Mechanical Properties of a Particle-Strengthened

Polyurethane Foam

S. H. GOODS,1 C. L. NEUSCHWANGER,2 L. L. WHINNERY,1 W. D. NIX3

1
Sandia National Laboratories, Livermore, California 94550

2
Hewlett-Packard Company, Palo Alto, California 94304

3
Department of Materials Science and Engineering, Stanford University, Stanford, California 94350

Received 16 March 1999; accepted 15 June 1999

ABSTRACT: Quasi-static compression tests have been performed on polyurethane foam

specimens. The modulus of the foam exhibited a power-law dependence with respect to
density of the form: E* } ( r *) n , where n 5 1.7. The modulus data are described well
by a simple geometric model (based on the work of Gibson and Ashby) for a closed-cell
foam in which the stiffness of the foam is governed by the flexure of the cell struts and
cell walls. The compressive strength of the foam is also found to follow a power-law
behavior with respect to foam density. In this instance, Euler buckling is used to
explain the density dependence. The modulus of the foam was modified by addition of
gas-atomized, spherical, aluminum powder. Additions of 30 and 50 wt % Al measurably
increased the foam modulus, but without a change in the density dependence. However,
there was no observable increase in modulus with 5 and 10 wt % additions of the metal
powder. Strength was also increased at high loading fractions of powder. The increase
in modulus and strength could be predicted by combining the Gibson–Ashby model,
referred to above, with a well-known model describing the effect on modulus of a rigid
dispersoid in a compliant matrix. © 1999 John Wiley & Sons, Inc. J Appl Polym Sci 74:
2724 –2736, 1999

Key words: mechanical properties; foam; particle-strengthening; modulus

INTRODUCTION (that is, whether or not the foam is open- or

closed-cell), are used in applications involving
Foams are widely used in a variety of applications cushioning, thermal insulation, and structural
for the advantages they bring in energy absorp- support.
tion characteristics, thermal properties, and spe- The application for which a particular polymer
cific strength. A closed-cell foam consists of two foam is suited also depends on the intrinsic prop-
phases: a solid phase, from which the structure of erties of the polymer (which determine whether
the foam is formed, and a gaseous phase, which the foam is flexible or rigid). For example, foam
may be derived from any of several sources, either derived from a polymer having a glass transition
physical or chemical. Polymeric foams are the temperature (T g ) near ambient will exhibit
most common and, depending on their structure greater ductility and rate sensitivity in its me-
chanical response than a foam derived from a
Correspondence to: S. H. Goods. polymer with a high T g . 1 The properties of foams
Contract grant sponsor: U.S. Department of Energy; con- also depend on the structure and geometry of the
tract grant number: DE-AC04-94AL85000.
cellular structure. A foam comprised of a closed-
Journal of Applied Polymer Science, Vol. 74, 2724 –2736 (1999)
© 1999 John Wiley & Sons, Inc. CCC 0021-8995/99/112724-13 cell structure will be more thermally insulating
2724
 PARTICLE-STRENGTHENED POLYURETHANE FOAM 2725

than an open-cell foam. Lastly, foams with cell nate (TDI), which has been identified a potential
structures having high aspect ratios will exhibit human carcinogen. Rather, it incorporates a mod-
anisotropic behavior with respect to a variety of ified diphenylmethane diisocyanate (MMDI).5 It
mechanical and physical properties.2,3 is a rigid, closed-cell, water-blown polyurethane
A foam can also be a composite material incor- foam formulated from the following constituents.
porating other solids into the host polymer in
order to modify certain physical (modulus, ther- 1. Voranol 490™, a polyether polyol, made
mal conductivity, or coefficient of thermal expan- from polypropylene oxide and a sucrose/
sion [CTE]) or mechanical (strength, toughness) glycerin base, available from Dow Chemi-
properties. Examples of such additives or fillers cal. The manufacturer specifies the follow-
include metal, glass, and ceramic powders. These ing properties: Density (25°C), 1.1 g/cm3;
composite foams can possess even better strength typical hydroxyl number, 490 mg KOH
to weight ratios and energy absorption properties equiv/g of resin; functionality, 4.3 (calcu-
than the baseline foams formed solely from the lated); average molecular weight, 460 g/mol;
matrix polymer.4 viscosity (25°C), 5572 centipoise (cps).
The present work examines the effect of in- 2. DC193™, a silicone glycol copolymer sur-
creasing loading fractions of an aluminum pow- factant from Air Products with an average
der additive on the mechanical properties of a hydroxyl number of 75.
structural polyurethane (PU) foam. The principal 3. Polycat 17™, a tertiary amine catalyst (tri-
consideration for the use of a metallic filler phase, methyl-N-hydroxyethyl propylene diamine)
such as aluminum, is the effect that the additive available from Air Products with an average
may have on the insulating qualities of the foam. hydroxyl number of 400.
For example, metal powders or fibers can signifi- 4. Distilled water, added in various amounts
cantly increase the thermal conductivity of foams. as a chemical blowing agent to produce
This can be advantageous when the foam is used carbon dioxide. The amount of water added
in a storage or shipping container to encapsulate determines the foam density as the water
a structure that is self-heating. In such an in- controls the rise within the foam.
stance, the ability to tailor the dissipative prop- 5. Isonate 143L™, modified MMDI available
erties of the foam is important in order to keep from Dow Chemical. The manufacturer
the internal temperature of a container accept- specifies the following properties for
ably low. Isonate 143L: Isocyanate equivalent
It is this interest in modifying the thermal weight, 144.5 g; content by weight, 29.2%;
characteristics of a foam that motivated the cur- functionality, 2.1; viscosity (25°C), 33 cps.
rent work. If a metal-loaded foam is to be used in
such an application, the influence of the filler Components 1– 4 were stirred together by
phase on mechanical properties must also be con- hand. The isocyanate was then added, and the
sidered (as the foam must also protect a compo- complete mixture is stirred for 1 min using a
nent from shock or impact). Specifically, the Conn mixing blade. The addition of the isocyanate
present work examines the effect of increasing initiates the rise of the foam by reacting with the
loading fractions of a 2-mm aluminum powder water to form CO2. The mixture was poured into
additive on the modulus and strength of a poly- a waxed cylindrical mold and sealed with a hand-
urethane foam called CRETE. The intrinsic me- tightened clamp. This tooling allowed the foam to
chanical properties of this foam, without the ad- expand and pack to a density twice that of the
dition of a filler phase, have been presented in expected free-rise density.
detail elsewhere5 and, along with additional data, The rise time for CRETE is approximately 8
are reviewed here for the sake of completeness. min from the time of the pour. After the reaction
was completed, the molds were cured at 66°C for
a minimum of 4 h. After unmolding, right circular
EXPERIMENTAL cylinders 28.7 mm in diameter and 50.8 mm long
were cored from the foams. The cylindrical axis of
these corings was always parallel to the rise di-
Formulation and Processing of CRETE Foam
rection of the foam. Care was taken to ensure that
The foam chosen for this study is called CRETE. the cored samples did not contain any skin or any
CRETE is formulated without toluene diisocya- material from within 3 mm of the surface. The
2726 GOODS ET AL.

to the formulation after components 1– 4 were

hand-stirred. The fraction of filler (wt %) added to
the CRETE matrix was defined as

x
wt % 5 (1)
x 1 tot

where wt % is the weight percent of filler present

in the composite, x is the weight of filler added,
and tot is the total weight of all liquid constitu-
ents.
The formulation was once again stirred by
hand. After the Al powder was completely wetted,
the foam reaction was initiated with the addition
Figure 1 Size distribution for aluminum powder
of the isocyanate. The components were then
shown as differential and cumulative volume fractions
mixed as before, poured into a mold, cured, and
versus particle diameter.
machined into cylindrical test specimens of the
same dimensions, as described above.
Foam, which remained after the right circular
ends were machined to yield flat surfaces, and the samples were cored out, was cut into small strips
cored sample densities were measured. and then snapped by hand to reveal a fracture
surface for subsequent microscopy. The micro-
Formulation and Processing of Al-Filled CRETE structure of the unfilled and Al-filled foams, as
Foam well as their fracture morphology, were examined
using a JEOL 840 scanning electron microscope.
Spherical, gas-atomized aluminum powder was
purchased from Valimet Inc., Stockton, CA. A
Mechanical Testing
Microtrac® analysis was conducted by the powder
supplier to verify the mean particle size and dis- All mechanical testing was conducted under am-
tribution. The results of this analysis are pre- bient laboratory conditions on specimens ranging
sented in Figure 1. From Figure 1, it can be seen in foam density (r*) (Terms or values annotated
that the median particle diameter is 3 mm and by an asterisk refer to parameters of the foam;
that more than 50 vol % of the powder consists of other terms annotated by “PU” refer to parameters
spheres between 2 and 4 mm in diameter. No related to the solid polymer.) from 0.05 to 0.91
powder particles were greater than 12 mm in di- g/cm3. Since the density of solid polyurethane
ameter. (rPU) is 1.2 g/cm3,6 this range corresponds to nor-
The as-received powder tended to agglomerate. malized densities (r*/rPU) from 0.04 to 0.76 (for
To minimize this and to ensure uniform incorpo- the unfilled foams). Mechanical properties in com-
ration into the foam constituents, the powder was pression were evaluated on the right cylinders
sieved using a Ro-Tap Sieve Shaker Model B from described above using a conventional Instron me-
Tyler Inc., Mentor, OH. Testing sieves meeting chanical test frame. Tests were conducted in dis-
ASTM E-11 specifications were used with a placement control at a constant initial strain rate
screen stacking sequence of 100, 200, and 325 of 1.7 3 1024 s21.
mesh. Sieving was conducted in batches of four Moduli (E*) were calculated as the slope of the
100-mL scoops of powder for 1 h. Powder, sieved linear portion of the compression loading curves.
to 2100/1200, was collected, and the process was The modulus of CRETE measured in tension has
repeated. This procedure resulted in powder that been reported in a previous study.5 Those results
was more easily blended with the liquid constit- are included here for the sake of completeness.
uents and in foams that were more uniform with Macroscopic failure did not occur under quasi-
respect to powder distribution within the cell static compressive loading for unfilled foam spec-
walls and struts. imens and low loading fractions (less than 30 wt
The Al-filled foam was formulated in an anal- %) of the Al-filled foams. In these cases, tests were
ogous manner to the unfilled foam detailed above. generally run to strains of approximately 0.3. Col-
The exception was that the Al powder was added lapse stress, defined as the plateau in stress sub-
 PARTICLE-STRENGTHENED POLYURETHANE FOAM 2727

contact between adjoining cells. These contact ar-

eas are continuous, thin polymer films, which, in
fact, form the walls that enclose the cells. The
intersection of several cells results in the forma-
tion of a cell strut, the principal structural mem-
ber of the foam. A representative polymer strut is
magnified in the insert in Figure 2(a) to illustrate
this important feature. This high-magnification
image reveals that the cell walls are very thin,
only a few microns in thickness (thus, the major-
ity of polymer resides in the struts). Increasing
the CRETE density resulted in smaller cells, as
illustrated in Figure 2(b) (note the difference in
magnification). The structure is still uniform with
an average cell diameter of 250 mm. At this higher
density, the cell membranes are thicker on aver-
age than those seen in Figure 2(a).

Microscopy of Al-Filled CRETE

Scanning electron microscopy (SEM) analysis of
the microstructure of the Al-filled foam revealed
that, overall, the morphology of the cell structure
was unaffected by the presence of the Al powder.
That is, for any given foam density, the size and
shape of the cells were the same for the loaded
foams compared to the unloaded foams. At high
magnifications, a uniform distribution of alumi-
num particles is found throughout the foam struts
and cell walls of the foam. Figure 3(a)–(c) are
scanning electron micrographs of the fracture
surfaces of three Al-loaded specimens having a
constant foam density of 0.28 g/cm3 6 0.01 and
Figure 2 SEM micrograph of CRETE foam (a) r* with increasing concentration of aluminum pow-
5 0.12 g/cm3 (a typical polymer strut is shown in der (5, 10, and 30 wt %, respectively). In each
greater detail) and (b) r* 5 0.36 g/cm3. instance, a single strut is isolated to highlight the
distribution of the aluminum powder.
The solid polymer struts are bounded by con-
cave foam cells. With a concentration of only 5 wt
sequent to the initial linear loading region, was
%, it is seen in Figure 3(a) that the aluminum
also measured as a function of specimen density.
powder particles are widely separated. Increasing
the fraction of aluminum to 10% yields a some-
what higher density of particles within the struts
RESULTS
and cell walls, as shown in Figure 3(b). It is also
apparent that a number of powder particles have
Microscopy of Unfilled CRETE
pulled out of the polymer upon fracture leaving
Scanning electron microscopy was conducted on behind small dimples. Figure 3(c) shows the typ-
foam specimens to determine their structure and ical distribution of Al powder in a strut at 30 wt
morphology. Figure 2(a) shows the typical micro- %. The faint features decorating the cell walls are
structure of a foam specimen having a density of Al particles that reside just below the free sur-
0.12 g/cm3. The cell microstructure is quite uni- faces of the internal cell walls.
form and equiaxed with respect to the rise direc-
tion. Using a line intercept technique,7 the aver- Compression Testing of Unfilled CRETE
age cell diameter is approximately 0.7 mm. The A typical compression test is shown in Figure 4
dimples on the cell walls represent the areas of for a foam having a density of 0.14 g/cm3. After an
2728 GOODS ET AL.

Figure 4 Compression test results for a foam, r* 5

0.14 g/cm3. Yielding is abrupt and is followed by a
sustained plateau region.

initial linear loading regime, the foam specimens

show relatively abrupt yielding, followed by a sus-
tained plateau region over which there is little
increase in stress with increasing strain. The ini-
tial linear region defines the modulus of the foam.
The broad plateau region results from the col-
lapse or cell wall buckling of the foam3 and, as
previously indicated, is referred to as the collapse
stress. At the lower densities, such as that shown
in Figure 4, the stress after the plateau actually
drops, giving rise to a yield pointlike behavior. At
larger strains, subsequent to this plateau region,
the stress increases as the foam begins to densify.
Moduli, derived from the compression tests, as
a function of foam density are shown in Figure 5,
along with tensile data from a previous study.5
All of the data can be fit with a power-law expres-
sion with respect to density of the form, as fol-
lows:

E*}~ r *! n (2)

where E* is the modulus of the foam, r* is the

foam density, and n is the density exponent. Over
the range of density shown in Figure 5, the data
are well fit for a density exponent of n 5 1.7.
Figure 6 shows the same data plotted on a log–log

Figure 3 (a) Foam specimen (0.29 g/cm3) containing

5 wt % Al. At this low loading, the aluminum powder
particles occupy only a small volume fraction of the the debonding of the powder from the polymer strut. (c)
strut. (b) Foam specimen (0.28 g/cm3) containing 10 wt Foam specimen (0.29 g/cm3) containing 30 wt % Al. A
% Al. Along with the higher concentration of powder high concentration of particles are seen randomly dis-
particles, dimples are visible. These features arise from tributed within the strut and the cell walls.
 PARTICLE-STRENGTHENED POLYURETHANE FOAM 2729

Figure 5 Density dependence of the modulus in ten-

Figure 7 Power-law dependence of the collapse
sion and compression.
stress.

scale with the density normalized to that of the

solid polymer, 1.2 g/cm3. Thus, a specimen having an important parameter in the design of cushions
a normalized density of 1 is equivalent to a fully for shock or impact mitigation because it repre-
dense monolithic polymer. In Figure 6, the data sents the onset of the mechanical instability
fall on a straight line with a slope of 1.7, confirm- of the foam microstructure.3 It, too, exhibits a
ing the power-law behavior. The intersection of power-law dependence with respect to foam den-
the best-fit curve of the data and the ordinate for sity, although the density exponent, 2.1, is some-
r*/rPU 5 1 can be taken to be the modulus of the what higher than that exhibited by the modulus.
solid polymer. By doing so, we arrive at a value of When plotted on logarithmic axes, the data shown
the solid modulus of 2.5 GPa. This is within the in Figure 7 would fall along a straight line similar
range of reported values for E PU of 1.6 to 2.7 to the data shown in Figure 6 (although with a
GPa.6,8 scope equal to 2.1).
The collapse stress s *c for the foam specimens
as function of density is shown in Figure 7. This is Compression Testing of Al-Filled CRETE
In order to directly observe the influence of the
aluminum additive on the mechanical properties
of the unfilled foam, a common reference state
must be established. We choose to compare be-
havior between the unfilled and Al-filled foam on
the basis of polymer density. For the case of the
filled foams, the polymer density is calculated
from the measured aggregate density of the filled
specimen and the weight fraction of the alumi-
num added. For example, for a specimen having
an aggregate density of 0.68 g/cm3 that contains
10 wt % Al, the calculated polymer density is 0.63
g/cm3.
With this basis of comparison, the influence of
the Al powder mechanical properties is readily
seen in Figure 8. This figure shows a series of
Figure 6 Log–log plot showing power-law behavior compression stress–strain curves for loaded spec-
between modulus and normalized density. The inter- imens having a constant polymer density of 0.28
section of the best fit curve and the ordinate for r*/rPU g/cm3. It is clear from Figure 8 that progressively
5 1 is used as the modulus of the solid polymer here- increasing the fraction of powder in a foam that
after. has a constant polymer density increases both the
2730 GOODS ET AL.

fect sites, allowing for easier crack initiation and

propagation. At low concentrations, this effect is
small, and ductility is largely unaffected. This
same dewetting phenomena accounts for the dra-
matic yield point behavior shown in Figure 8 that
was exhibited by the specimen containing the
highest loading fraction of filler. As the filler de-
wets during deformation and the adhesion be-
tween it and the matrix is compromised, the com-
posite modulus is reduced. When tested at a fixed
displacement rate, the decreased modulus results
in a drop in the observed load.
The influence of the rigid filler phase on the
modulus of the foam over a wide range of polymer
density is shown in Figure 9. This figure com-
pares the original modulus data for the unfilled
Figure 8 Comparison of the compressive behavior of
foam shown in Figure 5 to specimens containing
the Al-filled CRETE to the unfilled CRETE foams.
controlled additions of the aluminum powder
filler. The abscissa is the calculated polymer den-
sity, as was described at the beginning of this
elastic modulus and the strength of the foam. In section. The solid line represents a best fit
all cases, the Al-loaded specimens have moduli through the unfilled foam data previously pre-
that are greater than that of the unloaded speci- sented in Figure 5. The corresponding data points
mens. The magnitude of the increase is depen- are omitted for clarity. The measured moduli for
dent upon the amount of filler added. Additions of the specimens containing 5, 10, 30, and 50 wt %
5 and 10 wt % have a small effect on modulus. In are also shown in this figure. It is seen that the
the examples shown in Figure 8, 5 and 10 wt % power-law dependence, n 5 1.7, denoted by the
additions of powder are found to increase the traces through each set of data, still describes the
modulus from 175 to 190 and 206 MPa. These density dependence of the modulus for each load-
represent 8 and 17% increases, respectively. The ing fraction of aluminum.
modulus of the foam specimen containing 30 wt %
filler exhibits a modulus of approximately 280
MPa, a 60% increase.
The aluminum filler also affects the collapse
stress of the foam. The collapse stress of the un-
filled foam is 5.3 MPa and 5 and 10 wt % additions
of aluminum increase s *c to 6.0 and 6.4 MPa (13
and 21% increases, respectively). It is also seen
from Figure 8 that at 30 wt % Al, the strength of
the foam is about 25% greater than that of the
unloaded foam, 8.3 MPa.
However, the sample loaded to 30 wt % Al fails
at a strain of 0.13 (for the other specimens, the
tests were terminated at about strains of about
0.25). Unlike the unloaded foam specimen or
specimens with low loading fractions, this speci-
men showed evidence of cracking with increasing
strain. Such cracking at low strains is uncharac-
Figure 9 Influence of Al powder additions on the
teristic of polyurethane foams of this density and
CRETE modulus. Modulus increases with loading per-
stems from the fact that some of Al powder is cent and still exhibits the 1.7 power law dependence on
either poorly bonded to the polymer matrix ini- CRETE density. The solid line represents the best fit
tially or dewets during deformation. Thus, while through the data in Figure 5. The abscissa is the cal-
the powder may act to strengthen the foam, some culated density of each foam specimen, regardless of
of the individual particles act as preexisting de- the loading fraction of aluminum, as defined by eq. (1).
 PARTICLE-STRENGTHENED POLYURETHANE FOAM 2731

It is clear from Figure 9 that aluminum load- the density dependence of the modulus for
ings of 5 and 10 wt % have only a small effect on CRETE (and other) foams. In their model, a foam
the modulus of the foam. It is only at the higher is described as an array of cubic cells consisting of
loading fractions of 30 and 50 wt % (volume frac- struts (that define the cell edges) and walls (that
tions of about 0.15 and 0.30, respectively) that a enclose the cells in a closed-cell foam). These cells
significant increase in modulus is observed. For are then staggered so that corners of one cell rest
example, at relatively low polymer densities (r upon the midpoint of adjacent cells. Such a struc-
5 0.3 g/cm3), the modulus of the loaded foams can ture does not correspond to the actual geometric
be more than doubled by the addition of 50 wt % characteristics of a real foam, nor can it be repro-
Al powder. duced to fill space. It does, however, capture the
That the modulus of the foam is found to in- critical physical processes that govern the defor-
crease at powder concentrations greater than 30 mation processes of a cellular structure. De-
wt % is not unexpected. The effect may be attrib- scribed in this fashion, the modulus of such a
uted to the formation of a nearly continuous net- structure is related to the elastic deflection of the
work of aluminum powder at these higher loading struts that are oriented normal to an applied far
fractions. For a random dispersion of spherical field load.
particles, it has been shown that a continuous For an open-cell foam, where all of the polymer
network is formed at the critical volume fraction resides in the cell struts, Gibson and Ashby arrive
30%.4 This volume fraction corresponds to an alu- at the following simple expression for the density
minum weight concentration of approximately 50 dependence of the modulus of a foam:
wt %, about where the most significant increase
in foam modulus was observed. E*
Es
<
r*
rs S D 2
(3)
DISCUSSION
Relationship Between the Modulus and Strength of Equation (3) predicts that a parabolic relation-
Foams and Their Structure ship should exist between the modulus of the
foam and its density. The data in Figure 5 sug-
The dependence of both the modulus and the elas- gest, however, that the power-law exponent is
tic collapse stress of a cellular foam can be under- less than 2. This discrepancy can be found in the
stood in terms of the mechanical properties of the fact that the foam has a closed-cell structure
polymer material from which the cell struts (and rather than an open-cell structure. As such, some
in the present case, the cell walls) are made and fraction of the polymer resides in the cell walls or
the deformation mechanics of the cellular struc- faces rather than solely in the struts.
ture itself. Elastic moduli are related principally If the fraction of polymer contained in the cell
to the bending stiffness of the members compris- struts is f, then the fraction contained in the cell
ing the cellular structure, while the elastic col- walls is (1 2 f). The modulus of a closed-cell foam
lapse is caused by the elastic buckling of these (ignoring internal gas pressure) results then from
same members. two contributions. The first component is strut
For the discussion presented below, the impor- bending, as for open-cell foams. The second com-
tant cell strut/wall properties are the solid poly- ponent arises from the stretching of the cell wall
mer density r s and its modulus E S . The impor- faces.10 –12 Gibson and Ashby3,9 derive the modu-
tant structural features for the analysis of the lus of a closed-cell foam, which accounts for these
modulus and the collapse stress are the normal- two components, as follows:
ized density of the foam r */ r s (as before, r* is the

S D S D
density of the foam) and whether or not the cells
E* r* 2
r*
are open or closed. In this regard, a parameter, f, < f2 1 ~1 2 f ! (4)
is defined as the fraction of material in the cell Es rs rs
struts. For an open-cell foam, f 5 1, while for a
closed-cell foam, where some of the polymer is in The first term on the right describes the contri-
the cell walls, it is less than 1. bution of the cell struts to the modulus, while the
second term accounts for the stretching of the cell
Modulus walls. Equation (4) predicts a pseudo power-law
Gibson and Ashby3,9 have put forth a simple relationship with respect to density, in which the
model that has been shown to accurately describe density exponent increases continuously from 1 to
2732 GOODS ET AL.

“Euler buckling load.” The derivation of the Euler

buckling load is a well-known problem in mechan-
ics14 and, when incorporated into the geometric
model of Gibson and Ashby, can be used to predict
the mechanical collapse of a foam as

s *c
Es
5C3 S DS S D D
r*
rs
2
11
r*
rs
1/2 2
(5)

where the constant C contains all of the physical

and geometric factors relevant to the foam poly-
mer and structure. In principle, the constant C
can be explicitly calculated. However, it is more
expedient to fit the measured collapse stress data
shown in Figure 7 to eq. (5), using C as a fitting
Figure 10 Comparison of modulus measurements to parameter. As shown in Figure 11, eq. (5) best fits
eq. (4). Note that density is normalized to the density of the data for C 5 0.02. As for the modulus predic-
the solid polymer. The partitioning factor is held con- tion, the prediction derived by Gibson and Ashby
stant at f 5 0.9. for the collapse stress accurately reflects the den-
sity dependence of the foam over the range of
densities measured.
2 with increasing r */ r s . Over the density range of
the measurements reported here, eq. (4) describes
the functional dependence of modulus on density Prediction of Al-Filled Foam Properties
with an exponent close to that which best fits the
The strut of a particulate-filled polyurethane
data shown in Figure 5.
foam is essentially a polymer–matrix composite
The data shown in Figure 5 can be quantita-
consisting of two or more separate constituents, a
tively compared to eq. (4) using known values for
continuous matrix phase and a discontinuous
r s and E s and f. For the density of solid polyure-
filler phase. The change in properties of a foam
thane, we use a value of r s 5 r PU 5 1.2 g/cm3.6
containing a rigid phase (from that of an unloaded
The value for the modulus of solid polyurethane is
foam) should track with the effect of the additive
less well established and varies considerably, de-
on the properties of the monolithic polymer. Thus,
pending on the precise formulation, processing
it should be possible, therefore, to predict the
conditions, and product form. Since the value for
the modulus of solid polyurethane is less certain,
we use the value obtained from Figure 6, E s
5 E PU 5 2.5 GPa. For f, we use a value of 0.9.
(Note: We have not independently measured f for
this foam system; the value used, f 5 0.9, is
typical for polyurethane foams.13) Using these
values, the comparison between eq. (4) and the
data in Figure 5 is shown in Figure 10. The figure
shows that the model represents the data well
over the range of densities examined experimen-
tally.

Collapse Stress
The dependence of the plateau stress on foam
density has also been address by Gibson and
Ashby.3,9 As the applied load increases, cell struts
that are parallel to the load become unstable.
This instability is termed “lateral buckling,” and Figure 11 Comparison of collapse stress measure-
the applied load necessary to cause it is called the ments to eq. (5) versus normalized density.
 PARTICLE-STRENGTHENED POLYURETHANE FOAM 2733

behavior of the Al-filled foam by understanding where M c and M m are any composite and matrix
the effect of the filler phase on the host matrix. moduli and matrix and u f is the volume fraction of
Much work has been done in regards to char- the filler phase.
acterizing the effect of filler phases on the prop-
erties of resin matrix composites. Palumbo et al. 7 2 5nm
have studied the mechanical behavior of a com- A5 (7)
8 2 10 n m
posite consisting of a fully dense epoxy matrix
and hollow glass microspheres (GMB).15 Results
of this work revealed that the modulus of the where n m is Poisson ratio of the matrix
composite decreased with increasing loading frac- and
tion of the GMB, which has a lower modulus than
the host epoxy matrix. In another study, Domeier ~M f /M m! 2 1
B5 (8)
et al. found much the same effect in a similar ~M f /M m! 1 A
epoxy/GMB composite system.16 In contrast,
Monette et al. observed an increase in Young’s
where M f is the modulus of the filler.
modulus when silica beads were dispersed
Specifically, for the Young’s modulus of a par-
throughout an epoxy matrix.17 The experimental
ticle-filled composite,
data revealed an increasing elastic modulus as

F G
the volume fraction of filler was increased. Simi-
lar findings were reported for an epoxy/alumina 1 1 AB u f
Ec 5 Em (9)
system.16 In both instances, the modulus of the 1 2 Buf
additives were greater than that of the matrix
yielding composite structures of overall greater where E c is the composite modulus and E m is that
stiffness. of the host matrix.
Beyond these experimental investigations, The measured moduli of traditional reinforced
there have been many other studies18,19 devoted polymer composites have been compared favor-
to modeling the elastic properties of composites. ably with predictions based on eq. (9).15,16
The two most widely used methods for predicting
the mechanical properties of particulate-rein- Modulus for Al-Loaded CRETE
forced composites are those developed by Hashin
and Shtrikman20 and Kerner.21 It has been shown in Figures 2 and 3 that the Al
In the latter reference, a composite is consid- powder is uniformly incorporated into the cell
ered to be a distribution of spherical particles strut and cell wall elements of the foam. There-
suspended in a homogeneous matrix. The analy- fore, these elements of the foam structure can be
sis assumes that there exists an average state of modeled as a composite with a matrix phase of
stress and strain within each particle when sub- solid polyurethane and a reinforcing phase of Al
jected to a hydrostatic stress. Kerner then formu- powder. As such, it should be possible to use the
lated expressions for the bulk and shear moduli of Halpin–Tsai equation in conjunction with the
a composite in terms of the bulk and shear moduli Gibson and Ashby model for the modulus of a
of both the matrix solid and filler phases, their closed-cell foam to predict the modulus of the
respective volume fractions, and Poisson’s ratio of Al-filled foam. In this case, the matrix modulus
the matrix. E m is the modulus of solid polyurethane E PU. E c
The resulting Kerner equations are quite com- then represents the modulus of the filler-loaded
plicated and can be difficult to apply. Halpin and strut (i.e., the composite) and is substituted for E s
Tsai22–24 showed that by grouping terms that per- (the modulus of the foam strut material). In an
tain to the Poisson effect and particle geometry analogous fashion, r s is now the density of the
and, also, separating terms, which depend on the PU–Al powder composite at each loading fraction
elastic properties of the constituents, the Kerner of aluminum and is termed rcomp. Finally, r* is
equations can be generalized to the following replaced by the total aggregate density of a spec-
form: imen, rsample. These new terms for the modulus
and density of the PU–Al powder composite and
specimens can be substituted into eq. (4).
M c 1 1 AB u f The final governing expression for the modulus
5 (6)
Mm 1 2 Buf of the powder-loaded PU foam is given as
2734 GOODS ET AL.

Kerner equation and Gibson and Ashby model

accurately predicts the density dependence of the
modulus as well as the increase in modulus for
each loading fraction. It accurately predicts that
there is little effect on foam modulus for low (5
and 10 wt %) loading fractions of aluminum.
Equation (10) also predicts the significant in-
crease in foam modulus for 30 and 50 wt %.

Collapse Stress for Al-Loaded CRETE

Equation (5) predicts that for foams of constant
density, the collapse stress should vary as a linear
function of the modulus of the strut and cell wall
material. By combining eqs. (5) and (9), the pre-
dicted collapse stress is

F GS D
1 1 AB u f r sample 2
s *c 5 0.02 3 E PU
1 2 Buf r comp
Figure 12 Combined Halpin–Tsai and Gibson and
Ashby predictions for the modulus of a powder-filled
foam.
S S DD
3 11
r sample
r comp
1/2 2
(11)

should track with the collapse stress values

shown in Figure 8. Such a comparison is shown in

F GF S D
Figure 13, where the data points represent the
1 1 AB u f r sample 2
measured collapse stress values from Figure 8,
E* < E PU f2
1 2 Buf r comp and the trace represents the expected collapse

1 ~1 2 f ! S r sample
r comp DG (10)
stress based on eq. (11). The trace representing
the prediction is always somewhat higher than
the actually measured collapse stress. This is to
be expected as the Halpin–Tsai model assumes
Substituting appropriate values into eqs. (7) and
complete adhesion between the filler phase and
(8), the parameters A and B can be calculated. As
the matrix. Since both the microscopy in Figure 3
before, the modulus of solid polyurethane is taken
to be 2.5 GPa. The modulus of aluminum is 68.9
GPa.25 Poisson’s ratio for polyurethane is taken
as 0.3526 and the partitioning factor f was once
again held constant at 0.9.
Figure 12 compares the experimentally mea-
sured moduli to the predicted moduli for Al-
loaded foam. The data for the aluminum filled
foam are the same as those previously presented
in Figure 9. The solid line corresponding to the
unfilled foam represents the modulus predicted
by eq. (4). The data points have been omitted for
clarity. The additional traces show the predicted
density dependence of the modulus from eq. (10)
for each loading fraction of aluminum. Note that
both the data and the prediction for the 5 wt %
aluminum samples are not shown since both show
little difference from the unfilled foam. It is seen
that the final governing equation captures the Figure 13 Comparison of combined Halpin–Tsai and
effect of the aluminum powder on the foam mod- Gibson and Ashby predictions for the collapse stress of
ulus well. The combined Halpin–Tsai form of the a powder-loaded foam.
 PARTICLE-STRENGTHENED POLYURETHANE FOAM 2735

and the mechanical properties shown in Figure 8 porated into the Gibson and Ashby foam model.
suggest less than complete adhesion, it is not By combining these two models, expressions that
surprising that the full strengthening potential of accounted for both the composite nature of the
the filler is not realized. Notwithstanding this, solid and the overall foam structure were created.
the agreement between the data and the collapse These expressions were shown to reasonably pre-
stresses predicted by eq. (11) is quite good. As dict the effect of a rigid filler phase on both the
with the modulus, this generally good agreement modulus and strength of the filled foam.
suggests that the behavior of the powder loaded
foam can be modeled as that of a cellular compos- The authors thank Mike Tootle for his help in sample
ite. preparation and testing and Nancy Yang and Eric
Kleinschmidt for microscopy support. This work sup-
ported under U.S. Department of Energy contract #DE-
CONCLUSIONS AC04-94AL85000.

Scanning electron microscopy revealed that the

structure of the CRETE foam formulation was REFERENCES
uniform with respect to cell size and shape.
Addition of aluminum powder did not alter the 1. Klempner, D.; Frisch, K. C. Handbook of Polymer
uniformity of the foam structure. A random Foams and Foam Technology; Oxford University
nonagglomerated distribution of aluminum Press: New York, 1991; pp. 47–94.
powder throughout the cell struts and walls was 2. Hilyard, N. C.; Cunningham, A. Low Density Cel-
evident. lular Plastics, Physical Basis of Behaviour; Chap-
The modulus of the unfilled foam was in- man and Hall: London, UK, 1994; p. 104.
creased with the addition of aluminum powder. 3. Gibson, L. J.; Ashby, M. F. Cellular Solids, Struc-
The experimental data tended to group by loading ture, and Properties; Pergamon Press: New York,
1988.
fraction with 5 and 10 wt % additions showing
4. Bhattachara, S. K. Metal-Filled Polymers: Proper-
only minor increases in modulus. However, as the ties and Applications; Marcel Dekker: New York,
weight fraction of filler was increased to 30 and 50 1986; pp. 5–15, 172.
wt %, the foam modulus increased significantly. A 5. Goods, S. H.; Neuschwanger, C. L.; Henderson, C.;
2 to 3-fold increase in modulus can be realized at Skala, D. M. J Appl Polym Sci 1998, 68, 1045.
the highest loading fractions examined in this 6. Roff, W. F.; Scott, J. R. Fibres, Films, Plastics and
study. The density dependence of the modulus Rubbers—A Handbook of Common Polymers, But-
was unaffected by the additions of the filler phase. terworths: London, UK, 1971; p. 445.
Over the range of densities examined the 7. Underwood, E. E. Quantitative Stereology; Addi-
strength of the filled foam was greater than that son-Wesley Publishing Co.: Reading, MA, 1970.
of the unfilled foam for all loading fractions of 8. Plastics Digest, Vol. 2; D.A.T.A.: Englewood, CO,
1994, pp. 1301–1304.
aluminum. However, high loading fractions pro-
9. Gibson, L. J.; Ashby, M. F. Proc R Soc London 1982,
duced a yield point behavior and caused the foam A382, 43.
to be susceptible to cracking. This was the result 10. Patel, M. R.; Finnie, I. J Mater 1970, 5, 909.
of the poor adhesion between the aluminum pow- 11. Green, D. J. J Am Ceram Soc 1985, 68, 403.
der and the polyurethane matrix. 12. Christensen, R. M. J Mech Phys Solids 1968, 34,
For the unfilled foam, the good agreement be- 563.
tween the measured modulus and the collapse 13. Reitz, D. W.; Schuetz, M. A.; Glicksman, L. R.
stress with the predictions of eqs. (4) and (5) sug- J Cell Plast 1984, 20, 104.
gests that a model available in the literature, 14. Timeshenko, S.; Young, D. H. Elements of Strength
based on a simple, idealized cell geometry, can be of Materials, 4th ed.; D. van Nostrand & Co. Inc.:
useful in these mechanical and physical proper- New York, 1962; pp. 212, 342.
15. Palumbo, M.; Donzella, G.; Tempesti, E.; Ferruti,
ties of foams. By considering the strut of the filled
P. J Appl Polym Sci 1996, 60, 47–53.
foam to be a composite consisting of a rigid phase 16. Domeier, L.; Skala, D. M.; Goods, S. H.; Neusch-
within a polyurethane matrix, the increase in the wanger, C. L.; Tootle, M. L.; Lu, W. SAND98-8204,
modulus and strength of the Al-loaded foam could Sandia National Laboratories, Livermore, CA,
be understood. A well-known expression that de- 1998.
scribes the effect of rigid, spherical additives on 17. Monette, L.; Anderson, M. P.; Wagner, H. D.; Muel-
the modulus of monolithic composites was incor- ler, R. R. J Appl Phys 1994, 75, 1442.
2736 GOODS ET AL.

18. Paul, B. AIME Trans 1960, 218, 36. 24. Tsai, S. W. US Government Report AD 834851,
19. Bohme, R. D. J Appl Polym Sci 1968, 12, 1097. 1968.
20. Hashin, Z.; Shtrikman, S. J Mech Phys Solids 1963, 25. Smithells, C. J.; Brandes, E. A. Metals Reference
11, 127. Book, 5th ed.; Butterworths: Woburn, MA;
21. Kerner, E. H. Proc Phys Soc London 1956, B69, 808. 1976.
22. Halpin, J. C. J Comp Mater 1969, 3, 732. 26. Birley, A. W.; Haworth, B.; Batchelor, J. Physics of
23. Halpin, J. C. Primer on Composite Materials Anal- Plastics—Processing, Properties and Materials En-
ysis, 2nd ed.; Technomic Publishing: Lancaster, gineering, Oxford University Press: New York,
PA, 1992; p. 161. 1991; p. 255.