to suppress unwanted peaks and
experimental artifacts arising from imperfections
in the hardware were developed and refined.
Chemical problems could be solved more elegantly
and with greater confidence. Most chemists, how-
ever, had to wait for commercial spectrometer
development to catch up with the innovators.
3.3. Onward and upward
The development of superconducting magnets
was continued throughout the 1970s. Varian intro-
duced the SC-300 in 1971 and the SC-360 in 1973,
operating at 300 and 360MHz. Unfortunately,
neither instrument was engineered for factory pro-
duction (the market having been estimated at only
3 or 4 instruments per year). Even with the entire
engineering team working on production, delivery
dates soon stretched out to a year. Reluctantly, the
products were withdrawn, and work began on
designing a completely new research superconduct-
ing NMR spectrometer, incorporating every
technical advance in electronics and computing.
That effort came to fruition in 1978 with the intro-
duction of the XL-200, the first NMR spectrometer
with disk-based software and data storage. The use
of two computers, programmed in concurrent
PASCAL and connected by a high speed bus, to
acquire and process data obviated the problems of
time-sharing and increased throughput. The basic
architecture of the XL-200 has continued to the
present time as a model for today’s much more
powerful instruments.
During the same period in Europe, Oxford
Instruments and Bruker both pursued higher field
strength. The WH-270 was joined by the WH-300,
and in 1974, by the WH-360. The last spectrometer
gained wide acceptance as complex molecules
began to reveal more structural information with
the increased chemical shift dispersion. In rapid
succession, the WH-400 and WH-470 appeared in
1974 and 1976, followed by the WH-500 in 1979. In
reaching those milestones, techniques for handling
the niobium-tin superconducting materials were
developed and refined, and the solenoid designs
modified.
Most people thought that 500 MHz (11.7T,
where T is the Tesla, equal to lo4 gauss) would be
the maximum field for the niobium-tin compound.
Consequently, the 500MHz instrument soon
became the standard for biological NMR.
Although they were quite expensive for organic
chemists, several of the leaders in that field
managed to acquire them as the decade drew to a
close.
4. The 1980s
4.1. Computers come ojage
The explosion in computing speed, memory, and
cost reduction began slowly in the 1970s but
rapidly gained momentum. By the middle of the
1980s the stage was set for widespread acceptance
of 2D-NMR as a practical experimental technique.
With adequate speed and memory, and with soft-
ware to set up the experiment, acquire data, and
offer convenient options for processing and plot-
ting, organic chemists began to rely on the new
technique. Because the highly selective 2-D experi-
ments provided more information, the structures of
more complex molecules could be worked out with
confidence, their spectral data completely assigned
and understood.
 J.N. ShoolerylProgress in Nuclear Magnetic Spectroscopy 28 (1995) 37-52
The large chemical shift range in high fields
requires very efficient broad-band excitation for
spin decoupling. In the early 1980s a new method
for decoupling and spin-locking (MLEV) was
developed [25] based on earlier concepts for
generating tailored distributions of spectral excita-
tion [26] and for removing the dipolar broadening
in solids [27]. The new technique used trains of
precise pulses with repeating sequences of pulse
widths and phases. It reduced the artifacts being
generated in 2-D experiments and stimulated
further development of even more complex pulse
sequences.
The design of new 2-D experiments benefited
greatly from the realization that the use of the pro-
duct operator formalism of quantum mechanics
made it possible to predict the precise behavior of
the spin system. That made it easier to create a
desired coherence of precessing spins and to trans-
fer that coherence from one set of spins to another.
Pulse sequences could be designed to operate more
efficiently and generate fewer artifacts.
4.2. The injuence of biochemistry
For decades, biochemists had sought to deter-
mine the molecular conformations of proteins in
solution. With the increased potential provided
by NMR, funding for high-field instruments
and for the development of experimental NMR
methods in biochemistry increased rapidly and an
all-out assault on the problem began. Only a few
leaders in this field were able to marshal the
resources needed to keep improving spectrometer
hardware, devising new pulse sequences, and
upgrading computer hardware and software to
keep pace with new developments.
The two-dimensional displays of data from pro-
teins often contain too many overlapping peaks to
be interpreted readily. It was natural to extend the
principles of the 2-D experiments to 3-D and higher
dimensional experiments. Adding another evolu-
tion period in the pulse sequence provides an
additional axis along which peaks can be located.
The drawback is that the acquisition of data is
multiplied hundreds of times, with corresponding
demands on data storage and processing time.
Only after the mid-1980s were the data systems of
41
spectrometers powerful enough to make such
experiments practical.
In 1977, working with Aksel Bothner-By at
Carnegie Mellon Institute in Pittsburgh, Josef
Dadok [28] constructed the first 600 MHz spectro-
meter, using a magnet wound by Intermagnetics
General Corp., Guilderland, NY, with a strong,
flexible tape coated with a niobium-tin film. Diffi-
culties in making resistance-free joints made it
necessary to operate the magnet in a non-persistent
mode, with a continuous supply of power. That
resulted in high helium consumption and a need
for stabilization, both problems that could be
tolerated only in a developmental instrument.
However, the data generated at 600MHz created
a great demand among biochemists, and in 1987,
Oxford Instruments completed the development of
a 600 MHz magnet operating in a persistent mode.
The benefits of increased field strength were not
lost on chemists, and by the end of the decade,
600 MHz instruments were beginning to appear in
pharmaceutical research laboratories and in lead-
ing academic organic chemistry departments.
4.3. Indirect detection
In 1979, Mtiller [29] had developed a novel 2-D
pulse sequence for correlating the chemical shifts of
two coupled spins: I, with a strong magnetic
moment, and S, with a weak magnetic moment.
The pulse sequence made it possible to obtain the
data with the full sensitivity of the I spins. Bio-
chemists soon adapted this experiment for indirect
detection of the weak N-l 5 nuclei in proteins. After
a proton 90” pulse, 15NH spin coupling evolves,
followed by a “N pulse to create double quantum
coherence for the amide nitrogens and the attached
protons. When converted back to single quantum
coherence by a second 15N pulse, the proton peaks
can be detected with the sensitivity characteristic of
protons, but with the peak intensities modulated at
the N-l 5 chemical shift frequency. Mtiller’s experi-
ment was later modified to detect C-13 chemical
shifts through detection of the protons bonded
directly to the carbons [30]. The resulting Hetero-
nuclear Multiple Quantum Correlation (HMQC)
2-D plot gives the same data as the carbon-detected
proton-carbon chemical shift correlation 2-D plot
48 J.N. ShooferylProgress in Nuclear Magnetic
(HETCOR) but with greater sensitivity. Corre-
lations through long-range couplings can also be
mapped by using the closely related Heteronuclear
Multiple Bond Correlation (HMBC) pulse sequence
[3 11.Those experiments make it possible to establish
all of the observable bonding pathways between pro-
tons and C-13 nuclei through two or three bonds.
4.4. Total structure elucidation
An experiment developed in 1981 [32], whim-
sically given the acronym 2-D INADEQUATE,
was able, in principle, to define all of the carbon-
carbon bonds, thus establishing the complete
carbon skeleton in a single experiment. That
seemed like the ultimate triumph for NMR but,
because it required two C-13 nuceli in natural
abundance bonded to one another, the sensitivity
was a hundredfold less than other carbon-detected
2-D experiments. From a practical point of view,
the data were tantalizingly just out of reach for
most samples, and chemists had reluctantly written
the experiment off.
At the end of the decade, a general approach to
total structure elucidation had been applied to at
least one unknown natural product [33]. The basis
of this method was to establish a known sub-
structure, e.g., an aromatic ring, and to obtain
HMQC and HMBC 2-D correlation data defining
all of the observable proton-carbon bonding path-
ways. As the assignment of the 2-D peaks pro-
ceeded, one could add additional carbons and
heteroatoms to the initial fragmentary structure,
weeding out inconsistencies in the assignments by
using all of the observed pathways through each
bond. Ultimately, a unique total structure could
be built up, one carbon atom at a time. Finally, a
measurement of interproton distances, using NOE
difference or NOESY data allowed the molecular
conformation to be specified.
While not foolproof, the method has general
applicability and can be supplemented by additional
data, such as COSY, TOCSY, and 2D-INADE-
QUATE, if needed. Although data acquisition is
lengthy, modern spectrometers can be set up to per-
form all of the experiments unattended. The whole
process illustrates the synergistic effect of the
improvements that have occurred over forty years.
Spectroscopy 28 (1995j 37-52
5. The 1990s
5.1. It’s not over yet
Fig. 5 shows a plot of the proton resonance
frequency vs. time from 1952 to 1994. The rate
of increase has been surprisingly constant, averag-
ing about 10 MHz year-’ for iron magnets and
18 MHz year-’ for superconducting magnets. The
large increases from 500 to 600 MHz, and from 600
to 750MHz required an average of seven years
each. One is tempted to extrapolate and predict a
900MHz frequency in the year 2000, which may
not be a bad estimate. Several projects are under-
way at this writing with 900 MHz as the objective.
The difficulty of the problems that must be
solved increases more rapidly than the frequency;
consequently, the development cost per megahertz
keeps rising. The next development may be accom-
plished with company-sponsored funding if one or
more companies can maintain a risk-taking level of
R&D, but almost all research in new superconduct-
ing materials is government-funded. We may well
be approaching the time when further increase in
frequency will depend upon the political climate as
much as the technological climate.
5.2. Pulsed field gradients
The 1990s have continued the accelerating
growth of improvements in electronics, computers,
and software development. As a result, many new
capabilities and conveniences have appeared in
NMR spectrometers. One promising development
is the use of pulsed field gradients as elements of
pulse sequences [34]. A field gradient applied for a
time, t, will result in a spatial distribution of the
phases of any coherently precessing nuclear
moments. An identical pulse of opposite sign
applied at a later time will unwind the phase dis-
tribution, and the signal can be detected as usual.
However, if some molecules diffuse during the time
between the pulses, they will not be refocussed and
their signal will be suppressed. The small size of the
water molecules leads to rapid diffusion compared
to proteins, so the technique is very effective in
suppressing the water peak.
The coherences of different spins, such as
 J.N. ShoolerylProgress in Nuclear Magnetic Spectroscopy 28 (1995) 37-52
Fig. 5. The proton resonance frequency (MHz)
beginning in 1952.
protons and carbon-13, behave differently under
the influence of the gradients. Pulses applied to
the coherently precessing spins can reorient them
and also determine the effect of a subsequent
gradient pulse. By proper combination of r.f. and
gradient pulses, coherence pathways can be selected
and unwanted coherences suppressed. This can be
accomplished during each repetition of the sequence,
eliminating the need for phase cycling. Very substan-
tial savings of time can result, making 2-D experi-
ments comparable to one-dimensional experiments
in the time required to solve chemical problems.
5.3. Small is beautiful: micro- and nano-NMR
probes
The story of NMR is the story of the quest for
greater sensitivity. In 1977, Varian Associates
introduced a microprobe for the Varian FT-80
49
YEAR
at the highest available magnetic field plotted as a function of time for fifty years
that optimized the coil geometry for small samples.
A capillary (1.7 mm) sample tube in a 2 mm coil
reduced the sample volume from 400 to 15 ~1, con-
centrating small samples by a factor of more than
25. The coil was wound from fine copper wire to
avoid distorting the magnetic field appreciably. A
proton spectrum of 1 pg of cortisone acetate was
obtained, and a carbon-13 spectrum from 1 mg of
sucrose was produced in an overnight data
acquisition.
When superconducting spectrometers replaced
the FT-80s microprobes were not among the
available accessories, partly because with the
higher sensitivity most chemists did not feel a
need for a microprobe. But more importantly, the
spectral quality from such probes was poor at high
fields, because the magnetic susceptibility of the
coil material could not be controlled.
Fifteen years passed. Probe technology
50 J.N. Shoolery/Progress in Nuclear Magnetic Spectroscopy 28 (1995) 37-52

improved through research into better materials, for visual analysis, The spectrum on the right
and the demand rose because chemists wanted to shows the patterns recognized by computer analy-
perform the time-consuming 2-D experiments on sis and displayed in the same format. The entire
smaller samples. The time had come for another carbon skeleton is determined in a single experi-
attempt to confine the sample. Microprobes similar ment, using several milligrams of sample. If 5-
in principle to the first model were offered by two 1Omg of a natural product can be dissolved in
companies in 1992. Modest gains in sensitivity were 40~1 of solvent, the 2-D INADEQUATE experi-
reported. ment becomes 2-D ADEQUATE!
I believed that maximum efficiency would
require confining the sample completely within 5.4. New .horizons
the receiver coil, resulting in severe distortion of
the field over the sample, due to the discontinuity The only safe prediction about the future is that
of magnetic susceptibility at the cell boundaries. To it will be different from that predicted. However,
get around the problem, I proposed tilting the cell three aspects of current research that may offer
and rotating it at high speed around an axis ongoing improvements in experimental NMR can
inclined to the strong field at the magic angle, be identified.
54.7”. Theory predicts that all effects arising from
magnetic dipoles will then average to zero.
The Varian Nano .nmr probe, announced in
1993, is based on that approach. Proton spectra
from as little as 75 ng of a polycyclic aromatic com-
pound have been obtained, and a carbon-13 spec-
trum from 100 pg of sucrose has been acquired in
12 h. The proton sensitivity will prove valuable in
drug metabolite studies, and in elucidation of the
structures of minute amounts of natural products,
such as insect pheromones.
A more important application may arise from a
property of the magic angle spinning; namely, the
removal of resolution degradation in hetero-
geneous samples such as cell suspensions or mole-
cules attached to solid polystyrene bead supports in
multi-step organic syntheses. Fig. 6 shows the
comparison of the proton spectra from a molecular
species covalently bonded to polystyrene beads
taken with a conventional probe and with a
magic angle spinning nanoprobe [35].
Finally, the carbon sensitivity combined with a
computational method of recognizing predictable
patterns of peaks embedded in random noise has
brought the organic chemists’ dream within their
grasp. The 2-D INADEQUATE experiment can
give interpretable results with a 12-fold reduction
in the sample requirement. Fig. 7 shows a 2-D
,~‘(‘1”~‘(~“‘1’~“1”“/“‘~1~“‘i’~“1(
INADEQUATE spectrum from 3.8 mg of sucrose 8 7 6 5 4 3 2 t PPY
acquired in 62 h of data acquisition. The spectrum
on the left employs conventional weighting func- Fig. 6. Proton spectrum at 500 MHz from 1. Conventional 5 mm
tions and displays the resulting signals and noise probe (top) and magic angle spinning nanoprobe (bottom).
 J.N. ShoolerylProgress in Nuclear Magnetic Spectroscopy
Fig. 7. 2-D INADEQUATE spectrum of 3.8 mg of sucrose in a 40 bl nanocell, using conventional data processing (left). Complete set of
C-C bonding patterns recognized by pattern recognition computer analysis of the same NMR data (right).
1. Research in superconducting materials,
especially high T, superconductors, may result in
still higher field strength spectrometers.
2. Startling improvements in sensitivity could
result from noise reduction if the temperature of
the receiver coil and input circuitry can be lowered
to 4 K, especially if a superconducting coil material
can be used successfully.
3. Spectrometer data systems will become faster
and more powerful, and software will extract more
of the information encoded in the precession of the
nuclei, either by distinguishing it from background
noise, or by separating it into interpretable forms.
Experimental NMR has developed to its present
state through patient, persistent building of a body
of knowledge and technology, and through inter-
spersed flashes of genius. It has been a great privilege
to have been part of the first fifty years and a great
pleasure to know that NMR today shows promise of
even greater progress in the coming half century.
References
[I] LT. Arnold, S.S. Dharmatti and M.E. Packard, J. Chem.
Phys., 19 (1951) 507.
[2] J.N. Shoolery. Anal. Chem., 65 (1993) 731A.
28 11995) 37-52
[3] H.S. Gutowsky and C.J. Hoffman, J. Chem. Phys., 19
(1951) 1259.
[4] J.T. Arnold, Phys. Rev., 102k (1956) 136.
[S] F. Bloch, Phys. Rev., 94 (1954) 496.
[6] W.A. Anderson and J.T. Arnold, 94 (1954) 497.
[7] M.J.E. Golay. Rev. Sci. Instrum., 29 (1958) 313.
[8] W.A. Anderson, Rev. Sci. Instrum., 32 (1961) 241.
[9] J.P. Lloyd, Bull. Am. Phys. Sot., 1 (1956) 92.
[lo] R.H. Dieke, Rev. Sci. Instrum., 17 (1946) 268
[I I] V. Royden, Phys. Rev., 96 (1954) 543.
[12] F. Bloch, Phys. Rev., 93 (1954) 944.
[13] A.L. Bloom and J.N. Shoolery, Phys. Rev., 97 (1955) 1261.
il4] R.H. Varian, U.S. Patent 3. 109, 138, filed July II, 1957;
issued Oct. 29, 1963.
[IS] W.A. Anderson, U.S. Patent 3,085, 125, filed July 11, 1957;
issued April 9, 1963.
[16] F.A. Nelson and H.E. Weaver, Science, 146 (1964) 223.
(171 E-L. Hahn, Phys. Rev., 80 (1950) 580.
(181 E.L. Hahn and D.E. Maxwell, Phys. Rev., 88 (1952) 1070.
[19] R.H. Varian, U.S. Patent 2,561,490, filed Oct. 28, 1948;
issued July 24, 1951,
[20] R.H. Varian, U.S. Patent 3.287,629. filed Aug. 29, 1956;
issued Nov. 22, 1966.
[Zl] R.R. Ernst and W.A. Anderson, Rev. Sci. Instrum., 37
(1966) 93.
[22] W.A. Anderson and R.R. Ernst, U.S. Patent 3,475,680,
filed May 26, 1965; issued Oct. 28, 1969.
[23] R.R. Ernst, J. Magn. Reson., 3 (1970) 10.
[24] W.P. Aue, E. Bartholdi and R.R. Ernst, J. Chem. Phys., 64
(1976) 2229.
1251M.H. Levitt and R. Freeman, J. Magn. Reson., 33 (1979)473.
52 J.N. ShoolerylProgress in Nuclear Magnetic Spectroscopy 28 (1995) 37-52
[26] B.L. Tomlinson and H.D.W. Hill, J. Chem. Phys., 59
(1973) 1775.
[27] J.S. Waugh, L.M. Huber and V. Haeberlen, Phys. Rev.
Lett., 20 (1968) 180.
[28] J. Dadok and A.A. Bothner-By, Progress Report on Spec-
troscopy at 141 kG, presented at the 19th Experimental
NMR Conference, Blacksburg, VA, 1978.
[29] L. Miiller, J. Am. Chem. Sot., 101 (1979) 4481.
[30] A. Bax and S. Subramanian, J. Magn. Reson., 67 (1986) 565.
[31] A. Bax and M.F. Summers, J. Am Chem. Sot., 108 (1986)
2093.
[32] A. Bax, R. Freeman and T.A. Frienkiel, J. Am. Chem.
Sot., 103 (1981) 1202.
[33] R. Mukherjee, B.A. Da Silva, B.L. Das, P.A. Keifer and
J.N. Shoolery, Heterocycles, 32 (1991) 985.
[34] R.E. Hurd, J. Magn. Reson., 87 (1990) 422.
[35] W.L. Fitch, G. Detre, C.P. Holmes, J.N. Shoolery and P.A.
Keifer, J. Org. Chem., 59 (1994) 7955.