Sunteți pe pagina 1din 107

CARBON-13 NUCLEAR MAGNETIC RESONANCE

COUPLED RELAXATION STUDIES OF


MACROMOLECULAR DYNAMICS

by
Russell A. Brown

A dissertation submitted to the faculty of


The University of Utah
in partial fulfillment of the requirements for the degree of

Doctor of Philosophy

Department of Chemistry
The University of Utah
March 1994
Revised June 2013

Copyright Russell A. Brown 1994, 2013


All Rights Reserved

THE UNIVERSITY OF UTAH GRADUATE SCHOOL

SUPERVISORY COMMITTEE APPROVAL

ofa~nadonsubnllttedby

Russell A. Brown

This dissenarion has been read by each member of the following supervisory committee
and by majority vote has been found to be satisfactory.

David M. Grant

~1i.~
Richard H. Boyd

William H. Breckenridge

William

tf. Epstein

Michael D. Morse

THE UNIVERSITY OF UTAH GRADUATE SCHOOL

FINAL READING APPROVAL

To the Oradwuc Council of the University of Utah:


I have read the dissertation of
Russell A. Brown
in its fmal
form and have found that (1) its format, citations and bibliographic style are consistent
and acceptable; (2) its illustrative materials including figures, tables, and chans are in
place; and (3) the final manuscript is satisfactory to the supervisory committee and is
ready for submission to The Graduate School.

r-.d.- ~
;

/97.7'

d~/H. ~
David M. Grant
0I8ir. Supervisory Commiace

e Majer Department

Peter J. Stang

CIIir/Dean

Ann W. Hart {.
Dean of The Graduare Scbool

ABSTRACT
The carbon-13 multiplet spin-lattice relaxation for the 13CH2 spin system is
studied outside of the limits of extreme narrowing. Various novel pulse sequences are
used to perturb the spin system away from thermal equilibrium and create initial
conditions for the relaxation. The relaxation of the spin system is simulated by applying
the Redfield formalism to relaxation of each element of the density matrix. This
approach simulates not only the spin-lattice relaxation that occurs following the
perturbing pulse sequence, but also the transverse relaxation that occurs during the
dephasing delays inserted into the pulse sequence. Since relaxation simulation is not
limited to spin-lattice relaxation, the relaxation matrix contains oscillating terms that
complicate solution of the Redfield differential equation. A novel transformation of
variables removes these oscillating terms and permits straightforward solution of the
differential equation.
Outside of the limits of extreme narrowing, the spectral lines broaden to a
sufficient extent that the individual lines of the multiplet cannot be completely resolved.
For this reason the computer simulation of the relaxation generates a theoretical spectrum
that is fit to the experimental spectrum via nonlinear least squares without considering
individual spectral lines. The fitting procedure uses the Favro diffusion model to obtain
molecular rotational diffusion coefficients describing the rate of rotation of the molecule
in a liquid. This technique has been utilized to probe the tumbling motion of a peptide in
aqueous solution.

CONTENTS
ABSTRACT ...................................................................................... .iv
LIST OF FIGURES .............................................................................. vi
LIST OF TABLES ............................................................................... .ix
ACKNOWLEDGMENTS ........................................................................ x
1.

INTRODUCTION .......................................................................... 1
1.1

2.

RELAXATION MATRIX COEFFICIENTS ............................................ 6


2.1

3.

References ........................................................................ 81

RELAXATION OF ABU23_[2- 13 C]GLy24_GCN4-pl.. ............................ 82


8.1

9.

References ........................................................................ 73

RELAXATION OF [2_ 13 C]GLYCINE ................................................. 74


7.1

8.

References ........................................................................ 62

SPECTRAL SIMULATION ............................................................. 63


6.1

7.

References ........................................................................ 54

PULSE SEQUENCE SIMULATION .................................................. 55


5.1

6.

References ........................................................................ 30

RELAXATION EXPERIMENTS ....................................................... 31


4.1

5.

References ........................................................................ 19

THE FAVRO DIFFUSION MODEL. .................................................. 20


3.1

4.

References ......................................................................... 5

References ........................................................................ 92

THE AXY SPIN SYSTEM............93


9.1

References.....96

LIST OF FIGURES
2.1. The energy level diagram including the eigenstates 11) through 18) of the
static Hamiltonian written in tenns of the spin-product states

1f3f3f3)

laaa) through

for a 13CH2 (AX2) spin system .................................................. 7

3.1. The coefficients for the chemical shielding anisotropy may be used in
equation [3.22] to calculate reduced spectral densitiy functions j ':I' ({J))
for second rank relaxation mechanisms ................................................ 28
4.1.

The sandwich sequence experiment perturbs and measures M 4 .................... 33

4.2. Precession and dephasing of the carbon transverse magnetization during the
delay periods of duration 'r = 1/4JCH are represented by clockwise rotation
of vectors LJ, L2 and L3 in the X,Y plane .............................................. 35

4.3. Precession and dephasing of the proton transverse magnetization during the
delay periods of duration 'r = 1/4JCH are represented by clockwise rotation
of vectors L4 and Ls in the x,Y plane .................................................. 36
4.4.

Precession and dephasing of the proton transverse magnetization during the


delay periods 'r = 1/4JCH are represented by clockwise rotation of vectors

L4 and Ls in the X,Y plane ............................................................... 38


4.5.

The INEPT experiment perturbs and measures Ms .................................. 41

4.6.

The proton inversion INEPT experiment perturbs M2 and permits


measurement this proton mode in the carbon NMR spectrum...................... 43

4.7.

The refocused INEPT experiment perturbs and measures M 3 ..................... .45

4.8. Precession and rephasing of the carbon transverse magnetization during the
delay periods of duration Il = 1/8JCH are represented by clockwise rotation
of vectors Lb L2 and L3 in the X,Y plane ............................................. .46
4.9. The multiple-quantum INEPT experiment perturbs and measures M4 ............ .48
4.10. Precession, dephasing and rephasing of the proton-proton double-quantum
coherences during the relaxation periods of duration D2/2 are represented
by clockwise rotation of vectors Qlo and Q2 in an x,Y coordinate system........ .49
4.11. The carbon inversion INEPT experiment perturbs and measures Ml .............. 51
5.1. The procedure direct provides a direct product operator............................. 56
5.2. Basis and rotation matrices defined in Maple......................................... 57
5.3. The procedure sim concatenates three matrices and simplifies the result.......... 59
6.1. Elements of the relaxation matrix R .................................................... 65
6.2. Elements of the relaxation matrix S .................................................... 68
7.1. Experimental data indicated by diamonds for Ml and by crosses for M4 are
shown for application of the sandwich sequence experiment to
[2_13C] glycine in glycerol-ds at 303 K ................................................ 76
7.2. The principal axis system of the molecular rotational diffusion for glycine....... 77
7.3. Semilog plots of Rl./T and RII/T versus T-l for [2_13C]glycine in
glycerol-ds demonstrate an Arrhenius relationship across a temperature
range from 288 K to 323 K ............................................................. 80
8.1. A helical wheel diagram of the leucine zipper peptide GCN4-p 1................... 83

vii

8.2. Theoretical and experimental spectra are shown for application of the
multiple-quantum INEPT experiment to abu23_[2-I3C]gly24-GCN4-pl in
D20 at 303 K .............................................................................. 86
8.3. Experimental data indicated by diamonds for MI' by crosses for M3 and by
squares for M 5 are shown for application of the refocussed INEPT
experiment to abu 23 _[2-I3C]gly24-GCN4-pl in D20 at 308 K ..................... 88
8.4.

Semilog plots of Rl./T and RII/T versus T-I for the leucine zipper peptide
abu 23 _[2-13C]gly24-GCN4-pl in D20 demonstrate an Arrhenius relationship
across a temperature range from 293 K to 308 K .................................... 91

viii

LIST OF TABLES
3.1. Eigensystem of the Asymmetric Rotor (/ = 2) ........................................ 24
4.1. Perturbed State of the Spectral Lines ................................................... 53
6.1. Nonlinear Least-Squares Fitting Parameters .......................................... 71
7.1. Rotational Diffusion Coefficients for [2_13C]Glycine in Glycerol-dg ..............78
8.1. Rotational Diffusion Coefficients for Abu23 -[2- 13C]Gly24_GCN4-pl.. .......... 90

ACKNOWLEDGMENTS
The research reported in this dissertation was supported by National Institutes of
Health grants CAOl607, GM08521 and 5P30CA42014.

CHAPTER 1
INTRODUCTION
The measurement of spin-lattice relaxation in coupled nuclear spin systems via
13C nuclear magnetic resonance (NMR) spectroscopy has been used to study numerous
chemical systems (1-5). Accurate measurement of spin-lattice relaxation in liquids may
be used to obtain molecular dynamical parameters such as rotational correlation times or
rotational diffusion coefficients. These rotational parameters may be obtained from
relaxation studies because molecular tumbling causes temporal modulation of the dipoledipole interaction and chemical shielding anisotropy (6), two well established spin
relaxation mechanisms. Although many different spin systems have been probed via
relaxation studies, the AX2 spin system associated with a methylene 13CH2 moiety is
particularly amenable to these studies and provides a wealth of molecular dynamical
information.
The relaxation process may be modeled using Redfield theory (7) wherein the
time derivatives of the density matrix elements O'~,(t) are expressed using the
following coupled differential equation

In this equation, the terms 0'<;],(00) represent the thermal eqUilibrium values of the
density matrix elements, and the terms

i( WIC -

wIC'

- W A + W A' )t RICIC,A.A'

form a time-

dependent relaxation matrix in the interaction representation of the static magnetic field
(8). This equation admits two simplifications. First, in the eigen basis the explicit time
dependence is removed for

diagonal tenns when


m/C - m/C,

= mAo -

IC= IC'

mAo"

and A = A', and for off-diagonal terms whenever

Second, a secular approximation eliminates the remaining

rapidly oscillating tenns when m/C

-m/C,:;I!:

mAo -mAo"

The surviving terms form a

block-diagonalized relaxation matrix in the eigen baisis. Matching this relaxation matrix
to experimental data describing the time evolution of the spin system provides a solution
to equation [1.1] and determines values for ~.
The differential equation, solved in the eigen basis, is used to generate a
theoretical NMR spectrum that is fit to the experimental spectrum via least-squares.
Since the populations of the eigenstates are not experimentally observable, the
interpretation of the solution to the differential equation is facilitated by transfonnation
from the eigen basis into a magnetization mode basis. Magnetization modes may be
defined either as linear combinations of the eigenstates or as linear combinations of the
spectral line intensities. One convenient set of magnetization modes is expressed as
linear combinations of the spectral lines in the carbon and proton spectra of the 13CH2
moiety (9,10) as follows

[1.2]

In equation [1.2], Ll, L2 and L3 are the intensities of the carbon triplet lines
listed in order from high to low frequency, and L4 and Ls are the intensities of the
proton doublet lines, also listed in order from high to low frequency. Moreover,

intensities L2, L4 and L5 may be expressed as the sum of two components that typically
are not separately observable in isotropic liquids, e.g., L2

=L2 + L 2.

In isotropic

liquids, M}, M3 and M5 are directly observable via carbon NMR spectroscopy, whereas

M2 and M5 are directly observable via proton NMR spectroscopy. M6 is observable


only in proton NMR spectra of ordered phases such as the nematic phase of liquid
crystals (11,12). M7 is not observable but has the physical interpretation of the
population differences between the triplet and singlet proton states. M 4 has been
observed indirectly via carbon NMR difference spectroscopy (13) using a slightly
different magnetization-mode basis (2) than formulated in equation [1.2].
The aim of a coupled relaxation experiment is to perturb one or more
magnetization modes away from thermal equilibrium by using a sequence of radio
frequency pulses, and then to sample the temporal evolution of the modes as they return
to thermal equilibrium through spin-lattice relaxation. Macromolecules present several
challenges for coupled relaxation studies. First, the solubility of macromolecules is
typically on the order of one to 10 mM. This concentration range results in an NMR
spectrum having a lower signal to noise ratio than is obtainable with more soluble
molecules. The signal to noise ratio may be improved by utilizing pulse sequences that
transfer magnetization from the sensitive IH nucleus to the insensitive 13C nucleus, thus
amplifying the 13C signal by up to a factor of four. Second, relaxation is so rapid in
macromolecules that substantial relaxation occurs during the pulse sequence. This
problem is solved by modeling the relaxation of the spin system both during the pulse
sequence and during the subsequent spin-lattice relaxation period. Third, in the NMR
spectrum of macromolecules the spectral lines broaden to an extent that they overlap and
cannot be completely resolved. This difficulty is overcome by using computer
simulation to produce a theoretical 13C NMR spectrum that is fit to the experimental
spectrum via nonlinear least squares minimization. The fitting procedure fits the
theoretical spectrum to the experimental spectrum without considering individual spectral

lines, thus the overlapping spectral lines present no problem. Fourth, the motion of
macromolecules in liquids occurs outside of the extreme narrowing limit. This slow
motional regime gives rise to a large number of spectral density functions /I because the
spectral density functions are frequency dependent for this motional regime (14). Since
the relaxation matrix coefficients R~ in equation [1.1] are composed of linear
combinations of the spectral density functions, the use of these spectral density functions
as fitting parameters can result in a large number of fitting parameters. However, it is
possible to reduce substantially the number of fitting parameters by using the Favro
diffusion model (15) to calculate the spectral density functions from a small number of
parameters describing the molecular rotational diffusion and the chemical shielding
anisotropy. This approach has the additional advantage that it expresses the solution to
equation [1.1] in terms of rotational diffusion coefficients describing the rate of
molecular rotation diffusion about an orthogonal set of coordinate axes. These axes are
known as the principal axes of the molecular rotational diffusion tensor.
Subsequent chapters of this thesis describe the experimental and simulation
methods required for coupled relaxation studies. Chapter 2 discusses the expression of
the relaxation matrix coefficients ~ in terms of the spectral density functions

/I.

Chapter 3 shows how the Favro diffusion model relates the spectral density functions to
molecular rotational diffusion coefficients and chemical shielding anisotropy parameters.
Chapter 4 presents several new spin-lattice relaxation experiments for perturbing
magnetization modes and measuring their temporal evolution. Chapter 5 describes
algebraic techniques for simulating relaxation experiments. Chapter 6 discusses numeric
simulation and theoretical NMR spectra generation. Chapter 7 discusses the results of
relaxation experiments used to study the dynamics of [2_13C]glycine dissolved in
glycerol-ds. Chapter 8 discusses the results of relaxation experiments used to study the
dynamics of the leucine zipper peptide abu23_[2-13C]gly24-GCN4-pl dissolved in D20.

1.1 References
1.

R. R. VoId and R. L. VoId, Prog. NMR Spectrosc. 12, 79 (1978).

2.

D. M. Grant, C. L. Mayne, F. Liu, and T. X. Xiang, Chern. Rev. 91, 1591


(1991).

3.

H. W. Spiess, "NMR Basic Principles and Progress" (P. Diehl, E. Fluck, and R.
Kosfeld, Eds.), Vol. 15, p. 55, Springer-Verlag, New York, 1978.

4.

L. G. Werbelow and D. M. Grant, "Advances in Magnetic Resonance" (J. S.


Waugh, Ed.), Vol. 9, p. 189, Academic Press, New York, 1977.

5.

D. Canet, Prog. NMR Spectrosc. 21, 237 (1989).

6.

J. McConnell, "The Theory of Nuclear Magnetic Relaxation in Liquids", p. 99,


Cambridge University Press, Cambridge, 1987.

7.

A. G. Redfield, IBM Res. Develop. 1, 19 (1957).

8.

K. Blum, "Density Matrix Theory and Applications", p. 55, Plenum Press, New
York,1981.

9.

Z. Zheng, C. L. Mayne, and D. M. Grant, J. Magn. Reson. 103, 268 (1993).

10. L. G. Werbelow and D. M. Grant, J. Chern. Phys. 63, 544 (1975).


11. L. G. Werbelow, D. M. Grant, E. P. Black, and J. M. Courtieu, J. Chern. Phys.
69, 2407 (1978).
12. P. Diehl and C. L. Khetrapal, "NMR Basic Principles and Progress" (P. Diehl, E.
Fluck, and R. Kosfeld, Eds.), Vol. 1, p. 1, Springer-Verlag, New York, 1969.
13. M. M. Fuson and J. H. Prestegard, J. Magn. Reson. 41, 179 (1980).
14. M. Goldman, "Quantum Description of High-Resolution NMR in Liquids", p. 232,
Oxford University Press, Oxford, 1988.
15. L. D. Favro, Phys. Rev. 119, 53 (1960).

CHAPTER 2
RELAXATION MATRIX COEFFICIENTS
It is possible to define the relaxation matrix coefficients R::::~ in terms of the
reduced spectral density functions J :;,' (mcl)' In this nomenclature, J1. and J1.' specify
relaxation mechanisms such as dipole-dipole interactions or chemical shielding
anisotropy, s and s' designate either single spins or pairs of interacting dipolar spins,
and

mICA. designates a transition frequency between eigenstates I( and A. The specific

equations relating

R::::;;' to J ::' (mICA.) are (1-3)

[2.1]
S,s'

~)-l)m(l(lf~(J1.,s)IA)(A'lf~m(J1.',s')II(')J:;" (mcl)
m=-l

In these equations

DICA. represents a Kronecker delta.

~f is an interaction constant

specific to the relaxation mechanism J1. and the spin (or spin pair) s. The eigenstates II()
and

IA) are defined in terms of the eight spin-product functions of the AX2 spin system

as shown in Figure 2.1, e.g., eigenstate 17) =

(I/l/la) + l/la/l)/..fi .

In equation [2.1] summation occurs over all components m of rank I irreducible


spherical tensor operators T(J1.,s) and T(J1.',s'). Also, the reduced spectral density
functions

J :;" (mcl)

used in this equation are derived from the irreducible spherical

tensors F(J1.) and F(J1.'), as discussed in Chapter 3. The spin tensor T(J1., s) and the

18) = 1f3f3f3)

14) = laf3f3)

.~

L5

L 4"

17) =

{f(I f3f3a) +If3a(3) )

~~

12) = {f(laaf3 ) + laf3a))

L'2

r=
L"2

.~

L5

15) = lf3aa)

I~

16) = {f(lf3f3 a) -1f3a(3))

(f

13) = ~2(1aaf3) -l af3a))

L4

.~

11) = laaa)

Figure 2.1. The energy level diagram including the eigenstates 11) through 18) of the

laaa) through 1f3f3f3) for a


For the spin-product states, the frrst a,f3 spin label gives

static Hamiltonian written in tenns of the spin-product states


13CH2 (AX2) spin system.

the state of the carbon spin while the second and third labels give the states of the two
identical protons. L 1, L 2, L2 and L3 represent the single-quantum transitions giving
rise to the carbon lines, while L4, L4, L; and L5 represent the single-quantum
transitions giving rise to the proton lines.

8
spatial tensor F(p) define a Hamiltonian via the scalar product
1{Jl

= L~fLL(-I)mF~(p)f~m(P)

[2.2]

1 m

Since experimental observables are measured in the laboratory coordinate


system, equation [2.1] and consequently
in this coordinate system.

T(p,s)

The spin tensor

and F(p), must ultimately be defined

T(p,s)

typically contains no spatial

information, although in the case of the chemical shielding anisotropy it contains time
independent spatial infonnation describing the static magnetic field Bo. Hence it may be
written in an arbitrary Cartesian coordinate system and subsequently transformed into
the laboratory coordinate system without difficulty. In contrast, the spatial tensor F(p),
or more correctly the time dependent spatial tensor F(p, t), characterizes the temporal
dependence of the molecular tumbling. This tensor cannot be evaluated in the laboratory
coordinate system for every individual molecule in an ensemble. A solution to this
problem is to obtain

F(p,t) via rotation of the time independent spatial tensor

~ (p)

F~(p,t) = L ~ ~'.m[.o-1(t)]!F ~, (p)

[2.3]

m'=-l
~ (p) is defined in the principal axis system of the molecular rotational diffusion.

~ ~'.m[.o-l(t)] represents the elements of the time dependent Wigner rotation matrix that
transforms the tensor from the principal axis system of the molecular rotational diffusion
into the laboratory coordinate system (4). Use of equation [2.3] permits estimation of
ensemble averages for

F(p,t),

as discussed in Chapter 3. For now, it is sufficient to

recognize that F(p, t) may be obtained from ~ (p).


For relaxation mechanisms arising from either dipole-dipole interactions or
chemical shielding anisotropy, the tensors ~ (p) and T(p, s) are second rank spherical
tensors (5,6). In order to illustrate how to derive expressions for these tensors, the
chemical shielding anisotropy will be taken as an example, and ~ (CSA) and T(CSA) will

be derived. The chemical shielding Hamiltonian (including isotropic and anisotropic


contributions) defined in an arbitrary Cartesian coordinate system for nucleusj is

[2.4]

In this equation

Box, Boy and Boz represent the Cartesian components of the static

magnetic field Bo.


nucleus j,

rj

11, 11

and

11 are the spin angular momentum operators for

represents the gyromagnetic ratio for nucleus j, and OJ represents the

chemical shielding tensor, a second rank Cartesian tensor.


Inspection of equation [2.4] allows direct derivation of the spin tensor T( CS, j).
First, T(CS,j) is created via a direct product

BOX]

[Boil

T(CS,j) =[Boy

[i1 11 il] = BOYY

Boz

Boll

[2.5]

Then T(CS,j) is transformed into its irreducible spherical components in an arbitrary


Cartesian coordinate system (7)

[2.6]

"(2)
" . i [Boy!l
" . + Box!;
" ']}
T2
(CS,j) = 21 {Box!"1. - Boy!;

10
The spin tensor T( CS,j) contains spatial infonnation in the form of the
components Box, BOy and Boz of the static magnetic field Bo. Therefore, T( CS, j) may
be transfonned into the laboratory coordinate system by substituting for these
components in equation [2.6]. In the laboratory coordinate system where the z axis is
the quantization direction, Box = BOy = 0 and Boz = Bo. Making these substitutions in
equation [2.6] yields

fJO) (CS,j) = -{fijBo


fJl)(CS,j) = 0

f~~)(cS,j) = -!:.i4Bo
2

j;(2)(CS ).) - [2jj B

'-'{3'z

A(2)(

.)

A(2)

Tl CS,)

[2.7]

Aj
=_1
+ 2 IBo

T2 (CS,))=O
Only the anisotropic portion of the chemical shielding Hamiltonian gives rise to
relaxation. It is therefore necessary to derive ~ (CSA, j) in order to detennine which
components of T(CS,j) contribute to the anisotropic chemical shielding and are included
in T( CSA, j). From the Cartesian tensor OJ, it is possible to create three irreducible
Cartesian tensors of rank zero, one and two (8)

[2.8]
1 0
= O"iso [ 0

11

In equation [2.8] O"iso' a uv and

'UV are defined as

[2.9]

In equation [2.8], the zero rank: tensor containing O"iso is isotropic. It is therefore
invariant to molecular tumbling and does not contribute to relaxation. The fIrst rank
tensor containing
'uv

a uv is antisymmetric and traceless. The second rank tensor containing

is symmetric and traceless, as can be appreciated from equation [2.9]. Both the fIrst

and second rank tensors can in principle contribute to relaxation. However, theoretical
studies indicate that the fIrst rank term is small compared to the second rank term, except
in the case of a distorted

1t

bond (9). No such distortion is present in the molecules

presented in this dissertation, so the fIrst rank term is not considered further.
The Cartesian tensor C5j is transformed into its irreducible spherical components
in an arbitrary Cartesian coordinate system (7)
1 (
)
r;;
--.J3
O",X;t+O"yy+O"zz j=--v30"iso

(0) _

0"0

ag) = - ~(axy -ayx)j ~ 0


(1(1)
1 --

0"&2)

-'!'[a
-0
2 zx - a xz +
- z'(a zy - a yz )] j -

~[3'zz -('xx +'yy +'zz)t

(1~;) =+~ ['xz + 'zx i('yz + 'zy )

[2.10]

12

Equation [2.10] may be simplified by rewriting OJ in the principal axis system of the
chemical shielding, designated by a tilde. In this principal axis system, the second rank
Cartesian tensor

~j

is diagonal, i.e.,

'xy ='xz ='YZ =0, hence


[2.11]

The shielding anisotropy parameter ..1 and the asymmetry parameter 17 are defined as (10)

[2.12]

In equation [2.12] labeling of the principal axes follows the convention

[2.13]

Other conventions are possible (10).


Now OJ is rotated from the principal axis system of the chemical shielding into
the principal axis system of the molecular rotational diffusion. The components of the
rotated spherical tensor cg; (CSA, j) are calculated as
I

~ ~ (CSA,j) = l~ ~',m(.f.rl) a-~,

[2.14]

m'=-l

In this equation, ~ ~"m(DT1) represents the elements of the Wigner rotation matrix
elements and DTI represents the Euler angles

-aj'

-f3j

and

-rj

effecting the rotation.

13

It is convenient to express the Wigner rotation matrix elements as (11)

~ ~"m(.QTl) =~ ~',m(-rj,-/3j,-aj)
= irjm' d~',m(-/3j) iajm

=[e -ir jm'd~"m (-/3j) e-ia jm ]*


=

[e -iajm d~,m'(/3j) e-irjm']*

= [~ ~m' (a j' /3 j' r j )]

[2.15]

=~ ~~m,(.Qj)
This transformation is possible because the reduced Wigner rotation matrix elements

d~.m'(/3j) are real. Since equation [2.11] indicates that the fIrst rank: components of OJ
equal zero, substituting equation [2.15] into equation [2.14] and expanding gives

[2.16]

Substituting for the Wigner rotation matrix elements in equation [2.16] and factoring
L1 j ..J811:/15 from the result yields the components of ?; (CSA,j)

X{+3sin2/3j + 211j sin/3j[isin2rj cos/3j cos2rj]}

g;

ll) (CSA,j) = ~ 9~11: e2iaj


X{3sin 2 /3j + 11j[cos 2rAI + cos 2 /3j) 2i sin 2rj cos/3j]}

[2.17]

14

Since the spatial tensor rg; (CSA,}) comprises only second rank components, the spin
tensor T( CSA,}) will similarly comprise only the second rank components. Factoring

Bo/2 from the second rank components of T(CS,}) given in equation [2.7] yields
fJ2)(CSA,})

=.JI ij

fR)(CSA,}) = +i4

[2.18]

fE)(CSA,}) = 0
Written in the laboratory coordinate system, the chemical shielding anisotropy
Hamiltonian is
~ CSA

= L~fSA

L(-lr F~2)(CSA,}) f~;1(cSA,})

[2.19]

m=-2

The interaction constant ~fSA is defmed by combining the leading coefficients from
equation [2.4] with the constants Bo/2 and Lij ..J81r/15 factored from equations [2.7]
and [2.17] respectively
~.CSA = ~21r
-nrBo Li
J

15

[2.20]

At this point relaxation due to dipole-dipole interactions will be considered and


expressions for the spatial tensor rg; (DIP) and the spin tensor T(DIP) will be derived.
The dipolar Hamiltonian defined in an arbitrary Cartesian coordinate system for
interacting dipolar spins i and} is (5,6)

~ IJpIP = rirjn
[iix iiY ii].
3
z
rIJ

In this equation,

ri and rj

DXX
D

DXZ]

yx

Dyz

zx

D zz ij

[D

[il]ij

[2.21]

ij

represent the gyromagnetic ratios for nuclei i and}, and

'ij

is

15

the distance between these nuclei. Dij is defined in terms of the unit vector eij pointing
from nucleus i to nucleus j

-3exey
1- 3eyey

[2.22]

-3eze y
When Dij is represented in the principal axis system of the dipolar interaction
where ex = ey = 0 and ez = 1, it comprises only the second rank Cartesian tensor

Dij

[1 01 0]0

= 0

[2.23]

o 0 -2
This tensor may be transformed into its irreducible spherical components (7)

D~O) =0

no(l) --

D(l) -

l-

D~2) =-.J6

[2.24]

(2) - D(2) - 0
Dl
2-

The single nonzero component D~2) is rotated from the principal axis system of the
dipolar interaction into the principal axis system of the molecular rotational diffusion.
Factoring -.J241r/5 from the result yields the spatial tensor ?J (DIP,ij)

[2.25]

16
In equation [2.25] Y~(/3ij,aij) are the spherical hannonic functions (12). The spherical
polar angles /3ij and aij orient the dipole-dipole axis relative to the principal axis system
of the molecular rotational diffusion. Comparison of equations [2.17] and [2.25]
reveals that CZF (DIP,ij) =CZF (CSA,j) when 71j = O. This equivalence arises because (13)

~ :::0 (a,/3, y) = ~ 2/+1


41r Y~(/3,a)

[2.26]

Hence the expression for CZF (CSA,j) is a more general equation subsuming the
expression for the axially symmetric CZF (DIP,ij).
An expression for the spin tensor

TiflP

may be derived by proceeding in a

manner analogous to the derivation of equation [2.18]. Factoring 1/2 from the result
and keeping only the second rank components yields

j
i O(2)(DIP ,i'J) = _1_[4jij
jj - ji-+
jj]
.J6 zz - ji+A(2)

..

_[AiAj

AiAj]

Tl (DIP,lj) = + IIz + IzI

[2.27]

fE)(DIP,ij) = j~j1
As the spin tensor T(DIP,ij) contains no spatial information, its definition in an arbitrary
Cartesian coordinate system is equivalent to its definition in the laboratory coordinate
system. Hence there is no need to transform it into the laboratory coordinate system.
The interaction constant ~fIP is defined by combining the leading coefficients
from equation [2.21] with the constants -..J241r/5 and 1/2 factored from equations
[2.25] and [2.27] respectively

)!DIP _ ~61r YiY/l

<:'ij

- -

-5 - - 3 r.I)..

[2.28]

In addition to the chemical shielding anisotropy and dipolar Hamiltonians, a

17

random field Hamiltonian may be used to model relaxation mechanisms such as spinrotation and intermolecular dipole-dipole interactions (1). For this Hamiltonian the spin
tensor T(RDM,}) for nucleus} is a first rank tensor (6)

fJI)(RDM,}) =
T(l) (RDM

i1

J') - =+= {1 j j
, - V2

[2.29]

Once the spin tensors for the various relaxation mechanisms are defined by
equations [2.18], [2.25] and [2.29], expansion of equation [2.1] yields expressions for
the relaxation matrix coefficients
functions

.f:;" (mcl)'

R.:/:;;'

in terms of the reduced spectral density

Automatic generation of these expressions by a computer

algebra system (14) is facilitated if equation [2.1] is rewritten specifically for each
relaxation mechanism. For the chemical shielding anisotropy Hamiltonian, equation
[2.1] becomes a summation over nuclei i and}
2

K;Ec,A.' (mcl) = L L ~fSA ~fSA


i

where

L (_I)m
m=-2

[2.30]

.f iSC (mICA.) designates eSA-eSA reduced spectral density functions.


For the dipolar Hamiltonian, equation [2.1] becomes a summation over

interacting dipolar spins i and}, and k and I

i<jk<l

where

m=-2

[2.31]

.f J.~(mKA.) designates dipolar-dipolar reduced spectral density functions.


Since the eSA and dipolar Hamiltonians both comprise second rank tensors,

18
cross-correlation spectral densities arise coupling these two mechanisms (1). In this
case equation [2.1] becomes a summation over nucleus k and interacting dipolar spins i
and}

i<j k

m=-2

x{(IClf~ (DIP,i})IA)(A'lf~m (CSA,k )IIC')

[2.32]

+(IClf~(CSA,k)IA)(A'lf~m(DIP,i})IIC')} rf iff (mICA)


where

rf f,3 (mICA) designates dipolar-CSA reduced spectral density functions.


Finally, for the random field Hamiltonian, equation [2.1] becomes a summation

over nuclei i and}

where

rflt(md)

[2.33]

m=-l

designates random-random reduced spectral density functions. In

this last equation ~RDM is left undefmed, which does not present a problem since for the
random-field interactions ~fDM ~rM may be absorbed by

rf i~R (md),

as discussed in

Chapter 6.

2.1 References
1.

R. R. VoId and R. L. VoId, Prog. NMR Spectrosc. 12, 79 (1978).

2.

D. M. Grant, C. L. Mayne, F. Liu, and T. X. Xiang, Chern. Rev. 91, 1591


(1991).

3.

S. A. Smith, "Simulations of Pulsed Nuclear Magnetic Resonance Experiments",


Doctoral Dissertation, University of California Santa Barbara, 1991.

4.

R. N. Zare, "Angular Momentum", p. 177, John Wiley & Sons, New York, 1988.

5.

S. A. Smith, W. E. Palke and J. T. Gerig, Conc. Magn. Reson., 4, 107 (1992).

19
6.

S. A. Smith, W. E. Palke and J. T. Gerig, Cone. Magn. Reson., 4, 181 (1992).

7.

M. Mehring, "Principles of High Resolution NMR in Solids", 2nd Ed., p. 290,


Springer-Verlag, New York, 1983.

8.

U. Haeberlen, "High Resolution NMR in Solids, Selective Averaging", Supp. 1,


p. 8, Academic Press, New York, 1976.

9.

J. Facelli, A. M. Orendt, D. M. Grant and J. Michl, Chem. Phys. Lett., 112, 147
(1984).

10. J. Mason, Solid State Nucl. Magn. Reson., 2,285 (1993).


11. R. N. Zare, "Angular Momentum", p. 85, John Wiley & Sons, New York, 1988.
12. R. N. Zare, "Angular Momentum", p. 8, John Wiley & Sons, New York, 1988.
13. R. N. Zare, "Angular Momentum", p. 97, John Wiley & Sons, New York, 1988.
14. B. W. Char, K. O. Geddes, G. H. Gonnet, B. L. Leong, M. B. Monagou and S.
M. Watt, "Maple V Reference Manual", Springer-Verlag, New York, 1991.

CHAPTER 3
THE FAVRO DIFFUSION MODEL
The reduced spectral density functions

.1 ':f' (m)

used in equations [2.30]

through [2.33] are related to the spectral density functions JII#',


,(m) via (1)
mtn ,ss
11#'

J""",n
, ,(m)=(-l)

where

o-m,m' is a Kronecker delta.

11#'
o-mm'.1 ps'
(m)
,~,

[3.1]

The spectral density functions JII#',


,(m) are halfmtn ,ss

interval Fourier transforms (2,3) of the correlation functions GJIP',


,('t')
mtn ,ss

-J

J:!',n,(m) = (F~ (Jl,t)F~'(Jl',t + 't')e-im't'd't'

[3.2]

o
In this equation F~ (Jl, t) designates the I,m components of a time dependent spatial
tensor F(Jl,t) defmed in the laboratory coordinate system for relaxation mechanism Jl.
The brackets ( ) specify an average over an infinite number of time pairs t and t + 't' for
one molecule. The spatial tensor F(Jl,t) may be obtained from the spatial tensor ~ (Jl)
defmed in the principal axis system of the molecular rotational diffusion. A time
dependent rotation from the principal axis system of the molecular rotational diffusion
into the laboratory coordinate system produces F(Jl, t). This rotation is represented as
I

F~(Jl,t) = L~ ~',m[ n-1(t)].'J" ~, (Jl)

[3.3]

m'=-l
In this equation, ~ ~',m[n-l(t)] represents the elements of the Wigner rotation matrix

21
and .o-1(t) represents the time dependent Euler angles effecting the rotation. Note, the
use of .o-1(t) instead of .o(t) is necessary because the spatial tensor ~ (J.L) is rotated
instead of the coordinate axes. Equation [3.3] may not be evaluated exactly because
each molecule reorients in a random manner under the effects of Brownian motion (4)
and therefore the precise form of .o-1(t) cannot be known.
In view of the difficulty in specifying .0-1(t), a number of simplifying
approximations are used to calculate the time average contained in the spectral density
function JJlIL',
,(m). Methods for its calculation have been described (1,5) and are
min ,ss
summarized as follows. Substituting equation [3.3] into equation [3.2] yields

[3.4]

The spatial tensors ~ ~ (J.L) and ~ ~: (J.L') are time independent since they are defined in
the principal axis system of the molecular rotational diffusion. Therefore, they may be
factored out of the time average ( ) to give

-1 ]

[-1

])

JlIL'
I [
l'
' ,)
GmIn',ss'(r)
= ~(
~ ~ n,m.o
(t) ~ n',m'.o
(t + -r) ~ nI (J.L)~ n'I (J.L
non

The

time

dependence

is

now

only

in

the

correlation

[3.5]

function

(~~,m[.o-1(t)]~ ~:,m,[.o-1(t+-r)]) of the Wigner rotation matrix elements (3). It is


convenient to rewrite this correlation function as

(~~m[ .o-l(t)]~ !:,m,[n- 1(t+ -r)]) =


(_I)m-n(~ :m,-n[n(t)]~ ~:n' [.o(t+ -r)])
using the relations

[3.6]

~~m(n-1)=(-1)m-n~:m,_n(n)=~;::n(.o). This correlation

function implies an average over all possible angles

n (spin orientations) weighted by

22
the probability of occurrence of each angle

(~~m.-n[ Q(t)]~ ~~n'[ Q(t + ~)]) = ff~ ~m.-n[ Q(t)]~ ~~n'[Q(t + ~)]

[3.7]

xp[ Q(t)] p[ Q(t) IQ(t + ~)] dQ(t) dQ(t +~)

In equation [3.7] p[Q(t)] is the probability of orientation Q(t) at time t.


p[Q(t) IQ(t + ~)] is the conditional probability that the molecule will have the

orientation Q(t +~) at time t +

~,

given that at time t it has the orientation Q(t).

Equation [3.7] may be simplified under the ergodic principle, which states that the time
average over an infinite number of time pairs t and t + ~ for one molecule is equivalent
to the average over an ensemble of molecular orientations at time t = 0 (6). Applying
the ergodic principle to the probabilities in equation [3.7] gives
p[Q(t)] dQ(t) = p[Q(O)] dQ(O) = p(Qo) dQ o

[3.8]

p[Q(t)IQ(t+~)] dQ(t+~) = p(Q o IQ,~) dQ


Substitution of equation [3.8] into equation [3.7] yields

(~~m.-n[Q(t)] ~ ~~n'[Q(t + ~)]) =

f f~ ~m.-n(Qo)ig ~~n,(Q) p(Qo) p(Qo IQ,~) dQo dQ


In this equation
fluid,

f dQo = 81r

fp(Qo) dQo = 1 due to normalization (1,3).

[3.9]

In an isotropic

under the assumption that all possible initial orientations are equally

probable. Hence p( Q o) = 1/81r 2 is a normalization constant. Also, p( Q o Q, ~)


designates the conditional probability that the molecule will have the orientation Q at time
~,given

that at time zero it has the orientation Q o. Favro (7) has reported a method for

obtaining

p( Q o IQ, ~) from the following differential equation that is analogous to the

wave equation for a rigid rotor

23

[3.10]

In equation [3.10], Lk is the angular momentum operator about the kth axis and Rkk is
the kth component of the diagonalized molecular rotational diffusion tensor. This
equation must satisfy the initial condition p(.Q o 1.0,0) = 8(.0 0 ,.0), where

8is a Dirac

delta function (7-10). The solution to equation [3.10] is the Green's function (11)

p(.Qo 1.Q,'r) = L'P:(.Qo)'P,,(.Q)e-m,,'r

[3.11]

"
'P" (.0) are the eigenfunctions and

m" are the eigenvalues of the asymmetric rotor. For

a completely asymmetric rotor, the eigenfunctions 'P,,(.Q) cannot be written in closed


form. However, exact eigenfunctions 'Pl,i.Q) may be obtained by linear combination
of the symmetric rotor eigenfunctions tpf.q(.Q) for low values of p (7)

'Pl,q(.Q) = Laf(lC) tpf.q(.Q)

[3.12]

The coefficients a!(lC) and eigenvalues m: are given in Table 3.1 for p = 2 (1,5,9).
Since Wigner rotation matrix elements are eigenfunctions of the symmetric top
Hamiltonian, the symmetric rotor eigenfunctions may be expressed as (12)
mP (.0) = (_I)q-r ~ 2p +
T'r,q
8n2

1 ~!

q, _r(.0)

[3.13]

Substituting this expression sequentially into equations [3.12], [3.11] and [3.9] gives

(~:m,-n[.Q(t)]~ ~~n,[.Q(t+ 'r)]) =


L 2 p +} e-m~'rL(-I)q-r(-I)q-r' a!*(lC) a;,(lC)
p,q." 64n
r.r'

[3.14]

24

Table 3.1. Eigensystem of the Asymmetric Rotor p = 2

a~2)(1(')

W(2)

1('

IC

r:

-2

-1

-2

-~

Rxx + Ryy + 4Rzz

-1

-~

~
0

Rxx + 4Ryy + Rzz

cosX

~.
2 smx

2R-2~R2 -PA 2

~.
2 smx

4Rxx + Ryy + Rzz

J{cos x

sin X

J{cos x

2R+2~R2 -PA 2

Note: R, R2 -PA 2 and X are defined as follows:

1
X =-tan

-1[ .J3(Rxx - Ryy) ]


2Rzz - Rxx - Ryy

25
In equation [3.14], (_l)q-r(_l)q-r' = (_1)2 q-r-r' = (-lr r- r'. Moreover, the integrals

J~ ~m.-n(no)~ !:'-r(no) dn o and J~ ~~n,(n)~ !q._r,(n) dQ are evaluated using


the orthogonality properties of Wigner rotation matrix elements (13), yielding
I
1'*
\~ -m.-n
[n(t)]~ m'.n'[
n(t + -r)] ) =

[3.15]
8n20 0
) ( 8n 2
)
x( 21+ 1 I.p -m.-q 8-n.-r 21' + 1 0,I.p 0,
m ,-q0,
n ,-r ,

The Kronecker deltas and summations in equation [3.15] permit several simplifications.
The Kronecker delta o-n,-r removes all terms except the term having r = n from the
summation over r. Similarly, 0n'.-r' removes all terms except the term having r' = -n'
from the summation over r'. The two Kronecker deltas O_m,-q and Om'.-q remove all
terms except the term having q = m = -m' from the summation over q. This term is
designated by introducing the new Kronecker delta 0m,-m' = O-m,m"

These

simplifications give
/
I
[
1'* [
) = O-m,m' (-1) n'-n
\~
-m,-n
net)] ~ m',n'
n(t]
+ -r)

[3.16]

One further simplification is possible. The Kronecker deltas Ol,p and Or,p
remove all terms except the term having p = 1= l' from the summation over p. This
simplification means that no spectral density functions involve interactions of different
rank. For example, no cross correlation spectral densities arise coupling the dipolar and
random field relaxation mechanisms. Replacing bothp and l' with 1 in equation [3.16]
selects the term having p = 1= l' , yielding

26

(~~m,_n[D(t)]~ ~~n'[D(t+-r)]) = O-m,m' (~~):~n


[3.17]

xLa~* (J() a:'n,(J()e -mi-r


I(

Substituting equation [3.17] sequentially into equations [3.6] and [3.5], and noting that

a~ (J() is real yield


GJ.lIl.', ,(-r) = 8
' (_l)m
mm ,ss
-m,m 21 + 1

XL LL(-lt' e-mi-ra~(J() a:'n,(J() ~ ~ (J.l)~ ~, (f.l')


I(

[3.18]

n'

The spectral density functions JIlIL',


mm ,ss,(m) are Fourier transforms (14) of GIlIL',
mm ,ss,(-r)

,
(_l)m
JJ.lIl., ,(m)=8
,--""
mm ,ss
-m,m 21 + 1 .J

1 e-mI("e-lv-l"d-r
00

l ".

."."..

I(

xLa~(J()~ !(f.l)L(-lt' a:'n,(J()~ !'(f.l')


n

[3.19]

n'
1

_ ( l)m 8
1 "" 1 ( ) 1* ( ' )
ml(
- -m,m' 21 + 1 .J cl( f.l cl( f.l ( 1
2
I(
ml( + m

)2

The imaginary part of the Fourier transform may be neglected because it produces only a
small constant frequency shift (15). The coefficients

c~(f.l) and ct (f.l')

C~* (f.l') = L (-1 t' a~n' (J() ~ !, (f.l')

are given by

[3.20]

n'

Equations [3.1] and [3.19] reveal the reduced spectral density functions

J :;,' (m) to be
[3.21]

27
Equation [3.21] reduces to the sum of five Lorentzian functions for 1=2

1 2

.j W;' (m) = - '" c(2)(J.l) c(2) (J.l')


ss
5 4- I(
I(
(
1(=-2

m(2)
I(

(2)2 +m 2

[3.22]

ml(

where the coefficients c~2) (J.l) are given by


2

c~2)(J.l) = lai2)(K")~ i2) (J.l)

[3.23]

n=-2

Equation [3.23] may be evaluated by combining the components of a spatial tensor

~ ~2) (J.l) with the coefficients ai2)(K") read across row K"ofTable 3.1 to give

C&2)(J.l) = -~SinX ~ ~~) (J.l)+ cosX~ g2) (J.l) -

~ sinx~ ~2) (J.l)

d2) (J.l) = ~ ~ ~i) (J.l) + ~ ~ 1(2) (J.l)


C~;)(J.l) = -~~ ~i) (J.l) + ~ ~ f) (J.l)
c~2) (J.l) = ~ cos X ~ ~~) (J.l) + sin X ~ g2) (J.l) + ~ cos X ~ ~2) (J.l)

[3.24]

c~~(J.l) = -~~ ~~) (J.l) + ~~ ~2) (J.l)


Equation [3.24] may then be expanded using a specific defmition for ~ i2) (J.l) to yield
equations for the coefficients c~2) (J.l). For example, using the definition for ?f (CSA, j)
given in equation [2.17] yields the coefficients shown in Figure 3.1.
The equations for cf)(J.l) given in Figure 3.1 are an important result having
applicability to both chemical shielding anisotropy and dipole-dipole interactions. For
the case of chemical shielding anisotropy, c~2)(CSA,j) may be calculated from the
asymmetry l1j and the Euler angles aj' Pj and rj' relating the principal axis system of
the chemical shielding tensor to the principal axis system of the molecular rotational

28

C&2) (CSA,j)

=-

~ l~" {.J3sinx[sin 2Pj cos2a j + ~ ([1+ cos2Pj lcos2a j cos2rj

-2sin2aj sin2rj cosf3j)] - cos X[(3cos2 f3j -1) + 11j sin 2 f3j COS2rj]}

cfltCSA,j)= ~ I~" {.J3cosx[sin2 Pjcos2aj + ~ [I + cos2Pj lcos2ajCOS2rj


-2sin2aj sin2rj cosf3j )] + sinX[(3cos 2 f3j -1) + 11j sin 2 f3 j COS2rj]}

C~~)(CSA,j) = i~ 1~1r {sin 2 f3j sin2aj


+ ~ [cos2rAI +cos2Pj )sin2aj +2Sin2rjCOSPjCOS2a]}

Figure 3.1. The coefficients for the chemical shielding anisotropy may be used in
equation [3.22] to calculate reduced spectral density functions J :;,' (m) for second rank
relaxation mechanisms.

29
diffusion. For the case of dipole-dipole interactions, c~2)(DIP,i}) may be calculated by
setting

T1.
.,}

=0

'}

=At.. and f3. ={}..


'l'1}

I} ,

where the spherical polar angles {}I}.. and

At..
'l'1}

orient

the dipole-dipole axis relative to the principal axis system of the molecular rotational
diffusion.
The equations for c~2)(J.L) from Figure 3.1 may be used in conjunction with
equation [3.22] to calculate the reduced spectral density functions

.1 ':j!' (co).

For

example, one use of these equations is the calculation of the dipolar-CSA reduced
spectral density functions

.1 iff (co)

involving nucleus k and interacting dipolar spins i

and}. The coefficients c~2)(DIP,i}) and c~2)* (CSA,k) may be calculated using equation
[3.22]

[3.25]

Note, the products c~2)(DIP,i}) c~2)*(CSA,k) used in equation [3.25] are real numbers,
not complex numbers. The eigenvalues cof-) used in equation [3.25], and the angle X
used in Figure 3.1 are defined in tenns of the rotational diffusion coefficients Rxx ' Ryy
and Rzz , as shown in Table 3.1.
Once the reduced spectral density functions "

:'j!' (co) have been calculated they

may be used to calculate the relaxation matrix elements ~;;., via equations [2.1] and
[2.30] through [2.33]. Pursuing the example of the dipolar-CSA cross-correlation
reduced spectral density functions "

iff (co), the relaxation matrix elements R~C;:u, may

be calculated via equations [2.1] and [2.32]. The interaction constants ~flP and ~fsA
used in equation [2.32] are calculated from the chemical shielding anisotropy A k , the
internuclear distance 'ij and the gyromagnetic ratios ri' rj and

rb as indicated by

equations [2.20] and [2.28].


To summarize, the Favro diffusion model provides a method for calculating the

30
relaxation matrix coefficients R::::~,. For the case of dipolar-CSA cross-correlation
involving nucleus k and interacting dipolar spins i and j. the relaxation matrix
coefficients R~C;).),: are calculated in terms of angles ak. 13k. Yk. tPij and Oij.
gyromagnetic ratios Yi' Yj and Yk' chemical shielding anisotropy L1 k asymmetry 17k>
internuclear distance lij and rotational diffusion coefficients Rxx ' Ryy and Rzz

3.1 References
1.

S. A. Smith, "Simulations of Pulsed Nuclear Magnetic Resonance Experiments",


Doctoral Dissertation, University of California Santa Barbara, 1991.

2.

R. R. VoId and R. L. VoId, Prog. NMR Spectrosc.12, 79 (1978).

3.

D. M. Grant, C. L. Mayne, F. Liu, and T. X. Xiang, Chem. Rev. 91. 1591


(1991).

4.

J. McConnell, "The Theory of Nuclear Magnetic Relaxation in Liquids", p. 74,


Cambridge University Press, Cambridge, 1987.

5.

L. G. Werbelow and D. M. Grant, "Advances in Magnetic Resonance" O. S.


Waugh, Ed.), Vol. 9, p. 189, Academic Press. New York, 1977.

6.

D. Chandler, "Introduction to Modern Statistical Mechanics", p. 239, Oxford


University Press, Oxford, 1987.

7.

L. D. Favro, Phys. Rev. 119, 53 (1960).

8.

H. Shimizu,l. Chem. Phys. 37, 765 (1962).

9.

W. T. Huntress, "Advances in Magnetic Resonance" (1. S. Waugh, Ed.), Vol. 4,


p. 1, Academic Press, New York, 1970.

10. H. W. Spiess, "NMR Basic Principles and Progress" (P. Diehl, E. Fluck, and R.
Kosfeld, Eds.), Vol. 15, p. 128, Springer-Verlag, New York, 1978.
11. J. Mathews and R. L. Walker, "Mathematical Methods of Physics", 2nd Ed.,
p. 267, Addison-Wesley, New York, 1970.
12. R. N. Zare, "Angular Momentum", p. 105, John Wiley & Sons, New York, 1988.
13. R. N. Zare, "Angular Momentum", p. 101, John Wiley & Sons, New York, 1988.
14. E. O. Brigham, "The Fast Fourier Transform and its Applications", p. 27, PrenticeHall, Englewood Cliffs, 1988.
15. J. McConnell, "The Theory of Nuclear Magnetic Relaxation in Liquids", p. 45,
Cambridge University Press, Cambridge, 1987.

CHAPTER 4
RELAXATION EXPERIMENTS
A spin-lattice relaxation experiment perturbs the spin system away from thermal
equilibrium and subsequently uses

13e NMR spectroscopy to sample the time evolution

of the spin system as it returns to thermal equilibrium. The perturbation of the spin
system, accomplished via application of a sequence of radio frequency pulses, may be
expressed in terms of perturbation of the spectral lines or in terms of perturbation of the
magnetization modes. Typically, a series of several different relaxation experiments is
performed to study the time evolution of the spin system. Each experiment is designed
to perturb one or more magnetization modes in a particular manner and hence create
unique initial conditions for the relaxation. Historically, four different experiments have
been employed to study relaxation of the AX2 spin system (1). One type of perturbation
is provided by the carbon inversion recovery experiment that inverts all three carbon
lines (Ll' L2 and L3), thus perturbing magnetization mode MI. A second type of
perturbation is provided by the hard pulse experiment that inverts both proton lines (L4
and Ls) and perturbs M 2 . Another kind of perturbation is obtained with the i-pulse
experiment (2) that inverts the central carbon line (L2) and perturbs M3. Yet another
type of perturbation is provided by the INEPT experiment (3-5) that inverts one of the
two proton lines and perturbs Ms. The combination of these four experiments samples
the temporal evolution of the spin system by perturbing Mh M2, M3 and Ms, then by
measuring the carbon modes M 1, M3 and Ms constructed from spectral lines L 1, L2 and
L3, as indicated in equation [1.2].

Improved nonlinear least-squares fits of the experimental data result when a

32

larger number of magnetization modes can be perturbed and measured. A new


relaxation experiment perturbs M4 then measures its time evolution (6). The quantum
mechanical operator defining M4 is i~[i.~1 j~2 + j~1 i!!2] (7,8).

In this notation,

i!!1 i~2 and i~1 i!!2 specify proton-proton zero-quantum coherences, and
i~ =(i; - iff)/2 specifies antiphase orientation of the proton-proton zero-quantum
single transitions selected by j~ and i~ (9). Thus, the creation of antiphase, protonproton zero-quantum coherences perturbs M4 away from thermal equilibrium. The pulse
sequence shown in Figure 4.1 creates these zero-quantum coherences. It resembles the

sandwich sequence (10) for converting in-phase single-quantum coherences into


antiphase multiple-quantum coherences.
The pulse sequence begins with a delay time Dl of sufficient duration to ensure
that the spin system is at thermal equilibrium along the z axis. Then simultaneous
carbon and proton nl2 pulses ending at time a place both the carbon and proton
magnetization in the x,Y plane. This transverse magnetization is depicted with vectors
corresponding to the carbon spectral lines L 1, ~ and L3, and the proton spectral lines L4
and L5, in Figures 4.2 and 4.3 respectively. During the delay period between times a
and b, the carbon and proton magnetization vectors precess in the rotating frame (11)
depicted as the x,y plane. The vectors corresponding to high frequency spectral lines
precess more rapidly than those corresponding to low frequency lines. Specifically, Ll
precesses more rapidly than L2, L2 precesses more rapidly than L3, and L4 precesses
more rapidly than L5. These differential rates of precession cause the vectors to dephase
by 900 during the delay period of duration 't' = 1/4JCR , where JCR is the carbon-proton
scalar coupling constant. As shown in Figure 4.2b representing the magnetization at
time b, Ll and L3 have dephased by +900 and -900 respectively, relative

to~.

Also, L4

has dephased by +900 relative to L5, as shown in Figure 4.3b. Simultaneous carbon and
proton n pulses applied between times b and c then refocus the effects of chemical
shielding and magnetic field inhomogeneity, but not the effect of scalar coupling. The

33

-(q
2

n(q>+n)

-(q>+n)
2

-(ref)
2

Figure 4.1. The sandwich sequence experiment perturbs and measures M4. The carbon
and proton magnetization vectors dephase during the delay periods of duration
't

= 1/4JCH between

a and b, and between c and d. The angles q> and e represent the

phases at which the radio frequency pulses are applied.

34
result of these pulses is represented graphically in Figures 4.2c and 4.3c. The vectors
have been rotated 1800 about the x axis followed by an interchange (12) of L1 for L3,
and L4 for L5. During the second delay period of duration r = 1/4JcH between times c
and d, the vectors continue to dephase. At time d, vectors L1 and L3 assume an
antiphase orientation relative to L2, as shown in Figure 4.2d. Also, L4 assumes an
antiphase orientation relative to L5, as shown in Figure 4.3d. Simultaneous carbon and
proton nl2 pulses then convert all of these carbon and proton antiphase single-quantum
coherences into antiphase zero-quantum and double-quantum coherences.

The

appearance of antiphase proton-proton zero-quantum coherences perturbs M 4 to a


nonequilibrium value. At this point the spin system is permitted to relax for a relaxation
period D2. During this relaxation period the zero- and double-quantum coherences relax
towards their thermal equilibrium values of zero, and longitudinal magnetization
develops along the z axis due to spin-lattice relaxation. Following D2, carbon and
proton nl2 read pulses applied at a reference phase convert both the longitudinal carbon
magnetization and the multiple-quantum coherences into transverse magnetization. The
free induction decay is then acquired and Fourier transformed to create the carbon
spectrum.

Repeating this pulse sequence with different values of D2 permits

measurement of the time evolution of lines Ll, L2 and L3, beginning with the initial
values created by the perturbing pulse sequence and ending with their thermal
equilibrium values.
As previously mentioned, the quantum mechanical operator defining
magnetization mode M4 is

izC[iJ!li!f2 + i!f iJ!2].


1

Therefore M4 is composed of only

proton-proton zero-quantum coherences. The sandwich sequence also creates protonproton double-quantum coherences, as well as carbon-proton zero- and double-quantum
coherences.

These additional multiple-quantum coherences ordinarily would be

converted into carbon transverse magnetization by the read pulses. However, phase
cycling is used to eliminate all multiple-quantum coherences except the proton-proton

35

....------+---...

-..j~

L1

+ L2 + L3

Figure 4.2. Precession and dephasing of the carbon transverse magnetization during the
delay periods of duration

'f

= 1/4JCH are represented by clockwise rotation of vectors

L1, L2 and L3 in the x,y plane.

36

J - - - - - + - - - -...... L 4 + L5

Figure 4.3. Precession and dephasing of the proton transverse magnetization during the
delay periods of duration

L4 and L5 in the x,y plane.

'r

= 1/4JCH are represented by clockwise rotation of vectors

37
zero-quantum coherences.
The term phase cycling refers to variation of the angles cp and () (see Figure 4.1)
defining the axes about which radio frequency pulses rotate the carbon and proton
magnetization vectors. Commonly used phases are +x, +y, -x and -yo Figures 4.3 and
4.4 demonstrate how a simple example of phase cycling can be used to eliminate
magnetization. Both figures depict dephasing proton single-quantum coherences. The
difference between these figures is the phase of the proton n pulse applied at time b. In
Figure 4.3b a n(+x) pulse is applied, whereas in Figure 4.4b a n(+y) pulse is given.
Figures 4.3d and 4.4d illustrate the different effects of the n( +x) and n(+y) pulses. In
Figure 4.3d the directions of L4 and L5 are inverted relative to Figure 4.4d. The average
of these two different orientations of L4 and L5 equals zero. Hence this simple form of
phase cycling may be used eliminate the proton single-quantum coherences.
A phase cycle that eliminates the additional multiple-quantum coherences from
the sandwich sequence experiment alternates the cpand (}phases as cp = (+x,+y,-x,-y)
and () = (+x,+y,+x,+y) for each reference phase of the read pulses. The carbon-proton
zero-quantum coherences are removed by the identical cp and () phases during the first
two steps of the cycle, followed by a n phase difference during the final two steps of the
cycle. The proton-proton and carbon-proton double-quantum coherences are removed
by the nj2 increments to the cp and () phases between successive steps of the cycle.
Vector diagrams such as Figures 4.3 and 4.4 prove useful in analyzing phase cycles or
creating new pulse sequences. A more powerful approach uses the product operator
formalism (13,14) to predict the effects of the pulse sequence on the density matrix.
This formalism may be implemented with a computer algebra system (15) that increases
the speed and accuracy of the prediction relative to manual methods.
One feature distinguishing the sandwich sequence experiment depicted in Figure
4.1 from the other relaxation experiments mentioned above is the addition of a proton

nj2 read pulse to the carbon nj2 read pulse found in the other experiments. The

38

I - - - - - - + - - - -.....-l L4 + L5

Figure 4.4. Precession and dephasing of the proton transverse magnetization during the
delay periods 1" = 1/4JCH are represented by clockwise rotation of vectors L4 and L5 in
the x,y plane. The orientation of the proton magnetization at time d is determined by the
phase of the proton npulse applied at time b. Compare to Figure 4.3.

39

addition of the proton read pulse permits indirect observation of M4 in carbon NMR
spectra (16) through transformation of L1> L2 and L3. These spectral lines are not
transformed when only a carbon n/2 read pulse is applied in the absence of a proton

n/2 read pulse. Hence modes M I , M3 and Ms assume their normal definition in terms
of the linear combinations of these spectral lines indicated in equation [1.2]. Addition of
the proton n/2 read pulse to the carbon n/2 read pulse changes the information
contained in these spectral lines, and consequently accesses different magnetization
modes. These modes may be expressed in terms of the following linear combinations of
the transformed spectral lines L1, L'2 and L;
MI = Li +L'2 +L;

M4 = L1 -L'2 +L;

[4.1]

Ms =L1 -L; =0
Thus addition of the proton n/2 read pulse leaves MI unchanged, permits observation
of M4 in lieu of M3, and precludes observation of Ms.
The combination of the carbon inversion, hard pulse, I-pulse, INEPT and
sandwich sequence experiments has been used to study [2_13C]glycine with satisfactory
results. The concentration of glycine in glycerol was sufficiently high to produce an
adequate signal to noise ratio. In contrast, when these experiments were employed to
study the abu 23 _[2-13C]gly24-GCN4-pl peptide, poor signal to noise ratios were
obtained with all but the INEPT experiment. The reason for the improved performance
of the INEPT experiment is suggested by its acronym: Insensitive Nuclei Enhanced by
Polarization Transfer. The IH nucleus is about four times more sensitive to the external
magnetic field than the 13C nucleus. It is possible to exploit the relative sensitivity of the
IH nucleus in carbon NMR spectroscopy, even though the proton spectral lines L4 and
Ls are not observed. Transferring the proton magnetization from the two protons of the

40

13CH2 group to the 13C nucleus results in amplification of the carbon signal by up to a

factor of four. The INEPT experiment shown in Figure 4.5 makes use of this technique
during the perturbing pulse sequence.
The pulse sequence begins with a delay time Dl of sufficient duration to ensure
that the spin system is at thermal equilibrium along the z axis. Then the proton nI2(x)
pulse ending at time a places the proton magnetization along the y axis. This transverse
magnetization is depicted with vectors corresponding to the proton spectral lines L4 and
L5

in Figure 4.4a. During the delay period of duration

'r

= lj4JCH between times a and

b, L4 dephases by +90 0 relative to L5, as shown in Figure 4.4b. Simultaneous carbon

and proton n(y) pulses applied between times b and c then refocus the effects of
chemical shielding and magnetic field inhomogeneity, but not the effect of scalar
coupling, as shown in Figure 4.4c. During the second delay period of duration
'r

= lj4JCH between times c and d, the vectors continue to dephase, resulting in

antiphase orientation of vectors L4 and Ls along the x axis at time d, as shown in Figure
4.4d. At this point a proton nl2 (y) pulse places the antiphase proton magnetization
along the z axis, creating a J -ordered spin state (17) that transfers the proton
magnetization to the 13C nucleus. The carbon vectors Ll and L3 are in antiphase along
the z axis, and their magnitudes are increased by approximately four. The carbon vector
~

is also oriented along the z axis, parallel to L3, but does not receive any of the proton

magnetization transfer. The antiphase orientation of Ll and L3 perturbs M5 away from


its thermal equilibrium value, as can be appreciated from equation [1.2]. At this point
the spin system is permitted to relax for a relaxation period D2. Then a carbon nl2 read
pulse converts the longitudinal carbon magnetization into transverse magnetization, and
the free induction decay is acquired.
The polarization transfer from the protons to the carbon is accomplished by
means of the nl2 -

'r -

n-

'r -

nl2 proton pulse sequence, together with the carbon n

pulse applied in concert with the proton n pulse. This combination of pulses and delay

41

n
-(ref)
2

n{y)

n(-y)

;(x)

'H

;(y)

rl~__ ~r-l~__ ~rl~__


T__

Figure 4.5.

T__

D_2________________

bed

The INEPT experiment perturbs and measures M 5.

magnetization vectors dephase during the delay periods of duration

The proton

= 1J41CH between

a and b, and between c and d. This experiment transfers magnetization from the protons
to the carbon, resulting in amplification of the carbon signal.

42
periods is the INEPT pulse sequence (as opposed to the INEPT relaxation experiment
shown in Figure 4.5). It is possible to embed this pulse sequence in a variety of novel
relaxation experiments, thereby effecting polarization transfer and amplification of the
carbon signal in these experiments. For example, the experiment depicted in Figure 4.6
incorporates the INEPT pulse sequence after the relaxation period D2. This experiment
perturbs and measures M2. The pulse sequence begins with a delay time Dl of sufficient
duration to ensure that the spin system is at thermal equilibrium along the z axis. Then a
proton n(x) pulse inverts the proton magnetization, after which the spin system is
permitted to relax for a relaxation period D2. Following the relaxation period, the
INEPT pulse sequence places the relaxing proton magnetization in the x,y plane, permits
it to dephase to antiphase orientation, then places it back along the z axis and transfers it
to the

l3e nucleus.

At time d, the final proton n/2( -y) pulse of the INEPT pulse

sequence is accompanied by a carbon n/2 read pulse. Thus, as the antiphase proton
magnetization is transferred to the l3e nucleus, the antiphase carbon magnetization along
the z axis is placed in the x,y plane, and the free induction decay is acquired.
Examination of the density matrix shows that the INEPT pulse sequence,
together with the carbon n/2 read pulse transforms the spectral lines L 1, L2 and L3.
Thus the following magnetization modes may be accessed from the transformed spectral
lines L\, L~ and L~

Ml = -(L\ +L~ +L~)


M3 = -(L\ -L~ +L~)

[4.2]

M2 =-(L\ -L~)
The negative signs of Mh M2 and M3 may be removed through normalization to the
thermal equilibrium value of M h which in this experiment carries a negative sign. So
the combination of the INEPT pulse sequence and the carbon n /2 read pulse negates M 1

43

n(x)
'H

~(-x)

n(y)

-(ref)
2

n(y)

~(-y)

n __
~ n"""-----'~n"""-----'~n"'---_
a

Figure 4.6.

The proton inversion INEPT experiment perturbs M2 and permits

measurement this proton mode in the carbon NMR spectrum. The proton magnetization
relaxes due to spin-lattice relaxation during the relaxation period D2. The ensuing
INEPT pulse sequence transfers the proton magnetization to the 13e nucleus.

44

and M3, and permits observation of -M2 in lieu of Ms. This relaxation experiment may
be used in place of the hard pulse experiment that perturbs but does not measure M2.
One advantage of this experiment is that it permits indirect measurement of M2 in the
carbon spectrum. Another advantage is that the carbon lines are amplified by
polarization transfer. Since this experiment incorporates both proton inversion and
INEPT pulse sequences, it is called the proton inversion INEPT experiment.
The polarization transfer experiment shown in Figure 4.7 perturbs and measures
M3, and may be used in place of the J-pulse experiment. It incorporates the refocused
INEPT pulse sequence (17,18). The pulse sequence begins with a delay time D1 of
sufficient duration to ensure that the spin system is at thermal equilibrium along the z
axis.

Between times a and e the INEPT pulse sequence transfers the proton

magnetization to the

13C

nucleus. The carbon 7r/2{y) pulse accompanying the final

proton 7r/2(y) pulse of the INEPT pulse sequence places the carbon magnetization
along the x axis at time e. The carbon magnetization is amplified by polarization
transfer, with L1 and L3 in antiphase orientation, as shown in Figure 4.8e. During the
delay period of duration 8 = 1j8JCH between times e and f, L1 rephases by 90 relative
to L3, as shown in Figure 4.8f. Simultaneous carbon and proton 7r(y) pulses applied
between times f and g then refocus the effects of chemical shielding and magnetic field
inhomogeneity, but not the effect of scalar coupling, as shown in Figure 4.8g. During
the second delay period 8 = 1j8JCH between times g and h, L1 rephases by another 90
relative to L3. At time h, the vectors L1 and L3 are in phase along the -y axis, and ~ is
90 out of phase along the x axis, as shown in Figure 4.8h. At this point a carbon

n/2(x) pulse places L1 and L3 along the z axis, leaving L2 in the x,y plane. The inphase orientation of L1 and L3 perturbs M3 away from its thermal equilibrium value, as
can be seen from equation [1.2]. At this point the spin system is permitted to relax for a
relaxation period D2. Then a carbon n/2 read pulse converts the longitudinal carbon
magnetization into transverse magnetization, and the free induction decay is acquired.

1r(-Y)

abc

n~ n ~

1r

"2 (x)

l'

1r(Y)

tJ.

1r2 (+x)
-x
D2

2"1r (ref)

r1JlL....--_ _

1r

2"(Y)

~(y) 1r(:i)

l'

= 1/4JCH between a and b, and between c and d. The carbon

andh.

magnetization vectors rephase during the delay periods of duration tJ. = 1/8JCH between e andj, and between g

dephase during the delay periods of duration

Figure 4.7. The refocused INEPT experiment perturbs and measures M3. The proton magnetization vectors

lH

13e

1r(=i)

+:>.
VI

46

+ L3 ....1 - - - -.........- - - - - - 1

Figure 4.8. Precession and rephasing of the carbon transverse magnetization during the
delay periods of duration

= 1/8JCH

L1, L2 and L3 in the x,y plane.

are represented by clockwise rotation of vectors

47

At the beginning of the relaxation period D2, vector L2 remains in the x,y plane
but may be eliminated through phase cycling. The carbon1'(+Y,+x) and 1'/2(+x,-x)
pulses applied at times/and h eliminate L2. The phase cycling of the carbon 1'(-y,-x)
pulse applied at time b is provided merely to compensate for pulse imperfections in the

carbon1'(+y,+x) pulse applied at timet, and is not required to remove~.


The polarization transfer experiment shown in Figure 4.9 perturbs and measures
M4 , and may be used in place of the sandwich sequence experiment of Figure 4.1. This
mUltiple quantum INEPT experiment is identical to the refocused INEPT experiment up

to time h, including the phase cycling to eliminate L2. At time h, simultaneous carbon
and proton 1'/2 (+ x, -x) pulses convert all of the carbon single-quantum coherences into
antiphase proton-proton zero- and double-quantum coherences. No phase cycling is
directed at eliminating the double-quantum coherences, so both the zero- and doublequantum coherences relax during the relaxation period D2. The proton-proton zeroquantum coherences do not precess about the z axis during D2. In contrast, the protonproton double-quantum coherences precess and dephase during D2. Since D2 has a
variable duration, the double-quantum coherences dephase to a variable extent,
destroying the phase coherence of the double-quantum coherences. This variable
dephasing may be avoided by inserting a proton

1t' pulse

at time j, the midpoint of the

relaxation period D2. This pulse refocuses (18) the double-quantum coherences at time
I, as demonstrated in Figure 4.10.

At time i the double-quantum coherences are in antiphase, as represented by


magnetization vectors Ql and Q2 in Figure 4.10i. During the relaxation period of
duration D2/2 between times i andj, these vectors dephase to an extent dependent upon
the length of D2 /2, as shown in Figure 4.10j. A proton 1'(-Y) pulse applied between
times j and k then refocuses the effects of chemical shielding, magnetic field
inhomogeneity and scalar coupling. The result of this pulse is represented graphically in
Figure 4.10k, where the vectors Ql and Q2 have been rotated 1800 about the y axis.

1r (x)

'f

1r(-Y)
'f

'f

1r

IJ.

"2(Y)

~(y)

i
g

:~)

1r(Y)

1r(

IJ.

IJ.

h
i

D2/2

(+x)
2 -x

1r

1r(+x)
2 -x

1r(-Y)

D2

D2/2

1r

"2 (x)

7r

"2 (ref)

'f

= 1/4JCH between a and b, and between c and d. The

i andj, and then rephase during the relaxation period between k and I.

between g and h. The proton-proton double-quantum coherences dephase during the relaxation period between

carbon magnetization vectors rephase during the delay periods of duration IJ. = 1/8JCH between e and i, and

vectors dephase during the delay periods of duration

Figure 4.9. The multiple-quantum INEPT experiment perturbs and measures M4. The proton magnetization

IH

l3

1r(=~)

00

49

Figure 4.10. Precession, dephasing and rephasing of the proton-proton doublequantum coherences during the relaxation periods of duration D2/2 are represented by
clockwise rotation of vectors Q 1, and Q2 in an X,Y coordinate system. This coordinate
system is not the x,y plane of the transverse magnetization.

50
Note, the proton 1l'( -y) pulse, unaccompanied by a carbon 1l' pulse, will only rotate the
vectors 1800 about the y axis but not interchange them (12). During the relaxation period
of duration D2/2 between times k and I, these vectors rephase to the same extent that
they dephased between times i andj. At time I the vectors Ql and Q2 are again in
antiphase, as shown in Figure 4.1Ol. In macromolecules, the refocusing effect of the
proton 1l'pulse compensates for the different relaxation rates of Ql and Q2 (19).
Since D2 is the relaxation period, longitudinal magnetization develops along the z
axis, and the proton-proton zero- and double-quantum coherences relax towards their
thermal equilibrium values of zero. At time I a carbon 1l'/2 pulse applied at a reference
phase, and accompanied by a proton 1l'/2(x) pulse, converts both the longitudinal
carbon magnetization and the multiple-quantum coherences into transverse
magnetization, and the free induction decay is acquired. Since the carbon 1l'/2 read
pulse is accompanied by a proton 1l'/2 (x) pulse, the spectra1lines are transformed as in
equation [4.1]. Thus the proton-proton zero-quantum coherences that compose M4 may
be measured. Note, it is possible to eliminate the proton-proton double-quantum
coherences via phase cycling, and thus measure only M4. But including the doublequantum coherences increases the signal to noise ratio.
The polarization transfer experiment shown in Figure 4.11 perturbs and
measures M I> and may be used in place of the carbon inversion recovery experiment.
This experiment is identical to the multiple-quantum INEPT experiment up to the
relaxation period D2, including the phase cycling to eliminate L2. The state of the
density matrix at the beginning of D2 reveals that proton-proton zero- and doublequantum coherences are created, as in the multiple-quantum INEPT experiment.
Moreover, lines LJ, L2 and L3 are inverted and amplified via polarization transfer,
perturbing M 1 away from its thermal equilibrium value. Following the relaxation period

D2, a carbon 1l'/2 read pulse unaccompanied by a proton 1l'/2 read pulse results in
untransformed spectral lines. Hence MI> M3 and Ms may be measured instead of Ml

1r(-y)

"21r(y ) 1r(Y)

"21r(+X)
-x

2 -x

1r(+X)
D2

1r (ref)

abc

de

n n 1hf1Jl_ __

"21r (x)

l'

~(y) 1r(!i)

between g and h.

vectors dephase during the delay periods of duration

l'

= 1/4JCH between

a and b, and between c and d. The


carbon magnetization vectors rephase during the delay periods of duration ~ = 1/8JCH between e and j, and

Figure 4.11. The carbon inversion INEPT experiment perturbs and measures M 1. The proton magnetization

'H

13

1r(=i)

......

VI

52

and M4. Since this experiment combines features of the carbon inversion recovery and
INEPT experiments, it is called the carbon inversion INEPT experiment. Unlike the
carbon inversion recovery experiment, this experiment may not be used to estimate the
spin-lattice relaxation time T}. because in this experiment Ml does not return to thermal
equilibrium via an approximately monoexponential decay.
Taken as a group, the polarization transfer experiments presented above perturb
and permit measurement of magnetization modes Mh M2, M3, M4 and Ms. Table 4.1
indicates the perturbed state of L 1, L2 and L3 for each experiment, and includes for
comparison the perturbed state of these spectral lines in the nonpolarization transfer
experiments. The thermal equilibrium values for these lines are Ll
L3

= 1f2.

=1f2,

L2

=1 and

Since the proton gyromagnetic ratio YH is approximately four times as large

as the carbon gyromagnetic ratio Ye, the perturbed values of linesL I and L3 may be
amplified by up to a factor of four. In actual experiments a factor of three to four is
obtained, because some loss of magnetization results from transverse relaxation
occurring during the delay periods

'f

and A.

4.1 References
1.

D. M. Grant, C. L. Mayne, F. Liu, and T. X. Xiang, Chem. Rev. 91, 1591

(1991).
2.

F. Liu, C. L. Mayne, and D. M. Grant, J. Magn. Reson. 84, 344 (1989).

3.

G. A. Morris and R. Freeman, J. Am. Chem. Soc. 101, 760 (1979).

4.

D. P. Burrum and R. R. Ernst, J. Magn. Reson. 39, 163 (1980).

5.

G. A. Morris and R. Freeman, J. Am. Chem. Soc. 102, 428 (1980).

6.

R. A. Brown, R. H. Price and D. M. Grant, J. Magn. Reson. (in press).

7.

Z. Zheng, C. L. Mayne, and D. M. Grant, J. Magn. Reson. 103,268 (1993).

8.

L. G. Werbelow and D. M. Grant, J. Chem. Phys. 63,544 (1975).

9.

R. R. Ernst, G. Bodenhausen, and A. Wokaun, "Principles of Nuclear Magnetic


Resonance in One and Two Dimensions", pp. 33 and 271, Oxford University
Press, Oxford, 1988.

53

Table 4.1. Perturbed State of the Spectral Lines

Polarization Transfer Experiments

Nonpolarization Transfer Experiments

Mode

Ll

L2

L3

Mode

Ml

-YH
-

-YH
-

-YH
-

MI

M2

YH
- -4

M3

--+-

M4

-YH
-

Ms

YH
- -4

2Ye

4Ye

1
2

2Ye

YH
2Ye

1
4

2Ye

2Ye

4Ye

1
4

YH
---

L2
1
2

1
4
1
4

L3

1
2

1
4
1
4

M2

YH
-- -4

M3

1
4

1
2

-YH
-

M4

1
4

1
2

2Ye

2Ye

2Ye

1
2

Ll

1
4

1
4

YH
-- -4

2Ye

Note: For perturbation with polarization transfer, Mb M2,

M3, M4 and Ms are studied

by the carbon inversion INEPT, proton inversion INEPT, refocused INEPT, multiplequantum INEPT and INEPT experiments respectively. For perturbation without
polarization transfer, MI, M2,

M3, and M4 are studied by the carbon inversion recovery,

hard pulse, J-pulse and sandwich sequence experiments respectively.

54
10. R. R. Ernst, G. Bodenhausen, and A. Wokaun, "Principles of Nuclear Magnetic
Resonance in One and Two Dimensions", pp. 403 and 450, Oxford University
Press, Oxford, 1988.
11. T. C. Farrar and J. E. Harriman, "Density Matrix Theory and its Applications in
NMR Spectroscopy", p. 56, Farragut Press, Madison, Wisconsin, 1991.
12. R. K. Harris, "Nuclear Magnetic Resonance Spectroscopy", p. 165, John Wiley &
Sons, Inc., New York, 1986.
13. O.W. Sj1Irensen, G. W. Eich, M. H. Levitt, G. Bodenhausen and R. R. Ernst,
Prog. NMR Spectrosc. 16, 163 (1983).
14. T. C. Farrar and J. E. Harriman, "Density Matrix Theory and its Applications in
NMR Spectroscopy", p. 101, Farragut Press, Madison, Wisconsin, 1991.
15. B. W. Char, K. O. Geddes, G. H. Gonnet, B. L. Leong, M. B. Monagou and S.
M. Watt, "Maple V Reference Manual", Springer-Verlag, New York, 1991.
16. M. M. Fuson and J. H. Prestegard, J. Magn. Reson. 41, 179 (1980).
17. T. C. Farrar and J. E. Harriman, "Density Matrix Theory and its Applications in
NMR Spectroscopy", p. 138, Farragut Press, Madison, Wisconsin, 1991.
18. R. K. Harris, "Nuclear Magnetic Resonance Spectroscopy", p. 177, John Wiley &
Sons, Inc., New York, 1986.
19. L. E. Kay, T. E. Bull, L. K. Nicholson, C. Griesinger, H. Schwalbe, A. Bax and
D. A. Torchia, J. Magn. Reson. 100, 538 (1992).

CHAPTER 5
PULSE SEQUENCE SIMULATION
Computer algebra systems provide a powerful means of simulating pulse
sequences and analyzing the effects of phase cycling. Algebraic description of the time
evolution of the density matrix is particularly helpful when the pulse sequence creates
multiple-quantum coherences. Computer algebra systems are capable of simulating the
rotation of magnetization vectors in response to radio frequency pulses. It is also
possible to simulate the precession and dephasing of magnetization vectors in the
rotating frame (1) during delay periods. However, relaxation effects are beyond the
scope of these systems because relaxation must be simulated using numeric techniques.
Two commonly used computer algebra systems are Maple (2) and Mathematica (3).
This chapter discusses techniques for simulating pulse sequences using Maple.
The linear algebra commands of Maple provide a rich set of tools for simulating
the time evolution of the density matrix under the application of radio frequency pulses
and delay periods. However, Maple lacks the direct product operator required to apply
the product operator formalism (4). The procedure direct shown in Figure 5.1 provides
the direct product capability. Using this procedure it is possible to construct operator
matrices from the limited set of basis matrices and rotation matrices shown in Figure 5.2
(5). An operator matrix may be used to express the response of the density matrix to a
radio frequency pulse.

For example, a carbon 1C/2(+x) pulse is represented

mathematically as a unitary transformation of the density matrix 0'


[5.1]

56

direct := proc(A,B)
local 10calA,localB,i,j,k,l,m,n,0,p,C,t;
if 2 < nargs then # Recursive call for more than 2 args.
RETURN (' direct' (' direct' (A,B ),args[3 .. nargs]))
fi:,
ifnargs <> 2 then ERRORCwrong number of arguments') fl;
if type(A,'matrix') then localA := A
else localA := eva1m(A) # Recursive evaluation of arg A.
fl,
if type(B,'matrix') then localB := B
else localB := eva1m(B) # Recursive evaluation of arg B.
fl,
if type(1ocaIA,'matrix') and type(localB,'matrix') then
m := rowdim(localA);
n := coldim(1ocalA);
0:= rowdim(localB);
p := coldim(1ocalB);
C := array(1 .. m*o,l .. n*p);
for i to m do for j to n do for k to 0 do for I to p do
t := localA[i,j]*loca1B[k,I];
C[k+o*(i-l),l+p*(j-l)] := normal(t)
odododod
else ERRORCexpecting matrices as arguments')
fl,
subs('1ocalA' = 10calA,'localB' = 10calB,op(C));
if has(",'localA') or has(",'localB') then
ERRORCundefined elements in matrix')
fl,
RETURN (")
end:

Figure 5.1. The procedure direct provides a direct product operator.

57

# The following is an Iz matrix.

Iz := matrix([[l/2,0],[0,-l/2]]):
# The following is an E matrix.

Ie := matrix([[l,O],[O,l]]):
# The following is an 1+ matrix.

Ip := matrix([[O,l],[O,O]]):
# The following is an 1- matrix.

1m := matrix([[O,O],[l,O]]):
# The following is an alpha-polarization matrix.

la := matrix([[l,O],[O,O]]):
# The following is a beta-polarization matrix.

Ib := matrix([[O,O],[O, 1]]):

# The following procedure creates an x rotation matrix. In contrast


# to Farrar's matrices, Theta is negated.
Rx := proc(Theta)
local t;
if nargs <> 1 then ERROR(,wrong number of arguments') fi;
t := -Theta/2;
matrix([[cos(t), -I*sin( t)],[ -I*sin( t),cos(t)]])
end:
# The following procedure creates a y rotation matrix. In contrast
# to Farrar's matrices, Theta is negated.

Ry := proc(Theta)
local t;
if nargs <> 1 then ERRORCwrong number of arguments') fi;
t := -Theta/2;
matrix([[cos( t),-sin( t)],[sin(t),cos( t)]])
end:

Figure 5.2. Basis and rotation matrices defined in Maple.

58
In equation [5.1], the operator matrix effecting the rotation is constructed via the
direct product of the rotation matrix
identity matrices

R+x=90

in the first (leftmost) position and the

E in the second and third positions.

The rotation matrix

R+x=90

in the

first position specifies that the pulse rotates the carbon magnetization 900 about the +x
axis. The identity matrices

E in the second and third positions specify that the pulse has

no effect on either of the protons. This convention is identical to the convention used in
Figure 2.1 to express eigenstates in the spin-product basis. Equation [5.1] is also
expressed in the spin-product basis. This equation may be represented in the Maple
language as a unitary transformation of the density matrix sigma
sim(direct(Rx( 1C/2 ),Ie,Ie ),sigma,direct(Rx( -1C/2 ),Ie,Ie

[5.2]

Equation [5.2] references the procedure sim listed in Figure 5.3. This procedure
accomplishes the unitary transfonnation and simplifies the result.
A typical pulse sequence includes pulses and delay periods. A pulse may be
represented in Maple programs as in equation [5.2]. A delay period may be represented
mathematically as temporal evolution of the density matrix (6) under the influence of the
static Hamiltonian ~

e-i'3e otve+i'3e ot
IT

[5.3]

Representation of this equation in the Maple language requires a description of the time
.
od uct baSlS.

F or t he 13CH 2 mOIety,
.
. operator e -i '3e ot .m t he spm-pr
evoI uUon
t he
Hamiltonian ~ is defined mathematically as

[5.4]

under the simplification of the X approximation (7). In equation [5.4]

represents the

59

sim := proc(A,B,C)
10callocalA,localB ,local C,i,j ,m,n,D;
if nargs <> 3 then ERROR(,wrong number of arguments') fi;
if type(A,'matrix') then 10calA := A
else localA := evalm(A)
fi:,
if type(B,'matrix') then localB := B
else localB := evalm(B)
fl;
if type(C,'matrix') then localC := C
else localC := evalm(C)
fi,
if type(localA,'matrix') and type(localB,'matrix')
and type(locaIC,'matrix') then
D := multiply(localA,localB,localC);
m := rowdim(D);
n := coldim(D);
for i to m do for j to n do
D[i,j] := expand(simplify(D[i,j])
odod
else ERROR('expecting matrices as arguments')
fl,
subsClocalA' = 10calA,'localB' = 10calB,localC = 'localC',op(D;
if has(tt,'localA') or has(tt,'localB') or has(tt,'localC') then
ERROR(,undefmed elements in matrix')
fl,
RETURN(tt)
end:

Figure 5.3. The procedure sim concatenates three matrices and simplifies the result.

60
carbon resonance frequency and COf.I represents the proton resonance frequency.

JCH

and

are the carbon-proton and proton-proton scalar coupling constants respectively.

fz'

JHH
A

1+ and L are spin angular momentum operators.

It is possible to represent the Hamiltonian 'tieo in the Maple language by rewriting


equation [5.4] using the basis matrices given in Figure 5.2 for i z ,

i+

and

ham :=-Wc*direct(lz,le,le)
-Wh*(direct(Ie,Iz,Ie) + direct(Ie,Ie,Iz
+Jch*(direct(Iz,Iz,Ie) + direct(Iz,Ie,Iz
+Jhh*(direct(Ie,Iz,Iz) + (112)*(direct(le,Ip,In) + direct(Ie,In,Ip);

[5.5]

The resulting Hamiltonian operator matrix ham, written in the spin-product basis, is not
..
.
fior f ormmg
.

d lagon
a I. H ence It
IS.mappropnate
e -i 'tie 0 t because M ap Ie cannot
compute the exponential of a nondiagonal matrix. However, it may be diagonalized to

yield the Hamiltonian 'tie in the eigen basis. The eigenvalues of this matrix are used to
-i'tie'0 t m
.
.
.
.
.
-i 'tie'0 t IS
.
form e
the elgen basIs. The orne evoluoon operator e
then
transformed from the eigen basis to the spin-product basis as i'tie 0 t = Qei 'tie 0t Q-I
and substituted into equation [5.3]
Qe -i 'tie 0t Q-10 Qe+i 'tie 0t Q-I

[5.6]

The transformation matrix Q is formed from the eigenvectors of the matrix ham. The
eigenvectors and eigenvalues of ham may be obtained using standard Maple commands

eigsys := eigenvects(ham);

This approach yields the following eigenvalues

[5.7]

61
~

1
1
1
= --mc -mH +-JCH +-JHH
2
2
4

= --mc --JHH

1
1
= --mc + -JHH
2
4

1
2

3
4

1
1
1
.1.4 = --mc +mH --JCH +-JHH
224

111
As =-mc
-~ --JCH +-JHH
2
2
4
.1.6

1
2

1
4

=-mc +-JHH
[5.8]

=-mc --JHH
111

As =-mc +mH +-JCH +-JHH


2

The time evolution operator e-i ~ot is constructed from the eigenvalues Aj. It is the
matrix M having diagonal elements M jj = e-iAl and off-diagonal elements Mjk = O.
In the spin-product basis, the normalized eigenvectors of the matrix ham are
11) = [1

12) = [0 .../1/2
13) = [0

..J1/2

.../1/2

0
0

-.../1/2

14)=[0 0

0 1 0

15) = [0 0

16) = [0

17) = [0

0]

0
0

0]

.../1/2

0]

0]

1 0

0]

-.../1/2

.../1/2 .../1/2
o 0 1]

18) = [0 0 0 0 0

0]

[5.9]

0]

These eigenvectors form the columns of the transformation matrix Q. The matrices Q
and M = e-i ~ot are used in equation [5.6] to describe the time evolution of the density

62
matrix under the influence of the static Hamiltonian. The Maple expression is
sim(sim(Q,M,inverse(Q) ),sigma,sim(Q,-M,inverse(Q)

[5.10]

Another use for computer algebra systems is the evaluation of equations such as
[2.31] through [2.34]. The spin tensor operators T(,u,s) in these equations may be
constructed using the product operator formalism. For example, the tensor operator

fJi)(DIP,ij) from equation [2.27] is defined mathematically as

[5.11]
for the case where dipolar spin i is the carbon and dipolar spin j is the second proton.
However, prior to use in equations [2.32] or [2.33], this operator matrix must be
transformed from the spin-product basis to the eigen basis using the matrix Q. The
Maple language expression for the transformed operator matrix is
sim(inverse(Q),-(direct(lp,Ie,Iz) + direct(lz,Ie,Ip)),Q)

[5.12]

5.1 References
1.

T. C. Farrar and J. E. Harriman, "Density Matrix Theory and its Applications in


NMR Spectroscopy", p. 56, Farragut Press, Madison, Wisconsin, 1991.

2.

B. W. Char, K. O. Geddes, G. H. Gonnet, B. L. Leong, M. B. Monagou and S.


M. Watt, "Maple V Reference Manual", Springer-Verlag, New York, 1991.

3.

S. Wolfram, "Mathematica - A System for Doing Mathematics by Computer",


Addison-Wesley, New York, 1988.

4.

O.W. SlZSrensen, G. W. Eich, M. H. Levitt, G. Bodenhausen and R. R. Ernst,


Prog. NMR Spectrosc. 16, 163 (1983).

5.

T. C. Farrar and J. E. Harriman, "Density Matrix Theory and its Applications in


NMR Spectroscopy", p. 101, Farragut Press, Madison, Wisconsin, 1991.

6.

T. C. Farrar and J. E. Harriman, "Density Matrix Theory and its Applications in


NMR Spectroscopy", p. 110, Farragut Press, Madison, Wisconsin, 1991.

7.

R. K. Harris, "Nuclear Magnetic Resonance Spectroscopy", p. 54, John Wiley &


Sons, Inc., New York, 1986.

CHAPTER 6
SPECTRAL SIMULATION
Each of the relaxation experiments described in Chapter 4 creates unique initial
conditions for the relaxation and thereby promotes a particular time evolution of the spin
system back to thermal equilibrium. The data obtained by sampling the time evolution of
the carbon magnetization are used to solve equation [1.1]. The solution is accomplished
via nonlinear least-squares minimiiation (1) that obtains optimum values for fitting
parameters by matching a theoretical carbon spectrum to the measured spectrum. The
theoretical spectrum is generated by simulating the time evolution of the density matrix,
beginning with the pulse sequence that creates the initial conditions for spin-lattice
relaxation, and ending with the free induction decay. A particular pulse sequence may
be simulated as a series of pulses, with each pulse followed by a relaxation step. For
example, the INEPT relaxation experiment depicted in Figure 4.5 is simulated as
follows. Beginning with a thermal equilibrium density matrix, the initial proton 1C/2(x)
pulse is modeled by applying a unitary transform (2) to the density matrix. Following
this pulse, relaxation of the density matrix is simulated via equation [1.1] for a delay
period

'r

= 1/4JCH '

Next the simultaneous carbon and proton 1C(y) pulses are modeled

by a unitary transform, following which relaxation of the density matrix is simulated for
another delay period

'r.

Then another unitary transform models the proton 1C/2(y)

pulse. Next the relaxation is simulated for the relaxation period D2. After D2 the carbon

1C/2 read pulse is modeled by a unitary transform. Then the free induction decay is
simulated by relaxation of the density matrix. Simulation of the INEPT relaxation
experiment requires four unitary transforms, each followed by a relaxation step.

64
Relaxation occurs during the delay periods
during the free induction decay. For the

't',

during the relaxation period D2 and

13CH2 moiety,

relaxation is simulated through

application of equation [1.1] to each element of the eight by eight density matrix
describing the AX2 spin system. A 10 kilohertz secular approximation reduces the
required 64 by 64 relaxation supermatrix to seven block-diagonal matrices treating the
upper triangle of the Hermitian density matrix. These seven block-diagonal relaxation
matrices are: a 12 by 12 spin-lattice relaxation matrix treating the eigenstate populations
and proton-proton zero-quantum coherences, an eight by eight proton transverse
relaxation matrix, a six by six carbon transverse relaxation matrix, a four by four
carbon-proton zero-quantum relaxation matrix, a four by four carbon-proton doublequantum relaxation matrix, a two by two proton-proton double-quantum relaxation
matrix, and two one by one triple-quantum relaxation matrices~ With the exception of
the triple-quantum relaxation matrices, all of these matrices contain time-dependent
exponential terms, or slowly oscillating nonsecular terms, and therefore traditional
methods for solving the secular Redfield differential equation as a symmetric eigenvalue
problem (3) do not apply. However, a simple mathematical substitution (4) may be used
to eliminate the oscillating terms from any nonsecular Redfield differential equation
expressed in the eigenstate representation, and thereby render it amenable to solution as a
non symmetric eigenvalue problem.
The form of this substitution is illustrated using the six by six relaxation matrix
treating the six density matrix elements 0'15' 0'26'

(127' 0'36' (137

and

(148

representing

the carbon transverse magnetization. Rewriting equation [1.1] for these elements gives

do =Ro
dt

[6.1]

In this equation R is the relaxation matrix shown in Figure 6.1. The vector 0 is defined as
*With the exception of the 12 by 12 relaxation matrix, there are two instances of each
block-diagonal relaxation matrix, for a total of 13 block-diagonal relaxation matrices.

4815

e-i2JCHt

e-i(JCH +JHH)t

R 4827e- i./,CH t

e-i(JCH + JHH)t

4826

R 3727e- iJHH t

R3627

R2727

-iJCHt

R 4836 e

-iJHHt

R3736e

R3636

R 2736

R 2636eiJHH t

R 2627eiJHH t

e-i2JHHt
3726

R 3626e

R 2726e

R 1536eiJCHt

R 1527eiJCHt

i(JCH+JHH)t

4837

iJHHt

e-i(JCH -JHH)t

R3737

R3637e

iJHHt

2737e

R2637i 2JHHt

1537

3748

iJCHt
2748 e

+ J HH)t

R 4848

i(JCH - JHH)t

R3648iJCHt

R 2648i(JCH

R1548i 2JCHt

coupling constant JHH.

dependent exponential tenns are defined in tenns of the carbon-proton scalar coupling constant JCH and the proton-proton scalar

Figure 6.1. Elements of the relaxation matrix R. The coefficients R~ are defined by equation [2.1]. The nonsecular, time-

3715

R 2715e

-iJHHt

R 2626

i(JCH - JHH)t

R 3615e-iJCHt

1526

-iJHHt

e-i(JCH -JHH)t

-iJCHt

2615

R 1515

0'1
VI

66
0"15
0"26

0'=

0"27

[6.2]

0"36
0"37
0"48

The required substitution must eliminate the time-dependent exponential terms from all
elements of the relaxation matrix; however, a suitable substitution may be determined by
eliminating these exponential terms from any particular row of the matrix. Eliminating
the time dependence from row three of matrix R may be accomplished by substituting
,-r

Il.

VIS - ,..,15

eUCHt

UHHt

0"26 - P26 e
0"27

= P27

0"36

= P36

0"48

P4S e

[6.3]

-iJ,

CH

This substitution is an interaction representation (5) in the scalar coupling interaction.


The time derivatives for this substitution are

[6.4]

67

Equations [6.3] and [6.4] are substituted into equation [6.1], yielding an equation from
which the oscillating tenns are readily eliminated. Moving tenns such as ilCH and ilHH
to the right-hand side of this equation gives

dp =sp
dt

[6.5]

In this equation S is the relaxation matrix shown in Figure 6.2. The vector p is

PIs
P26
P27

p=

[6.6]

P36

1137
P48

The matrix S may be diagonalized as a non symmetric eigenvalue problem (6). The
temporal evolution (3) of the solution is expressed in tenns of the complex eigenvalues

Aj and eigenvectors Z j of the relaxation matrix S


6

p(t) = LCjZjeAjt

[6.7]

j=l

The coefficients Cj in equation [6.7] are detennined from the state of the density matrix
just prior to the relaxation step. Letting t

=0 at this time gives


6

p(o) = LCjZj

[6.8]

j=l

Equation [6.8] is simply a matrix multiplication that can be inverted to give

C=Z-lp(O)

[6.9]

68

R4827

R4836

Figure 6.2. Elements of the relaxation matrix S. The imaginary terms


model dephasing of the magnetization vectors.

iJCH

and iJHH

69

In equation [6.9], C is a vector of the coefficients Cj , and Z is a matrix of the


eigenvectors Z j

Equation [6.7] may be used to calculate the temporal evolution of the six density
matrix elements P15' P26' P27' 1>36' 1>37 and P48 representing the carbon transverse
magnetization in the scalar coupling interaction representation. The temporal evolution
of the other density matrix elements may be calculated in an analogous manner with
different block-diagonalized relaxation matrices. The temporal evolution models the
effects of spin-lattice relaxation and transverse (spin-spin) relaxation. Moreover, since
the relaxation matrix S includes imaginary terms along the diagonal, dephasing of the
magnetization vectors is modeled as well. Temporal evolution of the density matrix
elements is applicable to relaxation simulated for the

'r

and D2 periods. In contrast,

relaxation during the free induction decay is simulated as a frequency evolution of the
density matrix in order to produce a frequency spectrum instead of a free induction
decay. The frequency evolution may be obtained by a Fourier transform of equation

j)

[6.7]. This Fourier transform is expressed in terms of the real parts Re( A

and

imaginary parts Im( Aj ) of the eigenvalues Aj

[6.10]

Inspection of equation [6.10] reveals that the real part Re(Aj) of each eigenvalue Aj
determines the line width of a complex Lorentzian function, while the imaginary part

Im( Aj ) determines the frequency shift of that Lorentzian (7).

The frequency shift of the

spectral lines corresponds to the dephasing of the magnetization vectors, since both arise
from the imaginary terms along the diagonal of the relaxation matrix S. The spectral

70

lines L 1(/), L 2(/) and L 3 (/) of the carbon triplet, listed in order from high to low
frequency, are given by

L1(f) = P4S(/)
2

L2 (f) = L2 (f) + Vi. (f) = P27 (f) - 1>36 (f)


2

[6.11]

L3 (/) = P15(f)
2

The fitting software does not fit the spectral lines separately, but rather sums
them to compute the theoretical spectrum for comparison to the experimental spectrum.
Note, fitting to spectra as opposed to integrated line intensities (8) results in a ten fold
improvement in the marginal standard deviations of the fitting parameters. The fitting
parameters obtained from the nonlinear least-squares minimization include parameters
describing the molecular rotational diffusion, the dipole-dipole interactions, and the
chemical shielding anisotropy (CSA), as shown in Table 6.1.
Three rotational diffusion coefficients Rxx , Ryy and Rzz describe the rate of
rotation of the molecule about the principal axes of the molecular rotational diffusion.
Five additional parameters define the CHI> CH2 and HH dipole-dipole interactions.
Three of these parameters are the Euler angles

a, Pand r describing the orientation of

the 13CH2 moiety relative to the principal axis system of the molecular rotational
diffusion. This orientation is used to generate the spherical polar angles

~ij

and 8ij

applied as aj and Pj in the equations of Figure 3.1. The two other dipole-dipole
parameters are the carbon-proton distance

rCH

and the valence angle 8HCH ' These two

parameters are used to generate the dipole-dipole distances 1";,j used in equation [2.29].
Five more parameters define the chemical shielding tensor associated with the 13C
nucleus, as well as the orientation of that tensor relative to the principal axis system of
the molecular rotational diffusion. These five parameters are the anisotropy Li c , the

71

Table 6.1
Nonlinear Least-Squares Fitting Parameters

Moiety

Molecule
13 CH2
13e

Parameters

a
ae

Rxx

Ryy

Rzz

P
Pc

r
re

TeH 8HCH
Lie 17e
LiH

72

asymmetry l7c, and the three Euler angles

a c , Pc and r c' They are used in the

equations of Figure 3.1. Finally, the chemical shielding tensors associated with each of
the two protons are described with one additional parameter, L1 H

A number of

assumptions are required to justify one parameter only. It is assumed that the protons

=L1H1 = L1 H2 . Moreover, the proton CSA is assumed to


be axially symmetric, so 17H1 =l7H2 = O. Finally, the proton CSA tensors are assumed

have equal anisotropy, so L1H

to be oriented along the CH bond vectors, hence the required Euler angles are obtained
from the orientation of the 13CH2 moiety.
In addition to the fitting parameters discussed above, the minimization directly
fits seven random-field reduced spectral densities that cannot be parameterized via the
Favro diffusion model.

These spectral densities include two carbon-carbon

autocorrelation spectral densities


autocorrelation spectral densities
correlation spectral densities
correlation spectral density

.1 ~~(mc)

.1 ~(mH)

and

and

.1 ~(O),

.1 ~(O),

two proton-proton cross-

.1 :f.H2(mH) and .1 :fH2(O) , and

.1 ~~(O).

two proton-proton

a carbon-proton cross-

Since the minimization fits these reduced spectral

densities directly, they absorb the product of the undefmed interaction constants ~fDM
and ~fDM.
Other fitting parameters are required to superimpose the theoretical and
experimental spectra. For each relaxation experiment, a frequency parameter /!if shifts
the theoretical spectrum such that the theoretical and experimental spectra are aligned at
the center of L2. For each experiment, a phase parameter IIp rephases the theoretical
spectrum to obtain an optimum match to the experimental spectra.
Still other fitting parameters model the imperfections of the spectrometer. For
each relaxation experiment, a

Do parameter scales the absorption and dispersion spectra.

Only one parameter is necessary because CYCLOPS phase cycling of the reference
phase utilizes the absorption and dispersion acquisition channels symmetrically. Two
other parameters

Da and Dd model the baseline drift in the absorption and dispersion

73
acquisition channels respectively. In addition, for each relaxation experiment, four
parameters

1' 2' 3

and

model the efficiency of the carbon 1C/2 and 1C pulses and

the proton 1C/2 and 1C pulses, respectively. The efficiency of a pulse may drop below
100% due to imperfect pulse width calibration or radio frequency field inhomogeneity.
Certain pulse sequences, such as the carbon inversion recovery sequence, do not use all
four types of pulses and so do not require all four efficiency parameters.

6.1 References
1.

J. J. More, B. S. Garbow and K. E. Hillstrom, "User Guide for MINPACK-1",


p. 17, Argonne National Laboratory, Argonne, 1980.

2.

M. Goldman, "Quantum Description of High-Resolution NMR in Liquids", p. 61,


Oxford University Press, Oxford, 1988.

3.

A. L. Rabenstein, "Elementary Differential Equations with Linear Algebra", 3rd


Ed., p. 313, Academic Press, New York, 1982.

4.

R. A. Brown, R. H. Price and D. M. Grant,l. Magn. Reson. (in press).

5.

K. Blum, "Density Matrix Theory and Applications", p. 55, Plenum Press, New

York, 1981.
6.

E. Anderson et al., "LAPACK Users' Guide", p. 28, Society for Industrial and
Applied Mathematics, Philadelphia, 1992.

7.

E. O. Brigham, "The Fast Fourier Transform and its Applications", p. 27, PrenticeHall, Englewood Cliffs, 1988.

8.

D. M. Grant, C. L. Mayne, F. Liu, and T. X. Xiang, Chern. Rev. 91, 1591


(1991).

CHAPTER 7
RELAXATION OF [2. 13 C]GLYCINE
The slow tumbling motion of macromolecules in liquids occurs outside of the
extreme narrowing limit. In order to study this motional regime while avoiding the
additional experimental complexity of a macromolecule, relaxation experiments were
applied to [2_13C]glycine dissolved in glycerol-ds. The viscosity of glycerol is highly
temperature dependent, permitting the selection of slow tumbling motion of glycine by
temperature adjustment (1). A sample of [2_13C]glycine was dissolved in glycerol-ds at
a concentration of 165 mM, and 0.4 ml of this sample was placed in a five mm o.d.
NMR tube. Deuterium locking was accomplished by means of acetone-d6 in a 2.5 mm
o.d. capillary tube. Relaxation studies were performed across a temperature range from
288 K to 323 K, using a 9.4 T Varian XL-400 spectrometer. Perturbation of the AX2
spin system and subsequent measurement of spin-lattice relaxation was accomplished via
five relaxation experiments. Specifically, the carbon inversion recovery, hard pulse, Jpulse, sandwich sequence and INEPT experiments were used. For each experiment, the
pulse sequence was preceded by a delay period Dl equal to 30 times the decoupled
carbon T 1 of the spin system. Similarly, the observe pulse was preceded by a relaxation
period D2 ranging from zero to 30 times T 1 These two periods ensured adequate
sampling of the spin system as it returned completely to thermal equilibrium.
Data from all relaxation experiments performed at a particular temperature were
processed simultaneously by the nonlinear least-squares minimization software. As
described in Chapter 6, the least-squares fit was accomplished by matching a theoretical
NMR spectrum to an experimental spectrum. Each spectrum included approximately

75
300 data points (150 from the absorption mode spectrum and 150 from the dispersion
mode spectrum). Each of the five relaxation experiments conducted at a given
temperature included 18 spectra, owing to the use of 18 different relaxation periods D2.
Hence, for each temperature 90 different spectra resulted in about 27 thousand data
points for the least-squares fit. In order to facilitate interpretation of the least-squares fit,
the theoretical and experimental spectra were integrated between frequency limits
corresponding to the spectral lines Ll, ~ and~. These integrated line intensities were
then used to construct magnetization modes that were plotted against the relaxation
period D2, as shown in Figure 7.1.
The minimization software obtained values for the fitting parameters, with the
exception of those parameters that were locked to predetennined values. The parameters
a c , Pc,

r c'

Ltc and TIc defining the chemical shielding tensor associated with the 13C

nucleus, as well as the orientation of that tensor relative to the principal axes of
diffusion, were locked to values taken from the NMR literature (2). The parameters

rCH

and 8HCH defining the geometry of the methylene moiety were locked to 1.20 A and
109 respectively. The principal axes of diffusion were defined so as to coincide with
the principal axes of the methylene moiety, as depicted in Figure 7.2. Thus the
parameters

a, Pand r defining the rotation of the methylene moiety relative to the

principal axes of diffusion were locked to zero. The rotational diffusion of the glycine
molecule was found to be consistent with that of a symmetric top, hence only the two
rotational diffusion parameters R1. =R1QC =Ryy and R II =Rzz are required to describe the
rotational diffusion. The nonlinear least-squares fit determined the values for the
rotational diffusion parameters R1. and R II , the proton chemical shielding anisotropy
parameter Lt H , the random-field spectral densities ,1 ~~(md, ,1 ~(O), ,1

:'Mm

H ),

,1 ~(O), ,1 :f,H2(mH), ,1 :f,H2(0) and ,1 ~(O), and the 6, e, Ilf and IIp parameters.
The fitted values for the rotational diffusion parameters R1. and R II are shown in
Table 7.1. The temperature dependence of these parameters follows the Arrhenius

76

1.4

r----r---,------r----r----r---r---~--~-___,.--...,

1.2
1

0.8
0.6
0.4
0.2

o ........................................................................................................................................................................
-0.2

'--_-.1.._ _. . . L - _ - - - 1_ _........_ _..L.-_---L_ _- ' - - _ - - - I L . . . - _ - - L . _ - - - - J

0.2

0.4

0.6

0.8

1.2
Relaxation Period (s)

1.4

1.6

1.8

Figure 7.1. Experimental data indicated by diamonds for Ml and by crosses for M4 are
shown for application of the sandwich sequence experiment to [2_13C]glycine in
glycerol-ds at 303 K. A nonlinear least-squares fit of the experimental data is indicated
by solid curves. Intensities are normalized to the total carbon magnetization Ml at
thermal equilibrium (D2 = 4.0 s).

77

Figure 7.2. The principal axis system of the molecular rotational diffusion for glycine.
The x and z axes lie in the CCN plane. The x and y axes lie in the HCH plane. The x
axis bisects the HCH valence angle. Glycine is depicted in zwitterionic fonn, with NHr
and CO; moieties.

78

Table 7.1. Rotational Diffusion Coefficients for [2_13C]Glycine in Glycerol-dga

Rl.

288

0.0312

0.00143

0.0888

293

0.0547

0.00096

0.173

0.00242

298

0.0643

0.00911

0.219

0.00264

303

0.0908

0.00129

0.286 0.00287

308

0.123

0.00232

0.355

0.00431

313

0.149 0.00321

0.435

0.00635

318

0.168

0.00415

0.539

0.0107

323

0.179 0.00267

0.673

0.00767

0.00365

aRotational diffusion coefficients are given in units of 109 rad s-l.

79
relation

R= kT e-AGt/kT = kT eASt/k-Mlt/kT
1i
1i

[7.1]

predicted by statistical mechanics of liquids (3,4). This temperature dependence is


demonstrated in Figure 7.3. From equation [7.1] and the slopes and intercepts of the
linear regression lines shown in Figure 7.3, it is possible to calculate rotational activation

=8.55 and MI~ =9.19 kcal mol-I, and rotational entropies of


ASi =2.26 and AS~ = 6.62 cal mol- l K-I. The correlation coefficients for the
regression lines are r.L =0.975 and ru = 0.981. These correlation coefficients indicate
energy barriers of Mli

reasonable adherence to the Arrhenius model across the entire temperature range.
Moreover, as the values for the rotational diffusion parameters are obtained through
application of the Favro diffusion model, the correlation coefficients also indicate
reasonable adherence to the Favro diffusion model. Note, the Arrhenius relationship
persists on the low-temperature side of the TI minimum, well beyond the limits of
extreme narrowing.
It is apparent from Table 7.1 that R.L is less than R II at each temperature studied.
This effect probably arises from the zwitterionic form of glycine depicted in Figure 7.2.
The zwitterion creates a net electric dipole oriented along the z axis. Rotation of the
zwitterion about the z axis does not reorient the dipole; however, rotation about the x or
y axis does reorient the dipole. Since

Mli = Mli, the rotational activation energy

barriers are about equal. Hence the energy required to reorient the dipole, as well as to
reorient the compensatory dipole in the solvent molecules, appears primarily as the
rotational entropy difference AAS t = ASi-

ASi = -4.36 cal mol- l K-I.

7.1 References
1.

L. E. Kay, T. E. Bull, L. K. Nicholson, G. Griesinger, H. Schwalbe, A. Bax and


D. A. Torchia, J. Magn. Reson. 100, 538 (1992).

80

3.ooo~--------------------.

1.000

o
0.100 -+-..,.-.,..-,.......,.......,.--,-....,.-..,.......,.-.,..-r--r......,........,......,.--r--..,.--r--r--r---l--r-'
3.050
3.150
3.250
3.350
3.450
liT x 103 K- 1

Figure 7.3. Semilog plots of Rl./T and RII/T versus T- 1 for [2_13C]glycine in
glycerol-dg demonstrate an Arrhenius relationship across a temperature range from
288 K to 323 K. A semilog plot of Tl versus T- 1 shows the Tl minimum, indicating
non extreme narrowing conditions.

81
2.

R. A. Haberkorn, R. E. Stark, H. van Willigen, and R. G. Griffin, J. Am. Chern.


Soc. 113, 2534 (1981).

3.

R. K. Harris, "Nuclear Magnetic Resonance Spectroscopy", p. 54, John Wiley &


Sons, Inc., New York, 1986.

4.

S. Glasstone, K. J. Laidler and H. Eyring, "The Theory of Rate Processes",


p. 522, McGraw-Hill, New York, 1941.

CHAPTER 8
RELAXATION OF ABU23 .[2. 13 C]GLy24.GCN4.pl
The leucine zipper peptide GCN4-pl is the dimerization region of the yeast
transcriptional activator protein GCN4 (1). It is composed of 33 amino acids and has a
molecular weight of four kiloDaltons. In aqueous solution it forms an a-helix that is
continuous for at least 32 of the 33 residues (2). The helix does not exist as a monomer
(3). Instead, two helices dimerize, twisting around one another to form a superhelix,
similar in structure to a two-stranded rope (1). This superhelix is roughly cylindrical in
shape, having a length of 49 A and a radius of 11

A.

The dimer forms because of the peculiar amino acid sequence of GCN4-pl

Ac-RMKQLEDKVEELLSKNYHLENEVARLKKLVGER

In this sequence, Ac represents acetylation of the amino terminus. As designated in


boldface letters in the sequence, leucine (L) repeats every seventh residue. Valine (V)
also repeats every seventh residue (with one exception), and the occurrences of valine
are shifted by four residues relative to the leucine positions. Between these hydrophobic
residues, the peptide is rich in hydrophilic residues such as lysine (K), arginine (R),
glutamic acid (E) and aspartic acid (D). The seven repeating residue positions may be
mapped onto a "helical wheel" diagram as depicted in Figure 8.1. One tum of an
a-helix requires three and one-half residues. Hence, the helical structure of the peptide
places the hydrophobic residues leucine and valine on one face of the helix (positions a
and d), while placing hydrophilic residues on the other faces (positions b, c, e,fand g).

83

L29
E22
K15
K8
R1

D7
S14
N21
K28

K27
E20
L13
E6

D7
S14
N21
K28

E6
L13
E20
K27

R1
K8
K15
E22
L29

Figure 8.1. A helical wheel diagram of the leucine zipper peptide GCN4-pl. The
positions of the amino acids are indicated by letters a through g. The amino terminal
amino acids are enumerated adjacent to the helical wheel, while the carboxy terminal
amino acids are enumerated toward the periphery of the diagram. The superhelical twist
is not shown. B represents L-a.-aminobutyric acid (abu).

84
Upon dimerization, the hydrophobic face of one helix is joined to the hydrophobic face
of a second helix, protecting both hydrophobic faces from the solvent. The hydrophilic
residues project out into the solvent and account for the relatively high solubility of
GCN4-pl.
The leucine zipper peptide GCN4-p 1 has been modified to permit 13C relaxation
studies. Alanine24 has been replaced with [2_13C]glycine so as to introduce a labeled
methylene moiety into the main chain of the peptide. The glycine is located nine residues
from the carboxy terminus. This position is far enough from the terminus to prevent any
fraying motion at the terminus from affecting measurements of the tumbling motion of
the peptide main chain. Also, it is assumed that the helical structure is sufficiently stiff at
this position that vibrations are minimized and can be neglected in the analysis of
tumbling motion. Valine23 has been replaced with L-a-aminobutyric acid (abu). This
amino acid replacement is unimportant for molecular tumbling studies. However, future
experiments are planned for studying side chain conformatonal dynamics using
[3-13C]abu23 . For the present, the tumbling motion of the peptide main chain has been
probed with abu23 _[2-13C]gly24-GCN4-pl.
The abu23 _[2-13C]gly24-GCN4-pl peptide was synthesized via solid-phase
techniques (4). The peptide was purified to 95% purity using HPLC. This process
removed short polypeptide fragments resulting from incomplete coupling of amino acids
during the synthesis. The molecular weight of the HPLC-purified peptide, measured
using electrospray mass spectrometry (5), was 4010.25 (the theoretical molecular weight
is 4010.65). The peptide was dissolved to a concentration of 1 mM in 99.9% D20 and
subsequently lyophilized three times in order to replace labile protons with deuterons.
The peptide was then dissolved to a concentration of nine mM in a 50 mM
CICD2COOD/CICD2COONa buffer having a pD of 2.95. This buffer was prepared
using 200 mM NaCI dissolved in 99.9% D20.

The peptide concentration was

confirmed via tyrosine absorbance at 274 nm (6). Helicity, and therefore the presence of

85

the dimer (3), was confmned using circular dichroism to measure the molar ellipticity at
222 nm (7).

The peptide was again lyophilized and redissolved to nine mM

concentration in 99.996% D20, in order to minimize the concentration of labile protons.


A 0.8 ml sample of the buffered peptide was placed in a five mm o.d. NMR
tube. Relaxation studies were performed across a temperature range from 293 K to 308
K, using a 9.4 T Varian XL-400 spectrometer. Perturbation of the AX2 spin system and
subsequent measurement of spin-lattice relaxation was accomplished via five different
relaxation experiments. Specifically, the carbon inversion recovery, proton inversion
INEPT, refocused INEPT, multiple-quantum INEPT and INEPT experiments were
used. The carbon inversion recovery experiment was chosen instead of the carbon
inversion INEPT experiment. Although the carbon inversion INEPT experiment results
in a larger signal to noise ratio than the carbon inversion recovery experiment, only the
carbon inversion recovery experiment permits estimation of the carbon T 1 of the spin
system. The proton Tl was estimated from the proton inversion INEPT experiment, and
found to be approximately three times longer than the carbon T 1 For each experiment
the pulse sequence was preceded by a delay period Dl equal to five times the proton Tl
of the spin system. Similarly, the observe pulse was preceded by a relaxation period D2
ranging from zero to five times the proton Tl.
Data from all relaxation experiments performed at a particular temperature were
processed simultaneously by the nonlinear least-squares minimization software. The
least-squares fit was accomplished by matching a theoretical NMR spectrum to an
experimental spectrum. The experimental spectrum included the two clearly-identifiable
multiplets shown in Figure 8.2. The theoretical spectrum was generated to fit lines

L~,

L1,

I!f and Lj, by calculating the relaxation for two multiplets offset from one another

by about 160 Hz. The multiplet giving rise to lines

L1 and L~ represented about 90%

of the spectral intensity, while the multiplet giving rise to lines

Li

and Lj represented

about 10% of the intensity. However, the relative intensity of the multiplet giving rise to

86

o ......

300

200

100

-100
Frequency (Hz)

-200

-300

-400

Figure 8.2. Theoretical and experimental spectra are shown for application of the
multiple-quantum INEPT experiment to abu23_[2-13C]gly24-GCN4-pl in D20 at 303 K.
The dimer gives rise to lines L1 and L~. The monomer gives rise to lines Li and L3.
Polypeptide fragments give rise to lines such as
The theoretical spectrum is

L;.

calculated for lines

L1, L~, Li

and

IJ3 , but not for line L;.

87

LT and L3 increased with temperature, from 7.7% at 293 K to 10.7% at 303 K.


Also, lines LT and L3 relaxed more rapidly than lines L~ and L~. These intensity and
lines

relaxation features are consistent with a monomer

dimer equilibrium that shifts

towards the monomer with rising temperature (3). Lines

LT

and

L3

belong to the

monomer and lines L~ and L~ belong to the dimer. In addition, the experimental
spectrum included other recognizable, but low-intensity, multiplets giving rise to lines
such as L~. However, the signal to noise ratio of these lines is too low to permit reliable
fitting. These lines were ignored for purposes of the least-squares fit, and probably
represent polypeptide fragments not removed by the HPLC purification.
At each temperature, 90 different spectra were acquired, resulting in about 27
thousand data points for the least-squares fit. In order to facilitate interpretation of the
least-squares fit, the theoretical and experimental spectra were integrated between
frequency limits corresponding to the spectral lines L1, ~ and L3. These integrated line
intensities were then used to construct magnetization modes that were plotted against the
relaxation period D2, as shown in Figure 8.3. Due to the increased width of the spectral
lines in the peptide, it was not possible

to

resolve the spectral lines arising from the

dimer from the spectral lines arising from the monomer. Hence, the nominal spectral
lines, and the magnetization modes created from them, included contributions from the
dimeric and monomeric forms of the peptide. However, these various contributions
were properly treated by fitting the theoretical spectrum to the experimental spectrum.
As in the case of [2_13C]glycine, the minimization software obtained values for
the fitting parameters, with the exception of those parameters that were locked to
predetermined values. The parameters a e , Pc, re, Lie' l1e and
values taken from the NMR literature (8,9). The parameters

rCH

LiH

were locked to

and 8HeH defming the

geometry of the methylene moiety were locked to 1.09 A and 1090 respectively. The
principal axes of diffusion were defined with the z axis along the long axis of the
cylinder and the x and y axes perpendicular to the long axis. The rotational diffusion of

88

1.5
<>

<>

0.5
0
o

-0.5
-1

-1.5
-2
-2.5

0.2

0.4

0.6

1
1.2
0.8
Relaxation Period (s)

1.4

1.6

1.8

Figure 8.3. Experimental data indicated by diamonds for MI. by crosses for M3 and by
squares for Ms are shown for application of the refocused INEPT experiment to
abu 23 _[2-I3C]gly24-GCN4-pl in D20 at 308 K. A nonlinear least-squares fit of the
experimental data is indicated by solid curves. Intensities are nonnalized to the total
carbon magnetization MI at thennal equilibrium (D2 =4.2 s).

89
the peptide dimer was found to be consistent with that of a symmetric top, hence only
the two rotational diffusion parameters R.1.

=Rxx =Ryy

and RII

=Rzz

are required to

describe the rotational diffusion of the peptide dimer. The peptide monomer probably
assumes a random coil form (3) with no symmetry axis, so a rigorous treatment of its
relaxation should require three rotational diffusion parameters. However, the leastsquares fits revealed that three parameters provided too many degrees of rotational
freedom for the monomer. It was therefore preferable to treat the monomer as a
symmetric top and use only two rotational diffusion parameters to describe its rotational
diffusion. Hence, the parameters for which values were determined by the nonlinear
least-squares fit are the rotational diffusion parameters R.1. and RII for the dimer, another
pair of rotational diffusion parameters R.1. and R II for the monomer, the random-field

:a

.1 ~~(cod, .1 ~~(O), .1 (coH) , .1


.1 ::'H2(0) and .1 ~~(O), and the ~, e, AI and IIp parameters.

spectral densities

:a

(0),

.1 ~H2(COH)'

The fitted values for the rotational diffusion parameters R.1. and R II for both the
monomer and the dimer are shown in Table 8.1. Figure 8.4 depicts semilog plots of
R.1./T and RII/T versus T-1 for the dimer. These plots demonstrate that the temperature

dependence of R.1. and R II follows the Arrhenius relation expressed in equation [7.1].
The correlation coefficients for the regression lines shown in Figure 8.4 are

7.1.

= 0.896

and 'il = 0.974. From the regression lines, the calculated rotational activation energy
barriers are
are

MIl =1.29 and MIt =12.4 kcal mol-

The calculated rotational entropies

ASl =-26.5 and ASt =12.8 cal mol-1 K-1.


As was the case for [2_13C]glycine, R.1. is less than RII at each temperature

studied. The rotational entropy difference AAS';

=ASl- ASt =-39.3

cal mol- 1

K-1

may be indicative of electric dipoles aligned along the z axis that are reoriented by
perpendicular rotation. The numerous hydrophilic side chains depicted in Figure 8.1
may give rise to these dipoles.

90

Table 8.1. Rotational Diffusion Coefficients for Abu23 _[2-13C]Gly24-GCN4-pl a.b

R.l.

293

0.00690 0.00011
0.0325 0.0038

0.0139 0.0002
0.134 0.019

298

0.00706 0.00004
0.0282 0.0024

0.0159 0.0002
0.149 0.016

303

0.00736 0.0001
0.0293 0.0014

0.0004
0.197 0.004

308

0.0002
0.0223 0.0000

0.0391

0.00812

0.0273

0.0007
0.266 0.000

aRotational diffusion coefficients are given in units of 109 rad s-l.


bPor each temperature, the upper and lower lines list parameters for the dimer and
monomer respectively.

91

0.2,-------------------------------------~

R II / T

10-6 rad s-l K- 1

R.J T

10-6 rad s-l K- 1

0.1

o.02

-+-~r__T'"""T""_r_.,.....,__,__,__~I"'""""I"_.......,._~r__T'"""T""...,._.,.....,__,__,__..,.......

3.2

3.25

3.3

3.35

3.4

3.45

Figure 8.4. Semilog plots of R.1./ T and Rill T versus T- 1 for the dimeric form of
the leucine zipper peptide abu 23 _[2- 13C]gly24_GCN4-p1 in D 20 demonstrate an
Arrhenius relationship across a temperature range from 293 K to 308 K.

92
8.1 References
1.

E. K. O'Shea, J. D. Klemm, P. S. Kim and T. Alber, Science 254, 539 (1991).

2.

T. G. Oas, L. P. McIntosh, E. K. O'Shea, F. W. Dahlquist and P. S. Kim,


Biochemistry 29,2891 (1990).

3.

E. K. O'Shea, R. Rutkowski and P. S. Kim, Science 243, 538 (1989).

4.

B. Merrifield, Science 232, 341 1986).

5.

I. Jardine, Nature 345, 747 (1990).

6.

C. R. Cantor and P. R. Schimmel, "Biophysical Chemistry", Part II, p. 364, W.


H. Freeman and Company, New York, 1980.

7.

C. R. Cantor and P. R. Schimmel, "Biophysical Chemistry", Part II, p. 425, W.


H. Freeman and Company, New York, 1980.

8.

R. A. Haberkorn, R. E. Stark, H. van Willigen, and R. G. Griffin, 1. Am. Chem.


Soc. 113, 2534 (1981).

9.

B. Voigtsberger and H. Rosenberger, Phys. Stat. Sol. 35, K89 (1976).

93

CHAPTER 9
THE AXY SPIN SYSTEM
23

13

24

Relaxation data that were obtained for the abu -[2- C]gly -GCN4-p1
leucine zipper peptide demonstrated unequal rates of relaxation for the two
protons (1) and hence suggest that the two protons are not magnetically equivalent.
Therefore, an AX2 spin system does not provide a sufficiently accurate model for
the methylene

13

CH2 moiety of this peptide. A more accurate model is the AXY

spin system wherein spin A represents the carbon atom and spins X and Y
represent the two non-equivalent protons.

In the spin-product basis, the

Hamiltonian ! for this AXY spin system is defined mathematically as


H

! = C !C H1 ! 1 H! ! ! + CH !C ! 1 + CH !C ! !

1
H H
H
H
+ HH ! 1 ! ! + HH ! 1 !H! + !H1 ! ! 9.1
2
Comparison of equation [9.1] to equation [5.4] reveals that equation [5.4]
contains a single proton frequency H whereas equation [9.1] contains two proton
frequencies H! and H! due to the different chemical shielding of the two
protons.

Each equation contains only a single carbon-proton scalar coupling

constant CH because in both the AX2 and AXY spin systems each methylene
proton forms the same type of chemical bond with carbon atom and the carbonproton scalar coupling is mediated by the bonding electrons, not by the magnetic
environment.

94

The eigenvalues of the AXY spin system are


1
1
1
1
1
! = C H1 H2 + CH + HH
2
2
2
2
4
1
1
1 !
!
! = C HH
HH + H2 H!
2
4
2
1
1
1 !
!
! = C HH +
HH + H2 H!
2
4
2
1
1
1
1
1
! = C + H1 + H2 CH + HH
2
2
2
2
4
9.2
1
1
1
1
1
! = C H1 H2 CH + HH
2
2
2
2
4
1
1
1 !
!
! = C HH +
HH + H2 H!
2
4
2
1
1
1 !
!
! = C HH
HH + H2 H!
2
4
2
1
1
1
1
1
! = C + H1 + H2 + CH + HH
2
2
2
2
4
In the spin-product basis, the un-normalized eigenvectors are
1 =1
2 =

3 =

4 =0
5 =0
6 =

7 =

!
H! H! + HH
+ H 2 H !

HH
!
H! H! HH
+ H 2 H !

0
0
0

0
0
0

1
0
0

0
1
0

8 =0

HH
0 0
0 0

0
0

9.3

!
H! H! HH
+ H 2 H !

HH
!
H! H! + HH
+ H 2 H !

HH
0

95

The square root

!
HH
+ H 2 H !

in equations [9.2] and [9.3]

complicates the evaluation of equation [5.6] by computer algebra systems such as


Maple (2). The special case of H! H! = HH eliminates this square root and
simplifies the eigenvalues and eigenvectors of the AXY spin system. Under this
simplification, the Hamiltonian ! is
H

! = C !C H ! 1 H + HH ! ! + CH !C ! 1 + CH !C ! !
1
H H
H
H
+ HH ! 1 ! ! + HH ! 1 !H! + !H1 ! ! 9.4
2

This equation contains the single proton frequency H of the first proton because
the frequency of the second proton is H + HH . The eigenvalues of the
Hamiltonian ! of equation 9.4 are
1
1
1
! = C H + CH HH
2
2
4
1
2 2+1
! = C
HH
2
4
2 21
1
! = C +
HH
2
4
1
1
3
! = C + H CH + HH
2
2
4 9.5
1
1
1
! = C H CH HH
2
2
4
1
2 21
! = C +
HH
2
4
1
2 2+1
! = C
HH
2
4
1
1
3
! = C + H + CH + HH
2
2
4

96

The normalized eigenvectors of the Hamiltonian ! of equation 9.4 are


1 =1
2 =0
3 = 0

0 0 0
1 2

42 2
1+ 2

0 0
1
42 2
1

0
0

4+2 2
4+2 2
4 =0 0 0 1 0 0 0 0
5 =0 0 0 0 1 0 0 0
1+ 2
6 = 0 0 0 0 0
4+2 2
1 2
7 = 0 0 0 0 0
42 2
8 =0 0 0 0 0 0 0 1

0
0

0
0
9.6

1
4+2 2
1
42 2

0
0

When H! = H! equation [9.2] becomes equation [5.8] and equation [9.3]


becomes the un-normalized form of equation [5.9]. The energy level diagram for
the AXY spin system is the same as the energy level diagram that is shown for the
AX2 spin system in Figure 2.1.
9.1 References
1. R. A. Brown and D. M. Grant, J. Magn. Reson. B106, 253, (1995).
2. L. Bernardin, et al., Maple Programming Guide, Maplesoft, 2015.

S-ar putea să vă placă și