Composite Structures 263 (2021) 113693

Contents lists available at ScienceDirect

Composite Structures
journal homepage: www.elsevier.com/locate/compstruct

letter

Mechanical performance of 3D printed interpenetrating phase composites

with spinodal topologies
Yunfei Zhang a, Meng-Ting Hsieh b, Lorenzo Valdevit a,b,⇑
a
Department of Materials Science and Engineering, University of California, Irvine, CA 92697, USA
b
Department of Mechanical and Aerospace Engineering, University of California, Irvine, CA 92697, USA

A R T I C L E I N F O A B S T R A C T

Keywords: The mechanical response of interpenetrating phase composites (IPCs) with stochastic spinodal topologies is
Spinodal topology investigated experimentally and numerically. Model polymeric systems are fabricated by Polyjet multi‐
Interpenetrating phase composite (IPC) material printing, with the reinforcing phase taking the topology of a spinodal shell, and the remaining volume
Mechanical metamaterial filled by a softer matrix. We show that spinodal shell IPCs have comparable compressive strength and stiffness
Triply periodic minimal surface (TPMS)
to IPCs with two well‐established periodic reinforcements, the Schwarz P triply periodic minimal surface
Energy absorption
(TPMS) and the octet truss‐lattice, while exhibiting far less catastrophic failure and greater damage resistance,
particularly at high volume fraction of reinforcing phase. The combination of high stiffness and strength and a
long flat plateau after yielding makes spinodal shell IPCs a promising candidate for energy absorption and
impact protection applications, where the lack of material softening upon large compressive strains can prevent
sudden collapse. Importantly, in contrast with all IPCs with periodic reinforcements, spinodal shell IPCs are
amenable to scalable manufacturing via self‐assembly techniques.

1. Introduction such a way that both phases are independently self‐supporting and
load‐bearing results in improved mechanical properties compared to
Composite materials attain their superior combinations of proper- traditional discontinuously reinforced composite materials, as the
ties from the synergistic contribution of each constituent phase. In topological interconnectivity allows each constituent phase to best
the vast majority of structural composites, only one phase (the rein- contribute to the overall properties of the composite. The synergistic
forcement) is stiff and strong and responsible for load bearing, while role of two phases can be engineered to result in intriguing combina-
the other phase (the matrix) provides load transfer between elements tions of properties. As a few examples, (i) in Al2O3/Al IPCs, the stiff-
of the reinforcement phase and guarantees structural stability of the ness of the composite was shown to be superior than that of Al,
composite. In these traditional designs, only the matrix phase is con- while its toughness and structural integrity were superior than that
tinuous throughout the component, whereas the reinforcement ele- of monolithic Al2O3 [19]; (ii) while traditional aluminum foams
ments (fibers, whiskers, particulates…) are topologically isolated. undergo irrecoverable plastic deformation throughout their stress pla-
While these designs can be very efficient under specific loading condi- teau, aluminum foam–polyurethane IPCs display extensive recoverable
tions (e.g., unidirectional fiber‐reinforced laminae under tensile uniax- deformation under cyclic loadings, thanks to the stabilizing effect of
ial load), their efficiency significantly suffers under more complex the soft elastomeric phase [17]; (iii) metal‐polymer interpenetrating
multi‐axial states, where the lack of a continuous reinforcement phase phase nanocomposites based on nanoporous titanium can be tuned
inevitably introduces soft and weak directions. to match the elastic modulus of human bones, showing great potential
Interpenetrating phase composites (IPCs), composed of two phases for implant applications [18]; (iv) finally, molecular dynamics simula-
that are topologically interconnected throughout the microstructure tions have indicated that silicon carbide–aluminum IPCs display
[1], represent an exciting alternative to traditional designs and have unique toughening mechanisms [10]. All these studies have consis-
drawn a lot of research interest in recent years. Multiple studies, both tently shown that IPCs are promising candidates for structural and
numerical [2–10] and experimental [11–18], have consistently multi‐functional applications. It is worth noting that the vast majority
demonstrated that the combination of reinforcement and matrix in of IPCs in these studies are manufactured through conventional meth-

⇑ Corresponding author at: Department of Materials Science and Engineering, University of California, Irvine, CA 92697, USA.
E-mail address: valdevit@uci.edu (L. Valdevit).

https://doi.org/10.1016/j.compstruct.2021.113693
Received 26 August 2020; Revised 31 December 2020; Accepted 2 February 2021
Available online 12 February 2021
0263-8223/© 2021 Elsevier Ltd. All rights reserved.
Y. Zhang et al. Composite Structures 263 (2021) 113693

ods, such as powder metallurgy [12] and infiltration processes specific stiffness and strength, when compared to both cellular materi-
[13,20]. While efficient and generally scalable, these approaches ran- als with spinodal solid topologies and truss‐based lattice materials
domly distribute controlled amounts of the two phases, and do not [46], and perform on par with cellular materials with triply periodic
allow full topological control of the composite architecture. minimal surface (TPMS) topologies [47,48]. The similarity with TPMS
Recent advancements in additive manufacturing have made it pos- topologies can be attributed to the very tight distribution of curvatures
sible to manufacture IPCs with controlled and complex topologies in spinodal topologies, with the vast majority of surface patches pos-
[21–25]. Notable examples that have been recently extensively inves- sessing near‐zero mean curvature and negative Gaussian curvature,
tigated are IPCs based on triply periodic minimal surfaces (TPMS). the signature feature of TPMSs. Remarkably, the intrinsically stochas-
TPMS are mathematically defined periodic surfaces which have zero tic nature of spinodal shell topologies (and hence the deviation from
mean curvature everywhere on the surface, resulting in locally mini- minimal surface characteristics) not only does not significantly depress
mal surface area in each defined unit cell. Examples of TPMS include mechanical properties, but rather results in considerable imperfection
the Schwarz P surface, the Schwarz D surface and the gyroid surface insensitivity [46]. Whether the unique combination of scalable manu-
[26]. In recent years, the mechanical performance of TPMS‐based facturing and mechanical performance shown by cellular materials
structures has drawn much research interest. For example, cellular with spinodal topologies also translates to interpenetrating phase com-
materials based on TPMS have been shown to be mechanically effi- posites with spinodal topologies remains to be demonstrated.
cient compared to truss‐based cellular materials, by virtue of their In this paper, we investigate the mechanical properties of interpen-
smooth and regular topology, resulting in low local stress concentra- etrating phase composites with spinodal shell reinforcement, with
tions and hence efficient load transfer [27,28]. Recent studies have emphasis on stiffness, yield strength and energy absorption under
shown that similar benefits in mechanical properties extend to non‐linear deformations. Spinodal shell‐based IPCs are compared with
TPMS‐based IPCs, where either the interface between the two solid composites with other reinforcement topologies, namely (1) spinodal
phases is a TPMS or the reinforcement phase is a thickened TPMS shell solids, where both phase topologies are directly obtained by spinodal
structure, embedded in a softer matrix. In particular, excellent combi- decomposition [42,43]; (2) octet lattices, the most widely studied cel-
nations of high stiffness, strength and energy absorption have been lular architecture [49–51]; and (3) Schwarz P surfaces, one of the most
demonstrated [29–34], and TPMS‐based IPCs have been shown to pos- mechanically efficient TPMS topologies [30,52]. While a wide range of
sess better mechanical properties than their truss‐based counterparts truss lattice and TPMS topologies exist, and many have been character-
[31]. In addition, TPMS‐based IPCs have been shown to possess multi- ized mechanically, the octet lattice and the Schwartz P surface
functional extrema [35,36]. However, due to the periodic nature of approach upper bounds of performance for the two classes of periodic
TPMS, all TPMS‐based IPCs are generally difficult to manufacture in reinforcement topologies, respectively, and are thus ideal candidates
a scalable fashion, thus limiting their potential applications. Far better for assessing the mechanical efficiency of spinodal shell reinforce-
scalability can be achieved by incorporating a stochastic reinforcement ments. In order to accurately control all topologies, all IPCs in this
phase that can be self‐assembled through spinodal decomposition of study are produced by multi‐material jetting, a recently developed
two materials followed by material conversion. additive manufacturing technique. While the resulting materials are
Spinodal decomposition is a thermodynamic transformation where polymer/polymer composites with stiffness and strength far below
a homogeneous (solid or liquid) solution separates spontaneously into those of any structural material, the significant difference in mechan-
two coexisting phases [37]. The result is an interpenetrating phase ical properties between the two constituents allows extraction of
composite with very large interfacial area between the phases, and mechanistic understanding that can be readily extend to other classes
hence a characteristic length scale (domain size) much smaller than of composite materials, including ceramic/metal composites.
the sample size; while thermodynamics tends to reduce the interfacial
area between the two phases (hence increasing the characteristic 2. Materials and methods
length scale), this growth can be arrested by reducing the temperature
of the system (in the case of solid phases) [38] or by jamming the 2.1. Numerical generation of the reinforcement topologies
interface with particles that are immiscible in both phases (in the case
of liquid phases) [39,40]. A number of materials conversion tech- 2.1.1. Spinodal solid topology
niques can be subsequently used to (i) eliminate one of the phases The spinodal solid topology is generated numerically with the
and converting the remaining phase to the desired material (hence approach reported in [46] and detailed in Appendix A. A short synop-
producing a cellular material with spinodal solid topology) [41], (ii) sis of the procedure is presented here. A 50% dense spinodal solid cel-
converting both phases to the desired materials (resulting in an IPC lular topology is generated by solving the Cahn–Hilliard evolution
with spinodal solid topology) [42,43], or (iii) eliminating one phase, equation [37], one of the classic evolution models for spinodal decom-
coating the other phase with the desired material and finally eliminate position. The Cahn–Hilliard equation can be written as:
the second phase as well (resulting in a cellular material with spinodal
@u df ðuÞ
shell topology) [40]. As a notable example, in a recent study bicontin- ¼ Δ½  θ2 Δu ð1Þ
@t du
uous interfacially jammed emulsion gels (bijels) are formed and pro-
cessed into sacrificial porous nickel scaffolds for chemical vapor where u(x, y, z, t) is the concentration of the material and void
deposition to produce freestanding three‐dimensional turbostratic gra- phases (1 ≤ u ≤ 1, with u = −1 indicating solid material and
phene (bi‐3DG) monoliths with spinodal shell topologies, possessing u = 1 indicating void space) at a coordinate (x,y,z,), t is the evolution
2
exceptionally high specific surface area and exceeding 100,000 unit time, f ðuÞ ¼ 14 ðu2  1Þ is a double‐well free energy function, θ is the
cells [44]. In all cases, the inherent self‐assembly of spinodal topolo- width of the interface between the two phases and Δ is the Laplacian
gies provides a route to fabricate micro‐ or nano‐architected materials operator. The equation is solved in space and time over a cubic domain
with macroscopic dimensions, with a level of scalability unmatched by with edge length of N, via a finite difference algorithm. The evolution
any additive manufacturing technique [44,45]. As spinodal topologies time controls the characteristic feature size (λ), a measure of the
are bicontinuous, such micro/nano‐architected materials could be domain size of the topology. As time progresses, the features coarsen
infiltrated with a second phase via deposition or infiltration processes, to reduce interfacial energy, thus increasing λ. We choose to extract
potentially providing a uniquely scalable fabrication process for shell‐ our topologies at the evolution time corresponding to λ = 15N (loosely
based IPCs. corresponding to a sample with 5  5  5 unit cells), as this number of
Previous studies have shown that cellular materials with spinodal unit cells was shown in previous studies to provide a homogenized
shell topologies are exceptionally efficient from the perspective of

2
Y. Zhang et al. Composite Structures 263 (2021) 113693

response [46]. To produce spinodal solid topologies with volume frac-

tions different from 50%, a thresholding technique is used. See Appen-
dix A for details.

2.1.2. Spinodal shell topology

The spinodal shell topology is extracted from the interface between
the solid and void phases, produced as explained in Section 2.1.1. The
volume fraction of reinforcement, V f , is chosen by assigning the appro-
priate thickness to the shell, t s , so that V f ¼ t s A=V, with A the surface
area of the shell and V the sample volume. The procedure is schemat-
ically depicted in Fig. 1 and described in more detail in Appendix A.
Spinodal shell topologies extracted from a 50% dense spinodal solid
topology have negative Gaussian curvature throughout the shell, with
mean curvature close to zero everywhere – thus approaching the geo-
metrical characteristics of TPMSs. While spinodal shell topologies with
non‐zero mean curvature can be generated by extracting the surface of
solid spinodal topologies with densities different from 50%, previous
studies have demonstrated that their mechanical properties are infe-
rior [46]. Hence in this work we will limit our attention to spinodal
shell topologies with near‐zero mean curvature throughout.

2.1.3. Schwarz P TPMS topology

The Schwarz P surface is created using the level set approximation
technique from the implicit surface:
cos ðxÞ þ cos ðy Þ þ cos ðzÞ ¼ 0 ð2Þ
where (x,y,z) is a coordinate point in three‐dimensional Euclidean
Fig. 2. IPC topologies under consideration: (a) Reinforcement phase, (b)
space. The minimal surface corresponding to the above equation is
matrix phase, (c) assembled interpenetrating phase composite and (d) 3D
generated using Minisurf, a software used to generate minimal surface
printed composite sample for spinodal shell, spinodal solid, octet lattice and
CAD files [53,54]. Cubic samples with 3  3  3 unit cells are gener- Schwarz P shell IPCs.
ated. As explained in Section 2.1.2, the volume fraction of reinforce-
ment, V f , is chosen by assigning the appropriate thickness to the
shell, t s , so that V f ¼ t s A=V, with A the surface area of the shell and 2.3. Fabrication approach
V the sample volume.
All samples are manufactured with a PolyJet 3D printer (Objet260
2.1.4. Octet lattice topology Connex3, Stratasys), which allows 3 different materials to be printed
The octet lattice is built using the commercially available CAD simultaneously, with intermediate (digital) formulations obtained by
modeling software package SolidWorks. Cubic samples with mixing appropriate amounts of the 3 constituent materials at any voxel
3  3  3 unit cells are generated. The volume fraction of reinforce- location. The reinforcement phase in IPCs is printed with
ment, V f , is chosen by assigning the appropriate diameter to the bars, VeroWhitePlus, a glassy photopolymer, while the soft matrix phase
pffiffiffi 2
d, according to V f ¼ 6 2πðdlÞ with l being the length of the strut [49]. was printed with Agilus30, a rubbery polymer. All samples are
30 × 30 × 30 mm cubes. Selected reinforcement‐only cellular mate-
2.2. Numerical generation of the IPC architectures rials are printed by replacing Agilus30 with support material, which is
subsequently removed by water jetting. All 3D printed samples are left
After the reinforcement phase topologies are generated, they are at room temperature for 4 days for curing.
imported into Geomagic Design X, a reverse engineering software that
can also be used to handle CAD files. After the appropriate shell thick- 2.4. Mechanical characterization
nesses and bar diameters are generated to result in the desired volume
fraction of reinforcement, each CAD model is subtracted from a cube of All mechanical tests are performed with an Instron 8800 mechani-
the same size with a Boolean cut operation to create the complemen- cal test frame equipped with a 100kN load cell. The constituent mate-
tary matrix phase. Finally, the CAD files are converted into STL files rials are characterized under both uniaxial tensile and compression
and transferred to the GrabCAD software, the control software for loading, using dog bone and cylindrical samples, respectively. Engi-
the 3D printer. This process, along with the printed samples, is illus- neering stress and strain are extracted as σ 0 ¼ F=A0 and ɛ 0 ¼ δ=l0 ,
trated in Fig. 2. with F and δ the raw force and displacement measurements, and A0
and l0 the original cross‐section and length of the sample, respectively.
True stress and true strains are obtained as σ t ¼ σ 0 ð1 þ ɛ 0 Þ and
ɛt ¼ lnð1 þ ɛ 0 Þ, respectively. The Young’s modulus is extracted as the
slope of the initial linear region.
All IPC composites and cellular materials are tested in compression
only. ASTM 695–15 for compressive properties of rigid polymer is fol-
lowed in all compression tests [55]. Engineering stress and strain are
extracted as explained above. To minimize the influence of the aniso-
tropic nature of 3D printing introduced by the layer‐by‐layer printing
Fig. 1. Schematic of approach for numerically generating the spinodal shell process, all compression tests are conducted in the direction parallel to
reinforcement phase for IPCs. the printing direction, which is known to be the strongest. A quasi‐

3
Y. Zhang et al. Composite Structures 263 (2021) 113693

static strain rate of 0.001 s−1 is used in all tests. All samples, except for provided in Fig. 3. Notice that the reinforcement phase material is
those subjected to cyclic tests, are compressed to 50% strain to mea- approximately three orders of magnitude stiffer than the matrix phase
sure Young’s modulus, 0.2% offset yield strength, and energy absorp- material; consequently, it is expected that the reinforcement phase will
tion. The latter is extracted as the area under the stress–strain curve up take the vast majority of the load in the composite samples. The matrix
to 50% strain. The cyclic test samples are compressed to 10% strain for phase, which is three orders of magnitude softer than the reinforce-
three cycles using the same quasi‐static strain rate as all in other tests. ment phase, still plays an important role on deformation mechanisms
and damage evolution. See Appendix C for details. It is worth noting
2.5. Finite elements modeling that both constituent materials show some tension–compression asym-
metry, with the reinforcement phase material about 20% stiffer and
Finite element meshes of spinodal shell and Schwartz P shell com- 32% stronger in compression than in tension. In previous studies, Ver-
posite samples generated through SimpleWare ScanIP are imported in oWhitePlus has been shown to exhibit significant size effects on
the commercial Finite Element package Abaqus, while the octet lattice strength [56]. While we observe similar size effects on the tensile
composites are meshed directly within Abaqus. Explicit quasi‐static properties of VeroWhitePlus dog bone specimen and the compressive
analyses are performed to extract stress–strain curves under uniaxial properties of VeroWhitePlus cellular samples with spinodal topology
compression, subject to the following boundary conditions: (i) each (Appendix B), we do not observe any significant size effect in our
composite sample is sandwiched between two frictionless rigid plates; experimental results on fully dense IPC samples. We attribute this lack
(ii) the top plate moves down and compresses the composite up 25% of size effects to the fact that all samples experience similar degree of
strain (to ensure quasi‐static response, the speed of the top plate is curing regardless of feature size, as the matrix and the reinforcement
adjusted to ensure that the overall kinetic energy is less 5% of the phases are printed at the same time in fully dense samples, and all sam-
internal energy throughout the simulation); (iii) all the side faces of ples are of the same size. The conclusion is that size effects on materi-
the composite sample are left unconstrained. als properties can be ignored in the following analysis of the
The reinforcement phase material, VeroWhitePlus, is modeled as an mechanical performance of IPC composites. For more details on size
elastoplastic material, with parameters fitted on the experimental effect, refer to Appendix B.
results in Fig. 3a. A Young’s modulus, E = 1234 MPa, is chosen as
the average of measured moduli in tension and in compression. The 3.2. Spinodal IPCs: The difference between solid and shell reinforcement
plastic model is chosen to represent the significant tension/compres- topologies
sion asymmetry (Fig. 3a): in compression, the response is perfectly
plastic, with a yield strength σ y = 72 MPa, taken as the average of We start by comparing the mechanical response of IPCs with spin-
the plastic flow stress over the plastic strain range of interest; in ten- odal shell and spinodal solid reinforcement topologies, for volume
sion, an initial yield strength of 53 MPa is used, followed by isotropic fractions of reinforcement between 20 and 50%. Compressive stress–-
hardening, σ y ¼ 53 þ 90ɛ p 0:58 MPa, with a maximum strength of strain curves are obtained as explained in Section 2.4, and depicted in
62 MPa. Finally, a damage model is used, with a maximum plastic Fig. 4a,b. While the response at low relative density is qualitatively
strain damage initiation, ɛ po ¼ 0:017, and a damage evolution based similar, IPCs with spinodal shell topologies display a much more grad-
on linear material softening (dp ¼ L  fɛ pf  ɛpo } = 0.52, where dp is ual failure as the relative density is increased, with much reduced load
the plastic displacement, Lis the elemental characteristic length, and drops and softening at large strains. Young’s modulus (E), yield
ɛ pf is the plastic strain at failure). strength (σ y ) and energy absorption (U) are presented as a function
Since tension/compression asymmetry is small up to ~ 25% strain, of the volume fraction of reinforcement in Fig. 4c,d. For IPCs with
the matrix material, Agilus30, is modeled as a hyperelastic material, spinodal shell reinforcement, we extract the power laws E ∼ V 1:4 f ,
with a Marlow strain energy potential fitted to the uniaxial compres- σ y ∼ V 1:6
f and U ∼ V 1:1
f , whereas for IPCs with spinodal solid rein-
sion engineering stress–strain curve in Fig. 3b. forcement we findE ∼ V 2:2 f , σy ∼ V f and U ∼ V 1:8
2:7
f . As the hard rein-
forcement material is much stronger and stiffer than the rubbery
3. Results and discussion matrix material (with more than three orders of magnitude difference
in stiffness, see Section 3.1), it is expected that the reinforcing phase
3.1. Mechanical properties of constituent materials will dominate the mechanical response of the composite; therefore,
the scaling laws obtained here can be compared to those of cellular
Constituent materials are first tested in tension and compression as materials, with the volume fraction of reinforcement representing
described in Section 2.4. Stress–strain curves for the two materials are 
the relative density, ρ. For truss‐based lattice materials, a scaling

Fig. 3. Tensile and compressive true stress – true strain curves for (a) VeroWhitePlus and (b) Agilus30.

4
Y. Zhang et al. Composite Structures 263 (2021) 113693

Fig. 4. Comparison between the mechanical response of IPCs with spinodal shell reinforcement and spinodal solid reinforcement. (a, b) Compressive stress–strain
curves. The volume fraction of reinforcement ranges from 20% to 50%. (a) IPCs with spinodal shell reinforcement. (b) IPCs with spinodal solid reinforcement. (c,
d) Mechanical properties as a function of volume fraction of reinforcement phase: (c) Young’s modulus and yield strength; (d) Energy absorption.

 1  1
E∼ ρ and σ y ∼ ρ denote mechanically efficient stretching‐ tion (U) are presented as a function of V f in Fig. 5d–f. While repeat
  tests at any single density were not conducted, the clear power‐law
dominated behavior, with E ∼ ρ 2 and σ y ∼ ρ 1:5 indicating less effi-
cient bending‐dominated behavior [57,58]. For spinodal cellular mate- behavior for all properties over the entire density range, with all data
   points narrowly banded around the average trend, confirms repeata-
rials, E ∼ ρ 1:2 and σ y ∼ ρ 1:2 for shell topologies and E ∼ ρ 2:3 and
 bility of the results (e.g., in the case of strength, all data fit within a
σ y ∼ ρ 2 for solid topologies [46], in good agreement with the scalings
98% confidence interval). Two key results clearly emerge: (i) The
found herein for composites, confirming that the reinforcement mate-
mechanical properties of IPCs with these three very different reinforce-
rial dominates the mechanical response. The implication is that IPCs
ment topologies (periodic truss and shell and stochastic shell) are
with spinodal shell reinforcement topologies are consistently superior
nearly identical over the entire range of V f , with the spinodal shell
in all metrics, with the advantage increasing significantly at lower vol-
IPC performing slightly worse than the others in stiffness and strength
ume fractions of reinforcement, where the thin shells with nearly zero
(we attribute this to manufacturing defects, as the reinforcement shell
mean curvature and negative Gaussian curvatures behave in a predom-
of spinodal IPCs is thinner than that of the other two geometries at
inantly stretching dominated manner and deform uniformly, with little
most volume fractions and is close to the resolution of the 3D printer
stress intensification. As V f is increased, the topological difference
at 5% volume fraction), and slightly better in energy absorption at
between solid and shell spinodal reinforcement blurs, and the proper-
high V f ; (ii) At V f > 35%, while the plastic and failure response of
ties of the two IPCs converge. Given the consistent superior mechani-
octet lattice and Schwartz P shell‐based IPCs is qualitatively similar,
cal response of IPCs with spinodal shell reinforcement topology, IPCs
and characterized by a sharp stress drop immediately after the ulti-
with spinodal solid reinforcement topologies are not investigated
mate strength, the stochastic spinodal shell‐based IPC displays a much
further.
more gradual failure mechanism, characterized by a nearly flat stress
plateau over the entire strain range.
3.3. The mechanical advantage of spinodal shell IPC compared to IPCs with
The first result clearly reveals that the bicontinuous nature of the
regular reinforcement topologies
phases is more important than the specific reinforcement topology in
determining stiffness, strength and failure initiation of the IPC compos-
It is instructive to compare the mechanical response of IPCs with
ite. As many applications require materials to remain within the elastic
spinodal shell reinforcement topology, which is intrinsically stochas-
regime, the implication is that the design space for stiff and strong IPCs
tic, with that of IPCs with periodic reinforcement topologies, in partic-
is very broad: a wide range of topologies will result in very similar
ular the octet lattice topology and the Schwartz P shell topology
mechanical behavior. Conversely, the post‐yielding deformation and
(Fig. 2). Compressive stress–strain curves obtained over a wide range
failure behavior of the IPC composite (particularly at high volume frac-
of volume fractions of reinforcement (V f ¼ 5  50%) are presented in
tion of reinforcement) is strongly affected by the topological arrange-
Fig. 5a–c. Young’s modulus (E), yield strength (σ y ) and energy absorp-
ment of the reinforcement and matrix phases, suggesting that topology

5
Y. Zhang et al. Composite Structures 263 (2021) 113693

Fig. 5. Mechanical response of IPC composites. (a–c) Stress–strain curves of IPCs with different reinforcement topologies at various volume fractions compressed
to 50% strain. Notice that spinodal shell IPCs don’t exhibit catastrophic load drops at high volume fractions of reinforcement. (d–f) Comparison of mechanical
properties of IPCs at different reinforcement volume fractions: (d) Young’s modulus; (e) Yield strength; (f) Energy absorption. Mechanical properties of spinodal
IPC at 5% were excluded from the scaling as the thickness of the reinforcement shell is close to the resolution of the 3D printer and is susceptible to manufacturing
defects.

optimization would play a substantial role in design of IPCs for energy odic topologies. This substantial difference in the failure mechanism
absorption and impact protection. results in a nearly flat stress–strain curve up to ɛ ∼ 0:3, followed by
To better understand this deformation and failure behavior, we a very gentle stress drop up to ɛ ∼ 0:4, when the test was interrupted.
compare both the stress–strain curve and deformation response of To further investigate the damage resistance of the spinodal shell
the IPCs with the three reinforcement topologies, at V f ¼ 50% IPC, loading–unloading compression experiments are performed,
(Fig. 6). Notice that the two types of IPCs with periodic reinforcements whereby IPCs with the three reinforcement topologies, at V f ¼ 50,
experience more catastrophic failure events than the spinodal shell are compressed to 10% strain for three cycles (Fig. 7). As the number
IPC. Soon after the ultimate strength (at ɛ ∼ 0:1), cracks in the rein- of cycles increases, all IPCs show increasingly visible cracks in the rein-
forcement are clearly visible in the Schwartz P shell IPC; at a strain forcement phase. The Schwartz P shell IPC and the octet lattice IPC
as low as ɛ ∼ 0:2, these cracks have multiplied and aligned along a both experience decreasing load‐bearing capacity as a result. On the
shear band, inducing catastrophic failure soon after. While cracks contrary, even with increasing visible fracture sites, the load‐bearing
are not visible at the surface of the octet lattice IPC at strains as low capacity of the spinodal shell IPC is largely unaffected by the cyclic
as ɛ ∼ 0:1, presumably those cracks exist at the interior of the sample; loading, demonstrating superior resistance to damage.
at ɛ ∼ 0:2, alignment of cracks along a shear band is clearly visible, We attribute these dramatic differences in deformation and failure
inducing the same catastrophic failure mechanism as in the Schwartz behavior to (i) the larger surface area and (ii) the stochastic nature of
P shell IPC. In fact, the stress–strain curves of these two IPCs are essen- the spinodal shell topology. The larger surface area of the spinodal
tially identical, throughout the entire strain range. By contrast, the shell compared to the other two periodic geometries results in
stochastic spinodal shell IPC does not exhibit any visible reinforcement increased reinforcement/matrix interaction, with the matrix prevent-
cracking until ɛ ∼ 0:3; even then, the cracks appear stochastically dis- ing/arresting crack propagation in the reinforcement phase. For a
tributed across the microstructure, and not banded as for the two peri- more detailed discussion on the influence of surface area, refer to

6
Y. Zhang et al. Composite Structures 263 (2021) 113693

Fig. 6. Deformation response of IPCs with different reinforcement topologies

at 50% volume fractions, compressed to 50% strain. (a) Stress–strain curve.
(b) Spinodal shell composites. (c) Schwartz P shell composites. (d) Octet
lattice composites. Notice that octet and Schwartz P IPCs at high volume
fraction of reinforcement experience more localized catastrophic failures than
spinodal shell IPCs.

Appendix D. In addition, the stochastic nature and complex shape of Fig. 7. Cyclic compression experiments on IPCs with different reinforcement
the spinodal shell reinforcement distribute the loads more efficiently topologies, at 50% volume fraction of reinforcement. All samples have been
throughout the topology, maintaining load‐bearing capacity after frac- compressed to 10% strain for 3 cycles. (a) Octet lattice composite. (b)
ture initiation and preventing the formation of catastrophic failure‐ Schwartz P shell composite. (c) Spinodal shell composite. All three topologies
inducing crack bands. In periodic geometries like Schwartz P or octet show increasingly visible fractures in the reinforcement phase as cycling
loading progresses. Both periodic IPCs show decreasing load bearing capacity
truss, failures first occur around stress concentration locations, and
as a result, while the stress–strain curve of the spinodal shell composite is
subsequently band along specific directions, leading to catastrophic
largely unaffected.
failure and loss of load bearing capacity. Conversely, for the stochastic
spinodal shell topology, stress is more uniformly distributed across the
entire structure. Even after fracture of some shell sections, the complex tational model and provide confidence in its ability to capture the
topology still allows the spinodal shell‐reinforced composite to main- onset and early evolution of damage. As finite elements simulations
tain nearly unchanged load bearing capacity. involving post‐failure behavior (including fracture) are very challeng-
The difference in failure response between the spinodal and the ing because of stress singularity and loss of uniqueness, a full numer-
two types of periodic reinforcement IPCs is less pronounced at low vol- ical description of the failure mechanisms all the way to
ume fractions of the reinforcement (Fig. 5), as the increased volume densification is beyond the scope of this work. Nonetheless, several
fraction of the matrix helps stabilize even periodic structures against conclusions can be extracted from an analysis of the stress state at
crack banding of the reinforcement. the early and intermediate phases of deformation. First we compare
To better understand and quantify the differences in deformation the von Mises stress distribution in the reinforcement phase for all
and failure behavior, finite element analyses are performed on the IPCs three IPCs at a strain of 0.1, roughly coinciding with the attainment
with the three reinforcement topologies, at a volume fraction of rein- of the maximum strength, in Fig. 8d–f. It is apparent that the spinodal
forcement V f ¼ 30%. The results are presented in Fig. 8. Notice that shell IPC shows small and uniformly distributed stress concentrations
the computational prediction of the stress–strain response is in good throughout the sample, whereas the two IPCs with the periodic rein-
agreement with the experimental results, well capturing the initial forcement topologies exhibit very large and interconnected stress con-
stiffness, the yield and ultimate strength and the beginning of the centrations. This is consistent with the experimental findings in Fig. 6
post‐failure behavior (Fig. 8a–c). This agreement validates the compu- and further supports the argument that the spinodal shell IPC is effec-

7
Y. Zhang et al. Composite Structures 263 (2021) 113693

Fig. 8. Comparison between experiment and simulation results. (a, b, c) Stress strain curve comparison between experiment and simulations, for IPCs at 30%
volume fraction of reinforcement: (a) spinodal shell IPC; (b) Schwartz P shell IPC; (c) octet lattice IPC. (d–f) von Mises stress map of the reinforcement phase,
extracted from the composite at 10% strain: (d) spinodal shell reinforcement; (e) Schwartz P shell reinforcement; (f) octet lattice reinforcement. (g–i) von Mises
stress map of the reinforcement phase at the half-way cross section, extracted from the composite at 15% strain: (g) spinodal shell reinforcement; (h) Schwartz P
shell reinforcement; (i) octet lattice reinforcement. Cracks are highlighted in black.

tive at avoiding catastrophic failure by preventing formation of rein- strains. Combined with a very large interfacial area, which ensures
forcement cracking bands. On the contrary, in IPCs with periodic rein- intimate reinforcement/matrix interaction, this feature makes spin-
forcement such as the octet and Schwartz P IPC studied here, fracture odal shell IPCs ideally suited for energy absorption and impact
of a member leads to a cascading crack propagation which results in a applications.
near‐instant loss of load bearing capacity. We then compare the von
Mises stress distribution in the reinforcement phase in the middle of 4. Conclusions
the sample for all three IPCs at a strain of 0.15, when surface cracks
start to appear in Fig. 8g–I. It can be seen that the stress distribution In summary, we have fabricated and mechanically investigated a
of the spinodal shell IPC is still uniform and largely unaffected by new type of interpenetrating phase composite (IPC) with spinodal shell
cracks, while the two periodic IPCs experience drop in load bearing reinforcement topology, comparing their mechanical response to that
capacity as a result of cracks at stress concentration points. These sim- of well‐established mechanically efficient periodic IPCs with octet lat-
ulation results are also in good agreement with the cyclic experiments tice and Schwartz P shell reinforcement topologies. Polymeric model
shown in Fig. 7 and demonstrate the excellent damage resistance of systems were produced by Polyjet multi‐material additive manufactur-
the spinodal shell IPC. Movies of the evolution of the von Mises stress ing; while the mechanical properties of the two polymeric phases (and
distribuaption in the reinforcement phase for all three IPCs generated hence the resulting composites) were inferior to those of any practical
in the simulation can be found in Appendix E: Supplementary Data. structural material, this technique allowed unbiased comparison of dif-
Collectively, these experimental and numerical results clearly illus- ferent topologies. We have shown that while all three types of IPCs
trate that the stochastic nature of spinodal shell topologies enables the perform nearly identically in terms of initial stiffness, yield strength
establishment of a uniform stress field throughout the reinforcement, and energy absorption over a wide range of volume fraction of rein-
which persists even after the onset of reinforcement cracking. The lack forcement (5–50%), spinodal shell IPCs are far more robust than any
of substantial stress intensifications promotes a stochastic distribution other IPCs, exhibiting greater damage resistance as well as a much
of initial cracking locations and prevents the catastrophic occurrence more uniform deformation and gradual failure. This unique feature
of crack banding, resulting in a flat stress plateau through very large is attributed to: (1) uniform distribution of stresses and strains in shell

8
Y. Zhang et al. Composite Structures 263 (2021) 113693

topologies, stemming from fairly uniform distribution of negative UC Irvine and SIMULIA. All fabrication was carried out at the Institute
Gaussian curvature across the entire surface; (2) larger surface area for Design and Manufacturing Innovation at UC Irvine.
at a given volume fraction of reinforcement, resulting in increased
matrix support on the load bearing reinforcing phase; (3) the stochas- Appendix A. Generation of spinodal topologies
tic nature and complex shape of spinodal topologies, which not only
act as crack barriers and locally inhibit crack propagation and banding, The generation of spinodal topologies follows an approach dis-
but also continue to provide load‐bearing capacity even after fractures cussed in detail in [42,51], and summarized here for completeness.
of some members occurs. The combination of excellent mechanical Spinodal decomposition can be described by the Cahn‐Hillard evolu-
efficiency, on par with those of the best IPCs with periodic reinforce- tion equation [37]:
ment, great damage resistance, gradual deformation and failure mech-
@u df ðuÞ
anisms and potential for scalable manufacturing makes spinodal shell ¼ Δ½  θ2 Δu ðA1Þ
@t du
IPCs exceptional candidates for damage tolerance and energy absorp-
tion applications where a prolonged compressive stress plateau after where u(x,y,z,t) is the concentration of the two phases A and B at a
maximum stress is desired. While quantitative assessment of impact coordinate (x,y,z) (1 ≤ u ≤ 1, with u = −1 indicating only phase A,
performance requires high‐strain rate testing and are beyond the scope or void space, and u = 1 indicating only phase B, or solid material),
2
of this work, previous studies have demonstrated that impact perfor- t is the decomposition time, f ðuÞ ¼ 14 ðu2  1Þ is a double‐well free
mance ranking of cellular materials can be obtained by the quasi‐ energy function, θ is the width of the interface between the two phases
static experiments performed in this work [59,60]. This strongly sug- and Δ is the Laplacian operator. Equation (A1) is solved numerically
gests that spinodal shell IPCs would be excellent performers for impact with a finite difference scheme over a cubic volume with edge length
protection. N = 100, which is discretized into a lattice of mesh size,
Most importantly, unlike periodic IPCs, spinodal shell IPCs can in ℓ=N/100 = 1. Let um ijk denote the discrete value of the phase field
principle be scalably manufactured at various length scales, using variable u(i,j,k,m τ) at nodal point (i,j,k), with τ the integration time
self‐assembly approaches followed by material conversion techniques. step, chosen to be sufficiently small to achieve convergence
Possible self‐assembly approaches include spinodal decomposition of (τ = 0.005 was used here), and m the time step. After discretization
block copolymers [45], interfacially jammed colloidal suspensions (bi- with a finite difference scheme, equation (A1) can be written as:
jels) [40,44] and selective etching of bimetallic alloys [41]. These
approaches allow ready fabrication of polymeric, metallic or ceramic
mþ1
uijk  um
ijk 3
¼ Δ½ðum
ijk Þ  uijk  θ Δuijk 
m 2 m
ðA2Þ
macro‐scale samples with domain sizes at the micro or nano‐scale, τ
resulting in architected materials with enormous surface area and fur- where θ is the thickness of the interface between the two phases
ther improving mechanical properties by virtue of well‐established and Δ is the Laplacian operator. The following boundary conditions
size effects on the constituent materials [61–63]. As spinodal topolo- are applied to solve equation (A2):
gies are bicontinuous, such micro/nano‐architected materials could
be infiltrated with a second phase via deposition or infiltration pro- uði; j; k; mτÞ ¼ uði þ L; j; k; mτÞ 3:1Þ
cesses, potentially providing a uniquely scalable fabrication process
uði; j; k; mτÞ ¼ uði; j þ L; k; mτÞ 3:2Þ
for shell‐based IPCs.
Finally, we emphasize that, while the multi‐material additive man-
uði; j; k; mτÞ ¼ uði; j; k þ L; mτÞ 3:3Þ
ufacturing approach used in this study resulted in polymer–polymer
composites with absolute mechanical properties far inferior to those A randomly generated initial condition, uði; j; k; 0Þ ¼ u0 ði; j; k; 0Þ ∈
of any structural material, the self‐assembly‐based processes envi- ½5; 5  104 –0, is used as a perturbation to exit unstable equilibium
sioned above could be used to produce ceramic‐polymer, cera- and start the decomposition kinetics. As the solution progresses, the
mic–metal, metal‐polymer and metal–metal composites that can in system phase separates at early times, and subsequently continues to
principle outperform most existing structural materials. Demonstra- coarsen; during the coarsening phase, the curvature of the interface
tion and characterization of such advanced spinodal shell‐based IPCs between solid and void decreases and the size of the single‐phase
will be the subject of future studies. domains increases. A cutoff um c is defined to separate phase A from
phase B, with the phase at a point (i,j,k) and a time t ¼ mτ defined as:
Data availability
ijk ¼ Hðuijk  uc Þ
Gm ðA4Þ
m m

The raw/processed data required to reproduce these findings can- where HðÞ represents the Heaviside function. To achieve a 50%
not be shared at this time as the data also forms part of an ongoing volume fraction (V) of phase A, the cutoff um m
c is adjusted so that uijk sat-
study. isfied the distribution given by:
1 N N N m 1 N N N  
Gmc ¼ Σ Σ Σ G ¼ Σ Σ Σ H um ijk  uc ¼V ðA5Þ
m
Declaration of Competing Interest
N 3 i¼1 j¼1 k¼1 ijk N 3 i¼1 j¼1 k¼1
The authors declare that they have no known competing financial Here the decomposition time that controls the characteristic fea-
interests or personal relationships that could have appeared to influ- ture size (λ) is set to provide λ = 1/5N, a geometrical condition that
ence the work reported in this paper. has been shown to offer the best overall mechanical performance for
spinodal shell‐based cellular materials [46].
Acknowledgments The spinodal shell topologies are subsequently derived by extract-
ing the surface from spinodal solid cellular topologies with volume
This work was supported by the Office of Naval Research (program fraction, V = 50%, resulting in a shell topology with negative Gaus-
Manager: D. Shifler, Grant No. N00014‐17‐1‐2874). The ABAQUS sian curvature and near‐zero mean curvature throughout the domain.
Finite Element Analysis software is licensed from Dassault Systemes While shell‐topologies with non‐zero mean curvature can be derived
SIMULIA, as part of a Strategic Academic Customer Program between from spinodal solid topologies with V–50%, previous studies have

9
Y. Zhang et al. Composite Structures 263 (2021) 113693

Fig. B2. Feature size of the reinforcing phase as a function of the volume
fraction of reinforcement, for different topologies. The yellow horizontal line
represents the threshold below which size effects are expected.

Appendix B. Size effect on the mechanical properties of 3D

printed consituent materials

The mechanical properties of the photopolymers used in this study,

namely the glassy VeroWhitePlus and the rubbery Agilus30, have not
been as thoroughly investigated as other commonly used structural
materials for additive manufacturing. As several materials for additive
manufacturing are known to exhibit size‐dependent mechanical prop-
erties, we investigate the tensile properties of dog bone samples
printed with VeroWhitePlus (the reinforcement material in our study),
as a function of sample size. The results are reported in Fig. B1a, and
show a clear size effect for sample dimensions lower than ~ 1 mm, with
smaller samples being less stiff and strong, but far more ductile. The
results are consistent with a recent study, which has shown a well‐
defined size threshold for the strength of VeroWhitePlus: above this
threshold, no size effect is observed, whereas below this threshold, a
smaller‐is‐weaker effect is present [56]. This size threshold is esti-
mated to be approximately 50 times the printing layer thickness,
which is in good agreement with the threshold identified in our dog
bone experiments. The size threshold is shown as a horizontal line in
Fig. B2, overlaid with the reinforcement feature sizes of IPC samples
tested in this study. According to Fig. B2, we would expect to see no
size effect for most IPCs examined in this study, with the exception
of IPCs with spinodal shell topologies for reinforcement volume frac-
tions smaller than 30%. To verify this assumption, we proceed to test
in compression two 30% dense spinodal shell cellular materials, with
identical topology but different shell thicknesses, t 1 ¼ 0:75mm and
t 2 ¼ 1:5mm. The latter is printed with two times the overall dimension
of the former, so to have the same topology and volume fraction but
different shell thickness. As shown in Fig. B1b, a significant size effect
exists, quantitatively consistent with the dog bone results in Fig. B1a
and the threshold depicted in Fig. B2. Finally, we test in compression
two spinodal shell IPCs, identical to those in Fig. B1b but with the void
space occupied by the Agilus30 matrix. The results, reported in
Fig. B1. Size effects. (a) Stress–strain curves of tensile samples with different Fig. B1c, remarkably show effectively no size effect on stiffness or
thickness; (b) Compressive stress–strain curves of spinodal shell cellular
strength, albeit the effect on ductility is pronounced. This lack of size
(reinforcement only) sample at 30% volume fraction of reinforcement, for
effect is also visible in Fig. 5 in the main manuscript, where no change
different shell thicknesses; (c) Compressive stress–strain curves of spinodal
shell IPC at 30% volume fraction of reinforcement, for different shell
of slope in the strength VS volume fraction curves is observed for spin-
thicknesses. odal shell composites, in disagreement with the predictions of Fig. B2.
This apparent disagreement can be explained as follows. A recent
demonstrated that near‐zero mean curvature topologies provide the study on size effect in 3D printing polymers has shown that the size
best overall mechanical performance for spinodal shell‐based cellular effect observed on thinner samples emanate from the layer‐by‐layer
materials [46]. printing process [64]. As the sample thickness increases, previously

10
Y. Zhang et al. Composite Structures 263 (2021) 113693

Fig. C1. Deformation mechanisms comparison between composite and cellular (reinforcement only) samples with 30% volume fraction at different strain levels:
(a) Octet lattice composite, (b) Octet lattice reinforcement, (c) Schwartz P shell composite, (d) Schwartz P shell reinforcement, (e) Spinodal shell composite, (f)
Spinodal shell reinforcement. For all reinforcement topologies, the composite shows more uniform and less catastrophic deformation, while the reinforcement only
sample shows more localized deformation and failures. Addition of matrix also changed the failure mode of spinodal shell composite from shell bending or
buckling of the reinforcement only sample to failure from shell fracture because of the incompressibility and support of the matrix.

11
Y. Zhang et al. Composite Structures 263 (2021) 113693

ing failure. Interestingly, while the spinodal shell reinforcement‐only

sample deforms in a more uniform manner compared to the other
two periodic geometries, the addition of the matrix still helps stabilize
the reinforcement shell, and changes the failure mode of the spinodal
shell IPC from shell bending and buckling (as observed in the
reinforcement‐only sample, Fig. C1e) to shell fracture (Fig. C1f). This
matrix contribution is attributed to the large surface area of the spin-
odal shell topology and the incompressibility of the matrix: while lar-
ger surface area leads to increased support from the matrix,
incompressibility promotes membrane deformation of the shell mem-
bers, while preventing bending and buckling. The overarching conclu-
sion is that the matrix, which is three orders of magnitude softer than
the reinforcement, barely contributes to initial stiffness and strength of
the composite, but has a pronounced effect on deformation mecha-
nisms and damage evolution.

Appendix D. Influence of surface area of the reinforcement phase

in IPCs

Fig. D1. Surface area of the reinforcing phase for IPC samples with different One possible reason for the unique properties of spinodal shell IPCs
reinforcement topologies. is the increased surface area of the spinodal shell relative to the other
reinforcement topologies, which leads to increased interaction and
support from the matrix. As discussed in Appendix A, the surface area
printed layers are exposed longer to UV light, which leads to higher and curvature of spinodal geometry is affected by decomposition time
curing degree and better mechanical properties. In agreement with t, with increasing t resulting in deceased surface area and curvature.
this study, we believe that the absence of size effect in our IPC compos- The decomposition time t can be controlled to generate different ratio
ites stems from the fact that the reinforcing phase and the matrix phase between the characteristic length over the cubic length, λ=L (Fig. D2).
are printed on the same layer in the same scan over the entire sample, In all spinodal samples tested in the manuscript, λ=L ¼ 1=5 is used as it
so that printing composites is qualitatively similar to printing solid has been shown to provide the best overall mechanical properties to
blocks. As all composite samples have similar size and therefore simi- spinodal shell cellular materials [46]. The surface area of the 3D print-
lar number of layers, the polymer in them is cured to a similar degree. ing models for all three topologies is extracted in Netfabb, a 3D print-
The important implication is that size effects on the base material ing software, and plotted in Fig. D1. Notice that the spinodal shell
properties, while generally important, can be ignored in the analysis topology has significantly more surface area than both Schwartz P
of the mechanical properties of the IPC composites discussed in this and octet lattice topologies, especially at lower volume fractions of
study. reinforcement. As seen in Fig. 6, even at high volume fraction, where
the octet lattice has similar surface area, spinodal shell IPCs are still
better at avoiding catastrophic failures and have better energy absorp-
Appendix C. Comparison of the deformation mechanisms in
tion as a result. This is attributed to the stochastic nature and complex
composite and cellular (reinforcement only) samples
shape of the spinodal shell topology.
The influence of curvature and surface area on mechanical proper-
Past research has shown that the soft phase in a 3D printed compos-
ties of spinodal shell IPCs deserves further investigation. Some prelim-
ite helps stabilize the reinforcing cellular materials which leads to
inary results are shown in Fig. D3. Samples with λ=L ¼ 1=8 have the
more uniform deformation and increased load bearing capacity [34].
highest surface area and curvature, while samples with λ=L ¼ 1=3 have
Fig. C1 illustrates the deformation mechanisms of IPC and cellular
the lowest. It can be seen that samples with λ=L ¼ 1=8 seem to perform
(i.e., reinforcement‐only) samples, for all three reinforcement topolo-
slightly better mechanically than samples with λ=L ¼ 1=5, and signifi-
gies, i.e. octet lattice, Schwartz P and spinodal shell. For all three rein-
cantly better than samples λ=L ¼ 1=3. One possible explanation is that
forcement topologies, the general finding is that reinforcement‐only
samples with λ=L ¼ 1=3 are more susceptible to defects from manufac-
samples fail in a more localized and catastrophic fashion, whereas
turing. It has been found in our previous study that samples with
the deformation of IPC samples is stabilized by the matrix, which dis-
λ=L ¼ 1=8are less sensitive to imperfection than samples with
tributes stress/strain more uniformly across the samples and arrests
λ=L ¼ 1=3 because of its increased surface area and more stochastic
crack propagation upon individual member fracture, ultimately delay-

Fig. D2. λ=L ¼ 1=8, λ=L ¼ 1=5 and λ=L ¼ 1=3 spinodal shell cellular materials at 30% volume fraction.

12
Y. Zhang et al. Composite Structures 263 (2021) 113693

Fig. D3. Mechanical properties of spinodal shell IPCs with different curvature/surface area. Stress–strain curves for (a) λ=L ¼ 1=8; (b) λ=L ¼ 1=5; (c) λ=L ¼ 1=3.
(d) Young’s modulus. (e) Yield strength. (f) Energy absorption.

nature [46]. Further experiments and simulation are needed to better metal dealloying. Scr Mater 2019;163:133–6. https://doi.org/10.1016/j.
scriptamat.2019.01.017.
understand this phenomenon.
[15] Liu S, Li A, Xuan P. Mechanical behavior of aluminum foam/polyurethane
interpenetrating phase composites under monotonic and cyclic compression.
Appendix E. Supplementary data Compos Part Appl Sci Manuf 2019;116:87–97. https://doi.org/10.1016/
j.compositesa.2018.10.026.
[16] Zheng Y et al. Synthesis and mechanical properties of TiC-Fe interpenetrating
Supplementary data to this article can be found online at phase composites fabricated by infiltration process. Ceram Int 2018;44
https://doi.org/10.1016/j.compstruct.2021.113693. (17):21742–9. https://doi.org/10.1016/j.ceramint.2018.08.268.
[17] Liu S, Li A, He S, Xuan P. Cyclic compression behavior and energy dissipation of
aluminum foam–polyurethane interpenetrating phase composites. Compos Part
References Appl Sci Manuf 2015;78:35–41. https://doi.org/10.1016/
j.compositesa.2015.07.016.
[1] Clarke DR. Interpenetrating phase composites. J Am Ceram Soc 1992;75 [18] Okulov AV, Volegov AS, Weissmüller J, Markmann J, Okulov IV. Dealloying-based
(4):739–58. https://doi.org/10.1111/j.1151-2916.1992.tb04138.x. metal-polymer composites for biomedical applications. Scr Mater
[2] Wegner LD, Gibson LJ. The mechanical behaviour of interpenetrating phase 2018;146:290–4. https://doi.org/10.1016/j.scriptamat.2017.12.022.
composites – I: modelling. Int J Mech Sci 2000;42(5):925–42. https://doi.org/ [19] Agrawal P, Sun CT. Fracture in metal–ceramic composites. Compos Sci Technol
10.1016/S0020-7403(99)00025-9. 2004;64(9):1167–78. https://doi.org/10.1016/j.compscitech.2003.09.026.
[3] Jhaver R, Tippur H. Processing, compression response and finite element modeling [20] Zhu J, Wang F, Wang Y, Zhang B, Wang L. Interfacial structure and stability of a
of syntactic foam based interpenetrating phase composite (IPC). Mater Sci Eng A co-continuous SiC/Al composite prepared by vacuum-pressure infiltration. Ceram
2009;499(1):507–17. https://doi.org/10.1016/j.msea.2008.09.042. Int 2017;43(8):6563–70. https://doi.org/10.1016/j.ceramint.2017.02.085.
[4] Abueidda DW, Dalaq AS, Abu Al-Rub RK. Micromechanical finite element [21] Compton BG, Lewis JA. 3D-printing of lightweight cellular composites. Adv Mater
predictions of a reduced coefficient of thermal expansion for 3D periodic 2014;26(34):5930–5. https://doi.org/10.1002/adma.201401804.
architectured interpenetrating phase composites. Compos. Struct. [22] Wang X, Jiang M, Zhou Z, Gou J, Hui D. 3D printing of polymer matrix composites:
2015;133:85–97. https://doi.org/10.1016/j.compstruct.2015.06.082. A review and prospective. Compos Part B Eng 2017;110:442–58. https://doi.org/
[5] Lee J-H, Wang L, Boyce MC, Thomas EL. Periodic bicontinuous composites for high 10.1016/j.compositesb.2016.11.034.
specific energy absorption. Nano Lett 2012;12(8):4392–6. https://doi.org/ [23] Hong S et al. 3D printing of highly stretchable and tough hydrogels into complex,
10.1021/nl302234f. cellularized structures. Adv Mater 2015;27(27):4035–40. https://doi.org/
[6] Agarwal A, Singh IV, Mishra BK. Numerical prediction of elasto-plastic behaviour 10.1002/adma.201501099.
of interpenetrating phase composites by EFGM. Compos Part B Eng [24] Palaganas NB et al. 3D printing of photocurable cellulose nanocrystal composite
2013;51:327–36. https://doi.org/10.1016/j.compositesb.2013.03.022. for fabrication of complex architectures via stereolithography. ACS Appl Mater
[7] Chen Y, Wang L. Periodic co-continuous acoustic metamaterials with overlapping Interfaces 2017;9(39):34314–24. https://doi.org/10.1021/acsami.7b09223.
locally resonant and Bragg band gaps. Appl Phys Lett 2014;105(19):. https://doi. [25] Zhang M et al. 3D printed Mg-NiTi interpenetrating-phase composites with high
org/10.1063/1.4902129191907. strength, damping capacity, and energy absorption efficiency. Sci. Adv. 2000;6
[8] Li G et al. Simulation of damage and failure processes of interpenetrating SiC/Al (19):eaba5581. https://doi.org/10.1126/sciadv.aba5581.
composites subjected to dynamic compressive loading. Acta Mater [26] Schoen AH. Infinite periodic minimal surfaces without self-intersect. NASA
2014;78:190–202. https://doi.org/10.1016/j.actamat.2014.06.045. Technical Note D-5541 1970.
[9] Del Frari G, Shadlou S, Wegner LD. Comparison of the elastic and plastic [27] Rajagopalan S, Robb RA. Schwarz meets Schwann: Design and fabrication of
behaviours of two interpenetrating phase composites with HCP inspired biomorphic and durataxic tissue engineering scaffolds. Med Image Anal 2006;10
morphologies. Int J Mech Sci 2020;186:. https://doi.org/10.1016/j. (5):693–712. https://doi.org/10.1016/j.media.2006.06.001.
ijmecsci.2020.105891105891. [28] Bonatti C, Mohr D. Mechanical performance of additively-manufactured
[10] Xie L et al. Enhancement of toughness of SiC through compositing SiC–Al anisotropic and isotropic smooth shell-lattice materials: Simulations &
interpenetrating phase composites. Nanotechnology 2020;31(13):. https://doi. experiments. J Mech Phys Solids 2019;122:1–26. https://doi.org/10.1016/j.
org/10.1088/1361-6528/ab6468135706. jmps.2018.08.022.
[11] Kouzeli M, Dunand DC. Effect of reinforcement connectivity on the elasto-plastic [29] Wang L, Lau J, Thomas EL, Boyce MC. Co-continuous composite materials for
behavior of aluminum composites containing sub-micron alumina particles. Acta stiffness, strength, and energy dissipation. Adv Mater 2011;23(13):1524–9.
Mater 2003;51(20):6105–21. https://doi.org/10.1016/S1359-6454(03)00431-2. https://doi.org/10.1002/adma.201003956.
[12] Chen Y et al. Preparation, microstructure and deformation behavior of Zr-based [30] Dalaq AS, Abueidda DW, Abu Al-Rub RK. Mechanical properties of 3D printed
metallic glass/porous SiC interpenetrating phase composites. Mater Sci Eng A interpenetrating phase composites with novel architectured 3D solid-sheet
2011;530:15–20. https://doi.org/10.1016/j.msea.2011.08.063. reinforcements. Compos Part Appl Sci Manuf 2016;84:266–80. https://doi.org/
[13] Qi Y, Chen G, Li Z, Chen L, Han W, Du Z. A novel approach to fabricate ceramic/ 10.1016/j.compositesa.2016.02.009.
metal interpenetrating phase composites by ultrasonic-assisted spontaneous [31] Al‐Ketan O, Al‐Rub RKA. Mechanical Properties of a New Type of Architected
infiltration. Ceram Int 2021;47(2):2903–7. https://doi.org/10.1016/j. Interpenetrating Phase Composite Materials. Adv. Mater. Technol. 2017;2
ceramint.2020.09.121. (2):1600235. https://doi.org/10.1002/admt.201600235.
[14] Okulov IV, Geslin P-A, Soldatov IV, Ovri H, Joo S-H, Kato H. Anomalously low [32] Al-Ketan O, Adel Assad M, Abu RKAl-Rub. Mechanical properties of periodic
modulus of the interpenetrating-phase composite of Fe and Mg obtained by liquid interpenetrating phase composites with novel architected microstructures.

13
Y. Zhang et al. Composite Structures 263 (2021) 113693

Compos. Struct. 2017;176:9–19. https://doi.org/10.1016/ [49] Deshpande VS, Fleck NA, Ashby MF. Effective properties of the octet-truss lattice
j.compstruct.2017.05.026. material. J Mech Phys Solids 2001;49(8):1747–69. https://doi.org/10.1016/
[33] Dalaq AS, Abueidda DW, Abu Al-Rub RK. Finite element prediction of effective S0022-5096(01)00010-2.
elastic properties of interpenetrating phase composites with architectured 3D [50] Mohr D. Mechanism-based multi-surface plasticity model for ideal truss lattice
sheet reinforcements. Int. J. Solids Struct. 2016;83:169–82. https://doi.org/ materials. Int J Solids Struct 2005;42(11):3235–60. https://doi.org/10.1016/j.
10.1016/j.ijsolstr.2016.01.011. ijsolstr.2004.10.032.
[34] Mansouri MR, Montazerian H, Schmauder S, Kadkhodapour J. 3D-printed [51] O’Masta MR, Dong L, St-Pierre L, Wadley HNG, Deshpande VS. The fracture
multimaterial composites tailored for compliancy and strain recovery. Compos toughness of octet-truss lattices. J Mech Phys Solids 2017;98:271–89. https://doi.
Struct 2018;184:11–7. https://doi.org/10.1016/j.compstruct.2017.09.049. org/10.1016/j.jmps.2016.09.009.
[35] Torquato S, Donev A. Minimal surfaces and multifunctionality. Proc R Soc Lond [52] Zheng X, Fu Z, Du K, Wang C, Yi Y. Minimal surface designs for porous materials:
Ser Math Phys Eng Sci 2004;460(2047):1849–56. https://doi.org/10.1098/ from microstructures to mechanical properties. J Mater Sci 2018;53
rspa.2003.1269. (14):10194–208. https://doi.org/10.1007/s10853-018-2285-5.
[36] Torquato S, Hyun S, Donev A. Multifunctional composites: Optimizing [53] Hsieh M-T, Valdevit L. Minisurf – A minimal surface generator for finite element
microstructures for simultaneous transport of heat and electricity. Phys Rev Lett modeling and additive manufacturing. Softw Impacts 2020;6:100035. https://doi.
2002;89(26):. https://doi.org/10.1103/PhysRevLett.89.266601266601. org/10.1016/j.simpa.2020.100026.
[37] Cahn JW, Hilliard JE. Free Energy of a Nonuniform System. I. Interfacial Free [54] Hsieh M-T, Valdevit L. Update (2.0) to MiniSurf—A minimal surface generator for
Energy. J Chem Phys 1958;28(2):258–67. https://doi.org/10.1063/1.1744102. finite element modeling and additive manufacturing. Softw Impacts 2020;6:.
[38] Hakamada M, Mabuchi M. Microstructural evolution in nanoporous gold by https://doi.org/10.1016/j.simpa.2020.100035100035.
thermal and acid treatments. Mater Lett 2008;62(3):483–6. https://doi.org/ [55] D20 Committee, “Test Method for Compressive Properties of Rigid Plastics,” ASTM
10.1016/j.matlet.2007.05.086. International. 10.1520/D0695-15.
[39] Stratford K, Adhikari R, Pagonabarraga I, Desplat J-C, Cates ME. Colloidal [56] Bell D, Siegmund T. 3D-printed polymers exhibit a strength size effect. Addit
jamming at interfaces: A route to fluid-bicontinuous gels. Science 2005;309 Manuf 2018;21:658–65. https://doi.org/10.1016/j.addma.2018.04.013.
(5744):2198–201. https://doi.org/10.1126/science.1116589. [57] Fleck NA, Deshpande VS, Ashby MF. Micro-architectured materials: past, present
[40] Lee MN, Mohraz A. Bicontinuous macroporous materials from bijel templates. Adv and future. Proc R Soc Lond Math Phys Eng Sci 2010;466(2121):2495–516.
Mater 2010;22(43):4836–41. https://doi.org/10.1002/adma.201001696. https://doi.org/10.1098/rspa.2010.0215.
[41] Seker E, Reed ML. Nanoporous Gold: Fabrication, Characterization, and [58] Gibson LJ, Ashby MF. Cellular Solids: Structure and Properties. Cambridge
Applications. Materials 2009;2(4). https://doi.org/10.3390/ma2042188. University Press; 1999.
[42] Biesiekierski A et al. Extraordinary high strength Ti-Zr-Ta alloys through [59] Sutherland LS, Guedes CSoares. The use of quasi-static testing to obtain the low-
nanoscaled, dual-cubic spinodal reinforcement. Acta Biomater 2017;53:549–58. velocity impact damage resistance of marine GRP laminates. Compos. Part B Eng.
https://doi.org/10.1016/j.actbio.2017.01.085. 2012;43(3):1459–67. https://doi.org/10.1016/j.compositesb.2012.01.002.
[43] Findik F. Improvements in spinodal alloys from past to present. Mater Des [60] Evans AG, He MY, Deshpande VS, Hutchinson JW, Jacobsen AJ, Carter WB.
2012;42:131–46. https://doi.org/10.1016/j.matdes.2012.05.039. Concepts for enhanced energy absorption using hollow micro-lattices. Int J Impact
[44] Garcia AE et al. Scalable synthesis of gyroid-inspired freestanding three- Eng 2010;37(9):947–59. https://doi.org/10.1016/j.ijimpeng.2010.03.007.
dimensional graphene architectures. Nanoscale Adv 2019;1(10):3870–82. [61] Izard AG, Bauer J, Crook C, Turlo V, Valdevit L. Ultrahigh energy absorption
https://doi.org/10.1039/C9NA00358D. multifunctional spinodal nanoarchitectures. Small 2019;15(45):1903834. https://
[45] Portela CM et al. Extreme mechanical resilience of self-assembled doi.org/10.1002/smll.201903834.
nanolabyrinthine materials. Proc Natl Acad Sci 2020;117(11):5686–93. https:// [62] Crook C, Bauer J, Guell Izard A, de Oliveira CS, de Souza e Silva JM, Berger JBL
doi.org/10.1073/pnas.1916817117. Valdevit. Plate-nanolattices at the theoretical limit of stiffness and strength. Nat.
[46] Hsieh M-T, Endo B, Zhang Y, Bauer J, Valdevit L. The mechanical response of Commun. 2020;11:1579. https://doi.org/10.1038/s41467-020-15434-2.
cellular materials with spinodal topologies. J Mech Phys Solids 2019;125:401–19. [63] Bauer J et al. Additive manufacturing of ductile, ultrastrong polymer-derived
https://doi.org/10.1016/j.jmps.2019.01.002. nanoceramics. Matter 2019;1(6):1547–56. https://doi.org/10.1016/
[47] Han SC, Lee JW, Kang K. A new type of low density material: Shellular. Adv Mater j.matt.2019.09.009.
2015;27(37):5506–11. https://doi.org/10.1002/adma.201501546. [64] Hong SY et al. Experimental investigation of mechanical properties of UV-Curable
[48] Han SC, Choi JM, Liu G. A Microscopic Shell Structure with Schwarz’s D -Surface. 3D printing materials. Polymer 2018;145:88–94. https://doi.org/10.1016/j.
Sci. Rep. 2017;7(1). https://doi.org/10.1038/s41598-017-13618-3. polymer.2018.04.067.

14