NMR For The People

Nuclear Magnetic Resonance
for the People
October 31, 2011
Contents
I For the People 1
3 5 9 11 13 23
1 Introduction 2 Precessing Tops and the Faraday Detector 3 The Zeeman Interaction 4 The Chemical Shift Interaction 5 Magnetic Resonance, Coherence, and Relaxation 6 The Bloch Equations 6.1 6.2 6.3
Free Precession . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 An RF Pulse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 The Bloch Decay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 33
7 The Fourier Transform 7.1 7.2
Absorption and Dispersion Mode Lineshapes . . . . . . . . . . . . . . . . . . . . . . . . . 37 Phase Corrections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 7.2.1 7.2.2 Zeroth Order Phase Correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 First Order Phase Correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 45 49
8 Measuring Chemical Exchange - The Modied Bloch Equations

9 Inhomogeneous External Magnetic Fields and T2
ii 10 Measuring Relaxation Times
CONTENTS 51
10.1 Spin-Lattice Relaxation Times . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 10.1.1 The Saturation Recovery Experiment . . . . . . . . . . . . . . . . . . . . . . . . . 52 10.1.2 The Inversion Recovery Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . 53 10.2 Spin-Spin Relaxation Times . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 10.2.1 The Spin Echo Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 10.2.2 Carr-Purcell Meiboom-Gill sequence . . . . . . . . . . . . . . . . . . . . . . . . . . 56 11 Measuring Translational Diusion Coecients 59
11.1 Pulsed Field Gradients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 12 Coherence Transfer Pathways 13 Limitations of the Bloch Equations 63 69
13.1 Nuclei with Electric Quadrupole Moments . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 13.2 Coupling between distinguishable nuclei . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 13.2.1 Magnetic Dipole Coupling between Nuclei . . . . . . . . . . . . . . . . . . . . . . . 71 13.2.2 Indirect J Coupling between Nuclei . . . . . . . . . . . . . . . . . . . . . . . . . . 72 14 Interpreting Relaxation Times 75
14.1 Time Correlation and Spectral Density Functions . . . . . . . . . . . . . . . . . . . . . . . 75 14.2 Relaxation via Dipolar Couplings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 14.2.1 Nuclei with Identical Resonance Frequencies . . . . . . . . . . . . . . . . . . . . . . 78 14.2.2 Nuclei with Dierent Resonance Frequencies . . . . . . . . . . . . . . . . . . . . . 78 14.2.2.1 In the Motional Narrowing Regime . . . . . . . . . . . . . . . . . . . . . 79
14.2.3 Steady-State Overhauser Eect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 14.3 Quadrupolar Relaxation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 14.4 Nuclear Shielding Anisotropy Relaxation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 15 Multi-Dimensional NMR 83
15.1 2D Exchange and 2D NOESY NMR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 15.1.1 Transient nOes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 16 Magnetic Resonance Imaging P. J. Grandinetti, October 31, 2011 89
CONTENTS 17 Inside the Spectrometer
iii 93
17.1 The Magnet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 17.1.1 Improving Field Homogeneity - Shims . . . . . . . . . . . . . . . . . . . . . . . . . 94 17.1.2 Improving Field Drift - The Lock . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 17.2 A Primitive NMR Spectrometer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96 17.3 The Probe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 17.3.1 Review of RF circuit components . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 17.3.1.1 The Resistor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 17.3.1.2 The Inductor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99 17.3.1.3 The Capacitor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 17.3.2 Impedance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 17.3.3 A Tuned Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 17.3.4 A Matched Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104 17.4 The Duplexer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108 17.4.1 Cross-Diodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108 17.4.2 Transmission lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 17.5 The Transmitter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114 17.5.1 Specifying Power Levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114 17.5.2 Controlling the Transmitter Phase and Amplitude . . . . . . . . . . . . . . . . . . 116 17.5.2.1 The Mixer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116 17.6 The Receiver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 17.6.1 Quadrature Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120 17.6.2 Single Channel Detection and the Redeld Trick . . . . . . . . . . . . . . . . . . . 124 17.6.3 Discretely Sampled Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
17.6.4 Signal Averaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129 17.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130 17.8 Problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
II
For the Workers
133
137 141 P. J. Grandinetti, October 31, 2011
18 The Faraday Detector 19 Vector Rotation
iv 20 The Bloch Equations
CONTENTS 145
20.1 The Laboratory Frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145 20.2 The Rotating Frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 20.3 The Detector Coil in the Rotating Frame . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 20.4 The RF Pulse in the Rotating Frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150 21 Signal Processing 157
21.1 The Fourier Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157 21.1.1 Fourier Transform Pairs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157 21.1.1.1 Gaussian Decay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157 21.1.1.2 Gaussian Echo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158 21.1.1.3 Square Pulse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158 21.1.2 Fourier Transform Theorem Pairs . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 21.1.2.1 Time and Frequency Scaling . . . . . . . . . . . . . . . . . . . . . . . . . 159 21.1.2.2 Time and Frequency Shifting . . . . . . . . . . . . . . . . . . . . . . . . . 159 21.1.3 The Convolution Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 21.1.4 Magnitude and Phase Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160 21.2 Baseline Correction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161 21.3 Zero Filling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162 21.4 Lineshape Transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163 21.5 Apodization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164 21.6 Signal-to-Noise Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166 22 The Quantum Picture 167
22.1 Diracs Formulation of Quantum Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . 169 22.1.1 Probabilities and Inner Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169 22.1.2 Operators and Outer Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172 22.1.3 Expectation Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175 22.2 Spin > 1/2 Nuclei . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183 22.3 Coupling between nuclei . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186 22.3.1 Weak Couplings - AX system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187 22.3.2 Strong Couplings - AB system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188 22.4 Observables with a Continuous Set of Outcomes . . . . . . . . . . . . . . . . . . . . . . . 190 P. J. Grandinetti, October 31, 2011
CONTENTS
v 22.4.0.1 Probability, Charge and Current Density . . . . . . . . . . . . . . . . . . 190
22.5 Quantum Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191 22.5.1 Interaction Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197 22.5.1.1 Double Rotating Frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204 22.5.2 Observables with a Continuous Set of Outcomes . . . . . . . . . . . . . . . . . . . 205 22.6 Transitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206 23 The Density Operator 209
23.1 The Projection Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209 23.2 The Density Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213 23.3 Thermal Equilibrium Density Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214 23.3.1 The Density Operator in the Rotating Frame . . . . . . . . . . . . . . . . . . . . . 215 23.4 Ensemble of Uncoupled Spins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216 23.4.1 The Bloch Decay . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217 23.4.2 The Spin Echo Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217 23.5 Ensemble of Heteronuclear Coupled Spins . . . . . . . . . . . . . . . . . . . . . . . . . . . 218 23.6 Ensemble of Weakly Homonuclear Coupled Spins . . . . . . . . . . . . . . . . . . . . . . . 218 23.6.1 The Bloch Decay for Two Coupled Spin 1/2 Nuclei . . . . . . . . . . . . . . . . . . 222 23.6.2 The Spin Echo Experiment For Two Coupled Spin 1/2 Nuclei . . . . . . . . . . . . 222
P. J. Grandinetti, October 31, 2011
Part I
For the People
Chapter 1
Introduction
Since NMR was rst demonstrated in condensed matter in 1945, the technique has experienced tremendous growth. NMR was commercialized in the 1960s and by the end that decade NMR was routinely used as a central tool by synthetic organic chemists. Compared to other spectroscopies NMR spectra are
H3C(25) 13 14 21 (26) C H3 9 C(22)H3 5 10 8 23 18 H3(22)C 18 4 6 11 20 C(22)H3 16 17 12 HO 3 1 7 2 15
150
100 Frequency (ppm)
50
Figure 1.1: 1 H-decoupled
13
C NMR spectrum of Cholesterol. There are 26 resonances in this spectrum,
which when numbered from left to right, are assigned to the corresponding numbered carbon atom in Cholesterol based on its resonance frequency. 3
CHAPTER 1. INTRODUCTION
embarrassingly easy to interpret and use as a tool for determining molecular structure. The frequency of each line tells you not only what type of atomic nuclei are present but can also distinguish among the same type of nuclei in dierent chemical bonding environments. Additionally, the intensity of each signal is quantitative. What could be simpler? With increasing magnetic eld strengths, the separation between frequencies (in Hertz) increases, and thus, the ability to resolve overlapping peaks improves and larger molecules can be studied. Higher magnetic eld strengths also oer signicant improvements in sensitivity. In the early years most NMR spectrometers were continuous-wave instruments. Today, practically all NMR experiments are performed as pulsed time domain Fourier transform experiments. We will examine the reasons for this situation later. It would be dicult and unnecessary to discuss all the NMR experiments (pulse sequences) that have been developed and applied over the years. The aim of this text is to give you a basic understanding of the underlying theoretical and experimental aspects of the technique. With this understanding you should be able to follow many articles in the NMR literature and know how to implement most NMR experiments on a modern NMR spectrometer and analyze the data. In this chapter, we will present a simple classical model for understanding the NMR experiment, intentionally avoiding the use of quantum mechanics so that a broader range of students can more quickly understand many key concepts behind the technique. In later chapters, we will review the necessary aspects of the quantum theory and connect them to the concepts described in this chapter.
Chapter 2
Precessing Tops and the Faraday Detector

Lets start with the example of a macroscopic top spinning in a gravitational eld.
Z
The top will precess about the direction of the gravitational eld, with a characteristic frequency determined by parameters such as the mass, moments of inertia and the gravitational eld strength, g. If we eliminate g, that is, place the top in a zero gravity environment, it would stop precessing. If we then inserted a bar magnet inside our macroscopic top and placed it in a magnetic eld, while still in a zero gravity environment,
Z
N S
g=0
CHAPTER 2. PRECESSING TOPS AND THE FARADAY DETECTOR
the magnetic top would precess about the direction of the magnetic eld with a characteristic frequency 0 that is linearly proportional to the strength of the magnetic eld B0 . In NMR, 0 is called the Larmor frequency, and the time of one precession is called the Larmor period. How might one measure the precession frequency for such a magnetic top? One way is to exploit Faradays Law, which tells us that a changing magnetic eld will induce a current in a surrounding loop of wire. Thus, we place a coil of wire of radius rcoil around our spinning magnetic top.
Z
B
Y
S
X
N
Faradays Law tells us that the Electromotive Force (EMF, i.e., voltage) induced in the coil will be related to the change in magnetic ux with time. E = d dt (2.1)
Here E is the EMF and is the magnetic ux. In our example, the magnetic dipole vector of the top will change with time according to
0 t
Y X
(t) = || [sin cos(0 t + 0 ) ex + sin sin(0 t + 0 ) ey + cos ez ] ,
(2.2)
where || is the length of the precessing vector, is the angle between the precessing vector and the z -axis, 0 is the initial phase of the precessing vector, and 0 is the precession frequency. With our precessing magnetic dipole at the origin the EMF induced in the coil will be Ex (t) = P. J. Grandinetti, October 31, 2011 dx (t) 0 = 0 || sin sin(0 t + 0 ). dt 2rcoil (2.3)
7 From this signal we can measure the precession frequency. Lets consider the factors that inuence the amplitude of this EMF signal. First, we see that the amplitude is directly proportional to the magnetic dipole moment strength, ||. Next, there is a scaling by sin , implying that the closer the magnetic dipole precesses to the z -axis, the smaller the EMF signal. There is also a scaling by the inverse of the coil radius, so that the EMF signal amplitude decreases with increasing coil radius. And nally, we note that the EMF signal amplitude increases with increasing precession frequency. Thus, we see part of the sensitivity advantage to having the highest possible static elds where the precession frequency will be greatest.
CHAPTER 2. PRECESSING TOPS AND THE FARADAY DETECTOR
Chapter 3
The Zeeman Interaction

Now we can make the connection to NMR. Every NMR active nucleus (i.e., its spin angular momentum quantum number, I , is not zero) can be thought as a microscopic magnetic top, with a magnetic dipole moment given by = I. (3.1)
Here, I is the nuclear spin angular momentum vector and is the nuclear gyromagnetic ratio. Values of the gyromagnetic ratio for selected nuclei are given in Table 3.1. When you place an NMR active nucleus in a magnetic eld, B0 , its magnetic dipole moment, , couples to the magnetic eld, with an energy of interaction given by E = B0 . (3.2)
For nuclei, this coupling, called the Zeeman coupling, causes a precession about the magnetic eld direction at a frequency 0 = B0 , where B0 is the strength of the magnetic eld vector, B0 .
10
CHAPTER 3. THE ZEEMAN INTERACTION
Nucleus
Spin
Natural Abundance (Percent)
gyromagnetic ratio (107 rad/T-s)
Quadrupole Moment Q (barns) 2 .8 103
Frequency at B0 = 2.35 Tesla 0 /2 (MHz) 100 15.351 106.663974 76.179437 14.716086 38.863797 32.089 25.145004 7.226329 10.136783 13.561 94.094003 26.466 6.121642 26.056888 19.867184 40.480737 9.809 24.001253 30.496576 6.256819 27.835 4.047878 4.653601 1.841000
1H 2H 3H 3 He 6 Li 7 Li 11 B 13 C 14 N 15 N 17 O 19 F 23 Na 25 Mg 27 Al 29 Si 31 P 35 Cl 69 Ga 71 Ga 67 Zn 87 Rb 107 Ag 109 Ag 235 U
1/2 1 1/2 1/2 1 3/2 3/2 1/2 1 1/2 5/2 1/2 3/2 5/2 5/2 1/2 1/2 3/2 3/2 3/2 5/2 3/2 1/2 1/2 7/2
99.985 0.015 0.000137 7.59 92.41 80.42 1.108 99.634 0.37 0.037 100 100 10 100 4.70 100 75.53 60.108 39.892 4.1 27.835 51.839 48.161 0.7200
26.7519 4.1066 28.5349779 -20.3801587 3.9371709 10.3977013 8.5843 6.7283 1.9337792 -2.712 -3.6279 25.181 7.08013 -1.63887 6.9762715 -5.3188 10.8394 2.6240 6.438855 8.181171 1.676688 8.786400 -1.0889181 -1.2518634 -0.52
-0.808 -40.1 4.1 102 0.02044 2.6 102 0.10 0.1994 0.1403 0.10 0.171 0.107 0.150 0.132 4.936
Table 3.1: NMR Properties for selected nuclei obtained from www.webelements.com. 1 barn = 1024 cm2 .
Chapter 4
The Chemical Shift Interaction

The slight variations in the resonance frequencies in Fig. 1.1 are caused by the nuclear shielding eect. This eect arises from surrounding electrons that slightly shield the nucleus from the full strength of the external magnetic eld. Nuclear shieldings depends on a number of factors, and often increases with increasing local electron density around a nucleus. Thus, we write the precession frequency as = (1 )B0 , (4.1)
where is a dimensionless quantity, called the isotropic nuclear shielding, and is typically on the order of 106 . While an NMR spectrometer can measure the precession frequency with extremely high precision and accuracy, a direct measurement of using Eq. (4.1) is problematic since it requires equally precise and accurate measurements of B0 , which is complicated by the need to take into account the samples bulk magnetic suseptibiliy and shape. For this reason, it is more convenient to use the chemical shift, which is dened in terms of the dierence between NMR frequency of a given resonance and the NMR frequency of a resonance in a reference compound, that is, sample = sample reference reference (4.2)
which, dened in terms of the nuclear shielding, is sample = reference sample . 1 reference (4.3)
Often chemical shifts are reported in units of parts per million, obtained by multiplying by 106 . In solution state NMR, an internal reference, that is, a compound homogeneously mixed into the sample, 11
12
CHAPTER 4. THE CHEMICAL SHIFT INTERACTION
is often used. This is preferred over external references, since it eliminates the need to correct for bulk magnetic suseptibilities dierences.
Chapter 5
Magnetic Resonance, Coherence, and Relaxation

One of the main objectives in the NMR experiment is to measure the precession frequency of every chemically distinct nucleus in the sample. To do this, we place a coil around a sample in a strong magnetic eld and measure the oscillating voltage induced in the coil by all the nuclear spins precessing. There is just one complication with this approach. While all the nuclear magnetic dipole vectors will precess about the magnetic eld direction when the sample is placed in a magnetic eld, there will be a random distribution of angles from the z -axis, and angles in the x-y plane, in other words, a random distribution of and . Imagine approximately 1020 nuclear magnetic dipole vectors pointing in random directions in three dimensions and all of them precessing about the external magnetic eld.
Y X
The vector sum of all the microscopic nuclear spin magnetic dipole vectors has a projection in the x-y plane of zero! Alas, there is no signal to detect with our Faraday detector. 13
14
CHAPTER 5. MAGNETIC RESONANCE, COHERENCE, AND RELAXATION Fortunately, there is a solution. After the sample comes to thermal equilibrium in the magnetic eld,
there are a few more nuclear magnetic dipole vectors on the +z half of the sphere than the z half of the sphere, or if you change the sign of the gyromagnetic ratio, more vectors on the z half than the +z half. Thus, there is a small non-zero vector sum pointing along the z -axis (i.e., along the magnetic eld direction). Here, we say the spin system has Zeeman spin order. To a good approximation, the size of this vector from Zeeman spin order is related to the strength of the magnetic eld, according to Meq = 1 N 2 2 I (I + 1)B0 , 3 kB T is Plancks constant,
where N is the total number of nuclei, I is the nuclear spin angular momentum, kB is Boltzmanns constant, and T is the temperature.
Since this magnetization vector is along the z -axis it will give no signal, but, if we can move this net magnetization vector away from the z -axis, then it will precess and induce a signal in our receiver coil1 . One simple experimental solution is Magnetic Resonance, developed in the 1930s by I. I. Rabi, and for which he received the Nobel Prize in Physics in 1944. We start with the net magnetization vector lined up with the B0 vector.
Z net magnetization vector B0 - magnetic field vector
Then, we apply a second magnetic eld, B1 , along the y -axis (which, in practice, is thousands of times smaller than B0 ), so that the total external magnetic eld vector now is ever so slightly tilted away from the z -axis.
1 In
principle, one approach to do this would be to hop the magnetic eld direction away from the magnetization
suddenly. Then the net magnetization (still along the z -axis) would precess around the new magnetic eld direction (no longer aligned with the z -axis) and could be detected by a receiver coil. In practice, this is quite dicult, since you would need to move the magnetic eld direction much faster than one precession period, which in NMR would mean less than a nanosecond.
15
Z B0 Beff
B1
The net magnetization vector will begin precessing about this new eective magnetic eld direction. After the magnetization has precessed 180 around the new eective magnetic eld, we can switch the small B1 to lie along the y axis.
Z Beff B0
B1
After switching the B1 direction, the magnetization vector will precess along a wider cone about the newer eective magnetic eld direction. Once the net magnetization has processed 180 around this eective magnetic eld direction, we switch the B1 vector back along +y , and repeat the process. Eventually, we can take the net magnetization vector from along +z and move it completely into the x-y plane. All we needed to accomplish this, was to apply a small alternating magnetic eld along the y -axis that alternates with the same frequency that the net magnetization vector naturally precesses around the magnetic eld direction. This process is called magnetic resonance. It allows us to take the net magnetization vector from along the z -axis and place it in the x-y plane2 .
2 Clearly
to do a magnetic resonance experiment you need to know the precession frequency. You may be thinking that
if you already knew the precession frequency why would you even bother doing the experiment? It turns out that you dont need to know the exact resonance frequency to do magnetic resonance. You can still move the magnetization vector away from the z -axis if youre slightly o resonance. Once there is a detectable magnetization component in the transverse
16
CHAPTER 5. MAGNETIC RESONANCE, COHERENCE, AND RELAXATION

Z Z
magnetic resonance
Once the magnetization vector is completely in x-y plane, we can turn o the alternating B1 eld and the magnetization will precess in the x-y plane and induce an EMF signal in a coil. We can even use the same coil that we use to receive the signal to generate the alternating B1 magnetic eld. By taking the magnetization vector from the z axis to the x-y plane we have turned the Zeeman order of the spin system into a coherence. The concept of coherence is central to understanding nearly all NMR experiments, so it is worthwhile to examine this concept more carefully. In NMR, coherence implies a correlated phase relationship among the orientations of many nuclear magnetic dipole vectors. Dened in this way, coherence has no meaning for a single nuclear spin3 and is only a property of many nuclear spins in an ensemble. Consider an analogy to the Mexican Wave4 , a popular activity at football stadiums around the world. Lets say it takes one minute for the wave to travel around the stadium. One spectator alone standing up and then down once a minute will not create an observable wave. Likewise, all the spectators in the stadium standing up and down once a minute but each at random times will also not create an observable wave. It is only when there is a certain correlation between every spectators standing cycle will there be an observable wave. The Mexican Wave requires coherence among many spectators to be observable. In NMR there are relaxation processes that will destroy (randomize) the correlated phase relationship among the precessing nuclear magnetic dipole vectors in the sample, causing the spins to lose coherence. These relaxation processes, which randomize the phase relationships among the individual spin magnetic dipole vectors, eventually lead to the net magnetization being completely along the z -axis again. Two time constants often used to characterize these processes are:
plane, you can measure the frequency very accurately with your detector coil. 3 Coherence can also dened as a correlated phase relationship for a single spin at dierent times. Do not confuse this usage with the one above, where theres a relationship among many spins at a given time. 4 First became popular during the 1986 World Cup in Mexico. Spectators jump to their feet with arms outstretched and sit down again as neighbors in the stand rise up.
17 T2 : Spin-Spin (Transverse) Relaxation Time - Time scale that coherence is lost in the x-y (transverse) plane.
T1 : Spin-lattice (Longitudinal) Relaxation Time - Time scale that the equilibrium net magnetization along the z (longitudinal) axis is restored.
So far, weve only considered the interaction of the nuclear magnetic dipole moment with the external magnetic elds, such as the static B0 , and the time dependent B1 (t). Generally, a nucleus will experience a variety of interactions between all its electric and magnetic moments and its atomic and molecular surroundings (see section 13). NMR Relaxation arises from the time dependence of these interactions due to random molecular motion. For simplicity, well consider here just the interactions of an atoms nuclear magnetic dipole moment with the magnetic dipolar moments on neighboring atoms. Imagine a nuclear magnetic moment vector precessing about the external B0 eld on a cone of angle . As neighboring molecules, containing their own nuclear magnetic moments, tumble and translate, they will generate a local oscillating magnetic eld that can act on our nuclear magnetic moment vector and change not only the angle about which it is precessing, but also advance or retard , the phase of its precession. Although this process is happening to every nucleus in the sample, the phase and amplitude of the local eld uctuations experienced vary from nucleus to nucleus. This is the mechanism that destroys the phase coherence among the precessing nuclear spin magnetic moments and results in the T2 decay of the transverse magnetization components, as well as the T1 recovery of the equilibrium magnetization along the z axis. Note that T1 relaxation is aected only by local elds that change the angle , and not , whereas T2 relaxation is aected by local elds that change either or . A nice feature of this simple picture for NMR relaxation is that it qualitatively explains the dependence of relaxation times on the time scale for molecular motion. When molecules are tumbling rapidly, such that the molecular correlation time is much shorter than the Larmor period, then the local eld oscillation frequencies will be o-resonant, and their ability to move a nuclear spins magnetic moment vector away from the angle will be diminished. Likewise, if the molecular correlation time is much longer than the Larmor period, the local eld oscillations will again be o-resonant and less eective in reorienting the nuclear magnetic moment vector. Its only when the molecular correlation time is close to the Larmor period that we will nd the most ecient relaxation, that is, the shortest T1 value. Thus, we observe a minimum in the T1 value as a function of molecular correlation time when 0 1. P. J. Grandinetti, October 31, 2011
18
Compound water ethanol glycerol glycerol chloroform-d 0.5 M sucrose in H2 O methyl-cyclohexane
Phase liquid liquid liquid liquid liquid liquid

13
Nucleus
1 1 1 2 2
Temperature 20 C 20 C 20 C 20 C 25 C 27 C 30 C

Larmor Frequency 29 MHz 29 MHz 29 MHz 9.2 MHz 9.2 MHz 15.08 MHz 15.09 MHz
T1 value or range 2.3 seconds 2.2 seconds 23 milliseconds 2 seconds 1.35 seconds 1-2 seconds 9-18 seconds
H H H D D C
13
liquid
Table 5.1: Spin-lattice relaxation times, T1 , for selected substances.
Log()
Because T2 relaxation is enhanced by any process that changes , the phase of the spin precession it will also be aected by variations in static local elds, which can slow down or speed up the precession of the nuclear magnetic moment vectors. Such static local elds do not aect the precession angle , only its phase, . Thus, while T1 relaxation is unaected by static local elds, T2 relaxation will not increase like T1 in the long molecular correlation time limit. Eventually, the molecular correlation time becomes so slow that the sample becomes a rigid lattice, and then T2 remains relatively constant. Note that T1 will always be greater than or equal to T2 . P. J. Grandinetti, October 31, 2011
19 Theres a problem with our classical picture of the nucleus as a microscopic spinning magnetic top. If we tried to observe the motion of an individual nuclear magnetic dipole vector precessing we would run into some quantum weirdness. For example, lets say we wanted to measure the z -component of the nuclear magnetic dipole vector. If the nucleus behaved like a classical magnetic top, then we would expect for a sample of randomly oriented of nuclei that we should measure a random distribution of z -components that vary continuously between || and ||. That is, if was pointing along the +z axis we would expect to measure z = ||,
Z
Y X
and if was pointing along x or y then we would expect to measure z = 0.

Z
Y X
Instead, we nd that for a random sample of spin 1/2 nuclei we only measured two possible values for z . Strange indeed! The two values we measure are equal in magnitude but opposite in sign. In fact, for a spin 1/2 nucleus || = 3 /2, and the two measured values of z are + /2 and /2.
Randomly oriented spin 1/2 nuclei measure z
z = + h/2 group
z = h/2 group
It turns out that the same is true for the x and y components. Instead of a continuous range of values from + 3 /2 to 3 /2 you can only measure x or y as /2. What happened to all the values P. J. Grandinetti, October 31, 2011
20
between + 3 /2 and 3 /2? Good question! Once you acknowledge the wave-like properties of matter one can show that observable quantities that appear continuous in the macroscopic realm can become quantized in the microscopic world. Generally, the presence of a magnetic moment implies an underlying circulating charge with angular momentum. The magnitude of the magnetic moment is = I (I + 1) . Generally, for a nucleus with spin I , there will be 2I + 1 quantized observable
outcomes of mI = I, (I 1), . . . , (I 1), I , and the measured values of x , y , or z will be mI . We will Dirac notation, where these these observable outcomes (or states) are shown using a bra and ket notation, given by mI | and |mI , respectively. Using this notation, we relate a transition mI mI to the notation |mI mI |. Additionally, we relate the number of nuclei in an ensemble found in the state mI to the notation |mI mI |. Recalling that the Zeeman interaction contribution to the energy was given by Eq. (3.2), and therefore the energy of the two outcomes vary linearly as a function of magnetic eld
Energy
B0
So far, we have considered nuclei of spin I = 1/2, with two observable outcomes mI = |1/2 . Approximately 70% of the NMR active nuclei, however, have nuclear spin values of I > 1/2. For example, for a nucleus with spin I = 1, a measurement of the z component in a sample of randomly oriented nuclei would result in only three equally probably outcomes + , 0, and , associated with |mI = |+1 , |0 , and |1 . P. J. Grandinetti, October 31, 2011
21
Randomly oriented spin 1 nuclei measure z
z = + h group 33.3%
z = 0 group 33.3%
z = h group 33.3%
Strangely enough, the angular momentum of our spinning magnetic nucleus occur with half-integer values, which would have no analog in the macroscopic world. It was only when Paul Dirac combined quantum theory with Einsteins theory of relativity that the origin of half-integer spin were nally understood5 .
5 For
a enlightening discussion on the origins of spin angular momentum see the article Whence Spin by Nicholas
Zumbulyadis, Concepts in Magnetic Resonance, 3, 89-107(1991). Also, there are some quantum mechanical subtleties in this denition of spin angular momentum that are described nicely in Marvin Chesters excellent book, Primer of Quantum Mechanics.
22
How might one measure the z-component of a nuclear magnetic dipole vector? You could take the same approach that O. Stern and W. Gerlach did in 1921 when they measured the magnetic moment associated with an neutral silver atom. A neutral silver atom has an electron conguration of [Kr] 5s1 4d10 . With this conguration the magnetic dipole moment of the neutral silver atom arises almost completely from the unpaired electron in the 5s orbital. Additionally, since an electron in this orbital has no orbital angular momentum, the angular momentum associated with the neutral silver atoms magnetic dipole moment comes entirely from the intrinsic angular momentum of the electrons spin. In order to separate each neutral silver atom according to the z -component its magnetic moment, Stern and Gerlach sent a beam of neutral silver atoms emitted from an oven of silver vapor through a region of space with a strong magnetic eld gradient before impacting on a photographic plate. Why a region of space with a strong magnetic eld gradient? Recall that the potential energy of a magnetic dipole moment interacting with a magnetic eld is E = B, and the force acting on a particle is F= dE . dr (5.2) (5.1)
Since the silver atom is neutral there are no Lorentz forces and only the force associated with a magnetic dipole passing through a magnetic eld gradient acts on the silver atom, F= dB . dr (5.3)
Chapter 6
The Bloch Equations

We can calculate the magnetization vector in the NMR experiment with a vector sum of the individual nucleis magnetic dipole moments according to
N
M(t) =
j
j (t).
(6.1)
In 1946 Felix Bloch proposed a phenomenological set of equations to describe the precession and relaxation of the net magnetization vector in the NMR experiment. Given a magnetization vector, M(t) = Mx (t)ex + My (t)ey + Mz (t)ez , it will evolve according to dM(t) = (t) M(t) [R] [M(t) Meq ] , dt where (t) = (1 )B(t), 1/T2 0 and [R] = 0 1/T2 0 0 (6.3) (6.2)
0 0 1/T1
Meq = Meq ez
In the Bloch equations the [R] [M(t) Meq ] term describes the relaxation decay of x-y transverse magnetization and the growth of z longitudinal magnetization (similar to rst-order chemical kinetics). The (t) M(t) term describes the change in M(t) as it precesses about the B(t) direction. That is, 23
24
CHAPTER 6. THE BLOCH EQUATIONS
the magnetization vector, M(t), precesses about the instantaneous direction of the vector (t) with an instantaneous precession frequency specied by the length of the vector (t). The precession of the magnetization vector has so far been described with respect to a stationary reference frame associated with the lab. What if we stood on a turntable at the origin of our magnetization vector, and the turntable was spinning at the same angular velocity as the magnetization vector?
-X -Y X Y
Then the lab frame would appear (to us) to be rotating in the opposite direction, and the magnetization would appear stationary in this rotating frame. You can then imagine that the fast precessing and relaxing magnetization vector in the lab frame would not be precessing at all if your rotating frame exactly matched the precession frequency. That is, in the rotating frame the magnetization vector sum would just move from y to z , or if T1 Meq along z . Transforming the Bloch Equations into this frame we obtain d M(t) = e (t) M(t) [R]{M(t) Meq }, dt T2 the vector would shrink to zero along y and then grow to
(6.4)
where the magnetization vector, M(t), precesses in the rotating frame about a direction with a lower precession frequency given by e (t) = (t) rot . (6.5)
Here rot arises from going into the rotating frame. With Eq. (6.4), we can write the rate of change for the magnetization vector components in the rotating frame. Lets rst start with the static magnetic eld applied alone, and then consider the static eld and an oscillating magnetic resonance eld. P. J. Grandinetti, October 31, 2011
6.1. FREE PRECESSION
25
6.1
Free Precession
In the presence of a static magnetic eld alone the rotating frame precession frequency will be = 0 (1 ) rot , and the general solution for the M (t) x My (t) Mz (t) magnetization in the rotating frame will be Mx (0) cos t My (0) sin t et/T2 = My (0) cos t + Mx (0) sin t et/T2 Mz (0)et/T1 + Meq (1 et/T1 ) (6.6)
, (6.7)
are the magnetization vector components in the rotating frame. , and Mz , My where Mx
When describing free precession it is often convenient to dene the magnetization vector in NMR in terms of spherical basis vectors, that is,
1 M(t) = M+1 (t)e+1 + M0 (t)e0 + M , 1 (t)e
where eq are dened by 1 e1 = (e x iey ), 2 and the components Mq are given by 1 M 1 = (Mx iMy ), 2
M0 = Mz .
e0 = e z,
(6.8)
(6.9)
Note the sign dierence in the denition of e1 and M1 . With this denition the general solution for free precession of the transverse components becomes
it t/T2 M e . 1 (t) = M1 (0)e
Using spherical basis vectors gives us a more compact means for describing the length, orientation in the x-y plane, and sense of rotation about the z -axis of the transverse components of the magnetization
vector. Most importantly, note that other than being multiplied by a phase factor eit , the M 1
components are unaected by rotation about the z -axis.

Another advantage to using the spherical basis is that the components M+1 and M 1 are directly
related to the transitions + 1 2
1 1 2 and 2
1 +2 , respectively, in an ensemble of uncoupled spin I = 1 +2 1 2 1 2
1 2
nuclei. Additionally, the component M0 is related to + 1 2
1 2 , that is, the dierence in
populations of states in an ensemble of uncoupled spin I =
nuclei. P. J. Grandinetti, October 31, 2011
26
CHAPTER 6. THE BLOCH EQUATIONS Finally, the signal detected in the NMR experiment, written in terms of the magnetization compo-
nents in the rotating frame, is given by

S (t) = kM+1 (t)eiref ,
(6.10)
where ref is the receiver reference phase and k is a constant that depends on the receiver coil geometry and the receiver frequency.
6.2
An RF Pulse
Next, we consider the more dicult case of the static eld and a magnetic eld oscillating at a frequency |rf |, which is always dened as positive. In the lab frame, we have B(t) = 2B1 cos(|rf |t + rf )ex + B0 ez . Taken into a rotating frame we obtain1 an time independent expression for e ,
e = 1 [cos e x + sin ey ].
(6.11)
Laboratory Frame
Z B0 Beff(t)
Rotating Frame
Z*
Y X X* Y*
B1 Bx(t)
eff
2B1cos(|rf|t + rf)
B1 time
1 Here,
we have dened the nutation or Rabi frequency as 1 = |B1 |, and have redened the phase of radio frequency
(rf) as = (sign ) rf to yield an expression for e that is independent of the sign of .
6.2. AN RF PULSE If the rf phase is set to = 0, then we get a solution for the transformation of M(0) into M(t)
Mx
27
My Mz
Rx (1 t) My cos 1 t Mz sin 1 t Mz cos 1 t + My sin 1 t
Mx
rf pulse along x -axis ( = 0),
(6.12)
Thus, the magnetic resonance approach to moving the magnetization from along z to the x -y plane is seen as a simple rotation about a single axis in the rotating frame. A magnetization vector initially along the +z axis will be rotated counterclockwise to the y axis after a time 1 t = /2.
Z* Z*
Y* X*
1 t = /2
X*
Y*
1
The time is called a 90 (or /2) pulse length in NMR. The maximum signal will be detected in the receiver coil after a 90 pulse. If you turned the rf pulse on long enough so that 1 t = , then you could also rotate the magnetization from +z to z .
Z* Z*
Y* X*
1 t =
X*
Y*
1
This is called a -pulse (180 -pulse). In the specic case when magnetization is taken from +z to z , it is called an inversion pulse. After an inversion pulse, there will be no detectable signal in the receiver coil. In practice, the strength of the 1 is determined experimentally by systematically increasing the pulse length until a full oscillation in signal strength is observed. P. J. Grandinetti, October 31, 2011
28
CHAPTER 6. THE BLOCH EQUATIONS Similarly, if we set = /2, then we would have a transformation of M(0) into M(t) according to
Mx
My Mz
Ry (1 t) My Mz cos 1 t Mx sin 1 t
Mx cos 1 t + Mz sin 1 t
rf pulse along y -axis ( = /2).
(6.13)
Setting 1 t = /2 with = /2 gives a counterclockwise rotation about the y -axis which will rotate a magnetization vector along +z to the +x axis.
Z* Z*
Y* X*
1 t = /2
Y* X*
Finally, lets consider the eect of an rf pulse of arbitrary phase on the magnetization vector expressed in terms of the spherical basis vector components. We nd the eect of a /2 pulse of arbitrary phase to be
M+1
R (/2) i i i i M M0 2 M+1 e 1e 2 1 i 1 i i2 M M + M e + M e 1 1 0 +1 2 2 2 and the eect of a pulse of arbitrary phase to be i2 M M 1 e +1 R () M0 M0 . i2 M M e 1 +1
1 i2 2 M1 e
i i M0 e +1 2 M+1 2
(6.14)
(6.15)
Notice that the transverse components of the magnetization vector in the spherical unit vector bases,
and M that is, M+1 1 , are swapped by a pulse,
In the literature, one often sees the following notations in NMR: (/2)y which means a 1 t = /2 pulse of B1 along the y -axis of the rotating frame (/2)x which means a 1 t = /2 pulse along x-axis of rotating frame (/2)y or (/2)y which means a 1 t = /2 pulse along y axis of rotating frame P. J. Grandinetti, October 31, 2011
6.3. THE BLOCH DECAY ( )x which means a 1 t = pulse along the x-axis of the rotating frame Generally, one writes ( ) where = 1 t and is the rf pulse phase, that is, = = = = 0 90 180
29
x y x y
270
6.3
The Bloch Decay
Now lets consider the trajectory of the magnetization vector in the simplest NMR experiment: A single (/2)x pulse followed by acquisition of the NMR signal. Since we will ultimately be detecting the complex
magnetization component M+1 (t) in the rotating frame with our receiver coil it is more convenient to
follow the trajectory of the magnetization vector expanded in terms of the spherical basis vectors. The evolution of the magnetization during this experiment will go as follows: M = Meq e0 .
M0 (/2)x
i i M = Meq ey = Meq e1 + Meq e+1 2 2

My M 1 M+1
free evolution i i M(t) = Meq eit et/T2 e1 + Meq (1 et/T1 ) e0 + Meq eit et/T2 e+1 . 2 2
(t) M 1 (t) M0 (t) M+1
We obtain the signal detected in this experiment by plugging the M+1 (t) component into Eq (20.14).
Setting the receiver phase, ref , to zero, we obtain i S (t) = kMeq eit et/T2 = S (0)eit et/T2 . 2 (6.16)
The time dependent signal after a single /2 pulse is called a Bloch Decay, or more commonly known as the Free Induction Decay (FID). P. J. Grandinetti, October 31, 2011
30
/2
Real
time
Imaginary
6.3. THE BLOCH DECAY
31
Problems
1. Derive Eq. (18.6) starting from Eqs. (18.1)-(18.4). 2. Solve the Bloch equations for free evolution of the magnetization vector in the lab frame. dMx (t) = 0 My (t) Mx (t)/T2 , dt dMy (t) = 0 Mx (t) My (t)/T2 , dt dMz (t) = [Mz (t) Meq ]/T1 . dt
3. In this chapter we dened four frequencies: 0 , 1 , |rf |, rot , and two phases: and ref . Give the physical signicance of each of these variables, and describe the relationships, if any, that may exist among them. 4. Given the initial conditions M(0) = Meq ez solve the Bloch equations in a rotating frame (rot = rf ) for the free evolution of the magnetization vector under a magnetic eld of B(t) = 2B1 cos rf tex + B0 ez . Assume that |rf | = |0 | and that the relaxation can be ignored, i.e., 1/T1 = 1/T2 = 0. In which direction is the magnetization vector pointing at the times 1 t = /2 and 1 t = ?
6. Given the initial conditions M(0) = Meq ez solve the Bloch equations in the frame rotating for the free evolution of the magnetization vector under a magnetic eld of B(t) = 2B1 cos rf tex + B0 ez . Assume that rot = rf = 0 1 and that the relaxation can be ignored, i.e., 1/T1 = 1/T2 = 0. In which direction is the magnetization vector pointing at the times 21 t = /2 and 21 t = ?
7. Given the initial conditions M(0) = Meq ex solve the Bloch equations for the free evolution of the magnetization vector under an external magnetic eld of B = B0 ez , and calculate the time dependent signal induced into a coil in the x-z plane that is surrounding the time dependent magnetization vector.
32
Chapter 7
The Fourier Transform

The time domain free induction decay and the NMR spectrum are related by the Fourier transform. Most everyone is familiar with the idea behind this transform. For example, in music a sound is represented as a note on a sta representing frequencies. When you hear a note your ear is sensing oscillations in air density, and your brain recognizes what frequency it is.
amplitude
time
louder softer
*
frequency
The louder the note, the bigger the amplitude of the sound wave you hear. Also the higher the note, the higher the frequency of sound oscillations you hear. 33
34
CHAPTER 7. THE FOURIER TRANSFORM
amplitude
time
louder
softer
*
frequency
So we can view a musical note as either a time domain signal or frequency domain signal. Now, if you have a good musical ear, you might be able to listen to more than one note simultaneously and be able to distinguish more than one frequency.
amplitude
time
frequency
What Fourier realized was how to mathematically relate some oscillating signal (like sound) and transform it into a plot of amplitude versus frequency, that is
S ( ) =
S (t)eit dt.
(7.1)
We integrate S (t) over all time for each to get S ( ). S (t) and S ( ) are called the time and frequency domain signals, respectively. The inverse Fourier transform S (t) = 1 2
S ( )eit d
(7.2)
takes you from the frequency domain signal to the time domain signal. S ( ) and S (t) form a Fourier transform pair. Lets take the signal S (t) = cos t = as an example. P. J. Grandinetti, October 31, 2011 1 it e + eit 2 (7.3)
35
time
The Fourier transform of this signal is calculated according to
S ( ) =
S (t) eit dt =
1 2
eit eit dt +
1 2
eit eit dt,
and becomes S ( ) = where if x = 0 ( x) = 0 if x = 0 So the Fourier transform of S (t) = cos t looks like 1 1 ( ) + ( ), 2 2
frequency
i.e., a spike at and and zero everywhere else. In this context a negative frequency has little meaning. However, in the context of the rotating frame, we can connect this with the sense of circular motion of magnetization vectors. Heres an example of counter clockwise motion in the complex plane.
y r x -r r x r y -r r sin t r cos t
time
time
r exp( i t)
36
A point in the complex plane is dened z = x + iy or in polar coordinates z = rei . The Fourier transform of this complex signal is reit r ( )
FT
frequency
Likewise clockwise motion in the complex plane is described by

y r x -r r x r y -r -r sin t r cos t
time
time
}
+
r exp(-i t)
and the Fourier transform of this complex signal is reit r ( ).

FT
frequency
In a rotating frame those magnetization vectors precessing slower than the rotating frame will appear to rotate in the opposite direction of those precessing faster than the rotating frame. P. J. Grandinetti, October 31, 2011
7.1. ABSORPTION AND DISPERSION MODE LINESHAPES
37
7.1
Absorption and Dispersion Mode Lineshapes
Now lets look at something more realistic - more like NMR, i.e., a signal that starts at time t = 0 and decays exponentially with time. S (t) = eit et/T2
X Y
time
time
In this case the limits of our integral go from zero to +, not to +.
S ( )
=
0
eit et/T2 eit dt 1/T2 ( ) +i (1/T2 )2 + ( )2 (1/T2 )2 + ( )2
Our spectrum has a real and an imaginary part. The real part is called the Absorption mode spectrum and the imaginary part is called the Dispersion mode spectrum. A( ) = ; 2 + 2 Absorption Mode Lineshape (7.4) D( ) = 2 2 ; + Here = 1/T2 . Dispersion Mode Lineshape
Real 2/T2 Imaginary
Absorption Mode
Dispersion Mode
2/T2
38
As you can see from these equations, while the lineshapes associated with the real and imaginary parts of the NMR spectrum are dierent, they contain the same information. They are related by the KramersKronig relation A( ) = 1

D( ) d
and
D ( ) =
A( ) d .
(7.5)
The tails of the dispersion lineshape extend further out than the absorption lineshape. These lineshapes are characteristic of any damped oscillator problem. The names absorption and dispersion come from optical spectroscopies. Generally, the specic form of the absorption- and dispersion-mode lineshape functions in an NMR spectrum can vary depending on the system under study. Our particular example has an exponentially decaying signal, whose well-known Fourier transform results in a Lorentzian lineshape in the Absorption mode. For a Lorentzian lineshape the full width at half height, , is equal to 2/T2 . Another common Fourier transform pair is the Gaussian decaying signal, whose Fourier transform is a Gaussian lineshape in the Absorption mode. Gaussian lineshapes tend to arise when there is a distribution in resonance frequencies. That is, when the signal arises from many transitions, each having ever so slightly dierent transition frequencies, then the overall lineshape becomes Gaussian. This is often explained in terms of the Central Limit Theorem1 . Lorentzian lineshapes are observed in liquid state NMR spectra, whereas Gaussian lineshapes are more common in solid-state NMR spectra.
7.2
Phase Corrections
Whenever you have complex signals you always have the issue of the phase relationship between the real and imaginary parts. Recall the relationships z = x + iy = rei , where = tan1 y x and r= x2 + y 2 . (7.7) (7.6)
In our context of circular motion in the complex plane we have a vector rotating where r is the length of the vector and is the angle from the x axis.
1 This
theorem says that when you have many dierent contributions to the lineshape, each with their own characteristic
lineshape, then in the limit that you have a innite number of contributions the nal lineshape will be Gaussian, even if none of the individual lineshapes are Gaussian. An important condition is that the component have the same order of magnitude and that no single source dominates all the others.
7.2. PHASE CORRECTIONS

O
39
Lets look at how this picture is used in NMR, and how it relates to phase corrections of NMR spectra. Immediately after applying a /2 pulse along the y axis the magnetization previously along z will now be along the x axis, and begins precessing in the x-y plane of the rotating frame. If we plotted the tip of the net magnetization vector in the x-y plane as it evolved we would see
path of tip of magnetization vector as it precesses
x detector
time
y detector
time
7.2.1
Zeroth Order Phase Correction
What happens if the magnetization doesnt start out exactly lined up with the x-axis at t = 0, but rather is oset from the x-axis by an angle ? P. J. Grandinetti, October 31, 2011
40
x detector
time
y detector
time
The complex Fourier transform of this signal may look like
Real
Imaginary
and neither of these look like absorption or dispersion mode signals. In this case we can calculate what is in the real and imaginary parts. It is given by S ( ) = [A( ) + iD( )]ei which can be expanded out to S ( ) = [A( ) cos D( ) sin ] + i[A( ) sin + D( ) cos ] . real part imaginary part (7.9) (7.8)
One can show that when = /2 (i.e., magnetization starts precessing from along the y axis), then the Fourier transform of the resulting signal will be S ( ) = D( ) + iA( ). P. J. Grandinetti, October 31, 2011 (7.10)
7.2. PHASE CORRECTIONS
41
Since the conventional in NMR is to report only the absorption mode spectrum it is standard procedure to apply a phase correction to the spectrum (or d) so that the real part of the spectrum contain only the absorption mode spectrum. The spectrum is corrected by the simple application of a zeroth order phase correction, S ( ) = S ( )ei0 , (7.11)
where 0 is adjusted until 0 = and the spectrum contains pure absorption mode lineshapes in the real part.
7.2.2
First Order Phase Correction
Now lets consider when we have two or more resonances present in the spectrum. In a given time the magnetization for site 2 will have rotated further around than the magnetization for site 1. For example, think of two athletes running around a track. At the starting line they are in-phase with each other. After they start running, the faster runner will be further along the track than the slower runner so there will be a phase dierence. In principle, this is no problem since they both start out at the same point and the Fourier transform gives pure absorption mode spectra for both sites in the real part and pure dispersion mode spectra for both sites in the imaginary part.
2 1
X Real Imaginary
1 2
y
Even if our detector is 45 away from the starting point then we phase correct the spectrum to get the pure absorption mode in real and pure dispersion mode in the imaginary part. The problem comes in if youre late and miss the starting point. That is, you turn your detector on at the some time t0 after the starting point. Then what you may see is . . . P. J. Grandinetti, October 31, 2011
42
1 1
X Real
Imaginary
y
Site 2 had precessed signicantly ahead of site 1 before the receiver was turned on. Now the phase we need to make site 1 have a pure absorption mode spectrum in the real part is not the same as the phase needed for site 2. The phase correction we need can be calculated from the frequency of each site. Thus, we dene our phase correction as linearly dependent on , ( ) total phase = 0 0th order + t0 1st order (7.12)
where t0 is the time we were late in starting the detector. Each frequency in the spectrum gets a dierent phase correction. Thus, we apply S ( ) = S ( )ei0 eit0 , (7.13)
to obtain a spectrum with both sites in pure absorption mode in the real part of the spectrum. Applied to our example, we obtain
1 2
Real Imaginary
Actually, you get pure absorption mode lineshape in the real part plus a sinc-type function distortion in the baseline. Remember that the time shift theorem says that a time shift can be eected by a phase P. J. Grandinetti, October 31, 2011
7.2. PHASE CORRECTIONS shift in the frequency domain. S (t t0 ) S ( )eit0

F.T.
43
(7.14)
So, in eect, by applying the phase shift to get pure absorption mode spectrum you are shifting the time domain signal so that the time origin is where it should have been, which we can see from the convolution theorem is like multiplying a correctly acquired (i.e., acquisition started at t = 0) signal by a step function.
S1(t)
F. T.
S 1 ( )
(Convolution)
(Multiplication)
S2(t)
1 0
F. T.
S 2 ( )
=
F. T.
ST()
ST(t)
Heres an algorithm for phasing a spectrum containing multiple resonances. Take, for example, the unphased spectrum below.
Apply zeroth order phase correction until one peak is completely absorption mode lineshape. P. J. Grandinetti, October 31, 2011
44
one peak "phased"
Since this peak (at 1 ) should look the same when the whole spectrum is phased, then we know any further phase corrections should not aect this peak. To force this constraint we set (1 ) = 0 + 1 t0 = 0. With this equation we can solve for 0 obtaining 0 = 1 t0 , and substituting this expression back into Eq. (7.12) we obtain ( ) = ( 1 )t0 . (7.15)
Now the only adjustable variable is t0 , and we can vary it until all the peaks are pure absorption mode.
In this approach, 1 is sometimes called the pivot frequency.
Chapter 8
Measuring Chemical Exchange - The Modied Bloch Equations

During a chemical reaction, nuclei can move between chemically nonequivalent environments. Depending on the reaction rate, these processes will manifest themselves dierently in the NMR lineshape [1]. In this section, we will examine the simplest example, a single spin 1/2 nucleus experiencing no other spin couplings, and undergoing chemical exchange between two chemically nonequivalent sites, that is, k1 A k1 In this situation we write the two coupled Bloch equations dMA = (1 A )MA (t) B(t) [R]{MA (t) MA,eq } k1 MA (t) + k1 MB (t), dt dMB = (1 B )MB (t) B(t) [R]{MB (t) MB,eq } + k1 MA (t) k1 MB (t). dt We will further simplify our treatment by neglecting the eects of spin relaxation, assuming equal populations of A and B, and that k = k1 = k1 . Solving this set of linear, rst-order dierential equations, one obtains S ( ) = k (A B )2 , ( A )2 ( B )2 + 4k 2 2 (8.1) B.
where = (B + A )/2, as analytical expression for the absorption mode spectrum. In Fig. 8.1 are simulations of the predicted NMR lineshape in a simple two site exchange, based on Eq. (8.1). In the 45
46CHAPTER 8. MEASURING CHEMICAL EXCHANGE - THE MODIFIED BLOCH EQUATIONS slow exchange limit (low k ), the individual resonances for each site are resolved with linewidths, if we include spin-spin relaxation, of 1/T2 + k . As the rate of exchange increases the individual lines broaden and coalesce together. In the fast exchange limit the two lines collapse into a single line appearing at the
2 2 average frequency of the two sites and has a linewidth of 1/T2 + 1 2 /k . The decrease in sensitivity in
the intermediate regime is more apparent in the spectra on the right of Fig. 8.1, where the lineshapes are normalized to have the same area, a situation that more accurately reects the experimental sensitivity variations Given that rate constants often follow the Arrhenius equation k = A exp(Ea /RT ), (8.2)
then the evolution of lineshapes with increasing temperature would parallel those in Fig. 8.1 with increasing k . A full lineshape analysis as a function of temperature provides a simple way to measure the activation energy, Ea , for the exchange process.
47
k=10000
k=3200
k=400
k=200
k=100
k=50 k=10
k=2
-400 -200 0 200 400 -400 -200 0 200 400
Figure 8.1: NMR lineshape for a two site system (A = B = 200 Hz) undergoing chemical exchange as a function of rate constant k in s1 . The intensity of the spectra on the left are scaled to the same amplitude to more clearly illustrate the lineshape transformation. The decrease in sensitivity in the intermediate regime is more apparent in the spectra on the right where the lineshapes are normalized to have the same area; a situation that more accurately reects the experimental sensitivity variations. P. J. Grandinetti, October 31, 2011
48CHAPTER 8. MEASURING CHEMICAL EXCHANGE - THE MODIFIED BLOCH EQUATIONS
Chapter 9
Inhomogeneous External Magnetic

Fields and T2
So far we have assumed that the external magnetic eld was the same for all spins throughout the sample. In practice, however, there will be some spatial variations in the external magnetic eld strength across the sample. Lets consider the eect of an inhomogeneous external magnetic eld on our NMR spectrum. A spatial variation in the external magnetic eld makes the precession frequency depend on position in sample, that is 0 (r) = (1 )B0 (r). The total signal in the rotating frame becomes an integral over the volume of the sample, Stotal (t) = (r)ei(r)t et/T2 dV,
where (r) is the density of nuclei at r. Even without knowing the functional form of the spatial variation in magnetic eld and thus frequency, it is easily concluded that the decay of the total signal will be faster in an inhomogeneous eld compared to a homogeneous eld. For example, if there is a distribution of frequency clearly at long times the individual signal oscillations will be out of phase and thus will be destructively interfering at long times. Whereas at short times the oscillations from dierent signal will still be relatively in phase and will add up more constructively than at later times. The Fourier transform of a time domain signal from a sample in an inhomogeneous magnetic eld will appear wider than the signal of a sample in a homogeneous eld. Without knowing the functional form for the spatial eld variations, it is impossible to predict the absorption mode lineshapes. Thus in the presence of an 49
50
CHAPTER 9. INHOMOGENEOUS EXTERNAL MAGNETIC FIELDS AND T2
inhomogeneous magnetic eld the full width at half height, , can not longer be simply related to T2 , i.e., = 2/T2 .
2/T2
Often times, NMR spectroscopists will measure the full width at half height of an inhomogeneous broad ened lineshape and use this number to report a quantity called T2 , dened as T2 = 2/. One should not conclude, however, that a reported T2 means that lineshape is Lorentzian.
Chapter 10
Measuring Relaxation Times

Since NMR relaxation arises from the uctuating elds produced by molecular motion, it is possible to use NMR relaxation times to characterize and quantify this motion. In this section we will examine two common approaches for measuring the spin-lattice relaxation time, T1 , introduced with the phenomenological Bloch Equations. It is important to note, however, that the Bloch equations assumption of a single exponential recovery time is not valid for nuclei that have any time independent couplings, as described in the last section. That is, the Bloch Equations assume that all couplings, such as the quadrupolar, dipolar, and J coupling, are averaged to zero by molecular motion, or selective rf irradiation (i.e., spin decoupling). If this is not true, multi-exponential recoveries can occur and a single exponential recovery time, i.e., T1 , would no longer be appropriate. The problem is still tractable, but a more detailed analysis of the data is required, often combined with other experiments, and will not be described in this section. If your main goal is to obtain a proper value for the equilibration time so that the signal-to-noise ratio is acceptable and the integrated intensities are quantitative then it is not a problem if your recovery is not exponential. Performing the saturation or inversion recovery experiment described below is sucient to determine the equilibration time needed for all the magnetization to recover, i.e., 99% of Meq .
10.1
Spin-Lattice Relaxation Times
In this section, we will assume the nuclei relax with single exponential recoveries, and describe two experiments for measuring the spin-lattice relaxation time T1 : the saturation recovery method and the inversion recovery method. 51
52
CHAPTER 10. MEASURING RELAXATION TIMES
10.1.1
The Saturation Recovery Experiment
In the saturation recovery experiment the magnetization is rotated from z to the x-y plane and allowed to completely dephase during a period that is set to approximately 2 or 3 times T2 . This process is repeated many times so there exists no net magnetization in any direction, and the magnetization is saturated. During the time t1 the magnetization recovers, growing along the z -axis. This recovery is monitored by applying a /2 pulse after t1 and measuring the signal intensity.
( )
/2
/2
t1
t2
n
repeat sequence inside parantheses n times.
From the Bloch equations we know the z -component of the magnetization will recover according to Mz (t) = Meq (1 et1 /T1 ) + Mz (0)et1 /T1 . Inserting the initial condition created by the saturation pulse train, Mz (0) = 0, into this expression we obtain Mz (t1 ) = Meq (1 et1 /T1 ). Below is a plot of the percent recovery of the longitudinal (Mz ) magnetization as a function of t1 given in multiples of T1 .
100% 80% 60%
Meq
Mz(t1)
40% 20% 0% 0 2 4 6 8 10
t1/T1
10.1. SPIN-LATTICE RELAXATION TIMES
53
At t1 = 0 we have Mz (0) = 0 and at t1 = we have Mz () = Meq . The saturation recovery experiment provides a less precise measurement of T1 than the inversion recovery experiment, described in the next section. Unlike the inversion recovery experiment, however, the saturation recovery experiment doesnt require an equilibration time, and for samples with long T1 values (greater than 30 seconds) this can lead to signicant shorter experiment times.
10.1.2
The Inversion Recovery Experiment
In the inversion recovery experiment the magnetization is rotated from the +z axis to the z axis. The magnetization recovers and grow back towards the +z axis during a t1 period. The recovery of the magnetization is monitored by applying a /2 pulse after t1 and measuring the signal intensity.
equilibration time
/2
t1
t2
Again, using the Bloch equations we have Mz (t) = Meq (1 et1 /T1 ) + Mz (0)et1 /T1 . Here we have Mz (0) = Meq so Mz (t) = Meq (1 2et1 /T1 ). Below is a plot of the percent recovery of the longitudinal (Mz ) magnetization as a function of t1 given in multiples of T1 .
100% 50%
Meq
Mz(t1)
0% 2T1 -50% 4T1 6T1 8T1 10T1
-100%
At t1 = 5T1 then Mz (5T1 ) = 0.993Meq . The inversion recovery experiment has a wider range of signal variation that makes it more precise. A quick approach for estimating T1 is to locate the time t1 = P. J. Grandinetti, October 31, 2011
54
when Mz ( ) = 0. One can then show that T1 = / ln 2. The disadvantage of the Inversion Recovery sequence is that you need an equilibration time of a least ve times the longest T1 . Since you dont know T1 then you have to guess, and this makes the process somewhat iterative.
10.2
Spin-Spin Relaxation Times
In theory one can obtain T2 by taking half the inverse of the full width at half height in Hertz of a resonance in an NMR spectrum. Unfortunately, the line widths of resonances in NMR are often dominated by the inhomogeneities in the magnetic eld rather than T2 .
10.2.1
The Spin Echo Experiment
It turns out that in many systems it is possible to measure the true T2 even in the presence of inhomogeneous broadening. One approach is to apply a second rf pulse before detecting the signal. Consider the following sequence
equilibrate (/2)x t1 ( )x t2
How is the signal detected during t2 aected by this two pulse sequence? Lets examine the path of the magnetization vector under this sequence of events. To simplify our discussion, we will assume that t1 + t2 T1 , and neglect spin-lattice relaxation. The evolution of the magnetization during this
10.2. SPIN-SPIN RELAXATION TIMES experiment will go as follows:
55
M = Meq e0 .
M0 (/2)x
i i M = Meq e+1 + Meq e1 2 2

M+1 t1 evolution
M 1
i i M(t1 ) = Meq ei(r)t1 et1 /T2 e+1 + Meq ei(r)t1 et1 /T2 e1 2 2
M+1 (t1 ) M1 (t1 )
( )x i i M(t1 ) = Meq ei(r)t1 et1 /T2 e+1 + Meq ei(r)t1 et1 /T2 e1 . 2 2
M+1 (t1 ) M1 (t1 )
t2 evolution i i M(t1 , t2 ) = Meq ei(r)(t1 t2 ) e(t1 +t2 )/T2 e+1 + Meq ei(r)(t1 t2 ) e(t1 +t2 )/T2 e1 . 2 2
M+1 (t1 ,t2 ) M1 (t1 ,t2 )
Notice that the eect of the ( )x pulse is to swap the M+1 and M1 coherences. Focusing on the transverse magnetization we see that when t1 = t2 = we get the remarkable result i i M( ) = Meq e2 /T2 e+1 + Meq e2 /T2 e1 . 2 2
M+1 ( ) M1 ( )
When both t1 and t2 are increased while keeping t1 = t2 , the evolution of the signal will be independent of (r). In this two pulse experiment on samples with large inhomogeneous line broadenings the signal can even appear dead after the /2 pulse but after a pulse the signal returns in what is called a spin echo1 .
1 Sometimes
called the Hahn spin echo, after its discover Erwin Hahn.
56
/2
time
t1 t1 = t2
That is, independent of the frequencies (r) present, the magnetization vectors for all sites return to the starting point (the y axis) when t1 = t2 . The only decay of the echo top (refocussed magnetization) is due to T2 relaxation. This sequence gives us a method of measuring T2 in the presence of eld inhomogeneities. The spin echo arises in many other contexts in NMR, and will be a key concept in understanding numerous multi-dimensional NMR experiments. A popular analogy for the spin echo experiment is to consider a group of runners lined up at the starting line of a circular track. When the whistle is blown (the /2 pulse), the runners begin running at various speeds depending on their abilities. After a time t1 , the whistle is blown again (the pulse) and the runners all stop and run in the opposite direction. At a time t2 = t1 after the second whistle the runners, if they ran at the same speed in both directions, will all pass the starting line simultaneously as they continue running in the opposite direction. This lining up of the runners as they pass the starting line is analogous to what happens to the magnetization in the spin echo experiment. There are two possible complications with measuring T2 values with the spin echo experiment. The rst occurs for nuclei experiencing homonuclear J -couplings. Here, the echo tops are modulated as a function of t1 due to evolution under the J -coupling. This evolution cannot be refocused by the pulse. The second complication arises from molecular diusion. If a molecule diuse from one region of the sample to the next in a eld gradient, then it may have dierent resonance frequencies during t1 and t2 . Here there will be an incomplete refocusing of the echo.
t2
10.2.2
Carr-Purcell Meiboom-Gill sequence
Carr and Purcell proposed a multiple echo sequence to minimize the eects of translational diusion when measuring T2 , which, as modied by Meiboom and Gill and shown below, is commonly called the P. J. Grandinetti, October 31, 2011
10.2. SPIN-SPIN RELAXATION TIMES CPMG sequence.

(/2)x y y y y y y
57
time
58
Chapter 11
Measuring Translational Diusion Coecients

The dependence of the echo intensity on molecular diusion can be exploited as a means to measure translational diusion coecients [2, 3]. By applying linear magnetic eld gradient, B(r) = B0 + G r, across the sample and performing the spin echo experiment one can solve the modied Bloch equation, dM(r) = M(r) B(r, t) [R]{M(r) Meq } + D2 M(r). dt to obtain an analytical expression for the echo intensity, given by 2 S (, G) = S (0, 0) exp D 2 G2 3 2 /T2 . 3 (11.2) (11.1)
By applying a systematic variation in magnetic eld gradient strength, G, as well as the echo dephasing time, , one can measure both the diusion coecient, D, and spin-spin relaxation time, T2 .
/2
rf
Gradient
59
60
CHAPTER 11. MEASURING TRANSLATIONAL DIFFUSION COEFFICIENTS
11.1
Pulsed Field Gradients
A more robust method for measuring diusion coecients employs pulsed eld gradients [4] instead of a static eld gradient. In this approach, as illustrated below, the gradient is pulsed on before and after the pulse, each time for a duration of .
/2
rf
Gradient
The analytical expression for the echo intensity is given by S (, G) = S (0, 0) exp 2 G2 2 D( /3) 2 /T2 , (11.3)
where is the time between the two gradient pulses. Since the detection of the echo signal occurs in a relatively homogeneous magnetic eld a high resolution spectrum with higher sensitivity is obtained. This is particularly advantageous for making simultaneous measurements of diusion coecients in samples containing mixtures of molecules. Additionally, the experimenter has the additional option of varying the gradient pulse duration, , gradient pulse interval, , or gradient strength, G, when measuring the diusion coecient. A plot of log[S (2 )/S (0)] versus 2 ( /3)G2 should yield a straight line passing through the origin with a slope of 0.4343 2 D. With pulsed eld gradients, values of D as low as 109 cm2 sec1 can be measured. This is about two orders of magnitude slower than D values measurable with static eld gradients. It is also possible, with this sequence, to measure ow: a coherent motion of all molecules with a uniform velocity vector v. The eect of ow is to cause a phase modulation of the echo tops. Combined with the attenuation of the echo tops due to diusion the analytical expression for the echo intensity is given by S (, G) = S (0) exp i G v (G )2 D( /3) 2 /T2 . (11.4)
Finally, recall that the CPMG sequence can be used to minimize the eects of translational diusion (and ow) when measuring T2 . In the presence of a gradient the CPMG echo intensities will be given P. J. Grandinetti, October 31, 2011
11.1. PULSED FIELD GRADIENTS by 1 S (t = 2n ) = S (0) exp 2 G2 2 Dt t/T2 . 3
61
(11.5)
By reducing the time, , between pulses, the eect of the diusion can be minimized, and the T2 can be measured more reliably.
62
CHAPTER 11. MEASURING TRANSLATIONAL DIFFUSION COEFFICIENTS
Chapter 12
Coherence Transfer Pathways

Consider the two pulse NMR experiment shown below.
t1
t2
Is this the Spin Echo experiment of section 10.2, or the T1 inversion recovery experiment of section 10.1.2? You might say it depends on the rotation angle used for the two pulses. That is, the spin echo experiment uses a /2 and a for the rst and second pulse, respectively, whereas, the T1 inversion recovery experiment uses a and a /2, respectively. In a sense, that is correct. But what if both pulses were 2/3 rotations? Would we observe a spin echo signal, an inversion recovery signal, or both? The answer is that we would observe both. This illustrates a common problem in multiple pulse NMR spectroscopy. There can be many dierent types of signals associated with the same pulse sequence, some desired and some undesired. To separate desired from undesired signals it is rst necessary to understand how dierent signals originate from the same pulse sequence. Consider the spin echo experiment. If we followed the path of the magnetization vector component responsible for the spin echo signal through the sequence we nd that it starts out along the e0 direction, moves to the e1 direction during t1 , and then to the e+1 direction during t2 where it is detected. This pathway is dierent from the one associated with the inversion recovery pathway, where the detected magnetization vector component follows the path e0 e0 e+1 . Note that the magnetization vector components only change direction during a pulse, and, as we saw in Eq. (20.2), during free evolution their directions remain unchanged. 63
64
CHAPTER 12. COHERENCE TRANSFER PATHWAYS If we wrote out all magnetization vector pathways possible in the two pulse experiment for arbitrary
pulse lengths the resulting expansion would quickly become complicated. As you can see below, with each pulse the number of terms in the expansion grows exponentially.
M/Meq = e0
M(t1)/Meq = a0,- e-1
(1)
a0,0 e0
(1)
a0,+ e+1
(1)
t1
2
t2
M(t2)/Meq = a0,- a-,- e-1 + a0,- a-,0 e0 + a0,- a-,+e+1 + a0,0 a0,- e-1 + a0,0 a0,0 e0 + a0,0 a0,+e+1 + a0,+ a+,- e-1 + a0,+ a+,0 e0 + a0,+ a+,+e+1
(1)
(2)
(1)
(2)
(1)
(2)
(1)
(2)
(1)
(2)
(1)
(2)
(1)
(2)
(1)
(2)
(1)
(2)
After two pulses there will be 9 dierent terms contributing to the total magnetization, each multiplied by coecients that carry the history of each term. Tracing the origins of each of these terms back to the initial magnetization along e0 also reveals all the possible transition frequency modulations that can be present in a signal. These possibilities can be graphically represented in what is called a coherence transfer pathway.
(1)1
(2)2
t1
-1 0 +1
t2
Start along e0
2nd pulse changes coherence order. Shown here is one possible change in Pulse changes coherence order. coherence order from e-1 e+1 Shown here is one possibility where the coherence order goes from e0 e-1
This pathway is associated with the underlined terms in our two magnetization vector expansion. For the two pulse sequence on a spin 1/2 there are eight other such pathways. Below are examples of three others. P. J. Grandinetti, October 31, 2011
65
(1)1 (2)2
t1
-1 q 0 +1
t2
or
-1 q 0 +1
or
-1 q 0 +1
It is left as an exercise for the reader to draw the remaining pathways. The signals associated with each of the nine dierent pathways are not all unique. A pathway can give rise to a signal that is the complex conjugate of another. These complex conjugate pathways are easily identied since they will be mirror images of each other about the q = 0 coherence level. For example, the two pathways below are mirror images of each other about the q = 0 coherence level and therefore represent signals that will be complex conjugates of each other.
(1)1
(2)2
t1
t2
-1 q 0 +1
and
-1 q 0 +1
If you remove the redundancy of mirror image pathways, then the number of pathways that give rise to unique signals in a two pulse sequence on a spin 1/2 system is ve. If we consider only those that end on the q = +1 coherence level, that is, assume that our receiver is a perfect detector of the observable M+ , then we only need to consider three pathways. Before we consider these three pathways, we note that, for reasons related to the quantum mechanical treatment of NMR the convention in NMR is to draw the mirror image (or complex conjugate) pathway with the coherence levels labeled as p, where p = q . With that in mind, lets now consider the three pathways below. P. J. Grandinetti, October 31, 2011
66
CHAPTER 12. COHERENCE TRANSFER PATHWAYS

(1)1 (2)2
t1
+1
t2
p 0
-1 +1
p 0
-1 +1
p 0
-1
Each of these pathways describe a dierent NMR signal. For example, the Spin Echo experiment is described by pathway 1. Assuming that the absolute frequency associated with the p = 1 coherence levels doesnt change between t1 and t2 , then the change in the sign of the coherence level from p = 1 in t1 to p = +1 in t2 will lead to an echo formation during t2 when t2 = t1 . Similarly, the inversionrecovery T1 measurement experiment is described by pathway 2. The Anti-Echo experiment is described by pathway 3. It has this name because when t1 = 0 the Anti-Echo pathway signal in t2 , like the Echo pathway signal, contains only the right half of a spin echo signal, and with increasing t1 the anti-echo signal maximum moves backwards in t2 , occurring virtually at the negative times t2 = t1 , where it cannot be detected. Lets consider the pathways for the two other experiments weve discussed so far, the saturation recovery and the CPMG sequences. In the saturation recovery experiment we have
( )
+1 p 0 -1
/2
/2
t1
t2
During the train of n pulses we intend to saturate the magnetization, and desire no coherence transfer at all. Thus we write the pathway during the pulse train to reect no change in the coherence order1 During the recovery time, t1 , there will be recovery along the z axis, which we will measure with the
1 Technically,
one might argue that there is no magnetization along e0 direction after the pulse n pulses, when t1 = 0,
since the magnetization is saturated. That is correct. For practical purposes, however, we group this saturated state with the e0 magnetization components.
67 nal /2 pulse. The goal of the nal /2 pulse is to convert the M0 (p = 0) order into an M+ (p = 1) coherence that we can detect with our receiver coil. In the CPMG experiment there will be two pathways from which we will acquire signal. The signal acquired from sequential echo tops will alternate between the two pathways shown below.
(/2)x y y y y y y
+1 p 0 -1
time
Because the two pathways are mirror images their signals will be complex conjugates of each other. Although signal from a given pathway will be missing every other point, the signal from a single pathway with all points can be reconstructed by combining the signal of one pathway with the complex conjugate of the other pathway. An ensemble of coupled spin 1/2 nuclei, or an ensemble of uncoupled nuclei with spin I > 1/2, will have additional degrees of freedom that can be manipulated in the NMR experiment. These additional degrees of freedom are represented in coherence transfer pathways as coherence orders beyond p = 1. For example, a single spin I will have coherence orders extending over 2I p 2I . Generally, N coupled spins will have coherence orders extending over 2L p 2L, where L =
k Ik .
68
CHAPTER 12. COHERENCE TRANSFER PATHWAYS
Chapter 13
Limitations of the Bloch Equations

We have seen that a nucleus with a non-zero angular momentum possesses a magnetic dipole moment vector. For the 1 H isotope, which only contains a single proton of spin I = 1/2, the nucleus only has the three internal degrees of freedom associated with the three components of its magnetic dipole moment vector. The Bloch equations were designed to describe an ensemble of such spin 1/2 nuclei, each having only three internal degrees of freedom.
13.1
Nuclei with Electric Quadrupole Moments
In heavier isotopes and elements, the nucleus of the atom will contain additional protons and neutrons, and with them a corresponding increase in the available degrees of freedom for the nucleus. Although neutrons have no charge, they do have a spin angular momentum and a magnetic moment. With an increase in the number of nucleons we will nd additional degrees of freedom within the nucleus that can be manipulated in the NMR experiment. These additional degrees of freedom can be described in terms of higher order magnetic and electric multipole moments. Due to symmetry restrictions only certain magnetic and electric multiple moments are allowed as a function of total angular momentum. The allowed moments up to l = 4 (hexadecapole) are summarized below for isotopes with total angular momentum up to I = 2. 69
70
CHAPTER 13. LIMITATIONS OF THE BLOCH EQUATIONS
Nuclear Spin I=0 I=

1 2
l=0 monopole electric electric electric electric electric
l=1 dipole 0 magnetic magnetic magnetic magnetic
l=2 quadrupole 0 0 electric electric electric
l=3 octapole 0 0 0 magnetic magnetic
l=4 hexadecapole 0 0 0 0 electric
I=1 I=
3 2
I=2
Every multipole moment, l, will bring 2l + 1 additional degrees of freedom to the nucleus in the NMR experiment. The Bloch equations, designed to describe an ensemble of uncoupled spin 1/2 nuclei, are inadequate for uncoupled nuclei with spin I > 1/2, that is, with more than three degrees of freedom. It is possible to modify the Bloch Equations to describe this higher dimensional motion, but this motion is just as easily described using the well-established machinery of quantum mechanics. In later chapters, we will look in detail at how quantum mechanics describe the NMR experiment. It is also useful to note that just as the nuclear magnetic dipole moment interacts with external magnetic elds, the nuclear electric quadrupole moment interacts with external electric elds, or more specically, surrounding electric eld gradients. These electric eld gradients are generated by orbiting electrons as well as neighboring nuclei. Additionally, just as the vector elements of the nuclear magnetic dipole moment precess in an externally applied magnetic eld, the tensor elements of the nuclear electric quadrupole moment will precess in an external electric eld gradient1 . Measuring the frequencies of these oscillations allows us to measure the electric eld gradient surrounding a nucleus, which in turn provides information about the local geometry and bonding around that atom.
1 One
might wonder if it is possible to directly detect the oscillations in the electric quadrupole moment tensor elements
with a quadrupole condensor and measure the distribution of electric eld gradients experienced by the nuclei in the sample. Although such experiments have been suggested [5], the direct measurement of the oscillations of the nuclear electric quadrupole moment are much less sensitive than the indirect measurement of their oscillations with NMR through their eect on the magnetic dipole vector oscillations.
13.2. COUPLING BETWEEN DISTINGUISHABLE NUCLEI

Vzz
71
Magnetic dipole moment causes atomic nucleus to precess in a magnetic field
Electric quadrupole moment causes atomic nucleus to precess in an electric field gradient
Vyy Vxx
In principle, there will be couplings between the higher nuclear multipole moments and surrounding electric and magnetic elds, but in practice, attempts to measure such couplings have been inconclusive [6].
13.2
Coupling between distinguishable nuclei
So far in our discussion we have only considered isolated nuclei interacting with external magnetic and electric elds. In samples containing nuclei with abundant NMR active isotopes there will also be interactions between nuclear magnetic moments.
13.2.1
Magnetic Dipole Coupling between Nuclei
The classical interaction energy between two magnetic moments 1 and 2 is given by 1 2 3(1 r)(2 r) , r3 r5
E=
and where the r is the internuclear vector, and 1 and 2 are the vectors describing the direction of each magnetic moment. P. J. Grandinetti, October 31, 2011
72
In the absence of any external magnetic elds the lowest energy

1 +1 2, 2
1 1 1 +2 , +1 2 and 2 , 2
and highest total
and
1 1 2, +2
congurations are shown below.

2 2 1 2 1
Lowest Energy configurations
Highest Energy configurations

1 +2 , +1 2 , and highest:
In the presence of an innitely large external magnetic eld the lowest:

1 1 2 , 2 , total energy congurations are shown below.
B0
2 1 1 2
Lowest Energy configuration
Highest Energy configurations
13.2.2
Indirect J Coupling between Nuclei
In addition to the direct dipolar coupling between nuclei there will also be an indirect J coupling between nuclei. The mechanism behind J coupling is more theoretically involved than the dipolar coupling. Qualitatively, we can think of J coupling as a two step process. First, the magnetic moments of valence electrons orbiting nucleus A get polarized in the same direction as the A nucleus magnetic dipole moment. The polarized valence electrons then move to nucleus B, producing a small magnetic eld that shifts the resonance frequency of nucleus B. The same process occurs in reverse, and in this manner, A and B become indirectly coupled. P. J. Grandinetti, October 31, 2011
13.2. COUPLING BETWEEN DISTINGUISHABLE NUCLEI
73
e-
In either case, the total magnetic eld experienced by a nucleus will vary depending on the spin state of neighboring nuclei. In such a situation we again nd that the Bloch equations are inadequate for coupled nuclei because they have more than three degrees of freedom. Additionally, the direct dipoledipole and the indirect J coupling depend on the orientation of the internuclear vector and bonding orbitals, respectively. In the liquid and gas state this orientation dependence is averaged away by rapid molecular reorientations. Direct dipole-dipole couplings average to zero in liquids, whereas J couplings average to a non-zero scalar value.
74
Chapter 14
Interpreting Relaxation Times

Constructing a theoretical model to interprete NMR relaxation times can be relatively simple or completely intractable, depending on the sample. The general approach starts with determining which uctuating couplings contribute to the relaxation process. For spin 1/2 nuclei these couplings typically arise from magnetic dipole interactions, whereas for spin I > 1/2 they can arise from electric quadrupole as well as magnetic dipole interactions. For each of these possibilities we can obtain an expression for the relaxation times, and the overall spin relaxation rates will be the sum from all the contributions, 1 = T1 1 T1 +
paramagnetic
1 T1
+
quadrupole
1 T1
+
dipole
1 T1
+ .
chem. shift
(14.1)
Like resistors in parallel, we see that relaxation can easily be dominated by one coupling that is larger than the rest. Typically, the relaxation is dominated by uctuating unpaired electron-nuclear dipolar couplings, followed by quadrupolar couplings, dipolar couplings, and chemical shift anisotropy. Also important is whether the uctuating couplings experienced by a nucleus are coming from other nuclei or electrons in the same molecule (intramolecular couplings) or other molecules (intermolecular couplings). Chemical shift and electric quadrupole couplings are generally intramolecular, whereas dipolar couplings can be intra- or inter-molecular.
14.1
Time Correlation and Spectral Density Functions
Fluctuations in couplings arise from uctuations in the distances and orientations of atoms and molecules nearby the nucleus in question. Thus, we introduce the concept of a time correlation function to quantify the time scale on which these uctuations occur. For example, the time correlation function for an 75
76
CHAPTER 14. INTERPRETING RELAXATION TIMES
intermolecular distance vector can be written G( ) = r(t) r(t + ) , where the angle brackets represent an ensemble average. In the same manner one can imagine a time correlation function for a nuclear spin coupling, C , that depends on the same uctuating internuclear vector, GC ( ) = C (r(t)) C (r(t + )) , which we more simply write as GC ( ) = C (t) C (t + ) . We can also write the time-correlation function of two dierent couplings, GCD ( ) = C (t) D(t + ) . This is called a cross-correlation function, whereas GC ( ) is called an auto-correlation function. The spectral density function is the Fourier transform of the time correlation function,
J ( ) =
0
G( )ei d
J()
slow motion
intermediate motion
fast motion
0
One of the simplest models is that of random isotropic molecular tumbling where G( ) = G(0)e| |/c , where c is the reorientational correlation time for the molecular tumbling. The Stokes-Einstein-Debye relation predicts the reorientational correlation time as c = P. J. Grandinetti, October 31, 2011 s V , kT (14.2)
14.2. RELAXATION VIA DIPOLAR COUPLINGS
77
where s is the solvent viscosity and V the solvated volume of the rotating molecule. If we assume that the Stokes-Einstein-Debye relation holds, then the NMR relaxation times can be used to obtain information about molecular size or local viscosity. This equations also provides an understanding of why it is dicult to obtain high resolution solution-state NMR studies of molecules with large molecular weights. With increasing molecular size, there will be a corresponding increase in molecular reorientation times which in turn leads to decreased T2 , and excessive line broadenings.
14.2
Relaxation via Dipolar Couplings
To describe time correlation functions for translational and reorientation motion it is convenient to use the three functions F0 (r, , ) = b (1 3 cos2 ), 3 F1 (r, , ) = b sin cos exp(i), 2 3 F2 (r, , ) = b sin2 exp(i2), 4 where b= 0 I S . 4r3 (14.4)
(14.3)
Using these denitions we further dene the time correlation and spectral density functions,
Gn ( ) = Fn (t) Fn (t + ) ,
and
Jn ( ) =
0
Gn ( )ei d.
(14.5)
One can further show that the spectral density functions with dierent n are all related to a single function j ( ), called the reduced spectral density function, according to b2 j ( ) = 15 15 15 J0 ( ) = J1 ( ) = J2 ( ). 12 2 8 (14.6)
If we assume the spectral density function arises from a simple model of isotropic molecular tumbling, having a single reorientational correlation time c , then j ( ) is given by j ( ) = In the limit of extreme narrowing, 0 2c . 2) (1 + 2 c (14.7)
1, we obtain j ( ) 2c , (14.8) P. J. Grandinetti, October 31, 2011
78
14.2.1
Nuclei with Identical Resonance Frequencies
When there is a dipolar coupling between two nuclei with the same angular momentum, I , and resonance frequency, 0 , then the spin lattice relaxation arising from the modulation of the dipolar coupling is given by d(Mz1 + Mz2 ) 1 eq eq = (Mz1 Mz ) + (Mz2 Mz ) , 1 2 dt T1 where 1 I (I + 1) b2 = [3j (0 ) + 6j (20 )] . T1 15 Similarly, for the transverse component, we have M+11 + M+12 d(M+11 + M+12 ) = , dt T2 where 1 = I (I + 1) b2 T2 1 3 3 j (0) + j (0 ) + j (20 ) . 10 2 15 (14.12) (14.11) (14.10) (14.9)
Finally, note that in the Motional Narrowing Regime we obtain 1 1 I (I + 1) b2 2 c . T1 T2 (14.13)
14.2.2
Nuclei with Dierent Resonance Frequencies
When there is a magnetic dipolar coupling between two nuclei having dierent resonance frequencies, for example, a nucleus of spin I and a nucleus of spin S , the longitudinal recovery of the magnetization for I and S nuclei follows the coupled dierential equations, also known as the Solomon equations: dMzI dt dMzS dt where 1 S (S + 1) b2 = [j (I S ) + 3j (I ) + 6j (I + S )] , II 15 T1 and 1
IS T1
1 1 eq eq (MzI Mz ) IS (MzS Mz ), I S II T1 T1 (14.14)
1 1 eq eq (MzI Mz ) SS (MzS Mz ), I S SI T1 T1
(14.15)
I (I + 1) b2 [j (I S ) 6j (I + S )] . 15
(14.16)
SS SI Expressions for T1 and T1 can be obtained from the expressions above by interchanging the indices SS II IS SI I and S . The times T1 and T1 are called auto-relaxation times, and T1 and T1 are called cross-
relaxation times. P. J. Grandinetti, October 31, 2011
14.2. RELAXATION VIA DIPOLAR COUPLINGS Similarly, for the MI transverse component one nds dM+1I M+1I = , dt T2 where 1 2 3 1 3 1 = S (S + 1) b2 j (0) + j (S ) + j (I ) + j (I + S ) + j (I S ) . I 15 15 10 15 30 T2 14.2.2.1 In the Motional Narrowing Regime
79
(14.17)
(14.18)
If we assume the spectral density function arises from a simple model of isotropic molecular tumbling, having a single reorientational correlation time c , then one can show that 1 II T1 Thus, we dene mzI = and obtain dmzI 1 1 eq II (mzI meq zI ) (mzS mzS ) , dt 2 T1 1 1 dmzS eq SS (mzI meq zI ) (mzS mzS ) . dt 2 T1 (14.21) (14.22) Mz S Mz I , and mzS = , I (I + 1) S (S + 1) (14.20) 1
IS T1
2S (S + 1) . I (I + 1)
(14.19)
We still have two coupled dierential equations but now the the cross-relaxation times are no longer important and only the two auto-relaxation times are needed to describe the motional narrowing limit.
14.2.3
Steady-State Overhauser Eect
If the S nucleus is continuously irradiated, i.e., decoupled, while the I spin is detected then 1 1 dMzI eq eq = II (MzI Mz ) IS (Mz ) I S dt T1 T1
ss Under these conditions MIz (t) eventually will reach a steady state value of MIz when
(14.23) dMIz = 0. This dt (14.24)
leads to
eq ss Mz = MIz + I II T1 eq Mz . S IS T1
Rearranging this expression and substituting back into the previous expression we obtain dMzI 1 ss = II (MzI Mz ). I dt T1 (14.25) P. J. Grandinetti, October 31, 2011
80
These last two equations indicate that under continuous irradiation of the S spins, the I magnetization will adjust to a steady state value with a single exponential relaxation time. The enhancement ratio of the steady-state to equilibrium magnetization is given by
N OE
ss Mz I eq = 1 + IS , MzI
(14.26)
where IS , given by IS = S I 6j (S + I ) j (S I ) , j (S + I ) + 3j (I ) + 6j (S + I ) (14.27)
is called the steady-state nuclear Overhauser enhancement. In the motional narrowing regime (small c ) the enhancement ratio becomes
N OE
1+
S , 2I
(14.28)
whereas in the slow motion regime (0 c > 1) we nd

N OE
2 2 2 2 (I S )(3I + 5I S 10S ) . 4 2 2 3 4 3I + I S 10I S + 10S
(14.29)
14.3
Quadrupolar Relaxation
In the special case of I = 1 a single exponential recovery with time constant T1 arises from uctuations of quadrupolar coupling is given by 1 3 = T1 80 and 1 1 = T2 160 1+ 1+
2 q 3
e2 qQ
[j (0 ) + 4 j (20 )],
(14.30)
2 q 3
e2 qQ
[9 j (0) + 15 j (0 ) + 6 j (20 )].
(14.31)
Here e2 qQ/h is the quadrupole coupling constant and q is the asymmetry parameter for the quadrupole coupling. For nuclei with spin I > 1, there will generally be a multi-exponential recovery, and the assumption of a single exponential recovery with time constant T1 is not valid. In the extreme narrowing case and with the assumption of isotropic molecular tumbling, we obtain a single exponential recovery time for arbitrary spin I , which is given by 1 1 3 2I + 3 = = T1 T2 40 I 2 (2I 1) 1+
2 q 3
e2 qQ
c ,
(14.32)
where c is the molecular reorientation correlation time. P. J. Grandinetti, October 31, 2011
14.4. NUCLEAR SHIELDING ANISOTROPY RELAXATION
81
14.4
Nuclear Shielding Anisotropy Relaxation
When the electron density around a nucleus is not spherically symmetric the nuclear shielding is a tensor quantity which depends on the orientation local electron density with respect to the external magnetic eld. In such a situation the nuclear shielding requires two additional parameters, || and to describe the full tensor. Fluctuations in the nuclear shielding interaction will cause relaxation, described in terms of a single T1 and T2 according to 1 1 = B0 || j (0 ), T1 15 and 1 = B0 || T2 1 2 j (0) + j (0 ) . 45 30 (14.34) (14.33)
In the extreme narrowing case and with the assumption of isotropic molecular tumbling we obtain 1 2 2 2 = B0 || T1 15 and 7 2 2 1 B0 || = T2 45
2 2
c ,
(14.35)
c .
(14.36)
82
Chapter 15
Multi-Dimensional NMR
Multi-dimensional NMR originated with Jean Jeener back in the early 1970s and was developed by many others over the last few decades, including Richard Ernst who received the Nobel Prize in Chemistry in 1991.
In two dimensional NMR experiments we dene two time domains t1 and t2 , that are usually divided by some perturbation (typically an rf pulse), and increment both times independent of each other. Now depending on the perturbations you apply and the particulars of the spin system youre exciting, some spins that have one particular frequency during t1 will have a dierent frequency during t2 . A 2D spectrum provides us with a map of the correlations between the spins frequencies in the dierent time domains.
83
84
CHAPTER 15. MULTI-DIMENSIONAL NMR
preparation period
mixing period
t1
preparation period mixing period
t2
t1
preparation period
t2
Frequency (1)
1 0
mixing period
t1
t2
-1 -2 -2 -1 1 0 Frequency (2)
2
-2
-1
-2
-1
Frequency (1)
Frequency (2)
Lets look at a couple examples where such correlations can be exploited to obtain information unavailable in a one-dimensional NMR experiment.
15.1
2D Exchange and 2D NOESY NMR
We saw in section 8 how the NMR resonances of two nuclei undergoing chemical exchange will broaden and then coalesce into a single resonance as the rate of exchange increases. The two-dimensional exchange experiment is used to follow chemical exchange processes that occur on a much slower time scale, where the exchange rate has little eect on the lineshapes. The sequence and coherence transfer pathway is shown below.
(/2)1
(/2)2
(/2)3
t1
+1 0 -1
tm
t2
Lets consider this experiment with a system consisting of two chemically distinct but uncoupled spin 1/2 nuclei. Well label these two sites A and B. One can show that without exchange the signal for these P. J. Grandinetti, October 31, 2011
15.1. 2D EXCHANGE AND 2D NOESY NMR two sites in the experiment above is
85
S (t1 , t2 ) = SA (0, 0)eiA t1 eiA t2 e(t1 +t2 )/T2
(A)
+ SB (0, 0)eiB t1 eiB t2 e(t1 +t2 )/T2 .
(B )
If we did a double Fourier transform of this signal with respect to t1 and t2 , that is,

S (1 , 2 ) =
0 0
S (t1 , t2 )ei1 t1 ei2 t2 dt1 dt2 ,
we would get a two-dimensional spectrum that would look schematically like the one below.
Projection onto 2 axis
1 A
B
Peaks for the two sites A and B will appear along the diagonal line 1 = 2 in the two-dimensional spectrum. On the top and right side of the two-dimensional spectrum are the one-dimensional projections (sums) of the two-dimensional spectrum onto each axis. This two-dimensional spectrum is a plot of how these two one-dimensional spectra (projections) are correlated. Lets now consider again the two site exchange process where a nucleus is exchanging between sites A and B. When this exchange process is occurring much slower that the individual T2 times for each site then our two-dimensional spectrum will contain two additional peaks as shown below. P. J. Grandinetti, October 31, 2011
86
1 A
B
These two new peaks are called cross-correlation peaks and appear on the o-diagonal. For example, the signal at (1 , 2 ) = (A , B ) corresponds to a nucleus that was in the A environment during t1 and during tm exchanged over to the B environment. Sites along the diagonal are called auto-correlation peaks and correspond to nuclei which were in the same environment for both t1 and t2 . Clearly when tm = 0 there should be no cross peaks and as tm is increased, the cross peaks will grow in intensity at the expense of the diagonal peaks. Assuming k1 = k1 = k and that the sites A and B are equally populated the intensities of the auto-correlation and cross-correlation peaks as a function of mixing time, tm , are given by aAA = aBB = 1 [1 + exp(2ktm )] exp(tm /T1 ), 2 1 [1 exp(2ktm )] exp(tm /T1 ), 2 (15.1)
aAB = aBA =
(15.2)
respectively. From this result one nds that the exchange rate constant can be extracted from the ratio of the cross peak to diagonal peak as a function of tm , that is, IAA aAA 1 ktm = . IAB aAB ktm Below are plots showing typical variations of the auto- and cross-peak intensities as a function of mixing time, when T1 = 100 milliseconds and k = 30 sec1 . P. J. Grandinetti, October 31, 2011
15.1. 2D EXCHANGE AND 2D NOESY NMR

1.0 0.8 0.6 1.0 0.8 0.6
87
aAA
0.4 0.2 0.0 100 200 300
aAB
0.4 0.2 0.0
400
100
200
300
400
tm
tm
As mentioned in section 8, by measuring rate constants, k , as a function of temperature one can obtain the activation energy for the exchange process.
15.1.1
Transient nOes
Not only can chemical exchange generate cross-peak intensity, but relaxation mediated by dipolar coupling between nuclei can also do the same. Recalling our expressions for auto- and cross-relaxation we dene auto and cross-relaxation rates for two nuclei of the same isotope (same gyromagnetic ratio and angular momentum) but dierent resonance frequencies as
11 22 Rauto = 1/T1 = 1/T1 ,
(15.3)
and
12 21 Rcross = 1/T1 = 1/T1 ,
(15.4)
and rewrite the Solomon equations dMz1 dt dMz2 dt =

eq eq Rauto (Mz1 Mz ) Rcross (Mz2 Mz ), 1 2
(15.5) =
eq eq Rcross (Mz1 Mz ) Rauto (Mz2 Mz ). 1 2
Solving these equations we nd that there will be an exchange of longitudinal magnetization during the mixing time according to
m M1z cosh(Rcross m )eRauto m M1z + sinh(Rcross m )eRauto m M2z ,
(15.6) (15.7)
m M2z sinh(Rcross m )eRauto m M1z + cosh(Rcross m )eRauto m M2z .
88
Thus, we see that cross-peak intensity will be generated by nuclei that cross relax each other. From these equations we can obtain expressions for the auto- and cross-peak amplitudes, aauto (m ) = cosh(Rcross m )eRauto m , across (m ) = sinh(Rcross m )eRauto m . At short mixing times these amplitudes can be approximated as aauto (m ) 1, across (m ) Rcross m . Recalling our earlier expression for the cross-relaxation rate Rcross = 1
IS T1
(15.8) (15.9)
(15.10) (15.11)
I (I + 1) 15
0 I S 4r3
[j (I S ) 6j (I + S )] ,
(15.12)
we nd for the cross-peak amplitude in the motional narrowing limit that across (m ) 1 . r6 (15.13)
Chapter 16
Magnetic Resonance Imaging

Nuclear magnetic resonance can also be used as a tool for imaging samples. By applying a linear magnetic eld gradient across the sample, B(r) = B0 + G r, and neglecting relaxation we write the signal in the rotating frame, integrated over the sample volume as S (t) = (r)ei Grt dr, (16.1)
where dr implies integration over the volume of the sample. We could identify this as the time domain signal after applying a /2 pulse. On the other hand, if we dene the reciprocal space vector k with k = Gt/2, we can describe this signal as the nuclear spin density in reciprocal space, S (k) = (r)ei2kr dr, (16.3) (16.2)
Written in this form, the volume integral the signal is recognized as a three-dimensional Fourier transform of the magnetization density in real space. Thus, if we apply a three dimensional inverse Fourier transform to S (k) we obtain the three dimensional real space image of the sample, that is, (r) = S (k)ei2kr dk. (16.4)
There are numerous approaches for doing magnetic resonance imaging. For additional details, the reader is directed to the text by Callaghan listed at the end of the chapter. To illustrate one approach consider the sequence below. 89
90
CHAPTER 16. MAGNETIC RESONANCE IMAGING
(/2)x r.f.
t
time
Gz Gy Gx
In this experiment a single /2 creates transverse magnetization. Three magnetic eld gradients are then applied sequentially along three orthogonal directions. The objective in this experiment is to independently and linearly increase the value of k in all three directions, so that a Fourier transform with respect to all three reciprocal space dimensions leads to a three dimensional real space spin density plot, according to
(x, y, z ) =
S (kx , ky , kz )ei2(kx x+ky y+kz z) dkx dky dkz
(16.5)
From Eq. (16.2) we see that we can vary the size of k through either the size or the duration of the applied eld gradient. In the sequence above one increases kz and ky by increasing the gradient strengths Gz and Gy with constant durations z and y . Increases in kx are obtained by acquiring the signal as a function of time with a xed gradient strength Gx . The medical applications, alone, of magnetic resonance imaging are staggering, and form the basis of a highly protable industry. Below is an example of a two-dimensional MRI scan of a human head. P. J. Grandinetti, October 31, 2011
91
92
CHAPTER 16. MAGNETIC RESONANCE IMAGING
Chapter 17
Inside the Spectrometer

The chapter is intended to give the beginning NMR spectroscopist a fundamental understanding of the principles behind the operation of an NMR spectrometer. More detailed descriptions of the components within the NMR spectrometer can be found in other texts. After mastering the contents of this chapter you should be able to deal with the most routine problems that crop up in daily operations. At the very least, you should be able to speak more intelligently with the company repairman.
17.1
The Magnet
One piece of equipment you need for most NMR experiments is the magnet. In the old days water cooled electromagnets were used. Without building a cooling lake the maximum elds you could obtain were about 3.7 Tesla.
Lines of magnetic flux
93
94
CHAPTER 17. INSIDE THE SPECTROMETER These days the most eect way to generate high elds is to use superconducting wire technology.
The superconducting wire, usually niobium-based, is wrapped around a tube and the NMR experiment is performed in the tube at the center of the coil. Current superconducting magnets operate at liquid Helium temperatures ( 4 K). Therefore, you usually never see the superconducting wire wrapped around the tube because the wire is immersed in a dewar of liquid Helium. Usually the dewar of the liquid Helium is immersed in a dewar of liquid nitrogen ( 77 K). The dewars are constructed, however, so that the usable magnetic eld region is at room temperature inside the inner bore of the magnet. Two very important criteria for any NMR magnet are (1) eld drift and (2) eld homogeneity. Field drift is a measure of the time variations in the magnetic eld strength, and eld Homogeneity is a measure of the spatial variations in the magnetic eld strength. NMR has very stringent requirements on both of these criteria. For most NMR magnets the eld drift is less than 40 ppb/hour and eld is homogeneous to within 20 ppb over the sample region ( 20 mm in an 89 mm bore magnet).
17.1.1
Improving Field Homogeneity - Shims
Even with values as good as these, many NMR experiments require even more stringent requirements. In this case supplemental coils called shim coils1 are added. One way to describe a spatially inhomogeneous magnetic eld is with a multipole expansion about the origin,
B (r) B0 (0) =
l=1
4 2l + 1
+l
al,m rl Yl,m (, ).
m=l
(17.1)
where Yl,m (, ) are the spherical harmonics. Shim coils are often designed to produce magnetic eld gradients that counteract specic contributions in this expression. Although the al,m and Yl,m (, ) are complex numbers, one can combine terms to obtain the real combinations of spherical harmonics shown in the table below.
1 The
word shim comes from the days of electromagnets when physicists and engineers would place small pieces of
metal (called shim stock) in the magnetic eld region to move the magnetic ux lines around and make the eld more homogeneous near the sample. Today, current is adjusted in the shim coils to improve the eld homogeneity.
17.1. THE MAGNET
95
Order, l 0 1 1 1 2 2 2 2 2
Cartesian Coordinates z0 x y z 1
2 1 2 2 (3z
Spherical harmonics 4 Y0,0 2/3 (Y1,1 + Y1,1 )r 2/3 i(Y1,1 + Y1,1 )r 4/3 Y1,0 r 4/5 Y2,0 r2 4/10 (Y2,2 + Y2,2 )r2 4/10 i(Y2,2 + Y2,2 )r2 4/10 i(Y2,1 + Y2,1 )r2 4/10 (Y2,1 + Y2,1 )r2
r2 )
3(x2 y 2 ) 3xy 3yz 3zx
On a commercial spectrometer there will be a series of knobs2 labeled with the Cartesian coordinates shown above. Each of these knobs control the currents in a shim coil designed to eliminate that particular contribution to the eld inhomogeneity. A properly shimmed magnet can have eld homogeneities as good as 1 ppb. Another approach to eliminate eld inhomogeneities is to spin the sample. This will make the m = 0 terms in the expansion time dependent, and if the eld gradients are not so large, and the sample spinning is sucient rapid, all the m = 0 terms will be averaged away,
B (r) B0 (0) =
l=1
4 2l + 1
al,0 rl Yl,0 (, ).
(17.2)
For this reason, coils for shimming inhomogeneities up to fourth order, l = 4, along z , that is, m = 0, are common on spectrometers, whereas coils for shimming inhomogeneities involving x or y are usually only available up to second or third order, l = 2 or 3. If the gradients are so large that the sample spinning rate is insucient, then spinning sidebands will appear in the spectrum. Fortunately, the shim currents can be controlled via the spectrometer software, the process of optimizing to obtain the best eld homogeneity can be automated.
17.1.2
Improving Field Drift - The Lock
To improve magnetic eld drift most spectrometers employ a lock channel. A lock adjusts the current in the shim coils so that the resonance frequency of a reference nucleus (typically, 2 D in a solvent molecule) does not vary with time. The lock channel design is a scaled down version of an NMR spectrometer whose signal is used solely as feedback in holding the static eld constant.
2 More
modern spectrometers dont have knobs, but control all the shim currents via software.
96
CHAPTER 17. INSIDE THE SPECTROMETER
17.2
A Primitive NMR Spectrometer
As a starting point in understanding how NMR spectrometers work lets start with the simplest possible design that can function as an NMR spectrometer and understand its operation. We will then explore a series a renements leading to a design that shares many of the features of todays commercial instruments.
Radio Frequency Source
Output of Freq. Source is Continuous
N
Transmitter Switch Sample
Receiver Switch
Receiver
There are 6 essential components in our primitive NMR spectrometer: (1) a radio frequency (rf) source tuned to the resonance frequency of the nuclei, (2) a switch (or gate) for turning the rf irradiation on and o, (3) a magnet to polarize and split the nuclear spin energy levels, (4) a transmitter and detector coil containing the sample, (5) a switch (or gate) in front of the receiver for protection, and (6) the receiver, which could simply be an oscilloscope in this design. With the exception of the magnet, constructing such an instrument is fairly straightforward. Because the time scale of the NMR experiment is often on the order of microseconds, the switching times for our gates need to be on the order of nanoseconds for precise time resolved measurements. Computer controlled low power RF gates having such switching speeds are readily available commercially. Although a computer is not essential for our primitive spectrometer, and were not available in the early days of NMR, they have become an essential component in any modern spectrometer. In the context of our primitive spectrometer the computer would be a convenient addition for controlling the timing for the opening and closing of the transmitter and receiver gates. The simplest NMR experiment consisting of a single pulse and detection of the free induction decay would consist of following three steps in our primitive spectrometer: Step 1 2 3 Transmitter switch state OFF ON OFF Receiver switch state OFF OFF ON Duration 10 seconds 4 microseconds 100 milliseconds
17.3. THE PROBE
97
This is an elementary version of what NMR spectroscopists call a Pulse Sequence or Pulse Program. An important part of a pulse sequence is the duration of each step, or event. Modern NMR spectrometers have many more computer controlled switches than just the receiver and transmitter ON/OFF switch. These instruments also have complex pulse sequence languages that allow the NMR spectroscopist to write sequences containing loops, if statements, and other possibilities. A design, even as primitive as the one above would actually work for detecting strong signals like the 1 H resonance in liquid H2 O. More interesting samples, however, would represent a much greater challenge to this spectrometer in regards to sensitivity. In addition, we will need more exibility in our spectrometer if it is to be capable of multinuclear investigations. On the issue of sensitivity we now look at the NMR probe. Later we will consider modications to the transmitter and receiver that allow for more exible operation of the NMR spectrometer in terms of multinuclear experiments.
17.3
The Probe
In our primitive spectrometer the coil has two purposes: (1) to produce the oscillating B1 eld that rotates the magnetization, and (2) detects the precessing magnetization after the pulse. Thus, we call it a transceiver coil.
B0 E(t)
Lets rst examine what is needed to enhance to ability of this coil as a detector. These same changes, in turn, will also enhance the eciency of the coil in producing higher B1 elds for a given rf power level. We start with the question: How do we measure E (t)? To measure voltage E (t) we measure the current through a resistor, as in the circuit below. P. J. Grandinetti, October 31, 2011
98
meter resistance transceiver coil meter
Before we give a more detailed answer to this question lets rst review some basics about voltages and currents in electronic components like the resistor, capacitor, and inductor.
17.3.1
Review of RF circuit components
In this section we will review the current-voltage-frequency relationships for the resistor, inductor, and capacitor, as it relates to the operation of our NMR probe. 17.3.1.1 The Resistor
Lets start with the resistor. If we applied an oscillating voltage V (t) = V0 cos t to a resistor in the circuit below
V(t) V(t) R I(t)
time
time
we would nd across the resistor an oscillating current given by I (t) = V (t) V0 = cos t. R R
The maximum current through the resistor is related to the maximum voltage according to Ohms law V0 = Imax R. P. J. Grandinetti, October 31, 2011
17.3. THE PROBE 17.3.1.2 The Inductor
99
Now lets apply the same oscillating voltage to an inductor (like our transceiver coil)3 as in the circuit below.
V(t) V(t) L I(t)
time
time
Here we nd that the current is given by

t
I (t) =
t0
V (t) V0 ds = sin t L L
Notice here that: 1. The current oscillation is 90 out of phase with the voltage oscillation. That is,
V (t) = V0 cos t
and
I (t) =
V0 cos(t /2). L
2. Ignoring the phase dierence, we nd that the maximum current is related to the maximum voltage according to Imax = V0 . L
The inductor behaves like it has a frequency dependent resistance of L, and has a current that is 90 out of phase with the voltage oscillation. In general, low frequency currents pass through inductors and high frequencies are blocked.
3 One
can usually estimate the inductance of a coil from its dimensions. The inductance, L, in Henries is given by L= r2 n2 , 9r + 10l (17.3)
where r is the radius of coil in inches, n is the number of turns, and l is the coil length in inches.
100
in
out out
XL = L
in
You may now see the problems of using an inductor alone as a transceiver coil. As a transmitter of the oscillating B1 eld we need to push lots of current through the coil to generate a large B1 eld. However, for a xed oscillating voltage source (i.e., xed power), the current decreases with increasing frequency. As a receiver coil, the oscillating current we measure from the E (t) due to the precessing magnetization decreases rapidly as the frequency of precessing magnetization increases. It turns out there is a simple solution to this problem. Before we look at this solution, however, lets rst take a look at the voltage-current-phase relationships for capacitors, and second, dene the concept of impedance for helping us follow the voltage-current-phase relationships of capacitors, inductors, and resistors in a circuit. 17.3.1.3 The Capacitor
What if we applied the same oscillating voltage to a capacitor as in the circuit below.
V(t) V(t) C I(t)
time
time
What would be the current owing across the capacitor? In this case we would nd that I (t) = C dV (t) = CV0 sin t. dt (17.4)
Again, there are two important things to notice: 1. The current oscillation is 90 out of phase with the voltage oscillation. That is, V (t) = V0 cos t P. J. Grandinetti, October 31, 2011 and I (t) = CV0 cos(t + /2)
17.3. THE PROBE 2. Ignoring the phase dierence, the maximum current is related to the maximum voltage by Imax = V0 . 1/(C )
101
In short, the capacitor behaves like it has a frequency dependent resistance of 1/(C ) and has a current oscillation that is 90 out of phase with the voltage oscillation. Because it is 90 out of phase it is called reactance (not resistance). We say that a capacitor will block (react against) low frequency currents, but allow high frequencies to pass. This is the opposite behavior of the inductor.
Xc = 1/(C)
in
out
in
out
17.3.2 Impedance
We just saw that the relationships between the maximum current and maximum voltage across a resistor, capacitor, and inductor were given by Imax = V0 , R Imax = V0 , 1/(C ) and Imax = V0 , L
respectively. How do we derive a relationship for current and voltage at all times, not just for the maximum values; and one that also includes the phase information? We do this by dening a complex voltage Vc (t) = V0 eit , where the actual voltage is obtained by taking the real part, that is, V (t) = {V0 eit } = V0 cos t. Now if we wanted to phase shift V (t) by 90 we would simply multiply the complex voltage Vc (t) by ei/2 , e.g., V (t) = {V0 eit ei/2 } = V0 cos(t + /2). We can applying this to the current-voltage relationship for capacitors we can rewrite Eq. (17.4) as I (t) = Vc (t)ei/2 1/(C ) = V0 cos(t + /2). 1/(C ) P. J. Grandinetti, October 31, 2011
102 Since ei/2 = i we can simplify to I (t) =
Vc (t) i/(C )
Vc (t) ZC
where ZC = i/(C ). Similarly, we dene a complex current Ic (t) = I0 cos t = {Ic (t)} where Ic (t) = I0 eit . Then for a capacitor we can nally write Ic (t) = Vc (t) , ZC
and it now looks like our familiar Ohms law, describing the relationship between current and voltage at all times, including the phase information. Applying the same approach to an inductor we have Ic (t) = Vc (t) , ZL
where ZL = iL. We call this generalized resistance Z the impedance. In summary, the impedance for resistors, capacitors, and inductors are ZR = R , ZC = i/(C ), ZL = iL. (resistor) (capacitor) (inductor)
Using these impedance denitions, Ohms law can be generalized to include inductors and capacitors. For components in series we have ZT = Z1 + Z2 + Z3 + , and for components in parallel we have 1/ZT = 1/Z1 + 1/Z2 + 1/Z3 + .
17.3.3
A Tuned Circuit
Now were ready to solve our problem with the transceiver coil. To make our problem more realistic we include the wire resistance as in the circuit below. P. J. Grandinetti, October 31, 2011
17.3. THE PROBE
103
V(t)
transceiver coil
For this circuit we start with Ic = Vc /ZT , where ZT = ZR + ZL = R + iL. With a little algebra we obtain Ic = Vc R iL R iL Vc . = = Vc 2 R + iL R + iL R iL R + 2 L2
How does the amplitude of the real current depend on frequency? To take the real part of the current we multiply it by its complex conjugate and take the square root:
= Ic Ic
V . R2 + 2 L2
The denominator in this expression tells us that the current amplitude decreases as the frequency increases.
200 180 160
||
140 120 100 80 60 500 1500 2500 /2 3500 4500
Only at = 0 can we get the highest current amplitude. Given that our NMR signal will be oscillating at frequencies in the megaHertz range this is not an optimum situation. We can solve this problem, however, by adding a capacitor in series as shown below. P. J. Grandinetti, October 31, 2011
104
V(t)
transceiver coil
C
For this circuit we have ZT = ZR + ZL + ZC = R + i L 1 C
2 If we set ZL = ZC , that is, 0 L = 1/(0 C ) or 0 = 1/(LC ) then we have
I = V /R
at
0 = 1/ LC
and we get the highest current amplitude at 0 , not = 0.

200 180 160
||
140 120 100 80 60 500 1500 2500 /2 3500 4500
Making the capacitor variable gives us the added exibility for tuning the probe to the resonant frequency of other nuclei.
17.3.4
A Matched Circuit
With that problem solved lets put our voltage meter back into the circuit and measure the oscillating voltage due to the precessing magnetization at 0 . P. J. Grandinetti, October 31, 2011
17.3. THE PROBE

transceiver coil resistance
105
Rt
Rm
Voltage meter transceiver coil inductance
Ct
Variable Capacitor for Tuning
By adding the meter into the circuit we have increased the total impedance to ZT = Rm + Rt + i L and at the resonant frequency 0 = 1/ LC we have ZT = Rm + Rt , and the current is I (0 ) = Thus, the voltage across the meter is Vm (0 ) = I (0 )Rm . Wed like the amplitude of the meter voltage oscillation to be the highest possible for a given current. That is, we want maximum power transfer from the coil to the meter. Recalling that power is the product of current and voltage, wed like to maximize Pm (0 ) = I (0 ) Vm (0 ) where Pm (0 ) is the power transferred to the meter at 0 , and Vm (0 ) is the voltage across the meter at the frequency 0 . We obtain Vm (0 ) from the relation Vm (0 ) = I (0 )Rm giving us Pm (0 ) = I (0 )2 Rm . Using Eq. (17.3.4) in this expression we obtain Pm (0 ) = V (0 )2 Rm . (Rm + Rt )2 P. J. Grandinetti, October 31, 2011 V (0 ) . Rm + Rt 1 C
106
To nd the relationship between Rm and Rt that gives maximum Pm (0 ) we set 2V (0 )2 Rm V (0 )2 dPm (0 ) = 0. = 2 dRt (Rm + Rt ) (Rm + Rt )3 Solving this expression gives the condition Rt = Rm for the maximum power transfer. When Rt = Rm we say that the impedance of the tuned circuit is matched to the receivers impedance. The receivers impedance (resistance) is usually xed at 50 . Since the tuned circuit impedance is xed at Rt when 0 = 1/ LC then Rt needs to be 50 to have maximum power transfer. Rt is generally less than 1 . Adding a variable resistor, Rv , to the transceiver circuit would give you a matched circuit (i.e., Rt = Rm + Rv ) but this doesnt really help since we will lose power in resistant heating of Rv . This solution also doesnt help in transferring all the power possible into the meter. What do we do? We add another capacitor to our circuit to match the impedance and better couple the power into the resistor, as shown below.
Rt
Receiver 50
Cm
L
Ct
Transceiver Circuit
In this series tuned, parallel matched circuit we have 1 = iCm + ZT and after a little algebra we get Rt + i ZT = L 1 Ct 1 Cm L + 1 Cm 1 Ct
2 Cm Rt 2
1 1 Rt + i L Ct
2 R2 2 Cm t
1 L Ct
If the capacitors Ct and Cm can be adjusted so that this impedance is completely real (no imaginary part) and equal to 50 at a frequency 0 then we will have maximum power transfer between the sample and the receiver, or between the transmitter and the sample. P. J. Grandinetti, October 31, 2011
17.3. THE PROBE
107
Before leaving the transceiver (i.e., the NMR probe) lets look at another transceiver circuit shown below, the parallel tuned, series matched circuit, that is popular.
Cm
Rt
Receiver 50
Ct
L
Transceiver Circuit
We will leave it as an exercise for the student to work out the impedance of this circuit. To understand the simple picture of whats going on here lets ignore the matching capacitor, Cm , and resistance, Rt .
Rm
Ct
Then we have 1 1 = i C Z L In this case when C = .
1 then Z = . That is, there is an innite impedance and therefore no L current would ow through the meter. If the precessing magnetization creates an oscillating voltage across the inductor, then where is the current? It stays in the capacitor-inductor closed loop. To couple the power out of a parallel circuit, a matching capacitor is used.
You should now have a basic understanding of the rf circuits that are typically found in an NMR probe. Most NMR spectrometers are built so that probes can be easily interchanged, and often are P. J. Grandinetti, October 31, 2011
108
purchased with a number of dierent NMR probes optimized for performing a wide variety of liquid and solid-state NMR experiments.
17.4
The Duplexer
Our tuned and matched circuit will give us signicantly higher sensitivity and shorter /2 times. To get the shortest possible /2 times, however, we will want to place a high power rf amplier after our transmitter gate and before our probe. With the addition of the high power amplier we need to add an extra switch - the duplexer switch, that switches the probe between the transmitter and the receiver, and is necessary to protect the receiver from the high power rf pulse of the transmitter.
dB
Transmitter Switch
High Power Amplifier
Duplexor Switch Receiver Switch Receiver
Because the duplexer is switching between the high power rf from the transmitter and the low power rf from the probe there is an inexpensive and simple passive circuit that can be used to rapidly perform this switching. To understand how this works, however, we need to review two important devices - (1) the cross diodes, and (2) quarter-wave transmission lines.
17.4.1
Cross-Diodes
. A diode only allows current to ow in one direction (i.e. direction
First we consider the diode: of arrow). P. J. Grandinetti, October 31, 2011
17.4. THE DUPLEXER

Current Flow
109
No Current Flow
A plot of current ow versus applied voltage for a diode looks like
20 mA
10 mA -100 v -50 v 1v -1 A 2v Applied Voltage
-2 A Current Flow
Note the axes ranges. For a diode current ows forward after a voltage of greater half a volt is applied. If we took two diodes and connected them antiparallel we would have
=
Symbol for Cross diodes
Then a cross-diode pair would have the property that current could ow in either direction as long as the voltage is greater than half a volt in magnitude. P. J. Grandinetti, October 31, 2011
110
20 mA
10 mA -2 v -1 v 1v -10 mA 2v Applied Voltage
-20 mA Current Flow
One way we can use this cross-diode arrangement is to protect the receiver from seeing the high power rf pulse using the circuit below.
from probe
Receiver 50
A high voltage pulse would turn the diodes on and go to ground instead of going into the receiver with its 50 impedance. Another use of cross-diodes are to block low power noise from the transmitter from entering the probe.
Transmitter 50
to probe
Broadband noise from transmitter can easily saturate NMR signal and needs to be eliminated. With this arrangement as long as the noise voltage doesnt exceed the threshold voltage of the diode it will be blocked from going to the probe. Often times when the signal-to-noise ratio unexplainably drops it is often a blown cross-diode here that is the problem. Solid-state NMR experiments which use long high power pulses such as cross-polarization are often responsible for blown out diodes.
17.4.2
Transmission lines
Before talking about /4 lines lets consider transmission lines in general. We saw that for maximum power transfer it is desirable to have the impedance of all our devices (i.e. probes, transmitter, receiver) P. J. Grandinetti, October 31, 2011
17.4. THE DUPLEXER
111
matched. So we will want to connect all these devices together making sure the impedance is 50 everywhere.
Transmitter 50
If the impedance of the transmitter is 50 then the current and voltage oscillations will be in-phase
transmission line (usually a coaxial cable)
Load (Probe) 50
V(t)
time
Will the load (Probe) see a 50 impedance? That is, will the load see the current and voltage oscillations in phase?
I(t)
time
In order to match the impedance of the source and load the characteristic impedance of the transmission line must be given by Z0 = Zs Zl =
50 50 = 50 ,
where Zs is the source impedance and Zl is the load impedance. Now lets talk about /4 lines. If a transmission is terminated by a load that doesnt match its characteristic impedance then the voltage and current waves are partially reected and standing waves are set up in the transmission line. If we terminated a transmission line with a short, i.e. connected the inner conductor to the outer conductor
Outer Conductor
End shorted so that Zl = 0

Inner Conductor
then all the voltage and current oscillation will reect and set up a standing wave.
/4
Transmitter 50
V(x)
I(x)
112
At the shorted to ground end the voltage must be zero and the current must be maximum at the grounded end. Such constraints at the shorted end force the voltage to be maximum and the current to be zero at a distance /4 away from the shorted end, that is, where the source is connected. Thus it looks like the cable has an innite impedance at point where the source is connected, Z= V (0) Vmax = = I (0) 0 amps
So the source cant tell the dierence between nothing connected and a /4 with a shorted and connected. What if we didnt short the end, but left the two conductors unconnected?
/4
Transmitter 50
V(x)
I(x)
If the end is open, then the voltage must be maximum, and current must be zero at the open end. At the source, a distance /4 away, this translates into a zero voltage and a maximum current. Thus the source sees a zero impedance Z= V (0) 0 volts = =0 I (0) Imax
So all the power of the source is being sent to ground. In summary 1. /4 length cables with shorted ends look like an innite impedance at the source.
2. /4 length cables with open ends look like a zero impedance (short to ground!) at the source. Sort of counter intuitive if youve only thought about DC circuits. Now we can go back to our original problem, constructing a duplexer, which switches the probe between the transmitter and receiver. Consider the circuit below: P. J. Grandinetti, October 31, 2011
17.4. THE DUPLEXER

Probe 50 Ohms Transmitter 50 Ohms /4 Receiver 50 Ohms
113
Cross Diode blocks noise from transmitter but lets high voltage pulse through unattenuated.
When the high voltage rf pulse is coming from the transmitter it turns on all cross-diodes. So when the pulse is on, the transmitter sees
Probe 50 Ohms Transmitter 50 Ohms
Pulse On
/4
Receiver 50 Ohms
At the tee (marked with dot) the transmitter sees the 50 load of the probe, and an innite load in front of the receiver. So no pulse rf goes into the receiver. On the other hand, when the rf transmitter pulse is o, then the weak signal from the probe is not able to turn on the diodes and therefore sees a dierent picture:
Probe 50 Ohms Transmitter 50 Ohms /4 Receiver 50 Ohms
Here the cross-diodes are all o so probe signal goes only to receiver. Since the /4 length depends on frequency (i.e. = c/ ) then any time you change the NMR frequency (i.e. when changing to a dierent nucleus), the duplexer has to be changed (a dierent /4 cable). With the wrong /4 length, then part of the transmitter power will go to ground and not to the probe. If the cross-diodes are blown ( current ows in both directions with no resistance in blown P. J. Grandinetti, October 31, 2011
114
diodes), then transmitter noise will saturate the magnetization and the signal from the probe will not go only to the receiver and sensitivity will suer. Our primitive spectrometer is now very close to what you would nd on a commercial spectrometer. Lets redraw a simplied block diagram.
Probe
Transmitter
Duplexor
Receiver
Weve looked inside the probe and the duplexer, now lets open the transmitter and receiver boxes.
17.5
The Transmitter
So far we have considered the transmitter as shown below:
dB
Transmitter Switch
17.5.1
Specifying Power Levels
The addition of a high power rf amplier brings up an important issue of how to specify rf power levels in the spectrometer. For an amplier the gain is usually given in decibels (dB). This is a logarithmic scale and is calculated according to dB = 20 log10 (Vpp )out (Vpp )in
where Vpp is the voltage peak to peak. So if the rf source has Vpp = 0.632 volts then Vpp after a 50 dB gain would be (V pp)out = (V pp)in 10dB/20 = (0.632 volts) 1050/20 = 200 volts P. J. Grandinetti, October 31, 2011
17.5. THE TRANSMITTER
115
Decibels is a measure of gain, and doesnt provide an absolute measure of the rf power level. One common unit for specifying absolute power level is watts. Starting with P =
2 Vrms , R
and
Vpp Vrms = , 2 2
where R = 50 , we calculate the power level of an rf source with output of Vrms = 0.2236 volts as P = (0.2236 volts)2 = 0.001 watts or 1 milliwatt. 50
After the 50 dB amplier, this power level would be amplied to (200 volts/(2 2))2 P = = 100 watts. 50 RF levels are also specied in units of dBm, given by the expression dBm = 10 log P (milliwatts) 1 milliwatt
In other words, the dBm is the gain in terms of dBs with respect to a power level of 1 milliwatt. So if the power level of our rf source was 1 milliwatt, then the power in dBm would be zero. After a 50 dB amplier we have 100 watts, or in dBm that would be 50 dBm. Many high power ampliers in NMR have a maximum input level of 0 dBm. Anything higher will overdrive an amplier and possibly cause damage to the amplier. Checking the rf power levels is often an important step in trying to debug problems with the spectrometer. When checking power levels, however, always remember this important rule - Never connect the output of a high power amplier directly to the oscilloscope. The oscilloscope cannot handle that much power, and you will damage the oscilloscope. Always stick a high power attenuator (Check power rating) between the amplier and the oscilloscope.
50 dB
30 dB Attemuator (high power rating)
With the setup above you would (1) measure the voltage peak-to-peak on the oscilloscope, (2) convert this to dBm, and (3) add 30 dB to get the output power level of the amplier. P. J. Grandinetti, October 31, 2011
116
17.5.2
Controlling the Transmitter Phase and Amplitude
For many NMR experiments it is necessary to vary the strength of the B1 eld. This can be accomplished by adding a computer controlled attenuator into our transmitter as shown below.
attentuator
dB
TTL control line to pulse programmer
TTL control lines to pulse programmer
The number of TTL control lines determines the precision of amplitude control. For example, with 8 control lines we can have 28 = 256 amplitude levels. Now when we write a pulse sequence for this spectrometer, we will need to specify the states of the eight switches controlling the pulse amplitude in addition to the transmitter and receiver switches for each pulse sequence event. Adding rf phase control is a bit more complicated. It would be nice to add a computer controller phase shifter in the same way as we added the attenuator. This would work if our spectrometer only transmitted at one frequency. Computer controller in-line phase shifters like that are only tuned to one frequency. So if we changed the transmitter frequency (to look at a dierent nucleus), then the phase shifts would all be wrong. Another solution would be to replace the rf source in the diagram above with a computer controller digital rf source. Such devices exist and can be programmed to output any frequency with any phase (or amplitude). Indeed, this would be the perfect solution. Unfortunately, digital rf sources commercially available at the time of this writing cannot change the phase fast enough when operating at the highest NMR resonance frequencies. Fortunately, an old and simple solution to this problem exists and is called Heterodyning. It allows us to use either approach, that is, the computer controlled in-line phase shifter or the computer controlled digital rf sources. In a heterodyne transmitter, we do all our phase shifting at a constant intermediate frequency (IF), and then mix this phase shifted IF up to whatever frequency we need. To explain how heterodyning works in NMR we need to introduce another electronic component - the mixer. 17.5.2.1 The Mixer
For our purposes we can consider the mixer as simply a box with three connections. If you put an oscillating voltage at frequency LO into the LO connection, and an oscillating voltage at frequency RF P. J. Grandinetti, October 31, 2011
17.5. THE TRANSMITTER
117
into the RF connection, then a voltage oscillating at the sum and dierence frequency of LO and RF is obtained at the IF connector.
LO
LO IF
RF
RF = LO IF
IF = LO RF
It also works with RF and IF connections switched.
LO
LO IF
RF
RF = LO IF
IF
A mixer is a very versatile device. For example, if your NMR signal was oscillating at 400MHz, then you could use a mixer and a band pass lter to shift your NMR signal oscillations down or up to whatever frequency you want. For example, below we take an NMR signal from the probe oscillating at 400 MHz and mix it with a rf source oscillating at 430 MHz to obtain, after a 30 MHz bandpass lter, the NMR signal oscillating at 30 MHz.
430 MHz continuous wave rf oscillator (Signal)400 MHz

LO RF IF
30 MHz BandPass Filter
(Signal)30 MHz
An important feature of the mixer is that phase changes on the inputs show up at the output. P. J. Grandinetti, October 31, 2011
118
430 MHz rf oscillator with phase L
30 MHz rf oscillator with phase I
LO RF IF
400 MHz BandPass Filter
400 MHz rf oscillation with phase L-I
This is an important feature we can use in our transmitter design to do phase shifting at any frequency. For example, in a spectrometer with a Larmor frequency of O , we can use two rf sources, one with an adjustable phase I and a frequency of 30 MHz and the other with a frequency of O + 30 MHz and constant phase L = 0, and mix them together to get an rf source oscillating at O with phase I .
30 MHz + 0 rf oscillator with phase L= 0
30 MHz rf oscillator with phase I
IF
LO RF
0
BandPass Filter
0 oscillation with phase I
Thus our transmitter can be built to give us any phase we want at any frequency by using a mixer and frequency source with fast computer controlled phase. P. J. Grandinetti, October 31, 2011
17.6. THE RECEIVER
119
Radio Frequency Source 0 + IF
LO IF
RF
attentuator
dB
IF Source with phase
TTL control lines TTL control to pulse programmer line to pulse programmer
TTL control lines to pulse programmer
As with the digital attenuator the number of TTL control lines determines the precision of phase control. For example, with 8 control lines we can have 28 = 256 phase levels. Our initial primitive NMR spectrometer had a pulse sequence that just turned on or o two switches. (i.e. The receiver and transmitter on/o switches) With our amplitude and phase controlled transmitter many more switch states need to be specied for each pulse sequence event.
17.6
The Receiver
Finally to nish our spectrometer we need to look into the receiver. Coming out of the NMR probe we often have signals on the order of a tenth of a watt. Measuring this signal directly would be dicult, therefore we place a ultra-low noise preamplier with about 30-60 dB gain immediately after the duplexer.
Probe 50
Duplexor
dB
to rest of receiver
This preamplier usually sits in a box next to the magnet and the probe. Once the signal is amplied at the rst stage, it is then transferred along the 50 cable to the main receiver in the spectrometer console. Inside the receiver is an analog-to-digital converter (ADC). An ADC converts a voltage to a number in the computer. P. J. Grandinetti, October 31, 2011
120
17.6.1
Quadrature Detection
Consider the receiver design below.

Analog to Digital Converter
Signal from First Stage Preamp
Variable Attentuation & Gain
To Computer
Gain Controls
With the receiver design below we would need an ADC fast enough to handle the highest NMR Larmor frequency. Unfortunately, the price of ADCs go up as they get faster and faster. Fortunately, we can modify our receiver so that we dont need such an expensive and fast ADC. Although Larmor frequencies are on the order of tens or hundreds of MHz, the range of frequencies for a given isotope are often less than a few hundred kilohertz. This is a very important piece of information. If we plotted the 1 H NMR precession frequencies for a sample it might look like the spectrum below:
0 MHz
400 MHz
This width in Hertz may only be a few hundred kiloHertz
Thus, an easy solution is to use our mixer to shift the signal frequencies down to a lower frequency range as shown below.
399.9 MHz
LO RF IF
Analog to Digital Converter
To Computer
~ 400 MHz
Gain Controls
17.6. THE RECEIVER
121
Then the signal will oscillate at a much slower frequency and we can digitize the signal with a much slower and cheaper ADC.
0 MHz
400 MHz
In fact, wed like to get away with the absolute slowest ADC possible. Thus it would be nice if we could shift all the NMR frequencies as close to zero as possible. What would happen if we shifted so much that some frequencies shifted below zero (i.e. became negative)? Those frequencies would be folded (reected) back as positive frequencies.
peak reflected about zero to other side
0 MHz
400 MHz
This is a problem if we dont know which peaks are reected. We need a way of distinguishing the signs of the frequencies relative to the reference as they are mixed down. This can be done by splitting the signal into two parts and mixing each of them down with the same reference frequency but with dierent reference phases. For example, in the diagram below we see that mixing a signal oscillating at with a reference oscillating at |ref | will produce a signal oscillating at |ref |. P. J. Grandinetti, October 31, 2011
122
cos(|ref|t+)
IF
LO RF
Low Pass Filter
cos[(||-|ref|)t + ]
cos(||t + )
IF RF LO
Low Pass Filter
sin[(||-|ref|)t + ]
sin(|ref|t+)
Based on the output of a single mixer we would not be able to determine whether was greater than or less than ref . By splitting the signal and mixing it with two references having the same frequency but phases that are 90 out of phase, however, allows us to distinguish both positive and negative frequencies relative to the reference. This approach of detecting two signals, both mixed down to the much lower (audio) frequency range is called quadrature detection. This can be implemented as shown below:
X signal
Low Pass Filter
(Signal)lab
frame
IF RF LO
Reference frequency
/2 phase shift
cos(|ref|t+)
RF LO IF
Low Pass Filter
Y signal
Using X and Y output we construct a complex signal that allows us to distinguish positive and negative frequencies. X (t) iY (t) S (t) = X (t) + iY (t) P. J. Grandinetti, October 31, 2011 when < 0, when > 0.
17.6. THE RECEIVER
123
Later we will discuss this more in the context of the Fourier transform. Were almost done with the receiver. One last point is that most NMR receivers are also heterodyned like the transmitter. That is, they mix down to an intermediate frequency before mixing down to the lower (audio) frequency range. This helps to simplify the design of receiver by allowing more narrow band components to be used. Below is the nal block diagram of the receiver.
To Computer
ADC X Channel
IF + 0
Low Pass Filter

TTL control line to pulse programmer
LO RF IF
IF
BandPass Filter
Quadrature Mixing Low Pass Filter
(IF)
Gain Controls
ADC Y Channel
To Computer
That completes our NMR spectrometer design. By now, you should be able to understand most of the switches and knobs on any well designed NMR spectrometer. P. J. Grandinetti, October 31, 2011
124
17.6.2
Single Channel Detection and the Redeld Trick
Some older spectrometers are not equipped with quadrature detection, and only detect a single channel. In such a case the mixed down signal is given only by Eq. (20.7). After eliminating the fast oscillating terms and dening the reference and receiver phase according to Eqs. (22.25) and (20.13) we obtain S (t) = k iref M+1 . (t)eiref + M 1 (t)e 2 detects p = 1 detects p = +1 (17.5)
For each p = 1 resonance there will be a mirror image p = +1 resonance in the spectrum.
p = +1
p = -1
This is not a problem if all p = 1 peaks have positive frequencies, however, this is usually not the case.
p = -1 p = +1
p = +1 p = -1
In the example above both p = +1 and p = 1 resonances are mixed. What do we do? The Redeld trick! We make the receiver phase time dependent ref (t) = 0 s t then the signal becomes S (t) = k i0 is t M+1 (t)ei0 eis t + M e . 1 (t)e 2 detects p = 1 detects p = +1 (17.7) (17.6)
17.6. THE RECEIVER
125
Now we see that the p = 1 signal is shifted to more positive frequencies by s , and the p = +1 signal is shifted to more negative frequencies by s .
p = -1 p = +1 s s
p = +1 s p = -1 s
which leads to
p = +1 p = +1
p = -1 p = -1
(s + )
(s ) 0 (s )
(s + )
The goal is to make s large enough so that all p = 1 resonances are separated from p = +1 resonances. In general s is set equal to the spectral width of p = 1 spectrum. Therefore we sample twice as fast compared to the quadrature case, and make the receiver phase ref (t) = 0 2t t (17.8)
The constant phase 0 is used during phase cycling in the same way as with quadrature detection.
17.6.3
Discretely Sampled Data
There are two essential issues that need to be understood when using any ADC. The rst is that in order for the ADC to accurately turn a voltage into a number, the voltage cant exceed the limits of the ADC.
Upper Voltage Limit for ADC
Signal Voltage
time
Lower Voltage Limit for ADC
126
If signal exceeds the ADC limits then clipping occurs as shown below on the right:
"Clipping" - distorts signal and adds high frequency harmonics
Signal Voltage Signal Voltage
time
time
To avoid problems with clipping in our NMR receiver we place a variable gain preamplier to deal with samples of varying sensitivity. Often, the receiver gain is controlled by the spectrometer software. It is best to adjust the receiver gain to have the biggest signal without clipping to minimize digitizer noise. The second issue is how fast does the ADC have to be? That is, how soon after an ADC turns a voltage into a number can it turn the next voltage into another number? The answer to this depends on how fast the oscillations in the signal are. The ADC has to be able to sample the voltage oscillations of the signal faster than the shortest voltage oscillation period.
V(t)
time
The Nyquist theorem tells us that c =
1 2 s
= 1/2t must be greater than the highest frequency
oscillation present in signal. c is called the Nyquist critical frequency. For example, in the gure below we see that the ADC t is too long for this signal, the numbers generated by ADC make signal look like it has lower frequencies than it actually has.
V(t)
time
When a sample is undersampled, then frequencies are aliased to lower frequencies. Another way to understand this process is through the convolution theorem. The ADC takes the continuous function, S1 (t), like the free induction decay shown below, and multiples it by the continuous time domain picket P. J. Grandinetti, October 31, 2011
17.6. THE RECEIVER
127
fence function, S2 (t), with spikes every t, also shown below, to obtain the digitized signal ST (t). In the frequency domain we view this process as the convolution of the Lorentzian lineshape centered at with the frequency domain picket fence, with unit spikes every s = 1/t, to produce the digitized spectrum.
t=0
Real
S1(t)
time
F. T.
Imaginary
S 1 ( )
0
(Convolution)
(Multiplication)
S2(t)
0
F. T.
time
S2()
s
0
A/t s
frequency
=
ST()
s
t=0
Real
ST(t)
time
F. T.
+s
Imaginary
Since the frequency domain signal in the digitized spectrum is repeated every s it is common practice to only display the spectrum of a single s window. Generally, if a continuous function, S (t), sampled at an interval t, happens to be bandwidth limited to frequencies c , then the function S(t) is completely determined by its samples Sn , where
S (t) =
n=
Sn t
sin[2 c (t nt)] . (t nt)
(17.9)
This means that if all the frequencies present in S (t) are within c then the digitized S (t) is a good approximation. Consider, for example, the convolution below. P. J. Grandinetti, October 31, 2011
128
S 1 ( )
0 1 2
Convoluted with
s
S 2 ( )
0
A/t
frequency
ST()
1s 2s
0 1
2 +C
1+s 2+s
If a continuous function, S (t), sampled at an interval t is not bandwidth limited to frequencies c , then the spectral intensities that lie outside of c will be aliased into that range.
S 1 ( )
0 1 2
Convoluted with
s
S 2 ( )
0
A/t
frequency
ST()
0 1
23
S
2
2+ 1+
S S
2 22 1 12
S S S
1+2
17.6. THE RECEIVER
129
Because 2 was undersampled, it appears in our central window at 2 s , that is, it is aliased into the window. Obviously, if we are to competently sample a continuous signal, all of its frequency components must lie within c . If so, then we can assume that the Fourier transform outside c is equal to zero.
17.6.4
Signal Averaging
One of the biggest advantages of digitizing your signal is that you can signal average (actually summation) to improve your signal-to-noise ratio. Often experiments produce signals so small that they are overwhelmed by the noise of the system. This usually means that the sensitivity of the technique is insucient to examine the sample in question. However, if the signal is there, no matter how small, and if the noise is random, then there is a good chance that weak signals can be improved by signal averaging. During signal averaging the signal will grow linearly with the number of scans Sn = n S1 , where Sn is the total signal after n scans and S1 is the signal of 1 scan. In contrast, the noise, if it is truly random, will add less eciently, growing at a rate proportional to the square root of the number of scans, Nn = n N1 ,
where Nn is the R.M.S. noise after n scans, and N1 is the R.M.S. noise after one scan. Thus after n scans we have n S1 (S/N )n = = n n N1 S N
1
So, the signal-to-noise ratio grows as the square root of the number of scans. This is a very important result for an experimentalist. If each scan takes the same amount of time, then in order to double your S/N ratio you will need to average your signal four times longer. To Quadruple your S/N youll need 16 times longer signal averaging. Therefore, your S/N with averaging isnt cheap as far as time is concerned. This is why improvements in S/N of a single scan can be something worth working towards. For example, in
13
C NMR in solids the S/N was nearly prohibitively low. Back in the early 1970s, however, Pines,
Gibby, and Waugh developed an approach called cross-polarization for improving the 13 C S/N ratio by a factor of approximately four, and at the same time shorten the time required between scans signicantly. Just the factor of four improvement meant that the S/N of a two week signal averaging experiment could now be obtained in less than a day. The fact that the time between scans was also reduced by a factor P. J. Grandinetti, October 31, 2011
130
of almost 2 or 3 meant that in one day we can now get the signal-to-noise ratio that would have taken one and a half months of signal averaging!
17.7
Summary
Bringing everything together we can think of our spectrometer in terms of the following block diagram:
Probe
Others, e.g. gradients, temperature control, etc...
Transmitter
Duplexor
Receiver
Computer
Doing an NMR experiment involves setting up a sequence of events for the transmitter, receiver, and other devices attached to the NMR probe. For the transmitter each event involves setting the phase, amplitude, and frequency of the rf pulses, and for the receiver setting the gain and ADC timings. The product of this sequence of events is a signal transferred from the receiver to the computer. In the next chapter we will discuss the dierent ways the acquired signal can be processed.
Further Reading
A. Derome, Modern NMR Techniques for Chemistry Research, Pergamon Press, 1987. E. Fukushima and S. B. W. Roeder, Experimental Pulse NMR. A Nuts and Bolts Approach, Addison-Wesley Publishing, 1981. D. D. Tracante, Impedance: What it is, and Why It Must Be Matched, Concepts in Mag. Reson., 1, 73-92 (1989). P. J. Grandinetti, October 31, 2011
17.8. PROBLEMS
131
D. D. Tracante, Introduction to Transmission Lines: Basic Principles and Applications of Quarter-Wavelength Cables and Impedance Matching, Concepts in Mag. Reson., 5, 57-86 (1993).
17.8
Problems
1. Starting with the expression for the series tuned, parallel matched circuit impedance: R+i ZT = L 1 Ct Cm L 1 Ct 1 Ct
2
Cm R2
2
2 R2 + 1 C 2 Cm m L
nd the relationship between Cm and Ct when ZT = 50 . Assume L and R are xed. 2. Find the relationship between Cm and Ct when ZT = 50 for the parallel tuned, series matched circuit. Again, assume L and R are xed. 3. Given the time dependent voltage Vcoil (t) = kMeq cos 0 t et/T2 induced in a coil due to a precessing magnetization, use the Fourier transform
S ( ) =
0
S (t)eit dt,
to calculate the frequency dependent voltage induced in the coil, i.e., Vcoil ( ). 4. Given the following circuit
sample R
Receiver 50
(a) Calculate the frequency dependent voltage in the receiver Vrcvr ( ) assuming the frequency dependent voltage induced in the coil is Vcoil ( ) from the question 3 above. (b) How might you modify this circuit so that Vrcvr ( ) would be larger?
132
Part II
For the Workers
133
135
-X -Y X Y
136
Chapter 18
The Faraday Detector

Faradays Law tells us that the Electromotive Force (EMF, i.e., voltage) induced in a coil coil of wire of radius rcoil
B
Y
S
X
N
will be related to the change in magnetic ux with time, according to
E =
d . dt
(18.1)
Here E is the EMF and is the magnetic ux. In our example, the magnetic dipole vector of the top will be changing with time according to 137
138
Z
CHAPTER 18. THE FARADAY DETECTOR
0 t
Y X
(t) = || [sin cos(0 t + 0 ) ex + sin sin(0 t + 0 ) ey + cos ez ] .
(18.2)
where || is the length of the precessing vector, is the angle between the precessing vector and the z -axis, 0 is the initial phase of the precessing vector, and 0 is the precession frequency. With our precessing magnetic dipole at the origin the resulting time dependent magnetic eld produced at a point r from the dipole1 will be 0 3((t) er )er (t) , (18.3) 4r3 where er is a unit vector in the direction of r. The time-dependent magnetic ux through the surface of B (r, t) = the coil is given by the surface integral (t) = B (r, t) n da, (18.4)
where n is a unit vector normal to the face of the coil and da = rdrd.
Z
r d dr
Substituting Eq. (18.3) into (18.4) one obtains (t) =

1 We
0 4
rcoil 0
dr r2
d (t) n.
0
(18.5)
assume we are working in the near static zone. See Classical Electrodynamics, Jackson, 2nd Edition, Chapter 9.
139 If the coil is in the y -z plane, as shown above, then n = ex and the dot product in Eq. (18.5) selects out only the ex components in (t). Thus, we obtain x (t) = and the EMF induced in the coil will be Ex (t) = dx (t) 0 = 0 || sin sin(0 t + 0 ). dt 2rcoil (18.6) 0 || sin cos(0 t + 0 ), 2rcoil
From this signal we measure the precession frequency. Notice that if we put the coil in the x-z plane we would get the same result except the EMF signal is 90 out of phase, Ey (t) = 0 0 || sin cos(0 t + 0 ). 2rcoil
If we had two coils, one in the x-z plane, and one in the y -z plane, then we could measure not only the precession frequency, but also the sense (direction) of the precession. For example, consider the projection of the precessing magnetic eld vector in the x-y plane.
y Counter Clockwise r
+
Clockwise
r
+
x -r r x r y -r
time
x -r x
time
r y -r
time
time
Without knowledge of the initial orientation of the vector it would be impossible to determine the sense of the rotation from either Ex (t) or Ey (t) alone. Knowledge of both Ex (t) and Ey (t) together, however, allows one to distinguish between clockwise and counterclockwise circular motion. It is rare that NMR spectroscopists use two coils to detect the signal in the NMR experiment. We will revisit this concept, however, with the introduction of the rotating frame in section 6.1. Finally, we note that if the coil was in the x-y plane no EMF signal would be detected, since the ux would be time independent, i.e., z = and therefore Ez (t) = dz = 0. dt 0 || cos , 2rcoil
140
CHAPTER 18. THE FARADAY DETECTOR
Chapter 19
Vector Rotation
How can we mathematically describe the rotation of a vector through an angle d? We use the crossproduct dA = d A, where dA is the change in the vector A, and d = nd is a vector whose direction determines the rotation axis and whose length is the angle through which A is rotated1 . n is a unit vector in direction of rotation axis.
Z d = n d
d
' Y
Writing the cross-product as a determinant ex dA = d x Ax

1 In
ey d y Ay
ez d z Az .
this text we will use a right handed coordinate system and dene all positive angles of rotation as a counter-clockwise
rotation about the axis of rotation.
141
142 and expanding we obtain
CHAPTER 19. VECTOR ROTATION
dA = (Az dy Ay dz )ex + (Ax dz Az dx )ey + (Ay dx Ax dy )ez . Lets consider rotation about the z axis. Here we start with d = (0, 0, d) and obtain dAx = Ay , d dAy = Ax , d dAz = 0. d
These two coupled dierential equations can be decoupled using the spherical basis, dA1 1 = d 2 which have the solution A1 () = A1 (0)ei . In the Cartesian basis we have the solution A() = Rz ()A(0), where cos sin cos 0 0 Rz () = sin 0 0 . 1 dAx dAy i d d 1 = (Ay iAx ) = iA1 , 2
We summarize the transformation of A(0) into A() with Ax (0) Ax () = Ax (0) cos Ay (0) sin Rotation about the z -axis through an angle , (19.1)
Ay (0) Az (0)
Rz () Ay () = Ay (0) cos + Ax (0) sin Az () = Az (0)
A vector originally lying on the x-axis will be rotated about the z -axis will rotate counterclockwise in the x-y plane to +y , then to x, then to y , and back to +x.
Z -X
-Y
Notice that the complex spherical basis components, A1 , serve not just as a mathematical tool for solving our equation. Recall from section 18 that knowing both the x and y component of the precessing magnetization vector allows us to determine whether the sense of direction for the magnetization P. J. Grandinetti, October 31, 2011
143 precession is clockwise or counterclockwise. Thus, they provide a more compact means for describing length and orientation in the x-y plane, and sense of rotation about the z -axis. Additionally, note that other than getting multiplied by a phase factor ei , the A1 components are unaected by rotation about the z axis. Now lets consider a rotation about the x-axis, d = (d, 0, 0), where we start with dAx = 0, d and obtain the solution A() = Rx ()A(0), where 1 0 0 dAy = Az , d dAz = Ay , d
Rx () = 0 0
cos sin . sin cos
The transformation of A(0) into A() is Ax (0) Ax () = Ax (0) Rotation about the x-axis through an angle , (19.2)
Ay (0) Az (0)
Rx () Ay () = Ay (0) cos Az (0) sin Az () = Az (0) cos + Ay (0) sin
A vector originally on the y -axis and rotated about the x-axis will rotate counterclockwise in the y -z plane to +z , then to y , then to z , and back to +y .
Z -X
-Y
One can also show that A1 (0) A1 () =

Rx ()
1 1 A1 (0)(1 + cos ) + A1 (0)(1 cos ) iA0 (0) sin . 2 2
Finally, for a rotation about the y -axis, d = (0, d, 0), we start with dAx = Az , d dAy = 0, d dAz = Ax , d P. J. Grandinetti, October 31, 2011
144 and obtain A() = Ry ()A(0), where
CHAPTER 19. VECTOR ROTATION
Ry () =
cos 0 sin
0 1 0
sin 0 cos
and the transformation of A(0) into A() is Ax (0) Ax () = Ax (0) cos + Az (0) sin Rotation about the y -axis through an angle . (19.3)
Ay (0) Az (0)
Ry () Ay () = Ay (0) Az () = Az (0) cos Ax (0) sin
A vector originally lying on the z -axis and rotated about the y -axis will rotate counterclockwise in the x-z plane to +x, then to z , then to x, and back to +z .
Z -X
-Y X
One can also show that A1 (0) A1 () =

Ry ()
1 1 A1 (0)(1 + cos ) A1 (0)(1 cos ) A0 (0) sin . 2 2
Chapter 20
The Bloch Equations

20.1 The Laboratory Frame
In the frame of the laboratory the Bloch equations are given by dM = (t) M(t) [R] M(t) Meq , dt Here (t) = (1 )B(t), 1/T2 0 and [R] = 0 1/T2 0 0 (20.1)
0 0 1/T1
Meq = Meq ez
In the Bloch equations the [R] [M(t) Meq ] term describes the relaxation decay of x-y transverse magnetization and the growth of z longitudinal magnetization. The (t) M(t) term describes the change in M(t) as it precesses about the B(t) direction. Expanding Eq. (20.1) we obtain x (t) M [Mx (t)y (t) My (t)z (t)] Mx (t)/T2 My (t) = [Mx (t)z (t) Mz (t)x (t)] My (t)/T2 z (t) M [My (t)x (t) Mx (t)y (t)] [Mz (t) Meq ]/T1
For the magnetic eld vector we can consider two cases: (1) we include only the static magnetic eld and (2) we include both the static magnetic eld and alternating magnetic resonance eld. The rst one 145
146
is the easiest of the two to consider, so lets start there and write B(t) = B0 ez . Dening the Larmor frequency as 0 = (1 )B0 , we can expand the cross-product and the matrix multiplication of Eq. x (t) 0 My (t) Mx (t)/T2 M My (t) = 0 Mx (t) My (t)/T2 z (t) [Mz (t) Meq ]/T1 M whose solution gives Mx (t) [Mx (0) cos 0 t My (0) sin 0 t] et/T2 , (20.1) to get , (20.2)
My (t) Mz (t) from which we can also obtain
= [My (0) cos 0 t + Mx (0) sin 0 t] et/T2 Mz (0)et/T1 + Meq (1 et/T1 )
M1 (t) = M1 (0)ei0 t et/T2 . Starting with a magnetization vector pointing in any direction on the sphere, these equations will describe its counterclockwise (0 > 0) or clockwise (0 < 0) precession about the magnetic eld direction, its shrinking projection in the x-y plane due to T2 dephasing, and the growth of its z -component back to Meq value due to T1 relaxation. Notice that the M1 components are unchanged by the Larmor precession about the z axis other than the T2 decay and a phase factor ei0 t . If we include the alternating magnetic eld then the equations become more complicated, B(t) = 2B1 cos(|rf |t + )ex + B0 ez . We dene the peak to peak amplitude of the oscillating magnetic eld along ex as 2B1 to simplify our theoretical discussions later. Thus, we obtain x (t) M 0 My (t) Mx (t)/T2 My (t) = 0 Mx (t) 2B1 Mz (t) cos(rf t + ) My (t)/T2 z (t) M 2B1 My (t) cos(rf t + ) [Mz (t) Meq ]/T1
While these equations can be solved using the same approach, the solution is messy. By transforming into the rotating frame the solution to the problem is more straightforward. P. J. Grandinetti, October 31, 2011
20.2. THE ROTATING FRAME
147
20.2
The Rotating Frame
How do we mathematically relate the magnetization vector in the laboratory and rotating frames? First, recall that a vector is the same no matter what coordinate system is used to describe it. That is,
M = Mx ex + My ey + Mz ez = Mx ex + My ey + Mz ez .
Mx , My , and Mz describe the projections of the magnetization vector M onto the lab frame unit vector
axes, whereas Mx , My , and Mz describe the projections of the same magnetization vector M onto the
rotating frame unit vector axes. The unit vectors in the lab frame, ex , ey , and ez , are related to the
unit vectors in the rotating frame, e x , ey , and ez , according to ex = e x cos(rot t + rot ) ey sin(rot t + rot ), ey = e y cos(rot t + rot ) + ex sin(rot t + rot ),
(20.3)
ez = e z. The rotating frame frequency rot will need to match the spins precession frequency in both magnitude and sign, i.e., sense of direction. We have dened the rotating frame so that a positive rotating frame frequency, rot > 0, describes a frame rotating in a counterclockwise direction about the z axis. The initial rotating frame phase, rot , is a constant term with no experimental signicance, and we include it here only for a later theoretical convenience. The rate of change of M as seen in the rotating frame will dier from the rate of change seen in the lab frame, and the dierence is the eect of transforming into the rotating frame, d M = dM d M, and the time rate of change is obtained by dividing by dt d M dM = rot M. dt dt Note that rot = d/dt; the magnitude of rot is the angular frequency of the rotating frame and the direction of rot is the axis about which rotating frame rotates. Recalling dM/dt given by the Bloch equation we have d M = (t) M rot M [R]{M Meq }. dt Using the relation A B = B A we obtain d M = e (t) M [R]{M Meq }, dt (20.4)
148 where we dene an eective frequency by
e (t) = rot .
(20.5)
With Eq. (20.4), we can write the rate of change for the magnetization vector components in the rotating frame. Lets rst start with the static magnetic eld applied alone, and then consider the static eld and an oscillating magnetic resonance eld. First, we note that the static magnetic eld written in the rotating frame coordinate system is the same as in the lab frame coordinates, i.e., B0 = B0 e z lab frame = B0 e z, rotating frame
and also rot = rot e z so that e = (0 rot )ez . Also, note that the equilibrium magnetization vector
is Meq = Meq e z. Thus, the rate of change of M in the rotating frame is

x (t) M
(t)/T2 (t) Mx (rot 0 )My
My (t) z (t) M
(rot 0 )Mx (t) My (t)/T2 . (t) Meq ]/T1 [Mz
If we move into a frame rotating at rot = 0 , then all the oscillatory motion is removed and the magnetization vector simply shrinks in the x -y plane according to T2 , and grows along z according to T1 as expected. In the rotating frame we have the general solution for free precession
Mx (t)
Mx (0) cos t My (0) sin t et/T2
, (20.6)
My (t) Mz (t) from which we can derive
= My (0) cos t + Mx (0) sin t et/T2 Mz (0)et/T1 + Meq (1 et/T1 )
it t/T2 M e , 1 (t) = M1 (0)e
where = 0 rot . P. J. Grandinetti, October 31, 2011
20.3. THE DETECTOR COIL IN THE ROTATING FRAME
149
20.3
The Detector Coil in the Rotating Frame
Lets now consider the detector in the rotating frame. If we assume n = ex and plug the expression for ex in Eq. (20.3) into Eq. (18.5) we obtain x (t) = 0 Mx (t) cos(rot t + rot ) My (t) sin(rot t + rot ) . 2rcoil
Using Faradays Law, we take the time derivative of this expression. With rot 0 the time dependences
in Mx (t) and My (t) are too slow to contribute signicantly to the time derivative. Thus, we obtain
k (t) sin(rot t + rot ) + My (t) cos(rot t + rot ) , Sx (t) = Mx 2 2 where k =

0 2rot r . The NMR precession frequency, and thus the rotating frame frequency, for a coil
given nucleus is typically on the order of a few megaHertz. While this frequency can vary depending on its local environment, the range over which 0 varies is typically on the order of a few kiloHertz. Since low frequency detectors are less expensive than high frequency detectors, it is common practice to shift the frequency of the NMR signal down by |ref | before detecting. To further simplify matters, we often set |ref | equal to |rot |. We will discuss the instrumentation necessary to achieve this frequency shifting in Chapter 17. Mathematically, this process is equivalent to splitting the signal into two parts, the rst obtained by multiplying the signal detected in our receiver coil by 2 cos(|ref |t + ref ) to obtain k cos Sx (t) = Mx (t) sin(rot t+rot ) cos(|ref |t+ref )+My (t) cos(rot t+rot ) cos(|ref |t+ref ) , (20.7) 2 and the second obtained by multiplying by 2 sin(|ref |t + ref ) to obtain k sin Sx (t) = Mx (t) sin(rot t+rot ) sin(|ref |t+ref )+My (t) cos(rot t+rot ) sin(|ref |t+ref ) . (20.8) 2 How we obtain this component will depend on the sign of the gyromagnetic ratio [7]. Thus, we combine these two real signals in the spectrometer into a complex signal given by S cos (t) + iS sin (t) when < 0, x x S (t) = S cos (t) iS sin (t) when > 0. x x
(20.9)
Substituting Eqs. (20.7), (20.8) and (20.16) into the signal expression for < 0 in Eq. (20.9) one obtains
i[(rot +|ref |)t+ref ] S (t) = k M+1 (t)ei[(rot |ref |)tref ] + M . 1 (t)e
In this case we set |ref | rot , and nd that the rst term in this signal will be slowly oscillating with frequencies associated with our rotating frame magnetization vectors, and the second term will contain P. J. Grandinetti, October 31, 2011
150
fast oscillating terms with frequencies at twice the rotating frame frequency. Passing this signal through analog or digital lters to eliminate the fast oscillating terms we are left with
S (t) = kM+1 (t)eiref eiref t ,
(20.10)
where ref = rot |ref |. In practice, we set ref = 0 and think of the signal detected by the spectrometer as the signal detected in the rotating frame. This is a common and working assumption among many NMR spectroscopists, however, there are some situations, for example, applying pulses with dierent o-resonance frequencies, where ref may not be zero. Ignoring such situations, we can write the signal as
S (t) = kM+1 (t)eiref
for < 0.
(20.11)
Similarly, in the case of > 0 one obtains

S (t) = kM+1 (t)eiref
for > 0.
(20.12)
Dening a reference phase analogous to Eq. (20.17), ref = (sign ) ref , we can obtain a independent expression for the NMR signal,
S (t) = kM+1 (t)eiref .
(20.13)
(20.14)
The advantage of detecting a complex signal is that we can distinguish the sign of the oscillation
frequencies in M+1 (t), which, depending on the chosen rotating frame frequency, could be positive or
negative, that is, above or below rot .
20.4
The RF Pulse in the Rotating Frame
To complete our transformation into the rotating frame lets consider the case of a large static magnetic eld along z -axis and a smaller oscillating magnetic eld along the x-axis at a frequency |rf |, that is always dened as positive. In the lab frame we have B(t) = 2B1 cos(|rf |t + rf )ex + B0 ez . P. J. Grandinetti, October 31, 2011 (20.15)
20.4. THE RF PULSE IN THE ROTATING FRAME
151
How do we write this in the rotating frame coordinate system? We plug Eq. (20.3) into (20.15) and obtain B(t) = B1 {cos[(|rf | rot )t + (rf rot )] + cos[(|rf | + rot )t + (rf + rot )]} e x + B1 {sin[(|rf | rot )t + (rf rot )] sin[(|rf | + rot )t + (rf + rot )]} e y + B0 e z. For a rotating frame that follows our precessing magnetization vector we will set rot = 0 . Thus, to a good approximation, we can neglect the fast rot + |rf | oscillations in B(t) for positive Larmor frequencies, i.e., < 0, or neglect the fast rot |rf | oscillations for negative Larmor frequencies, i.e., > 0. The fast oscillation terms do not signicantly aect the motion of the magnetization vector. Setting |rot | = |rf |, keeping only the time independent terms, and multiplying Be by , we obtain B cos( )e B sin( )e when < 0, 1 rf rot x 1 rf rot y e = B cos( + )e + B sin( + )e when > 0.
1 rf rot x 1 rf rot y
Here we have a bit of a dilemma. Wed like to have one set of equations to describe the motion of the magnetization, regardless of the sign of the gyromagnetic ratio. This problem can be remedied by adopting the following convention [79]: First we set 0 when < 0, rot = when > 0, and obtain e |B | cos e + |B | sin e 1 rf y 1 rf x = |B | cos e |B | sin e
1 rf x 1 rf y
(20.16)
when < 0, when > 0.
Then we dene the phase of the nutation axis as = (sign ) rf , to yield an expression for e that is independent of the sign of ,
e = 1 cos e x + 1 sin ey ,
(20.17)
where we have dened the nutation or Rabi frequency1 1 = |B1 |.

1 Unlike
(20.18)
0 , we do not include the chemical shielding, , in this denition, since the eects of are usually insignicant
compared to the B1 inhomogeneity.
152
We now obtain the Bloch Equations in the rotating frame as2 (t) M sin Mx /T2 M 1 z x My (t) = 1 Mz cos My /T2 z 1 Mx sin + 1 My cos (Mz Meq )/T1 M (t)
. (20.19)
If we look at the solution of Eq. (20.19) over a short time scale, then we can assume that T1 and T2 are innitely long and drop the relaxation terms. If the phase of the rf is set to = 0, then we get a
2 Another
solution is to switch from a right- to left-handed coordinate systems when changing the sign of the gy-
romagnetic ratio. With such an approach, the Bloch equations would also appear identical, regardless of the sign of the gyromagnetic ratio, except the unit vectors would belong to coordinate systems with dierent handedness. Such an approach, however, is not convenient because of its impact on our later quantum mechanical description of NMR.
time = 0
pulse with phase = (sign )
Spins Nutate Counter-Clockwise around
Spins Nutate Counter-Clockwise around

frame=0
<0
>0
Spins Nutate Counter-Clockwise around Spins Nutate Clockwise around
frame=
-X* -X*
-Y*
Magnetization after /2 pulse
Y*
-Y*
Y*
X* -X*
Rotating Frame
-Y*
Magnetization after /2 pulse
X*
X*
Y*
Y*
-Y*
X*
-X*
-X
-X
Lab Frame
rf coil along x
-Y
neglected counter-rotating component
-Y
neglected counter-rotating component
Figure 20.1: To obtain a rotating frame where the phase of the nutation axis and the sense of the nutation are independent of the gyromagnetic ratio we select an initial phase for the rotating frame according to Eq. (20.16), and dene the phase of the nutation axis by Eq. (20.17). Illustrated above is an example of how the rf phase applied in the lab frame translates into the same phase in the rotating frame independent of the sign of gyromagnetic ratio. P. J. Grandinetti, October 31, 2011
20.4. THE RF PULSE IN THE ROTATING FRAME familiar set of equations for rotation about the x -axis
dMx = 0, dt dMy dMz = 1 Mz , and = 1 My , dt dt
153
whose solution yields the transformation of M(0) into M(t)

Mx
My Mz
Rx (1 t) My cos 1 t Mz sin 1 t Mz cos 1 t + My sin 1 t
Mx
rf pulse along x -axis ( = 0),
(20.20)
From these one can also derive

M 1 Rx (1 t)
1 1 M (1+cos 1 t) + M 1 (1 cos 1 t) iM0 sin 1 t 2 1 2
Similarly, if we set = /2, then we have another set of familiar equations for the rotation about the y -axis,
dMx = 1 Mz , dt dMy dMz = 0, and = 1 Mx . dt dt
which leads to the transformation of M(0) into M(t)

Mx
My Mz
Ry (1 t) My sin 1 t cos 1 t Mx Mz
Mx cos 1 t + Mz sin 1 t
rf pulse along y -axis ( = /2).
(20.21)
From these one can also derive

M 1 Ry (1 t)
1 1 M (1+cos 1 t) M 1 (1 cos 1 t) + M0 sin 1 t. 2 1 2
Lets also consider the eect of an rf pulse on complex vector components M 1 . A particularly useful and M result from these equations is that complex vector components M+1 1 are swapped by a pulse,
x M 1 M1 ,
and
M 1 M1 .
(20.22)
More generally, one can show that the eect of an rf pulse of arbitrary phase on the Cartesian basis vector components is 1 1 Mx M (1+cos t )+ ( M cos 2 + M sin 2 )(1 cos t )+ M sin sin t 1 1 1 x y z 2 2 x R (1 t) 1 2 My (1+cos 1 t) 1 My . 2 (My cos 2 Mx sin 2)(1 cos 1 t) Mz cos sin 1 t Mz Mz cos 1 t (Mx sin My cos ) sin 1 t (20.23) P. J. Grandinetti, October 31, 2011
154
From these expressions we nd the eect of a 1 t = /2 pulse of arbitrary phase to be 1 1 Mx Mx +2 ( Mx cos 2 + My sin 2)+ Mz sin 2 R (/2) 1 1 , My M ( M cos 2 M sin 2 ) M cos x z 2 y 2 y Mz My cos Mx sin and the eect of a pulse of arbitrary phase to be Mx cos 2 + My sin 2 Mx R () My My cos 2 + Mx sin 2 . M z Mz
(20.24)
(20.25)
Similarly, one can calculate the eect of an rf pulse of arbitrary phase on the spherical basis vector components, i i2 M+1 (1+cos 1 t)/2 M0 e (i sin 1 t)/ 2 + M M+1 e (1 cos 1 t)/2 1 R (1 t) i i M+1 e M0 cos 1 t M e ( i sin t ) / 2 (i sin 1 t)/ 2 + M0 1 1 i2 i M M+1 e (1 cos 1 t)/2 + M0 (1+cos 1 t)/2 e (i sin 1 t)/ 2 + M+1 1 (20.26) From these expressions we nd the eect of a /2 pulse of arbitrary phase to be 1 i 1 i2 i e e + M M M M+1 0 +1 2 2 2 1 R (/2) i i i i M M0 2 M+1 e 1e 2 i 1 1 i i2 M M + M e + M e 1 1 0 +1 2 2 2 and the eect of a pulse of arbitrary phase to be i2 M+1 M e 1 R () M0 M0 i2 M M e 1 +1
(20.27)
(20.28)
Dening a generic complex magnetization component Mq , where q = +1, 0, or 1, we can focus on
the eect of the rf pulse phase, , and condense Eq. (20.26) into the form
Mq 0

q1 =1
R (1 t)
1 iq1 cq0 ,q1 (t)Mq e , 1
rf pulse with phase .
(20.29)
where q1 = q1 q0 , and the coecient cq0 ,q1 (t) describes the eciency of the coherence transfer from q0 to q1 . Notice how the change in coherence order from q0 to q1 or q1 shows up in this expression.
Likewise, we can write the generic evolution of Mq in the absence of rf pulses as iq t Mq (t)Mq e , (t)z ,T1 ,T2
Free precession.
(20.30)
20.4. THE RF PULSE IN THE ROTATING FRAME
155
where the coecient (t) describes the decay or growth of the coherence or Zeeman order, respectively, due to relaxation processes during the evolution period. An important conclusion from Eqs. (20.29) and (20.30) is that coherence order, q , remains constant during unperturbed evolution periods, and can only change during a pulse. Both these expressions will play an important role when we discuss phase cycling in section 12. In nearly all remaining discussions in this text we will be working in the rotating frame.
156
Chapter 21
Signal Processing
21.1
21.1.1
21.1.1.1
The Fourier Transform

Fourier Transform Pairs
Gaussian Decay
An interesting aspect of a Gaussian decay is that its Fourier transform is also a Gaussian. Given a signal of the form S (t) = eit e its Fourier transform is
2 2
for
0 t ,
(21.1)
S ( ) =
0
eit e
2 2
eit dt =
exp 2 4 2
1 erf i
(21.2)

Real Imaginary Real F.T.
time time
Imaginary
157
158 21.1.1.2 Gaussian Echo
CHAPTER 21. SIGNAL PROCESSING
Given a signal of the form S (t) = eit e its Fourier transform is
2 2
for
t ,
(21.3)
S ( ) =
eit e
2 2
eit dt =
exp 42 2
(21.4)
Real Imaginary Real F.T. Imaginary
-time
+time
-time
+time
21.1.1.3
Square Pulse
Given a signal of the form S (t) = eit S (t) = 0 its Fourier transform is
T
for
0tT and t>T
(21.5) (21.6)
for t > T
S ( ) =
0
eit eit dt =
sin( )T 1 cos( )T +i
(21.7)
Real Real Imaginary F.T.

time time
Imaginary
~.6/T
21.1. THE FOURIER TRANSFORM
159
21.1.2
21.1.2.1
Fourier Transform Theorem Pairs

Time and Frequency Scaling S (at)
F.T.
1 S (/a) |a|
and S (a )
F.T.
1 S (t/a) |a|
(21.8)
21.1.2.2
Time and Frequency Shifting S (t t0 ) S ( )eit0

F.T.
and S ( 0 ) S (t)ei0 t
F.T.
(21.9)
The time and frequency shifting are both specic examples of the general theorem called the Convolution Theorem.
21.1.3
The Convolution Theorem
If S1 (t) and S2 (t) have the Fourier Transforms S1 ( ) and S2 ( ) respectively, then the Fourier transform of S1 (t) S2 (t) is equal to the convolution of S1 ( ) and S2 ( ).

ST =
S1 (t) S2 (t)et dt =
S1 ( )S2 ( )d
(21.10)
Lets consider the following example. We can think of the free induction decay of a single resonance as the product of the a non-decaying oscillation and an exponential decay, as shown below.
F. T.
S1(t) = eit
S1() 0
()
* S2(t) = e-t X
F. T.
S1() 0
()
=
S1(t) X S2(t) = eit e-t
F. T.
ST() 0
()
On the right, we see that in the frequency domain the nal spectrum is the convolution of the individual spectra of the non-decaying oscillation and an exponential decay. Using the convolution theorem it is easier to understand the eects of not acquiring the signal for a suciently long enough time. P. J. Grandinetti, October 31, 2011
160
F. T.
S(t) X 0 * 1 0 0
F. T.
There are many applications of the convolution theorem, and we will see many examples throughout this chapter and the next.
=
F. T.
0
21.1.4
Magnitude and Phase Spectrum
A complex spectrum S ( ) = A( ) + iD( ), can also be writen as S ( ) = M ( ) exp{i( )} where M ( ) = and ( ) = tan1 D ( ) A( ) (21.14) A( )2 + D( )2 (21.13) (21.12) (21.11)
M ( ) and ( ) are called the magnitude and phase spectrum, respectively. A nice feature of the magnitude spectrum is that it always contains positive peaks, even when the spectrum contains a number of resonances of mixed absorption and dispersion mode components. Also, the noise is a magnitude spectrum will be all positive. The disadvantage of using the magnitude spectrum is that resonance linewidths will be broader than in the pure absorption mode spectrum. For Lorentzian resonances, the full width at half height will be a factor of 3 greater in the magnitude spectrum compared to the pure absorption mode spectrum, as shown below. P. J. Grandinetti, October 31, 2011
21.2. BASELINE CORRECTION
161
Magnitude and Absorption
2 (Absorption) 2 3 (Magnitude)
Magnitude and Dispersion
On the other hand, if the spectrum has no dispersion mode components, as occurs in the Fourier transform of an spin echo signal, then the linewidths of the magnitude spectrum should be no larger than the pure absorption mode spectrum.
21.2
Baseline Correction
Sometimes your detector has a undesired d.c. signal that leads to a spurious peak at zero Hertz in the spectrum.
t=0
Real
F. T.
S0
S(t)
time Imaginary
Normally, this zero frequency peak is a spectrometer artifact, and needs to be eliminated. This can be achieved with a baseline correction. That is, we continue acquisition after the signal is dead so we can P. J. Grandinetti, October 31, 2011
162 measure S0 , and subtract it from the signal,
S (t) = S (t) S0 .
(21.15)
21.3
Zero Filling
Recall that the spacing between points in the frequency domain is related to the acquisition time in the time domain.
= 1/T,
where T is the total acquisition time. If we double the total acquisition time, then we half the spacing between frequency domain points. Zero lling is a means of halving the spacing between frequency domain points, without actually doubling the acquisition time. It is done by padding zeroes onto the end of the free induction decay. Zero lling in the time domain is equivalent to a trigonometric interpolation between points in the frequency domain. It is most useful to do when your time domain data doesnt contain the necessary 2n points, where n is an integer, for the fast Fourier transform algorithm.
In certain circumstances zero lling appears to give you a narrower linewidth (and improved resolution). Consider the case where the signal is not dead, i.e. truncated, before the end of the acquisition period, as shown below. P. J. Grandinetti, October 31, 2011
21.4. LINESHAPE TRANSFORMATION
163
# of ZeroFills 0
T time 2T
1 3T 2 4T 3 5T 4
time
time
time
time Frequency
Here a Fourier transform will give the convolution of the NMR spectrum from an untruncated time domain signal and the sinc function of a square pulse of length T . With discretely sample data, however, you will not see the sinc function oscillation in the frequency domain because its period matches the frequency point spacings. When you connect the dots, you see no sinc wiggles. On the other hand, if you zero lled to 2T , then when you connect the dots, you can see the sinc wiggles, and the linewidth will get narrower. Zero lling again will not give additional line narrowing, although the sinc function shape will become more well dened.
21.4
Lineshape Transformation
We know that the Fourier transform of a signal with an exponential decay is a Lorentzian lineshape. One could transform a Lorentzian into a Gaussian lineshape by multiplying the time domain signal by P. J. Grandinetti, October 31, 2011
164
an exponential growth that is opposite to its exponential decay and then by a Gaussian decay similar to the exponential decay. That is, given a time domain signal with an exponential decay S (t) = eit et and multiplying by e+t e
2 2
one obtains S (t) = S (t)e+t e

2 2
= eit e
2 2
whose Fourier transform is a Gaussian lineshape. Consider the example shown below.
F. T.
S(t)
S(t) e+t
F. T.
S(t) e+2t2
F. T.
Here, we multiply S (t) by e+t and then by e
2 2
and transform a Lorentzian lineshape into a Gaussian.
Notice how the noise overwhelms the spectrum in the intermediate step. This approach is sometimes used as a resolution enhancement as overlapping Lorentzian will appear more resolved when converted into Gaussian lineshapes.
21.5
Apodization
Given a free induction decay, one could improve the signal-to-noise ratio of its Fourier transform by replacing any signal after the d is dead with zeroes. P. J. Grandinetti, October 31, 2011
21.5. APODIZATION
165
S(t)
Replace with Zeroes Replace with Zeroes
S(t)
Of course, if you do this, you take the risk that theres a low intensity signal of narrow linewidth hidden under the noise that youll throw away. If ones condent, however, that only noise is being replaced with zeroes, then the sensitivity in the corresponding FT spectrum will improve. This approach is equivalent to multiplying the signal by a square pulse, so the FT spectrum is convoluted with a sinc function, and youll have lineshapes with some sinc function contribution. A more gentle way to achieve the same result is to multiply the data by an exponential decay instead of a square pulse.
e-t
X X
F. T.
S(t)
S(t) e-t
F. T.
This will reduce the noise signicantly at the tail of the decay and emphasize the signal at the beginning of the decay. When you transform this apodized signal the lineshape will have additional Lorentzian broadening instead of a sinc broadening. For maximum sensitivity enhancement, the optimum weighting function is the envelope of the signals decay. That is, if a signal is decaying exponentially, then the optimum weighting function is one that matches exactly the envelope of the signals decay. Of course, signals can have wildly dierent decays, but in every case, the optimum lter is the one that matches the P. J. Grandinetti, October 31, 2011
166
decay of the signal envelope. This optimum lter is often called a matched lter. In a real spectrum it is often not possible to choose the optimum lter for resonances of dierent widths, so some compromise is usually reached.
21.6
Signal-to-Noise Ratio
To calculate the signal-to-noise ratio of a spectrum or resonance one rst identies a region of the spectrum containing only baseline signal, i.e., no resonances. From the data in this region the standard deviation of the noise is calculated, 1 N 1
N
noise = where
(yi y )2 ,
i
(21.16)
y=
1 N
yi .
i
(21.17)
The average value of the baseline, y , should be zero to withing a few standard deviations units. If not there may be a small roll in the baseline. The baseline noise region should also be reasonably large for an accurate measure of noise . Using Eq. (21.16), the signal-to-noise ratio is dened as S 1 Smax = N rms 2 noise An older approach, is to measure the peak-to-peak noise value from the baseline region. Npp = ymax ymin y, and calculating the signal-to-noise ratio according to S N =
pp
(21.18)
(21.19)
5 Smax . 2 Npp
(21.20)
Further Reading
A. G. Marshall and F. R. Verdun, Fourier Transforms in NMR, Optical, and Mass Spectrometry. A Users Handbook, Elsevier, 1990.
Chapter 22
The Quantum Picture

There are additional measurement peculiarities with spin angular momentum that we still need to examine. Lets go back to the spin 1/2 case and take the z = + /2 group from our measurement above and measure their x components. It turns out that 50% of this group will have a x = + /2 and 50% will have x = /2. Now, does this mean that 50% of our original z = + /2 group had nuclear magnetic dipole vectors with both x = + /2 and z = + /2? Not exactly. Because if we took the 50% of the original z = + /2 nuclei that were measured to have x = + /2, and then measured their z component again, we would nd that only 50% of this group has z values of + /2 and the other 50% has z = /2!
z = + h/2 group measure x

50% 50%
x = + h/2 group measure z

50% 50%
x = h/2 group
z = + h/2 group
z = h/2 group
167
168
CHAPTER 22. THE QUANTUM PICTURE
It appears that for whatever reason there is a fundamental limitation at the microscopic level that keeps us from knowing x and z simultaneously. You can say that measuring x destroys any previous information about z . Thus, if you measured the z value of a nucleus and found it to be z = + /2, then you cannot say with certainty that a subsequent measurement of x would be + /2 instead of /2. Dont let this quantum measurement weirdness, however, make you think that everything is uncertain. After a measurement of z = + /2 you can always be certain that there is a 50% probability of measuring x = + /2 instead of x = /2. The main idea is that while we cannot always be certain about the outcome of a particular measurement, we can always be certain about the probability of measuring each of the possible observable outcomes. This idea of quantized observables is also not at all in conict with our picture of a spinning magnetic dipole vector undergoing continuous precession about an externally applied magnetic eld. For example, if we measured a nucleus to have z = + /2 and then we apply a -pulse along the x direction and remeasure z we can be certain that the probability of measuring z = /2 is 100%.
0%
z = + h/2 group
z = + h/2 group
()x
measure z
100%
z = h/2 group
Likewise, if we had instead applied a (/2)x pulse and remeasured z we would be certain that there will be a 50% probability of measuring + /2 instead of /2.
50%
z = + h/2 group
z = + h/2 group
(/2)x
measure z
50%
z = h/2 group
22.1. DIRACS FORMULATION OF QUANTUM MECHANICS
169
Similarly, if we applied a (/2)x pulse and instead measured y we can be certain that there will be a 100% probability of measuring y = /2.
0%
y = + h/2 group
z = + h/2 group
(/2)x
measure y
100%
y = h/2 group
Quantum mechanics tells us how to take this classical picture of spin precession and use it in terms of the probabilities of measured outcomes.
22.1
22.1.1
Diracs Formulation of Quantum Mechanics

Probabilities and Inner Products
Dirac showed that the way to predict these probabilities was to represent all the information that can possibly be known about a system in terms of a vector, called a state vector, in a complex vector space called a ket space. The number of dimensions in this ket space is equal to the total number of possible outcomes for an observable quantity. For a spin 1/2 nucleus there are only two observable values for x , y , or z , therefore the complex space is two-dimensional.
|z;>
dimension associated with z = - h /2
c-
dimension associated with z = + h /2

{
c+
state vector |> = c+ |z;+>
+ c- |z;>
|z;+>
170
In this example, |z ; + and |z ; represent vectors in the ket space corresponding to the z = + /2 and z = /2 outcomes, respectively. With these denitions the state vector is described entirely in terms of its projections onto the axes in the ket space. For example, in the case of a spin 1/2 nucleus we have | = c+ |z ; + + c |z ; . The fundamental idea here is that the projection of the state vector onto the axis associated with an observable outcome is a complex number that can be related to the probability of measuring that particular observable outcome. It must be emphasized that the coecients c+ and c , i.e., the projections of | onto |z ; + and |z ; are complex numbers. That is, c+ = a+ + ib+ = r+ ei+ ,
1 2 2 b+ /a+ . The square of the magnitude of each complex coecient = a2 where r+ + + b+ , and + = tan 2 2 is the probability, i.e., r+ is the probability of measuring z = + /2. Likewise r is the probability of
measuring z = /2. As the system evolves in time, the state vector changes direction in this complex space, and with it the corresponding probabilities of measuring each possible observable outcome. You might be wondering, why do we need a complex space? Why not a real space, where the projections are the probability? An advantage of the complex space is that we can use the complex projections of the state vector to specify not only the probabilities for z outcomes, but also for the x and y outcomes. In this complex space, the orthogonal ket vectors representing the x = + /2 and x = + /2 outcomes, i.e., |x; + and |x; , can be written as linear combinations of |z ; + and |z ; . The same is true for y and its orthogonal ket vectors |y ; + and |y ; . Later, we will show how the relationships among ket vectors associated with dierent observables can be derived. Generally, | contains information about the probability of measuring all possible observable outcomes, and can usually be written completely in terms of the orthogonal ket vectors associated with all the possible outcomes for a single observable. In the specic case of a spin 1/2 nucleus the ket space is two dimensional. For greater than spin 1/2 nuclei, or for coupled nuclei, the ket space will have more dimensions. For an N dimensional ket space the state vector is given by | =
j
cj |j ,
where j = 1, . . . , N.
Dirac also introduced another complex vector space that is dual to the ket space. It is called bra space. P. J. Grandinetti, October 31, 2011
171
<z;|
dimension associated with z = - h /2
state vector
c* -
dimension associated with z = + h /2

This state vector in this space contains no additional information to what is already contained in the state vector in ket space. You can think of the bra space as the complex conjugate of the ket space, and vice versa. The projections of the state vector onto the axes in the ket space, i.e., c and c+ are the complex conjugates of the projections of the state vector onto the axes in the bra spaces, i.e., c and c+ . For example, if c = a + ib , then c = a ib .
One advantage of having a dual bra and ket space is that for every ket | there exists a corresponding bra | so we can dene an inner product of a bra and a ket as | inner product = ( |) (| ) bra ket
The inner product is a complex number. As mentioned earlier, the dimensions associated with each possible observable outcome are orthogonal. Therefore, in our two dimensional complex space for the spin 1/2 nucleus, the kets |z ; + and |z ; (and likewise the bras z ; +| and z ; |) form an orthogonal basis set, so their inner products are1 z ; +|z ; + = z ; |z ; = 1, and z ; +|z ; = z ; |z ; + = 0.
The same is true for the complex unit vectors associated with the x and y observable outcomes.
1 Think
of this in the same way as the dot product (also called inner product) between the unit vectors ex , ey , and ez .
That is, ex ex = ey ey = ez ez = 1, but ex ey = ex ez = ey ez = 0.
{
c* +
<| = <z;+| c* + + <z;| c* -
<z;+|
172 Another property of inner products is that
| = | . The * means take the complex conjugate. For example, if | = a + ib, then |
= a ib.
We can use this formalism to relate the projection of the state vector onto the axis associated with an observable outcome to the probability of measuring that particular observable outcome. If z ; +| and |z ; + dene axes associated with the z = + /2 observable outcome, then the probability of measuring z = + /2 is Probability = z ; +| |z ; + = | z ; +| |2 . Substituting in our expression of Eq. (22.1.1) and using the property of orthogonal vectors, we obtain Probability = = z ; +| c+ |z ; + + c |z ; c+ z ; +|z ; + + c z ; +|z ;
z ; +| c + + z ; | c |z ; + , c + z ; +|z ; + + c z ; |z ; + ,
= c+ c +. Likewise the probability of measuring z = /2 will be2 Probability = z ; | |z ; = | z ; | |2 = c c . Thus, if we measured a nucleus to have z = + /2, i.e., | = |z ; + and we wanted to know the probability that the system would jump into |x; + when we measure x then the probability would be represented by 1 Probability of measuring x = + 2 is | x; +|z ; + |2 .
22.1.2
Operators and Outer Products
Now that weve dene how our state vector in a complex space can be used to obtain the probability of measuring a particular observable outcome, we move on to the next task of calculating the average or expectation value for a given observable. An observable (or measurable) quantity is represented by an operator in the complex vector space. An operator acts on a ket from the left or a bra from the right, for example |A,
2 Remember
| . or A
that the product of a complex number and its complex conjugate is real.
173
is an operator. When an operator acts on a state, the state can be transformed into another Here A state or linear combination of states in the complex space. There are a special set of state, known as eigenstates of an operator which are not transformed into other states when that operator acts on them. When an operator acts on its eigenstates, or more specically, eigenkets or eigenbras, the same state remains multiplied by a real number = aj |aj aj |A |aj = aj |aj . or A
and aj is an eigenvalue Here aj | and |aj represent eigenstates (eigenbras and eigenkets, repectively) of A . For example of the operator A z |z ; + = 1 |z ; + , 2 1 and z |z ; = |z ; , 2
here we recognize |z ; + and |z ; as the eigenstates of the operator z . The eigenvalues of z are /2. As we saw earlier, you cant know both x and z simultaneously, and in quantum mechanics this translates into saying that they do not have the same eigenstates. That is, 1 x |x; + = + |x; + , 2 but x |z ; + = a+ |z ; + + a |z ; When x operates on a z eigenstate it transforms it into a linear combination of z eigenstates. The same is true of z operating on an x eigenstate. Also, the eigenstates of y are not shared by either x or z . Thus, we say that x , y , and z are incompatible observables. Using the bra and ket spaces we can also determine an outer product of a bra and ket: | | outer product = (| ) ( |) bra ket
An outer product of a bra and a ket is regarded as an operator and is fundamentally dierent from the inner product | . In our two dimensional bra-ket space, there are only 4 independent outer products or operators, |z ; + z ; +|, |z ; z ; +|, |z ; + z ; |, and |z ; z ; |. All other operators and outer products in this two-dimensional bra-ket space must be a linear combination of these four outer products or operators. For example, z = 1 2 |z ; + z ; +| |z ; z ; | , P. J. Grandinetti, October 31, 2011
174 where, z |z ; + = = =
1 |z ; + z ; +| |z ; z ; | |z ; + , 2 1 1 |z ; + z ; +|z ; + |z ; z ; |z ; + , 2 2 1 |z ; + . 2
Another example is the identity operator = |z ; + z ; +| + |z ; z ; |, E where |z ; E = = = |z ; + z ; +| + |z ; z ; | |z ; , |z ; + z ; +|z ; + |z ; z ; |z ; , |z ; .
Later we will derive similar expressions x and y . For an N -dimensional bra-ket space there will be N 2 operators, which include
N
= E
j
|j j |.
It is also important to examine the algebra of operators. Operator addition is associative and commutative, + (B +C ) = (A +B ) + C, A and +B =B + A. A Operators multiplication, on the other hand, is associative but, generally, is not commutative. That is, (B C ) = (A B )C =A B C, A but B =B A A in general.
B and B A is called the commutator of A and B and is written as The dierence between A B ] = A B B A [A, P. J. Grandinetti, October 31, 2011
22.1. DIRACS FORMULATION OF QUANTUM MECHANICS B ] = 0 then A and B are said to commute. Note also that [A, B ] = [B, A ]. If [A, While an operator always acts on a ket from the left and on a bra from the right, that is | , A or |A,
175
these two operations may not give you a dual ket and bra. In general, a dual ket and bra are preserved if
dual | , A |A
is called the adjoint of A . The adjoint of an operator is formed by reversing the outer products where A in its expansion and taking the complex conjugate of the expansion coecients. Given
N N N N
= A
j k
bj,k |j k |,
then
= A
j k
b j,k |k j |.
=A then A is called a Hermitian operator. All operators corresponding to physical observables, If A such as x , y , and z , must be Hermitian.
22.1.3
Expectation Values
With our denitions of the outer and inner products we can calculate the expectation value A of a . We start with the denition Hermitian operator A | . A = |A We can rewrite this by inserting the identity operator A E | , A = |E in terms of the eigenstates of A , and then expanding E
N
= E
j
|aj aj |.
We then get A = |
j N
|aj aj | A
|ak ak | | =
k j k
|ak ak | . |aj aj |A
Recalling that aj |A|ak = aj |ak |ak = ak aj |ak = ak j,k , P. J. Grandinetti, October 31, 2011
176
that is, aj |ak = 0 unless j = k . Then aj |aj = 1, and our expectation value becomes
N N
A =
j
aj |aj aj | =
j
aj | |aj |2 .
and | |aj |2 is the probability of measuring the eigenvalue aj . Clearly Here aj is an eigenvalue of A the expectation value of A is simply a weighted average of aj values. Lets consider the z operator. Its expectation value is given by z = | z | . To calculate this expectation value you take the inner product of the state bra with the state ket after youve operated on either the state ket or bra with the z operator. Given | = c+ |z ; + + c |z ; , then | z | =
z ; +|c + + z ; |c z c+ |z ; + + c |z ;
and
| = z ; +|c + + z ; |c ,
1 |z ; so Recall that z |z ; = 2
| z |
= c+ c + ( /2) z ; +|z ; + + c c+ ( /2) z ; +|z ; , + c+ c ( /2) z ; |z ; + + c c ( /2) z ; |z ; .
Since the |z ; + and |z ; states are orthogonal, we are left with z = | z | = 1 (c+ c + c c ). 2
Since c+ and c are continuous complex variables, z can take on continuous values, even though in a single measurement only values of /2 are measured. We should also note that c+ and c are constrained by the fact that the total probability cannot exceed unity, | =
z ; +|c + + z ; |c
c+ |z ; + + c |z ;
= c+ c + z ; +|z ; + + c c+ z ; +|z ; + c+ c z ; |z ; + + c c z ; |z ; ,
which requires
| = c+ c + + c c = 1.
For an N-dimensional complex space we require cj c j = 1.

j
22.1. DIRACS FORMULATION OF QUANTUM MECHANICS To calculate x we again start with x = | x | .
177
Weve been working, however, in the eigenstates of z . To get x while continuing to use the z eigenstates, well need to know the eect of applying the x operator on the z eigenstates. How do we do this? We start by recalling that z = 1 2 |z ; + z ; +| |z ; z ; | ,
and realizing that we can write the same expression for x in terms of its eigenstates x = 1 2 |x; + x; +| |x; x; | .
To convert this expression in terms of eigenstates of z we need to nd the relationship between the eigenstates of z and x . To this end we start with |x; + = |x; + , insert the identity operator, |x; + , |x; + = E = |z ; + z ; +| + |z ; z ; | to get and then use E |x; + = |z ; + z ; +|x; + + |z ; z ; |x; + . In this expression, |x; + = z ; +|x; + |z ; + + z ; |x; + |z ; ,
complex coe. z eigenket
complex coe. z eigenket
we have achieved our primary task of expressing the |x; + state in terms of the |z ; + and |z ; states. Similarly, we obtain |x; = z ; +|x; |z ; + + z ; |x; |z ; . Our next task is to determine the value of these complex coecients. Recall that the coecients can be written z ; +|x; + = a + ib = rei , P. J. Grandinetti, October 31, 2011
178 where r2 = = = and z ; +|x; + z ; +|x; + (a + ib)(a ib), a2 + b2 ,
= z ; +|x; + x; +|z ; + ,
= tan1 b/a. If z is measured for a group of nuclei known to be x = + /2 the wed obtain a 50% split into z = /2.
x = + h/2 group measure z

50% 50%
z = + h/2 group
z = h/2 group
This tells us the probability that a system initially in the x = + /2 state will jump into either z = + /2 or z = /2 when z is measured is 50%. In quantum mechanics, this implies | x; +|z ; + |2 = 1/2, and | x; +|z ; |2 = 1/2.
Now, we know the magnitude of these two complex numbers, but not their phases, 1 x; +|z ; + = ei1 , 2 We can now write 1 1 |x; + = ei1 |z ; + + ei2 |z ; , 2 2 and 1 1 |x; = ei3 |z ; + + ei4 |z ; . 2 2 P. J. Grandinetti, October 31, 2011 and 1 x; +|z ; = ei2 . 2
22.1. DIRACS FORMULATION OF QUANTUM MECHANICS In quantum mechanics its only the relative phase that counts, so |x; + |x; = = 1 1 |z ; + + ei+ |z ; , 2 2 1 1 i |z ; + + e |z ; , 2 2
179
where + = 2 1 and = 4 3 . Using the additional constraint that x; +|x; = 0 we can show that ei+ = ei = eix and thus |x; + = 1 1 |z ; + + eix |z ; , 2 2 (22.1) |x; = 1 1 |z ; + eix |z ; . 2 2
Following the same approach one can derive 1 1 |y ; + = |z ; + + eiy |z ; , 2 2 (22.2) 1 1 |y ; = |z ; + eiy |z ; . 2 2 Now, is there anyway we can know x and y ? We know that if we measure y in in a group of spins selected with x = + /2 nuclei that we would get 50% of y = + /2 and 50% of y = /2. Thus, we can write | y ; |x; + |2 = | y ; |x; |2 = 1/2. Using this constraint and Eqs. (22.1) and (22.2) we obtain 1 1 ei(x y ) = 1/ 2, 2 which can be true only if x y = /2. The adopted convention is to take x = 0 and y = /2 (This choice agrees with right handed coordinate system). Therefore, we have |x; 1 1 |z ; + |z ; , 2 2 (22.3) |y ; = 1 i |z ; + |z ; , 2 2 P. J. Grandinetti, October 31, 2011
180 and nally x , y , and z can be written x = 1 2 1 2 1 2
|z ; + z ; | + |z ; z ; +| ,
i|z ; + z ; | + i|z ; z ; +| ,
|z ; + z ; +| |z ; z ; | .
As in section 6, we will also nd it convenient to dene the magnetic dipole moment operators in terms of the spherical basis vectors, where 0 = z , 1 and 1 = ( x i y ). 2
In the case of spin I = 1/2 the complex components 1 become 1 1 = |z ; z ; |. 2 We nish our calculation of x and y , starting with x , x = = = + | x | ,
z ; +|c x c+ |z ; + + c |z ; + + z ; |c
c+ c x |z ; + + c c x |z ; + z ; +| + z ; +| c+ c x |z ; + + c c x |z ; . z ; | z ; |
Using our denition above we nd z ; +| x |z ; + = z ; | x |z ; = 0, Thus, we obtain x = Similarly, one nds y = and also 1 = c c , 2 P. J. Grandinetti, October 31, 2011 1 (ic c + ic+ c ), 2 1 (c c + + c+ c ). 2 and z ; +| x |z ; = z ; | x |z ; + = /2.
22.1. DIRACS FORMULATION OF QUANTUM MECHANICS Summarizing these results for the Cartesian operators we have: x = 1 [c c + + c+ c ], 2 1 [ic c + ic+ c ], 2 1 [c+ c + c c ]. 2
181
(22.4)
Although the c and c+ are complex numbers, we know that x , y , and z must be real. So lets make certain of this. Writing c+ = aei and c = bei we have
2 c+ c + =a 2 c c =b

i( )
real, real, complex, complex.
c+ c
= abe
i( ) c c = abe
But since
i( ) c c + ei( ) ] = a b 2 cos( ), + + c+ c = ab[e
and
i( ) ic c iei ) ] = a b 2 sin( ), + ic+ c = ab[ie
we can write x y z = = = 1 a b 2 cos , 2 1 a b 2 sin , 2 1 (a2 b2 ), 2
where = . All three expectation values are real as expected.

2 2 Recalling the constraint c+ c + + c c = 1 which gives a + b = 1, we set a = cos /2 and b = sin /2
to further obtain a 2 + b2 and a 2 b2 = cos2 (/2) sin2 (/2) = cos , 2 cos(/2) sin(/2) = sin , P. J. Grandinetti, October 31, 2011 = cos2 (/2) + sin2 (/2) = 1,
2ab =
182 which nally gives x =
1 sin cos , 2 1 sin sin , 2 1 cos , 2
z and also
1 1 = sin ei . 2 2 We have not only shown that the expectation values of x , y , and z are real, but also that the expectation values describe a vector in a three dimensional expectation value space of length /2 whose direction is specied by the spherical coordinates and .
Z
<z>
Y
<y> <x>
This x-y -z space created from the expectation values of x , y , and z , provide a somewhat more intuitive picture for the state of the nuclear magnetic moment than the state vector in bra-ket space. If this expectation value vector was along z ,
Z
Y X
then the expectation value of x and y is zero, and z is /2. If the vector was in the x-y plane, midway between x and y axes,
Z
Y X
22.2. SPIN > 1/2 NUCLEI then the expectation value of x and y are
1 2
183 /2 and z = 0.
It is tempting to completely equate this space with the real space of our microscopic magnetic top. This is not completely correct, but from a practical point of view works out okay. Remember, that in this expectation value space the vector is of length /2, but in the real space the nuclear magnetic dipole vector length is 3 /2. In the real space, you can imagine that the nuclear magnetic dipole vector can have any arbitrary orientation of and , but whenever x , y , or z is measured only 1 2 can be measured. In the expectation value space you can imagine that the vector has exactly the same orientation of and as the nuclear magnetic dipole vector, but its length is shorter ( /2) and its projection on the x, y , and z axes tells you average value you would measure if you performed many measurements on identically prepared nuclei.
22.2
Spin > 1/2 Nuclei
In all discussions that follow we will drop the label z in our bra-ket space basis vectors, and label the states by the m value, that is, |m = |z ; m and m| = z ; m|,
and, unless noted otherwise, assume we are working with the z eigenstates. As we noted earlier, nuclear magnetic dipole moment vector is proportional to the nuclear spin angular momentum vector according to = I. We dene the magnetization vector in the NMR experiment as the vector sum of the expectation value vectors of the individual nucleis magnetic moments,
N N
M(t) =
j
(t)
=
j
I(t) j .
(22.5)
For an arbitrary spin, I , the angular momentum vector operator, I, is given by I I1 1 1 I I (I +1) m(m 1)|m 1 m| + 2 I (I +1) m(m +1)|m +1 m| x 2 m =1 I m = I I I1 i i I (I +1) m(m 1)|m 1 m| 2 I (I +1) m(m +1)|m +1 m| Iy = 2 . m=I m=1I I m|m m| Iz
m=I
(22.6)
184
These three operators follow the cyclic commutation relations x , I y ] = iI z , [I y , I z ] = iI x , and [I z , I x ] = iI y . [I (22.7)
So far, we have only dened the cartesian angular momentum operators in terms of the outer product operators. It is also common to dene the angular momentum operators in terms of the spherical basis vectors with3 0 = I z , I 1 1 = y ). and I (Ix iI 2 (22.8)
, dened as Alternatively, the raising and lowering operators, I = 2I 1 = I x iI y , I , with I 1 . The raising are often used. Be careful not to confuse the raising and lowering operators, I and lowering operators are dened in terms of outer products as = 1 I 2 and + = 1 I 2
I 1 I
I (I + 1) m(m 1) |m 1 m|,
m=1I
I (I + 1) m(m + 1) |m + 1 m|.
m=I
More generally, a nucleus of spin I can be described by irreducible spherical tensor operators, Tl,p , of rank l and order p. The rank of the irreducible tensor operators ranges from l = 0 to l = 2I . For a given rank, l, there are 2l + 1 operators with p values of p = l, (l 1), . . . , (l 1), l. They are called irreducible because of the invariance of the rank, l, during rotations about the x, y , or z axis. In terms of the spin angular momentum operators we have 0,0 = 1 T E, 2I + 1 1,p = and T 3 p , I (22.9)
I (I + 1)(2I + 1)
where p = 1, 0, or +1. In terms of outer products, the irreducible spherical tensor operators for an arbitrary spin I are given by l,p = T
3 Conversely,
2l + 1 2I + 1 m
I, m2 |I, l, m1 , p |I, m2 I, m1 |,
(22.10)
1 ,m2
we have 1 x = +1 I 1 I I 2 and i y = +1 + I 1 . I I 2
22.2. SPIN > 1/2 NUCLEI with an inverse relationship of |I, m2 I, m1 | =

l
185
2l + 1 l,p . I, m2 |I, l, m1 , p T 2I + 1
(22.11)
Here I, m2 |I, l, m1 , p is called the Clebsch-Gordon Coecient. Two useful aspects of the irreducible spherical tensor operator are the selection rule 1 when m = m m1 |Tl,p |m2 = m2 + p|Tl,p |m2 m1 ,m2 +p , where m,m = 0 when m = m , and the commutator z , T l,p ] = pT l,p . [I Note that the Tl,p operators with p = 0 are not Hermitian. For a spin I = 1 nucleus, there are three observable outcomes, m = 1, 0, and +1. The state z eigenstates will be function expanded in terms of the I | = c+1 |+1 + c0 |0 + c1 |1 . (22.14) (22.13)
(22.12)
In this three dimensional bra-ket space, there will be 9 independent outer products or operators. All other operators and outer products in this three-dimensional bra-ket space will be a linear combination of these nine outer products or operators. Using Eq. (22.6) one could obtain expressions for the angular momentum vector operator. These three operators plus the identity operator give us four operators, yet we know that there must be 9 operators to completely describe the spin 1 nucleus. Going beyond spin I = 1, there will be (2I + 1)2 outer products or operators for a spin I . What do these remaining operators physically represent? The answer lies within the structure of the nucleus. With increasing total angular momentum of an isotope additional degrees of freedom can be described in terms of higher order magnetic and electric multipole moments. A nucleus with spin I can have up to l = 2I multipole moments. Every multipole moment, l, will bring 2l + 1 additional degrees of freedom to the nucleus in the NMR experiment. The electric quadrupolar moment possessed by nuclei with spin I > 1/2 is characterized by a traceless second-rank tensor, q, and can be related to the second rank irreducible tensor operators according to q 2,p = 3 eQI T2,p 2 I (2I 1)
where e is the charge on the proton, QI is the quadrupole moment constant whose nuclear state is characterized by I and other parameters that we generically denote by . The ve T2,p elements are P. J. Grandinetti, October 31, 2011
186 given by 1 2 z 2,0 = 2 , 3I T I 6
2,1 = 1 I z I + I I z , T 2
2,2 = 1 I 2 . T 2
In terms of Cartesian basis vectors, we can also dene the second-rank tensor operator T xx T xy T xz T = yx T yy T yz T T , zx T zy T zz T where xx + T yy + T zz = 0, T and ij = T ji T where i, j = x, y, z.
The connection between the spherical tensor and Cartesian operators is xx = 1 (T 2,2 + T 2,2 ) T 2 yy = 1 (T 2,2 + T 2 zz = T
2 3 T2,0 , 1 T , 6 2,0 1 2,2 ) T , T 6 2,0
2,2 T 2,2 ), xy = T yx = i (T T 2 xz = T zx = 1 (T 2,1 T 2,1 ), T 2 yz = T zy = i (T 2,1 + T 2,1 ), T 2 (22.15)
and the reverse relations are 2,0 = T

3 2 Tzz ,
2,1 = (T xz iT yz ), T 2,2 = 1 (T xx T yy i2T xy ). T 2 We see that the traceless second-rank quadrupole moment tensor represents ve additional degrees of freedom for our nuclear spin.
22.3
Coupling between nuclei
When there are couplings between nuclei, our quantum mechanical system must be expanded to include every dipole coupled nuclear spin. Lets considering two coupled spin 1/2 nuclei. For each spin 1/2 x , I y , and I z . nucleus there are only two possible outcomes for I |1 |2 = = c+ |+1 + c |1 c+ |+2 + c |2
(2) (2) (1) (1)
for the rst spin, for the second spin,
When the two spins are coupled they form a single system with more possible observable outcomes, and therefore a higher-dimensional state space. We classify the coupling between nuclei with the same gyromagnetic ratio as homonuclear, and between nuclei with dierent gyromagnetic ratios as heteronuclear. We further classify couplings between P. J. Grandinetti, October 31, 2011
22.3. COUPLING BETWEEN NUCLEI
187
nuclei as weak (AB system) or strong (AX system) depending on whether the dierence in resonance frequencies of the two nuclei is greater or less, respectively, than the size of the coupling between them. Clearly, all heteronuclear couplings will be weak, whereas, homonuclear couplings depend on the system, and can sometimes even be intermediate between strong and weak.
22.3.1
Weak Couplings - AX system
When the coupling between the spins is weak, the new state space is best described by a direct product space of the two spin 1/2 complex spaces, |I1 , I2 , m1 , m2 = |I1 , m1 |I2 , m2 . Here |I1 , m1 and |I2 , m2 are the state spaces for the rst and second spin, respectively. That is, m1 = 1/2 and m2 = 1/2. The symbol means take the direct product. Thus, for two coupled spin 1/2 nuclei the product state space is four dimensional, | = c+1 ,+2 |+1 , +2 + c+1 ,2 |+1 , 2 + c1 ,+2 |1 , +2 + c1 ,2 |1 , 2 , and there are 16 possible outer products (operators) possible, that is, |I1 , I2 , m1 , m2 I1 , I2 , m1 , m2 |. The 16 possible Cartesian operators can be constructed by a direct product of the 4 Cartesian operators of each spin subspace. That is, 1 E 2 , E 1,y E 2 , I 1,z E 2 , I 1 I 2,x , E 1,y I 2,x , I 1,z I 2,x , I 1 I 2,y , E 1,y I 2,y , I 1,z I 2,y , I 1 I 2,z , E 1,y I 2,z , I 1,z I 2,z . I
1,x E 2 , I 1,x I 2,x , I 1,x I 2,y , I 1,x I 2,z , I
1,x I 2,x is written I 1,x I 2,x . Also, E 1 and E 2 are usually dropped, Usually the symbol is dropped, e.g., I 1,x is understood to mean I 1,x E 2 . This way of representing the operators of the spin for example, I system is called the Product Operator Formalism. Alternatively, we can dene the direct products of the raising and lowering operators 1 E 2 , E 1,z E 2 , I 1,+ E 2 , I 1 I 2, , E 1,z I 2, , I 1,+ I 2, , I 1 I 2,z , E 1,z I 2,z , I 1,+ I 2,z , I 1 I 2,+ , E 1,z I 2,+ , I 1,+ I 2,+ . I P. J. Grandinetti, October 31, 2011
1, E 2 , I 1, I 2, , I 1, I 2,z , I 1, I 2,+ , I
188
The coherence order associated with the direct product of operators is the sum of the individual coherence 1, I 2, is p = 1 1 = 2, whereas I 1,+ I 2,z is p = +1 + 0 = +1. orders, for example, for I 1 and I 2 , e.g., I 1,z I 2,z when labeling homonuclear Also note that the convention is to use the symbols I and S , e.g., I z S z , when labeling heteronuclear coupled nuclei. coupled nuclei, and I
22.3.2
Strong Couplings - AB system
In spin systems experience strong couplings we will nd it more convenient to use the coupled space representation, where |L, M =
I1 ,I2 ,m1 ,m2
|I1 , I2 , m1 , m2 I1 , I2 , m1 , m2 |L, M ,
where I1 , I2 , m1 , m2 |L, M is the Clebsch-Gordon Coecient. For two strongly coupled spin 1/2 nuclei we have two spaces called the triplet and singlet states, one for each value of L, respectively. In the triplet, L = 1, space we have the three states, |1, 1 , |1, 0 , and |1, 1 ,
and in the singlet, L = 0, space we have only one state, |0, 0 . In terms of outer products, we will have nine for the L = 1 space |1, 1 1, 1| |1, 1 1, 0| |1, 1 1, 1| |1, 0 1, 1| |1, 0 1, 0| |1, 0 1, 1| |1, 1 1, 1| |1, 1 1, 1| |1, 1 1, 1|
(312 ) of rank l = 0, 1, which can also be described in terms of irreducible spherical tensor operators, T l,p and 2, where (312 ) = T l,p
p1 ,p2
(1) T (2) l1 , p1 , l2 , p2 |l, p T l1 ,p1 l2 ,p2
(22.16)
For the L = 0 space, the subsystem is described by only one outer product, |0, 0 0, 0| (12) . Note that the approximation of a strong or just the l = 0 irreducible spherical tensor operator, S 0,0 coupling eliminates six degrees of freedom from the coupled spin system. P. J. Grandinetti, October 31, 2011
22.3. COUPLING BETWEEN NUCLEI
189
1/2 x 1/2
1 1
+1/2 +1/2 1 +1/2 -1/2 -1/2 +1/2 1x1 +1 +1 2 +2 1 +1 0 0 +1
1 0 1/2 1/2
0 0 1/2 -1/2 -1/2 -1/2 1 -1 1

m1 m2 m1 m2 m1 m2
Coefficents
J M
J M
... ...
2 +1 1/2 1/2
1 +1 1/2 -1/2 +1 -1 2 0 1/6 2/3 1/6 1 0 1/2 0 -1/2 0 0 1/3 -1/3 1/3 0 -1
4 +2 3/14 4/7 3/14 3 +2 1/2 0 -1/2 2 +2 2/7 -3/7 2/7 +2 -1 +1 0 0 1 -1 +2 4 +1 1/14 3/7 3/7 1/14 3 +1 3/10 1/5 -1/5 -3/10 2 +1 3/7 -1/14 -1/14 3/7 1 +1 1/5 -3/10 3/10 -1/5 +2 -2 +1 -1 4 0 1/70 8/35
2x2 +2 +2
4 +4 1 +2 +1 +1 +2 4 +3 1/2 1/2 3 +3 1/2 -1/2 +2 0 +1 1 0 +2
-1 +1
2 -1 1/2 1/2
1 -1 1/2 -1/2 -1 -1 2 -2 1
-1 0
3 0 1/10 2/5 0 -2/5 -1/10
2 0 2/7 1/14 -2/7 1/14 2/7
1 0 2/5 -1/10 0 1/10 -2/5
0 0 1/5 -1/5 1/5 -1/5 1/5 +2 -1 +1 0 0 1 -1 +2 4 -1 1/14 3/7 3/7 1/14 3 -1 3/10 1/5 -1/5 -3/10 2 -1 3/7 -1/14 -1/14 3/7 1 -1 1/5 -3/10 3/10 -1/5 0 -2 -1 -1 -2 0 4 -2 3/14 4/7 3/14 3 -2 1/2 0 -1/2 2 -2 2/7 -3/7 2/7 -1 -2 -2 0 4 -3 1/2 1/2 3 -3 1/2 -1/2 -2 -2 4 -4 1
0 0 18/35 -1 +1 -2 +2 8/35 1/70
Figure 22.1: Clebsch Gordon Coecients. A 8/15 use 8/15.
is understood to be over every coecient, e.g., for
190
22.4
Observables with a Continuous Set of Outcomes
When the quantum mechanical observable does not have a discrete set of outcomes but instead a continuous set, such as the position operator r, then we can expand the state function in terms of the continuous set of possible outcomes. For the position operator r this becomes

| =
|r
r | | dr =
r | |r dr
The coecient r | is called the wavefunction and dened as (r) = r| . Likewise, we can also expand the state function in terms of the continuous set of possible outcomes as for the canonical momentum operator p

| =
|p
p | | dp =
p | |p dp
The coecient p | is dened as (p) = p| . The canonical momentum operator written in terms of the position operator eigenstates is = ih. p The relationship between the position-space and momentum space wavefunctions is the Fourier transform: 1 2 1 (p) = 2 (r) = 22.4.0.1 dp exp dr exp ip r ip r (p), (r).
Probability, Charge and Current Density
If we wanted to calculate the probability as a function of r, that is, the probability density we would use the operator (r) = |r r| to obtain (r) = =
| |r r| |

| r
r |dr
|r r|
r | |r dr
|r r| = | (r)|2 .
22.5. QUANTUM DYNAMICS
191
For a particle of charge e we can then easily calculate the charge density as e| (r)|2 . We can further calculate the current density from the classical denition j = (e/c) v, where v is the velocity. Using that p = mv, we obtain the operator 1 +p [ (r, t) p (r, t)]. j (r) = 2m Using this operator one obtains the expectation value j (r) = e i (r) (r) (r) (r) . 2mc
In NMR we will need to consider our systems in the presence of a magnetic eld. In these cases we will need to use the kinematical momentum operator denoted by e =p A( r), c where A( r) is the magnetic vector potential related to the magnetic eld by B = A( r). Thus if we were to calculate the current in the presence of a magnetic eld we would use the operator 1 + j (r) = [ (r, t) (r, t)]. 2m Using this operator one obtains the expectation value for the current density in a magnetic eld j (r) = e i e2 A( r) (r) (r). (r) (r) (r) (r) 2mc mc2
1 If the external eld is uniform then A( r) = 2 (B r), and we obtain
j (r) =
e i 1 e2 | (r)|2 (B r). (r) (r) (r) (r) 2mc 2 mc2
Thus we see that the presence of a magnetic eld will induce an extra current. This is called the diamagnetic current.
22.5
Quantum Dynamics
Weve come a long way in our description of the fundamentals of quantum mechanics, yet all we have done to this point allows us to draw a vector in an expectation value space that corresponds to the magnetization vector of real space. We still have to show how quantum mechanics can be used to describe the time evolution of the state vector and then we can calculate the time evolution of the vector in expectation value space. Lets start by adding a time label to our state vector, |(t) . P. J. Grandinetti, October 31, 2011
192
If we knew the state vector as a function of time, we could calculate the expectation value of all the observables as a function of time. So the problem is, given the state vector at some instant t0 in time, what is the state vector at some later instant in time t1 . That is, given |(t0 ) what is |(t1 ) assuming t1 > t0 ? What we need is the operator that transforms |(t0 ) into |(t1 ) . In terms of an outer product we need |(t1 ) (t0 )|. Apply this to |(t0 ) and you will get |(t1 ) . This operator is called a propagator and is usually written (t1 , t0 ) = |(t1 ) (t0 )|, U so that (t1 , t0 )|(t0 ) . |(t1 ) = U (t1 , t0 ) is obtained from solving the Of course, this really doesnt answer anything. The propagator U Schr odinger equation i d U (t, t0 ), U (t, t0 ) = H dt
is the Hamiltonian operator, i.e., the total energy operator for the system, where H | . Etotal = |H It is through the Hamiltonian operator that we can calculate how the state function will evolve in time. =H . Remember, The Hamiltonian corresponds to an observable and therefore is Hermitian, i.e., H all observable operators are Hermitian operators. How do we know what our Hamiltonian operator is? We start with the energy of a classical magnetic dipole subjected to a magnetic eld E = B, and replace all the observables associated with our system by the corresponding operators = B. H If we take B as along the z axis we have = z B0 = 0 I z . H z B0 = I P. J. Grandinetti, October 31, 2011
193
The eigenstates of the Hamiltonian for a nuclear magnetic dipole in a static magnetic eld pointing along z eigenstates. z are the I What are the eigenvalues of this Hamiltonian? For the |+ state we have E+ = = = |+ , +| H z |+ = 0 +|I z |+ = 1 0 +|+ , +| 0 I 2 0 /2.
So E+ = 0 /2 and likewise E = 0 /2. If you plotted these energies as a function of B0 you obtain would obtain (for > 0) the energy level diagram below.
Energy
B0
If you change the sign of then these levels are ipped. It is the Hamiltonian operator, through the Schr odinger equation, that generates the time evolution of our state function. Lets focus on the time dependence of |(t) , starting with the solution to the Schr odinger equation, i Rearranging and integrating we have
(t1 ,t0 ) U (t0 ,t0 ) U
(t, t0 ) dU U (t, t0 ). =H dt
(t, t0 ) dU i = (t, t0 ) U
t1 to
Hdt,
and ln Exponentiating both sides we obtain
(t1 , t0 ) U i = H (t1 t0 ). U (t0 , t0 )
(t1 , t0 ) = U (t0 , t0 ) exp i H (t1 t0 ) . U P. J. Grandinetti, October 31, 2011
194
(t0 , t0 ) is the propagation from time t0 to t0 and should be the identity operator so we have U (t1 , t0 ) = exp i H (t1 t0 ) . U We write the time evolution of |(t0 ) to |(t1 ) as i (t1 t0 ) |(t0 ) . |(t1 ) = exp H How do we work with an exponentiated operator? As long as we operate on the operators eigenstates then H |j = Ei |j and for an exponentiated operator we it is straightforward. If |j is an eigenket of H have i i (t1 t0 ) |j = exp Ej (t1 t0 ) |j . exp H , Thus, if we write |(t0 ) in terms of the eigenstates of H
N
|(t0 ) =
j
cj (t0 )|j ,
we can calculate |(t1 ) (t1 , t0 )|(t0 ) , = U = i exp H (t1 t0 )

N N
cj (t0 )|j ,
=
j N
i (t1 t0 ) |j , cj (t0 ) exp H i cj (t0 ) exp Ej (t1 t0 ) |j .
=
j
, the time evolution results in a change in Starting with |(t0 ) written in terms of the eigenkets of H the phase of the complex coecients with time. i cj (t1 ) = cj (t0 ) exp Ej (t1 t0 ) . It is important to emphasize that the magnitudes of the complex coecients do not change with time, implying that
cj (t1 )c j (t1 ) = cj (t0 )cj (t0 ).
Therefore, the expectation values of the total energy do not change with time (as expected). In fact the ) expectation values of all observables that are compatible with the Hamiltonian (i.e. commute with H P. J. Grandinetti, October 31, 2011
195
will be independent of time. It is only the expectation values of the non-commuting operators that will, in general, vary with time. Lets look at our example of a spin I nucleus with the Hamiltonian = 0 I z , H/ i.e., a spin I nucleus subjected to a magnetic eld along the z -axis. How do the expectation values of x , I y , and I z vary in time? Lets start with I x |(t) = (0)|U (t, 0)I x U (t, 0)|(0) (t)|I where
z t (t, 0) = ei0 I . U
We will focus on evaluating the term

z t z t i0 I (t, 0)I x U (t, 0) = ei0 I . Ix e U
By performing a series expansion of an exponentiated operator one can show

iB eiB Ae = A + i[B, A] +
i2 2 2!
[B, [B, A]] + +
in n n!
[B, [B, [B, . . . [B, A]]] . . .],
(22.17)
is an Hermitian operator and is a real parameter. Using the commutation relationships in where B Eq.(22.7) one obtains
z i I x + (i ) [I z , I x ] + (i )2 1 = I ei Iz I xe 2! y iI
z , [I z , I x ]] + (i ) [I
y iI x I
1 3!
z , [I z , [I z , I x ]]] + [I
x I y iI
x 1 + I y + = I x cos I y sin . = I 2! 3! y and I z we nd Taking the same approach for I Rotation about the z -axis through an angle , (22.18)
x I
x cos I y sin I
y x sin ei Iz I ei Iz = I cos + I y z z I I
196
x , I y and I z , as We can then obtain expressions for the time dependences of I x (t) I y (t) I z (t) I = x (0) cos 0 t I y (0) sin 0 t, I y (0) cos 0 t + I x (0) sin 0 t, I z (0) . I
Ignoring relaxation, this is the same solution we obtained for the Bloch equations. These equations x , I y , and I z for our precessing nuclear magnetic dipole in an describe the expectation values of I external magnetic eld. We can also dene the magnetization vector in the NMR experiment as the vector sum of the expectation value vectors of the individual nucleis magnetic moments,
N N
M(t) =
j
(t)
=
j
I(t) j ,
(22.19)
thus, completing the connection between the quantum and classical descriptions of NMR. For an arbitrary spin, I , we can also calculate the time dependence of the irreducible spherical tensor operator, Tl,p , expectation value starting with l,p |(t) = (0)|U (t, 0)T l,p U (t, 0)|(0) . (t)|T We can take a dierent approach to evaluate the general transformation
z i I ei Iz T . l,p e
l,p in terms of the outer products of the I z eigenstates, we have Expressing T l,p = T
m,m
|m
l,p |m m|. m |T
Combining this expression and the selection rule in Eq. (22.12) we nd l,p ei Iz = ei Iz T
m
l,p |m m|ei Iz . ei Iz |m + p m + p|T
An exponentiated operator acting on its eigenket just gives back the exponentiated eigenvalues times the eigenket, l,p ei Iz = ei Iz T
m
l,p |m m|eim , ei(m+p) |m + p m + p|T
which simplies to l,p ei Iz = T l,p eip . ei Iz T P. J. Grandinetti, October 31, 2011

(22.20)
197
Here we have the general expression for the free precession under a Zeeman Hamiltonian of coherences of order p. Notice that the oscillation frequency is the product of the coherence order p and the Larmor frequency. Next, we will look at the quantum mechanics of including the eect of applying an oscillating (radio frequency) magnetic eld to rotate the expectation value vector about the x or y axes. Before we do that, however, we need to step aside and examine a few issues in quantum mechanics concerning time dependent Hamiltonians.
22.5.1
Interaction Representation
In our description of quantum mechanics so far, all the information about the system at any instant in time is completely contained in the state function. This approach to time evolution is called the Schr odinger Representation. In the Schr odinger Representation of quantum mechanics it is only the state function that changes with time and the Hamiltonian and all other observable operators are time independent. There is another approach called the Heisenberg Representation. In the Heisenberg Representation it is the observable operators that carry the time dependence, and the state function is time-independent. x we would do the following in the Schr For example, to calculate the expectation value of I odinger Representation, (t, 0)|(0) , |(t) = U and x (t) = (t)|I x |(t) = I x U (t, 0)|(0) (0)|U (t, 0) I . (t, 0), (t)| = (0)|U
x is an example of a time independent operator in the Schr The operator I odinger Representation. Now we can easily regroup this expression and dene a time dependent operator, i.e., x (t) I = = (t, 0)I x U (t, 0) |(0) , (0)| U x (t)|(0) . (0)|I
x (t) is an example of a time dependent operator in the Heisenberg Representation. SumThe operator I marizing, the time evolution in the Schr odinger Representation is represented by a time dependent state function, (t, 0)|(0) |(t) = U Schr odinger Representation, P. J. Grandinetti, October 31, 2011
198
and time evolution in the Heisenberg Representation is represented by a time dependent operator x (t) = U (t, 0)I x (0)U (t, 0) I Heisenberg Representation.
Additionally, the Hamiltonian operator in a conservative system will be time independent in either representation. In a non-conservative system, the Hamiltonian operator is time dependent. Whether the Hamiltonian is conservative or non-conservative depends on how the system is dened. For example, when we derived the Hamiltonian for the spin 1/2 nuclear magnetic dipole vector subjected to a magnetic eld, we started with the classical interaction energy, B. E = to get the Hamiltonian. and replace only the observable with its quantum mechanical operator We didnt replace the magnetic eld vector by its quantum mechanical operator equivalent. We could have done so, however, it would have unnecessarily complicated our quantum mechanical system, and we would no longer have had a simple two-level quantum mechanical system. By keeping the magnetic eld as a classical object in our system we have made a semi-classical approximation. A time dependent magnetic eld in our semi-classical Hamiltonian, as would be the case when we apply an oscillating (radio frequency) magnetic eld B(t) = 2B1 cos(|rt |t + rf )ex + B0 ez , leads to a time dependent non-conservative Hamiltonian4 , (t)/ = 2B1 cos(|rt |t + rf )I x B0 I z . H Given this time-dependent Hamiltonian we are now faced with trying to solve the Schr odinger equation i (t, t0 ) dU (t)U (t, t0 ), =H dt
(t, t0 ). This is often a dicult challenge. We take an approach similar to to obtain the propagator, U the rotating frame approach that we used for the Bloch equations. The approach we use is to switch
4 Relaxation
eects in NMR are also introduced in a semi-classical approach. Relaxation arises because of random
uctuations in the lattice create random uctuations in the direction and magnitude of the magnetic eld direction. (This occurs via couplings between the spin and the lattice not discussed yet.) As a practical limitation, it is essentially impossible to account for these random uctuations from rst principles and include relaxation in a single time-independent Hamiltonian.
199
to an interaction frame representation. The interaction frame representation is an intermediate frame between the Schr odinger Representation and the Heisenberg Representation. Given a Hamiltonian of the general form =H 0 + H 1 (t), H which in our case is divided into 0 / = 0 I z , H 1 (t)/ = B1 2 cos(|rf |t + rf )I x . and H
0 is One can transform the state function into an interaction frame where the time dependence due to H removed, that is t) = W (t)|(t) , ( where
z (t) = ei(rot t+rot )I W , z (t) = ei(rot t+rot )I and W .
and
t) = (t)|W (t), (
(t)W (t) = 1. From earlier discussions, one can see that the W (t) transformation generates Note that W a rotation about the z -axis that will remove the spin precession due to the static magnetic eld only (like what happens when we went into rotating frame for the Bloch equations). In the interaction representation the state vector evolves under an eective Hamiltonian, which can be derived as follows. Starting with t) = W (t)|(t) , ( multiplying by i and taking the derivative of both sides t) d ( i dt =i (t) dW (t) i d|(t) |(t) + W dt dt
(t) ( t) into the rst term, and i d|(t) = H (t)|(t) for the second we can substitute |(t) = W dt term to obtain t) d ( (t) dW (t) ( t) + W (t)H (t)|(t) . i =i W dt dt (t) ( t) into the second term gives Again substituting |(t) = W t) d ( i dt =i (t) dW (t) ( t) + W (t)H (t)W (t) ( t) , W dt P. J. Grandinetti, October 31, 2011
200 and after rearranging we obtain t) d ( i dt = i
(t) dW (t) + W (t)H (t)W (t) W dt
t) . (
t) evolves according to Thus in the interaction frame the state function ( t) d ( i where ef f (t) = W (t)H (t)W (t) + i dW (t) W (t). H dt The second term is akin to the ctitious eld of the rotating frame. Lets now consider the NMR Hamiltonian (t)/ = 0 I z B1 2 cos(|rt |t + rf )I x . H ef f (t), starting with We calculate the second term in H (t) dW z z ei(rot t+rot )I = irot I . dt (t) on the right one obtains Multiplying both sides by i and W i ef f (t) we start with To calculate H (t)H (t)W (t) W = = (t) W z B1 2 cos(|rt |t + rf )I x W (t), 0 I (t) dW (t) = rot I z . W dt (22.23) (22.22) dt
ef f (t) ( t) , =H
(22.21)
(t)I z W (t) B1 2 cos(|rf |t + rf ) W (t)I x W (t) , 0 W
and using Eq. (22.18) one can show that (t)I x W (t) = I x cos(rot t + rot ) + I y sin(rot t + rot ), W and (t)I z W (t) = I z . W Thus, we can write (t)H (t)W (t) = 0 I z B1 2 cos(|rt |t + rf )[I x cos(rot t + rot ) + I y sin(rot t + rot )]. (22.24) W P. J. Grandinetti, October 31, 2011
22.5. QUANTUM DYNAMICS Combining Eqs. (22.24) and (22.23) into Eq. (22.22) we obtain
201
ef f (t)/ = 2B1 [I x cos(|rt |t+rf ) cos(rot t+rot )+I y cos(|rt |t+rf ) sin(rot t+rot )]+ (0 rot )I z , H which can be rearranged to ef f (t)/ H = x cos((|rf | rot )t + (rf rot )) + I y sin((|rf | rot )t (rf rot ))} B1 {I x cos[(|rf | + rot )t + (rf + rot )] + I y sin[(|rf | + rot )t + (rf + rot )]} +B1 {I z + (0 rot )I For a rotating frame that follows our precessing magnetization vector we will set rot = 0 . Thus, to ef f (t) for positive Larmor a good approximation, we can neglect the fast rot + |rf | oscillations in H frequencies, i.e., < 0, or neglect the fast rot |rf | oscillations for negative Larmor frequencies, i.e., > 0. Setting |rot | = |rf |, keeping only the time independent terms, and recalling our denition 1 = |B1 |, rot we obtain 0 = when < 0, when > 0, (22.25)
I 1 x cos rf + 1 Iy sin rf ef f / = H I 1 x cos rf 1 Iy sin rf
when < 0 when > 0
Again, we dene the phase of the nutation axis convention of = (sign ) rf , and obtain ef f / = 1 I x cos + 1 I y sin . H x , I y , and I z vary in time in this interaction So now we ask: How do the expectation values of I frame? Setting = 0, so that ef f / = 1 I x , H one can use Eq. (22.17) to obtain x I x I Rotation about the x-axis through an angle , (22.26)
y z sin ei Ix I ei Ix = I cos I y z z cos + I y sin I I
202 From these expressions we obtain x (t) I y (t) I z (t) I Setting = /2, so that = = = x (0) , I
y (0) cos 1 t I z (0) sin 1 t, I z (0) cos 1 t + I y (0) sin 1 t. I
ef f / = 1 I y , H one obtains x I x cos + I z sin I Rotation about the y -axis through an angle , (22.27)
y ei Iy I ei Iy = I y z cos I x sin I Iz From these expressions we obtain x (t) I y (t) I z (t) I = = =
x (0) cos 1 t + I z (0) sin 1 t, I y (0) , I z (0) cos 1 t I x (0) sin 1 t. I
Again, neglecting relaxation, this is the same result we get from the Bloch equations in the rotating x , I y , and I z in the frame. They describe the rotation of the vector in expectation value space of I interaction frame. One can also derive
x i I (1+cos )/2 I (1 cos )/2 + iI z sin , =I ei Ix I e
(22.28)
and
i I y (1+cos )/2 + I (1 cos )/2 I z sin , ei Iy I =I e
(22.29)
from which we obtain the useful result A particularly useful result from these equations is that complex + and I are swapped by a pulse, vector components I
x I I ,
and
y . I I
(22.30)
More generally, we can calculate the eect of an rf pulse of duration 1 t = and phase using the rotation operator
z i I x iI ( ) = exp i (I x cos + I y sin ) = eiI e e z. R
203
Applied to the angular momentum operators, written in terms of the Cartesian basis vectors, we obtain 1 1 x y sin 2)(1 cos 1 t)+ I z sin sin 1 t I I (1+cos t )+ ( I cos 2 + I x 1 x 2 2 1 1 1 ( ) R ( ) = R I I (1+cos t ) ( I cos 2 I sin 2 )(1 cos t ) I cos sin t 1 x 1 z 1 , 2 y 2 y y z z cos 1 t (I x sin I y cos ) sin 1 t I I (22.31) and in terms of the raising and lowering operators we obtain + (1+cos )/2 I ei2 (1 cos )/2 + I z ei (i sin ) + I I ( ) = 1 ( ) (1+cos )/2 I + ei2 (1 cos )/2 + I z ei (i sin ) R . I R I z cos + ei (i sin )/2 ei (i sin )/2 z I I I I From these expressions we nd the eect of a 1 t = /2 pulse of arbitrary phase to be 1 1 x I I + ( I cos 2 + I sin 2 )+ I sin y z 2 x 2 x 1 1 1 R (/2)Iy R (/2) = 2 Iy 2 (Iy cos 2 Ix sin 2) Iz cos , z y cos I x sin I I and in terms of the raising and lowering operators we obtain /2 I ei2 /2 + iI z ei + I I + (/2) = 1 (/2) /2 I + ei2 /2 + iI z ei R I I R + ei /2 iI ei /2 z iI I
(22.32)
We can also calculate the eect of an rf pulse of arbitrary phase on the irreducible tensor operators. For this task we can use the well known and tabulated Wigner rotation matrices, whose elements are given by Dm ,m (, , ) = l, m |eiIz ei Iy ei Iz |l, m .
(l )
(22.33)
Here , , and are the Euler angles, and are used to specify an arbitrary rotation of a object. The Dm ,m (, , ) elements can be written as a product of three functions, each of which depends on only one argument , , or ,
(l ) l) im2 Dm (, , ) = eim1 d( . m1 ,m2 ( ) e 1 ,m2 (l) (l)
(22.34)
The function dm1 ,m2 ( ) is called the reduced Wigner matrix element. Expressions for these elements are given in Table 22.1 for l = 1 and 2. The general rotation of a irreducible spherical tensor operator is calculated according to
l
1 (, , )T l,p D (, , ) = D 0
p1 =l
(l) l,p . Dp (, , )T 1 1 ,p0
(22.35)
204
(1)
dm1 ,m2 ( ) m1 = 1 m1 = 0 m1 = 1 dm1 ,m2 ( ) m1 = 2 m1 = 1 m1 = 0 m1 = 1 m1 = 2

(2)
m2 = 1
1+cos 2 sin 2 1cos 2
m2 = 0
sin 2
m2 = 1
1cos 2 sin 2 1+cos 2
cos
sin 2
m2 = 2
(1+cos ) 4 sin (1+cos ) 2 2 1 3 2 2 sin sin (1cos ) 2 (1cos )2 4
2
m2 = 1
) sin (1+cos 2 2 cos2 +cos 1 2 3 2 1 2
m2 = 0
3 2
m 2 = 1
cos ) sin (12
m2 = 2
(1cos )2 4 cos ) sin (12 2 3 1 2 2 sin ) sin (1+cos 2 2 (1+cos ) 4
sin2
sin cos
2
3 2 sin cos (3 cos2 1) 2 3 2 1 2
2 cos
3 2
cos 1 2
sin cos
2 cos
cos 1 2 sin (1cos ) 2
sin cos
3 2
sin2
2 cos2 +cos 1 2 sin (1+cos ) 2
Table 22.1: Reduced Wigner rotation matrix elements for l = 1 and 2. To calculate the eect of an rf pulse of duration 1 t = and phase we set = where one can show that ( ). , , = R D 2 2 We can then calculate the eect of the rf pulse to be
l 2
and =
( ) = 1 ( )T l,p R R 0
p1 =l
l) ip1 eip1 /2 d( , p1 ,p0 ( ) Tl,p1 e
(22.36)
where p1 = p1 p0 . Here we obtain an expression similar, but more general than Eq. (20.29), showing how the rf pulse phase is scaled by the change in coherence order p during an rf pulse. Using quantum mechanics we havent changed our picture much at all from the classical description of magnetization vectors moving in a three dimensional space. We have now developed, however, much of the mathematical machinery that we will use to explore the higher dimensional spaces of spin greater than 1/2 and coupled nuclei.
22.5.1.1
Double Rotating Frame
For a heteronuclear two spin system, the Zeeman Hamiltonian will be 0 / = I I z + S S z . H P. J. Grandinetti, October 31, 2011
205
In this case, it is useful to transform into an interaction frame called the double rotating frame, using
I S S t+I (t) = ei(rot rot )Iz ei(rot t+rot )Sz . W
In this interaction frame, the eective Hamiltonian where both the I and S nuclei can be irradiated is ef f / = 1,I I x cos I + 1,I I y sin I + 1,S S x cos S + 1,S S y sin S . H
22.5.2
Observables with a Continuous Set of Outcomes
The Hamiltonian for a single particle in the absence of an electromagnetic eld is 2 = p (r), H +V 2m (r) is the potential energy operator. Writing the momentum operator in terms of position we where V have = H
2
2m
(r). 2 + V
Using the Schrodinger equation on the state ket for a single particle expanded in terms of the position eigenkets we obtain r|i and obtain i
2 d (r) (r, t). (r, t) = 2 (r, t) + V dt 2m 2 d |(t) = r| (r)|(t) , |(t) = r|H 2 + V dt 2m
From this one can derive the time independent Schrodinger equation
2
2m
(r) (r) = E (r). 2 + V
In the presence of an external magnetic eld our Hamiltonian becomes 2 (r), H= +V 2m = ih e A (r). In this case the time independent Schrodinger equation is where c
2
2m
(r) (r) + 2 + V
2 e ) + e A 2 (r) = E (r). +p A (A p 2mc 2m2 c2
This will be one of the starting points when considering the origins of the chemical shift Hamiltonian. P. J. Grandinetti, October 31, 2011
206
22.6
Transitions
Lets briey discuss the concept of a transition in NMR (or any spectroscopy). A transition changes the populations of levels in a quantum mechanical system. By level we mean the eigenstate of the Hamiltonian. For example, the state function can be expanded |(0) = c1 |1 + c2 |2 where |1 and |2 are the eigenstates of H . The populations of an eigenstate is the magnitude of the complex coecient in front of that eigenstate in this expansion of the state function. That is, the population of |1 is c1 c 1 . As long as the Hamiltonian doesnt change with time then the populations of the eigenstates will not change with time. Put another way, no transitions can occur if H is always time independent. A transition only takes place when the Hamiltonian is changed and the old eigenstates are no longer valid.
H
old populations
H'
transitions between eigenstates of H taking place here
H
new populations depend on how long H' was on.
If we continue to expand the state function | in terms of the old Hamiltonian eigenstates (not the same as new Hamiltonian eigenstates), then during the new Hamiltonian the magnitude of the complex coecients in front of the old eigenstates (i.e. population of old eigenstates) will change with time. Once the Hamiltonian is switched back to the old Hamiltonian then the populations of the old eigenstates will stop changing in time and will be locked at the value they had at the instant the old Hamiltonian came back on. For example, in the interaction (rotating) frame, if the populations (i.e., c+ and c ) were such that the expectation value of z was + /2, then turning the rf Hamiltonian on will change the populations so that z changes. While the rf Hamiltonian is on, we say a transition is taking place. When the rf is turned o, then the transition is over. If the pulse was on so that the expectation value was rotated /2, then the new populations are dierent after the pulse than before. If the pulse was on so that the expectation value was rotated 2 , then the new populations are the same before and after the pulse (transition). Of course, after the pulse, relaxation processes will return the ensemble average expectation value back to along the +z axis. You could think of the random uctuations in the lattice as creating another kind of time dependent Hamiltonian that induces transitions. The steady state eect of these random uctuations is an ensemble average expectation value along +z axis. P. J. Grandinetti, October 31, 2011
22.6. TRANSITIONS
207
H
Z
H'
Z
H
Z
Y X X
Y X
208
Chapter 23
The Density Operator

23.1 The Projection Operator
Earlier we obtained the equation for calculating the expectation value of an observable operator O, that is, = |O | , O given | =
j
cj |j .
The state function | is expanded in terms of the eigenstates of an operator, usually H . We are now going to rearrange some of our basic equations of quantum mechanics and develop a slightly dierent approach for calculating the expectation value of an operator. Lets start with a calculation of O for a two level system where | = c1 |1 + c2 |2 . The expectation value of an operator O in a nite-level system is | |O =
1|c 1 + 2|c2 O c1 |1 + c2 |2
= c1 c 1 1|O |1 + c2 c1 1|O |2 + c1 c 2 2|O |1 + c2 c2 2|O |2
209
210 The trace of a matrix is the a11 a12 a21 a22 Tr a31 a32 . . . . . .
CHAPTER 23. THE DENSITY OPERATOR sum of its diagonal elements: a13 a23 = a11 + a22 + a33 + . a33 . .. . . .
An important property of traces is that the sum of matrix traces is equal to the trace of the sum of matrices: ) + Tr(B ) + Tr(C ) + = Tr(A +B +C + ). Tr(A Another important property is that the order of the matrix products can be cyclically permuted without aecting the trace: B C D ) = Tr(B C D A ) = Tr(C D A B ) = Tr(D A B C ). Tr(A Figure 23.1: The trace
This can be rewritten as the trace of the product of two matrices | = Tr |O |1 1|O |1 2|O |2 1|O |2 2|O c1 c 1 c2 c 1 c1 c 2 c2 c 2 .
In this case one can easily show that the trace of the product of these two matrices is the expectation . The matrix of j |O |j is just the matrix representation of the operator O in the value of the operator O |j basis set. You can easily show that the matrix of coecient products is the matrix representation of the operator = | | . P In the two-level (spin 1/2) case one can write = | | = P c1 |1 + c2 |2
1|c 1 + 2|c2
= c1 c 1 |1 1| + c1 c2 |1 2| + c 1 c2 |2 1| + c2 c2 |2 2|.
23.1. THE PROJECTION OPERATOR You can verify that = P as Therefore, we rewrite O = Tr(P O ). O |1 1|P |1 2|P |2 1|P |2 2|P = c1 c 1 c2 c 1 c1 c 2 c2 c 2 .
211
is called the Projection operator. It carries all the information that the state function | where P (t), carries about observable expectation values. Just as |(t) depends on time so does P (t) = |(t) (t)|. P , and the Schr The time dependence of |(t) is determined by the Hamiltonian H odinger equation. Recall that (t, 0)|(0) , |(t) = U Thus (t) = |(t) (t)| = U (t, 0)|(0) (0)|U (t, 0), P which becomes (t) = U (t, 0)P (0)U (t, 0). P Now we come to the proof of an important relationship. If (t) = |(t) (t)|, P then and (t, 0). (t)| = (0)|U
(t) dP d = |(t) dt dt
(t)| + |(t)
d (t)| . dt
Recalling the Schr odinger equation d |(t) |(t) = (i/ )H dt we can write (t) dP dt = |(t) (i/ )H (t)| + |(t) , (t)|(i/ )H and d (t)| = (t)|(i/ )H, dt
|(t) (t)| |(t) (t)|H = (i/ ) H P (t) P (t)H . = (i/ ) H Which can be rewritten as (t) dP i = [H, P (t)]. dt (23.1) P. J. Grandinetti, October 31, 2011
212
CHAPTER 23. THE DENSITY OPERATOR
This is the Liouville-von-Neuman equation for the Projection operator. Instead of using |(t) to describe (t) to describe the system, and the time dependence of the expectation values the system, we can use P can be calculated according to: (t) = T r{P (t)O }, O Lets look at the specic expansion of P (t) in the case of a two level system. We have (t) P = +
c (t)c (t)| | + c (t)c+ (t)| +| c+ (t)c (t)|+ | + c+ (t)c+ (t)|+ +|.
(23.2)
, I x , I y , and I z in terms of outer products one obtains Substituting the denitions for E (t) = 1 E P 2 +
[c (t) c + (t) + c+ (t) c (t)] Ix
+ i [c+ (t) c (t) c (t) c+ (t)] Iy
[c+ (t) c + (t) c (t) c (t)] Iz .
Of course, we recognize these terms in the brackets from our earlier discussions of expectation values of spin 1/2 cx (t) cy (t) cz (t) (t) as Thus, we write P (t) = 1 E + [I x cos (t) + I y sin (t)] sin (t) + I z cos (t). P 2 (t) in terms of these Hermitian operators is convenient because the trace of the product of Expanding P orthogonal Hermitian operators is zero. That is, I ) = Tr(I I ) = 0 Tr(E On the other hand, for abitrary spin I we have x I x ) = Tr(I y I y ) = Tr(I z I z ) = I (I + 1)(2I + 1)/3, Tr(I In terms of the raising and lowering operators we obtain I ) = Tr(I I z ) = 0, Tr(I P. J. Grandinetti, October 31, 2011 and I ) = 2I (I + 1)(2I + 1)/3, Tr(I (23.4) (23.3) where , = x, y , or z and = .
= c (t) c + (t) + c+ (t) c (t) = ic+ (t) c (t) ic (t) c+ (t) = c+ (t) c + (t) c (t) c (t)
= sin (t) cos (t), = sin (t) sin (t), = cos (t).
23.2. THE DENSITY OPERATOR Thus, we calculate for the three expectation values for a spin 1/2 as x I y I z I So, for a spin
1 2
213
(t)I x } = Tr{P (t)I y } = Tr{P (t)I z } = Tr{P
1 sin (t) cos (t), 2 1 = sin (t) sin (t), 2 1 = cos (t). 2
(t) as a vector moving in a three-dimensional system we can think of the operator P
space, just like the expectation value vector moved in a three-dimensional expectation value space.
23.2
The Density Operator
(t) gives the same results as using |(t) . So whats the point of In the last section we saw that using P (t) instead of |(t) ? Consider two isolated (not coupled) spin 1/2 nuclei. What is their total using P x ? It is I x = I x I Using | we would write x = 1 |I x |1 + 2 |I x |2 . I x Remember that each spin has its own state function |1 and |2 , respectively, while the operator I we would write is the same for both. Using P x = Tr{P 1 I x ) + Tr(P 2 I x }. I Recalling the property of traces } + Tr{B } = Tr{A +B }, Tr{A we can rewrite Eq. (23.5) as x = Tr{P 1 I x + P 2 I x } = Tr{(P 1 + P 2 )I x }. I We further dene total = P 1 + P 2 P and write x = Tr{P total I x }. I total . All the information about the two spin total is contained in one operator P P. J. Grandinetti, October 31, 2011 (23.5)
1
x 2 . + I
214
CHAPTER 23. THE DENSITY OPERATOR Of course, this result can be generalized to the case of N isolated spins. That is, N N x = j I x } = Tr j I x = N Tr{I x }. I Tr{P P
j j
Here we dene the density operator as 1 = N

N
j . P
j
It is used to describe an ensemble average of isolated systems. is the same for each isolated Most important is how to calculate the time evolution of (t). Since H (t, 0) is the same for each isolated spin. Therefore, spin in the ensemble, then U (t) = 1 N
N
j (t) = 1 P N
(t, 0)P j (0)U (t, 0). U

j
(t, 0) and U (t, 0) are the same for all values of i, they can be factored out of the sum, Since U 1 N j (0) U (t, 0), (t, 0) P (t) = U N j giving (t, 0) (t, 0). (t) = U (0)U (t, 0) for single spin is also valid for the ensemble average spin. As we did with The propagator U Eq. (23.1), we can derive the Liouville-von Neuman eqaution for the density operator, d (t) i ]. = [ (t), H dt (t), Thus, just like P (t) is a vector moving in a total expectation value space, called Liouville Space. The ensemble average of expectation values of any observable will be given by (t) = N Tr{ }. O (t) O
23.3
Thermal Equilibrium Density Operator
From quantum statistical mechanics1 , we nd at thermal Equilibrium that eq =

1 See
eH/kB T , Z
J. J. Sakurais book Modern Quantum Mechanics for this proof.
23.3. THERMAL EQUILIBRIUM DENSITY OPERATOR
215
where Z = Tr(eH/kB T ) is the partition function. In NMR, where the Zeeman Hamiltonian is usually order of magnitude greater than other spin Hamiltonians, we have eq = In the high temperature unit, i.e., kB T the exponential to get eq + B0 I z /kB T z E E B0 I = + . kB T (2I + 1) kB T (2I + 1) Tr(E ) e
z /kB T B0 I z /kB T B0 I )
Tr(e
z ||, we can perform a Taylor series expansion of || B0 I
and I z . This agrees with our At thermal equilibrium the density operator is a linear combination of E is invariant under evolution, we can ignore it, keeping in picture from the Bloch Equations. Since E mind that its always there for the ride. Thus, we write the equilibrium density operator as eq 0 z . I kB T (2I + 1)
From this expression and Eq. (23.3), we can calculate the equilibrium magnetization, z ) Meq = N Tr( eq I 1N 3
2 2
I (I + 1)B0 . kB T
23.3.1
The Density Operator in the Rotating Frame
Before we use the rotating frame Hamiltonian, we need to transform the density operator into the rotating frame. Lets rst recall how this was done for the state function |(t) . The state function in the rotating frame (i.e., interaction frame) was given by (t)|(t) , t) = W ( t) is the state function in the rotating frame, and W (t) is the transformation from the lab where ( frame to rotating frame. The density operator in the lab frame, (t), can be similarly transformed to the density operator in the rotating frame, (t), according to (t) (t) (t) = W (t)W We can take the thermal equilibrium density operator z , eq bI where b= B0 , kB T (2I + 1) P. J. Grandinetti, October 31, 2011
216
z (t) = ei(rot t+rot )I into the rotating frame, with W giving us
i(rot t+rot )I z z . = bI eq bei(rot t+rot )Iz I ze
We can obtain the signal in the lab frame, where the detector coil is along the x axis, starting with a calculation of the magnetization component along the coil axis as x }. My (t) = N Tr{ (t) I Following the same arguments of section 20.3 one obtains kN + }eiref . (t) I S (t) = Tr{ 2 (23.6)
. In From Eq. (23.4) one sees that the only operator in (t) that can give a non-zero trace is I other words, only coherences of order p = 1 in the density operator will be detected in the NMR signal. This is the reason that the convention in NMR is to draw the mirror image (complex conjugate) coherence transfer pathway. That is, the coherence transfer pathway followed in terms of the density operator components is the complex conjugate of the coherence transfer pathway followed in terms of the observable magnetization components. Written in terms of irreducible spherical tensor operators gives S (t) = kN I (I + 1)(2I + 1) 1,1 }eiref . Tr{ (t) T 3 (23.7)
l,p , which is dierent than for Hermitian operators, is The trace relation for T l ,p T Tr(T ) = (1)p1 l1 ,l2 p1 ,p2 . 1 1 l2 ,p2 Using an irreducible tensor operator with l = 1 and p = +1 as the observable, we nd, more specically, that only rst rank, l = 1 coherences of order p = 1 in the density operator will be detected in the NMR signal.
23.4
Ensemble of Uncoupled Spins
We will examine the application of the density operator in calculating the signal in the two experiments, the Bloch Decay and Spin Echo experiments. P. J. Grandinetti, October 31, 2011
23.4. ENSEMBLE OF UNCOUPLED SPINS
217
23.4.1
The Bloch Decay
z . We will ignore the proportionality constant, b. We start with an equilibrium density operator, eq = I During the pulse we evolve under the rotating frame Hamiltonian = I z + 1 (I x cos + I y sin ), H/ where is dierence between resonance frequency and rf eld frequency (i.e. rotating frame frequency) and 1 is the rf eld strength. To simplify our evolution calculation during the pulse we assume that 1 , and use an rf pulse along x in the rotating frame, so that = 1 I x , H/ during the pulse. Immediately after the pulse we have + + i I . x (/2)I z R 1 (/2) = I y = i I (0)+ = R x 2 2 Note that the order of the rotation operators are reversed compared to Eq. (??), as they should be, since we are propagating the density operator instead of the operator associated with the observable. Since we will only detect coherences with p = 1, we can eliminate the p = +1 coherence. After the = I z , to give pulse, the density operator evolves under the Hamiltonian H/
(t) = eiIz t
i i it I eiIz t = I . e 2 2
Setting the receiver phase, ref , to zero, we obtain our signal as S (t) = iI (I + 1)(2I + 1)kN it e . 3 2
Of course, our quantum mechanical treatment doesnt include relaxation, so we can add this phenomenologically to obtain S (t) = S (0)eit et/T2 ,
I +1)kN where S (0) = iI (I +1)(2 3 2
23.4.2
The Spin Echo Experiment
Lets calculate the density operator evolution in a two pulse sequence with the coherence transfer pathway of p = 0 +1 1, that is, the spin echo experiment. Again, we assume that 1 pulse along x in the rotating frame, so that = 1 I x , H/ P. J. Grandinetti, October 31, 2011 , and use an rf
218 during the pulse, and = I z H
when the rf pulse is o. Immediately after the pulse we have

1 + + i I . x (/2)I z R x y = i I (0)+ = R (/2) = I 2 2
We eliminate the undesired p = 1 coherence during t1 and calculate the density operator evolution = I z , to give under the Hamiltonian H/
(t1 ) = eiIz t1
i it1 i z t1 iI = I . I + e +e 2 2
After the pulse we have i it1 (t1 )+ = I , e 2 and after a t2 evolution period we have i it1 it2 e . (t1 , t2 ) = I e 2 Setting the receiver phase, ref , to zero, we obtain our signal as S (t1 , t2 ) = S (0, 0)eit1 eit2 ,
I +1)kN where S (0, 0) = iI (I +1)(2 3 2
. When t1 = t2 then S (t1 , t2 ) = S (0, 0), and, neglecting relaxation,
the signal is as if no time evolution had occurred.
23.5
Ensemble of Heteronuclear Coupled Spins
In the examples that follow, we will consider the evolution of two coupled spin 1/2 nuclei with dierent gyromagnetic ratios experiencing the following lab frame Hamiltonian, = S z + 2 I z + 2J I z S z . H/
23.6
Ensemble of Weakly Homonuclear Coupled Spins
In the examples that follow, we will consider the evolution of two coupled spin 1/2 nuclei experiencing the following Hamiltonian, = 1 I 1,z + 2 I 2,z + 2J I 1 I 2 , H/ P. J. Grandinetti, October 31, 2011
23.6. ENSEMBLE OF WEAKLY HOMONUCLEAR COUPLED SPINS which leads to the propagator
1,z +2 I 2,z +2J I 1 I 2 )t (t) = ei(1 I U .
219
How do we use such a propagator? Remember the solution to this problem lies in applying the propagator only to states or operators expanded in terms of eigenstates of the Hamiltonian. The 16 operators weve chosen to describe our system are easily expanded in terms of the states |m1 m2 , but are these states the eigenstates of our Hamiltonian? In this case, the answer to that question is, unfortunately, no. Rewriting our Hamiltonian in the form = 1 I 1,z + 2 I 2,z + 2J H/ we nd that H |m1 m2 gives for each term 1,z |m1 m2 1 I 2,z |m1 m2 2 I 1,z I 2,z |m1 m2 +2 J I m1 1 |m1 , m2 1 1,z I 2,z + 1 I 1, I 2,+ I1,+ I2, + I 2 2
m2 1 |m1 , m2
2J m1 m2 |m1 , m2 J 4 J 4 3 m2 (m2 1) 4 3 m2 (m2 + 1) 4 3 m1 (m1 + 1)|m1 + 1, m2 1 4 3 m1 (m1 1)|m1 1, m2 + 1 4
1,+ I 2, |m1 m2 + J I
1, I 2,+ |m1 m2 + J I
. What do you do if youve It is the last two terms that prevent |m1 m2 from being an eigenstates of H ? You have to nd the transformation expanded everything in a basis set that isnt the eigenbasis of H , but | is expanded in terms of |j , then between basis sets. For example, if |j are eigenstates of H you need to know the transformation between |j and |j , that is, |j , |j = W Thus, if |j , Ej = j |H then we can write Ej = |j = { j |W }H {W |j } = j |W H W |j . j |H P. J. Grandinetti, October 31, 2011 = where W
k
|j j |.
220
This suggests another approach. Instead of transforming our state function, lets transform the Hamiltonian. In other words, e
i
Ej t
}e = j |eiHt |j = { j |W
Ht
|j } = j | e {W
H W t W
|j
transformed
So instead of having to transform our state function or density operator, we can instead transform the Hamiltonian. The matrix representation of the transformed Hamiltonian will be diagonal in our original makes our Hamiltonian diagonal in the basis |m1 m2 . We now need to know what transformation W |m1 m2 basis. In general there are 3 dierent approaches to this problem. . 1. Find an exact analytical solution for W (i.e., use a computer). 2. Find an exact numerical solution for W (i.e., use perturbation theory). 3. Find an approximate analytical solution for W Shown in appendix ?? is how the exact analytical solution is found in this case. In the weak coupling limit, that is, when the dierence between 1 and 2 is much greater than 2J then the approximate solution is H W / 1 I 1,z + 2 I 2,z + 2J I 1,z I 2,z . W We will assume we are working in the weak coupling limit in the examples that follow. In the weak coupling limit the propagator is given by
2,z )t 2,z +2J I 1,z I 1,z +2 I (t) = ei(1 I . U
Since each term in the exponent commutes with the others we can split this propagator into three parts
1,z t i2 I 2,z t i2J I 1,z I 2,z t (t) = ei1 I e e . U
(23.8)
We have already discussed how to apply the rst two parts of this propagator. Lets now focus on the various coherences under the J coupling Hamiltonian. Applied to the p = 1 coherences we have ei2J I1,z I2,z t

1, I 2, I
1,z I 2,z t ei2J I =
2,z t 1, ei2J I I 1,z t 2, ei2J I I
How do we deal with the left over operator in the exponent? We start with the relationship eiaIz = cos(aIz ) i sin(aIz ), P. J. Grandinetti, October 31, 2011
23.6. ENSEMBLE OF WEAKLY HOMONUCLEAR COUPLED SPINS and do a series expansion of each to obtain cos(aIz ) sin(aIz )
1 2 z E so For spin 1/2 we know2 that I =4
221
= =
z )2 z )4 (aI (aI + + , 2! 4! 3 5 z (aIz ) + (aIz ) + . aI 3! 5!
cos(aIz ) sin(aIz ) Therefore
= =
(a/2)2 (a/2)4 + + = cos(a/2), 2! 4! 3 (a/2) (a/2)5 z = 2I z sin(a/2). a/2 + + 2I 3! 5!
z sin(a/2), eiaIz = cos(a/2) i2I and we obtain e

2,z t 1,z I i2J I
(23.9)
1, I 2, I
e
2,z t 1,z I i2J I
1, cos Jt i2I 1, I 2,z sin Jt, I 2, cos Jt i2I 1,z I 2, sin Jt I
. (23.10)
In a coupled spin system we also have to consider the evolution of the other p = 1 coherences 2,z t i2J I 2 I I 2 I I e 1, 2,z 1, 2,z 1,z I 2,z t ei2J I = ei2J I1,z I2,z t 1,z t 2, ei2J I 2I1,z I2, 2I1,z I Using Eq. (23.9) we obtain 1, I 2,z 1, I 2,z cos Jt iI 1, sin Jt 2 I 2 I 2,z t 1,z I ei2J I . ei2J I1,z I2,z t = 1,z I 2, cos Jt iI 2, sin Jt 2I1,z I2, 2I
(23.11)
It is also instructive to consider the evolution of p = 0 and p = 2 coherences under a weak J coupling. For p = 0 coherences, one can easily show that 1,z I 2,z ei2J I1,z I2,z t I 1, I 2, I and for p = 2, one can also show that
1,z I 2,z t 1, I 2, ei2J I 1, I 2, . =I ei2J I1,z I2,z t I
1,z I 2,z I 1, I 2, I
, (23.12)
i2J I 1,z I 2,z t = e
(23.13)
Thus, two spin zero and double quantum coherences do not evolve under the two spin J coupling.
2 In
2 = matrix form we have I z
1/2 0
0 1/2
1/2 0
0 1/2
1/4 0
0 1/4
222
23.6.1
The Bloch Decay for Two Coupled Spin 1/2 Nuclei
Lets now consider the single pulse Bloch Decay experiment for two coupled spins in the presence of the chemical shift and J-coupling interactions. At equilibrium we have 1,z + I 2,z . eq = I During an rf pulse we assume that 1 frame to obtain = 1 (I 1,x + I 2,x ) cos + (I 1,y + I 2,y ) sin . H/ Keeping only the p = 1 terms, immediately after the pulse we have i i 1 x (/2) x (0)+ = R eq R (/2) = I 1, + I2, . 2 2 Evolution of the density operator under the propagator in Eq. (23.8) yields i 1, ei1 t + I 2, ei2 t ) cos Jt + i2(I 1, I 2,z ei1 t + I 1,z I 2, ei2 t ) sin Jt . (t) = (I 2 1,+ + I 2,+ to obtain our signal From this expression we take the trace with I S (t) = S (0)(ei1 t + ei2 t ) cos Jt,
. This expression can be rearranged to obtain where S (0) = ikN 2 2
1 , 2 , 2J and use an rf pulse along the x axis in the rotating
S (t) = S (0)(ei(1 J )t + ei(1 +J )t + ei(2 J )t + ei(2 +J )t ). Adding in a T2 decay the Fourier transform of this signal leads to the spectrum below:
2J
2J
1
23.6.2
The Spin Echo Experiment For Two Coupled Spin 1/2 Nuclei
Now we calculate the density operator evolution in a two pulse spin echo experiment with the coherence transfer pathway of p = 0 +1 1, for two weakly coupled spin 1/2 nuclei. P. J. Grandinetti, October 31, 2011
23.6. ENSEMBLE OF WEAKLY HOMONUCLEAR COUPLED SPINS Immediately after the rst pulse we keep only the p = +1 coherences, leaving i 1 x (/2) x 2,+ . (0)+ = R eq R (/2) = I1,+ + I 2 Applying the J coupling, as well as the oset evolution yields
223
i 1,z t1 2,+ ei2 t1 ei2J I I1,+ ei1 t1 ei2J I2,z t1 + I . (t1 ) = 2 1,z and I 2,z operators in their exponentiated form. ImmeTo simplify our treatment we will leave the I diately after the pulse we will have i 1,z t1 2, ei2 t1 ei2J I , I1, ei1 t1 ei2J I2,z t1 + I (t1 )+ = 2 where we have used
1 x ( )I R x , R ( ) = I z t1 1 x ( )ei2J I Rx ( ) = ei2J Iz t1 . and R
During t2 evolution we obtain i 1,z t2 i2 t1 i2J I 1,z t1 2, ei2 t2 ei2J I (t1 , t2 ) = I1, ei1 t2 ei2J I2,z t2 ei1 t1 ei2J I2,z t1 + I e e . 2 If we acquire only the echo tops, when t1 = t2 = , then our density operator expression simplies to i 1,z 2 2, ei2J I (t1 = t2 = ) = I1, ei2J I2,z 2 + I . 2 1,z and I 2,z operators gives Expanding the exponentiated I i 1, I 2,z sin J 2 + I 2, cos J 2 i2I 1,z I 2, sin J 2 . I1, cos J 2 i2I (t1 = t2 = ) = 2 1,+ + I 2,+ , leaves Keeping only the p = 1 terms observable after the trace with I i 2, cos J 2 . (t1 = t2 = ) = I1, cos J 2 + I 2 From this we calculate our signal and obtain S (t = 2 ) = S (0, 0)2 cos Jt = S (0, 0)(eiJt + eiJt ).
. Adding a T2 decay to this signal yields the spectrum below. where S (0, 0) = ikN 2 2
(23.14)
2J
224
Problems
1. Using the denitions x = y = z = prove that a. [x , y ] = i z , |z ; + z ; | + |z ; z ; +| , 2 i 2 |z ; z ; +| |z ; + z ; | ,
|z ; + z ; +| |z ; z ; | , 2
b. [z , x ] = i y ,
c. [y , z ] = i x .
2. Using the denition = x iy , prove that a. [ , z ] = ,
b. [+ , ] = 2 z ,
c. [+ , x ] = z .
3. Show that a. [aA, bB ] = ab[A, B ],
b. [A, B + C ] = [A, B ] + [A, C ],
c. [A, BC ] = [A, B ]C + B [A, C ].
23.6. ENSEMBLE OF WEAKLY HOMONUCLEAR COUPLED SPINS 4. Given 1 1 |x; = |z ; + eix |z ; , 2 2 1 1 |y ; = |z ; + eiy |z ; . 2 2 Show that the constraint 1 | y ; |x; + | = | y ; |x; | = , 2 leads to the constraint: 1 1 |1 + ei(x y ) | = , 2 2 and x y = . 2 5. What is the interaction energy between a nuclear magnetic dipole moment and magnetic eld if the state function for the nucleus is 1 1 | = |z ; + + |z ; ? 2 2 Use H = z B0 . 6. Using x (t) y (t) z (t) Show that x (t) y (t) z (t) = = = x (0) cos 0 t y (0) sin 0 t, y (0) cos 0 t + x (0) sin 0 t, z (0) . = = = 1 sin cos(0 t + ), 2 1 sin sin(0 t + ), 2 1 cos , 2
225
226 7. Given x I y I z I + I I show that x eiIx = I x , a. eiIx I

x i I y cos I z sin , =I b. eiIx I ye
= = = = =
1 |z ; + z ; | + |z ; z ; +| , 2 i |z ; z ; +| |z ; + z ; | , 2 1 |z ; + z ; +| |z ; z ; | , 2 |z ; + z ; |, |z ; z ; +|,
z cos + I y sin , z eiIx = I c. eiIx I

z i I ei . =I c. eiIz I e
8.
Calculate the density operator as a function of t1 and t2 for the pulse sequence
/2 - t1 - - t2 for two weakly J coupled spin 1/2 nuclei. Show that the signal at times t1 = t2 is only modulated by the J-coupling. 9. How would you design a two-dimensional NMR experiment to correlate the J-coupling only spectrum to the normal one pulse/acquire spectrum?
Further Reading
A. Abragam, Principles of Nuclear Magnetism, Oxford University Press, 1961 C. P. Slichter, Principles of Magnetic Resonance, Springer-Verlag, 1980 R. R. Ernst and G. Bodenhausen and A. Wokaun, Principles of Nuclear Magnetic Resonance in One and Two Dimensions, Oxford University Press, 1987 M. H. Levitt, Spin Dynamics, Basics of Nuclear Magnetic Resonance, Wiley, 2001 P. T. Callaghan, Principles of Nuclear Magnetic Resonance Microscopy, 1991
Bibliography
[1] J. I. Kaplan. Chemical exchange eects on spectra. In David M. Grant and Robin K. Harris, editors, Encyclopedia of Nuclear Magnetic Resonance, pages 12471256. John Wiley & Sons, 1995. [2] D. W. McCall, D. C. Douglass, and E. W. Anderson. Self-diusion studies by means of nuclear magnetic resonance spin-echo techniques. Ber. Bunsenges. Physik. Chem., 67:336340, 1963. [3] P. Stilbs. Fourier transform pulsed-gradient spin-echo studies of molecular diusion. Prog. Nucl. Magn. Reson. Spectros., 19:145, 1987. [4] E. O. Stejskal and J. E. Tanner. Spin diusion measurements: Spin echoes in the presence of a time-dependent eld gradient. J. Chem. Phys., 42(1):288292, 1965. [5] M. Bloom and M. A. LeGros. Direct detection of two-quantum coherence. Can. J. Phys., 64:1522 1528, 1986. [6] M.-Y. Liao and G. S. Harbison. The nuclear hexadecapole interaction of iodine-127 in cadmium iodide measured using zero-eld two dimensional nuclear magnetic resonance. J. Chem. Phys., 100(3):1895 1901, 1994. [7] M. H. Levitt. The signs of frequencies and phases in NMR. J. Magn. Reson., 126:164182, 1997. [8] R. R. Ernst, G. Bodenhausen, and A. Wokaun. Principles of Nuclear Magnetic Resonance in One and Two Dimensions. Oxford University Press, Oxford, 1987. [9] Malcolm H. Levitt. Spin Dynamics: Basics of Nuclear Magnetic Resonance. Wiley, New York, 2001.
227

NMR For The People

Hochgeladen von

Dokumentinformationen

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

NMR For The People

Hochgeladen von

Copyright:

Verfügbare Formate

Nuclear Magnetic Resonance

for the People

October 31, 2011

7 The Fourier Transform 7.1 7.2

8 Measuring Chemical Exchange - The Modied Bloch Equations

ii 10 Measuring Relaxation Times

CONTENTS 17 Inside the Spectrometer

For the Workers

18 The Faraday Detector 19 Vector Rotation

iv 20 The Bloch Equations

v 22.4.0.1 Probability, Charge and Current Density . . . . . . . . . . . . . . . . . . 190

P. J. Grandinetti, October 31, 2011

For the People

H3C(25) 13 14 21 (26) C H3 9 C(22)H3 5 10 8 23 18 H3(22)C 18 4 6 11 20 C(22)H3 16 17 12 HO 3 1 7 2 15

100 Frequency (ppm)

Figure 1.1: 1 H-decoupled

C NMR spectrum of Cholesterol. There are 26 resonances in this spectrum,

P. J. Grandinetti, October 31, 2011

Precessing Tops and the Faraday Detector

CHAPTER 2. PRECESSING TOPS AND THE FARADAY DETECTOR

(t) = || [sin cos(0 t + 0 ) ex + sin sin(0 t + 0 ) ey + cos ez ] ,

P. J. Grandinetti, October 31, 2011

CHAPTER 2. PRECESSING TOPS AND THE FARADAY DETECTOR

P. J. Grandinetti, October 31, 2011

The Zeeman Interaction

CHAPTER 3. THE ZEEMAN INTERACTION

Natural Abundance (Percent)

gyromagnetic ratio (107 rad/T-s)

Quadrupole Moment Q (barns) 2 .8 103

1H 2H 3H 3 He 6 Li 7 Li 11 B 13 C 14 N 15 N 17 O 19 F 23 Na 25 Mg 27 Al 29 Si 31 P 35 Cl 69 Ga 71 Ga 67 Zn 87 Rb 107 Ag 109 Ag 235 U

P. J. Grandinetti, October 31, 2011

The Chemical Shift Interaction

CHAPTER 4. THE CHEMICAL SHIFT INTERACTION

P. J. Grandinetti, October 31, 2011

Magnetic Resonance, Coherence, and Relaxation

P. J. Grandinetti, October 31, 2011

P. J. Grandinetti, October 31, 2011

CHAPTER 5. MAGNETIC RESONANCE, COHERENCE, AND RELAXATION

P. J. Grandinetti, October 31, 2011

CHAPTER 5. MAGNETIC RESONANCE, COHERENCE, AND RELAXATION

Compound water ethanol glycerol glycerol chloroform-d 0.5 M sucrose in H2 O methyl-cyclohexane

Phase liquid liquid liquid liquid liquid liquid

Table 5.1: Spin-lattice relaxation times, T1 , for selected substances.

and if was pointing along x or y then we would expect to measure z = 0.

Randomly oriented spin 1/2 nuclei measure z

CHAPTER 5. MAGNETIC RESONANCE, COHERENCE, AND RELAXATION

Randomly oriented spin 1 nuclei measure z

P. J. Grandinetti, October 31, 2011

CHAPTER 5. MAGNETIC RESONANCE, COHERENCE, AND RELAXATION

P. J. Grandinetti, October 31, 2011

The Bloch Equations

CHAPTER 6. THE BLOCH EQUATIONS

6.1. FREE PRECESSION

components are unaected by rotation about the z -axis.

related to the transitions + 1 2

1 +2 , respectively, in an ensemble of uncoupled spin I = 1 +2 1 2 1 2

nuclei. Additionally, the component M0 is related to + 1 2

1 2 , that is, the dierence in

populations of states in an ensemble of uncoupled spin I =

nuclei. P. J. Grandinetti, October 31, 2011

nents in the rotating frame, is given by

(rf) as = (sign ) rf to yield an expression for e that is independent of the sign of .

P. J. Grandinetti, October 31, 2011