This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:
IP Address: 150.214.230.47
This content was downloaded on 25/05/2014 at 20:05
Please note that terms and conditions apply.
A new force-constant model for graphite
View the table of contents for this issue, or go to the journal homepage for more
1977 J. Phys. C: Solid State Phys. 10 2295
(http://iopscience.iop.org/0022-3719/10/13/006)
Home Search Collections Journals About Contact us My IOPscience
J . Phys. C: Solid State Phys., Vol. 10, 1977. Printed in Great Britain. @ 1977
A new force-constant model for graphite
A P P Nicholson and D J Bacon
Department of Metallurgy and Materials Science, University of Liverpool, PO Box 147,
Liverpool L69 3BX
Received 26 November 1976
Abstract. Earlier force-constant models of graphite either assume the lattice vibrations
parallel and perpendicular to the basal plane to be independent, which leads to a zero
value for the shear modulus C,,, or require an external pressure to be applied for a finite
value of C,, to be obtained. In order to avoid these difficulties, and to account for the elastic
constants, it was found necessary to consider a model allowing for general interactions out
to third-nearest neighbours in the plane and second-nearest neighbours in adjacent planes.
The best fit obtained to experimental data uses a model which consists of eight pairwise
potential constants and three many-body valence constants. This model gives an exact fit
to all five elastic constants and a good fit to the phonon dispersion data. The combination
of pairwise potential and valence constants would make the model suitable for predicting
the displacement fields of vacancy clusters using the lattice statics method.
1. Introduction
A common feature of early force-constant models of graphite (Krumhansl and Brooks
1953, Yoshimori and Kitano 1956, Young and Koppel 1965) is that vibrations parallel
and perpendicular to the basal plane are independent, i.e., there is no coupling between
these modes of vibrations. This assumption considerably simplified the calculation of
the phonon spectrum and hence the specific heat since the 12 x 12 secular determinant
could now be reduced to three 4 x 4 determinants for any wavevector. Such decoupling
in a layered structure is not unreasonable for predicting properties such as the specific
heat, which are not very sensitive to details of the phonon dispersion curves. Unfortunately
these decoupled models can only have limited application since the lattice structure is
inherently unstable; this is due to lack of resistance to shear (C44 is always zero).
Phonon dispersion data for graphite have been extremely difficult to obtain owing
to the lack of good quality large single crystals. The first measurements of Dolling and
Brockhouse (1962) for modes propagating along the c axis have been improved upon
by Nicklow et a1 (1972) (hereafter referred to as NWS), who also made some measure-
ments for the first time of modes propagating in the basal plane. NWS analysed their
data with a force-constant model using pairwise interactions up to the third in-plane
neighbour and the first out-of-plane neighbour. Subsequently, Mani and Ramani (1974)
(hereafter referred to as MR) reinterpreted the NWS data in terms of a valence-constant
model of the same range. I n both these models a finite value was assigned to C44. As
we shall show, this implies non-vanishing stresses in the lattice, which thus require an
external pressure to be imposed to maintain equilibrium. This is somewhat unreasonable
since it is usual to assume that experimental data have been obtained at zero pressure.
2295
2296 A P P Nicholson and D J Bacon
Thus a proper application of the equilibrium conditions to these models makes C,,
zero, reducing them to the decoupled type used by earlier workers, and destroying the
fit to the phonon dispersion data.
The first objective of this work has therefore been the construction of a stable force-
constant model of graphite which is consistent with the NWS phonon dispersion data.
Secondly, we wished to use the resulting model in lattice-statics calculations of the
mechanical properties of vacancies, as has been done previously using a two-dimensional
single layer model (Nicholson and Bacon 1975a, b). Since such properties are dominated
by elastic contributions, we require the new model to predict the correct elastic constants.
I n our previous work (1975a), we were limited to studies of single isolated vacancies
and their interactions, for the valence-constant model used could not give any value
for the forces simulating the vacancy. These were calculated for the single vacancy
only using the defect-molecule method (1975b). To allow studies to be made of larger
defect clusters, we therefore permit the first derivatives of the total potential with
respect to any bond length or other invariant coordinate to be nonzero.
I n the next section the effect of invariance and equilibrium conditions on force
constants is described; the results are applied to a new model consisting of twenty
independent force constants. Pairwise potential contributions to this model are
considered and the nonzero C,, error in the NWS and MR models discussed. I n 5 3 the
limitations involved in fitting this model to the data are discussed, and the final successful
version consisting of pairwise-potential and valence-type interactions is described. The
results obtained are compared with the NWS and MR models in the final section.
2. The new force constant model
The graphite structure, which has space group D&,, is shown in figure 1. Referred to the
Cartesian axes A, shown, the primitive lattice vectors are
where a = 2*46A and c = 6.70A are the lattice parameters. The atoms labelled
1, 2, 3, 4 in the figure are all in the same unit cell, namely the origin cell with index
1 = l , a, + 12a2 + 13u, equals zero. Atoms labelled 1, l, l, etc., are on the same sub-
lattice. A feature of the graphite structure is the lack of inversion symmetry at the
atomic sites, although there is a centre of inversion at the point (0, 0, 4 4 ) for example.
We have followed standard practice in defining the force-constant matrix by
where subscripts U, refer to the Cartesian axes, u(1k) is the displacement from equilibrium
of the kth atom in cell 1, Q is the total potential energy of the lattice, and the derivatives
are evaluated at the equilibrium configuration.
In setting up any force-constant model for an infinite lattice it is necessary to ensure
that the force constants satisfy certain invariance relations and equilibrium conditions.
From the continuity of the potential energy, 0, wemust have
+,&lk ; lk) = 4@,(1k ; lk),
A new force-constant model for graphite 2297
and from its (numerical) invariance to rigid-body translations wemust have
These conditions are easily satisfied and we need consider them no further. The re-
maining conditions on the force constants arise from (i) (numerical) invariance of @
to rigid body rotations, (ii) invariance of the form of @ to symmetry operations belonging
to the space group of the crystal, and (iii) the structure must be an equilibrium configura-
tion. I t can be shown that any perfect lattice model of graphite satisfying space-group
symmetry ((ii) above) will automatically satisfy the rotational invariance condition (i).
1
Figure 1. The graphite lattice, showing the Cartesian axes and atom labelling used here, and
the angle q5 required in the definitions of pairwise potential contributions and of the valence
constant A.
The equilibrium condition for an infinite lattice ((iii) above) is in two parts. Firstly
the net force on each atom must vanish at the equilibrium position. This is automatically
the case in the harmonic approximation. Secondly, the configuration must correspond
to vanishing stresses. I n general it is not possible to determine the pressure, leaving
five anisotropic stress components to be put to zero, namely o12, 013, 023,
and ( o ~ ~ - o ~ ~ ) . I t can be shown that the space-group symmetry condition (ii) is sufficient
to ensure that all except the last of these components are always zero. Thus we have
at most two relations amongst the force constants to apply after making maximum
use of the crystal symmetry to reduce the number of independent force constants,
namely
and the pressure
(O11 - 033) = 0,
p = -'2 3( O11 + 633) = 0.
(3a)
(3b)
The results above were a by-product of an analysis of the effect of crystal symmetry
on various lattice sums and the resulting well-known bracket functions [ ~ p , y l ] and
A P P Nicholson and D J Bacon
(a/?, ?A) used in evaluating the elastic constants by the method of long waves (Born and
Huang 1954). Graphite has five independent elastic constants, C,,, c66, C,,, C13, and
C, , , and neither of the Cauchy relations C,, = C66 and C13 = C,, are obeyed. Lack
of inversion symmetry at the atomic sites is sufficient to ensure C, #C,,, which can
thus be achieved even with pairwise potentials, as in the NWS model. For C,,, C,,
and C, , however there are no sublattice shift contributions, i.e., (a/?, ?A) is zero for any
index equals three. Since for pairwise interactions the Cap, ?A] functions are symmetric
in all four indices,many-body interplanar interactions are necessary to ensure C, , #C,,.
Table 1 . General force-constant matrices for graphite.
Atoms involved in interaction
Three 1st-neighbours at distance
r i =aJJ3
e.g., M =1, N =2 (in-plorie)
Six 2nd-neighbours at distance
r 2 =a
e.g., M =1, N =1' (in-plane)
Three 3rd-neighbours at distance
r 3 =2ol,/3
e.g., M =1, N =2' (in-plane)
Two 4th-neighbours at distance
r4 =cl2 e.g., M =1, N =3
(1st-neighbour out-of-plane)
Twelve 5th-neighbours of atom 1 at distance
r 5 =(a213 +
(2nd-neighbour out-of-plane)
Six 5th-neighbours of atom 2 of
sublattice 4
e.g., M =2, N =4"
e.g., M =1, N =4'
Force-constant matrix for
atoms M and N (see figure 1)
(; A1 0 4 : )
(-i: p: 5,)
(i' i3 i3)
(i4 1. 5 )
(i: is 5:) +
0 c,
t By inversion symmetry we have, q5zB(0. 2; 13) =+pa( O 1 ; 1 - a, - a24).
We therefore initially considered a model (hereafter referred to as model I) which
consisted of the most general interactions allowed by the crystal symmetry between
up to third neighbours in the plane and second neighbours in adjacent planes, with
one limitation imposed. Second-neighbour in-plane interactions between atoms on
sublattice 1, and the similar interactions between atoms on sublattice 2, were assumed
to be exactly equivalent. The resulting 21 force constants of model I are defined in table 1.
The notation used is compared with that of earlier models in table 2. The entries for
A , and B, for the MR model result from certain inconsistencies in their work ; these are
discussed later in &I. Expressions for the elastic constants for model I are given in the
Appendix.
Since the nature of the interactions in model I has not yet been specified, an expression
for the pressure cannot be obtained, and we are left with one relation between the 21
force constants, equation (3a). I t has been shown by Huang (1950) that the vanishing
of the five anisotropic stress components together with ten conditions relating
A new force-constant model f or graphite 2299
to rotational invariance are equivalent to fifteen extra symmetry conditions on the
[ap, ?,I] functions, namely
[.P, = [;,A, 4. (4)
The only one of these Huang conditions not automatically satisfied by a graphite
lattice is the one equivalent to (3a), [l l , 331 = [33, 111. For model I this relation takes
the form
4a2(C, + 6C2 + 4C, + 2C, + C;) = c2[2A, + 6( A, + B,) + 3(A; + Bi ) ] ,
( 5)
thus giving 20 independent constants. Allowing for misprints in both papers (which
did not affect their numerical results), equation (5) contains within it equation (1) of
NWS and equation (2.5) of MR.
Table 2. The relation of the force constants of model I to those of previous models and to
pairwise potential contributions.
Model I NWS MR Young and Pairwise potential
Koppel (1965) contributions?
E1
Y1
a2
Y2
- 2
@a
Y3
E4
Y4
0
0
0
0
0
81
82
P3
- +( E, - 38,)
+(h, - P I )
P 2
- 4 2
83
Y 1
a2
Y 2
E3
Y3
E4
Y4
0
0
0
0
0
K
6fJ
2$/3
- 3/44
pi4
- , 4 9
0
0
0
0
K
0
0
0
0
0
Tl
T2
t The parameters Ri and 7; are defined in equation (6). C and S are cos 4 and sin 4 respectively
where rj is shown in figure 1 i.e., tan 4 =2a/ J 3c.
All previous models have used either pairwise potentials or valence-type potentials,
or both to represent the interactions. Valence-type potentials are particularly useful
since they automatically satisfy the equilibrium conditions, which depend solely on
first derivatives of the total potential energy, @. The contributions which a pairwise
potential V( r) makes to the force constants of model I are shown in table 2. The ten
potential parameters are defined by
where r n is the perfect lattice distance between nth neighbours (as defined in table 1).
We have assumed that the pairwise contribution to the second-neighbour out-of-plane
interaction (r = r J is independent of the sublattices to which the atoms belong. The
2300 A P P Nicholson and D J Bacon
pressure can now easily be evaluated (Born and Huang 1954), and the equilibrium
conditions (3) can be written as
o,, = (4/,/(3)c)(T, + 6T + 4T3 +3T5) = 0,
03.7 = (c/J (3) aI(T4 + 9TJ = 0
(74
(7b)
and
I t can immediately be seen that in the NWS and MR models where T, is zero we must
also have T4zero for zero pressure. I n these models T4is the sole contribution to force
constant A,, and C, , = (c/J (3)a2)A4:Thus C, , must be zero to give zero pressure in
the NWS and MR models.
3. Values for the force constants
Initially we performed a least-squares fit of 19 of the 20 independent force constants
of model I to all the available experimental data, namely the phonon dispersion data
of NWS, the Raman frequencies (Tuinstra and Koenig 1970, Brillson et a1 1971), and the
five elastic constants (Blakslee et a1 1970, Seldin and Nezbeda 1970). (Force constant
D, cannot be determined by these data). The solution obtained was unsatisfactory on
two counts. Firstly the fit obtained was not particularly good, especially for the elastic
constants. Secondly and more importantly, the values predicted for some of the force
constants were physically unrealistic : A ,, for example. expected to have the largest
value, came out as one of the smallest. Accordingly we performed the calculations
again with an exact fit to the elastic constants and with certain physically-reasonable
constraints applied to the permissible values of the force constants. These were
(a) A , > B, > C, $ C, > A, > 0,
(b) A , > B, > A, , B, > A, > B,, and C, > C, > C,, and
(c) A, > A, , B,, A; or B;, and
C, > C, or C,.
Despite these extra constraints, we were unable to find a satisfactory solution. In
fact too many force constants were being fitted to too little data. I t was discovered
that an excellent fit to the NWS data for the phonon dispersion along r - A - A could be
obtained simply by specifying a fit to the two lowest and the highest optical frequencies
at the centre of the zone. However, the lowest of these frequencies, o,,(T), given as
50 cm-byNWS,wasfoundto beincompatiblewiththevalueofC,,of0~4 x 10ON m-,
at least when using our model I or any sub-model of it. We therefore decided to fit to
C, , , reducing WTo(r) to about 42cm-. Of the remaining phonon dispersion data, it
was felt that the greatest reliance could be placed on the lowest frequency at the edge
of the zone, point M. This point had been very carefully examined by NWS, and we
think it only reasonable to fit our model to it. Thus, there are only nine items of data to
fit a model to, five elastic constants and four phonon frequencies.
The simplest model we could envisage involved the eight independent pair-potential
parameters discussed in the previous section together with at least one three-body
valence constant ; this is required to give the observed Cauchy discrepancy C,, #C,,
and must be interplanar. Mani and Ramanis (1974) major criticism of the NWS model
was the unphysical nature of purely pairwise in-plane interactions which. because
they are axially symmetric, imply that B, = C,. A , = C,. and B, = C,. We are in
A new force-constant model f or graphite 2301
full agreement with this criticism and therefore allowed the possibility of at least one
in-plane valence constant being nonzero.
After trying various combinations, an excellent fit was finally obtained with the
following valence constants.
(i) An interplanar constant A giving the energy change due to changes in the angle
q5 shown in figure 1 as
2A@(/1) = (l/S4)A(r,Aq5)z,
where
S = sin q5 and r , = a/J3,
Of all the single constants defining a three-body interaction of this triplet of atoms,
this was the only one that gave a reasonable value for the lowest optical frequency at
(ii) The in-plane adjacent-bond-angle constant p as used by Young and Koppel
(1969, and called k , by MR.
(iii) An in-plane adjacent-bond-length constant v, equivalent to the constant k,,
used by MR. v was required in order to make B, less than A,.
The relations between force constants and valence constants given by MR in their
equation (3.2) appear to be incorrect. Using the same definitions of valence constants
k,, k,,, k,e, k,, and k, as in their equation (3.1), and with 1, p, and v as defined above, we
obtained for the valence-constant contributions to the model I constants the following :
and a positive value for C, .
A1 = k, - 2)) - J(3)kr, + 2A/S2,
Bi = 6~ - J(3)kre3 C1 = 6k,/r:,
A2 = -&3p + v - J(3)kr,), Bz = & + 311 + J(3)krJ,
Cz = - k, / r: , Dz
(8)
i tJ (3)~ - 4( 3) ~ + krd,
A, = 3A/C? A, = -A, C, = A, E, = -SA/C and F, = -CA/$
where r , = a/J 3 is called ro by MR, and C = cos q5 and S = sin 4, where tan q5
= 2a/,/(3)c. As shown in table 2, the first neighbour constants a1 and /3, used by MR
differ from the normal definition. We obtain,
A useful check can be made on equation (8) by confirming that the resulting force
constants satisfy the Huang condition (5).
With eleven ( = 8 + 3) constants and nine data points, the fitting was made by
putting T3 = 0 and expressing all the constants as functions of T,. T, was then given
a value in the middle of the region defined by the constraints on the force constants
described earlier in this section and by optimising the fit achieved to the lowest phonon
frequencies at r and M, i.e., o&) and oTA(M).
Values for the 13 parameters in this successful model, hereafter referred to as model
11, are shown in table 3. These values satisfy the equilibrium conditions, equation ( 3) ,
which with T3 = 0, leaves 10 independent constants. The resulting values for the model
I constants are compared with those for previous models in table 4. A similar comparison
2302 A P P Nicholson and D J Bacon
Table 3. Values obtained for the 13 parameters of model I1 in units of N m-
Tl 231.950 R, 241.746 1, -0.486
T2 -38.533 R , 122.440 p 8,000
T3 0 R3 -18.830 IJ -4.400
T4 0.250 R, 0.997
75 -0.250 R, 1.045
is made in table5, where values for the elastic constants and certain phonon frequencies
are presented. The phonon dispersion curves obtained for the lines r - A - A and
- C - M are compared with experiment in figures 2 and 3 respectively.
4. Discussions and conclusions
A study of the results presented in tables 3-5 and figures 2 and 3 show that the original
objectives of this work as outlined in the Introduction have been achieved, albeit with a
simpler model than the one initially outlined. It can be seen that model I1 satisfies all
the required equilibrium conditions, gives an excellent fit to the experimental data,
and allows calculations of the first derivatives of the total potential to be made very
easily. The phonon dispersion curves obtained are very similar to those ofNWS (allowing
for finite pressure in their case). A comparison of the pairwise contributions to our
model, shown in table 3, and the NWS force constants (table 4) shows some similarity
Table 4. Values obtained for the model I force constants and for previous models in units
of N m-'.
~~
Force constant Model I NWS? M Rt
323,362
279.950
231.950
-33.533
9 1.440
- 38.533
22.517
- 18.830
0
0
0.529
0.997
0.433
-0.250
0.362
- 0.259
- 1,612
-0.053
0.848
0.465
362.00
199.00
199.00
- 52.00
133.00
- 52.00
0
- 3.70
28.80
28.80
0.77
5.80
0
0
0
0
0
0
0
0
4.971
625.941
46.863
48.933
60.260
7.423
31.758
- 22.779
-22.179
0
0.626
5~724
0
0
0
0
0
0
0
0
t Finite pressure values
A new force-constant model f or graphite 2303
in the general pattern, except for the relative sizes of R, and Tl ; this is reflected in the
high value obtained for C, , by NWS.
As noted in 52, there are certain inconsistencies in the MR model. Except for the
comparison they made with the NWS model (MR table 4), the definition used throughout
their paper of the first-neighbour interaction constants E , , p,, y, was for a pair of atoms
such as 3 and 4 shown in figure 1 of this paper. The values which should have been
used in their table 4 are therefore a; = -&al - 3p,) = 0.04977 and = 3(3a1 - p,)
= 6.25941. For symmetry reasons this has very little effect on the phonon dispersion
curves, but the unrealistic nature of these values is seen in the predictions for the elastic
Table 5. Elastic constants (10" N m-') and phonon frequencies (em-') for graphite.
Experiment
Model I1 Reference NWSt MRt
1063
3.654
0.401
1575
15753
1407
1405
127f
1575
1406
90
30
476$
44:
1.502
42$
106 (a, b)
44 (a, b)
3.65 ( U, b)
1.50 ( U, b)
0.40 (a, b)
1588 (c. d )
1575 ( c, d)
50 (e)
-
-
127 (e)
144
463
3.71
=c44
0.46
1575
1575$
45$
1396
1395
1271
1575$
1397
903
33$
473$
43
91
3.65:
0.40$
=c44
1576
15753
42
644
630
126
1574
636
89
30
224
t Values quoted require finite pressure.
$ Experimental value used to fit force constants.
References ( a) Blakslee et al (1970) ( b) Seldin and Nezbeda (1970) (c) Brillson et al (1971)
(d) Tuinstra and Koenig (1970) ( e) Nicklow et al (1972).
constants in table 5. With C,, = 43 and c66 = 91 in the usual units, we obtain CI2 =
- 139 ; thus C,, e ICl2[, and the MR model is unstable even with a pressure applied
to allow nonzero C44. These problems would not have been encountered by MR if the
expressions used relating valence to force constants (MR equation (3.2)) had been
correct ; using the expressions we have derived, equation (8), we found C, = 95 and
c66 = 42, which is quite reasonable. Unfortunately the phonon dispersion data cannot
then befitted without a high value of 270 N
Our new model has two major limitations. Firstly, no account has been taken
of possible electronic effects, such as the effective charge associated with the highest
frequency mode at the centre of the zone, which is infrared active (Brillson et al 1971).
Secondly the range of values of the force constants is very high, the ratio of largest to
smallest being about 6500. Naturally we tried to avoid this situation, but this was not
found possible with the limited data available.
for ( A3 + B3) .
2304 A P P Nicholson and D J Bacon
-- ' E
2 156f
4
3
2
1
1.0 0.8 0.6 04 0.2 0
q'qmox
Figure 2. Dispersion curves for graphite for wave-
vectors q along the line r - A - A of the Brillouin
zone; that is from the centre of the zone along the
[OOl] axis to the centre of the top face. -,
present work ; open circles, neutron measurements
of NWS ; full circle, optical measurements (Tuinstra
and Koenig 1970, Brillson et al 1971).
q'qrn,,
Figure 3. Dispersion curves for graphite for wave-
vectors q along the line r - C - M of the Brillouin
zone; that is, from the centre of the zone along the
[loo] axis to the point at the centre of a prism face.
-, present work ; open circles, neutron measure-
ments of NWS.
The model predicts quite reasonable values for the (unrelaxed) radial forces induced
by the presence of a vacancy. The nearest-neighbour force is - 20 eV A- compared
with a value of - 11 eV A- ' obtained by the defect molecule method (Nicholson and
Bacon 1975b). These values are not strictly comparable since the model assumptions
are different, but it suggests our new model will give reasonable results when used in the
lattice statics calculations for vacancy clusters. This work is in progress and will be
reported at a later date.
Acknowledgments
The authors acknowledge valuable discussions with Dr J W Martin. The work was
supported by a grant from the UKAEA, Harwell, and the encouragement of Mr J H W
Simmons is acknowledged.
A new force-constant model f or graphite 2305
Appendix
Of the 21 force constants defined in table 1 for model I, 17 are involved in the expressions
for the five independent elastic constants C,,, c66, C,,, C,, and C44. Force constants
C,, C,, C, are related to some of these 17 constants by the Huang condition, ( 5 ) ; the
elastic constants and the phonon dispersion curves are independent of force constant
D,. Three elastic constants are given by
C33 = (c/J 3aZ)(C4 + 6C5 + 3C3,
C44 =(c/J 3a2) [A4 + 3(A5 + Bs) +
+
C13
(-2,~) (D, + E5 - 0;) - C44.
For C, , and C66 the inner displacement contributions (up, yA) are nonzero, and it is
most convenient to use the following equations:
cl, - c66 = (1/J3c)
x [ ( A, - B, ) - 6(A2 - BJ + 4(A3 - B3) + 2(A, - B, ) + ( A; - B; ) ] ,
;(cl1 + c66) = (I 1, 11>f (1/J3c)
x [ A, + B, + 6(A, + B,) + 4(A3 + B3) + 2(A, + B5) + A; + B;].
Let
L = A , + B, + A, + B, ,
X = A , - B, - 2(A3 - B3) + 2(A, - B5) - 2(A; - 89,
M = A, + B, , N = A; + B;,
and
Y = A, - B, - 2(A3 - B, ) - 2(A, - B,).
Then the function (1 1, 11) is given by
J3[(L + 2M + $A4)XZ + ( L + 2M + 4N)Y2 - 2(L - 2M)XYl
(11.11) =
16c[3LM + 4NA4 + ( L + 2M) ( A, + +N)]
For models where second-neighbour out-of-plane interactions are zero, this expression
simplifies considerably to
References
Blakslee 0 L, Proctor D G, Seldin E J , Spence G B and Weng T 1970 J . Appl. Phys. 41 3373-82
Born M and Huang K 1954 Dynamical Theory of Crystal Lattices (Oxford: Clarendon Press)
Brillson L, Burskin E, Maradudin A A and Stark T 1971 The Physics of Semimetals and Narrow-Gap
Dolling G and Brockhouse B N 1962 Phys. Rev. 128 1120-3
Huang K 1950 Proc. R. Soc. A 203 178-94
Krumhansl J and Brooks H 1953 J. Chem. Phys. 21 1663-9
Mani K K and Ramani R 1974 Phys. Stat. Solidi (b) 61 659-68
Nicholson A P P and Bacon D J 1975a Phys. Stat. Solidi (a) 28 613-25
- 1975b Carbon 13 275-282
Semiconductors eds Carter D L and Bate R T (London: Pergamon Press) p 187
2306 A P P Nicholson and D J Bacon
Nicklow R, Wakabayashi N and Smith H G 1972 Phys. Rev. B 5 4951-62
Seldin E J and Nezbeda C W 1970 J. Appl. Phys. 41 3389-400
Tuinstra F and Koenig J L 1970 J. Chem. Phys. 53 1126-30
Yoshimori A and Kitano Y 1956 J. Phys. Soc. Japan 11 352-61
Young J A and Koppel J U 1965 J. Chem. Phys. 42 357-64