Foam microrheology: from honeycombs to random foams

Andrew M. Kraynik
Engineering Sciences Center MS 0834
Sandia National Laboratories
Albuquerque, New Mexico 87185–0834 USA

Douglas A. Reinelt
Department of Mathematics
Southern Methodist University
Dallas, Texas 75275–0156 USA

April 9, 1999

We dedicate this paper to Henry M. Princen for his seminal work in foam rheology.
Henry was born in Eindhoven; his Dutch name is Thijs.
Both of his parents were born in ’s Hertogenbosch.

Abstract
Foam includes a broad range of materials from shaving cream to the flexible polyurethane that
cushions our seats. This overview covers two decades of research on the rheology of liquid foam
from a micromechanical point of view. These highly structured, multiphase fluids exhibit rich
rheological response that can be related to geometry and mechanics at the cell level. Properties of
interest include the shear modulus, yield stress, and non-Newtonian viscosity. Theories based on
the simple liquid honeycomb in 2D illustrate cell-level mechanisms that are also important in 3D.
These include energy storage in expanding surfaces that causes elasticity, irreversible topological
transitions within the foam structure that produce yield phenomena, and the interplay between cell
distortion and film-level viscous flow that is responsible for viscoelasticity. Static 3D structures
ranging in complexity from the Kelvin cell to the elegant Weaire-Phelan structure to random
polydisperse foams are calculated with the Surface Evolver, a computer program developed by
K.A. Brakke. Excellent agreement with experimental data on foam structure and shear modulus is
demonstrated. Simulations involving large quasistatic deformations of Kelvin and Weaire-Phelan
foams in simple shearing flow are compared with 2D results. The geometry and rheological
consequences of Plateau borders in wet foams are described. The fluid mechanics of bubbles
growing in a viscous fluid reveals the evolution of foam structure that controls behavior in the
solid state. Simulated soap froth structure is used as a template to develop finite element models of
cellular solids. The micromechanical approach that is described has established a firm theoretical
foundation for developing structure-property-processing relationships for foamed polymers.

1
1 Introduction
The flexible polyurethane that cushions our seats and the extruded thermoplastics that insulate our
homes are familiar examples of foamed polymers that drive large and profitable businesses. These
and many other foams, which may be liquid or solid, mundane or esoteric, support a broad range of
applications that motivate the development of foam science and foam technology (Hilyard & Cun-
ningham, 1994; Prud’homme & Khan, 1996; Gibson & Ashby, 1997; Sadoc & Rivier, 1999). The
inherent multidisciplinary character of this field resonates with the theme and philosophy of PPS-15.
Foams are protypical engineered materials. Their cell-level structure and constituents determine
macroscopic properties. While many commercial products are cellular solids, their structure typically
evolves in the fluid state where gas bubbles grow in viscous liquids. These heterogeneous systems
undergo various degrees of expansion and shear during processing. Understanding the rheology of
these complex fluids in these complex flows is necessary to design processing equipment and establish
procedures to control the microstructure and performance of the final solid foam. We will focus on the
development of microrheological models that probe these connections and provide a firm theoretical
foundation for understanding structure-property-processing relationships. We would like to know,
for example, what foamed material and cell structure are optimal for a given application and how to
control the evolution of that structure during processing. The schematic in Fig. 1 illustrates the flow
of a gas-charged melt through a die during thermoplastic foam extrusion (Kraynik, 1981)
In broad outline, theoretical developments in this field progress from the simplest structures to
more complicated ones, as indicated by the subtitle of this article “...from honeycombs to random
foams.” Like most complex materials, the structure of foams involves multiple length scales; and the
function of foams often involves many physical phenomena. Foam science and foam technology draw
chemists, physicists, mathematicians, and engineers from many disciplines. Foam micromechanics
spans traditional fluid mechanics and traditional solid mechanics (Kraynik, 1988; Weaire & Fortes,
1994; Kraynik et al., 1999). Foams continue to present significant scientific and technical challenge;
and this challenge goes hand-in-hand with substantial opportunity.

Figure 1: Schematic of thermoplastic foam extrusion.

2 Structure of dry soap froth

From a geometric point of view, one can argue that a dry soap froth has the ‘simplest’ cell-level
structure of any 3D foam. The liquid volume fraction is zero in the dry limit. Liquid drainage is
absent and gas diffusion between cells can be neglected under static conditions. In a hypothetical dry
foam, the thin liquid films degenerate to mathematical surfaces that become the faces of polyhedral


cells and the Plateau borders collapse to form cell edges. The average cell volume determines the

2
 Figure 2: Kelvin cell, Weaire-Phelan foam, and random monodisperse foam with 125 cells.

only characteristic length scale


. The foam structure can be modeled as a continuous network of

surfaces, both sides of which have uniform surface tension . The equilibrium conditions known as
Plateau’s laws (Plateau, 1873; Taylor, 1976) require each face to have constant mean curvature. Three
faces meet at equal dihedral angles of 

along cell edges; and four edges join at equal tetrahedral
angles of     !#"%$ 
at each cell vertex.
The Surface Evolver, which has been developed, maintained, and freely distributed by Brakke
(1992), has become the standard computer program for calculating the minimal surfaces in a 3D
foam; and it is capable of solving many other problems. The Evolver is available over the internet
from http://www.susqu.edu/facstaff/b/brakke/evolver/.
Unless otherwise stated, all of the structures in this article are spatially periodic. This means that
&
space can be filled with a representative volume of foam (unit cell) that contains cells. The special
&
case, equals one, is reserved for the Kelvin cell (Kelvin, 1887) shown in Fig. 2. The Kelvin cell is
the only structure known that satisfies Plateau’s laws and forms a perfectly ordered foam in which all
cells have identical shape and orientation.
Kelvin believed that his minimal tetrakaidecahedron (14-hedron) divided space into equal-volume
cells with the least possible surface area. Weaire & Phelan (1994) used the Surface Evolver to compute
the elegant counterexample that is also shown in Fig. 2. The Weaire-Phelan (WP) foam has 0.3380%
less surface area than the Kelvin foam. It contains eight different monodisperse cells: two pentagonal
dodecahedra and six 14-hedra that have twelve pentagonal faces and two hexagonal faces. The WP
foam belongs to a class of two dozen or so structures known as tetrahedrally close-packed (TCP)
to crystallographers and Frank-Kasper to metallurgists and material scientists (Rivier, 1994). All of
the TCP foams contain two or more polyhedra of four types. Each of these unique -hedra ( ' '

!( ")( *!( +
) contains twelve pentagons and '  

hexagons that do not occupy adjacent faces on a

particular polyhedron.
The Kelvin cell and TCP foams are useful models of 3D structure but they do not possess the
topological disorder of random soap froths. Matzke (1946) used an eye dropper to carefully prepare
foams that he believed to be monodisperse and classified all of the different polyhedra that he observed
under a microscope. Data on six hundred cells from the foam interior are summarized in Fig. 3. The
Kelvin cell does not have the most common face: a pentagon; and TCP foams do not have quadrilateral
faces, which are also very common.

3
 Figure 3: Distributions of polyhedra with ' faces and polygons with , sides.

We have performed computer simulations to calculate the microstructure of random, monodis-

perse foams. Software developed by John M. Sullivan (University of Illinois) was used to calculate
Voronoi polyhedra. The initial Voronoi seeds were generated by two methods: random sequential
adsorption (RSA) and random close packing of hard spheres (RCP). In RSA, a randomly generated
seed is accepted in the unit cell only if the distance to existing seeds is greater than the sphere diam-
eter, which is chosen to be as large as possible to pack &
spheres in a cube with spatially periodic
.- /012*
constraints. Relatively loose packings of monodisperse spheres with volume fractions up to

dia National Labs), produced much denser configurations with

.- /01+3"
were produced. Molecular dynamics simulations, using software written by Frank B. Van Swol (San-
. The standard deviations on
Voronoi polyhedra volumes were about 4653708 
for RSA and 4953/0: "
for RCP. The Surface Evolver
was used to produce random monodisperse foams by relaxing the Voronoi structures under the con-
straint of equal cell volumes (see Fig. 2). This process involves many topological transitions before
structures that satisfy Plateau’s laws are achieved. Typical distributions of polyhedra with faces '
,
and polygons with sides, are also shown in Figs. 3. Matzke found averages of 13.70 faces/cell and
5.124 edges/face. Ten simulations based on RCP with &; 22
produced 13.86 faces/cell and 5.134
edges/face. Five simulations based on RSA with &;<*0 

gave 13.95 faces/cell and 5.140 edges/face.

The computed distributions and averages for random monodisperse foams are in substantial agree-
ment with the experimental data. Notice that triangular faces are common among the Voronoi cells
but very rare in the relaxed foams. Matzke did not find any triangular faces; nor did he find a single
Kelvin cell, although others have. We found a few Kelvin cells among our random structures but they
were very rare, which is not surprising since pentagonal faces are so common.
We use the term polydisperse to describe foams that contain cells of different size, i.e., volume
in 3D and area in 2D. Random polydisperse foams have also been simulated. These foams contain
'
a broader distribution of -faced polyhedra than monodisperse foams: smaller cells tend to have
,
fewer faces and vice versa. But surprisingly, the distribution of -sided faces is very similar to the
monodisperse case even when the individual cell volumes vary by over a factor of ten. This would
not occur if the polydisperse foam contained many extremely small cells because tetrahedra with four
triangular faces would be common.

4
 Figure 4: Wet Kelvin foam with /0:) , Plateau Border segments: /0:) (=0:2) .
3 Structure of wet foam
Even though the films and Plateau borders in real foams have finite thickness, we will consider sit-
uations where the film thickness is zero. This is a reasonable assumption since the thickness set by
colloidal forces in thin liquid films is often much smaller than the bubble size and the radius of the
Plateau borders, which scales as >

. The Surface Evolver was used to calculate the geometry
of the wet Kelvin foam with /0:)
shown in Fig. 4. The Plateau border interfaces have tension 
and the films have tension
?from two interfaces. The bubble and the Plateau border have volume
(1- @
and

, respectively. The figure also contains an isolated segment of Plateau border, which
clearly indicates that the border is relatively thick when /0:) BA
, and that the volume of liquid is sig-
nificant in the ‘node’ that forms at the junction of Plateau borders. When , the borders can be
considered long and slender, and the volume of the node can be neglected; however, is not /0:)
small enough to give accurate results. These comments on foam structure are also relevant to the struts
in a solid foam with open cells. Because of the Plateau borders, accurate calculations of geometry of
foams with several cells are computationally intensive and have not been pursued.
In sharp contrast with the dry limit, a perfectly ordered wet foam can have more than one stable
structure since Plateau’s laws are not valid when is finite. A second structure corresponds to bubbles
compressed on a face-centered-cubic (FCC) lattice, which relates to closest packed spheres. The
polyhedron associated with FCC packing is the rhombic dodecahedron. Some of the nodes (Plateau
border junctions) of the wet rhombic dodecahedron involve eight Plateau borders instead of four.
More details on the structure of wet foams can be found in the section on their rheology.

4 Microrheology of 2D foam
In 2D, Plateau’s laws require polygonal cells whose edges are circular arcs that meet at equal dihedral
angles of 
 
. The cell edges correspond to liquid films with zero thickness. A perfectly ordered,
monodisperse foam contains identical regular hexagons and a polydisperse hexagonal foam contains
irregular hexagons but it also has perfect topological order. Under static equilibrium conditions, all
of the edges are straight and all of the internal cell pressures are equal in hexagonal foams. Because
all vertices are threefold in a dry foam, Euler’s law requires that the average number of sides per
cell is exactly six. A random foam contains several different types of polygons and is topologically
disordered; the cell edges are curved and the pressures are different.
The nonlinear elastic response of a polydisperse hexagonal foam is isotropic (Khan & Armstrong,
1986; Kraynik et al., 1991). In simple shear, for example, the shear stress is given by 4
D E
F D  L J KI  N0120 
4C G H E ?" (

;MI
>
OMI
(1)

5
Figure 5: Simple shearing flow of a perfect 2D foam. The shear stress and first normal stress difference
are scaled by P3Q Q
where is the initial edge length.

D E

where is the shear modulus, is shear strain, and is the average cell ‘volume.’ In general, the
shear modulus, stress, and energy density all scale as R2
ST I U
where is the dimension and is
VT I
the characteristic length scale. Equation (1) indicates that a soap froth with smaller cells is stiffer; but
D
the cell-size distribution is inconsequential as long as the cells are hexagonal. Weaire et al. (1986)
computed the shear modulus of random foams and always found smaller values for , which led them
to conjecture that (1) is an upper bound on the shear modulus of a dry 2D foam.

chose (see Fig. 5) gives periodic response with the smallest possible strain period
E3W X
 L 
Princen (1983) investigated simple shearing flow of a perfect 2D foam. The orientation that he
. The
rheological behavior exhibits several universal features of dry foams under quasistatic flow, regardless
of whether they are 2D or 3D, ordered or random. The stress-strain curves are piecewise continuous,
which corresponds to elastic-plastic behavior. Each branch of the curve represents large-deformation
elastic response of a foam with fixed topology, i.e., the behavior is reversible and cell neighbors do not
change. Each branch terminates when the foam structure violates Plateau’s laws. Stability is restored
by a cascade of local topology changes called T1s that result in a stable foam structure with different
cell neighbors. The jumps in stress and structure are not reversible.
Equation (1) is valid for all polydisperse hexagonal foams up to a critical strain where the length
of some edge goes to zero and produces a fourfold vertex. This defines an elastic limit and creates
the unstable situation that provokes T1s. In Princen’s case, the new structure looks exactly like the
original structure after every T1. The perfect structure and particular orientation that he analyzed
are responsible for other microrheological artifacts that are evident in Fig. 5 but not representative of
random foams. These include large stress fluctuations and strain-periodic behavior. It is even common
to get negative shear stress for other orientations of a perfect foam (Kraynik & Hansen, 1986).
For comparison, Fig. 6 contains the stress-strain curve for a random polydisperse foam with 256
cells (Herdtle, 1991). The fluctuations are substantially smaller but have not disappeared. This occurs
because a T1 cascade can involve many cells. A single edge length going to zero with strain triggers
a local T1 that only involves four cells. The subsequent structure may not relax to a stable foam; this

6
Figure 6: Shear stress for a random polydisperse foam with 256 cells (Herdtle, 1991). The stress is
scaled by the energy density of a perfect foam.

triggers another T1 and so on until stability is eventually restored. One can anticipate ‘smooth’ curves
&
that asymptotically approach a plateau when is very large; but large has yet to be quantified. This
plateau determines the dynamic yield stress or flow stress of the foam. The stress-strain curve can
exhibit an overshoot when the initial structure is very different from the stationary state.
Princen (1983) also studied simple shearing flow of a wet foam with finite liquid fraction . When
is sufficiently large, the stress and structure evolve smoothly with strain but the average stress is zero
(Reinelt & Kraynik, 1990). In this incongruous situation, the foam flows and has a shear modulus but
no dynamic yield stress. It seems unlikely that this behavior could occur in a random wet foam.
Kraynik & Hansen (1987) developed an ad hoc model for viscous effects when is small and
Y
topological transitions are fast. The excess tension in a film is given by

[ Y Z\ ^] Q`a _  a > ( ] E

QbNcJ
MI _  (2)
]
where Q is the capillary number, c is the viscosity of the continuous liquid phase, a is the dimen-
E
sionless film length, a _ is the film stretch rate, and _ is the foam shear rate. The excess tension is
positive when a film is stretching and vice versa. In contrast with the quasistatic case, different film
tensions produce unequal vertex angles that vary with time. This simple model produces relaxation
phenomena that can be found in more sophisticated treatments. When the capillary number is large,
the response never becomes time periodic but cell distortion continues to grow and films continue to
] Q
thin. The highly distorted microstructure is especially evident in the first normal stress difference . & 
The behavior for large provides a plausible mechanism for shear-induced film rupture and foam
breakage. Distortion and damage of the microstructure could be highly undesirable in a foam process.
When a perfect foam is so wet that topological transitions are smooth, the slow film-level viscous
flow can be modeled with the asymptotic theory of Mysels et al. (1959) that describes a thin film
being pulled from or pushed into a Plateau border. The excess tension is given by
M
[ Y Zd ] Y Q e (3)
where the local capillary number
] YQ is based on the relative film speed. The resulting foam viscosity
cgf
is given by

cgfhZNc ] Q  eI (4)
which indicates shear-thinning behavior (Reinelt & Kraynik, 1990).

7
 1.5
c

xy
b

σ
d
0.5
a

1.5

1
1
N

0.5

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5
γ

(a) (b)

(c) (d)

Figure 7: Simple shearing flow of a wet foam with /0:* and

] Q) (Santos, 1998).

to solve the Stokes equations for a wet foam with

Vi 08
Santos and Reinelt (Santos, 1998) used a finite-difference method that involves overlapping grids
. The foam is stabilized by a disjoining
j
pressure that prevents the local film thickness from going to zero; k
jV `l k   (5)

where l is the Hamaker constant. Their approach is able to resolve the time evolution of the Plateau
border shape and the film-level geometry for fast topological transitions (see Fig. 7). Calculations
such as these provide important information on the development of foam geometry during processing.
This includes anisotropy in cell shape as well as the distribution of material throughout the continuous
phase, which eventually solidifies to form a cellular solid.
Martinez & Kraynik (1992) used a boundary integral method to study the effect of viscous forces
on foam expansion from dilute gas bubbles to highly crowded cells. The initial structure consists
9m   701

of circular bubbles on a hexagonal lattice suspended in a viscous fluid. Starting from a typical gas
n
fraction, , the foam undergoes uniform expansion at constant volumetric expansion rate
9m
] QoCcJ
MI n 
. The only parameters in this low Reynolds number flow are


and a capillary number defined
] Q A
as where is the instantaneous bubble volume. When , surface tension
6mqp 0 1 +2
] Q0
effects dominate and the bubbles expand as circles until they nearly touch at . The
m 701+
results for are very different, as shown in Fig. 8. When
deviations from circular shape and develop ‘flat spots’ between neighbors. By
, the bubbles exhibit slight
, honeycomb
9m /01r

8
 Figure 8: Foam expansion at
] Q) : 9m /01*!(=01+!(=0s$%(=01r!(=)1!(t01* .

Figure 9: Final foam structure at

9m /012* : ] Q070:) (=08 0( :2 .

9mvu
microstructure has emerged; most of the liquid is found in thick films and the Plateau borders are
9mxw 01 +2
highly curved. The hexagonal microstructure continues to evolve as
9m
. The cells can never
be circular when
] Q ] Q
, but they can be highly distorted for smaller when viscous forces,
as measured by
films and Plateau borders when
9m /012*
, are large. Figure 9 shows the effect of
] Q
on liquid distribution between the
. The film thickness increases with and asymptotically
approaches a maximum value that corresponds to regular hexagonal bubbles. Santos (1998) has used
the finite-difference method to study foam expansion and found the same behavior.
A simple shell model for foam expansion is based on a circular bubble expanding in a ring of
,Bf
viscous fluid. The effective foam pressure is given by
9m
 ,.fh h   ,.y  z {H|c  9m n (6)
z n
where ,}y is the internal gas pressure and is the bubble radius. The coefficient of is the expansion
viscosity of the foam ~9f . Even though the shell model clearly cannot capture the evolving bubble
]
shape when Q and
9m are large, Eq. 6 does provide an excellent estimate of the expansion viscosity
]
when Q is    .
It would be very difficult to form an open-cell foam if the process remained in the large capillary
number regime up until the foam solidified. Simulations that combine simultaneous foam expansion
and shear would shed light on processing flows such as thermoplastic foam extrusion (see Fig. 1).

9
Commercial scale production of flexible slabstock polyurethane foam involves expansion of a thick
layer of well-mixed prepolymers that have been deposited on a moving belt. Simultaneous foam
expansion and extension cause anisotropic cells that are oriented along the ‘rise’ direction.
Pozrikides & coworkers (Li et al., 1996) and Loewenberg et al. (1999) are developing bound-
ary integral capabilities that enable simulations involving hundreds of bubbles or drops in 2D and a
few dozen in 3D. Research on adaptive mesh refinement algorithms also shows great promise. This
methodology is necessary to deal with highly curved interfacial regions and thin film formation, both
of which occur in highly expanded foam flows.

5 Microrheology of 3D foam
The Surface Evolver has been used to apply homogeneous deformations to spatially periodic foams,
compute the minimal structures, and then evaluate the complete stress tensor. The shear modulus,
large-deformation elastic behavior, and quasistatic response in simple shearing flow have been studied
for wet and dry foams (Reinelt & Kraynik, 1993, 1996, 1999; Kraynik & Reinelt, 1996a-1996d).

5.1 Dry foam

5.1.1 Shear modulus
The shear modulus of a dry foam scales as R2
€eI
. The Kelvin foam and various monodisperse TCP
foams (Weaire-Phelan, Friauf-Laves, and Bergman) have cubic symmetry, which means that there are
two independent shear moduli given by
D D  †…@…
  ‚  @  € ƒ> @
„( > (7)

where the ˆ‡:‰ are elastic constants defined by Love (1994) and Nye (1985). An effective isotropic
D D
shear modulus can be obtained by averaging over all foam orientations. The average can be
D
calculated in many ways, e.g., at constant strain (Voigt) or at constant stress (Reuss). The Voigt and
D
Reuss averages give relatively loose upper and lower bounds for . The shear moduli of various
monodisperse foams are given in Table 1 where is reported as the mean of the bounds, and the
D
range. The Kelvin foam is significantly more anisotropic than the TCP foams. The T foam, which
has 81 cells and many faces of different orientation, is essentially isotropic. Based on , the TCP
foams are significantly stiffer than the Kelvin foam. We have also calculated the shear moduli of sev-
eral random foams with 64 cells; these included monodisperse and polydisperse foams. These shear
moduli exhibit modest anisotropy, presumably because of system size, but all values fall in a narrow
range: 0s$ r<Š\0:r
. Perhaps even more important, there is no significant difference between results


for the monodisperse and polydisperse foams. Recalling that is simply the average cell volume,
this preliminary finding suggests that the shear modulus of a random dry foam may be insensitive to
cell-size distribution. This would be a very simple result. It is also interesting that the average shear
modulus of a Kelvin foam provides an excellent estimate of the shear modulus of the random foams.
Princen & Kiss (1986) measured the shear modulus of concentrated oil-in-water emulsions with
polydisperse drop-size distributions. They used their data to develop an empirical correlation that
includes the dependence on liquid volume fraction (see Fig. 15). Their correlation extrapolates to a
shear modulus of 0.82 in the dry limit, which agrees well with our simulations.
Theoretical evidence supporting the possibility that polydispersity has little or no influence on the
shear modulus of a dry foam can be found in 2D and 3D. Cell-size distribution has no influence on
the elastic behavior of polydisperse hexagonal foams (Kraynik et al., 1991). The shear modulus of a

10
 D D D
Kelvin 0.5706
 0.9646
> 0.7814 0.0256Š
Weaire-Phelan (A15) 0.8902 0.8538 0.8682 0.0002Š
Friauf-Laves (C15) 0.8448 0.8860 0.8693 0.0002Š
Bergman (T) 0.858 0.856 0.857
Random 0.78 0.08 Š

Table 1: Shear moduli of dry foams

bidisperse Weare-Phelan foam depends very little on the relative cell size and is neither a minimum
nor a maximum for the monodisperse case (Kraynik & Reinelt, 1996c).

Figure 10: Uniaxial extension of a Kelvin foam.

5.1.2 Large-deformation extension

The most symmetric distortion of a Kelvin foam involves uniaxial extension; Fig. 10 includes the
‹
‹ u @‹ Œ  /01
2*3"
evolution of cell geometry and tensile stress with Hencky strain . The area of the ‘square’ face
that is being pulled decreases with increasing strain but remains finite in the limit .
@‹ Œ
‹@Œ  
There is no stable solution that maintains contact between the original cell neighbors beyond , the
elastic limit. The stress-strain curve exhibits a maximum well before the turning point at . The

‹†   2 Ž
/01
20
unstable solutions on the curve below the turning point have the same topology but higher surface
area than their stable counterparts. The endpoint at , where the tensile stress is zero,
corresponds to a rhombic dodecahedron on an FCC lattice, which is unstable in the dry limit.

5.1.3 Simple shearing flow

In contrast with linear elastic behavior, microrheological response in simple shearing flow depends
E
strongly on the complexity of the foam structure. The stress and structure of a Kelvin foam are

strain-periodic behavior with the smallest strain period

E W   
t>
piecewise continuous functions of shear strain , as shown in Fig. 11. This foam orientation produces
. Each discontinuity in the stress-
strain curve corresponds to topological transitions that are preceded by shrinking faces. As in 2D
(see Fig. 6), each T1 cascade reduces surface energy, results in cell-neighbor switching, and provides
a cell-level mechanism for irreversible yield behavior during foam flow. There are two T1 cascades

11
 E
E W  G  

Figure 11: Simple shearing flow of a Kelvin foam; cell shapes for
0.80, 0.98 (before T1), 1.0, .
= 0, 0.30, 0.60 (before T1), 0.62,

Figure 12: Simple shearing flow of a Weaire-Phelan foam,

E W  .

Figure 13: Simple shearing flow of a random monodisperse foam with 72 cells.

12
per cycle; each is triggered by opposite edges of shrinking quadrilateral faces going to zero length,
which leads to unstable edge connectivity. These standard transitions are very different from the
point transition in extension, where all edges on a face shrink together. Symmetry imposes strong
restrictions on the type and outcome of topological transitions when the foam has perfect order; Kelvin
cells beget Kelvin cells.
The Weaire-Phelan (WP) foam has less symmetry and exhibits more diverse topological transi-
tions than the Kelvin foam in simple shearing flow. Since there are eight different polyhedra in the
unit cell, the Weaire-Phelan foam is far less constrained. Individual polyhedra change type as T1
cascades produce foams that contain a much greater variety of polyhedra and faces than existed orig-
inally. This occurs because individual faces, which begin as pentagons or hexagons, can gain or lose
one or two edges. The first T1 cascade alone involves polygons ranging from triangles to octagons,
which assemble to form many different polyhedra.
Figure 12 indicates that stress-strain fluctuations are also large for the Weaire-Phelan foam, which
is still a relatively small system. The most elementary local T1 involves five cells and a particular cell
can be altered several times during a T1 cascade (Schwarz, 1964). Consequently, the disturbance to
the foam structure and the corresponding jumps in stress and energy are large.
For some orientations, Weaire-Phelan foams return to their original topology. In other cases they
become Kelvin cells but subsequent T1 cascades produce topological disorder. A T1 cascade begins
with a cell edge going to zero length as strain increases. The process by which edges vanish in Kelvin
foams and in TCP foams can be smooth and continuous, or abrupt and discontinuous. When multiple
cells are involved, abrupt onset is often connected with symmetry-breaking bifurcations that provide
a mechanism for Kelvin cells to disorder. The bifurcations can also cause strain localization in which
a T1 cascade results in layers, two Kelvin cells thick, sliding past one another.
Figure 13 contains the stress-strain curve for a random monodisperse foam with 72 cells. Some
of the polyhedra in the initial structure have short edges that go to zero length at small strains. This
reduces the elastic limit and leads to strain-hardening elastic-plastic behavior. The foam structure
appears to be sufficiently complex to reduce stress-strain fluctuations and prevent negative shear stress
at large strains. These promising trends also occur in 2D (compare Figs. 6 and 13).
Gopal & Durian (1995) have used a multiple-light scattering technique called diffusing-wave spec-
troscopy (DWS) to study nonlinear bubble dynamics during foam flow. They observe localized stick-
slip like rearrangement of bubbles that undoubtedly refers to T1 cascades. Simulation of large random
foams will eventually provide connections with the DWS experiments.

5.2 Wet foam

5.2.1 Shear modulus
Figure 14 contains wet Kelvin cells and wet rhombic dodecahedra (RD) with the same liquid content:
/0: "
. There is an example where each structure has isotropic stress. The shear moduli of both
foams are graphed in Fig. 15 and compared with the empirical correlation of Princen & Kiss (1986).
D
Both foams are stable over some overlapping range of . The smaller shear modulus of the Kelvin
foam

decreases rapidly as approaches 0.11. This is an indication that the smaller (original
quadrilateral) films are shrinking to zero area prior to being consumed by surrounding Plateau borders.
When these borders converge to form eight-way junctions, the bubbles lose contact with all next
nearest neighbors on the BCC lattice and the wet Kelvin cell becomes unstable.

13
Figure 14: A wet Kelvin cell on a BCC lattice is being stretched to form a wet rhombic dodecahedron
on an FCC lattice, /0: "
.

Figure 15: Shear moduli of wet Kelvin foams and wet rhombic dodecahedra (RD).

Figure 16: Tensile stress and energy for a wet Kelvin cell being stretched to form a wet rhombic
dodecahedron, /0:+
. Same deformation as in Fig. 14.

5.2.2 Large-deformation extension

As discussed in Section 5.1.2 above, a BCC lattice can be stretched to form an FCC lattice. Simarly,
a wet Kelvin cell can be stretched to form a wet rhombic dodecahedron. Figure 14 depicts this
process. The small film on ‘top’ of the Kelvin cell shrinks and eventually vanishes as the foam
stretches. This provokes a topological transition that results in a wet RD that has twelve films and
some four-way and eight-way Plateau border junctions (nodes). Further stretching takes the foam

14
 D
to FCC structure and isotropic stress. Figure 16 contains representative graphs of tensile stress and
/0:+
energy density for . The shear modulus

of each structure can be evaluated from the slope of
the stress-strain curve near the endpoints. The energy maximum that occurs when the stress changes
Bp 0:+3"
sign determines the energy barriers between the undeformed structures. These barriers are equal when

has lower energy than the wet RD when

vi 0:+3"
both undeformed structures have the same energy, which occurs when
and vice versa.
. The wet Kelvin cell

The process just described appears to be completely reversible; a wet RD can be compressed to
a wet Kelvin cell. Bubbles that were separated by an eight-way junction become neighbors when the
foam is compressed. This reverse topological transition is initiated when interfaces on opposite sides
of the eight-way node come into contact. The film that forms upon first contact grows in area with
further compression.

Figure 17: Simple shearing flow of a wet Kelvin foam with /0:+

. The bottom row gives a different
view of a thin film on a Kelvin cell vanishing and interfaces separating to form an eight-way node on
a wet RD.

5.2.3 Simple shearing flow

Simple shearing flow of wet Kelvin foams involves both of the topological transitions just described,
but they are not reversible. Figure 17 contains representative structures and a stress-strain curve.
Instead of two T1 cascades per cycle, like the dry case, there are four distinct topology changes in the
wet case; the foam alternates between wet Kelvin and wet RD. The first and third transitions involve
the ‘same’ shrinking film as the dry foam. The second and fourth transitions are provoked by contact
between opposite interfaces of an eight-way junction. Different from the reversible uniaxial extension,
the energy drops at each transition in simple shear. The magnitude of these jumps is much smaller
for wet foams than dry foams. The same is true for the average shear stress, which corresponds to the
(viscometric) yield stress of the foam. Consistent with measurements of yield stress by Princen (1985)
the average stress decreases very rapidly with .

15
 When is very small, we anticipate that the wet RD structures will be unstable intermediates that
lead to wet Kelvin structures. There will be two Kelvin branches in the stress-strain cycle, just like
the dry case.

5.3 Foam expansion

The interplay between viscous forces and surface tension and its influence on the evolving foam mi-
crostructure during foam expansion is fundamental to the science and technology of foam processing.
The natural starting point for a 3D analysis involves uniform expansion of a viscous fluid containing
spherical bubbles on a BCC lattice. Choosing an FCC lattice would of course permit tighter packing
of spherical bubbles but it would lead to instabilities involving eight-way nodes when
6m
approaches
unity. This would complicate attempts to study other physical phenomena that could control the com-
mercially important low-density regime. These include scalar transport of heat and mass, chemical
reaction, and rheological complexity of the bulk fluid and the interface.

nearest neighbors and form the precursors of hexagonal faces when

9m / L  r/01+2r 
Questions of stability aside, spherical bubbles on a BCC lattice make first contact with eight
. Under static

when
9m?p 01r2
conditions, which fall in the domain of the Surface Evolver, six next nearest neighbors make contact
; this completes the fourteen faces of the Kelvin cell. The stable equilibrium structures
beyond this point have been calculated.

] Q0/01* ,
9m 01r 0(=01 0(=012*!:)1r
Figure 18: Uniform expansion of bubbles on a BCC lattice evolving into Kelvin cells;
(Loewenberg et al., 1999).

9m ‘
Neglecting fluid inertia, the important parameters in the viscous free-surface flow problem in-
clude: the volume fraction , the ratio of gas viscosity to liquid viscosity, and the capillary
] Q
defined as
] b Q dcJ
y eI n  (8)

where
9y is the bubble volume and

n is the volumetric expansion rate of the foam.
] Q
Many qualitative features of the flow can be anticipated from the 2D analysis. A shell model may
9m
provide a good estimate of the foam expansion viscosity when
] Q
is large. Hydrodynamic interactions
between neighbors will cause bubble distortion when is just below 0.680 and large will promote
the formation of thick liquid films.

16
 Loewenberg et al. (1999) have made substantial progress on this problem by using the bounday-
integral method to solve the Stokes equations. Thus far, they have only considered the case
] Q0/01* ‘.
because the numerical implementation is simpler. Results of their simulations for are con-
tained in Fig. 18, which illustrates the evolution of bubble shape from spheres to Kelvin cells.

Figure 19: Bubble shape for

9m /01 
: when expansion at
] Q)/01* terminates and after 20 time units.
The bottom row contains two views of the equilibrium structure.

This process is illustrated in Fig. 19. The foam expands until

9m /01
Once the foam expansion has stopped, the bubble shape will continue to relax toward equilibrium.
and then the bubble shape relaxes
for twenty time units. The bottom row contains two views of the equilibrium structure calculated with
the Surface Evolver; the shape of a representative liquid region is also portrayed.

6 Micromechanics of solid foams with open cells

This article has focused on the microrheology of liquid foams. We conclude with an illustrative exam-
ple from the world of cellular solids. In nature and industry, liquid foams undergo phase changes to
produce solid foams so the cell-level structure of the former heavily influences the latter. This justifies
and motivates using the geometry of a soap froth and related materials as templates for developing
micromechanics models of cellular solids.
The primary microstructural feature found in low-density open-cell foams, such as flexible poly-
urethane, is the network of slender struts that meet most often at four-way joints. We are developing
finite element models by using beam elements to discretize the microstructure. The undeformed struts
are assumed to have uniform cross section and a straight axis, but these approximations can be relaxed.
The location of joints and the connectivity of struts are based on foam geometries calculated with the
Surface Evolver. Figure 20 contains a random, monodisperse soap froth and the corresponding model
of an open-cell foam. The general purpose finite element program ABAQUS is used to perform large-
deformation solid mechanics calculation of the foam structure and strut-level forces. This information

17
is then used to evaluate the macroscopic stress of the foam. The graph in Fig. 20 contains a typical
stress-strain curve for an unconfined foam subjected to uniaxial compressive stress. The struts are
’
shaped like Plateau borders and are composed of linear elastic material with Youngs modulus . The
nonlinear response is caused by large-deformation geometric effects.

Figure 20: Soap froth and corresponding open-cell foam. Undeformed and deformed finite element
mesh. Stress-strain curve for unconfined uniaxial compression.

Magnetic resonance imaging and micro-X-Ray CT are being used to analyze the cell-level struc-
ture of polyurethane foam (Pangrle et al., 1998). These data can be compared with statistics obtained
from simulations of foam structure. Beyond Surface Evolver results for minimal soap froths, fluid
mechanics calculations of foam structure evolution in process-related flows promise much-needed in-
formation. Combine all of this with finite element models for the mechanics of cellular solids and a
comprehensive package begins to emerge. The micromechanical approach has established a firm the-
oretical foundation for developing structure-property-processing relationships for foamed polymers.

ACKNOWLEDGEMENTS

We thank Ken Brakke for developing and maintaining the Surface Evolver. AMK also thanks Mitzi
Bower for valuable assistance with the graphics. Sandia is a multiprogram laboratory operated by
Sandia Corporation, a Lockheed Martin Company, for the the U.S. Department of Energy under con-
tract #DE-AC04-94AL85000. This work was also supported by the Dow Chemical Company under a
Cooperative Research and Development Agreement (CRADA).

18
References
Brakke, K.A., “The surface evolver,” Experimental Mathematics, 1, 141-165 (1992).
Gibson, L.J. and Ashby, M.F., Cellular Solids: Structure and Properties, 2nd Ed, Cambridge Univer-
sity Press, Cambridge (1997).
Gopal, A.D. and Durian, D.J., “Nonlinear bubble dynamics in a slowly driven foam,” Phys. Rev. Lett.
75, 2610-2613 (1995).
Herdtle, T., Numerical Studies of Foam Dynamics, PhD Thesis, University of California at San Diego,
La Jolla, CA (1991).
Hilyard, N.C. and Cunningham, A. (Eds.), Low Density Cellular Plastics, Chapman & Hall, London
(1994).
Kelvin, Lord (W. Thompson), “On the division of space with minimum partitional area,” Philos. Mag.
24, 503-514 (1887).
Khan, S.A. and Armstrong, R.C., “Rheology of foams. I. Theory for dry foams,” J. Non-Newtonian
Fluid Mech. 22, 1-22 (1986).
Kraynik, A.M., “Rheological aspects of thermoplastic foam extrusion,” Polymer Eng. Sci. 21, 80-85
(1981).
Kraynik, A.M., “Foam flows,” Ann. Rev. Fluid. Mech. 20, 325-357 (1988).
Kraynik, A.M. and Hansen, M.G., “Foam and emulsion rheology: A quasistatic model for large
deformations of spatially-periodic cells,” J. Rheology, 30, 409-439 (1986).
Kraynik, A.M. and Hansen, M.G., “Foam and emulsion rheology: A model of viscous phenomena,”
J. Rheology, 31, 175-205 (1987).
Kraynik, A.M. and Reinelt, D.A., “Linear elastic behavior of dry soap foams,” J. Coll. Int. Sci. 181,
511-520 (1996a).
Kraynik, A.M. and Reinelt, D.A., “Elastic-plastic behavior of a Kelvin foam,” Forma, 11, 255–270
(1996b). Also published in The Kelvin Problem: Foam Structures of Minimal Surface Area, D. Weaire
(Ed.), Taylor & Francis, London, 93-108 (1996b).
Kraynik, A.M. and Reinelt, D.A., “The linear elastic behavior of a bidisperse Weaire-Phelan foam,”
Chem. Eng. Comm. 148-150, 409-420 (1996c).
Kraynik, A.M. and Reinelt, D.A., “The microrheology of wet foams,” in Proceedings of XIIth Inter-
national Congress on Rheology, Ait-Kadi, A., Dealy, J.M., James, D.F. and Williams, M.C. (Eds.),
Quebec City, Canada, August 18-23, 625-626 (1996d).
Kraynik, A.M., Reinelt, D.A. and Princen, H.M., “The nonlinear elastic behavior of polydisperse
hexagonal foams and concentrated emulsions,” J. Rheology, 35, 1235-1253 (1991).
Kraynik, A.M., Neilsen, M.K., Reinelt, D.A. and Warren, W.E., “Foam micromechanics: Structure
and rheology of foams, emulsions, and cellular solids,” in Foams and Emulsions, Sadoc, J.F. and
Rivier, N. (Eds.), Proceedings of the School on Foams, Emulsions, and Cellular Materials, Cargese,
France, May 11-14, 1997, Kluwer, Boston, 259-286 (1999).
Li, X., Charles, R. and Pozrikides, C., “Simple shear flow of suspensions of liquid drops,” J. Fluid
Mech. 320, 395-416 (1996).
Loewenberg, M., da Cunha, F.R., Blawzdziewicz, J. and Cristini, V., Unpublished results (1999).
Love, A.E.H., A Treatise on the Mathematical Theory of Elasticity, Dover, New York (1994).
Martinez, M.J. and Kraynik, A.M., Unpublished results (1994).

19
Matzke, E.B., “The three-dimensional shape of bubbles in foam–an analysis of the role of surface
forces in three-dimensional cell shape determination,” Am. J. Botany, 33, 58-80 (1946).
Mysels, K.J., Shinoda, K. and Frankel, S., Soap Films: Studies of Their Thinning, Pergamon, New
York (1959).
Nye, J.F., Physical Properties of Crystals, Clarendon Press, Oxford (1985).
Plateau, J.A.F., Statique Experimentale et Theorique des Liquides Soumis aux Seules Forces Molecu-
laires. Gauthier-Villiard, Paris (1873).
Pangrle, B.J., Hammer, B.E., Bidault, N.P., Listemann, M., Stevens, R.E., Zhang, X.D., Macosko,
C.W., “Magnetic resonance imaging and micro-X-Ray CT based three dimensional analysis of poly-
urethane foams,” Proc. of Society of Plastics Industry Polyurethanes Conf., Dallas, Sept. (1998).
Princen, H.M., “Rheology of foams and highly concentrated emulsions, I. Elastic properties and yield
stress of a cylindrical model system,” J. Coll. Int. Sci. 91, 160-175 (1983).
Princen, H.M., “Rheology of foams and highly concentrated emulsions, II. Experimental study of the
yield stress and wall effects for concentrated oil-in-water emulsions,” J. Coll. Int. Sci. 105, 150-171
(1985).
Princen, H.M. and Kiss, A.D., “Rheology of foams and highly concentrated emulsions, III. Static
shear modulus,” J. Coll. Int. Sci. 112, 427-437 (1986).
Prud’homme, R.K. and Khan, S.A. (Eds.), Foams: Theory, Measurements, and Applications, Surfac-
tant Science Series, Vol. 57, Marcel Dekker, New York (1996).
Reinelt, D.A. and Kraynik, A.M., “On the shearing flow of foams and concentrated emulsions,”
J. Fluid Mech. 215, 431-455 (1990).
Reinelt, D.A. and Kraynik, A.M., “Large elastic deformations of three-dimensional foams and highly
concentrated emulsions,” J. Coll. Int. Sci. 159, 460-470 (1993).
Reinelt, D.A. and Kraynik, A.M., “Simple shearing flow of a dry Kelvin soap foam,” J. Fluid Mech.
311, 327-343 (1996).
Reinelt, D.A. and Kraynik, A.M., “Simple shearing flow of dry soap foams with TCP structure,”
J. Rheology, Submitted (1999).
Rivier, N., “Kelvin’s conjecture on minimal froths and the counter-example of Weaire and Phelan,”
Phil. Mag. Lett. 69, 297-303 (1994).
Sadoc, J.F. and Rivier, N. (Eds.), Foams and Emulsions, Proceedings of the School on Foams, Emul-
sions, and Cellular Materials, Cargese, France, May 11-14, 1997, Kluwer, Boston (1999).
Santos, J.G.G., A Numerical Study of Simple Shearing Flow of Foams, PhD Thesis, Southern Metho-
dist University, Dallas, Texas (1998).
Schwarz, H.W., “Rearrangements in polyhedric foam,” Recueil, 84, 771-781 (1964).
Taylor, J.E., “The structure of singularities in soap-bubble-like and soap-film-like minimal surfaces,”
Ann. Math. 103, 489-539 (1976).
Weaire, D. and Fortes, M. A., “Stress and strain in liquid and solid foams,” Advances in Physics, 43,
685-738 (1994).
Weaire, D., Fu, T-L. and Kermode, J.P., “On the shear elastic constant of a two-dimensional froth,”
Phil. Mag. B, 54, L39-43 (1986).
Weaire, D. and Phelan, R., “A counter-example to Kelvin’s conjecture on minimal surfaces,” Phil.
Mag. Lett. 69, 107-110 (1994).

20