Polymer International Polym Int 52:479–485 (2003)

DOI: 10.1002/pi.1203

Review
Defoamers: linking fundamentals to
formulations†
Robert Pelton1∗ and Ted Flaherty2
1 McMaster Centre for Pulp and Paper Research, Department of Chemical Engineering, JHE-136, McMaster University, Hamilton, Ontario,
Canada, L8S 4L7
2 Dorset Industrial Chemicals, 2725 Ford Blvd, Chateauguay, QU J6J 4Z2, Canada

Abstract: Antifoam (defoamer) chemicals are a crucial part of many industrial processes. Reviewed are
the types of defoamers and their mode of operation. Emphasized are the properties and location of the
hydrophobic particles which are the active foam breaking agents. The roles of surfactants in spreading,
defoamer emulsification and hydrophobic particle orientation are also discussed. Finally, a simple model
is described which simulates foam growth as functions of defoamer concentration, air hold-up, reactor
volume and air flow rate.
 2003 Society of Chemical Industry

Keywords: defoamer; antifoam; formation foam control; defoamer modelling

INTRODUCTION FOAM BASICS

Many commercial processes depend upon antifoam Macroscopic characteristics
addition to prevent excessive foam build-up. Two The physical chemistry of foams has fascinated physi-
important examples are fermentations and brownstock cists, chemists and children for generations. Although
washers in kraft wood pulp production. A large kraft most modern texts on colloids and surface science4
pulp mill will spend in excess of a million dollars a do a good job of describing foam fundamentals, the
year on defoamers. The defoamer effect is illustrated classic book by Boys, published first in 1911 and
in Fig 1 which shows laboratory measurements of now available in a Dover edition,5 is worth reading.
wash water flux through pulp pads.1 Pure water gave The major features of foam behaviour are well under-
the highest flow rate whereas, with surfactant, stable stood, and the very recent scientific literature deals
air bubbles blocked the pad, slowing drainage rates with unresolved issues, such as understanding what
and thus washer productivity by a factor of 10. The gives Champagne bubbles their unique characteristics6
negative effects caused by the presence of surfactant- or what forces enable solid particles to attach to air
stabilized air were partially eliminated by the presence bubbles.7
of defoamer. For many aqueous processes, such as fermentations,
The goal of this paper is to review the scientific it is convenient to divide the reactor into two zones in
basis of antifoam technology. The types of commercial which the dispersed air properties are very different.
antifoams are surveyed with emphasis on current The heart of the fermentor is the liquid zone where the
thinking on how antifoams function. Finally, a fermentation processes occur and in which dispersed
simple mathematical model for antifoam action is air is present as individual air bubbles, providing
discussed which may be useful for process models a source of oxygen. Mixing causes bubble/bubble
for design and/or control. The review is not exhaustive collisions. If during a collision, the thin film or lamella
but instead highlights recent results and is biased separating two bubbles ruptures, two bubbles become
towards our work. For more extensive reviews, see one. If the lamella does not rupture, the colliding
Garrett’s book, published in 1993.2 A much earlier bubbles will separate in flow.
review by Bryant summarizes foam control issues in The second zone is the foam phase which, like
fermentation.3 Although mechanical foam breaking the head on a glass of beer, has a high volume
devices exist, these are not considered in this work. fraction of air, is relatively stagnant and the lamellae

∗ Correspondence to: Robert Pelton, McMaster Centre for Pulp and Paper Research, Department of Chemical Engineering, JHE-136,

McMaster University, Hamilton, Ontario, Canada, L8S 4L7

E-mail: [email protected]
† Oral presentation—Paper presented at the Formula III Conference: New Concepts and strategies in formulation: from laboratory to

industry, 13–16 October, 2001, Hérault France

(Received 8 October 2002; accepted 20 December 2002)

 2003 Society of Chemical Industry. Polym Int 0959–8103/2003/$30.00 479

R Pelton, T Flaherty

Mass of filtrate (kg m-2) 600 bubbles whereas rupture to the headspace lowers the
Vacuum total foam volume. If the rate of bubble rupture to the
500 500 torr headspace is slower than the rate of air injection into a
150 torr Water
400 fermentor, the foam phase volume increases with time.
300 Detergent By contrast, if bubble rupture is fast relative to air injec-
& defoamer tion, there will be little, if any, visible foam phase. The
200 role of defoamers is to increase the rate of bubble coa-
100 Detergent lescence, bubble rupture and release to the headspace.
0
0 10 20 30 40 50 60 Microscopic (chemical) properties of foams
Time (s) Colloid science textbooks tend to treat foams as
one-dimensional colloids since the thicknesses of
Figure 1. Water drainage rate through a pad of wood pulp fibers
foam lamella are in the colloidal (nano) size range.
(with permission from Reference 1). Drainage rates were ten times
slower with detergent because of the presence of air bubbles Indeed, many of the concepts of colloidal stabilization,
whereas defoamer drainage was faster. such as the importance of double layer, steric
and van der Waals interaction, operate in foams.
Lamella Electrically charged species at the lamella surface exert
a repulsive force on the corresponding charged species
on the opposite lamella surface—this effect opposes
lamella drainage and thus is a foam stabilizing factor.
Plateau Border
or Strut Similarly, hydrated polymeric species adsorbed on one
air
lamella surface will repulse a similar layer on the
opposite surface. Colloid scientists call this effect steric
stabilization and, like electrostatic effects, it inhibits
lamella thinning and thus promotes foam stability.
By contrast, the ever-present van der Waals attractive
Figure 2. Intersection point of three bubbles in a foam phase. forces tend to promote foam lamella thinning.
The phenomena described in the last paragraph
are long-lived. Such concentrated foams have been are essentially static or equilibrium processes. In fact,
studied frequently. The key structural elements are foam stability is dominated by dynamic phenomena.
thin films, called lamellae, separating the bubbles, and For example, viscoelastic lamellae often obtained with
junction regions called Plateau borders where lamella adsorbed proteins, are far more robust than inelastic
are joined (see Fig 2).8 Plateau identified the following lamellae obtained with simple adsorbed surfactants.
two rules controlling foam geometry. The rules arise Products formulated for high foam stability, such as
because foams attempt to minimize the total air/water dish-washing detergents, contain mixtures of surface-
interfacial area. active agents which give viscous, elastic lamellae.
(1) Along an edge in a foam structure, three and only
three lamellae meet; the three lamellae are equally
ANTIFOAMS
inclined to one another all along their edges to
The chemical technology and patent literature contain
give a dihedral angle of 120◦ .
many references to both antifoams and defoamers.
(2) At a point, four and only four edges meet; the four
In some cases the implication is that these are
edges are equally inclined to one another in space;
different chemicals doing different jobs in different
hence the angle at which they meet is 109◦ .
ways. Our view is that antifoams and defoamers are
Foams undergo two important processes: water fundamentally the same. Perhaps the names reflect
drainage and bubble rupture or coalescence. Drainage an application philosophy; defoamers are added when
is the flow of water from the foam phase and is driven foam appears whereas antifoams are added to the
by gravity and by curvature-induced pressure gradi- broth to prevent foam accumulation. Herein the terms
ents (Laplace pressure). The literature contains many antifoam and defoamer are used interchangeably.
experimental and theoretical descriptions of foam Antifoams are found in nearly every water-based
drainage which are beyond the scope of this review. product and process including applications as diverse
The most important feature of foam drainage is that as soft drinks, anti-acid pills and laundry detergents.
it is a slow process compared with the acceptable The spectrum of antifoam suppliers is broad. Fortune
residence time of a foam bubble in a fermentation 500 companies compete with very small formulators.
operation or in a brownstock washers. In other words, Often small suppliers focus on specific markets
we cannot wait for foams to drain in most industrial whereas major suppliers are active across the board.
processes. Modern high performance antifoams are formulated
Bubble rupture occurs when the lamellae separating products containing many ingredients. When consid-
two bubbles or a single bubble and the headspace rup- ering defoamer types, the major distinctive property is
tures. Bubble/bubble coalescence leads to fewer, larger the solubility of the antifoam in water. Those which are

480 Polym Int 52:479–485 (2003)

 Defoamers: linking fundamentals to formulations

soluble in cold (relative to the process or application to the air bubble. The defoamer formulator must
temperature) water are sometimes called water-based achieve a delicate balance when choosing surfactants
or ‘cloud point’ antifoams.9 The latter name is a ref- for the defoamer. Surfactants are required to generate
erence to their proposed mechanism of operation (see and colloidally stabilize the oil-in-water defoamer
subsequent sections for mechanistic discussions). emulsion; however, if the surfactant is too effective
Water insoluble defoamer, herein called ‘oil-based’ at stabilizing the emulsion, the oil droplets will not
is the other major type. Oil-based antifoams are adhere to the air bubbles.
marketed as oil-in-water emulsions or as concentrates Upon adhering to an air bubble, spherical emulsion
(sometimes called compounds) which are dispersed droplets deform to give lenses on the water/air
in the process stream, which is the broth in the case interface. Figure 4 shows the cross-section of a dyed
of fermentations. Mineral oil and/or silicone oil is defoamer lens measured by confocal laser scanning
the major component, whereas the foam breaking microscopy.10 For these experiments, large (100-
active ingredient is a hydrophobic (water-hating) solid µm) defoamer oil droplets were placed on the top
which may be surface treated silica, wax particles, a surface of gelled surfactant solution. This experimental
silicone microgel, or a combination. Also present are configuration was a model for a lens on an air bubble.
surfactants. The functions of the antifoam components The gel immobilized the lens, permitting the imaging.
are discussed in subsequent sections. The shape of the lens (see Fig 4) reflects the balance
Selecting a defoamer for a given process or of surface tension and gravitation forces. The lens
product is ultimately an exercise in trial and error. has two interfaces (air/oil and water/oil) and a three-
However, the choices can be limited by considering phase contact line where air, oil and water meet. Most
the active solids contents, and the cost and efficacy of mineral oil-based defoamers contain small amounts of
commercial defoamers, which can span a broad range. free silicone oil introduced with the hydrophobic silica.
The oil-based defoamers usually have the highest Silicone oil is insoluble in water and mineral oil, so the
efficacy; however, oil-based defoamers ultimately lens shown in Fig 4 is likely to include a microscopic
contaminate processes and products because oil silicone oil phase, most probably near the air/mineral
and hydrophobic particles are not compatible with oil/water contact line. Therefore a defoamer lens is
water. By contrast cloud-point defoamers dissolve a complex structure containing as many as four fluid
when the process temperature is lowered. Moreover, phases (air, water, mineral oil, silicone) and a solid
cloud-point defoamers are not usually as effective. phase (silica or wax).
Requirements for food grade status decrease the Since the hydrophobic particles are the active foam
choices, although both oil-based and cloud-point food breaking agents, the location of the particles in the
grade products are available. defoamer lens is a crucial question. We used confocal
microscopy to view the position of fluorescent-labelled
hydrophobic silica in the defoamer lens.10 Figure 5 is
ANTIFOAM MECHANISMS an image looking down onto a defoamer lens on a
The essential function of a defoamer is that it must gelled surfactant solution. The dark domains are the
enhance the rate of lamella rupture. The addition fluorescent silica particles. Optical sectioning of the
of an effective defoamer to a fermentor will increase lens with confocal microscopy showed that most of
the average bubble size in the liquid phase due to the hydrophobic silica was on the oil surface near
the increased bubble coalescence rate as well as the
decrease in the volume of the foam head. Although Air
100 µm
there is a variety of defoamer types, all aqueous
defoamers contain surface active agents, contain or
form water insoluble phases, and they ‘wear out’, Defoamer
often requiring some form of continuous addition.
The following sections present a view of the defoaming
mechanism. Aqueous
Commercial defoamers are complicated formulated
Figure 3. The cross-section of a defoamer lens sitting on gelled
products containing surfactants, oil and one or surfactant (with permission from Reference 10).
more types of hydrophobic particles with an average
particle size in the range 1–10 µm. For aqueous
applications the defoamers are either added as water-
in-oil emulsions or as compounds which emulsify in
the process. Defoamer emulsions are coarser with
average emulsion droplet sizes in the range 5–50 µm.
The first step in the defoaming mechanism is the
collision of an emulsion droplet with an air bubble
resulting in the deposition of the defoamer onto the
bubble surface (see Fig 3). Not every collision leads
to the attachment of the defoamer emulsion droplet Figure 4. Step one: formation of a defoamer lens in the liquid phase.

Polym Int 52:479–485 (2003) 481

R Pelton, T Flaherty

rapidly adsorb components from the solution

which render them hydrophilic.
(3) Effective defoamers spontaneously spread to cover
the air/water interface with a very thin oil
layer.13,14 The remainder of the oil is present
in the lens.
(4) Hydrophobic particles smaller than 1 µm and
larger than 10 µm diameter are not very effective.
Presumably the former are too small to have long,
lamella-penetrating asperities whereas very large
particles are ineffective because it is not practical
to have them present in sufficiently large numbers.

Any successful defoaming mechanism must explain

Figure 5. Top view of a defoamer lens sitting on gelled surfactant these observations. Most of the early work concen-
(not visible). The dark objects are fluorescently labeled hydrophobic trated on molecular mechanisms in the thinnest part of
silica sitting on the aqueous oil interface (with permission from
the lamella—these however failed to explain many of
Reference 10).
the observations. Koczo and coworkers proposed that
defoamer lenses are too large to enter the lamella but
the three-phase contact line. For foam breaking it instead function in the plateau borders15 . In an effort to
is crucial that the hydrophobic particles migrate to verify this mechanism we conducted two-dimensional
the oil surface. By changing or removing surfactants, foam-breaking experiments in which a single layer of
it was possible to produce defoamer compositions soap bubbles on the surface of a liquid was observed
where the silica remained inside the oil phase—these under a microscope10 . Figure 7 shows some frames
compositions were ineffective defoamers.10 from a video. The defoamer emulsion contained an
The fundamental foam breaking event occurs when oil-soluble dye to enhance the observations. The first
two bubbles collide and coalesce because atleast frame shows three bubbles, two defoamer lenses and
one of the bubbles contains a defoamer lens. This defoamer emulsion droplets in the liquid phase. With
is illustrated schematically in Fig 6 which shows time, the two lens moved together at the edge of the
three bubbles in a foam phase coalescing into one. lamella in what would be the Plateau border of a
Bubble/bubble coalescence can also occur in the liquid three-dimensional foam. Between the third and fourth
phase, resulting in a larger average bubble size than pictures, the defoamer lens caused the small bubble
would be present without defoamers. For a given to coalesce with the larger one. We can only guess at
volume fraction of dispersed air (hold-up), the larger the key events which are sub-microscopic. It seems
the bubble size, the lower the air/water interfacial area reasonable to speculate that a lens with a surface cov-
and thus the lower the oxygen transport flux. ered with rugged hydrophobic particles bumps into
The microscopic details leading to lamella rupture the opposing lamella surface and, like a submerged
have been the subject of many experimental and iceberg hitting a ships hull, the result is catastrophic
theoretical investigations.2 There are a number of key rupture; this is illustrated in Fig 8.
experimental observations pertinent to the defoaming
mechanism. These are:
(1) The shape of the hydrophobic particle is impor- 100 µm
tant—smooth particles are ineffective whereas Defoamer
Lenses
particles with asperities and sharp edges break Air
foams.11
(2) Hydrophobic particles are much more effective Air Defoamer
when suspended in mineral or silicone oil.12 Pre- Air
Droplet

sumably bare, unprotected hydrophobic particles A) Time = 0 s B) Time = 5 s

C) Time = 9 s D) Time = 10 s

Figure 7. A sequence of captured video frames showing the

migration of a defoamer lens into the lamella between two air bubbles
Figure 6. Bubble coalescence: the foam-breaking step. causing their coalescence (with permission from Reference 10).

482 Polym Int 52:479–485 (2003)

 Defoamers: linking fundamentals to formulations

defoamer emulsion droplet onto a bubble to give a

Lamella lens (see Fig 3). The elements of the model are shown
in Fig 9.16 Gas enters the liquid phase as ‘primary
Hydrophobic bubbles’ which are assumed to have a radius, Rp .
Particle Piercing Also present in the liquid phase are antifoam emulsion
Lamella droplets which are assumed to be colloidally stable
Oil
with respect to homocoagulation, and monodisperse
in terms of particle size. Each emulsion droplet is
assumed to contain at least one hydrophobic particle
Water capable of inducing lamella rupture.
It is assumed that the foam above the liquid phase
Air contains two types of bubbles: primary bubbles from
the liquid phase, and secondary bubbles formed by the
coalescence of the primary bubbles.
The primary bubbles in the liquid phase are
Figure 8. An illustration of defoamer induced lamella rupture. assumed to collect antifoam emulsion droplets by
heterocoagulation. Upon entering the foam phase
The mechanisms by which cloud-point defoamers those primary bubbles carrying at least one defoamer
function is unknown.9 One might speculate that lens are then assumed to coalesce with neighbouring
a surfactant coacervate phase forms like an oil- primary bubbles yielding a secondary bubble with
based defoamer lens which can rupture lamellae. radius Rs . The secondary bubbles are assumed to
However, there are no ragged hydrophobic surfaces be large enough to rise through the foam phase and
in a coacervate. Perhaps they function at a molecular rupture in the headspace. Thus, only primary bubbles
level by displacing the foam stabilizing surfactants and which do not form secondary ones contribute to the
polymers from the air/water interface. volume of stable foam.
The mathematical formulation of the model is
based on statistical arguments. For example, if the
PROCESS MODELS secondary bubbles consist of 50 primary bubbles, then
Fermentations are particularly challenging applica- the probability of a secondary bubble not forming is
tions for antifoams—airflow rates are high and many the probability that none of 50 primary bubbles collect
fermentation broths contain proteins and carbohy- a defoamer lens while traveling through the liquid
drates which are excellent foam stabilizers. Obviously phase.
the antifoam must not be toxic for microorganisms. In The mathematical derivation has been published
some cases there is a potential for antifoams to inter- elsewhere16 and the main equations are:
fere with oxygen transport. Two obvious mechanisms
t
for this are: V (t) = R(t)dt (1)
t=0
(1) The average bubble size in the broth may increase
and gas hold-up can decrease due to antifoam- where V (t) is the total volume of foam and R(t) is the
induced bubble coalescence. The consequence is rate of foam rise (m3 s−1 ) given by
a lower total air/solution interfacial area which in
 
turn gives lower oxygen transport rates. 1 δ
(2) Oil or surfactants from the defoamer could be R(t) = r 1 − (2)
g
present as a spread film on the bubble surface
inhibiting oxygen diffusion.

Strategies are needed for efficient defoamer addi- foam

tion. Simply adding excessive defoamer is not only phase secondary gas
bubble
expensive, it has the potential to inhibit productivity.
Choice of the optimum defoamer addition strategy
is an exercise in process optimization and process
liquid antifoam emulsion
control, which are subjects beyond the scope of this phase droplet
review. However, it is relevant to consider simple,
primary gas
semi-empirical models for defoaming which could be bubble
incorporated into a fermentation process model.
Defoaming is a kinetic process requiring kinetic
frit
models. In particular, a realistic kinetic model must
describe the slowest, rate determining process. We
developed a model which predicts foam volumes gas
as functions of time. The key assumption was that
the rate determining step was the deposition of a Figure 9. Schematic diagram illustrating elements of model.

Polym Int 52:479–485 (2003) 483

R Pelton, T Flaherty

in which r is rate of air flow into the fermentor 0.9

(m3 s−1 ); δ is number of defoamer droplets captured

Foam volume (l)

by an air bubble over a period of one bubble residence 0.7
time (z seconds) of bubbles in the liquid phase; and 0.5 [AF] = 25
mg l-1
g is the maximum number of secondary bubbles
0.3
which could be created from the dispersed air in [AF] = 500
the liquid phase. δ is given by the following standard 0.1 [AF] = 0 mg l-1
heterocoagulation kinetics theory where the constant -0.1
K is the product of the concentration of primary 0 500 1000 1500
bubbles and the coagulation rate constant. E0 is the Time (s)
initial concentration of defoamer emulsion droplets
Figure 11. Simulated foam rise curves using eqns 1–4 varying only
and VL is the total liquid volume: the antifoam concentration and time (This was redrawn from
Reference 16).

δ = (exp(−K (t − z)) − exp(−Kt))E0 VL (3)

that our model may have some utility in process
analysis.
The parameter g is given by the following equation
where G is the volume fraction of air bubbles in the
liquid phase (ie the hold-up):
CONCLUDING REMARKS
VL G Antifoam technology is sophisticated and evolving.
g= (4) The major mechanisms are understood; however,
4
π R3s the detailed mechanistic contributions of some active
3
ingredients are not understood. Simple issues such
as the average number of bubbles ruptured per
The model was evaluated by comparison with hydrophobic particle are not understood. Consider
simple laboratory foam rise experiments in which 1 litre of dispersed air in a fermentor. If the average
foam volumes were measured as functions of time bubble diameter is 2 mm, the corresponding number
with a constant air flow rate. Figure 10 shows typical of bubbles is 239 000. If one antifoam emulsion droplet
foam volume versus time curves for various defoamer destroys two bubbles, the required volume of antifoam
concentrations. With low defoamer levels, the foam to handle 1 litre of air is 6 × 10−8 l of antifoam.
volume increased linearly whereas high defoamer Since few commercial defoamers are this efficient,
concentrations gave little foam initially; however, the there is scope for great improvements in defoamer
defoamer eventually was depleted and the foam growth efficacy.
accelerated. Formulation issues are particularly complex when
The model contains only two unknown and thus antifoams are included. Effective antifoams must
adjustable parameters, K and Rs . However these be sufficiently colloidally stable not to aggregate
are independent of defoamer concentration so all of or sediment. In contrast, if the dispersed defoamer
the curves in Fig 10 were fitted with a single value particles are too stable, the defoamer will not interact
of these parameters. Figure 11 shows the simulated with air bubbles and so will be inactive.
curves reflecting different values for E0 , the initial
defoamer concentration. Most of the main features
of the experimental curves were obtained suggesting
REFERENCES
1 Wang J, Pelton R, Hrymak AN and Yongmoon Kwon Y, New
insights into dispersed air effects in brownstock washing.
1000 Tappi 84:101–108 (2001).
50 2 Garrett PR, The Mode of action of Antifoams, in Defoaming
100 mg l-1 Theory and Applications, ed by Garrett PR, pp 1–118, Marcel
Foam volume (ml)

750 300 mg l-1 Dekker, New York (1996).

3 Bryant J, Anti-foam agents. Methods in Microbiology 2:187–203
(1970).
500
4 Evans DF and Wennerström H, The Colloidal Domain,
500 mg l-1 pp 591–597, Wiley–VCH (1999).
250 5 Boys CV, Soap Bubbles Their Colors and Forces Which Mold Them,
Dover, New York (1958).
6 Peron N, Cagna A, Valade M, Bliard C, Aguie-Beghin V and
0 Douillard R, Layers of macromolecules at the Champagne/air
0 1000 2000 Interface and the Stability of Champagne Bubbles. Macro-
Time (s) molecules 17:791–797 (2001).
7 Ralston J, Fornasiero D and Mishchuk N, The hydrophobic
Figure 10. Foam volume versus gas flow time for 0.5 % Ivory force in flotation—a critique. Colloids Surfaces A 192:39–51
Liquid . The labels give the concentrations of SAG 1010 antifoam. (2001).
The gas flow rate was 7 ml s−1 (This was redrawn from Reference 16). 8 Bikerman JJ, Foams, Springer-Verlag, New York (1973).

484 Polym Int 52:479–485 (2003)

 Defoamers: linking fundamentals to formulations

9 Blease TG, Evans JG, Hughes L and Loll P, The mode of action 13 Kulkarni RD, Goddard ED and Kanner B, Mechanism of
of Antifoams, in Defoaming Theory and Applications, ed by antifoaming action. J Colloid Interface Sci 59:469–476 (1977).
Garrett PR, pp 299–324, Marcel Dekker, New York (1996). 14 Jha BK, Christiano SP and Shaw DO, Silicone antifoam
10 Wang G, Pelton R, Hrymak A, Shawafaty N and Heng YM, On performance: correlation with spreading and surfactant
the role of hydrophobic particles and surfactants in defoaming. monolayer packing. Langmuir 16:9947–9954 (2000).
Langmuir 15:2202–2208 (1999). 15 Koczo K, Koczone JK and Wasan DT, Mechanisms for
11 Frye GC and Berg J, Antifoam action by solid particles. J Colloid antifoaming action in aqueous systems by hydrophobic parti-
Interface Sci 127:222–238 (1989). cles and insoluble liquids. J Colloid Interface Sci 166:225–238
12 Frye GC and Berg J, Mechanisms for the synergistic antifoam (1994).
action by hydrophobic solid particles in insoluble liquids. J 16 Pelton R, A model of foam growth in the presence of antifoam
Colloid Interface Sci 130:54–59 (1989). emulsion. Chem Engng Sci 51:4437–4442 (1996).

Polym Int 52:479–485 (2003) 485