TENSILE AND TEAR BEHAVIOR OF ANISOTROPIC

DOUBLE NETWORKS OF A BLACK-FILLED NATURAL

RUBBER VULCANIZATE
G. R. Hamed* and M. Y. Huang
Polymer Science Department, University of Akron, Akron, OH 44325-3909

ABSTRACT
Double networks of a black-filled natural rubber composition have been prepared by partially curing a sheet,
stretching it, and then completing cure. Upon release, a double network retracts to a residual extension ratio,
αr . Samples cut perpendicular to the stretch direction have stress–strain responses like the isotropic single
network, while parallel samples have enhanced stiffness and tensile strength, and reduced extensibility. Tensile
strength is rather weakly dependent on αr . Tear strengths of the double networks, determined using edge-cut
strip specimens, exceed that of the single network for low αr . However, when αr is high, double networks have
very low tear strengths. Consistent with previous studies, high tear strengths are associated with extensive
longitudinal cracking.

INTRODUCTION
Plastic films and fibers are oriented by stretching and drawing during processing to enhance
stiffness and strength. Molecular chain orientation is maintained after processing because the
chains are “frozen” by the glassy state and/or by crystallization. With rubbers, on the other
hand, orientation produced during simple processing will decay after processing, because of high
chain mobility. However, rubber vulcanizates with permanent chain orientation can be produced
by preparing, so-called, “double” or “composite” networks.1 11 Here, a rubber sample is lightly
crosslinked (first network), then, stretched to extension ratio, α0 , and further crosslinked to form
a “second” network. Upon release, the second network resists the retraction of the first network—
resulting in a sample with a residual extension ratio (set) of αr .1,3,9 Early studies of double
networks employed gum elastomers and were concerned with their elastic properties. If Gaussian
chain statistics and a classical phantom or affine elasticity model are used, double networks are
expected to be isotropic.7 10 However, more realistic elasticity models, such as Mooney–Rivlin,
can account for anisotropy in double networks of Gaussian chains.5 Recently, Santangelo and
Roland6,12 have shown that gum natural rubber double networks crystallize at lower strain and
can have enhanced tensile strength and fatigue life compared to conventional single networks.
In this paper, tensile properties and cut growth resistance of double networks based on a black-
filled natural rubber compound are discussed. Previously,13 we have shown the importance of the
development of strength anisotropy at a crack tip to resist facile propagation perpendicular to
loading. If sufficient anisotropy develops, substantial longitudinal cracking occurs at the crack tip
and catastrophic fracture is inhibited. These crack growth studies employed conventional, filled
NR vulcanizates composed of randomly oriented chains in the undeformed state. Now, we examine
crack growth in double networks which already have chain orientation prior to deformation.

EXPERIMENTAL
The rubber compound is designated NII3 and contains in phr: NR-100; N330-40; ZnO-5; stearic
acid-1; Ageriter Stalite S-1; TBBS-0.8 and sulfur-3.2.
Masterbatches were prepared in a water-cooled Banbury internal mixer having a total volume
of 250 mL. A fill factor of 0.8 was used, and the mixing speed was 97 rpm. The rubber first was
masticated in the Banbury for one minute before adding the carbon black and mixing for another

* Corresponding author. Tel: 330-972-6831; fax: 330-972-5290; e-mail: [email protected]

846
 DOUBLE NETWORKS OF A FILLED VULCANIZATE 847

minute; then, all ingredients, except for the curatives, were added, and mixing continued for five
minutes before dumping. Masterbatches were cooled to room temperature, before mixing with
curatives, within two hours.
Curatives were mixed into masterbatches on a 30 cm wide two-roll mill. Four masterbatches
from the Banbury were combined on the two-roll mill at a nip setting of about 2 mm and allowed
to form a rolling bank. The curatives were slowly added to the rolling bank with 3/4 cuts, then
the stock was removed from the mill as a narrow roll. The roll was then passed endwise into the
nip (about 1 mm) ten times. Finally, the stock was sheeted off at a nip of about 2 mm. Stocks
were stored at room temperature more than 24 hours before vulcanization.

VULCANIZATION CHARACTERISTICS

Vulcanization was characterized using a Monsanto Model R-100 oscillating disk rheometer, in
accordance with ASTM D 2084-88. The arc amplitude was 3◦ . Samples were tested at 135 and
150 ◦ C.
First cure.—The first vulcanizations were carried out using a Teflon-coated aluminum mold
at 135 ◦ C and 35 tons. Samples were cooled in air and stored at room temperature longer than
24 hours, but less than 96 hours, before stretching and applying the second cure.
Stretching.—A first-cured sample (100 mm in the stretch direction, 120 mm wide, and 2 mm
thick) was held in grips, which were connected to the Instron by pins. Two flat steel bars were
inserted into the sides of the grips [Figure 1(a), (b)]. The upper grip was fixed to the bars, while
the lower one was not. The rate of grip separation was 200 cm/min. When a desired extension
ratio, α0 , was reached, the lower grip was fixed to the bars to form a “frame” consisting of bars
and grips, which held the sample in the stretched state [Figure 1(c)]. The “frame” was removed
from the Instron and the stretched rubber subjected to a second cure within five minutes.
Second cure.—The stretched rubber was sandwiched between two Mylar films backed by steel
plates, and pressed under a load of 3 tons [Figure 1(d), (e)] at 150 ◦ C. After the second cure, the
sample was released [Figure 1(f)]. The time interval from the second cure to release was about
five minutes. The released sample was stored at room temperature for a minimum of 24 hours,
and then the residual extension ratio, αr , was measured.
Tensile testing.—Tensile tests were performed in accordance with ASTM D 412-89, using a
“C” type dumbbell die. Benchmarks were placed on the middle part of a specimen at a separation
of 25.4 mm. A minimum of three specimens was tested for each sample. The rate of grip separation
was 200 cm/min, and the testing was done at room temperature. The elongation was measured
directly using a ruler; the load at intervals of 50% elongation was noted using an event marker.
Cut growth.—Rectangular strips (12.5 mm wide) were cut-out using a die. Specimens were
cut at one edge at their midpoint using a razor blade. Care was taken to ensure that the cut was
perpendicular to the tensile testing direction and the tip was normal to the major surfaces. The
initial cut length was measured with a traveling microscope prior to testing. The specimen was
very slightly strained to facilitate observing the tip. The initial grip separation was 70 mm, and
the crosshead speed was 10 mm/min. Photographs were taken with a Nikon camera focused on
the cut tip during loading. The rupture path of a broken specimen was recorded photographically,
and the details of crack growth near the original cut tip were observed with an optical microscope.
Testing directions.—For the (uncut) normal tensile tests and the cut growth tests, specimens
were cut-out from double networks both parallel (k) and perpendicular (⊥) to the (pre-)stretch
direction (Figure 2). (Note that a k specimen contains a cut which is ⊥ to the pre-stretch direc-
tion). Specimens were cut from the portion of double network sheets, which had been determined
to undergo uniaxial extension. The test portion was selected by drawing a square grid on a sheet,
−1/2
stretching it, and noting the region where α⊥ = αk , within 10%.

RESULTS AND DISCUSSION

Table I gives initial (α0 ) and residual (αr ) extension ratios for double networks in which
the first crosslink density ν1c = 2.29 × 10−4 mole/cc and the final, total crosslink density, νT
= 5.58 × 10−4 mole/cc. The first crosslink density was determined after cure at 135 ◦ C, while νT
was determined after twice cure (i.e., 135 ◦ C then 150 ◦ C) of an unstretched conventional (i.e.,
848 RUBBER CHEMISTRY AND TECHNOLOGY Vol. 71

Fig. 1.—Photographs showing the preparation of uniaxial double network samples.

single) network. Crosslink densities are apparent values determined from swelling measurements
using the Flory–Rehner equation. As expected, αr increases as α0 increases.

TESTING: UNCUT SPECIMENS

Tensile stress–strain curves for these networks are shown in Figure 3. Strain is referenced
to specimen dimensions after (release from) the second cure. Specimens designated D1–D5 and
D10 –D50 were cut-out parallel and perpendicular, respectively, to the direction of strain during
the second cure. S denotes the response for the isotropic “single” network, obtained using a two-
step cure—first at 135 ◦ C then at 150 ◦ C, just like the double networks, but without stretching
during cure.
Parallel specimens are stiffer, especially at high strains, and somewhat stronger than the
isotropic control. Furthermore, modulus increases and breaking strain decreases as αr increases
(D1 → D5). Using phenomenological models of rubber elasticity, the stiffening of double networks
parallel to the pre-strain direction has been predicted.5 Parallel specimens of double networks at
ease may be modeled to consist of a first network extended to αr , and a second network compressed
to αr /α0 . Thus, the chains of the first network are already stretched, causing increased stiffness
 DOUBLE NETWORKS OF A FILLED VULCANIZATE 849

Fig. 2.—Schematic showing designation of cut-out specimens for double networks.

of parallel specimens compared to the isotropic single network. The higher the pre-extension of
the first network, the lower the elongation to break of parallel specimens.
All perpendicular samples from the double networks have stress–strain curves remarkably
similar to the isotropic single network, although stress at break (σb ) and strain at break (b )
decrease slightly from D10 (σb = 33.0 MPa, b = 500%) to D50 (σb = 27.0 MPa, b = 420%). With
the two network model, in the perpendicular direction at ease, the first network is compressed to
(1/αr )1/2 , while the second network is extended to (α0 /αr )2 . Table II lists at ease values of pre-
extension: (1) (αr ), for the first network of parallel specimens and (2) (α0 /αr )2 , for the second
network of perpendicular specimens, along with corresponding breaking elongations. Although
elongation to break generally decreases with pre-extension in each case, perpendicular samples
are more extensible than parallel specimens, when compared at similar pre-extensions. Thus,
breaking strains of the double networks are not uniquely dependent on finite extensibility of the
pre-extended first (parallel specimens) or second (perpendicular specimens) network.
It is convenient to define a parameter to further illustrate the anisotropy of the double net-
works:
σk ()
I ≡ (1)
σ⊥ ()

Table I
Initial and Residual Extension Ratios for the First Set of Double Networksa
Sample α0 αr

D1 1.47 1.30
D2 1.92 1.72
D3 2.28 1.96
D4 3.00 2.21
D5 4.10 2.80
a ν1c = 2.29 × 10−4 mol/cm3 ; νT = 5.58 × 10−4 mol/cm3
850 RUBBER CHEMISTRY AND TECHNOLOGY Vol. 71

Fig. 3.—Tensile stress–strain curves for the single (S) and double (D) network samples. D1, D2, D3,
D4, D5: Parallel; D10 , D20 , D30 , D40 , D50 : Perpendicular. Arrows indicate ultimate properties for single
network and perpendicular specimens.

Figure 4 shows that I150 (i.e., at 150% strain) is a linear function of the residual extension ratio;
as expected, the data extrapolate to I150 = 1.0, when αr = 1.0.
We note that parallel specimens of the double network are stiffer and stronger than the single
isotropic network. Of course, another way to increase the stiffness of a single network is simply
to increase the crosslink density by increasing the level of curatives. It would be interesting to
compare the strength of parallel specimens of a double network with that of a single conventional
network at similar stiffness.
It has been reported6 for gum NR vulcanizates that in the parallel direction double networks
are stiffer and have slightly lower tensile strength than a single network. The contrast with the
present results may be due to differences in crosslink apportionment, strain during the second
cure, and/or the role of filler.

Table II
Breaking Elongations for Parallel and Perpendicular Specimens
Along With At Ease Values of Pre-extension
← Parallel → ← Perpendicular →

First Second
Network Network
Breaking Pre-extension Breaking Pre-extension
Sample Strain (%) αr Sample Strain (%) (α0 /αr )2

D1 390 1.30 D10 500 1.28

D2 355 1.72 D20 470 1.25
D3 315 1.96 D30 500 1.35
D4 265 2.21 D40 455 1.85
D5 225 2.80 D50 420 2.14
 DOUBLE NETWORKS OF A FILLED VULCANIZATE 851

Fig. 4.—Anisotropy parameter, I150 , as a function of residual extension ratio.

To better compare the low strain behavior of the double and single networks, equilibrium
stress–strain measurements were carried-out in a “creep” mode on another set of double networks
(D7–D9), by placing small loads on (marked) specimens and determining the equilibrium extension
with a cathetometer. Table III shows the residual extension ratios for the double networks which
were tested, and Figure 5 shows the equilibrium stress–strain curves, plotted in Mooney–Rivlin
fashion. While anisotropy is clearly evident for the double networks at high strains (consistent
with Figure 3), it is noteworthy that up to about 30% strain (i.e., 1/α > 0.77), all samples
(isotropic, perpendicular, parallel) exhibit rather similar stiffness.

PRECUT SPECIMENS

Measurements were carried out on two other sets of double networks—one with low αr
= 1.35–1.46 and the other with high αr = 2.4–2.65. (Note: The range of αr in each case represents
the variability associated with the reproducibly of preparing double networks.) L(k) and L(⊥)
refer to specimens with low αr , which were cut either parallel or perpendicular to the direction of
stretching during the second cure. Likewise, H(k) and H(⊥) indicate high αr . Comparison with
Table I shows that L(k) and L(⊥) correspond approximately to D1 and D10 , respectively, while
H(k) and H(⊥) correspond to D5 and D50 , respectively.

Table III
Initial and Residual Extension Ratios for the Second Set of Double Networksa
Sample α0 αr

D7 2.00 1.71
D8 2.86 2.24
D9 3.88 2.59
a ν1c = 2.29 × 10−4 mol/cm3 ; νT = 5.58 × 10−4 mol/cm3
852 RUBBER CHEMISTRY AND TECHNOLOGY Vol. 71

Fig. 5.—Mooney-Rivlin plots for samples listed in Table III.

D7, D8, D9: Parallel; D70 , D80 , D90 : Perpendicular.

SINGLE CONVENTIONAL NETWORK

Figure 6 shows the effect of cut size on strength for the single network. Strength decreases with
increasing cut length, and drops quite abruptly at a critical cut length, lc , of about 1.0 mm. Note
that the cut lengths of specimens S-1 and S-2 are similar, but their strengths are quite different.
The strength of S-1 falls on the “tail” of the upper curve, while that of S-2 is on the “head” of
the lower line. S-1 is approximately 50% stronger than S-2. Figure 7 shows catastrophic rupture
paths for these specimens. S-1 has a much more extensive fracture path parallel to the loading
direction prior to catastrophic rupture than does S-2. Furthermore, examination of the crack tip
regions under the optical microscope revealed that S-1 has more extensive fine-scale longitudinal
cracking than S-2. Similar results and their significance have been reported previously.13

DOUBLE NETWORK: PARALLEL SPECIMENS

Figure 8 gives tensile strength against cut length for parallel samples with low residual strain.
(Tensile strength is defined as the breaking force divided by the unstrained, set cross-section
after the second cure.) The critical cut size is about 1.7 mm, nearly twice as long as that for the
isotropic network (Figure 6). Figure 9 shows catastrophic rupture paths for samples L(k)-2 and
L(k)-3 designated in Figure 8. L(k)-2 has extensive longitudinal cracking, while L(k)-3 exhibits
simple lateral tearing.
It has been proposed14 that below the critical cut size, the breaking extension is high and there
is wholesale, bulk crystallization prior to fracture, whereas, above the critical size, the breaking
extension is low and crystallization is confined to a region near the crack tip. Hence, crystallization
at lower strains would be expected to increase lc . Indeed, Roland and Warzel12 have shown that
a double network can crystallize at lower strain than the corresponding single network. For large
cuts, stresses at the cut tip are high and concentrated in a small volume. If a crack suddenly
breaks through this volume, it encounters material which has not crystallized. Rapid propagation
may preclude crystallization ahead of the crack tip, thus allowing catastrophic fracture to occur
 DOUBLE NETWORKS OF A FILLED VULCANIZATE 853

Fig. 6.—Strength versus cut length for conventional single network.

Arrows indicate two specimens examined microscopically in Figure 7.

rather readily. On the other hand, for small cuts, high stress, and hence crystallization, occurs
over a large volume. For catastrophic rupture, an advancing cut tip must tear through material
which is already strain-crystallized. Thus, strength is high.
The effect of cut size on tensile strength for parallel specimens of the double networks with high
residual strain is shown in Figure 10. Unlike the single isotropic network or the double networks
with low residual strain, there is no evidence of a critical cut size. Rather, strength decreases very
sharply with just a small cut. (If there is a critical cut size, it must be very small.) All precut
specimens fail by simple lateral crack propagation (Figure 11 shows the two specimens depicted
in Figure 10).
The fitted curves in Figures 6, 8 and 10 are replotted in Figure 12. The different behaviors are
quite dramatic! In the absence of a precut, the tensile strength difference among, S, L(k) and H(k)
is relatively small (consistent with Figure 3; recall that L(k) is similar to D1, while H(k) is similar
to D5). With a small edge cut just 0.1 mm in length, the strength of H(k) plunges from ∼35 MPa
to ∼10 MPa and the specimen fails by simple lateral cracking, with no (stabilizing) longitudinal
crack growth. On the other hand, the strengths of S and L(k) change only from 31–32 MPa to
26–27 MPa with the introduction of a cut 0.1 mm long. Furthermore, precut S and L(k) specimens
exhibit substantial longitudinal cracking prior to catastrophic rupture. Finally, it is noted that,
at intermediate cut size, the strength of L(k) exceeds that of S, while L(k) is somewhat weaker
than S for deeper cuts. This results from the “drop-off” in strength at the critical cut size being
about twice as large for L(k) (∼10 MPa) compared to S (∼5 MPa).
854 RUBBER CHEMISTRY AND TECHNOLOGY Vol. 71

Fig. 7.—Photographs showing catastrophic rupture paths for specimens:

(a) S-1 and (b) S-2, designated in Figure 6.

In the absence of an intentional precut, rubber test-pieces which have been “clicked-out” with
a die will inherently contain edge “flaws” introduced by the cutting action of the die. Typically,
fracture initiates from the edge flaw with the highest stress raising power—assuming that the
rubber does not contain a large internal defect due, e.g., to poor dispersion or an entrapped
air bubble. The size and shape of the inherent flaws in S, L and H are expected to be similar.
The normal tensile strength of rubber is defined to be that determined in uniaxial tension for
specimens which, of course, contain “inherent” flaws, but no intentional precut, while specimens
to determine tear strength are designed to contain a particular stress concentration. To facilitate
discussion, tensile strength will correspond to uncut specimens, while tear strength refers to those
with precuts. Based on this, S, L(k) and H(k) have similar tensile strengths, but the tear strength
of H(k) is much inferior to that of S and L(k), at least at a cut size of 0.1 mm. Nonetheless,
there must be some small cut size (between 0.1 mm and the inherent flaw size) at which the
tear/tensile(?) strength of H(k) exceeds that of S and L(k).

DOUBLE NETWORKS: PERPENDICULAR SPECIMENS

Firstly, we recall that the regular (uncut) stress–strain curves of the single network and the
perpendicular double networks with αr = 1.3–2.8 (Figure 3, Table I) are rather similar. Nonethe-
less, as shown below, edge cuts result in pronounced differences.
The effect of cut size on the tensile strength of L(⊥) (αr = 1.35–1.46) and H(⊥) (αr = 2.4–
2.65) specimens is shown in Figure 13, along with the single network response. The behavior
of L(⊥) converges with S at the shortest and longest cut sizes. In between, L(⊥) is somewhat
stronger than S and a critical cut size is not readily apparent, although the data suggest that
L(⊥) may have a value of lc (∼1.5–1.8 mm) like L(k) (see Figure 12). All L(⊥) specimens exhibit
auxiliary, longitudinal cracking prior to catastrophic rupture.
 DOUBLE NETWORKS OF A FILLED VULCANIZATE 855

Fig. 8.—Strength against cut length for parallel double network specimens with low αr = 1.35–1.46.
These are similar to the specimen designated D1 (αr = 1.30) in Figure 3. Arrows indicate two test
specimens examined microscopically in Figure 9.

The (uncut) tensile strength of H(⊥) is about 23 MPa compared to nearly 35 MPa for H(k),
while the breaking strain of H(⊥) (see D50 , Figure 3) is about 400% compared to about 200%
for H(k) (see D5, Figure 3). Both are extremely sensitive to small cuts, which propagate by
simple lateral tearing. As shown previously, the strength of H(k) plunges to just 10 MPa with a
cut size of 0.1 mm; likewise, the strength of H(⊥) drops dramatically to about 4 MPa, with a
0.1 mm cut.
To summarize, several important findings are demonstrated in Figures 3, 12 and 13:
(1) With no precut, H(k) specimens (similar to D5) have enhanced tensile strength and greatly
reduced strain at break, while H(⊥) specimens (similar to D50 ) have slightly reduced tensile
strength and elongation, relative to the single network S.
(2) With precuts, both H(k) and H(⊥) specimens have much inferior tear strengths compared
to S.
(3) With no precut, L(⊥) (similar to D10 ) has slightly better ultimate properties than S; L(k)
(similar to D1) is stronger but less extensible than S.
(4) When cut size is near lc of the single network, L(k) and L(⊥) have significantly enhanced
tear strengths relative to S.
Why are the double network specimens with large αr = 2.4–2.65 very cut sensitive (called
notch sensitivity in the plastics industry), while the single network and the double network with
small αr are rather cut insensitive? The ability for longitudinal cracking appears to be quite
important.
856 RUBBER CHEMISTRY AND TECHNOLOGY Vol. 71

Fig. 9.—Photographs showing catastrophic rupture paths for test specimens

L(k)-2 and L(k)-3, designated in Figure 8.

Consider a test specimen which contains a sharp razor cut in its edge, which is then deformed.
At very low strain, the sharp cut tip extends slightly to become a blunter crack tip. With further
deformation, the radius of the crack tip continues to increase and anisotropy in strength develops
at the crack tip as network chains increasingly align in the stretching direction. Crack propagation
can proceed by simple, lateral growth in approximately the same direction as the original cut or by
deviating significantly toward the direction of loading. We have shown previously13 very effective
behavior to resist catastrophic rupture is longitudinal crack growth, which turns back toward the
edge and arrests. Crack tip blunting and energy dissipation at the crack tip inhibit lateral crack
growth and allow anisotropy in strength to increase. Longitudinal crack growth results from a
critical degree of anisotropy and shearing stresses near the crack tip. Originally, it was thought
that H(k) tear specimens, because of substantial initial chain alignment of the “first” network
before testing, might show a high propensity for longitudinal cracking. Clearly, this hypothesis is
wrong! It appears that strain hardening at low strain amplifies stress concentration at the crack
tip prior to enough blunting; fracture proceeds by simple lateral crack growth. Rubbers with a
long “plateau” in their stress–strain response are expected to show good crack tip blunting; if
this is followed by substantial anisotropic structure development, then longitudinal cracking can
develop. Anisotropy is enhanced by strain crystallization (and carbon black reinforcement), which
promotes the formation of a fiber-like structure with interfibril weakness at the crack tip.
The previous explanation for the marked drop in strength of H(k) specimens with very small
cuts hypothesizes that crystallization at too low a strain may actually facilitate catastrophic
cut growth. (It is noteworthy that high density polyethylene has a lower notched Izod strength
than low density polyethylene,15 i.e., higher crystallinity, even when present initially, does not
necessarily ensure increased resistance to crack propagation.) However, high cut sensitivity is
also expected for vulcanizates which are incapable of or have insufficient strain-crystallization;
this possibility for H(k) seems unlikely in view of Roland and Warzel,12 who showed increased
strain-crystallizability of parallel specimens for double networks compared to isotropic single
networks. Nonetheless, we are presently characterizing the strain-crystallizability of parallel as
well as perpendicular specimens. Strain crystallizability could be reduced in double networks
with sufficiently high αr and may depend on the apportioning of crosslinks between the first
and second network. If a double network fractures when the first network has reached a critical
 DOUBLE NETWORKS OF A FILLED VULCANIZATE 857

Fig. 10.—Strength against cut length for parallel double network specimens with high αr = 2.4–2.65.
These are similar to the specimen designated D5 (αr = 2.80) in Figure 3. Arrows indicate two test
specimens examined microscopically in Figure 11.

extension, then for large αr , a double network may fracture at low elongation, when the second
network is still compressed or only little extended, and before the onset of crystallization.
Roland and Warzel did not characterize the strain-crystallizability of perpendicular specimens,
but did demonstrate that undeformed double networks thermally crystallize less readily than un-
deformed single networks. This curious finding may mean that perpendicular specimens, perhaps
at sufficiently high αr , will be less strain-crystallizable than parallel or isotropic specimens. The
low tear strength of H(⊥) is consistent with this hypothesis. The high tensile strength of H(⊥)
would then be attributed to “normal” carbon black reinforcement, as occurs in an amorphous
vulcanizate.
Although double networks with high αr have poor tear resistance, those with low αr can
exhibit improved tear resistance, relative to a simple conventional network. High tear strengths
may depend not only on the extent of crystallization, but also on the degree of blunting prior
to crystallization. Perhaps, before tip strain-crystallization, L(⊥) specimens blunt more than
L(k) specimens, but L(⊥) specimens may undergo less crystallization. The rather similar tear
strengths of L(k) and L(⊥) (see Figures 8 and 13) would then be coincidental, based on their
blunting/crystallization trade-off.
The very poor tear strength of H(⊥) compared to L(⊥) is startling, in view of their rather
similar stress–strain curves. (Refer to Figure 3, which shows that responses for perpendicular spec-
imens (D10 to D50 ) are similar over the entire range of αr .) In the absence of a precut (i.e., normal
tensile specimen), the tensile strengths of H(⊥) and L(⊥) are about 23 and 28 MPa, respectively.
However, with a cut depth of 0.5 mm, H(⊥) has a strength of only ∼1.0 MPa, while the strength
858 RUBBER CHEMISTRY AND TECHNOLOGY Vol. 71

Fig. 11.—Photographs showing catastrophic rupture paths for:

(a) H(k)-1 and (b) H(k)-2, designated in Figure 10.

of L(⊥) is ∼21 MPa! Perhaps H(⊥) specimens do not strain-crystallize significantly, while L(⊥)
ones do; similar resistances to deformation may be happenstance, resulting from compensating
differences in crystallizability and chain orientation. Clearly, studies of strain crystallizability are

Fig. 12.—Effect of cut length on the strength of parallel double

network specimens (Figures 6, 8 and 10).
 DOUBLE NETWORKS OF A FILLED VULCANIZATE 859

Fig. 13.—Effect of cut length on the strength of perpendicular double network specimens.
(L(⊥) corresponds to D10 and H(⊥) corresponds to D50 in Figure 3).

needed to gain a better understanding of the tear results. Furthermore, it would be interesting
to determine the hysteretic properties of L(⊥) and H(⊥) by stress–strain cycling near their break
points.
Further experimentation is required to optimize double networks for cracking resistance. Of
course, for a given composition (and hence υT ), the principal variables are α0 and the apportioning
of crosslinks between the first and second networks.

CLOSING COMMENTS
The results of this paper show that double networks can have enhanced tear strength relative
to conventional single networks. Besides these uniaxial double networks, it would be interesting
to study biaxial double networks, as well as double networks based on amorphous elastomers.
Additionally, it would be worthwhile to look at the low strain dynamic properties of double
networks: What happens to the Payne effect? Will double networks show reduced hysteresis as
is the case for strained single networks? Roland and Peng16 have carried out studies which may
be relevant to the first question. They found double networks had lower electrical conductivities
than the corresponding single network. This suggests that double networks have reduced filler–
filler contacts, which would be expected to result in a diminished Payne effect.

ACKNOWLEDGEMENT
The authors are grateful to EPIC for funding these studies and thank A. N. Gent for useful
comments.
860 RUBBER CHEMISTRY AND TECHNOLOGY Vol. 71

REFERENCES
1
R. D. Andrews, A. V. Tobolsky, and E. E. Hanson, J. Appl. Phys. 17, 352 (1946).
2
A. Greene and A. Ciferri, Kolloid-Z. 186, 1 (1962).
3
A. Greene, K. J. Smith, Jr., and A. Ciferri, Trans Faraday Soc. 61, 2772 (1965).
4
O. Kramer, R. L. Carpenter, V. Ty, and J. D. Ferry, Macromolecules 7, 79 (1974).
5
P. G. Santangelo and C. M. Roland, Rubber Chem. Technol. 67, 359 (1994).
6
P. G. Santangelo and C. M. Roland, Rubber Chem. Technol. 68, 124 (1995).
7
K. W. Scott and R. S. Stein, J. Chem. Phys. 21, 1281 (1956).
8
J. Scanlan and W. F. Watson, Trans. Faraday Soc. 54, 740 (1958).
9
P. J. Flory, Trans. Faraday Soc. 56, 722 (1960).
10
J. Scanlan, Trans. Faraday Soc. 57, 839 (1961).
11
K. V. Smith, Jr., A. Ciferri, and J. J. Hermans, J. Polymer Sci. Part A-2, Polym. Phys. 2, 1025 (1964).
12
C. M. Roland and M. L. Warzel, Rubber Chem. Technol. 63, 285 (1990).
13
G. R. Hamed, H. J. Kim, and A. N. Gent, Rubber Chem. Technol. 69, 807 (1996).
14
A. G. Thomas and J. M. Whittle, Rubber Chem. Technol. 43, 222 (1970).
15
H. Saechtling, “Saechtling Plastics Handbook,” 2nd Ed., Hanser Publishers, 1987, p. 164.
16
C. M. Roland and K. L. Peng, Rubber Chem. Technol. 64, 790 (1991).

[Received August 25, 1997; revised October 28 1998 ]