Pulsed Field Gradients

Pulsed- A short time period

Field- The B0 magnetic field
Gradient- A variation in some quantity with respect to another.

Pulsed Field Gradient = A short change in the magnetic field with respect to distance.
Normally applied in the z-direction although x and y pulsed field gradients exist as well.

PFGs have a time, a power, and a shape. The time is usually 100's of µs or a few ms, (gt1,
gt2 etc. is normally the parameter name on Varian spectrometers for gradient time 1, 2,
etc.). The power, reported in the literature in Gauss/cm, is converted to a DAC unit; the
conversion is spectrometer dependent (different hardware = different DAC to Gauss/cm
conversion). The shape is the shape of the pulse, to be discussed later.

For PFGs to be used, the instrument must have a gradient amplifier that creates the
pulses, a connection to the probe (in the case of a Varian, the current for the PFGs is
transmitted through the upper stack of the probe), and a probe with gradient coils. Not all
probes made for instruments have gradient coils, as the gradient coils usually limit the
temperature range of a probe (normally gradient coils cannot handle low temperature).

On recent NMR instruments (post ~1995), this is normal hardware; the routine
application of PFGs is easily the most significant development in the NMR field in the
last ~15 years. Few NMR experiments developed in the last 5-10 years do apply PFGs.
Nearly all previously developed NMR experiments have now been rewritten with PFGs.

The control of when PFGs are turned on and off are controlled by the software (pulse
sequence); Varian has an additional software override parameter (pfgon for pulsed field
gradient on, which is set to 'nny' for z-gradient, 'yyy' for x, y, and z-gradients).
PFGs have many uses in NMR:

1) Solvent Suppression- Gradient enhanced solvent suppression is a significant

improvement over standard solvent suppression techniques

2) Phase Cycle- Phase cycles can be reduced with PFGs. Many 2D phase cycles are
normally 8, 16, 32, or even 64 scans, meaning that you must acquire a multiple of that
many scans to acquire an experiment without artifacts. PFGs can often reduce these phase
cycles to 1, 2 or 4 scans by selecting specific coherence pathways.

3) Coherence Selection- Essentially in the same way as the phase cycle reduction and
solvent suppression, the gradients can greatly improve signal-to-noise in some
experiments (mostly proton-carbon or nitrogen correlation experiments) by specifically
selecting the proper magnetization for transfer.

4) Magnetic Resonance Imaging/NMR microscopy- Gradients are essential for either of

these NMR techniques. NMR microscopy is the application of magnetic resonance
imaging (MRI) principles to the study of small objects. Objects which are studied are
typically less than 5 mm in diameter. Resultant images can have 20 to 50 m resolution.

5) Diffusion NMR- NMR can be used to measure diffusion rates with PFGs.

6) Gradient Shimming- as discussed earlier

How do PFGs work?

The Gradient Echo:

Magnetization is initially rotated from the z-axis by a 90º hard pulse; a gradient pulse is
applied and causes dephasing of the signal dependent upon where in the NMR tube the
signal is. Chemically equivalent spins are now in the x-y plane but out of phase with
respect to each other; assuming the gradient is linear, the amount of dephasing can be
determined. In the same way that refocusing of magnetization occurs in normal spin-echo
experiments, the magnetization can be refocused by an inverted gradient pulse. The key
is that only selected magnetization will be refocused, the rest will be randomized.
Simple 1D gradient pulse sequence:

There is an RF pulse followed by a dephasing gradient. No signal is observed, as the

phases of the signals vary across the tube as a function of distance.

There is an RF pulse followed by a dephasing gradient then an inverted gradient pulse

which rephases the signal, a gradient echo. Full signal is observed.
Selective Refocusing

After a gradient pulse there is a spatially dependent Larmor frequency (ω(z)):

ω(z) = γ(B0 + Bg(z))

where Bg(z) is the spatially dependent magnetic field.

If the gradient is applied for a duration of τg seconds, the magnetization rotates through a
spatially dependent phase angle (Φ(z)):

Φ(z) = γ(B0 + Bg(z))τg = γB0τg + γBg(z)τg

where γBg(z)τg is the spatially dependent phase term caused by the gradient.

It can be shown that the quantum coherence order affects the rate of dephasing. Thus, this
expression becomes:

Φ(z) = pγBg(z)τg

where p is the coherence order (e.g. single quantum, double quantum).

If the coherence involves different nuclei (e.g. 1H and 13C), then:

Φ(z) = Bg(z)τg Σ(pγ)

To observe magnetization, the net phase must = 0, so to observe a double quantum

coherence to single quantum magnetization, the second gradient must be twice as intense
or twice as long.
Gradient Shimming

Earlier we discussed how to apply gradient shimming; now, we look back at it to examine
a little better why it works.

Shims are required to eliminate inhomogeneity of the magnetic field, as a magnet does
not have a perfectly homogenous magnetic field throughout the magnet. Thus, different
parts of the sample will see a slightly different field. The shims of the magnet are
supposed to make up for this problem. Gradient shimming uses magnetic resonance
imaging to get an image of the magnetic field, then adjust shims according to the image.
Since the purpose of shims is to make the field equal in all parts of the sample and the
goal of imaging is to observe a picture of the sample, combine the two and an image of
the magnetic field can show where the shims need to be adjusted.

All spins that contribute to a single line should have the same Larmor frequency except
that bad shimming causes slight differences- this is then translated to phase difference.
Gradients have the ability to separate out these differences as a function of distance from
the top (or bottom) of the sample for z-axis gradient shimming. In principle, x and y-axis
gradient shimming should work, but Varian spectrometers cannot do that yet.

First, an image of the probe is acquired, a reference map essentially, the gradient
shimming is done by trying to optimize response in a gradient echo experiment.

Errors in the local magnetic field cause phase differences between protons with
equivalent Larmor frequencies as a function of time in the x-y plane and distance in the
sample tube:
Gradient echo experiments are used to provide the spatial encoding of the phase errors as
a function of τ.

1D image of the sample:

Two Dimensional NMR

Goals of 2D NMR:

1) Disperse NMR signals over two dimensions, resolves problem of spectral overlap, as
in other 2D techniques (e.g. 2D electrophoresis which separates by charge and size)

2) Correlate NMR signals based upon measurable NMR parameters (e.g. chemical shift
of proton and carbon or chemical shift of J-coupled protons)

3) Observe resonances of a specific type (e.g. protons bound to carbons bound to

nitrogen, H-C-N)

4) Filter out resonances of a specific type (carbons bound to 2 other carbons)

What 2D NMR has yielded:

1) Full proton resonance assignments of MW ~20,000 molecules

2) High-resolution three-dimensional structures of MW ~10,000 molecules

3) Trivial resonance assignment of MW ~2,000 molecules

4) 3D and 4D NMR which has allowed full resonance assignment MW ~60,000

molecules and 3D structures of MW ~50,000 molecules
How do we achieve a second dimension?

It should be clear that the amplitude and phase of the magnetization detected during the
acquisition time is dependent upon the length of time that we wait between the two pulses
as magnetization is being lost due to T2 relaxation. The magnetization detected is given
by the equation:

M = M0 e-t1/T2
t1 = time, on the order of µs - ms

t2 = acquisition time period

T2 = transverse relaxation time

What if we arrayed the t1 time period by some set interval?

The frequency of the resonance stays the same, but the amplitude is being modulated by
the t1 time!

What does the amplitude modulation look like if you plot it as a function of the t1 time?
The second dimension looks like an FID!

Thus, both dimensions are Fourier transformed.

Standard pulse sequence setup for ALL 2D experiments:

Preparation ――― t1 ――― Mixing ――― Acquire

Coherence Transfer

The INEPT and DEPT discussion previously showed you could transfer magnetization
from proton to carbon to enhance signal intensity of the carbon resonance through the J-
coupled carbon and proton. That is heteronuclear coherence transfer- the magnetization
starts on 1H (the first pulse of the experiment is on 1H), and then by pulsing separately on
13
C and 1H, through use of spin-echo, the magnetization is transferred to 13C from 1H. At
the end of the experiment, 13C is detected. The effect of this in the case of the INEPT and
DEPT is that the larger population difference of 1H is transferred to 13C, increasing the
signal-to-noise of 13C. What if we tried to transfer magnetization from proton to proton
through J-coupling? Now, we are not able to pulse on 2 nuclei separately and
simultaneously as in the INEPT or DEPT experiment. If we could, this would be
homonuclear coherence transfer. The 90ºX --- t1 --- 90ºX --- Acquire sequence shown
previously accomplishes homonuclear coherence transfer.

Energy Level Diagram for AX System

H-C-C-H

Since each transition is affected by the other transitions, the amplitudes of the resonances
detected during t2 are modulated by all other transitions.
Jeener Experiment (COSY)

Contour Plot of the same experiment:

Quadrature Detection in Two Dimensions

Quadrature detection in the detect dimension of two-dimensional experiments is

accomplished in the same way as in 1D experiments. The indirect detect dimension, t1, is
similar.

Essentially there are 3 choices:

1) Absolute Value or Magnitude Spectrum- No quadrature detection in t1, all t1 FIDs are
real. On a Varian, the parameter phase = 0 yields a magnitude spectrum.

2) States (or States-Haberkorn or SHR)- As in the standard quadrature detection mode in

the detect dimension, there is a complex FID created by incrementing the phase of one
pulse by 90º and recording twice as many discreet t1 points, so there is phase difference
of 90º creating a real and imaginary FID for a complex FID in t1. The spectral width in t1
should be set the same as the spectral width in t2 for a 1H/1H experiment or in the same
manner; the spectral width is the real value. On a Varian, the parameter phase = 1,2 sets
States quadrature detection in t1.
Aliasing in States detection in t1 (if a signal is outside the spectral width):
3) TPPI (or the Redfield method)- Essentially the same as TPPI in the detect dimension
where the sampling rate is twice as fast as it should be and the phase of one pulse is
incremented in 90º increments in t1. The result is that the spectral width in t1 must be
twice as large as desired (because the sampling rate is twice as fast). On a Varian, the
parameter phase = 3 sets TPPI detection in t1.

Folding with TPPI for signals outside the spectral width:

(t2 acquisition is still States)

Two Dimensional Experiments

There are two basic types of two dimensional experiments: through-bond and through-
space.

Through-Bond Experiments:

Through-bond experiments correlate a spin 1/2 nucleus (virtually always proton) with
another spin 1/2 nucleus that are scalar (J) coupled to each other directly or indirectly
through another nucleus. Through-bond experiments can connect any spin 1/2 nucleus to
any other spin 1/2 nucleus where there is a measurable J-coupling (normally requires at
least ~1 Hz coupling). There are two types of through-bond experiments: homonuclear
and heteronuclear.

Homonuclear Through-Bond Experiments:

Homonuclear through-bond experiments correlate 2 spin 1/2 nuclei to each other (almost
always proton or fluorine) with each other.

COSY (COrrelated SpectroscopY) experiments correlate protons that are directly coupled
to each other (the protons are separated by less than ~3 bonds or 4 bonds in some cases).
In COSY experiments there will be a crosspeak in the 2D experiment that corresponds to
the two coupled protons. Coupling constants can be measured in some types of COSY
experiments, particularly DQF-COSY or E-COSY.

TOCSY (TOtal Correlation SpectrosopY; this used to be called HoHaHa for

Homonuclear Hartmann-Hahn) experiments correlate protons in the same spin system,
ones that are directly AND indirectly coupled to each other, meaning if proton A is
coupled to B and B is coupled to C, but A is not coupled to C, then in a COSY there will
a crosspeak from A to B and B to C but not A to C, whereas in a TOCSY, there will
crosspeaks from A to B and C, from B to A and C, and from C to A and B.

The INADEQUATE experiment (Incredible Natural Abundance Double Quantum

Transfer Experiment) normally correlates 13C and 13C as in a COSY like experiment

Heteronuclear Through-Bond Experiments:

Heteronuclear through-bond experiments correlate a proton and an X nucleus, usually

13
C, 15N, or 31P.
HSQC (Heteronuclear Single Quantum Coherence) and HMQC (Heteronuclear Multiple
Quantum Coherence) experiments correlate a proton and the X nucleus that it is
covalently bound to. There is no diagonal in this experiment, simply one peak for every
proton in the molecule that is bound to the X nucleus that you are indirectly detecting.
For example, in a 13C/1H HSQC, nitrogen bound protons will not be in the spectrum.

HMBC (Heteronuclear Multiple Bond Coherence) and Heteronuclear COSY experiments

correlate protons with X nuclei that are within 3-4 bonds. The HMBC filters out the X
nuclei that are directly bound to the proton, the Heteronuclear COSY does not.

HETCOR (Heteronuclear Correlation) is like HSQC/HMQC except the X nucleus is

detected rather than the proton

Through-Space Experiments:

Through-space experiments correlate protons with each other through-space, so protons

that are far apart in the molecule through-bond but are close in 3 dimensional space
(~< 0.5 Angstroms) will have an NOE to each other. There are two types of through-
space 2D experiments- NOESY (Nuclear Overhauser Effect SpectroscopY) and ROESY
(Rotating Overhauser Effect SpectroscopY). NOESY and ROESY each correlated
through space; the NOESY uses NOEs, the ROESY uses ROEs.
Glossary of Basic 2D Terminology

COSY Spectrum:

Diagonal = Peaks where νt1 = νt2

Crosspeak = Peak where νt1 = νt2

Glossary of Basic 2D Terminology

t2 = time domain of dimension that is being acquired (note the difference between t2 and
T2- the transverse relaxation time)

ν2 (or F2) = frequency dimension that is being acquired (some other software programs,
NOT VNMR refer to this as D1 or even F1)

t1 = time domain of second dimension, the one that is NOT being directly acquired (note
the difference between t1 and T1- the longitudinal relaxation time)

ν1 (or F1) = frequency dimension that is being acquired (some other software programs,
NOT VNMR refer to this as D2 or even F2)

direct detect dimension = ν2 = frequency dimension that is being acquired

indirect detect dimension = ν1 = frequency dimension that is NOT directly being

acquired

indirect detect experiment = an NMR experiment that records chemical shift of a nucleus
that is not being detected (such as a proton-carbon correlation where the acquired nucleus
is proton, while the carbon is indirectly detected during t1)
T1, t1, T2, t2

T1 = spin-lattice relaxation time constant

T2 = spin-spin relaxation time constant

t1 = indirect detect time dimension in 2D experiment

t2 = direct detect time dimension in 2D experiment

T1 affects the setting of pw and d1 in a 1D experiment and d1 in a 2D experiment. T1 can

affect the integration values in a 1D experiment acquired with more than one scan

T2 affects the setting of acquisition time (at) in a 1D or 2D experiment (although

acquisition time, at, in a 2D is limited by the amount of points that can reasonably be
acquired). The acquisition time is also affected by shimming in either case as shimming
affects T2* which is the observed time constant, which leads to the FID.

t1 = time in pulse sequence of 2D that indirect dimension is being acquired. When

Fourier transformed, t1 becomes F1. The length of time that t1 lasts is = ni/sw1 or the
number of increments (points) in the indirect detect dimension divided by the spectral
width in the indirect detect dimension.

t2 = time in pulse sequence of 2D that direct detect dimension is being acquired. . When
Fourier transformed, t2 becomes F2. The length of time that t2 lasts is = acquisition time
(at).

t1 and t2 are used in reference to pulse sequences and dimensions of a 2D.

T1 and T2 are discussed in reference to relaxation.

To further confuse you: in a 3D experiment, there would be t1, t2, and t3 with t3 now
being the dimension you are detecting. Also, on software other than Varian, the
dimension that you are detecting so t2 on a 2D or t3 on a 3D may be referred to as D1 or
F1 after transformation- in other the dimension names become reversed.
More Basic 2D Terminology

magnetization transfer = basically the transfer of the magnetization from one spin to
another, which is the critical ingredient for a useful 2D experiment

double resonance experiment = experiment with pulses on 2 nuclei (proton-carbon

correlations)

triple resonance experiment = experiment with pulses on 3 nuclei (usually for several
magnetization transfer steps)

stacked plot = way of plotting the 2D data that looks three dimensional, but is hard to
interpret

contour plot = way of plotting 2D data that looks like you are looking down at the plot, as
in a mountain range from the sky.

intensity plot = plot of intensity rather than contours, less three dimensional than contour
plot, does not look as well resolved as a contour plot, but alot faster to draw on the screen

digital resolution (t2) = 1/acquisition time for the t2 dimension

acquisition time (t2) = the acquisition time for the t2 dimension is the same as in a 1D
experiment, = number of points detected/(spectral width*2)

digital resolution in t1 should = 1/acquisition time for the t1 dimension, but there is no
real acquisition time in the t1 dimension. Thus, digital resolution in the t1 dimension =
2*(spectral width in the t1 dimension)/(number of complex points acquired in the t1
dimension). The Varian parameter for spectral width in the t1 dimension is sw1; the
Varian parameter for REAL points in the t1 dimension is ni (for number of t1
increments).
Thus, digital resolution in t1 = 2*sw1/2*ni = sw1/ni

acquisition time in t1 dimension = although there is no real acquisition time in t1

dimension, the equivalent to acquisition time = ni/sw1
artifacts = unwanted peaks arising from a variety of sources (full quantum mechanical
description of the specific pulse sequence would be required for discussion). Often
artifacts appear as weak intensity diagonals that are not in the center of the spectrum

axial peaks = apparent cross peaks to the center of the spectrum

t1 noise = noise in the t1 dimension (some of the major causes of the noise will be
discussed later)
Artifacts in 2D Experiments

1) Axial peaks-

A line of garbage peaks going across the spectrum in t2 at the center of the spectrum in
t1.

The axial peaks are normally suppressed with a full phase cycle and quadrature detection
in both dimensions.

2) t1 noise-

noise in the t1 dimension

Any instrumental problem that affects the intensity or phase of a resonance over time will
result in t1 noise as 2D experiments depend upon the modulation of phase and intensity
of resonances over time. Some problems are avoidable with proper setting of parameters,
others are instrumental and unavoidable. Instrumental (and thus unavoidable problems)
include magnet drift, thermal noise in the console, imperfect/inconsistent pulses, etc.

Important problems that are preventable:

1) Loss of lock or insufficient lock:

The lock power and gain should be set high enough to maintain lock throughout the
acquisition, but not too high to saturate it. Even brief loss of lock is a severe problem in a
2D- causes of brief loss of lock or just changes in the lock include moving metal objects
too close to the magnet during acquisition.

2) Temperature stability:

Resolution and thus intensity of resonances can be affected by temperature. Pulse widths,
tuning and shimming can be affected by temperature. Temperature regulation is critical
for t1 noise suppression.
3) Sample Heating:

Sample heating is separate from temperature stability as temperature stability refers more
to the temperature of the probe. Specific sample heating may not be significant enough to
heat the probe, but still be noticed on the sample. Sample heating arises from inputting
too much power to the probe for too long of a time. The primary reasons that sample
heating is possible is mistakes in acquisition parameter setup as decoupling (primarily
used in HSQC/HMQC type experiments) and mixing pulses (TOCSY and ROESY) have
moderate powers for long periods of time- milliseconds to seconds. The sample must be
allowed to cool between scans. Note that sample heating does not usually cause much t1
noise as the sample reaches a steady-state temperature.

4) Spinning:

If a spinner speed changes during the experiment, the lineshapes and thus the intensity
can change. The other (and bigger problem) with spinning induces a regular variation in
the quality factor of the probe as any asymmetry in the tube, sample, or spinner will be
reflected by a different response to a pulse- remember that the pulse is short in time scale
with respect to the spin speed, so the pulse could occur with the tube in a different
orientation each time.

5) Sample degradation:

Changes in the sample over time will obviously affect the spectra, mostly in t1.