Fundamentals of Hydrogels and Gelation

Simon B. Ross-Murphy and Henry McEvoy

Vnilever Research, Colworth Laboratory, Sharnbrook, Bedford MK44 I L Q , UK

(Manuscript received 15 April 1985; accepted 15 Jufy 1985)

A discussion of the properties of that class of materials known as ‘hydrogels’ must make particular Keywords
reference to (i) a working definition of what is meant by a ‘gel’, (ii) the manner in which this relates to the hydrogel, biopolymer, shear
underlying macromolecular structure, and (iii) how these properties are affected by the presence of water modulus, mechanical spectroscopy,
within the macromolecular network. The first point is answered by providing an operational definition of a percolation model, swelling
gel, which depends on the mechanical response of the material to oscillatory shear. This type of behaviour
is shown to be quantitatively related to the degree of connectivity of the network. The chemical aspects of
the problem address the relationship between this and the degree of reaction achieved by those chemical
or physical features responsible for crosslinking. Finally, the swelling properties of such materials are
discussed, and how these are complicated by the non-ideal thermodynamic properties of the solvent
water.

1 INTRODUCTION most aspects of this phenomenon have been well understood

for a number of years. Difficulties arise in the treatment of the
The term ‘hydrogel’ embraces two concepts; an understanding swelling properties of hydrogels due to the thermodynamically
of what materials can be classed as ‘gels’, and the relevance of non-ideal nature of water as a solvent. Attempts to deal with
the presence of water as a constituent of these materials. this problem have been largely empirical, although many inter-
While gel-like properties have been attributed to a consider- esting experimental observations have been reported.
able range of substances, classification as a hydrogel generally
implies the presence of a crosslinked network, whether synthe- 2 MECHANICAL CHARACTERISATION OF GELS
tic or biological in origin. Other papers in this issue, for exam-
ple, concentrate upon synthetic hydrogels for soft contact lens The mechanical properties of a deformable body are described
manufacture and biopolymer hydrogels as structurants for by relating the applied force to the deformation produced.
native and reconstituted biological materials. When the experimental geometry is chosen so that it is possible
to determine the force per unit area (stress) and the fractional
Most of us would claim the ability to recognise a gel, but
deformation (strain), then the constants of this relationship are
sensory evidence alone is rarely sufficient to distinguish
quantitative measures of properties specific to the material of
between a swollen, crosslinked network and a concentrated which the body is composed.
polymer solution. In what follows, a ‘gel’will be defined as the
former, the determination being in terms of its mechanical In general, the ratio of stess to strain is known as a modulus,
response to oscillatory shear, as detailed in the next section. and is a measure of the ‘rigidity’ of the material. Apparent
complications arise, however, if the strain is applied by
This experimental definition is particularly useful, because the oscillating one member of the measuring apparatus
quantity measured, the shear modulus (or more precisely, the sinusoidally. In this case, the measured stress is also sinusoidal,
real part of the complex shear modulus or storage modulus,
but the peak value does not always occur at the same time as
G‘) is intimately related to the connectivity of the polymeric the peak value of strain. It is then usual to measure both the
network. Quantitatively, the modulus is directly proportional phase angle between the applied strain and the measured
to the number of entities which can support stress, elastically
stress, and the amplitude ratio. The storage modulus G’ is then
active network chains (EANCs), here denoted N,. However,
found as the in-phase component of the stress wave, and G ,
as will be discussed in more detail, what constitutes an EANC
the loss modulus is the 90” out-of-phase component.‘
is not always obvious, nor are the methods used to count them
Both G’ and G depend on the frequency of oscillation.
trivial. Hydrogels, in particular, exhibit a wide range of types
However, for an elastic material, i.e. one in which very little
of bonding responsible for crosslinking. In addition to the
mechanical energy is dissipated, G‘, in particular does not
covalent bonds normally found in synthetic networks there are
depend very strongly on the frequency. It is this type of
the somewhat weaker intermolecular Van der Waal’s forces
behaviour which is exhibited by a crosslinked network (Fig.
and hydrogen bonds. Other types of bonding include electro-
l(a)) as opposed to a solution above the critical c* concentra-
static forces, both attractive and repulsive, the intermolecular
tion, where physical chain entanglement effects can occur.
linking of hydrophobic patches, and ion bridging interactions
especially involving multivalent cations. Such non-covalent It should be noted that in this figure, both G’ and the angular
bonding is involved, for example in the formation of globular frequency of oscillation, w (radians s-’) are displayed on a
protein gels, and the formation of multi-helical and other types logarithmic scale. This is a necessary feature of the operational
of ordered junction zones, found particularly with polymers of definition of a gel, that G’ should be relatively independent of
biological origin. Other papers in this issue will consider these frequency over a wide range, and that even at ‘zero frequency’
points more fully. it has a permanent ‘equilibrium modulus’.
From a thermodynamic standpoint, the swelling of a macro- This should be contrasted with the ‘mechanical spectrum’ for
molecular network by a solvent differs little from any other an entangled polymer solution shown in Fig. l(b). At the high
mixing process. It is true that allowance must be made for the frequency end the two have quite similar behaviour; however,
elasticity of the network, which resists the swelling process, but as the frequency is decreased the latter system flows as the

2 BRITISH POLYMER JOURNAL, VOL. 18, NO. 1 1986

 3 GELS AND NETWORKS
The manner in which the above operational definition of a gel
relates to the polymeric network structure requires a brief
discussion of the gelation process, via its basic paradigm, the
so-called f-functional random polycondensation (Fig. Z).5,6
Each molecule capable of participating in network formation
possesses a number of sites which can form crosslinks, here
simply depicted as hooks. The number of such reactive sites
log w
per primary molecule is termed the functionality. For synthetic
polycondensate precursors, this is generally well known, but
biopolymer gelation is often more akin to the vulcanisation of
preformed primary chains, i.e. f i s in general high (say -loo),
and rarely accurately known, if at all.
As crosslinking proceeds, a distribution of species develops
(dimers, trimers, etc.). Each oligomer has progressively more
functional groups capable of reacting, and as ‘aggregation of
aggregates’ occurs, a critical point is reached where one such
aggregate becomes effectively ‘infinite’, i.e. a gel forms. It
should be realised that at this point not all the material is
incorporated into this ‘infinite’ molecule, a proportion of
log w
monomer, dimer, etc. (the sol fraction) remaining uncon-
Fig. 1 Mechanical spectra for (a) a typical crosslinked gel network and (b) nected to the gel fraction. Further, in a finite volume of sam-
an entangled polymer solution (G’ --, G --). ple, the term ‘infinite’ merely implies that the gel fills this
volume. In fact, estimates can, be made of this molecular
weight, from estimates of the weight-fraction distribution of
time-scale of entanglement couplings becomes short compared the sol
to w.2
This practical determination of whether or not the system 4 NETWORK CONNECTIVITY
investigated is crosslinked relies largely on the frequency The relationship between this process and the mechanical
behaviour of G’. For the purposes of the present article, a properties can be illustrated by a crude electrical analogue
detailed discussion of the behaviour of G‘ is not appropriate. (Fig. 3). Obviously, since the battery voltage in the diagram is
Moreover, to calculate the complete concentration/frequency assumed constant, the current will depend on the resistance of
dependence of G’ and G would require a much more com- the network of copper strips. Knowing the resistance of one
plete understanding of the relaxation spectrum for a gel system such strip, this network resistance can be calculated from the
than currently exists. (cf. ref. 3). As far as the loss modulus G known laws of electrical circuit theory. The problem is, how is
is concerned, practically it is - 1 0 ~smaller in magnitude than
G‘ with the similar flat frequency dependence (Fig. l(a)).
During gelacme experiments, when G’ and G are monitored I 1

through the gel point, G“ is often observed to pass through a

maximum. Valles and Macosko4have suggested that this is due
to the presence of ‘dangling ends’ in the reacting network
system.

Percolation on o square l a t t l c e (f.4. d = 2 )

Random resistor Onologue

P
Fig. 2 Illustration off functional random polycondensation c f = 3 ) , (C+O
when a functionality is reacted). The angles are not meant to represent real Fig. 3 Electrical conductivity analogue of percolation on a two-dimensio-
bond angles. nal square lattice. A current will flow only ifp>p,.

BRITISH POLYMER JOURNAL, VOL. 18, NO. 1 1986 3

the current related to the number of strips added to, in this
case, a simple square lattice graph.
In Fig. 3, at least four strips are required to make an electrical
connection. However, if the strips are added at random, on I
average more than four such strips will be added before a
current can flow. This type of statistical mechanical behaviour
is known as a percolation model: in this case on a square
lattice, and the relationship between current and number of
strips or more accurately the proportion of ‘nodes’ connected
by strips, p , is depicted in Fig. 3. Here we have implicitly
defined the critical proportion of connected strips for a current
to flow as p‘. The analogy with gelation is evident - connec-
tivity across the container must be achieved before the net-
work can bear stress.
The analogy can be carried further still, if the number of
copper strips can be related to the fraction of all functionalities
reacted, a. Thus, no gel can form below some critical degree of Fig. 5 Sketch of a section of vulcanisate gel (a) immediately after stretch-
reaction. a,, which may be compared to our p c in Fig. 3. Of ing (b) after relaxation, (c) definition of an EANC, each of the two network
course, in a real chemical system, the flexibility of chains and junction points must have at least three ties (arrows) (d) an active network
rotation about bond angles makes the above model unrealistic, chain (heavy line) in a trifunctional network system. (For details see ref. 7.)
(Reproduced with pcrmission from Pure and Applied Chemisrry, 1975,43,
but a floppy tree-like lattice graph, or Bethe lattice (Fig. 4) is
1.)
the underlying model for the Flory-Stockmayer theory of gela-
tion. This states that a, is related to the functionality, f,by
a, = 1/(y- 1) (1) like chains. As the chains become shorter (or equivalently as
the conformational freedom is reduced), a becomes greater.’
Thus a,=0.5 for f=3, and a gel is formed when, on average, This is of major importance for biopolymer networks.“’
half of the functionalities of Fig. 2 have reacted.
For a network containing a given number of reacted func-
tionalities, the problem is to count how many chains on
average contribute t o its elastic properties by being stress
bearing, i.e. essentially, to relate the modulus to a. While this
calculation is not fundamentally difficult, the procedure is less
obvious than might appear at first sight, as can be illustrated by
referring to Fig. 5.In this diagram, part (a) is a schematic of a
gel network in a state of strain, as depicted by the linkages
being shown as straight lines. Part (b) shows the same network
some time later, when a certain amount of relaxation has
occurred.
The important point is that not only chain ends, but also more
complex of loosely attached material, can relax in this fashion.
It should also be evident that any unattached material (the sol
fraction discussed above) will not contribute directly to the
elastic properties at long times; indeed, this relaxation pro-
F I or y -St o c k m o y er ( f = 4) ceeds very rapidly compared to experimental time-scales.
-
= P e r c o l a t i o n on o Bethe lattice A most useful definition of an elastically active network chain
(Tree-like - no r i n g s )
(EANC) is that due to Scanlan and Case.”.’* First, an active
Fig. 4 Tree-like lattice graph (Bethe lattice). f = 4 in this case, cf. Fig. 2. junction point in a three-dimensional network is one from
which at least three ‘ties’ radiate, where a ‘tie’ is a link which
will lead eventually to the gel surface (Fig. 5(c)). Second, an
5 RELATIONSHIP BETWEEN DEGREE OF REACTION E A N C is a chain segment which has active junction points only
AND MODULUS’ at its two ends (Fig. S(d)). By examination of Fig. 5 it can be
seen that relaxation of chain segments is possible unless these
So far, an analogy has been made between the copper strips of
conditions are fulfilled.
Fig. 3 and the sections of polymeric chain between crosslinks.
In the electrical cases, each copper strip has the same conduc- Now consider the tree-like Bethe lattice of Fig. 4. Since a
tance. In the mechanical case, it might be argued that each fraction of all functionalities have reacted, the probability that
analogous chain segment has the same ‘modulus’, even though a given functionality, chosen at random, has reacted is also a.
they are not necessarily of the same length. However, a reacted functionality does not necessarily con-
stitute a tie, as defined above. Unless at least one ‘sub-tree’
This arises from the statistical mechanics of an isolated chain,
radiating from a given functionality is infinite (i.e. reaches the
which clearly shows that the force required to extend such a
gel surface), the functionality is said to be extinct. The pro-
chain is proportional to the ratio of its extended length to its
bability of this occurring, the extinction probability, U , can be
unperturbed length, i.e. a given force applied to any such chain
calculated in terms of a andf, and is used in the calculation of
produces the same fractional increase in length, regardless of the number of EANCs, N,. For illustrative case f=3, the
the unperturbed dimensions. The constant of proportionality
required result is
is akT, where k is Boltzmann’s constant, T the absolute temp-
erature, and a is a factor which is close to unity for ideal rubber- u = [(1-a)/.]’ (2)

4 BRITISH POLYMER JOURNAL, VOL. 18, NO. 1 1986

and Here the function g(C/Co, f) indicates a functional relation-
ship, which depends only on the quantities C/Co and f.
N,=?h[a(l-u)l3=%[(2a-l)/a]' (3)
The original Herman's treatment, corresponding to large f,
Since the storage modulus G' is insensitive to frequency, its gives a relationship similar to Eqn ( 5 ) , in which the right-hand
value at any suitable frequency can be identified with the side depends on C/C, alone. Indeed, Clark et al. have shown
equilibrium shear modulus, G,, discussed in elasticity theory, that this treatment actually gives a better fit to experimental
which is calculated from Eqn (3) via data than the more exact cascade approach." However, they
Ge=(Ne/Vrn,J aRT (4) do point out a number of problems which neither approach
attempts to address, at least not in the simplified form given
where V,,, is the volume per mole of primary chains, the other here. In terms of network topology no allowance has been
quantities taking the meaning ascribed previously. By examin- made for the possibility of cyclisation (for example, if two of
ing Fig. 4,it is also easy to see how, in principle, we obtain the the functionalities in Fig. 4 had reacted in such a way as to
mechanical spectrum of Fig. l(a); as well as the infinite paths produce a closed ring, rather than a continuation of the tree-
of the gel system, there are very many sub-paths of different like structure) Second, it is debatable whether a true state of
lengths from each node of the lattice graph to any other node, chemical equilibrium is ever attained (e.g. the modulus of
each of which will have a particular relaxation time of the gelatin gels is known to increase slowly over a period of
Rouse type.' weeks).
Formulae for more complex models of crosslinked primary The exact form of Eqn (5) is such that, for systems of the same
chains are derivable, although sometimes at the expense of functionality, plots of the log G against log C should be super-
more lengthy calculation, via for example, the cascade theory. 3osable on a 'master' curve by horizontal and vertical shifts of
amount log Co and log K , respectively. If the original Her-
mans formulation is used, all such curves should be super-
6 RELATIONSHIP BETWEEN MODULUS AND posable, regardless of functionality. This superposition is
CONCENTRATION demonstrated for a series of ten gels of bovine serum albumin
(BSA) in Fig. 6.16In this experiment, the equilibrium constant,
The procedure so far may be paraphrased thus: from the K , and hence C,, was varied by altering the thermodynamic
known degree of reaction and the functionality of the mono- conditions, viz. pH and salt content (cf. Eqn (7)). Note that CO
mer units calculate the extinction probability, hence the num- for these samples varies from 1% to 10% w/w.
ber of EANCs, hence the modulus. For a=a,, u is defined to
be unity, and thus from Eqns (3) and (4)G=O. This procedure Many authors have reported a power law dependence of
may be applicable when the polymeric network forms from an modulus on concentration, G K C " . ' ~ value
~ ~ Tn=2
~ ~has been
initial population of monomers, and the degree of reaction is frequently cited for gelationlSz1following the work of P00le.l~
known for example by titrating end groups for a simple For other systems exponents of 3-4,22 and even as high as 723
polycondensation.
However, the formulation of biopolymer gels is often
accomplished by heating and/or cooling aqueous solutions.
Examples might include the formation of a gelatin gel when a
concentrated solution is cooled,6 or the heat-set behaviour
observed for solutions of many globular protein^.'^,'^ In these
cases, while the gel modulus can be determined, the only other
information available which might be related to network struc-
ture is the concentration of polymer in the pregel solution. I

HermansI5approached this problem by treating the interaction

of two functionalities to form a crosslink as a dimerisation,
describable as a chemical equilibrium characterised by an
equilibrium constant, K . In this manner, it was possible to /
determine the degree of reaction, a , solely in terms of K and
the concentration, C. This implies the existence of a critical
concentration, C,, corresponding to a,, below which no gel will
form. At the same time, the equilibrium constant, K, is
assumed to depend upon temperature and solvent conditions,
but not on concentration or molecular weight. Hermans in fact
assumed a given number of reactive sites per repeat unit of the BSA gels
primary chain, i.e. the functionalityf, of the chain as a whole
mM NoCl pH6.0 pH7.0 pH8-0
was taken to be proportional to the molecular weight, M ,and
was hence large. The method has been generalised to any I00 O A -
functionality, but retaining the idea of a dimerisation 10' O A -
equilibrium. The final result can be paraphrased thus (for
I02
details see ref. 6): O A O

c (f- I)%( 1-u)'( 1-p)

G ' K ,=g( CICO, f)=-{
CO 2cf-2)
I I
10' I02 I 13
where K,=scale factor= K(f-l)(f-2)/aRT, v=extinction c/c,
probability and ,8=(f-l)av/(l-a+av) [ref. 71 and
Fig. 6 Shear modulus data for heat-set BSA globular protein gels reduced
Co=critical concentration =MU- 1)/ Kf (f-2)2 (6) to master curve form. For details see ref. 16.

BRITISH POLYMER JOURNAL, VOL. 18, NO. 1 1986 5

have been reported. Examination of Fig. 6 illustrates that this equilibrium swelling, the equilibrium swelling ratio, qm
may be merely a consequence of considering limited ranges of (= V / V o )is given by
concentration. In practice, for C/C,>lO, the Hermans for- . I

mulation does approach a C dependence, but as the gel point (9)

is approached from above, the slope of this log-log plot
becomes nominally infinite, so that the values of n in the range, In this e x p r e s s i ~ n N
, ~, ~is the number of EANCs per unit
say 2-8, reported above depend on the concentration range volume, which, as already shown, is intimately related to but
accessed. not equal to the density of crosslinks, and f(x) gives the
Lower exponents than 2 have also been observed,24and the thermodynamic terms, including the y, for water/polymer
observation has been rationalised using the lower functionality interaction. Thus, the greatest degree of swelling for given
cascade method.2s solvent conditions corresponds to the most lightly crosslinked
system, the complete absence of crosslinks implying infinite
swelling (i.e. complete dissolution).
7 LARGE DEFORMATION PROPERTIES
The arguments expounded so far, relating modulus to network
connectivity are only valid when the deformation is ‘infinitesi- 9 GELS SWOLLEN BY WATER (HYDROGELS)
mal’, i.e. changes in sample dimensions are small compared A number of additional complications arise when the solvent is
with the dimensions themselves. The large deformation stress- water. O n e is the most important considerations stems from
strain behaviour of rubbery networks has been based upon the fact that many of the polymers which form hydrogels are
classical rubber elasticity or variants thereof. For example, the polyelectrolytes. T h e mutual electrostatic repulsion of the
Mooney-Rivlin equationz6 charges on the polymer chains might b e expected to lead to
vastly greater swelling than is observed for a neutral polymer
o,=2(C,A+C,)(A- 1/P) (7) of similar topology. In practice, however, the presence of
describes deviations from rubber elasticity through the C, counter ions within the swollen network will screen the effect
term. Here aT=true stress (=Lo),L the tensile elongation of this repulsion, leading to lower swelling than might other-
ratio, and C, and C2 are coefficients. When C2+0, Eqn (7) wise be anticipated. For sufficiently great ionic strength,
reduces to that of classical rubber elasticity, and the shear ‘neutral’ behaviour should be observed, as for uncharged
modulus G+2C,. Experiments for swollen networks have con- networks.
firmed that in the limit of high swelling (or large volume
fraction of solvent) C2=0.26The molecular theory of rubber Various attempts have been made to quantify this behaviour,
elasticity has continued t o develop over the last few years, and but the thermodynamics of water/polymer systems is an uncer-
a review by FloryZ7puts much of the recent work into context. tain subject, even without the additional complications caused
Most literature deals with covalent networks of Gaussian by polyelectrolyte effects. Published values of y, for the same
chains, and deviations from ideal rubber elasticity are polymer system often vary considerably, and quantitative
attributed to constraints on chain movement^.^'-^^ The agreement with theory is not good unless additional terms are
Mooney-Rivlin equation is purely phenomenological and included in the swelling relationships. Nonetheless, the
suffers from a number of disadvantages, most notably that the degree of swelling for certain polyelectrolyte systems would
CI, C, coefficients take different values for different geome- seem to be unaccountably low, leading some authors to pro-
tries (i.e. tension, compression). The equation of Blatz, pose the introduction of additional crosslinks due to cation-
Sharda and Tschoeg130 (also phenomenological) usually specific bridging interactions. At the same time, the practical
returns the same constants for both experimental geometries uses of hydrogels frequently involve systems of gel beads. The
and may be written analysis of the swelling of such systems must be approached
with some caution. The diffusion of water into the pores
between beads, which must precede swelling of the beads
themselves, will be greatly affected by surface energy terms
(capillarity). In such cases, the attainment of a true swelling
for a tensile test. equilibrium may take a very long time.””.‘
For elastomers, n is always found in the range 1.8-2.0 (n=2 Perhaps the most interesting phenomena in this area are those
returns the classical equation). Higher values than 2 have been recently observed for polyacrylamide hydrogels by Tanaka
reported for biopolymer gels,31 and interpreted as a conse- and co-workers.3s.36The crosslinked polacrylamide has some
quence of the ‘stiffer’ nature of the constituent network chains. of its amide groups hydrolysed (pH 12) and is swollen in water.
Different amounts of acetone are added successively to the
8 SWELLING OF RUBBER-LIKE GELS IN NON-POLAR hydrogel sample immersed and swollen in excess water and the
SOLVENTS system allowed to re-equilibrate. A t first there is little effect,
and then above a particular percentage of acetone there is a
When an uncrosslinked polymer is mixed with a non-polar sudden and discontinuous de-swelling. For extensively hydro-
solvent, the free energy of mixing is expressed in terms of lysed gels, this occurred at -25% volume acetone, resulting in
various thermodynamic quantities, including the parameter -350-fold decrease in volume. It has been suggested that this
which represents the polymer-solvent interaction terms, the effect could be exploited, for example, in contact switches and
Flory-Huggins parameter x . ~ actuators. Very recently,36essentially the same effect has been
For B s y s t e m s ~ i sof
, course, 0.5. For crosslinked networks, it observed for polyacrylamide gels immersed in Cu (11) salt
is assumed that the mixing terms for polymer and solvent solutions, above a critical value of the C u (11) concentration.
remain essentially unaltered and the network swells in an
attempt to maximise the number of polymer-solvent contacts.
10 ACKNOWLEDGEMENTS
However, the swelling process is resisted by the elasticity of the
network, and a state of equilibrium is eventually reached. If V , We are grateful t o our colleagues Allan Clark, Bob Richard-
is the volume of the unswollen polymer, and V the volume at son and Lesley Linger for discussions and technical assistance.

6 BRITISH POLYMER JOURNAL, VOL. 18, NO. 1 1986

References 18 Hirai, N., Bull. Inst. Chem. Res. Kyoto, 1955,33, 31.
19 Poole, H.J., Trans. Farad. SOC.,1925, 21, 114.
1 Ferry, J.D., Viscoelastic Properties of Polymers, 3rd edn, 1980, New 20 Ferry, J.D., Adv. Protein Chem., 1948, 4, 1.
York: John Wiley and Sons. 21 Te Nijenhuis, K., Colloid Polym. Sci., 1981, 259, 322.
2 Graessley, W.W., Adv. Polyrn. Sci., 1974, 1. 22 Walter, A.T., J. Polym. Sci., 1954, 13, 207.
3 Kastner, S., Polymer, 1979, 20, 1327. 23 Ellis, H.G. & Ring, S.G., Carbohydr. Polymers, 1985, 5 , 201.
4 Valles, E.M. & Macosko, C.W., Macromolecules, 1979, 12,521. 24 Nelb, G.W., Kamykowski, G.W. & Ferry, J.D., Biophys. Chem.,
5 Flory, P.J., Principles of Polymer Chemistry, 1953, Ithaca, New York: 1981, 13, 15.
Cornell. 25 Clark, A.H. & Ross-Murphy, S.B., Adv. Polym. Sci., to be published.
6 Peniche-Covas, C.A.L., Dev. S.B., Gordon, M., Judd, M. & 26 Treloar, L.R.G., The Physics of Rubber Elasticity, 2nd edn. 1970,
Kajiwara, K . , Farad. Discuss. Chem. Soc., 1974, 57, 165. Oxford: Clarendon.
7 Gordon, M. & Ross-Murphy, S.B., Pure Appl. Chem., 1975,43,1. 27 Flory, P.J., Br. Polym. J . , 1985, 17, 96.
8 Stauffer, D., Coniglio, A. &Adam. M., Adv. Polym. Sci., 1982,4,103. 28 Edwards, S.F., Br. Polym. J., 1985, 17, 122.
9 Mark, J.E., Adv. Polym. Sci., 1982,44, 1. 29 Errnan, B., Br. Polym, J . , 1985, 17, 140.
10 Burchard, W., Br. Polym. J . , 1985, 17, 154. 30 Blatz, P.I., Sharda, S.C. & Tschoegl, N.W., Trans. SOC.Rheol., 1976,
11 Scanlan, J., J. Polym. Sci., 1960, 43, 397. 18, 156.
12 Case, L.C., 1. Polym. Sci., 1960,43, 501. 31 McEvoy, H., Ross-Murphy, S.B. & Clark, A.H., Polymer, 1985, 26,
13 Richardson, R.K. & Ross-Murphy, S.B., Int. J . Biol. Macromol., 1981, 1483, 1493.
3, 315. 32 DuSek, K., J . Polym. Sci. (Symp)., 1973, C42,701.
14 Clark, A.H. & Lee-Tuffnell, C.D., in Functional Properties of Food 33 Peska, J., Stamberk, J., Hradil, J. & Ilavskq, M. J., Chromatogr., 1967,
Macromolecules, eds Mitchell, J.R. & Ledward, D., in press, Barking: 32, 2766.
Applied Science. 34 Hofak, D., Sqec, F., Ilavskq, M., Bleha, M., Baldrian, J. & Kalal, J . , J .
15 Hermans, J., J. Polym. Sci. A , , 1965, 3, 1859. AppL Polym. Sci., 1981, 26,411.
16 Clark, A.H. & Ross-Murphy, S.B., Br. Polym. J . , 1985, 17, 164. 35. Tanaka, T., Sci. A m . , 1981,244, 110.
17 Clark, A.H., Richardson, R.K., Ross-Murphy, S . B . & Stubbs, J.M., 36 Ricka, J. & Tanaka, T., Macromolecules, 1985, 18, 83.
Macromolecules, 1983, 16, 1367.

BRITISH POLYMER JOURNAL, VOL. 18, NO. 1 1986 7