J. A. S.

Smith
School of Molecular Sciences Nuclear Quadrupole Resonance Spectroscopy
University of Warwick
Coventry CV4 7AL England General principles

Nuclear quadrupole resonance spectros-

copy is a branch of magnetic resonance spectroscopy
and is concerned with the absorption of radio waves by
matter in zero magnetic field. A unique feature of this
kind of spectroscopy is the remarkably simple instru-
ment that can be used to detect some signals, a circuit
that can be constructed by anyone with some experi-
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

ence of radio techniques at a cost comparable to that of

Downloaded via UNIV ILLINOIS URBANA-CHAMPAIGN on May 28, 2022 at 21:48:08 (UTC).

the cheapest transistor radio. We will return to the

subject of instrumentation, however, in the second of
this series of articles; the first deals with the basic
principles of nuclear quadrupole resonance spectros-
copy in solids, and subsequent parts will summarize
some recent applications to chemistry (see Editor’s Figure 1. Nuclear electric quadrupole moment.
Note). this integral will be zero, since the average value of z1
General Principles equals that of r2/3, but for a prolate spheroid the inte-
In order to understand the origin of nuclear quad- gral is finite and positive, and negative for an oblate
rupole resonance, we need to consider in some detail the spheroid; because of the presence of the fundamental
electrical interactions between the nuclei and electrons unit of charge on both sides of eqn. (1), Q itself has the
which form part of an atom or molecule. Starting.at dimensions of r2, viz., cm2, and is of the order of the
the simplest level, we may visualize the molecule as a nuclear area, viz., 10”24 cm2. Nuclear quadrupole mo-
ments will therefore be expressed in units of 10”24 cm2
charge cloud extending over several angstroms, in
which are embedded the nuclei with their own consid- and can be either positive or negative.
To return to Figure 1, we note that compression of
erably smaller charge distributions, about 10”4 A in
size. We then ask what symmetry is the nuclear the original sphere has removed positive charge from
two regions, marked —, and transferred-it to two other
charge distribution likely to have. Since all the nuclei
giving rise to nuclear quadrupole resonance have a regions, marked +. From a spherical charge distri-
magnetic moment which we explain in terms of nuclear bution, the compression has given two antiparallel clcc-
.tric dipoles. In an electric field, each of these expe-
angular momentum, we expect all quadrupolar nuclei
in our laboratory frame of reference to have one axis of riences a torque tending to align them along the field;
since the two turning torques are equal and opposite,
cylindrical symmetry, the "spin” axis, or the axis with
there is no net effect. However, in an electric field
respect to which the angular momentum is defined.
Such a charge distribution is represented in general not gradient which we assume to be axially-symmetric
about Oz in Figure 1, the torques on the two dipoles are
by a sphere but by an ellipsoid of revolution which we
not equal and a net turning torque exists which is
may derive from a sphere by compression or extension
along a given direction (Fig. 1). Compression gives an clearly proportional both to the electric field gradient
oblate spheroid, and extension a prolate spheroid; the and to the extent to which the nuclear charge distri-
bution departs from spherical symmetry, that is to eQ.
“quadrupole moment” Q of such a charge distribution
may be defined as Calling the electric field gradient in the Oz direction
(dli/dz) —eq, the net turning torque is proportional
=

eQ =
fpniSzn* —

>\d)dTw (1) in magnitude to |e2r^Q|. Since all quadrupolar nuclei

where pn is the nuclear charge density in a volume ele- also have angular momentum, the situation may be com-
ment dr„; e is the charge on the electron; and the inte-
—

pared to the precessional model for a nuclear magnet in

an applied magnetic field, a model useful in discussing
gral extends over the nuclear volume. For a sphere,
nuclear magnetic resonance; the field tries to turn the
Editor's Note : Professor Smith's review of Nuclear Quadru- magnet into parallelism, and the magnet responds in
some respects as would a gyroscope by processing about
pole Resonance spectroscopy will appear in four installments, of
which this is the first. Subsequent sections will appear in the the direction of the magnetic field with a precessional
“Topics in Chemical Instrumentation" column in the A-pages of frequency w (in radians sec”1) of ynH (y„ is a constant
this Journal for February, March, and April, 1971. Reprints known as the nuclear gyromagnetic ratio). This equa-
will be prepared for the series as a whole, and will be available in
limited quantity from the author and from G. W. Ewing, Depart- tion comes from Larmor’s theorem. By analogy, the
ment of Chemistry, Seton Hall University, South Orange, N. J. nuclear quadrupole responds to the torque by precessing
07079. about the direction of the maximum electric field gra-

Vo/u/ne 48, Number 1, January 1971 / 39

dient with afrequency which should be proportional to chanics, and we next discuss the results of such calcula-
\e2qQ/h\ (where h Planck’s constant). However,
=
tions. In order to do this, we need more precise defini-
there is an important difference from the magnetic case tions of both the nuclear quadrupole moment and the
which arises essentially because we have used different electric field gradient. Mathematically, we will pro-
coordinate systems to define q and Q; in the quadru- ceed by first deriving the full expression for the energy
polar case, the precessional frequency varies with orien- of a nuclear electric quadrupole in an electrostatic en-
tation, whereas in the magnetic case, the frequency is vironment, and then presenting the solutions of these
the same for all orientations. The model of quad- equations. For the intermediate stages, in which the
rupolar precession approaches the quantum-mechanical quantum-mechanical operator is first derived and the
one only in the limit of high quantum numbers. If, eigenvectors and eigenvalues obtained, the reader is
however, we imagine that the nuclear spin axis can be referred to the excellent introductory account by
“set” at one or more fixed orientations, conveniently Slichter (2).
defined by Jz, the component of the nuclear angular mo- Consider a quadrupolar nucleus which is part of an
mentum along Oz, then at one of these positions there atom or molecule and a volume element in that charge
will be one precessional frequency; this precession of the distribution dr„ dxdi/dz of charge density pn distant
=

nucleus carries with it, of course, the nuclear angular rn from our origin at 0 (midway between the two foci of
momentum and hence the nuclear magnetic moment, the ellipse, see Figure 1). The nuclear charge is imag-
thereby producing a small magnetic field at right angles ined to be under the influence of the electronic charge
to Oz rotating in phase with the nuclear quadrupole at distribution of the atom or molecule, which produces a
the same frequency. Now a linearly oscillating field potential at dr„ of V. The potential energy of this nu-
clear charge element in its electrostatic environment is
H =
2Hi cos at (2) then
can be resolved into two rotating fields —

V
H{cw) Hi cos at + iHi sin at
=
and for the complete nucleus
H(ccw) Hi cos at
<=
iHi sin at—

Potential energy -fpnV-dr„

=
(4)
where
Now V varies over the nuclear volume, but
we may ex-
H H( cw)-+ H( ccw) press this variation in terms of the values of V and its
=

first and higher differentials at the orgin by means of a

so that H(cwr) rotates clockwise, H (ccw) rotates
Taylor expansion, i.e.
counterclockwise; such a field is obtained by enclosing
the sample in a small induction coil through wrhich we.
send alternating current. That component of eqn. (3)
which rotates in phase with the precessing nucleus can
exchange energy with it—in classical terms, we have a
resonance process occurring at a specific frequency,
which wre label vQ (in Hz) or aQ (in radians sec-1), the where Vo, (5F/5k)o, etc, are to be evaluated at the
nuclear quadrupole resonance frequency. Now Q for origin O and (x, y, z) are coordinates in the nuclear
35C1 is close to —0.0802 X 10-24 cm2 (see the table), and frame of reference (labeled xnynzn in Figure 1). Since
in HC1 this nucleus finds itself in an electric field gra- we are interested in effects which vary with nuclear orien-
dient (eq) of +0.96 X 1024 e cm-3 from the adjacent tation, the first term involving F0 will not be relevant to
proton (using re = 1.278 A) and +23.41 X 1024 from the what follows and may therefore be omitted. The
electrons, giving a net value of +24.37 X 1024; the second term involves first differentials of V, i.e., electric
product e2qQ/h is then —67.85 MHz, corresponding to fields, since
transition frequencies in the radio frequency region.
In fact, for all quadrupolar nuclei in the periodic table, E =
—grad V
the range of possible frequencies is from a few kHz to or
several thousand MHz so that the induction coil which dV bV bV
generates the exciting magnetic field will be part of a ~EX, E„. ~EZ
= — =

bx by bz
radio frequency spectrometer. The observed fre-
quencies, which, we compare with the classical pre- so that one term, viz., pnxn(bV /bx)0, represents the con-
cessional frequencies of an electric quadrupole, are tribution of the electric moment of the volume element
governed by electric effects with the atom or molecule, x„ with respect to the origin at 0. When the integra-
but the mechanism whereby they are observed involves tion of eqn. (4) is therefore performed, the second term
the magnetic moment of the nucleus, that is they are will represent the interaction of the electrostatic dipole
magnetic dipole transitions in origin. moment of the nucleus in the electric field of its environ-
The reader will appreciate that the previous para- ment; now there is good evidence, both experimental
graphs have been concerned with a highly simplified and theoretical, that nuclear electric dipole moments
description of nuclear quadrupole resonance in terms of are zero, and since the induced moment is also taken as
classical mechanics, and for a fuller account of the clas- zero, we may also omit the second term in eqn. (5).
sical equations of motion and their solutions he is re- The third part is the electric quadrupole term, to which
ferred to an interesting article by Raich and Good (1). we now turn our attention (neglecting higher terms,
The quantitative expressions for the nuclear quadru- which may however be important in certain cases).
pole resonance frequencies and their relative intensities The full expression for the quadrupole potential en-
are usually derived by the methods of quantum me- ergy will be

40 / Journal of Chemical Education

 d2F 02F VXI -

Vn
Jb2V
r' _

bxt by2 / dz2 F„ (10)

q will be called the electric field gradient (strictly

speaking, this quantity is —eq) and can be either posi-
tive or negative, y is called the asymmetry parameter
and is a positive number; it measures the departure of
the electronic charge distribution at the nucleus from
in which we have nine terms involving the nuclear co-
ordinates and nine involving second derivatives of the cylindrical symmetry, q, the maximum principal com-
electronic charge distribution. Such quantities, strictly ponent, will be zero only at sites of very high sym-
metry, e.g., tetrahedral and octahedral; thus the quad-
speaking, are tensor's and should be manipulated by the
methods of tensor algebra: we will make the assump- rupole interaction at the 33S nucleus in gaseous SFe and
at the 35C1 nucleus in crystalline NaCl will be zero.
tion, which is discussed in the monograph by Nye (3),
that it will always be possible to select a new coordinate y will also be determined to some extent by symmetry
considerations. For example, in studying 35C1 nuclear
system (x, y, z) in which only the simple derivatives
quadrupole resonance, one is principally (but not al-
52V/£>x2, etc., are non-zero; these will now be called the
ways) concerned with an atom which is attached to
principal components of the second derivatives, and
eqn. (6) becomes
only one other; in HC1 and CH3C1 (at least, in the gas),
we have axes of infinite and three-fold symmetry, re-
spectively, so that in both molecules Vxx
-/ (P Vm and y is =

Potential energy =
+ + In C6H5C1, however, the C-Cl bond is on an axis
2)p"Xn\bx* A jo zero.
of at most 2-fold symmetry, so that we must expect y to
There is, in addition, the possibility of
Pilin'
<S)i dr„ (7) be non-zero.
conjugation of 3p electrons on Cl with the ^-electron
The remaining terms in eqn. (7) are related to the nu- system of the ring; the interaction will be strongest for
clear charge distribution and hence to its quadrupole those 3p electrons whose lobes are perpendicular to the
moment. The condition for a nucleus to possess an plane of the benzene ring, and this will introduce an
electric quadrupole moment is that 7 > 1, where 7 is the additional asymmetry and a further contribution to y.
nuclear spin quantum number. Many common nuclei Having therefore defined the principal parameters,
fulfil this condition, i.e., 14N(7 1), 35C1 and 37C1(7
= =
we may now turn to the solution of the full quantum-
3/2), 5flCo(J = 7/2); the proton, 'II, has 7 y2, but its =
mechanical equations for the potential energy. The
common isotope 2II is eligible, since 7=1. Similarly, detailed derivation of these solutions is given by Das
we have 170(7 =
5/2) and 33S(7 3/2), the common iso-
=
and Hahn (4) in their monograph; we present here the
topes being non-magnetic (7 0). Other isotopes,
=
conclusions in the form of the energy levels of the
like 67Fc, are inaccessible in their nuclear ground states, system, which we classify as in the precessional model
since 7 1/2, but have a nuclear quadrupole moment in according to the projection It of the nuclear spin
=

one or more of their nuclear excited states, and so can angular momentum in the direction of the maximum
be studied by Mossbauer spectroscopy. principal component of the electric field gradient which
Equation (7) may be simplified even further by the defines F2Z. The quantum number we therefore use is
methods of quantum mechanics, as is described by that appropriate to the operator 72, viz., m, the nuclear
Slichter (2). One convention that is thereby intro- magnetic quantum number, which also is used to
duced is to redefine the nuclear quadrupole moment in specify the magnetic levels in the applied field in nu-
eqn. (1) such that the integration is carried out in the clear magnetic resonance. This is not a complete de-
state in which the maximum component of the nuclear scription of the system, for it deals only with the longi-
magnetic moment lies along the z axis (that is, 7a 7). =
tudinal components of nuclear magnetization, or those
The quantities in round brackets in eqn. (7) become the which are parallel to the 2-axis, and neglects the trans-
expectation or average values of the principal compo- verse components. These are important in relaxation
nents of the electric field gradient over the ground- processes, and will be reconsidered in the second article
state wave-function of the molecule—like the dipole in this series. It turns out that for a nuclear spin of 7,
moment, therefore, the electric field gradient is a ground- there will be 2 7 levels, and the way in which these are
state property. If we neglect the effects of the very arranged depends on whether 7 is half-integral or inte-
small electronic charge distribution within the nucleus, gral.
we may use Laplace’s equation and write
cPF cPF Half-Integral Spins
ax2
+ + u (8)
by2 dz2 For 7 half-integral, viz., 35C1 and 37C1 (7 3/2), or
=

in which case we need only two quantities to define the 1277 (7=
5/2), the levels occur in pairs, leaving a degen-
electric field gradient (in addition, of course, to their eracy in m which is not removed even when y is finite.
direction cosines with respect to the principal coor- Considering first the case of zero y, we show in Figure 2
dinates); these we select to be the maximum principal the order of the levels for the cases 7 s/2, 6/2, and 7/2
=

component, which is by convention taken to be d2F/ and by their side the appropriate equations giving the
dz2 or F22(|F22| > \Vyy\ > \VXX\), and we then define the magnitude of the frequencies. The selection rule for
parameters magnetic dipole transitions is Am ±1, so that for
—

7 =
3/2 we observe just one transition at a frequency
02F
eq =
5^ ( =
F„) (9)
(ID
and

Volume 48, Number 1, January 1971 / 41

 ±3/2: When , is non-zero, the energy-level diagram for

H 1
now
=
is unchanged in appearance, but the energies
3/2
contain a term in ,, so that again one transition
± 1/2 :

frequency is observed given by the equation

(aI I =
3/2

’-mX'+iY'-
The quadrupoie coupling constant cannot therefore be
derived unless q is known, since neither (e2qQ/h) nor q
is separately derived; in practice, if q < 0.1, as is often
the case in 35C1 quadrupoie resonance, the error in
setting twice the frequency equal to the quadrupoie
coupling constant is small. In order to derive q for I =

V2, separate experiments are necessary in weak magnetic

fields, and these will be described shortly.
For I 5/2, 7/2, etc., with non-zero
=
q, closed solutions
to the quantum mechanical equations do not exist, ex-
cept in the limiting cases of low q. For I 5/2, and low =

lb) I =
5/2 tc> I = 7/2 q (say <0.25)

Quadrupoie energy levels and transitions for (a) I

3
Figure 2.
—

A. (b) I —

»i -
d + 0.0925V 0.63403,*)
(-£-)
~ -

SA, (d I = 7A (, 0). =

(15)
The quantity in parentheses is often referred to as the n -
jjj (“1“) (1 ~

°-20370,>2 + 0.16215V)
quadrupoie coupling constant and quoted in units of
MHz. For eqn. (11) it is the modulus that should be for I =
7/2

taken, but it will be convenient to omit this sign in sub-

(1 + 3.63333V 7.26070,*)
sequent equations. It can of course be either positive
ri = ~ -

or negative according to the product of the signs of q

and Q, but this is not determined from the spectrum, pt = ~

d -

0..56667,1 + 1 .85952,*) (16)

since a change in sign of (e2qQ/h) merely inverts the
order of the levels and does not change the frequency. (1 01001,>a 0.01804,*)
^
5 (“/(“)
= ~ ~

For 36C1 and 37Ci, therefore, the quadrupoie coupling

constant is just twice the observed frequency, and in The important conclusions from these equations are
molecules like HC1 or CH3C1 there is good evidence firstly that the harmonic ratio of the frequencies in eqns.
from experimental methods that do give signs (e.g., (12) and (13) is no longer observed and secondly that
microwave spectroscopy) that it is negative. For from the observed frequencies both q and(e2qQ/h) can
I =
5/2 (e.g.,
12
^b, mI), we observe two frequencies be separately derived provided the theoretical equa-
tions can be fitted to the observed frequencies. In
practice, this is done by using tabulations of frequency
ratios calculated for various values of 77; such tables
have been prepared for example by Cohen (5) for I =

SA, 7A> and 9A- As an interesting example of the solu-

which are in an harmonic ratio when q 0; thus in =
tion of such a problem we may quote the I27I quadrupoie
Sil4 at 77 °K, ml quadrupoie transitions are observed at resonance frequencies of 212.6 and 340.1 MHz in BI3
198.736, 199.999 and 397.430, 399.993 MHz, appro- (the temperature at which this measurement was car-
priate to nearly zero q and quadrupoie coupling con- ried out was not specified); these are clearly not in an
stants of 1324.77 and 1333.31 MHz, almost certainly harmonic ratio, so that q must he 11011-zero, as would be
negative. The reason for two constants rather than expected if there were considerable “back-donation”
just one presumably lies in the lower point symmetry of from orbitals of it symmetry on I to the unfilled 2p„
the molecule in the crystal compared to the vapor. orbital on B. Equation (16) is fitted by the values
For I = 1 /% (e.g., 69Co), we observe three frequencies (e2qQ/h) 1176MHzand, 0.456 (or 45.6%).
= =

Before concluding this discussion of quadrupoie fre-

quencies of nuclei with half-integral spin, we refer
briefly to two methods of measuring q for nuclei of
I —

3A (e.g., 35C1,37C1,75As, 79Br, 81Br). In the first, of

higher accuracy, a single crystal is required, often at
least 1 cm3 in volume, which is placed in a uniform but
weak magnetic field H of precisely known orientation
again in an harmonic ratio; thus in cobalticinium per- with respect to the crystal axes; by weak, we mean that
chlorate, [Co(C5H5)2]+C104", three groups of fre- the interaction of the nuclear magnetic moment with the
quencies are found near 12, 24, and 36 MHz, giving a magnetic field produces splittings or shifts of magnitude
quadrupoie coupling constant of 170 MHz (probably much smaller than the pure quadrupoie resonance fre-
positive, corresponding to positive q), and a near-zero quency. One effect of the magnetic field is to remove
77, as would be expected for the sandwich structure of the ± m degeneracy noted in Figure 2, so that four lines
the cation. now appear, corresponding to the selection rule Am =

42 / Journal of Chemical Education

 +3/2~* 7K of the order of 30-60 gauss, strong signals still show
some fine structure, and this arises essentially because
those crystallites which find themselves oriented with
the maximum principal component of the electric field
3/2 gradient (along z) perpendicular to II give rise to tran-
-

7^
sitions of maximum intensity, and that only the inner or
outer transitions of the two doublets are allowed ac-
cording as to whether H is parallel to Vxx or Vvv, respec-
J3
tively. From the shape of the observed fine structure
and its dependence on H, a value for q can be estimated.
The errors are greater than those of the best single
crystal measurements, and the directions of Vxx and/or
Vm are not, of course, found. Strong signals are
needed in the first place, and in addition the method
+
implies some assumptions about line shapes in nuclear
quadrupole resonance which may not be correct.
Integral Spins
For integral spins, of which the most studied at the
present time is I4N (I 1) and to which we devote all
=

our attention, the value of q can once again bo deter-

mined from the spectrum of the powder (single crystal
studies are needed to locate the directions of the prin-
Energy levels and transitions for / weak magnetic field.
Figure 3. 3/-2 'n a
cipal components). Figure 4 shows the behavior of the
=

± 1, as we show in Figure 3. The quadruplet is sym-

metrical about the original quadrupole resonance fre- } r~7 nr
quency in zero field, and the inner pair is conveniently
referred to as the a doublet and the outer the (f doublet.
Both the positions and intensities of these doublets
1

KtIH)
vary with orientation of the crystal in the applied mag- 1
netic field. For example, when H is perpendicular to -Ift9) u2
=

V the splitting between the a and j3 doublets is cal-

2Z, I(t?)H)
culated to be

va-ve= + l (cos 24,

-

„)] (17) lo) r =

0 lb) Tj vo

Figure 4. Quadrupole energy levels and transitions for J = 1: (a) q —

0
where <J> azimuthal angle with respect to the direc-
is the (b) q 7* 0.
tion of Vxx and g is the nuclear ^-factor (a known con-
stant, for each nucleus); the dependence on 4> can be
energy levels and transition frequencies for q 0 and =

neglected when q <0.25, so that va v$ is directly pro-

—

finite 17; in the latter case, we can observe three transi-

portional to q via known constants. Hence q can be tions at frequencies of
calculated to considerable accuracy provided the direc-
tion of Vzz (and possibly VXI) can be located; in this
technique, known as the field-frequeney method (6’),
-mi-
—1 mo-i)
fields of up to 500 gauss are used, and can be measured < *>
by a Hall-effect probe or a proton magnetometer. In
another and earlier variation, the method of zero- 3 ava/, , v\
+
n
vnr A
=

splitting locus (7), the investigator looks for the locus of 4 V

zero-splitting of the a doublet; the geometrical shape of When q 0, these frequencies go to just one at 3/4-
this locus can be shown to be related to q and its orien- In 1,2,4-triazole
(e2qQ/h),
tation to the directions of V2Z and Vxx. In this method,
homogeneous fields of up to 100 gauss are used for S5C1 N—CH
work, but considerable care is needed in determining HC N'
the orientation of the crystal with respect to the mag- \ /
netic field. N
H
In the second kind of technique (5), of lower ac-
curacy, a poly crystalline sample is placed in a small we find three frequencies for N' at 3.8759, 2.4G92, and
magnetic field H (about 50 to 100 gauss) which is 1.4070 MHz at, 77°K, which from cqn. (18) lead to a
parallel to the rf field Hi, i.e., parallel to the axis of the quadrupole coupling constant of 4.2301 MHz (almost
rf coil. Since the positions of the a and /3 doublets re- certainly negative) and q 0.6651. It is however
=

ferred to in the previous paragraph depend on the orien- fairly rare to observe i>lt since quite high values of q are
tation of the crystallites with respect to the static mag- needed to bring this frequency into a sensitive range.
netic field, at first sight one might expect a uniform In most 14N work, only v2 and v3 are observed, but this is
broadening of the line and in large fields a disappearance sufficient information to enable both q and the quad-
into the noise of the spectrometer. In practice, in fields rupole coupling constant to be separately deduced.
Volume 48, Number 1, January 1971 j 43
 In concluding this section, we should mention that To derive the total electric field gradient at the origin,
quadrupole coupling constants can also be derived by we must integrate over the right-hand side of eqn. (21)
studying the quadrupole splitting of the nuclear mag- to derive
netic resonance spectra of solids or of suitable solutions
in liquid crystal media; in fact, for 2H, with its very low F„ .
(23)
quadrupole moment (0.0027965 X 10-24 cm2) and
hence low quadrupole coupling constants (viz., 56
In quantum mechanical terms, we may use the Bohr
kHz in CD3C02-), such methods are the only direct
ones that have been used in condensed states. Such probability criterion to replace pdr by ^*^dr, at the
same time separating out from the right-hand side of
techniques, however, are outside the scope of these
eqn. (23) the contribution of the other nuclei, which
articles, and the interested reader is referred to other
reviews in the literature (9). effectively behave as point charges. If we multiply
both parts of the equation by the electronic charge e
Theoretical Interpretation of Quadrupole (unsigned) and distinguish between the two contribu-
Coupling Constants tions by their different sign, we have
In this section we attempt to answer the question:
What is the significance of the quadrupole coupling
constant in theoretical chemistry?
Consider a set of Cartesian axes embedded in a mole-
cule, with the origin located at the quadrupolar nucleus
whose quadrupole coupling constant we are studying Zfi is the total charge on any other nuclei in the mole-
(Fig. 5). A volume element dr of the molecular cule distant Rt from the one under consideration. The
nuclear term thus depends only on the geometry of the
1 molecule, from which it can be calculated directly; in
more refined calculations, it may be necessary to av-
erage this factor over the vibrational modes of the mole-
cule. The factor which is related to the electronic struc-
ture of the molecule is the first, so that part of the
answer to our question is that the term q in the quad-
rupole coupling constant measures at a fixed point
within an atom or molecule the average value of the
function (3 cos2 6 —

l)/r3 over the molecular charge dis-

tribution (and over the ground vibrational state).
Knowing atomic or molecular wave functions, we should
therefore be able to calculate quadrupole constants pro-
viding that the nuclear quadrupole moment Q is known,
and in these circumstances the parameter is then a test
of such wave functions. Like the dipole moment or
molecular quadrupole moment, therefore, q depends on
the ground state wave function of the molecule. In
Figure 5, Cartesian axes in a molecular electronic charge distribution. practice, its calculation is beset by many difficulties
which arise from our lack of knowledge of reliable
charge density p (including sign),1 produces at the origin atomic or molecular wave functions, and in order to
an electrostatic potential of proceed further, it is convenient to consider atoms and
molecules separately.
V pdr
=
(19)
Quadrupole Coupling Constants in Atoms
Hence Although no nuclear quadrupole resonance measure-
dV ments have been performed on atoms, accurate values
dt Q,)<r
-

-'fry (20) for the quadrupole coupling constants have been ob-
tained from the nuclear hyperline structure in atomic
since
spectra (10) or atomic beams (10, 11), and it is essen-
dr/dz =
z/r —

cos 8 tial to consider the significance of these results before

we proceed to the study of molecules.
and
We may first turn our attention to hydrogen and hy-
d“F /3 cos2 8 \
drogen-like atomic orbitals. From eqn. (24), we re-
-
1

p{—a—r
,
=
(21)
d?
quire the value of (3 cos2 8 l)/r3 at the nucleus av-
—

Similarly one may prove that eraged over the orbital. Now s orbitals have full
dsF /3 sin2 8 cos2 0 —
1\ spherical symmetry, so that the average value of (3
=

PK-h-) cos2 8 —

1) will be zero ((cos!0) = '/Y); s electrons

therefore have zero electric field gradient. For p
and orbitals, we consider 2pz (z being the axis used to define
d2F sin2 8 sin2 <j>
—
,
(22)
did ^3 *> 1
Note the difference between this symbol and that in eqns. (4),
so that eqns. (21) and (22) satisfy eqn. (8). (6), and (7) where pn refers to the nuclear charge density.

44 / Journal of Chemical Education

8), whose normalized wave function (omitting electron answered for the majority of nuclei, is: how reliable
spin) is are our estimates of q in atoms? One might expect the
most reliable estimates to come from the use of ana-
*(2p‘] =

4^(f„)6/2,"eXp (^) C0S 9

lytical Hartree-Fock atomic wave functions. Such
wave functions have been shown to give reasonable esti-
Hence from eqn. (24) the integral we need to solve is mates of electric field gradients in atoms (12), but the
question of accuracy has not been resolved, partly be-
d2V
dz2
=
V,
mm 3 cos2 8
X cause of well-known deficiencies in Hartree-Fock wave-
functions, particularly the failure to allow for electron
correlation. Very recently, Lyons, Pu, and Das (13)
cos2 0dr (25)
exp^ and other workers have attempted to allow for such
The volume element dr in Cartesian coordinates is re- effects by the use of Brueckner-Goldstone perturbation
lated to that in spherical polar coordinates by the rela- theory. We discuss these problems further in the next
tion paragraphs.
Alternatively, one could turn to other properties of
dr = dx dy dz = r2 sin 8 dr d8 d$ (26) the atom which depend on (1/r3), and use the “mea-
sured” value of this quantity in equations similar to
Inserting eqn. (26) into eqn. (25) and separating the those of eqn. (28) and (29) derived generally for atoms
triple integral into its component parts gives with one or more valence electrons. For open-shell
atoms, the function (3 cos20 l)/?’3 is identical with that
—

V“ =

-£i (|)B/0 r exp

( -f) *•
/0
COs29(3 sin! 9 -

X in one term in the expression for the dipolar interaction

between a nuclear and electron spin, the origin of hy-
/*2tt
sin 8d8 I perfine coupling in atoms. From measured hyperfine
(27)
couplings, we can therefore derive average values of
d</>
Jo
(3cos20— l)/r3. Unfortunately the equations relating
The first integral is (Z/a0)~2; the second 8/iS; and the the two quantities also contain approximations and as-
third is just 2t. Hence sumptions, so that the “measured” values of q are also
of uncertain reliability. However, the majority of
nuclear quadrupole moments quoted in the literature
have been obtained in this way, and Ramsey (11) has
given a critical discussion of their reliability.
Unfortunately, there is still one more factor which
A similar calculation using the normalized forms for makes matters even more complex, and this is the ex-
2px and 2p„ and calculating the electric field gradient istence of inner shell shielding. As Sternheimer (14)
along Oz gives was the first to recognize, the closed inner shells of the
V„(2 Vx) =
V„(2 Vv) =
—1/iV„(2pl) (29) atom cannot be regarded as spherically-symmetrical
charge distributions unresponsive to the electrical dis-
These equations enable us to draw several conclu- tortions imposed on them by the nucleus or the valence
sions. Firstly, the electric field gradient of one electron
in a p orbital depends on Zd, so that quadrupole reso- electrons, as we have assumed in the previous para-
nance frequencies are likely to be higher for nuclei of high
graph. If we have one valence electron in a p orbital,
as we have seen we may expect an electric field gradient
atomic number. Secondly, because of the inverse de- at the nucleus which we may call U2Z°; if, however, the
pendence on r3, electron density nearest to the nucleus inner shells are simultaneously distorted by this val-
is most likely to be effective; for this reason, the value
ency electron, they will produce an additional electric
of Vn for d orbitals is considerably less than for p field gradient at the nucleus, which we set equal to
orbitals of the same principal quantum number. —

yFzz0. Hence the total electrical field gradient ex-

Thirdly, from eqns. (28) and (29), the electric field perienced by the nucleus is
gradient of a filled p shell is zero; in general, closed-
shell atoms or ions in the gaseous phase will not have F«° -

7V„°
an electric field gradient at the nucleus and will there- or
fore give no information on the nuclear quadrupole mo-
ment. The same will be true for all atoms in states of S FS2°(1 -

7) (30)

symmetry; for example, the ground state of the ni- 7 is constant and is known as the Sternheimer polar-
a
trogen atom is 4S, so that any information on the 14N ization factor of the atom. In his early papers (14),
nuclear quadrupole moment must either be derived Sternheimer showed that the mechanism we have been
from excited states of the atom or from the coupling discussing here produces an entirely equal contribution
constants in molecules. to one in which the quadrupole moment of the nucleus
The critical stage of the calculation has now been is allowed to polarize the inner shells, and the resulting
reached. Experimentally, we measure the quantity quadrupolar distortion of their charge distribution then
(e2qQ/h), which contains the product of q and Q; having allowed to interact with the electric field gradient of the
a reliable theoretical value of q for an atom then enables p electron. While the former model may be more
us to derive Q for the nucleus of that atom; we may readily used in calculations for atoms, the latter en-
then use this value to derive q in compounds of the ables us to give a better conceptual description of the
atom and compare with values calculated from trial process. Consider a spherical shell of charge sur-
wave functions. This would be the ideal procedure, but rounding a quadrupolar nucleus (Fig. 6); the linesof force
the critical question, which unfortunately is still un- emanating from the nuclear quadrupole will tend to pro-
Volume 48, Number 1, January 1971 / 45
 external to the atom or ion, as in calculations of electric
field gradients in ionic crystals, the factor 7 used in such
circumstances can be very large; for example, values
foralarge negative ion like 02~ range from lOto —430!
—

The reliability of the polarization factors is then a

major factor in the calculation, to which we return in
the subsequent articles.
Quadrupole Coupling Constants in Molecules
At the present time, we have the somewhat para-
doxical situation that whereas for atoms we have in
general good experimental data and rather poorer theo-
retical calculations, for diatomic molecules the reverse
is often true. One basis for this rather sweeping gen-
eralization is that the "best” experimental data for the
theoretical chemist comes from measurements on gases,
Sternheimer polarization of electron shell.
since his calculations can then ignore solid state effects
Figure 6-
which are bound to arise in nuclear quadrupole reso-
duce two rather different kinds of distortion in the closed nance spectroscopy (and to which we return later) and
,

shell. In the first, the inner shell suffers a radial dis- the most accurate measurements in gases tend to be for
tortion by a repulsion of charge at A ,B and attraction at atoms. Another is the remarkable success of molec-
C,D; if Q > 0, from the point of view of a distant (i.e., ular orbital calculations based on self-consistent field
valence) electron, the quadrupole interaction with the methods and using highly extended basis sets of atomic
nucleus is reinforced, corresponding to antishielding orbitals in predicting molecular properties, particularly
(i.e., 7 is —ve). In the second, the inner shell suffers an for diatomic molecules. For example, such a calcula-
angular distortion in that charge at A and B tends to tion of the 36C1 quadrupole coupling constant in HC1 by
move towards C and D; if Q > 0, this effect reduces the Cade (15) has given the value —67.9 MHz, compared
quadrupolar interaction, corresponding to shielding with an experimental value (from microwave spectra
(i.e., 7 > 0 but less than one). By means of the angular in the gas) of —67.5 MHz. Even more accurate pre-
distortions, for example, an s orbital in the inner shell dictions of deuterium quadrupole constants can be
acquires a little of the appearance of a d orbital, i.e., made; such a calculation for HD (16), the coupling con-
ns —* d; on the other hand, a p orbital in the inner shell stant of which has been measured as 227 ± 4 kHz by
acquires "p-character” by virtue of the radial distor- molecular beam techniques (11), has enabled some of
tion (viz., np —* p) and “/-character” by virtue of the the first really reliable estimates of a nuclear quad-
angular distortion (viz., np —* /). These distortions are rupole moment to be made, namely, 0.0027965 X 10~24
known as “excitations,” and each occupied orbital in the cm2 for 2H. It is somewhat ironic that this should
atom will contribute by means of these excitations to come from calculations on a molecule rather than on an
the total electric field gradient, the actual value of y atom!
being a summed average over the electron distribution In view of the present rather gloomy picture of cal-
in each occupied orbital. It will be clear that for culations of the quadrupole coupling constants of
atoms other than H equations like cqn. (28) will need to atoms, one might ask why it is that calculations for
be multiplied by the factor 1 y before a reliable esti-
—*
molecules are apparently more successful. For HD,
mate of the true electric field gradient is obtained. In the simple answer is that present molecular wave func-
atoms, it is more usual to use the symbol Rq or R for 7. tions are so good—'and of course there is no inner shell
Now 1 Rq can be calculated from atomic wave func-
—

polarization. Otherwise the best results have been ob-

tions, although the majority of calculations hitherto in- tained for diatomic hydrides of first and second-row
volve approximations. Fortunately, in most open- elements and homonuclear diatomic molecules of the
shell atoms, Rq is not large, e.g., in Cl, 1
1
—

Rq is esti-
—

first row elements and some of their simpler polyatomic

mated to be 0.888 and in B, 0.857, corresponding to species; the present most reliable values of the quad-
shielding. Recently, some attempt has been made to rupole moment of 14N (+0.015 and +0.0166 X 10~24)
take many-body interactions into account in such cal- have come by combining theoretical calculations and ex-
culations by the use of Brueckner-Goldstone perturba- perimental results for molecular N2 and the molecules
tion theory (see, e.g., ref. (13)): for example, such a HCN, FCN, C1CN and CH2CN. Calculations of the
calculation for 7Li (Tb/,) has yielded a value for (1 Rq) —
1T0 quadrupole coupling constant in H20 or 14N in NFI3
of 0.8300, compared to a one-electron value of 0.8844; are extremely encouraging but not at present so reliable,
the new quadrupole moment of 7Li is then calculated to although the 2II quadrupole coupling constants of the
be —(3.9 ± 2.6) X 10~26 cm2 (the main uncertainty deuterated molecule are often well-reproduced. But
being probably in the experimental value of the quad- what about the Sternheimer polarization factor for the
rupole coupling constant!) in reasonable agreement with heavier atoms? In self-consistent-field calculations of
a value of —4,4 X 10~26 from molecular beam measure- the kind we are discussing, the basis set includes all
ments on BiH. Unfortunately, such calculations have orbitals of the two atoms, so that what we call the
been performed for very few atoms, so that estimates "core” orbitals are already heavily-distorted in the
of the polarization factor still represent yet another final electron distribution. There is therefore some
source of uncertainty in the derivation of nuclear quad- reason for supposing that polarization effects have been
rupole moments from atomic spectra. taken into account, and that the Sternheimer effect can
When the distortions are produced by charges wholly be ignored. The test will come when we have calcula-

46 / Journal of Chemical Education

tions of similar caliber for a set of molecules with a
a which will be written as a linear .combination of n or-
common quadrupolar nucleus, e.g., the series H Cl, thogonal atomic orbitals, fa. The jth member of this set
F-Cl, Cl-Ci, which could give us three independent we write as
estimates of the 35C1 nuclear quadrupole moment. In fa =
Cljfa + Cljfa + Czjfa + . . .
+ Cnjfa
this connection, the value of 170 quadrupole coupling
constants cannot be overestimated. In the meanwhile, C :;0>
=
(31)
the increasing interest of theoretical chemists in quad-
rupole coupling constants as an important parameter to and in general inner-shell orbitals are not included.
be derived from their calculations is certain in the
When this equation is substituted into eqn. (24), and
future to exert a considerable and beneficial influence on
the first term summed over all occupied orbitals, it is
the subject.
clear that three different kinds of basic integral will
We conclude this section with a list (see Table 1) of
arise. To clarify this differentiation, let us imagine
the “best-available” values of the quadrupole moments
of nuclei discussed in these articles.
that we are interested in the quadrupole coupling con-
stant of the nucleus of atom 1, so that [(3 cos2 6 l)/r3]i —

is to be evaluated at nucleus 1 (hence the subscript).

Table 1. Some Nuclear Quadrupole Moments The first kind of integral will be of the form
and Quadrupole Moment Ratios
Nu-
clear
Estimated
nuclear
CijCtj I fai
y-r3-fadr
Natural spin quadrupole
abundance quantum moment2 Isotopic1 and is a three-center integral, since atoms 1, 2, 3 are in-
Nucleus (%) no. (I) Q (X 10“om‘) ratio volved. The two-center integrals will therefore be of the
2H 1.56 X 10“* 1 + 0.0027965 form
“Li 7.43 1 -0.000741 )
>0.019
7Li 92.57 3A -0.039 j
+ 0.029 ^fadr
—

9Be 100 3A
“B 18.83 3 + 0.074 1
>2.084
”B 81.17 3A + 0.036 j
14N 99.635 i +0.0166 and the one-center integrals (generally the easiest to
17Q 3.7 X 10-2 6A -0.0301 evaluate)
27
A1 100 EA +0.151
«C1 75.4 3A -0.0802 1
>1,26878
37C1 24.6 Vs -0.0632 J

“Mil 100 "A + 0,355

“Co 100 TA + 0.40440
63Cu 69.09 3A
—

0,163 Itrl ()QQR

.U&UO
Note that the quadrupole coupling operator is a one-
“Cu 30.91 Vs -0,153 electron operator. Now the two most used approximate
7BAs 100 Vs + 0.32 methods in the literature differ in the way in which these
79Br 50.57 V 2 + 0.332
81Br 49.43 SA + 0.282 i [l.19707 various integrals are treated. In the method used by
I13In 4.16 9A + 1.145 k). 986365 Cotton and Harris {17), all three-center integrals are
1
lsIn 95.84 9
A + 1.165 J

121Sb 57.25 Vs -0.5310 k). 78447 neglected, some of the two-center integrals are allowed
123Sb 42.75 7
A -0,6810 j to cancel the nuclear term [the second in eqn. (24) ], the
127J 100 6A -0.785 remainder are approximated by a relationship involving
I3BBa 6.59 Vs + 0,182
13,Ba 11.32 3A + 0.283 j|1.534 the overlap integral Sik, and finally the Sternheimer
>8BRe 37.07 Vs + 2.8 U .056 polarization factor is assumed to vary negligibly from
18,Re 62.93 EA + 2.6 j
molecule to molecule. The equation they derive is that
197
Au 100 8A + 0.606
aosBi 100 9
A -0.4 the quadrupole coupling constant in the free molecule
2
No Sternheimer polarization factor has been applied, except {e^qQ/h) moi is equal to
for !H and possibly for 7Li and 14N.
6
This ratio is usually known to a considerably higher accuracy
(32)
than the individual nuclear quadrupole moments. V h A

Approximate Methods of Deriving Quadrupole Coupling where (e^Q/h)*^ contains the term cos2(?
Constants in Molecules
1)/?,3)(f>(dT in the atom and
—

Wavefunctions expressed in terms of extended basis

sets such as we have been discussing in the previous
section, simply do not exist at the present time for the
V- =
+ Cu
k>i
y }
(33)

vast majority of polyatomic molecules. It is natural,

being an expression for the occupation (including
therefore, that chemists should turn to more approxi-
mate methods of predicting quadrupole coupling con- overlap) of orbital <tn- If overlap is neglected, fi be-
stants in terms of such familiar parameters as orbital
comes simply
jv
populations, bond hybridization, and ionic character.
While the success of these methods has been rather lim- Y
ited, they are used in many discussions of quadrupole
coupling constants, and they will therefore be briefly the orbital population of <£, . It should be stressed
discussed in this section. however that many of the approximations involved in
Wc start from eqn. (24) and imagine to begin with deriving these equations are drastic ones. For com-
that our molecule will be described by a set of N plexes of the type MCI/1- (e.g., PtCh2-, PtCU2-), for
doubly-occupied orthogonal molecular orbitals, each of which the theory was originally designed, it has the con-

Folume 48, Number 1, January 1971 j 47

venient feature that it requires precisely those quanti- now a straightforward procedure to apply eqn. (38) to
ties that are derived in present molecular orbital cal- various molecules containing M-Cl bonds. Thus in
culations, e.g., the eigenfunctions and overlap integrals, molecular chlorine, 6=1 (neglecting overlap), n 2, =

but the value of such predictions of the quadrupole cou- and px py 2, so that the ratio comes to (1
= =
a2). —

pling constant will also depend on the approximations Since the quadrupole coupling constant in solid chlorine
made in deriving these eigenfunctions. (108.9 MHz) is almost the same as in the atom, this ap-
The second theory, much older, and one which has pears to predict that a 0, which is not unreasonable;
~

shaped the subject for over twenty years, is that due to unfortunately this simple picture also neglects the pos-
Townes and Dailey (18). They use what is essentially sibility of d orbital participation in the bonding. In
a localized-orbital approach, so that for a terminal nu- R-Cl, with no n bonding, px py 2, and if we make —
—

cleus, like 35C1 in R-Cl, we consider only orbitals on Cl n = 2

and the atom to which it is bonded. Three-center inte-
(e2qQ /h)mo\
grals are therefore zero, all two center integrals are set (1 «2)(2 b) (39)
- - -

(o2q0 /h )atom
against the nuclear contribution of the bonded atom,
and the Sternheimer polarization factor is again as- 6 will vary according to the relative electronegativity
sumed to vary negligibly from molecule to molecule (al- of the bonding atom with respect to that of Cl; thus on
though a small correction is sometimes applied when a valence bond picture, it can be related to the ionic
the Cl atom is believed to carry a small positive charge). character by expressing the wavefunction of the bonding
Everything therefore depends on how the Cl 3p orbitals orbital as
are occupied, i.e., on the 'ip orbital populations, a con-
clusion which is consistent with the equations of Cotton H =
(1 -

i)V^(R-Cl) + f7a*(R+Cl-)
and Harris when overlap is neglected. If moreover the assuming only ^(R+C1~) to make a significant con-
direction of the principal axes are known, and the same tribution; since in R +C1- the chlorine ion has spherical
axial system can be used to specify the p orbitals, a symmetry, its contribution to the electric field gradient
particularly simple situation arises. In R-Cl, for ex- (and hence its quadrupole coupling constant) will be
ample, qzz lies along the chlorine bond—or very nearly zero, so that (2 6) can be replaced by the factor
—

so—and we therefore take this as the z direction of our (1 i) and the ratio in eqn. (39) becomes simply
—

coordinate system. Taking the chlorine bonding

orbital \pi as a'3s-3p hybrid of s character a
(1 -

«»)(! -

i) (40)
When both 3px and 3pv participate equally in ir bonding
=
a^(3s) + (1 —
a2)l/ty(3 px) (34)
to R, the populations px and pv remain equal but fall
of population 6, the non-bonding orbital below 2; calling them 2 ir (ir being the ^-character of —

fi-i (1 a2)V^(3S) at(3p,)

= - -

(35)
the bond), on a valence bond picture we may use the re-
vised table of contributions to give
of population n, and the r orbitals
(e2qQ/h)moi
=
fi(3 px) (36)
=
(1 -

«2)(1 -

i) (41)
(c‘qQ/h )ntom
=
^(3pu)
neglecting the product ira2. When only one of the 3p
of populations px and py, we can set up the following orbitals, say 3pX) participates in the ir bonding (as ex-
table of one-center contributions neglecting overlap con- pected in chlorobenzene), the bond to chlorine ceases to
tributions and using eqn. (29) be axially symmetric and a similar analysis of the field
Total contribution gradients Vzx and Vyv shows that Vxx > Vvv and that
Electric field Orbital to electric field 3 (e2qQ/h)atom
Orbital gradient along z population gradient v
_
~

2* (42)
(e2qQ/h)moi
fa (1
-

o2)qa b 6(1 a2)q0

—

fa ot2qo n na2q0 Even if we exclude d orbital participation in the

fa “V*. Px Vx
bonding, these equations contain two experimental
—2q°
-

w Py quantities (e2qQ/h and ij) and at least three parameters

(a, i, 7r), with an additional factor if polarization
1

effects are to be allowed for. Some other experimental

q0 is the value of the integral
datum at least is required, and perhaps this feature
JV*[(3cosa0 —

l)/r5]^dT more than any other has been a serious defect in the

for a 3 pz orbital. The total electric field gradient is Townes-Dailey theory. Nevertheless, it has found
therefore widespread application in the interpretation of nuclear
quadrupole coupling constants, some of which will be
q =
+ ba1 ncd) -f l/i{px + py)]
—

?<,[(
—
b (37) —

considered in the subsequent articles in this series. Eor

an account of the Townes-Dailey theory, the interested
If we perform a similar calculation for atomic chlorine
reader is referred to a recent book by Lucken (19).
(in which the orbital populations are either 1 or 2), we
find that q in the atom is equal to —q0; hence multi- Literature Cited
plying both sides of eqn. (37) by (e2Q/h), we finally (1) Raich, J. C., and Good, R. H., Jh., Amer. J. Phys., 30, 356 (1963).
derive (2) Slichter, C. P., “Principles of Nuclear Magnetism,” Harper and Row,
New York, 1963; Chap. 6.
(e2gQ/fe)moi (e2qQ/h)moi
_

—109.7 ex2) -f- na2 + (3) Nye, J. F., “Physical Properties of Crystals,” Oxford Univ. Press,
(e2qQ/h),. tom Oxford, 1957; Chap. 1.
(4) Das, T. P., and Haiin, E. L., "Nuclear Quadrupole Resonance Spec-
VdPx + Py) (38) troscopy,” Academic Press, New York and London, 1958.
(5) Cohen, M. Ii., Phys. Rev., 96, 1278 (1954).
where (e2qQ/h),,u,m is measured as —109.7 MHz. It is (6) Rehn, V.,./. Chem. Phys., 38,749 (1963).

48 / Journal of Chemical Education