CHAPTER 5 HYPERFINE (A) ANISOTROPY 5.1 INTRODUCTION In many oriented systems there may be an anisotropy in the hyperfine splittings A as well as in g. Thus, not only does each hyperfine multiplet move as a unit when the orientation is changed, but simultaneously the spacing between its component lines changes. When the hyperfine anisotropy is sufficiently great, then the qualitative appearance of the spectrum is drastically changed by rotation of a single crystal through even a relatively small angle. We temporarily ignore simultaneous changes in A and in g until we reach Section 5.4. We also restrict ourselves to elec- tron spin S ¼ 1 2 and, for the most part, to consideration of hyperfine effects arising from a single nucleus. A very simple example of a strongly anisotropic hyperfine interaction is that of the V OH center [1,2] shown in Fig. 5.1, for which g is almost isotropic. This center in MgO consists of a linear defect 2 OAHO 2 in which a cation vacancy A separates a paramagnetic O 2 ion and the proton of a hydroxide impurity ion (by 0.32 nm). If the crystal is rotated in a (100) plane, taking u as the angle between the defect axis and the field B, the hydrogen hyperfine coupling A(u) is given by an expression of the form A ¼ A 0 þ (3 cos 2 u 1)dA (5:1) 118 Electron Paramagnetic Resonance, Second Edition, by John A. Weil and James R. Bolton Copyright # 2007 John Wiley & Sons, Inc.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
CHAPTER 5
HYPERFINE (A) ANISOTROPY
5.1 INTRODUCTION
In many oriented systems there may be an anisotropy in the hyperfine splittings A as
well as in g. Thus, not only does each hyperfine multiplet move as a unit when the
orientation is changed, but simultaneously the spacing between its component lines
changes. When the hyperfine anisotropy is sufficiently great, then the qualitative
appearance of the spectrum is drastically changed by rotation of a single crystal
through even a relatively small angle. We temporarily ignore simultaneous
changes in A and in g until we reach Section 5.4. We also restrict ourselves to elec-
tron spin S ¼ 12
and, for the most part, to consideration of hyperfine effects arising
from a single nucleus.
A very simple example of a strongly anisotropic hyperfine interaction is that of
the VOH center [1,2] shown in Fig. 5.1, for which g is almost isotropic. This
center in MgO consists of a linear defect 2OAHO2 in which a cation vacancy A
separates a paramagnetic O2 ion and the proton of a hydroxide impurity ion (by
�0.32 nm). If the crystal is rotated in a (100) plane, taking u as the angle
between the defect axis and the field B, the hydrogen hyperfine coupling A(u) is
given by an expression of the form
A ¼ A0 þ (3 cos2 u� 1)dA (5:1)
118
Electron Paramagnetic Resonance, Second Edition, by John A. Weil and James R. BoltonCopyright # 2007 John Wiley & Sons, Inc.
Specifically, consistent with Eq. 2.2, it was found experimentally that
A=gebe ¼ 0:0016þ 0:08475(3 cos2 u� 1) mT (5:2)
which ranges from 0.1711 mT for u ¼ 08, becoming zero when cos2
u ¼ (1 2 0.0016/0.08475)/3, to 20.08315 mT for u ¼ 908. The doublet splitting
(Fig. 5.1) is sufficiently small that it equals the magnitude of A/gebe, with no higher-
order terms needed (at 9–10 GHz). We see from Eq. 5.2 that, for this center, the
proton hyperfine splitting happens to be almost purely anisotropic. In most
systems, the isotropic contribution A0 is in fact of the same order of magnitude as
FIGURE 5.1 X-band EPR spectra of the VOH center in MgO. These spectra show almost
purely anisotropic hyperfine splitting. Lines arising from other (related) defects have been
masked. (a) Structure of the defect. The symmetry axis of the defect (tetragonal crystal
axis) is labeled Z. (b) Line components for B perpendicular to Z. (c) Line components for
B parallel to Z.
5.1 INTRODUCTION 119
dA. Then Eq. 5.1 is not applicable, and more complicated expressions are required
(Section 5.3.2).
To analyze anisotropic hyperfine effects properly, one must embark on detailed
consideration of the 3 � 3 hyperfine coupling matrix A, which describes the phys-
ical aspects phenomenologically. We shall see that this is not a trivial matter.
However, eventual attainment of parameter matrix A from a set of EPR measure-
ments yields a rich harvest, revealing much detail about the local geometric con-
figuration of a paramagnetic center and about the distribution of the nuclei and
unpaired electron(s) in it. In fact, it is primarily these hyperfine effects that cause
EPR spectroscopy to be such a rewarding structural tool.
5.2 ORIGIN OF THE ANISOTROPIC PART OFTHE HYPERFINE INTERACTION
The origin of the isotropic hyperfine interaction was discussed in Chapter 2. Inter-
action between an electron and a nuclear dipole some distance away was rejected
there as a source of the splittings observed in a liquid of low viscosity, since this
interaction is time-averaged to zero. However, in more rigid systems, it is precisely
this dipolar interaction that gives rise to the observed anisotropic component of
hyperfine coupling. The classical expression for the dipolar interaction energy
between an electron and nucleus separated by a distance r can be shown [3–5]
to be
Udipolar(r) ¼m0
4p
meT�mn
r3�
3(meT� r)(mT
n� r)
r5
� �(5:3)
Here r represents the vector joining the unpaired electron and a nucleus (Fig. 2.2).
Vectors me and mn are the classical electron- and nuclear-magnetic moments. For
both, mT . r ¼ r T . m. Superscript ‘T’ indicates the transpose (Section A.4). We see
that the energy of magnetic interaction between the spins varies as r23, and is inde-
pendent of the sign of r. Note that the dipolar interaction exists whether or not
there is an externally applied field.
For a quantum-mechanical system, the magnetic moments in Eq. 5.3 must be
replaced by their corresponding operators. For the sake of simplicity, we shall
here ignore the g anisotropy in the magnetic moment (Eq. 4.8). Thus both g and
gn are taken to be isotropic. The hamiltonian (using Eq. 1.9 for m in operator
form) thus is
Hdipolar(r) ¼ �m0
4pgbegnbn
ST� Ir3�
3(ST� r)(IT� r)
r5
" #(5:4)
That Hdipolar(r) describes an anisotropic interaction can be seen by expanding the
vectors in Eq. 5.4, yielding
120 HYPERFINE (A) ANISOTROPY
Hdipolar(r) ¼�m0
4pgbegnbn
r2 � 3x2
r5SxIx þ
r2 � 3y2
r5SyIy
�þ
r2 � 3z2
r5SzIz �
3xy
r5(SxIy þ SyIx)�
3xz
r5(SxIz þ SzIx)�
3yz
r5(SyIz þ SzIy)
�(5:5)
Coordinates x, y, z of the electron are taken with respect to axes fixed in the sample
(e.g., a crystal). The point nucleus is placed at the origin. Note that, as is discussed in
Section 5.3.2.3, the nucleus may be at the center of the electron distribution, or
removed from it.
On averaging the hamiltonian (Eq. 5.5) over the electron distribution (i.e., inte-
grating over the spatial variables), this becomes a spin hamiltonian, having the form
Hdipolar(r) ¼ �m0
4pgbegnbn�
Sx Sy Sz
� ��
hr2�3x2
r5i h�3xy
r5i h�3xz
r5i
hr2�3y2
r5i h�3yz
r5i
hr2�3z2
r5i
26666664
37777775�
Ix
Iy
Iz
26666664
37777775
(5:6a)
¼ ST�T�I (5:6b)
The angular brackets imply that the average over the electron spatial distribution has
been performed. Note that the dependence on electron-nuclear distance is r23 in all
elements, and that the average depends on which orbital the unpaired electron is in.
Note also that matrix T is symmetric about its main diagonal and is traceless.
The full spin hamiltonian requires the addition of the isotropic hyperfine term
A0ST . I, that is, the contact interaction (Eq. 2.39b) as well as the electron
Zeeman and nuclear Zeeman terms. Thus1
H ¼ gbeBT� Sþ ST�A� I� gnbnBT� I (5:7)
where the hyperfine parameter 3 � 3 matrix is
A ¼ A013 þ T (5:8)
Here A0 is the isotropic hyperfine coupling, and 13 is the 3 � 3 unit matrix. It is
useful to note that A0 is just tr(A)/3.
5.2 ORIGIN OF THE ANISOTROPIC PART OF THE HYPERFINE INTERACTION 121
When the crystal is rotated, that is, the unit vector n along B is changed, the value
of nT . A . n changes. In fact, from a set of such numbers [using the same procedure
as for the determination of matrix gg given in Section 4.4 (see also Eq. A.52b)], one
can arrive at the 3 � 3 symmetric hyperfine matrix A sym ; (AþAT)/2, to within a
factor of +1.2 This matrix (together with the matrix g and gn) contains all the infor-
mation needed to reproduce the EPR spectral positions and relative peak intensities
at all frequencies and crystal orientations.
Because magnetic-field units are often convenient, we have already defined (in
Eq. 2.48) the symbol a0 ¼ A0/gebe for the isotropic part of A. We now define
two analogous parameters useful when hyperfine anisotropy occurs, and which
are derivable from matrix T. Thus we have
a0 ¼ (A1 þ A2 þ A3)=3gebe (5:9a)
b0 ¼ ½A1 � (A2 þ A3)=2�=3gebe (5:9b)
c0 ¼ (jA2j � jA3j)=2gebe (5:9c)
Here Ai (i ¼ 1, 2, 3) denotes the principal values of A, ordered such that jA1j2 jA2j
and jA1j2 jA3j are larger than or equal to jA2j2 jA3j, thus selecting which pa-
rameter is A1, and (arbitrarily) taking jA2j2 jA3j to be non-negative. These new
hyperfine parameters, called the uniaxiality parameter b0 and the asymmetry (rhom-
bicity) parameter c0, are independent of a0 and vanish if there is no anisotropy. Note
that, because of the invariance of tr (A) to change of coordinate system, a0 (but not b0
and c0) is available from A without diagonalizing it. In many instances, as we shall
see, these parameters exhibit an intimate relationship to the fundamental quantum-
mechanical properties of individual atoms.
It is useful to realize that measurement of matrix A can yield the relative signs of
parameters a0 and b0. Often, when the relatively simple dipole-dipole model yields a
value of (b0)theor that is close in magnitude to that of (b0)expt , the actual sign of a0 can
be derived by assuming that the sign of (b0)expt is given by theory (Problem 5.11). The
sign of a0 may not be the one predicted by Eq. 2.38, that is, by the sign of gn (using
Table H.4), due to the core-polarization effect [9]. This features unpairing of inner-shell
electrons, often with inner s-type electron spins with a net polarization in the direction
opposite to that of the total spin population on the atom (Sections 9.2.4 and 9.2.5).
5.3 DETERMINATION AND INTERPRETATION OFTHE HYPERFINE MATRIX
5.3.1 The Anisotropic Breit-Rabi Case
In some instances, the hyperfine energy is not small compared to the electron
Zeeman energy, so that neither term in the spin hamiltonian (Eq. 5.7) can be
treated approximately. The result is the appearance of higher-order energy terms
(Section 3.6), leading to unequal spacings between the hyperfine components
122 HYPERFINE (A) ANISOTROPY
observed in the field-swept EPR spectra (e.g., see Fig. 5.2). Here, then, the general
Breit-Rabi type of approach (Appendix C) must be applied.
In practice, analytic mathematical solutions for the anisotropic Breit-Rabi
problem are not available. However, accurate numerical solutions (by computer)
are not difficult and yield all magnetic-resonance line positions as well as the relative
intensities. Let us now briefly consider another approach, in which anisotropy is
brought in as a perturbation on the isotropic hyperfine situation.
It can be shown that Relation C.26 for an isotropic S ¼ 12
situation can be modified
[10,11] to become
B�jMI j � BjMI j
2jMI j¼
nT�Asym� ngbe
1�nT�Asym� n
2hn
� �2(5:10)
where, as usual, g ¼ (nT . g . gT . n)1/2 (Eq. 4.12). The set of these relations yields
the elements of A sym ¼ (AþAT)/2 directly [when g is known, say, from an
even-even isotope (I ¼ 0) central spectrum]. We note that at each field orientation,
I (if I is an integer) or (2Iþ 1)/2 (if I is half of an odd integer) such quantities are
measurable, all (nominally) giving the same value. Equation 5.10 is valid when the
isotropic component jA0j is large compared to the hyperfine anisotropy.
FIGURE 5.2 Computer simulation of the 10.0000 GHz field-swept EPR spectrum of a
Ge3þ (S ¼ 12) center (denoted by [GeO4/Na]A
0) in crystalline a-quartz, obtained for
spin-hamiltonian parameters determined at 77 K. The spectrum extends from 205.0 to
505.0 mT. The central line arises from even-isotope species (I ¼ 0) of germanium, whereas
the 10-line hyperfine multiplet arises from 73Ge (I ¼ 92). The spectrum was calculated with
Bkz (¼optic axis c) and B1kx (¼electrical axis a) (simulated by M. J. Mombourquette and
J. A. Weil). The four-line 23Na superhyperfine structure is too small to be seen at the field
scale used.
5.3 DETERMINATION AND INTERPRETATION OF THE HYPERFINE MATRIX 123
For example, consider the analysis of the anisotropic splittings caused by the low-
abundance isotope 73Ge (I ¼ 9/2) in a Ge3þ center (S ¼ 1=2) found in crystalline
SiO2. The 10 hyperfine lines (Fig. 5.2) can be grouped in pairs (MI, 2MI), yielding
5 values for nT . A sym. n. These can be averaged. Equation 5.10 thus gives this
single number for a given n, despite the unequal hyperfine spacings.
Similarly, for hydrogen atoms trapped at low temperatures within cavities in
quartz crystals, the local electric fields cause anisotropy in A (i.e., admixture of
p, d, . . . orbitals into the nominal ground state) and in g [12]. The large isotropic
part of the hyperfine splitting constant makes it important (say, for 10 GHz EPR)
to use the Breit-Rabi formalism described above.
As implied, use of Eq. 5.10, together with fields and g factors measured at various
orientations of the crystal with respect to B, cannot yield A or AT. Rather, the
relation is valid in the approximation
A�AT � A0
2A11 � A0 A12 þ A21 A13 þ A31
2A22 � A0 A23 þ A32
2A33 � A0
264
375 (5:11a)
¼ A0(2Asym � A0 13) (5:11b)
¼ A0(Aþ AT � A0 13) (5:11c)
to the ‘square’ of A. Here A0 is tr (AþAT)/6; that is, it is the isotropic com-
ponent of A (and of AT). The magnitude of A0 must be large compared to
the anisotropic part for Eqs. 5.11 to hold. This analysis on its own does not
yield the sign of A0.
When the magnetic field B used is high enough that the higher-order effects
referred to above are negligible (this is usually the case), then we can turn to the
less general theory described in the following section.
5.3.2 The Case of Dominant Electron Zeeman Energy
Often, as pertains at sufficiently high magnetic fields, the electron Zeeman term
in Eq. 5.7 can be assumed to be the dominant energy term; that is, the electron
magnetic-moment alignment is much less affected by the nuclear magnetic
moment than by B. This allows one conveniently to quantize S along B; that
is, MS describes the eigenvalues of S projected along B. Furthermore, higher-
order hyperfine contributions (Sections 2.1, 3.6 and 5.3.1) can be taken to be
negligible. By inserting the electron-spin eigenvalue vector MS n for S in
Eq. 5.7, we obtain
H ¼ gbeBMS13 þ (MSnT�A� gnbnBnT) � I (5:12a)
; gbeBMS13 � gnbnBTeff� I (5:12b)
124 HYPERFINE (A) ANISOTROPY
As before, n is the unit (column) vector in the direction of B. Here we have
defined an effective magnetic field
Beff ¼ Bþ B hf (5:13)
acting on the nuclear magnetic moment, where
B hf ¼ �MS
gnbn
AT� n (5:14a)
and thus
B hfT ¼ �
MS
gnbn
nT�A (5:14b)
Vector Bhf represents the contribution to the magnetic field at the nucleus
arising from the electron-spin magnetic moment. We note that Beff is not
necessarily parallel to B (Fig. 5.3), and depends on MS. Thus the axis of quan-
tization changes during an EPR transition so that MI changes its meaning [13].
This is a generalization and correction to the erroneous view, generally held,
that MI is unchanged during a ‘pure’ electron spin flip. The magnitude of the
effective field is given by
Beff ¼ jBeff j ¼ ½(Bþ Bhf)T� (Bþ Bhf)�
1=2 (5:15a)
¼ B2 � 2MS
gnbn
(nT�A � n)Bþ1
(2 gnbn)2nT�A�AT� n
� �1=2
(5:15b)
We note from Eqs. 5.14 that the projection of the hyperfine field along
B is proportional to n T . A . n and the magnitude of the hyperfine field, to
FIGURE 5.3 Vector addition of the external field B and of the hyperfine field B hf for
S ¼ I ¼ 12. The superscripts a and b refer to MS ¼ þ
12
and MS ¼ �12, respectively. We note
that B hfa ¼ 2B hf
b. The 3 cases depicted above are: (a) B� Bhf; (b) B � Bhf; (c) B� Bhf.
5.3 DETERMINATION AND INTERPRETATION OF THE HYPERFINE MATRIX 125
[n T . A . A T . n]1/2 (Eqs. 5.15). The field magnitude Bhf can be very large; for
example, if the proton hyperfine coupling is �3 mT (a typical value), then
Bhf ffi 1 T. Remember that Bhf is the hyperfine field at the nucleus and not at
the electron. The latter would be only �2 mT in this case.
As is evident from Eq. 5.12b, it is most natural to quantize I along Beff. However,
this often is inconvenient, and hence various types of approximations are made,
depending on the physical circumstances. Several cases (Fig. 5.3) are considered
herein.
5.3.2.1 General Case In the general case [13,14], one finds the occurrence of
satellite lines. As an example, we deal with the S ¼ 12
system but leave I unspecified.
Referring to Fig. 5.3b, we consider the total resultant field Beff at the nucleus. Vector
I is quantized along Beffa for MS ¼ þ
12
and along Beffb for MS ¼ �
12. The energies
resulting from Eq. 5.12b, for I ¼ 12, are given by
Ua(e)aa(n) ¼ þ12
gbeB � 12
gnbnBeffa (5:16a)
Ua(e)ba(n) ¼ þ12
gbeB þ 12
gnbnBeffa (5:16b)
Ub(e)ab(n) ¼ �12
gbeB � 12
gnbnBeffb
(5:16c)
Ub(e)bb(n) ¼ �12
gbeB þ 12
gnbnBeffb
(5:16d)
The nuclear-spin eigenfunctions are not the same for MS ¼ þ12
and � 12, since the
axis of quantization for I is different in the two cases; here, as with Beff, the super-
scripts indicate the electron-spin state. By expressing jaa(n)i and jba(n)i as linear
combinations of jab(n)i and jbb(n)i, we can show that the relation between the
nuclear-spin states is
jaa(n)i ¼ cosv
2jab(n)i � sin
v
2jbb(n)i (5:17a)
jba(n)i ¼ sinv
2jab(n)i þ cos
v
2jbb(n)i (5:17b)
where v is the angle between Beffa and B eff
b (Sections A.5.2, A.5.5 and C.1.4).
The energy levels are given in Fig. 5.4 (also see Fig. C.2). The four possible EPR
transition energies are
DUa ¼ Ua(e)ba(n) � Ub(e)ab(n) ¼ gbeBþ 12
gnbn{Beffa þ Beff
b} (5:18a)
DUb ¼ Ua(e)ba(n) � Ub(e)bb(n) ¼ gbeBþ 12
gnbn{Beffa � Beff
b} (5:18b)
DUc ¼ Ua(e)aa(n) � Ub(e)ab(n) ¼ gbeB� 12
gnbn{Beffa � Beff
b} (5:18c)
DUd ¼ Ua(e)aa(n) � Ub(e)bb(n) ¼ gbeB� 12
gnbn{Beffa þ Beff
b} (5:18d)
126 HYPERFINE (A) ANISOTROPY
Since the intensities of the lines are proportional (Section C.1.4) to
|hMS0, MI
0jB1T . (gbeS 2 gnbnI)jMS, MI lj2, the relative intensities of the lines are
given by
Ia ¼ Id / sin2 v
2(5:19a)
Ib ¼ Ic / cos2 v
2(5:19b)
Thus all four transitions can be of comparable intensity (Fig. 5.4; herev � 708). Failure
to recognize this has led to incorrect assignments of hyperfine splittings. The treatment
shown above is still rather general, although neglecting higher-order terms (Section
5.3.1). It is instructive now to examine the preceding results for two limiting cases:
The Case of B� Bhf : Here v � 1808, and hence transitions a and d are the
strong ones. We see that gebe times the separation of these two lines is very
nearly the hyperfine energy, that is, jDUa 2 DUdj � 2jgnjbnBhf, where field
Bhf ¼ [nT . A . AT . n/4gn2 bn
2]1/2.
The Case of B� Bhf : Here v � 0 and hence transitions b and c of Fig. 5.4b
are strong. Now, gebe times the separation of these two lines is given by
FIGURE 5.4 (a) Energy levels at constant field for a system with S ¼ I ¼ 12
(g n . 0) when
B is close to Bhf (Fig. 5.3b), but with Beffa , Beff
b. Here a and d are the normally allowed
transitions; b and c are usually of much lower intensity. (b) Observed EPR lines at constant
(X-band) frequency, with relative intensities derived from Eqs. 5.19.
5.3 DETERMINATION AND INTERPRETATION OF THE HYPERFINE MATRIX 127
jDUc 2 DUbj � 2jgnjbnBhf0, where Bhf
0 ¼ jnT . A . n/(2gnbn)j. Note that this
result is consistent with Eq. 5.10 (see also Eq. 2.1), since hn � gbeB for sufficiently
large B.
We now turn to analysis of the anisotropy effects in these two limiting cases.
5.3.2.2 The Case of B � Bhf This case is the one most commonly encoun-
tered and thus is analyzed in detail.3 As before, in general B (taken along z) and
Bhf (�Beff) are not in the same direction (Fig. 5.3a). Thus S and I again are best
quantized along different directions (along B for S and along Bhf for I). The contri-
bution to Bhf can be resolved into two components that are parallel and perpendi-
cular to B. The latter defines axis x. Using Eq. 5.12b, the spin hamiltonian becomes
H � gbeBMS13 � gnbnBhfT� I (5:20a)
¼ gbeBMS13 � gnbn½B? Ix þ BkIz� (5:20b)
Note that Bhf ¼ [B?2þ Bk
2]1/2. For purposes of illustration, we consider the case of
I ¼ 12, but allow S to be arbitrary.
If the spin functions for I quantized along z are denoted by ja(n)l and jb(n)l,corresponding to MI ¼ þ
12
and � 12, then the nuclear-spin hamiltonian matrix in
terms of these is4 (Sections A.5, B.5 and C.1.2)
�H ¼ ha(n)j
hb(n)j
ja(n)i jb(n)i
gbeBMS �gnbnBk
2�
gnbnBk
2
gbeBMS þgnbnBk
2
2664
3775
(5:21)
On diagonalizing this matrix, the energies for this system are found to be
UMS,MI¼ gbeBMS þ jgnbnBhf jMI (5:22a)
¼ gbeBMS þ (nT�A�AT� n)1=2jMSjMI (5:22b)
and the energy eigenfunctions are admixtures of ja(n)l and jb(n)l. Here, as dis-
cussed in Section 5.3.2.1, MS and MI represent spin components taken along two
different directions, and MI changes sign during the electron spin flip. However, it
is convenient and conventional (although incorrect) to write Eq. 5.22a as
UMS, MI¼ gbeBMS þ AMSMI (5:22c)
where now MI is taken as constant when MS changes sign, and
A ¼ (n T . A . A T . n)1/2.
We now discuss the determination of matrix A from a set of EPR spectra taken at
a suitable set of crystal orientations.
128 HYPERFINE (A) ANISOTROPY
First, we present an illuminating example. Consider a single electron in a hybrid
orbital
jspi ¼ csjsi þ c pj pi (5:23)
centered on the interacting nucleus located at the origin. Here jcsj2þ jcpj
2 ¼ 1. We
take state jspl to be an s orbital admixed with a p orbital (Table B.1), whose axis (z0)is taken to lie in the xz plane at angle up from z (Fig. 5.5), the latter chosen to be
along B. Thus A is symmetric and uniaxial about this axis. Note that the direction
x is defined by the relative directions of axis z0 and B (and is arbitrary if up is 0
or 1808). We take g to be isotropic and neglect the nuclear Zeeman term in
Eq. 5.12a. Since S is quantized along z, terms in Sx and Sy in Eq. 5.5 may be
neglected. In the present case, the analogous situation does not hold for I, so that
the term in SzIx contributes. Using polar angle u and azimuthal angle f for r, one
can substitute r cos u for z, r sin u cos f for x, and r sin u sin f for y in Eq. 5.5.
The relevant effective hyperfine magnetic field components (along x and z; see
Fig. 5.5) are then given by
B? ¼ þMS
gnbn
3dA sin up cos up (5:24a)
Bk ¼ �MS
gnbn
½A0 þ dA(3 cos2 up � 1)� (5:24b)
FIGURE 5.5 The hybrid orbital jspl in a magnetic field B showing the vector r from the
nucleus to the unpaired electron, as well as relevant angles.
5.3 DETERMINATION AND INTERPRETATION OF THE HYPERFINE MATRIX 129
so that
Bhf ¼ {9(dA)2 sin2 up cos2 up þ ½A0 þ dA(3 cos2 up � 1)�2}1=2=2gnbn (5:25a)
The general form for A (Eqs. 5.27) was made available in 1960 [16]. Clearly, unless
B? vanishes, the correct nuclear-spin eigenfunctions for the spin hamiltonian
(Eq. 5.21) are admixtures of ja(n)l and jb(n)l. Note that it is the sign of dA/A0
that is important in Eq. 5.27a.
At constant microwave frequency, EPR transitions occur at the resonant fields
B ¼hn
gbe
�ge
g½(a0 � b0)2 þ 3(2a0 þ b0)b0 cos2 up�
1=2MI (5:28a)
¼hn
gbe
�
ge
g
�aMI (5:28b)
where a(up) ¼ A(up)/gebe (as in Eq. 2.48), a0 ¼ A0/gebe and b0 ¼ dA/gebe (Eqs.
5.9a,b). The sign of a can be taken as positive, since we are dealing only with first-
order hyperfine effects here.
130 HYPERFINE (A) ANISOTROPY
It is of interest to consider two limiting cases:
1. jA0j � jdAj. Here a is given [15] by
a ¼ jb0(1þ 3 cos2 up)1=2j (5:29)
2. jA0j � jdAj. The square root in Eqs. 5.28a may then be expanded to give
a ffi ja0 þ b0(3 cos2up � 1)j (5:30)
For intermediate cases, the general relation (Eqs. 5.28) must be used.
It would appear at first glance that A (up dependence in Eq. 5.27) does not average
to A0 for a molecule tumbling in a liquid. However, one must realize that it is the
hyperfine magnetic field at the nucleus, and not the energy, that is averaged over
all orientations. It is clear from Eq. 5.24a that B? averages to zero, whereas Bkaverages to 2MS A0/gnbn, as required. The energy for the tumbling system is not
obtained by averaging UMS,MI(Eqs. 5.27).
We now return to the general problem of obtaining A in the case B� Bhf. As in
Eqs. 5.12–5.14, the hyperfine interaction is considered in terms of the hyperfine field
Bhf at the nucleus. From Eq. 5.22, it is clear that the hyperfine part of the transition
energy DU is proportional to Bhf. The difference of transition energies DU is given
by gnbnBhf ¼ A, and is proportional to the magnitude [nT�A �AT�n]1/2 of vector
AT� n (Eq. 5.14a). With reference to the allowed (fixed-field) transitions k and m
of Fig. 2.4a, which occur at frequencies nk and nm, one has h(nk 2 nm) ¼ A.
For fixed-frequency spectra (Fig. 2.4b), the spacing between lines is Bm 2 Bk ¼
A/gbe at sufficiently high fields.
The procedure for evaluating the elements of the hyperfine matrix is analogous to
that for evaluating the g matrix in Section 4.4, since AT�n is a vector akin to gT� n.
where AA by definition is A�AT. Thus (Eq. 4.11b) one has
A2 ¼ ½ cx cy cz � �(AA)xx (AA)xy (AA)xz
(AA)yy (AA)yz
(AA)zz
24
35 �
cx
cy
cz
24
35 (5:31b)
The task at hand (compare with Eq. 4.15) is thus the evaluation of the elements of the
matrix AA, which is symmetric and hence contains only six independent com-
ponents.6 From Eq. 5.31b one obtains (Eqs. A.52)
5.3 DETERMINATION AND INTERPRETATION OF THE HYPERFINE MATRIX 131
A2 ¼ (AA)xx sin2 u cos2 fþ 2(AA)xy sin2 u cosf sinfþ
(AA)yy sin2 u sin2 fþ 2(AA)xz cos u sin u cosfþ
2(AA)yz cos u sin u sinfþ (AA)zz cos2 u (5:32)
We note that
A2 ¼ (gebea)2 (5:33)
where (ge/g)a is the experimental (first-order) splitting, which must be measured at
suitable orientations.
Once matrix AA has been obtained from the EPR spectra, the next task is to
diagonalize it. Note that all three of its principal values are non-negative. If we
take their square roots, we can obtain the magnitudes that are usually reported
in the literature. These are not necessarily those of the principal values of the
symmetrized hyperfine matrix (Aþ AT)/2. As already mentioned,1 the true
matrix A in general is asymmetric; that is, A = AT. In most of the literature,
it is at this point in the analysis that the hyperfine matrix is assumed to be sym-
metric, and it is that matrix (A) that is reported. This equals the true matrix A
reported only if in fact A ¼ AT for the latter. Luckily, knowledge of the
reported matrix usually suffices to fully characterize the EPR spectra of the
spin species studied, but this does not necessarily suffice when exact quantum-
mechanical modeling of the molecule is the objective.
As stated above, the magnitudes of the elements of the diagonal form are obtained
from the square roots of the principal values of AA. The relative signs of the prin-
cipal values become available when the fields B used are sufficiently large that the
nuclear Zeeman term in Eq. 5.7 affects the spectra. In some instances, signs and
likely asymmetry become available from quantum-mechanical modeling of the mol-
ecular species of interest.
Consider the especially simple system when we encounter uniaxial symmetry. In
this case, Eq. 5.32 becomes
A2 ¼ A?2 sin2 uþ A 2
k cos2 u (5:34)
where u is the angle between the unique axis and B.
Returning to the more general anisotropic case, we now apply the expressions
presented above to actual experimental hyperfine coupling data to obtain a matrix
A for the a-fluorine atom of the 2OOC22CF22CF222COO2 radical di-anion. This
species is obtained by irradiation of hydrated sodium perfluorosuccinate [17].
This p-type radical has its unpaired electron primarily in a non-bonding 2p
orbital on the trigonal carbon atom but, as we shall see, with appreciable spin popu-
lation also on the a-fluorine atom. Thus the s þ p example just presented (Eq. 5.23)
is relevant but is not quite general enough. The crystal structure is monoclinic, with
132 HYPERFINE (A) ANISOTROPY
unit-cell parameters a ¼ 1.14, b ¼ 1.10, c ¼ 1.03 nm and b ¼ 1068. Here b is the
angle between the a and the c axes. An orthogonal a0bc axis system is chosen,
taking a0 to be perpendicular to the bc plane. Figure 5.6a exhibits a typical
X-band EPR spectrum taken at 300 K, displaying the substantial a-fluorine splitting
as well as the smaller ones from the b-fluorine atoms. The g factors range from
2.0036 to 2.0060 but are herein treated as isotropic. In Fig. 5.7 the hyperfine split-
tings from the a atom are plotted as the magnetic field explores the a0b, bc and a0c
planes of the single crystal for both the allowed and the ‘forbidden’ lines.
With the crystal point group symmetry C2 at hand here, the radicals occur in
two different orientations (Section 4.5) related by two-fold axis b, as well as
translation (and possibly inversion). Thus, generally, spectra from both are
present and care must be taken in the analysis not to mix these up. How-
ever, these do superimpose exactly for B in the a0c plane or for B parallel to
b [18].
The elements of the a-19F AA matrices for the two sites can be obtained by
interpolation from these plots, using values at a set of special angles. Such data
are listed in Table 5.1 (one can average the duplicate measurements). However,
for better precision, least-squares fits should be made (using plots of A2 versus
rotation angle) to all the experimental data, including the forbidden lines.
FIGURE 5.6 (a) Second-derivative X-band spectrum of the perfluorosuccinate radical
dianion at 300 K for B k b at 9.000 GHz; (b) similar spectrum at 35.000 GHz showing the
greatly increased intensity of the forbidden transitions. [After M. T. Rogers, D. H. Whiffen,
J. Chem. Phys., 40, 2662 (1964).]
5.3 DETERMINATION AND INTERPRETATION OF THE HYPERFINE MATRIX 133
The matrix obtained from the limited data in Table 5.1 is
AA=h2 ¼
1:60 +4:78 0:64
16:36 +0:16
2:71
264
375� 104 (MHz)2 (5:35)
The choices in sign for two of the matrix elements are associated with the presence
of the two symmetry-related types of radical sites (Problem 5.6). Both matrices have
the same set of principal values.7
Note that the qualitative appearance (Fig. 5.7) of the plots of hyperfine splittings
versus orientation indicates the relative importance of off-diagonal elements of AA.
For example, if the relevant off-diagonal element is comparable in magnitude with
the diagonal elements it spans, then the plot of the splitting in the given plane is very
asymmetric about its center. However, if the off-diagonal element is relatively small,
then the plot is close to symmetric. Figure 5.7a is a good example of the former case
[i.e., (AA)xy is relatively large], whereas Fig. 5.7b represents the latter case [i.e.,
(AA)yz is small].
Either of the two matrices 5.35 may now be diagonalized by subtracting a
parameter l from each diagonal element and setting the resulting determinant
equal to zero (Section A.5.5). Expansion of the determinant yields the following
cubic equation:
l3 � 20:67l2 þ 51:56l� 1:30 ¼ 0 (5:36)
TABLE 5.1 Selected Hyperfine Splittings a and the Components of AA
for the a-Fluorine Atom of the 2OOC22CF22CF222COO2 Radical Ion
Plane Angle (deg) (A/h)2 (MHz2) Tensor Element
a0b 0 1.61 � 104 ¼ (AA)a0a0/h2
90 16.24 � 104 ¼ (AA)bb/h2
45 13.84 � 104
} Difference ¼
135 4.29 � 104 2(AA)a0b/h2
bc 0 16.48 � 104 ¼ (AA)bb/h2
90 2.72 � 104 ¼ (AA)cc/h2
45 9.67 � 104
} Difference ¼
135 9.99 � 104 2(AA)bc/h2
ca0 0 2.69 � 104 ¼ (AA)cc/h2
90 1.59 � 104 ¼ (AA)a0a0/h2
45 2.69 � 104
} Difference ¼
135 1.42 � 104 2(AA)a0c/h2
a Measured at 300 K with n ¼ 9.000 GHz. Only the data for one radical site are displayed.
Source: Data from M. T. Rogers, D. H. Whiffen, J. Chem. Phys., 40, 2662 (1964).
134 HYPERFINE (A) ANISOTROPY
The roots of this equation are 17.77, 2.87 and 0.025 (all �104 MHz2); hence
dAA=h2 ¼
17:77 0 0
2:87 0
0:025
264
375� 104 (MHz)2 (5:37)
The smallest principal value is not accurately determined from the present data.
Other orientations are required to obtain a more accurate value. By taking square
FIGURE 5.7 Angular dependence of the hyperfine line splitting (MHz) in the2OOC22CF22CF222COO2 radical at 300 K, yielding the data in Table 5.1. The uncertainty
of data represented by large circles is greater than that for the small circles. Curves are
drawn for only one of the two symmetry-related sites (the upper signs of the
direction-cosine matrix of Table 5.1). Dotted lines correspond to spectral lines with relative
intensity less than 20% of the total absorption intensity. (a)–(c) The microwave frequency
is 9.000 GHz. B is in the a0b, bc and a0c planes in (a)–(c); (d )–( f ) spectra analogous to
(a)–(c) but for a frequency of 35.000 GHz. [After M. T. Rogers, D. H. Whiffen, J. Chem.
Phys., 40, 2662 (1964).]
5.3 DETERMINATION AND INTERPRETATION OF THE HYPERFINE MATRIX 135
roots, we obtain (AþAT)/2; this equals A if as usual the latter is assumed to be
symmetric. Thus we have
dA=h ¼421:5 0 0
169:4 0
16
24
35 MHz (5:38)
where there is an ambiguity as to the sign of each principal element of dA. It is poss-
ible to obtain the correct signs for the diagonal elements of dA, if the nuclear Zeeman
term (the final term in Eq. 5.12a) is significant. This term accounts for the difference
between the 9-GHz separations in Figs. 5.7a–c, for which the nuclear Zeeman term
is negligible, and the 35-GHz separations of Figs. 5.7d– f, for which the full theory
must be used. All three signs turn out to be positive (see below). The matrix A in the
crystal coordinate system, obtained from the complete data set (Fig. 5.7), is pre-
sented in Table 5.2, as are its principal values and directions. Here small corrections
(Eqs. 5.15 and 5.16) were made to account for the effect of the nuclear Zeeman term.
In other words, the approximation B� Bhf is not quite adequate. All three principal
values were chosen to be positive.
As is now evident, the analysis of a complex spectrum, which may contain ‘for-
bidden’ transitions (e.g., lines for which DMS ¼+1, DMI ¼+1), is often aided by
using two different microwave frequencies. Figures 5.6a and 5.6b illustrate the spec-
trum of the 2OOC22CF22CF222COO2 radical at 9 and at 35 GHz, that is, cases
B� Bhf and B � Bhf. The latter spectrum clearly shows the ‘forbidden’ transitions.
We now demonstrate the use of a high microwave frequency in determining the
relative signs of hyperfine matrix elements. The measurements to be considered are
the ones made at 35.000 GHz. Thus here we revisit the case B � Bhf. The a-19F
hyperfine splittings are computed in the following manner.
The main-line splitting in the [100] direction is used as an example. We obtain
BhfT (Eq. 5.14b) from
�MSnT�A=h ¼ 2MS ½�23:5, + 51:7, �16:2� MHz (5:39)
TABLE 5.2 The a-19F Hyperfine Matrix Asym of the Perfluorosuccinate
Radical Dianion and Its Principal Values and Direction a
Matrix A/h
(MHz)
Principal Values
(MHz)
Direction Cosines
relative to Axes a0bc
46.9 +103.3 32.5 421 0.267 +9.964 0.011
— 392.7 +5.9 165 0.208 +0.068 0.976
— — 157.7 11 0.941 +0.258 20.219
a The upper and lower signs refer consistently to the two sets of radical sites. These data were obtained at
300 K with n ¼ 9.000 GHz.
Sources: Data from M. T. Rogers, D. H. Whiffen, J. Chem. Phys., 40, 2662 (1964); also see L. D. Kispert,
M. T. Rogers, J. Chem. Phys., 54, 3326 (1971).
136 HYPERFINE (A) ANISOTROPY
by use of matrix A/h in Table 5.2. With B ¼ 1.2475 T, we obtain (using Table H.4)