Computational Materials Science 139 (2017) 162–178

Contents lists available at ScienceDirect

Computational Materials Science

journal homepage: www.elsevier.com/locate/commatsci

Design of the P-surfaced shellular, an ultra-low density material with

micro-architecture
Ban Dang Nguyen a, Seung Chul Han b, Yoon Chang Jung b, Kiju Kang b,⇑
a
Electro-Optical Systems Institute, Viettel R&D Institute, Hanoi, Viet Nam
b
School of Mechanical Engineering, Chonnam National University, Gwangju 61186, Republic of Korea

a r t i c l e i n f o a b s t r a c t

Article history: The P-surface is an intersection-free smooth surface that has constant mean curvature everywhere on the
Received 21 March 2017 surface, and periodicity in three directions in 3D space. A cellular material composed of thin shells in this
Received in revised form 20 July 2017 configuration, named Shellular, reveals stretching-dominated deformation under external loading. The
Accepted 20 July 2017
thin continuous shell of a Shellular could play the role of a transfer interface between the two sub-
volumes, as well as a mechanical load support. This paper presents a preliminary study to design an opti-
mal shape of the P-surfaced Shellular. Analysis of the geometry and finite element analyses are performed
Keywords:
to investigate the effects of the geometric parameters and boundary conditions on their mechanical prop-
P-surface
Micro-architectured material
erties. It is shown that the geometry and mechanical properties of a P-surfaced Shellular can be expressed
Periodic cellular material in empirical equations with only two independent dimensionless parameters, i.e., the volume fraction
Elastic buckling and the ratio of the wall thickness to cell size; and that the effect of the volume fraction is substantial.
A low volume fraction is very beneficial to achieve high strength, and suppresses elastic buckling to delay
the transition of the failure mode from plastic yielding to elastic buckling as the wall thickness is
decreased.
Ó 2017 Published by Elsevier B.V.

1. Introduction
The triply periodic minimal surface (TPMS), which was first
described by the German mathematician, H. A. Schwarz in the
19th Century, is an intersection-free smooth surface that has zero
(or in the wider sense, a constant) mean curvature at every point
on the surface, and periodicity in three directions in space [1].
The surface with zero mean curvature provides a minimal surface
area for a given boundary, as observed in soap film formed within a
rigid frame. A TPMS partitions space into two disjoint but inter-
winding sub-volumes that are simultaneously continuous. Fig. 1
shows typical examples of the TPMSs, i.e., the P, G, D, I-WP, and
F-RD surfaces. Among the TPMSs, the P-surface has important fea-
tures over the others, which would be useful for applications as
ultralight structural materials and tissue engineering scaffolds.
Namely, P-surface has the least surface area and the highest fluid
permeability through the sub-volumes for a given overall volume
[2–4].
A thin shell is defined as a curved surface structure that sup-
ports external loads mainly through coplanar stresses [5]. If a thin
shell structure has a TPMS configuration, stress concentration due
⇑ Corresponding author.
E-mail address: kjkang@jnu.ac.kr (K. Kang).
to geometrical irregularity does not occur, and the structure effec-
tively resists against external load only through coplanar stresses,
because a TPMS has constant curvature over the surface. Therefore,
the thin shell structure in the TPMS configuration is another case
that reveals stretching-dominated deformation [6], in addition to
the material with periodic micro-architecture of trusses. In fact,
Rajagopalan and Robb [7] showed the uniform distribution of the
stresses in a TPMS shell subjected to compressive external load,
and they expected high strength of tissue engineering scaffold with
the TPMS shell. (Of course, at a critical state initiating plastic yield-
ing or elastic buckling after the deformation substantially pro-
gresses with load increase, the stresses become concentrated to
some areas. These states will be elaborated in Section 4.3.)
Through literature survey, we note that densities of most natu-
ral and engineering materials are limited down to 10
2 g/cc,
(which will be elaborated in Section 4.4). To obtain the lower den-
sity with a sufficient strength, a special geometry as well as high
mechanical properties of the constituent material is needed. There-
fore, we defined the ultralow density materials (ULDM) as the
materials that belong to the category. The typical examples are
Microlattice [8], Nanolattice [9,10], and Mechanical metamaterial
[11]. All of them were composed of periodic micro-architectures
of hollow trusses.

http://dx.doi.org/10.1016/j.commatsci.2017.07.025
0927-0256/Ó 2017 Published by Elsevier B.V.
 B.D. Nguyen et al. / Computational Materials Science 139 (2017) 162–178 163

Fig. 1. Typical examples of the triply periodic minimal surfaces (TPMSs), i.e., P, G, D, I-WP, and F-RD surfaces.

Han et al. [12] first introduced another concept of ULDM with a
periodic micro-architecture composed of smooth-curved thin
shells in a TPMS configuration, named Shellular. The first in
Fig. 2(a) shows the specimen. Han et al. [12] fabricated Shellular
based on a 3D photo-lithographical technique (self-propagating
polymer wave guide; SPPW) [13]. Namely, they used SPPW to form
a polymer template, then coated a Ni-P layer on the surface using
electroless plating, and finally etched out the template to obtain a
hollow structure of Shellular. Also, they demonstrated that the
Shellular specimens with a micro-architecture similar to
P-surface, mentioned above, had strength and Young’s modulus
comparable to the previous ultralow density materials, such as
Microlattice [9], Nanolattice [9,10], and Mechanical metamaterial
[11], shown in the rest three of Fig. 2(a). All of the previous ultra-
low density materials were fabricated according to the same
principle as that of the Shellular, based on templates formed by
3D photo-lithographical techniques; however, they had totally dif-
ferent micro-architectures from Shellular, i.e., hollow trusses ver-
sus smooth shells. Because of the intrinsic nature of SPPW, the
Shellular specimens fabricated by Han et al. [12] did not have a
configuration that exactly agreed with a mathematical P-surface;
the configurations were substantially elongated in the vertical
direction, and were rather seen to be composed of octahedral cells
with smoothened ridges. Fig. 2(b) shows the 3D tomography mea-
sured by a micro-CT on the real Shellular specimen, revealing peri-
odic unit cells [14]. Ban et al. sought an optimal design for the unit
164 B.D. Nguyen et al. / Computational Materials Science 139 (2017) 162–178

Fig. 2. (a) Specimens of Shellular [12], Microlattice [8], Nano lattice [9], Mechanical metamaterial [11], and (b) 3D tomography measured by a micro-CT on the real Shellular
specimen, revealing periodic unit cells [14].

cell model in a simplified form, as well as a realistic configuration Recently, the authors’ research group developed two alternative
for the maximum possible strength or modulus for a given density methods for forming the templates that could be used to fabricate
[15]. P-surfaced Shellulars. Fig. 3 depicts these methods. The first
 B.D. Nguyen et al. / Computational Materials Science 139 (2017) 162–178 165

Fig. 3. Two alternative methods for forming the templates that could be used to fabricate P-surfaced Shellulars under development in the authors’ research group, based on
(a) a wire-woven structure and (b) polymer beads arranged in a primitive cubic (PC) crystallographic pattern.

method of Fig. 3(a) is based on a wire-woven structure. After a equations are derived for the compressive strength, depending
wire-framework was woven into a diamond truss pattern [16], on the failure modes, and the Young’s modulus. Also, the upper
resin was applied and cured to fix the framework, and fill the inte- and lower bounds of the strength and modulus of P-surfaced Shel-
rior space. The surface profile in the P-surface could be achieved by lular are estimated according to the relative density, and are then
repeating the process to additionally coat and cure the resin with a compared with those properties of previous ultralow density
proper viscosity [17]. The surface tension of the resin made the materials with periodic micro-architectures.
surface minimal. The effect of gravity could be minimized by using
only a small amount of resin at each coating and periodically
2. Geometry of P-surface
turning over the structure during curing. The second method of
Fig. 3(b) used polymer beads arranged in a primitive cubic (PC)
The P-Surface located in a cube of the cell size of the P-surface,
crystallographic pattern, followed by resin infiltration and subse-
D, is approximately given by
quent chemical dissolution of beads to provide smooth connec-       
tions between the beads, resulting in the surface profile of the 2px 2py 2pz
Fðx; y; z; kÞ ¼ cos þ cos þ cos þ k ¼ 0;
P-surface [18]. Both structures were used as templates to fabricate D D D
Shellulars through hard coating and then etching out, in similar ð1Þ
fashion to other ultralow density materials with a periodic
 
micro-architecture. The wires used in the first method were of a where x; y; z 2
D2 ; D2 and k is a level set value that decides volume
polymer, to be specific, PLA (poly lactic acid). Therefore, the wire fraction, f. The volume fraction is defined as the ratio of the inner
framework was easily etched out by using chloroform after etching sub-volume to the overall volume. For example, in Fig. 1(a), the
out the outer resin that is infiltrated on the wire frame. overall volume and inner sub-volume are the volume of the mini-
This paper presents a preliminary study to be used to design an mum cube surrounding the outer surfaces of the TPMS (P-surface
optimal shape of P-surfaced Shellular. First, the geometry of the in this case) and the volume of the sub-volume bounded by the
P-surface and the parameters governing the geometry are inner (colored yellow1 in this case) surface, respectively. Fig. 4(a)
described. For unit cell models of P-surfaced Shellulars, finite ele- shows examples of the unit cells of P-surface family with various
ment analyses are performed to investigate the effects of the geo-
metric parameters and boundary conditions on their compressive 1
For interpretation of color in Fig. 1, the reader is referred to the web version of
strengths and Young’s moduli. Based on the results, empirical this article.
166 B.D. Nguyen et al. / Computational Materials Science 139 (2017) 162–178
values of volume fraction, f. Simple calculation using a CAD program
gives the relations between k and f, and the surface area, A, and f as
follows:
f ¼ 0:29k þ 0:5 and
A 4 2
¼
13:054ðf
0:5Þ
4:555ðf
0:5Þ þ 2:34; respectively:
D2
Therefore, the geometry of a P-surfaced Shellular with uniform
shell thickness, t, can be defined by only two independent param-
eters, f and t/D. Shellular with one of the TPMS configurations,
including the P-surface, is composed of anticlastic shells, quite sim-
ilar to the hyperbolic paraboloid. In fact, the mechanical resistance
of Shellular comes from its unique geometry of anticlastic shells
having convex and concave curvatures together [19,20]. By simply
multiplying A/D2 with t/D, the relative density is given as follows:
q  4 2
t
¼
13:054ðf
0:5Þ
4:555ðf
0:5Þ þ 2:34 : ð4Þ
qs D Microlattice. The periodic boundary condition is applied through
3. Mechanical analysis of the P-surfaced shellular
3.1. Modeling
It takes several steps to create a meshed model of a P-Surface
Shellular. First, Eq. (1) with the level set value, k, calculated for a
desired volume fraction, f, using Eq. (2) is input into an open source
code, K3Dsurf, to generate the point cloud data of the surface, as
shown in the first figure of Fig. 4(b). In the next step, a meshed
model is built up using the point cloud in Catia V5R20 with the
optional Digitized Shape Editor, as shown in the second figure of
Fig. 4(b). A model meshed by triangular elements is likely to result
in an overestimation problem in stiffness calculation [21]. There-
fore, the model is remeshed using HyperMesh (Altair Engineering,
Inc., MI, USA) to have quadratic elements; and finally, the thickness
is provided to form the final shell structure, using the surface
formed in the previous step as a middle surface. A half of the thick-
ness, t, is perpendicularly added on each side of the surface. The
third and fourth figures of Fig. 4(b) show the unit cell models of
the P-surface with the quadratic elements and the final Shellular
with solid elements, respectively. In the thickness direction, thin
models and thick models had 2 elements and 4 elements, respec-
tively. Additional analyses were carried out to confirm that other
models with the more elements through the thickness gave the
results similar to those with the current elements.
3.2. Finite Element Analysis (FEA)
The behaviors of the Shellular models under compression are
simulated by a FEA solver, ABAQUS v 6.9 to calculate the strengths
and Young’s moduli. Solid elements (specifically, type C3D8I) are
used to perform FEA on the models with four different volume
fractions of f = 0.23, 0.27, 0.35, 0.5 and seven wall thickness of
t = 15, 10, 5, 2, 1, 0.5, 0.2 lm. The size of the unit cells is fixed as
D = 1000 lm. For high accuracy, the aspect ratio of an element is
limited so as not to be larger than 10, and the greatest possible
number of elements are taken (for example, 2,000,000 for the
model with the thinnest wall). The properties of the constituent
material (Ni-P film) were given as follows: Young’s modulus of
Es = 210 GPa [22], Poisson’s ratio of ms = 0.3, and yield strength of
rys = 2 GPa.
Although the constituent material was assumed to be elastic-
perfectly plastic, to improve the convergence of the FEA, we pro-
vided a slow linear hardening after the initial yielding as the input
data of the material properties. Namely, the strength was input to
increase from the yield strength of 2 GPa (at e = 2 GPa/210
GPa  0.01) to 4 GPa at e = 1.0. This is a quite common technique
used when handling an elastic perfectly plastic material or when
ð2Þ
the post yielding properties are unknown.
The choice of appropriate boundary conditions is critical to
ensure results that are applicable to multiple-celled samples, and
that thus facilitate comparison with experimental results. The
ð3Þ nodes on the top face (+z face) of each unit cell model are subject
to downward displacement, and the nodes on the bottom face (
z
face) are fixed in the z direction, while all of the nodes on both
faces are constrained from rotations all in three directions. The
coordinate axes are indicated in Fig. 4(b). For the nodes on the
outermost lateral faces (i.e., ±x and ±y faces on the edges of the lat-
eral openings of each unit cell model, a periodic or free boundary
condition is applied to consider the upper or lower bound of the
mechanical properties of the P-Surface Shellular, respectively.
The free boundary condition literally lets the nodes be free from
any constraints, as done by Valdevit et al. [22] for analysis of
the linear constraint in Abaqus [21]. As an example, symbols of
the periodic boundary condition are indicated on the FEA model
at the end of Fig. 4(b). Refer to Wu et al. [23] for the technical
details. These boundary conditions are appropriate for the cases
with many cells in the x-y plane but a few cells or even one cell
in z direction, just like the Shellular and Microlattice specimens
shown in Fig. 2.
To calculate the compressive strength due to elastic buckling,
Eigenvalue buckling analyses are performed, similarly to Valdevit
et al. [22]. Loads are applied in two steps. First, a preload P is
applied to create a very small deformation prior to buckling. Then,
a perturbation load, DP, is applied by a small fraction (1%) of P. The
first estimate of the buckling strength is obtained from the small-
est positive eigenvalue k, by:
Pcr ¼ P þ kDP:
In the second step, to consider the potential geometric nonlin-
earities, an additional perturbation simulation is carried out with
a much higher preload that are 85% of the first buckling-load esti-
mate, and a perturbation load of a small fraction (1%) of the pre-
load. If the second step simulation converges, its results are used
to extract the effective buckling strength of the model. Otherwise,
the first estimate is taken as the buckling strength.
To calculate the compressive strength due to plastic yielding, a
separate elastic-plastic analysis is carried out according to J2-flow
theory, and the first peak in the stress-strain curve is taken as
the yield strength. To calculate Young’s modulus, a small displace-
ment dz is applied to the face +z (corresponding to ez =
0.001%)
and to obtain reaction force FR. Young’s modulus is calculated from
the displacement and reaction force, as if each model was a homo-
geneous material.
4. Results and discussion
Table 1 summarizes the results obtained from the FEA for all the
unit cell models of the P-surfaced Shellular. The data of the relative
density, compressive strength and Young’s modulus of each model
are listed as functions of the volume fraction, wall thickness, and
boundary condition. The compressive strengths due to plastic
yielding and elastic buckling for each model are independently cal-
culated, as mentioned above, and compared to each other. If one is
substantially higher than the other for the given wall thickness and
volume fraction, we conclude that the failure mode corresponding
to the higher value is no longer activated, and the data is skipped in
the table.
 B.D. Nguyen et al. / Computational Materials Science 139 (2017) 162–178 167

Fig. 4. (a) Configurations of unit cells of P-surface with various values of volume fractions, (b) several steps to create a meshed model of a P-Surface Shellular for finite
element analysis.

As mentioned above, the mechanical properties and failure versus the periodic boundary condition. The free boundary condi-
mechanisms are investigated under two different boundary condi- tion assumes the absence of neighbor cells, allowing the lateral
tions for the nodes on the outermost lateral faces, namely, the free surface to deform without any interference. Hence, the mechanical
168 B.D. Nguyen et al. / Computational Materials Science 139 (2017) 162–178

Table 1
All results obtained from the FEA for the unit cell models of P-surfaced Shellular. The data of the relative density, compressive strength and Young’s modulus of each model are
listed as functions of volume fraction, wall thickness, and boundary condition.

Volume Wall Thickness t Relative Density Free BCs Periodic BCs

Fraction f (lm) q/qs
Young’s Modulus Compressive Strength (MPa) Young’s Modulus Compressive Strength (MPa)
(MPa) (MPa)
Elastic Plastic Elastic Plastic
Buckling Yielding Buckling Yielding
0.23 0.2 3.88E
04 1.16E
02 3.40E
04 4.28E+00 2.01E
03
0.5 9.69E
04 9.39E
02 5.18E
03 1.22E+01 1.90E
02
1 1.94E
03 5.96E
01 4.62E
02 2.77E+01 8.33E
02
2 3.88E
03 4.16E+00 3.24E
01 6.57E+01 3.60E
01
5 9.69E
03 5.15E+01 4.58E+00 1.89E+00 1.94E+02 2.44E+00
10 1.94E
02 2.67E+02 6.44E+00 4.22E+02 7.34E+00
15 2.91E
02 6.51E+02 1.19E+01 1.19E+01
0.27 0.2 4.13E
04 4.64E
03 1.63E
04 3.59E+00 1.27E
03
0.5 1.03E
03 4.75E
02 1.66E
03 1.25E+01 8.18E
03
1 2.06E
03 2.35E
01 1.50E
02 3.55E+01 5.08E
02
2 4.13E
03 1.57E+00 1.41E
01 3.74E
01 8.57E+01 3.24E
01 7.36E
01
5 1.03E
02 1.91E+01 2.44E+00 1.81E+00 2.44E+02 3.62E+00 3.28E+00
10 2.06E
02 1.29E+02 5.46E+00 6.36E+02 8.32E+00
15 3.09E
02 3.72E+02 1.04E+01 1.04E+01
0.35 0.2 4.46E
04 2.94E
03 7.64E
05 3.47E+00 7.88E
04
0.5 1.12E
03 2.48E
02 8.60E
04 1.25E+01 5.44E
03
1 2.23E
03 1.38E
01 6.92E
03 3.55E+01 3.16E
02
2 4.46E
03 6.93E
01 4.84E
02 9.64E
02 9.30E+01 2.34E
01 4.40E
01
5 1.12E
02 7.58E+00 9.00E
01 8.60E
01 2.88E+02 2.82E+00 2.70E+00
10 2.23E
02 4.94E+01 2.66E+00 6.17E+02 9.78E+00
15 3.35E
02 1.50E+02 5.84E+00 1.64E+01
0.5 0.2 4.68E
04 6.38E
04 1.25E
05 2.37E+00 3.00E
04
0.5 1.17E
03 5.44E
03 1.78E
04 8.46E+00 1.76E
03
1 2.34E
03 3.82E
02 1.98E
03 2.39E+01 8.10E
03
2 4.68E
03 2.42E
01 1.98E
02 6.59E+01 6.36E
02
5 1.17E
02 2.73E+00 3.52E
01 5.00E
01 2.58E+02 8.78E
01 2.38E+00
10 2.34E
02 1.83E+01 3.18E+00 1.51E+00 6.36E+02 1.97E+01 6.92E+00
15 3.51E
02 3.22E+00 1.60E+01

resistance, i.e., strength or Young’s modulus is likely to be underes- strength. For example, at the lowest thickness of t/D = 0.0002, the
timated, and this is regarded as the lower bound. In contrast, the strength with f = 0.23 is 30 times as high as that with f = 0.5. Also,
periodic boundary condition assumes a uniform deformation at a low volume fraction, the transition of failure mode from plastic
regardless of cells. Namely, all the cells deform in an exactly iden- yielding to elastic buckling according to decrease of the wall thick-
tical pattern, ignoring any local irregularity. Hence, the mechanical ness is delayed.
resistance is likely to be overestimated, and this is regarded as the Through simple linear regression of the data, empirical equa-
upper bound. Here, we present the mechanical properties sepa- tions of the strengths due to plastic yielding and elastic buckling
rately under the free and periodic boundary conditions. are derived as follows:
 1:7  2
ro t 1
4.1. Mechanical properties under free boundary condition ¼ 0:45 and
ros D f

Fig. 5(a) shows variation of the compressive strengths of  3  4

ro t 1
P-surfaced Shellulars under free boundary condition according to ¼ ; respectively:
ros D f
the wall thickness. The strengths due to the plastic yielding are
shown together with those due to the elastic buckling. The Fig. 5(b) shows the equations fitted to all the discrete data,
strengths and wall thickness are normalized by the yield strength which reveals excellent agreement.
of the constituent material (Ni-P film), ros, and the cell size, D, Fig. 6(a) shows the Young’s moduli of P-surfaced Shellulars as
respectively. For a given volume fraction, f, the data of the functions of the wall thickness. Similarly to Fig. 5(a), the Young’s
strengths due to plastic yielding and those due to the elastic buck- modulus and wall thickness are normalized by the Young’s modu-
ling are indicated by the blind and open symbols of identical color, lus of the constituent material (Ni-P film), Es, and the cell size, D,
respectively. Solid and dashed lines indicate the fitted lines of the respectively. The lines fitted to the discrete data have an approxi-
data due to the two failure mechanism. The lines fitted to the mately constant slope of 2.6 in log-log coordinates. The effects of
strengths due to plastic yielding have an approximately constant the volume fraction are substantial. A lower volume fraction
slope of 1.7 in log-log coordinates; whereas those due to elastic results in a higher Young’s modulus. Specifically, at a given wall
buckling have a higher slope of 3. When the wall is very thick such thickness, the Young’s moduli with f = 0.23 is about 20 times as
as t/D = 0.015, the P-surfaced Shellular always fails due to plastic high as that with f = 0.5. Through simple linear regression of the
yielding. However, as the thickness decreases below a certain limit, data, an empirical equation of the Young’s modulus is derived as
the failure mechanism switches to elastic buckling and the follows:
strength decreases much faster. The effects of the volume fraction  2:6  3:5
E t 1
on the strength are substantial, particularly when the failure mode ¼ 0:45 :
is elastic buckling. A lower volume fraction results in higher Es D f
 B.D. Nguyen et al. / Computational Materials Science 139 (2017) 162–178 169

Fig. 5. (a) Variation of compressive strength of P-surfaced Shellulars under free boundary condition according to the wall thickness, (b) empirical equations of the
compressive strength fitted to the discrete data.
170 B.D. Nguyen et al. / Computational Materials Science 139 (2017) 162–178

Fig. 6. (a) Variation of Young’s modulus of P-surfaced Shellulars under free boundary condition according to the wall thickness, (b) empirical equations of the Young’s
modulus fitted to the discrete data.
 B.D. Nguyen et al. / Computational Materials Science 139 (2017) 162–178
Fig. 6(b) shows the equations fitted to all the discrete data,
which reveals excellent agreement.
4.2. Mechanical properties under periodic boundary condition
Fig. 7(a) shows the compressive strengths of P-surfaced Shellu-
lars under periodic boundary condition as functions of the wall
thickness. The strengths due to plastic yielding are not well fitted
linearly in the log-log coordinates at a given volume fraction. Also,
the effects of the volume fraction on the strength are negligible.
Rather, the whole data are approximately fitted to a single line
with a slope, 1.57 with some scatter. In contrast, the strengths
due to elastic buckling are rather well fitted linearly in the log-
log coordinates at each volume fraction. Also, the effects of the vol-
ume fraction on the strength are significant. Unlike under the free
boundary, the slopes of the fitted lines are not almost constant,
increasing with the volume fractions from 2.25 to 2.8 with an
average, 2.5. When the wall is very thick like t/D = 0.015, the
PAsurfaced Shellular always fails due to plastic yielding. However,
as the thickness decreases below a certain limit, the failure mech-
anism is switched to elastic buckling and the strength decreases
much faster. When the failure mode is elastic buckling, the lower
volume fraction results in the higher strength. However, the differ-
ences are not so significant as under the free boundary condition,
mentioned above. For example, the strength with f = 0.23 is about
10 times as high as that with f = 0.5. Also, the transition of failure
mode from plastic yielding to elastic buckling according to
decrease of the wall thickness is delayed at a low volume fraction.
Through simple linear regression of the data, empirical equation
of the strength due to plastic yielding is simply derived as
 1:57
ro t between these two buckling modes is also affected by the wall
¼ 5:15 ;
ros D thickness. For example, at a given volume fraction of f = 0.23, the
and the equation of the strength due to elastic buckling are derived
as
 2:5  2:5
ro t 1
¼ 35 : boundary condition, i.e., when the nodes at the lateral opening
ros D f
Fig. 7(b) shows the equations fitted to all the discrete data,
revealing fairly good agreement.
Fig. 8(a) shows the Young’s moduli of P-surfaced Shellulars as
functions of the wall thickness under the periodic boundary condi-
tion. The lines fitted to the discrete data have variable sloped,
1.18–1.44 in the log-log coordinates, but the effects of the volume
fraction are negligible. Rather, the whole data are approximately
fitted to a single line with a slope, 1.32 as follows:
 1:32
E t
¼ 1:244 : direction. This unique deformation pattern is the most apparently
Es D
Fig. 8(b) shows the equation fitted to the discrete data, reveal-
ing fair agreement with allowable error.
4.3. Failure behaviors
Fig. 9(a) depicts Mises stress distributed on P-surfaced Shellu-
lars, when the Shellulars fail by plastic yielding under free bound-
ary condition, i.e., no constraint on the nodes at the lateral opening
edges. Since the constituent material was assumed to be elastic
perfectly plastic, the area where the Mises stress became higher
than the given yield strength was simply indicated by a single grey
color without gradation despite the wide range of the stress.
Namely, the grey color indicates the area undergoing perfectly
plastic deformation. Because the wall thickness is high
(t = 10 lm), the Shellular fails by plastic yielding, as mentioned
171
above. Under the vertical compression, all of them reveal stress
concentration in the middle section, leading to the shell folding.
However, the details are substantially different depending on the
volume fraction. At the lowest volume fraction of f = 0.23, the high
stress is distributed around the top and bottom openings as well as
the middle section. This means that the small top and bottom
openings play an effective role in supporting the structure against
the external load. Therefore, the strength of the P-surfaced Shellu-
lar is higher than those with the higher volume fractions. That is, as
the f value increases, the stress level around the top and bottom
openings decreases and the stress becomes to be more concen-
trated around the middle section. Consequently, at the highest vol-
ume fraction of f = 0.5, the P-surfaced Shellular fails by collapse of
the middle section, and its strength is lower than those with the
lower volume fractions. Because these plastic deformations under
free boundary condition are a type of plastic buckling, and the esti-
mated compressive strengths are actually the plastic buckling
loads.
Fig. 9(b) depicts the buckled shapes of the P-surfaced Shellulars
with the thinnest wall (t = 0.2 lm) under free boundary condition.
The buckling modes are corresponding to the first Eigen values for
the given parameters. At a high volume fraction, the thin Shellular
tends to fail soon with transverse deformation near the top and
bottom (named Mode-I). In contrast, at the lowest volume fraction
of f = 0.23, the structure is more stable, revealing the high resis-
tance against buckling. When buckled, the transverse deformations
tend to appear around the middle section and lateral openings
(named Mode-II). At the volume fraction of f = 0.27, the transition
between these two buckling modes takes place and the transverse
deformation occurs both near the top and bottom openings and
around the middle section and lateral openings. The transition
P-surfaced Shellular with t = 2 lm buckles in Mode-I, but that with
t = 1 lm buckles in Mode-II.
Fig. 10(a) depicts Mises stress distributed on P-surfaced Shellu-
lars, when the Shellulars fail by elastic buckling under periodic
edges are fully constrained by neighbor. Because the wall thick-
ness is high (t = 10 lm), the Shellular fail by plastic yielding, as
mentioned above. Under the vertical compression, all of them
reveal the stress much well distributed over the entire body
compared to Fig. 9(a) under the free boundary condition. At the
lowest volume fraction of f = 0.23, the high stress is distributed
around all of the top, bottom, and lateral openings. The deforma-
tion pattern of the lateral openings is quite interesting. Namely,
the two lateral openings oppositely oriented to each other deform
in vertical direction, while the other two deform in horizontal
shown at the highest volume fraction of f = 0.5. When a hyperbolic
paraboloid-like shell composing the P-surface Shellular is sub-
jected to external load, the convex curvature is compressed,
whereas the concave curvature is tensioned, or vise versa [19],
as shown in the inset. To maintain the geometrical compatibility
with the neighbor cells under the periodic boundary condition,
these two modes of deformations occur alternatively around the
four lateral openings.
Fig. 10(b) depicts the buckled shapes of the P-surfaced Shellu-
lars with the thinnest wall (t = 0.2 lm) under the periodic bound-
ary condition. At the low three volume fractions, the buckling
modes corresponding to the first Eigen values reveal large trans-
vers deformation around the top and bottom. In contrast, at the
highest volume fraction of f = 0.5, the buckled shape is very similar
to that when failed by plastic yielding, shown in the last figure of
Fig. 10(a).
172 B.D. Nguyen et al. / Computational Materials Science 139 (2017) 162–178

Fig. 7. (a) Variation of compressive strength of P-surfaced Shellulars under periodic boundary condition according to the wall thickness, (b) empirical equations of the
compressive strength fitted to the discrete data.
 B.D. Nguyen et al. / Computational Materials Science 139 (2017) 162–178 173

Fig. 8. (a) Variation of Young’s modulus of P-surfaced Shellulars under periodic boundary condition according to the wall thickness, (b) empirical equations of the Young’s
modulus fitted to the discrete data.
174 B.D. Nguyen et al. / Computational Materials Science 139 (2017) 162–178

Fig. 9. (a) Mises stress distributed on P-surfaced Shellulars under free boundary condition when the Shellulars fail by plastic yielding with the thick wall (t = 10 lm), (b)
buckled shapes of the P-surfaced Shellulars with the thinnest wall (t = 0.2 lm) under free boundary condition.
 B.D. Nguyen et al. / Computational Materials Science 139 (2017) 162–178 175

Fig. 10. (a) Mises stress distributed on P-surfaced Shellulars under periodic boundary condition when the Shellulars fail by plastic yielding with the thick wall (t = 10 lm), (b)
buckled shapes of the P-surfaced Shellulars with the thinnest wall (t = 0.2 lm) under periodic boundary condition.
176 B.D. Nguyen et al. / Computational Materials Science 139 (2017) 162–178

Fig. 11. (a) Estimated compressive strength and (b) Young’s modulus of P-surface Shellulars, respectively, compared with those of previous ultra-low density materials.

4.4. Comparison with the previous ultralow density materials with
microarchitectures
In Fig. 11(a) and (b), the estimated mechanical properties of
P-surface Shellulars are compared with those of previous
ultra-low density materials such as Microlattice [8], Nanolattice
[9,10], Mechanical metamaterial [11], the Shellular with the
octahedron-like or elongated P-surfaced configurations, named
L-Shellular here [12], and also other conventional cellular
materials, i.e., engineering honeycombs [27], wire-woven metals
[28], natural materials [29], and foams [29]. A honeycomb is
usually regarded as one of the best artificial cellular materials.
Actually, according to Fleck et al.’s definition [30], a honeycomb
is the first micro-architectured material. Hence, when sandwiched
with two hard face sheets, the elastic modulus reaches almost an
ideal upper bound called the Hashin-Shtrikman bound [31].
The P-surface Shellular models are set to have a volume frac-
tion, f = 0.27 which is similar to that of the real specimens prepared
 B.D. Nguyen et al. / Computational Materials Science 139 (2017) 162–178
based on wire-weaving, as depicted in Fig. 3(a). In the figures, the
values of P-surface Shellulars are indicated by the solid lines. In
Fig. 11(a), their strengths estimated under the periodic boundary
condition are always higher than those under the free boundary
condition, which are regarded to represent the upper bound and
the lower bound, respectively. At a low relative density, which
measures the thin wall of shell, the failure occurs due to elastic
buckling and the strength rapidly decreases as the relative density
decreases. In contrast, at a high relative density, the failure occurs
due to plastic yielding and the strength slowly decreases as the rel-
ative density decreases. The transition between the two failure
modes is around the density of q/qs = 0.01. In the elastic buckling
dominated region, the strengths of the P-surfaced Shellulars are
comparable to those of L-Shellular and Microlattice. However, in
the plastic yielding dominated region, the strengths of the P-
surfaced Shellulars are higher than any of the previous ultralow
density materials with microarchitectures. This result means that
the real benefit of the unique geometry of P-surfaced Shellular
comes when the failure occurs due to plastic yielding.
In Fig. 11(b), in the most region of our interest, the Young’s
moduli of the P-surfaced Shellulars under the periodic boundary
condition are much higher than those under the free boundary
condition. In fact, the periodic boundary condition provides the
upper bound of the entire ultralow density materials, while the
free boundary condition does the lower bound. Namely, all the
Young’s modulus data are located between the upper and lower
bounds.
5. Concluding remarks
At the low density where the elastic buckling governs the fail-
ure, the strength of the P-surfaced Shellular is not significantly
high. This means that the mathematically ideal surface (TPMS) of
the geometry is not so effective to enhance the strength under
the uniaxial compression. However, we believe that under a 3D
loading such as hydrostatic tension or compression, the TPMS
geometry leads to quite uniform distribution of the stress over
the structure, as reported by Rajagopalan and Robb [7], which
substantially increases the strength. It is because the loading is
as equiaxial as the geometry is. In contrast, at the high density over
q/qs = 0.01, where the plastic yielding governs the failure, the P-
surface Shellular have noticeably higher strength than the previous
ultralow density materials, which means that the TPMS geometry
effectively suppresses the stress concentration even under the uni-
axial compression.
The periodic boundary condition always provides the higher
mechanical properties, particularly Young’s modulus, than the free
boundary condition does. This explains why any regional variance
such as local defects in geometry or material and uneven thickness
must be prohibited in a Shellular for the higher mechanical prop-
erties. Contrarily, cellular materials with truss-like microarchitec-
tures were reported to be tolerant to local defects [24,25]. Hence,
the ultralow density materials such as Microlattice [10], Nanolat-
tice [11,12], and Mechanical metamaterial [13] seem to be more
tolerant to local defects than a Shellular. Nevertheless, a Shellular
has a non-comparable benefit thanks to its unique geometry
comprising two disjoint but inter-winding sub-volumes as being
simultaneously continuous. Namely, the thin continuous shell of
a Shellular in a TPMS could play a role as a transfer interface
between the two sub-volumes, as well as a mechanical load sup-
port. For example, the interfacial shell could be used as a semiper-
meable membrane for osmosis in scaffolds for tissue engineering
[26].
In this paper, we presented a preliminary study to be used to
design an optimal shape of P-surfaced Shellular. The geometry of
177
P-surface and the parameters governing the geometry are
described. Finite element analyses are performed for unit cell mod-
els of P-surfaced Shellulars to investigate the effects of the geomet-
ric parameters and boundary conditions on their compressive
strengths and Young’s moduli. The conclusions are as follows:
(i) The geometry and mechanical properties of a P-surfaced
Shellular can be expressed in empirical equations with only
two independent dimensionless parameters, f and t/D.
(ii) When the wall is thick, the P-surfaced Shellular is likely to
fail due to plastic yielding, whereas, when the wall is thin,
the Shellular is likely failed due to elastic buckling. The
thickness at the transition between the two failure mecha-
nisms depends on the volume fraction.
(iii) The effect of the volume fraction is substantial. A low vol-
ume fraction is very beneficial to achieve high strength, par-
ticularly under the free boundary condition and also when
the failure is governed by elastic buckling under the periodic
boundary condition. The lower volume fraction suppresses
the elastic buckling to delay the transition of failure mode
from plastic yielding to the elastic buckling as the wall thick-
ness is decreased. In addition, the lower volume fraction
results in the higher Young’s modulus under free boundary
condition.
Acknowledgement
This work was supported by the National Research Foundation
of Korea (NRF) grant funded by the Korea government (MSIP)
(No. NRF-2015R1A2A1A01003702).
References
[1] S. Hyde, Z. Blum, T. Landh, S. Lidin, B.W. Ninham, S. Andersson, K. Larsson, The
Language of Shape, Elsevier, Danvers, MA, USA, 1996.
[2] Y. Jung, K.T. Chu, S. Torquato, A variational level set approach for surface area
minimization of triply-periodic surfaces, J. Comput. Phys. 223 (2007) 711–730,
http://dx.doi.org/10.1016/j.jcp.2006.10.007.
[3] Y. Jung, S. Torquato, Fluid permeabilities of triply periodic minimal surfaces,
Phys. Rev. E - Stat. Nonlinear, Soft Matter Phys. 72 (2005) 1–8 doi:10.1103/
PhysRevE.72.056319.
[4] J.K. Guest, J.H. Prévost, Optimizing multifunctional materials: Design of
microstructures for maximized stiffness and fluid permeability, Int. J. Solids
Struct. 43 (2006) 7028–7047, http://dx.doi.org/10.1016/j.ijsolstr.2006.03.001.
[5] S.P. Timoshenko, S. Woinowsky-Krieger, Theory of Plates and Shells, McGraw-
Hill, London, 1959.
[6] V.S. Deshpande, M.F. Ashby, N.A. Fleck, Foam topology: bending versus
stretching dominated architectures, Acta Mater. 49 (2001) 1035–1040.
[7] S. Rajagopalan, R.A. Robb, Schwarz meets Schwann: Design and fabrication of
biomorphic and durataxic tissue engineering scaffolds, Med. Image Anal. 10
(2006) 693–712.
[8] T.A. Schaedler, A.J. Jacobsen, A. Torrents, A.E. Sorensen, J. Lian, J.R. Greer, L.
Valdevit, W.B. Carter, Ultralight metallic microlattices, Science 334 (2011)
962–965.
[9] D. Jang, L.R. Meza, F. Greer, J.R. Greer, Fabrication and deformation of three-
dimensional hollow ceramic nanosctructures, Nat. Mater. 12 (2013) 893–898.
[10] L.R. Meza, S. Das, J.R. Greer, Strong, lightweight, and recoverable three-
dimensional ceramic nanolattices, Science 345 (2014) 1322–1326.
[11] X. Zheng et al., Ultralight, ultrastiff mechanical metamaterials, Science 344
(2014) 1373.
[12] S.C. Han, J.W. Lee, K. Kang, A new type of low density material Shellular, Adv.
Mater. 27 (2015) 5506–5511.
[13] A.J. Jacobsen, W. Barvosa-Carter, S. Nutt, Micro-scale truss structures formed
from self-propagating photopolymer waveguides, Adv. Mater. 19 (2007)
3892–3896.
[14] M.G. Lee, J.W. Lee, S.C. Han, K. Kang, Mechanical analyses of the ultralow-
density cellular metal Shellular, Acta Mater. 103 (2016) 595–607.
[15] B.D. Nguyen, J.S. Cho, K. Kang, Optimal design of ‘‘Shellular”, a micro-
architectured material with ultralow density, Mater. Des. 95 (2016) 490–500.
[16] M.G. Lee, G.D. Ko, J. Song, K. Kang, Compressive characteristics of a wire-woven
cellular metal, Mater. Sci. Eng. A 539 (2012) 185–193.
[17] S.C. Han, J.M. Choi, S.Y. Kang, K. Kang, An ultralow density material with
minimal surfaces fabricated based on wire-weaving, MRS 2016, XXV
International materials research congress(IMRC), held in Cancun, Mexico on
August 14–19, 2016, paper No. 178-Oral.
178 B.D. Nguyen et al. / Computational Materials Science 139 (2017) 162–178
[18] E.A. Bae, S.C. Han, T.S. Kim, K. Kang, Mechanical properties of B-Shellular
depending on geometry and fabricated process, MRS 2016, XXV International
materials research congress (IMRC), held in Cancun, Mexico on August 14–19,
2016, paper No. 178-Oral.
[19] L. Velimirovi, G. Radivojevi, D. Kosti, Analysis of hyperbolic paraboloids at
small deformations, Arch. Civil Eng. 1 (1998) 627–636.
[20] M. Farshad, Design and Analysis of Shell Structures, Springer, Netherlands,
1992.
[21] Abaqus Analysis User’s Guide, http://129.97.46.200:2080/v6.13/books/usb/
default.htm.
[22] L. Valdevit, S.W. Godfrey, T.A. Schaedler, A.J. Jacobsen, W.B. Carter, Compressive
strength of hollow microlattices: experimental characterization, modeling and
optimal design, J. Mater. Res. 28 (2013) 2461–2473.
[23] W. Wu, J. Owino, A. Al-Ostaz, L. Cai, Applying periodic boundary conditions in
finite element analysis, 2014 SIMULIA Community Conference, pp. 707–719.
[24] S. Hyun, J.E. Choi, K. Kang, Effects of imperfections on the mechanical
behaviors of a wire-woven bulk Kagome cellular metal subjected to
compression, Comput. Mater. Sci. 46 (2009) 73–82.
[25] J.C. Wallach, L.J. Gibson, Defect sensitivity of a 3D truss material, Scripta Mater.
45 (2001) 639–644.
[26] S. Tan, K. Kang, A Shellular material as tissue engineering scaffold, MRS 2016,
XXV International materials research congress (IMRC), held in Cancun, Mexico
on August 14–19, 2016, paper No. 137-Poster.
[27] T. Bitzer, Honeycomb Technology, Chapman & Hall, London, 1997.
[28] K.J. Kang, Wire-woven cellular metals: the present and future, Prog. Mater. Sci.
69 (2015) 213–307.
[29] L.J. Gibson, M.F. Ashby, Cellular Solids: Structure and Properties, Cambridge
Univ. Press, Cambridge, 1997, p. 192.
[30] N.A. Fleck, V.S. Deshpande, M.F. Ashby, Micro-architectured materials: past,
present, and future, Proc. R. Soc. A 466 (2010) 2495–2516.
[31] Z. Hashin, S. Shtrikman, A variational approach to the elastic behavior of
multiphase minerals, J. Mech. Phys. Solids 11 (1963) 127–140.