1 s2.0 S0020768315002152 Main

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 16

International Journal of Solids and Structures 69–70 (2015) 475–490

Contents lists available at ScienceDirect

International Journal of Solids and Structures


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

Mechanical properties of 3D re-entrant honeycomb auxetic structures


realized via additive manufacturing
Li Yang a,⇑, Ola Harrysson b, Harvey West b, Denis Cormier c
a
Department of Industrial Engineering, University of Louisville, 311 J.B. Speed Hall, Louisville, KY 40292, USA
b
Edward P. Fitts Department of Industrial & Systems Engineering, North Carolina State University, 400 Daniels Hall, 111 Lampe Dr. Raleigh, NC 27695, USA
c
Department of Industrial & Systems Engineering, Rochester Institute of Technology, 81 Lomb Memorial Dr. Rochester, NY 14623, USA

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

Article history: In this work, an analytical model of a 3D re-entrant honeycomb auxetic cellular structure has been estab-
Received 17 October 2014 lished based on both a large deflection beam model and a Timoshenko beam model. Analytical solutions
Received in revised form 26 April 2015 for the modulus, Poisson’s ratios and yield strength of the cellular structure in all principal directions
Available online 16 May 2015
were obtained, which indicate a wide range of mechanical property control via geometrical designs.
The results were compared with experimentation and finite element analysis, and it was verified that
Keywords: the analytical model provides a convenient and relatively accurate method in the prediction of the per-
Auxetic structure
formance for the auxetic cellular structures once the manufacturing related factors are adequately incor-
Negative Poisson’s ratio
Cellular structure
porated into the model. It was also found that the model provides less accurate predictions when
Additive manufacturing higher-order coupling effects such as warp locking becomes significant under lower structural symmetry.
Modeling Ó 2015 Elsevier Ltd. All rights reserved.
Design for additive manufacturing

1. Introduction 2006; Grima et al., 2006; Larsen et al., 1997; Prall and Lakes, 1996;
Theocaris et al., 1997; Williams et al., 2007). In many studies, finite
Auxetic cellular structures exhibit negative Poisson’s ratios so element analysis (FEA) was utilized to provide predictions on the
that, unlike regular cellular structures, they show lateral shrinkage auxetic behavior and other mechanical properties (Alderson et al.,
upon axial compression. This counterintuitive type of structure 2005; Bertoldi et al., 2009; Herakovich, 1984; Lee et al., 1996; Lira
was first introduced by Lakes (1987). Due to their excellent shear et al., 2009; Salit and Weller, 2009; Scarpa et al., 2000, 2004b;
stiffness (Lakes, 1987, 1993; Scarpa and Tomlin, 2000; Scarpa and Smith et aL., 2000; Smith et al., 2002; Theocaris et al., 1997; Yang
Tomlinson, 2000), indentation resistance (Alderson et al., 2009; et al., 2003). However, simulations with full scale 3D cellular models
Lakes, 1987; Lakes and Elms, 1993), high fracture toughness are often computationally prohibitive, making them inefficient for
(Bianchi et al., 2008; Lakes, 1993), high energy dissipation ability establishing comprehensive predictive models. In addition, there
(Bezazi and Scarpa, 2007, 2009; Scarpa et al., 2004b, 2005), and has generally been a lack of agreement between the experimental
unique acoustic absorption abilities (Howell et al., 1996; Scarpa results and the theory, which was largely caused by the stochastic
and Smith, 2004; Scarpa et al., 2004a), auxetic structures possess nature of the auxetic foams used in the experimentation (Alderson
appealing potential for various applications such as sandwich panel et al., 2005; Bezazi and Scarpa, 2007, 2009; Bianchi et al., 2008;
cores, energy and sound dampening structures, radome frames Herakovich, 1984; Howell et al., 1996; Lakes, 1993; Lakes and
(Scarpa and Smith, 2004), aerospace filler foams, and bio-implants. Elms, 1993; Scarpa and Smith, 2004; Scarpa and Tomlin, 2000;
Various efforts have been dedicated to the design and development Scarpa and Tomlinson, 2000; Scarpa et al., 2000, 2004a,b, 2005;
of auxetic structures over the decades, which roughly fall into two Smith et al., 2000).
categories: the modeling of auxetic behaviors and the experimental Due to the mathematical simplicity, analytical modeling with
characterization of auxetic structures. The majority of the published unit cell structures has been adopted by many researchers to
studies involving auxetic structure modeling are primarily focused establish comprehensive and in-depth design knowledge with sev-
on mechanism modeling (Almgren, 1985; Caddock and Evans, eral regular 3D cellular structures (Deshpande et al., 2001; Gibson
1989; Choi and Lakes, 1995; Evans and Caddock, 1989; Grima, and Ashby, 1997; Hutchinson and Fleck, 2006). When the bound-
ary effects are properly taken into account, this approach could
achieve satisfactory accuracies in predicting the mechanical prop-
⇑ Corresponding author. Tel.: +1 502 852 2197. erties of cellular structures (Andrews et al., 2001; Onck et al.,
E-mail address: li.yang.1@louisville.edu (L. Yang).

http://dx.doi.org/10.1016/j.ijsolstr.2015.05.005
0020-7683/Ó 2015 Elsevier Ltd. All rights reserved.
476 L. Yang et al. / International Journal of Solids and Structures 69–70 (2015) 475–490

2001). Furthermore, when boundary effects could be minimized, et al., 2009; Evans et al., 1994; Lee et al., 1996; Yang et al.,
analytical modeling also possesses an advantage over the homog- 2003). Fig. 1(a) shows the 2D re-entrant honeycomb structure.
enization methods due to the closed-form expressions (Gonella Unlike some of the 2D auxetic structures such as chiral honeycomb
and Ruzzene, 2008; Spadoni and Ruzzene, 2012). However, analyt- (Prall and Lakes, 1996) and double-tree lattice (Larsen et al., 1997),
ical modeling based on beam/truss theories with unit cell struc- the re-entrant honeycomb structures can be readily patterned into
tures has provided somewhat limited practical value due to the 3D structures with sufficient unit cell connectivities and auxetic
lack of manufacturing methods in realizing many of the complex behaviors in multiple principal directions. For example,
designs. With the increasingly common utilization of additive Schwerdtfeger et al. (2010) adopted the 2D re-entrant auxetic hon-
manufacturing (AM) in the fabrication of cellular structures, it eycomb structure and designed a 3D auxetic structures with
becomes possible to utilize analytical modeling methods to accu- hexagonal super-lattice pattern, which exhibits negative
rately design for the performance of an auxetic cellular structure Poisson’s ratios in multiple directions.
(Schwerdtfeger et al., 2010). The orthotropic 3D re-entrant honeycomb auxetic structure as
In this paper, a 3D re-entrant honeycomb structure that exhibits shown in Fig. 1(b) was first proposed by Evans et al. (1994) and
auxetic behavior in all three principal directions was modeled in was further investigated analytically in this study. The structure
detail. This auxetic structure exhibits orthotropic mechanical prop- can be represented by the unit cell shown in Fig. 1(c). From the
erties and could be readily patterned into 3D geometries. The same geometry of the unit cell, it is apparent that this type of auxetic
3D structural design was originally modeled via finite element structure exhibits orthotropy with negative Poisson’s ratios in
analysis, which did not establish any analytical model for the all three principal directions. The symmetry of the unit cell
design (Evans et al., 1994). In the previous studies by the authors, dictates that the mechanical properties along the x and y axes
Timoshenko beam theory was adopted for the modeling of are identical. Therefore, only properties in two directions (z and
mechanical properties of this 3D re-entrant auxetic cellular struc- either x or y) need to be analyzed to fully describe the structure’s
ture in one principal direction (the z direction in the current study), behavior.
which was subsequently verified experimentally (Yang et al., Without losing generality, it was assumed in the modeling that
2012b, 2013). However, in order to enable the model to be applied the cross sectional shape of the struts in the auxetic cellular struc-
for various materials and loading cases in multiple directions, a ture is square. There are four primary design parameters for the
multi-directional model that utilizes large deflection beam theory unit cell structure: the length of the vertical struts H, the length
must be established. Although a similar 2D re-entrant auxetic hon- of the re-entrant struts L, the re-entrant angle h, and the thickness
eycomb was previously modeled using large deflection beam the- of the strut cross section t. Fig. 2 illustrates the design parameters
ories, the analyses could not be directly applied to 3D geometries for the structure. Note that the strut cross sectional dimension is
(Levy and Goldfarb, 2006; Wan et al., 2004). In these modeling not shown in Fig. 2.
efforts, the maximum deflection angles of the structures were trea- Several assumptions were made for the modeling of the unit
ted as an input parameter that could be defined rather arbitrarily. cell. Firstly, the unit cell was assumed to be located in an infinite
As a consequence, some of the results presented in these studies structure (i.e. infinite numbers of unit cells in all three directions),
were physically infeasible. Therefore, a different approach must which eliminates boundary effects and also retains maximum
be taken. In the current study, the analytical model was established structural symmetry. Secondly, all of the joints in the structure
based on a large deflection model with maximum deflection angles were considered to be rigid, and the deformation of the structures
fully determined by the geometrical design parameters. The accu- is primarily contributed by bending of the re-entrant struts and the
racies of the large deflection model and Timoshenko small deflec- axial compression of the vertical struts. Thirdly, the torsional
tion model were then compared, and the characteristic behavior of
the re-entrant auxetic structure as functions of geometrical param-
eters was determined. In order to verify the theoretical modeling,
experimental studies were carried out with Ti6Al4V samples fabri-
cated via the electron beam melting (EBM) process. Boundary
effects as well as modeling issues introduced by manufacturing
characteristics were considered during the result comparison in
order to further extend the discussions to design for additive man-
ufacturing (DFAM) of cellular structures.

2. Structural design and simplification

The 2D re-entrant honeycomb was first proposed by Almgren


(1985), and subsequently studied by various researchers (Bertoldi Fig. 2. Design parameters for the re-entrant honeycomb structure.

(a) 2D re-entrant honeycomb (b) 3D lattice (c) Unit cell


Fig. 1. 3D re-entrant honeycomb auxetic structure.
L. Yang et al. / International Journal of Solids and Structures 69–70 (2015) 475–490 477

effects were not considered. In addition, the axial shrinkage of the treated as a cantilever beam that has one fixed end and one free
re-entrant struts due to the normal stress was ignored based on the end as shown in Fig. 3(b).
preliminary observation that its effect on beam deformation is neg- In Fig. 3(b), S is a curvilinear coordinate along the deflected strut
ligible compared to bending and shearing. The Cosserat effect was with the origin of A, u is the coordinate along the original strut also
also ignored following the same argument from previous studies with the origin of A, the angle u is the deflection angle with respect
with a similar approach (Onck et al., 2001). to the original strut alignment direction, and Dv and Du are the
In order to allow for the modeling of auxetic structures with changes in dimensions normal to the longitudinal direction and
less stiff polymer materials that are subject to large deflections along the longitudinal direction, respectively.
upon loading, a large deflection beam model was used for the basis The moment M can be obtained readily from equilibrium as
of modeling. The large deflection model takes nonlinear dimen-  
L
sional change into consideration and is expected to be more accu- M¼P  Du ð1Þ
2
rate in describing the deformation of polymer structures Belendez
et al. (2002). However, the large deflection model often requires The deflection angle is a function of the longitudinal position of
non-analytical algorithms for solutions, and therefore it is less effi- the strut, which is defined as u(s). Since the deflection does not
R
cient for structural design and optimization. Besides, no prior change the length of the strut, i.e. s ds ¼ 2L, according to the
knowledge exists on whether the large deflection model could Euler–Bernoulli theorem for beam bending,
achieve significant improvement with accuracy for metal cellular  
structures fabricated via additive manufacturing (AM) processes. @ u MðuÞ P L
¼ ¼  Du  u ð2Þ
Therefore this issue was investigated in detail in the study. @s EI EI 2
Several researchers employed the large deflection model for the where u is the distance of the calculated position from the ori-
2D re-entrant honeycomb and derived expressions for Poisson’s gin A, and I is the second moment of inertia of the cross-section.
ratios in the two principal directions (Levy and Goldfarb, 2006; The Euler–Bernoulli theorem does not take the shear deflection
Wan et al., 2004). In these efforts, the maximum deflection angles into account, which yields a good approximation for long beams.
were treated as input parameters, and the mathematical relation- In cases where the strut length is short, the shear contributes to
ship between the designs and the Poisson’s ratios were investi- the deflection of the re-entrant strut and must therefore be consid-
gated by considering different amounts of deflection angles. From ered. The deflection angle is then defined as (Hutchinson, 2001):
these results, it was concluded that the Poisson’s ratio of the 2D
re-entrant auxetic structures has significant dependence on the dv dv
u¼ þc¼ þc ð3Þ
strain level. However this conclusion must be subject to more scru- du ds
tiny, since the arbitrary values of deflection angles (and thus the where c is the shear strain, and v is the displacement of the strut
strain) discussed in these studies does not always have physical in the direction perpendicular to its longitudinal axis. For a beam
feasibility. Therefore, in the current study, a similar approach with a square cross section, c can be obtained by:
was taken with the maximum deflection angle determined by both
the stress levels and geometrical parameters, therefore ensuring 6P
c¼ ð4Þ
the physical feasibility of the solutions. Based on the solutions of 5GA
deflections, mechanical properties including Poisson’s ratios, effec- where G is the shear modulus of the solid, and A is the cross sec-
tive modulus and yield strength of the re-entrant auxetic struc- tional area of the re-entrant strut. Upon derivation, Eq. (3) becomes
tures were obtained, which are of primary interest to many  
2
applications. @ u d v MðuÞ P L
¼ 2¼ ¼  Du  u ð5Þ
@s du EI EI 2
Considering the relationship between the A-u and the A-S coor-
3. Deflections of the re-entrant struts under loading dinate systems from Eq. (2), the following geometrical relation-
ships could be readily obtained:
From the rigid joint assumption, the re-entrant struts could be
treated as beams that are fixed at both ends. Since the axial loading du
¼ cos h ð6Þ
is not considered, the effective loading condition is shown in ds
Fig. 3(a) for the re-entrant strut AC. The moment distribution on
dv
the strut is also shown in Fig. 3(a). From the moment distribution, ¼ sin h ð7Þ
it is obvious that the effective moment at the mid-point B of the ds
strut is zero. Therefore, the half strut (e.g. AB) could be further From Eqs. (5) and (6), it then follows:

(a) Loading of the re-entrant strut (b) Cantilever deflection of the half strut
Fig. 3. Loading and deflection of the re-entrant struts.
478 L. Yang et al. / International Journal of Solids and Structures 69–70 (2015) 475–490

@2u P du P Finally, from the results of the deflection angle uL=2 , the total
¼ ¼  cos u ð8Þ
@s2 EI ds EI displacement Dv and Du could be calculated as
rffiffiffiffiffi  rffiffiffiffiffi 
Further manipulation with Eq.(8) yields: EI EI
! ! Dv ¼ FðkÞ  Fða0 ; kÞ  2 EðkÞ  Eða0 ; kÞ ð14Þ
Z P P
@2u @u P @u @2u @u
¼  cos u ) ds rffiffiffiffiffiffiffi
@s2 @s EI @s @s2 @s
L 2EI 1=2
Z  2 Du ¼  sin uL=2  sin c ð15Þ
P @u 1 @u P 2 P
¼  cos u ds ) ¼  sin u þ C ð9Þ
EI @s 2 @s EI where F() and E() are elliptic integrals of the first and second
Considering the boundary condition, when s ¼ L
, define kind, respectively.
2
u ¼ uL=2 . At this location, the curvature of the beam becomes 0. From Du and Dv, the deflections of the beam along the global
coordination Dz and Dx can be obtained, which could then be used
Therefore, from Eq. (9) the constant can be determined as
to calculate the mechanical properties of the structures. The details
C ¼ EIP sin uL=2 , and Eq. (9) can be subsequently re-written as:
are given in the following sections.
rffiffiffiffiffiffi
@u 2P 1
¼ ðsin uL=2  sin uÞ2 ð10Þ 4. Modeling of mechanical properties
@s EI
Also notice that at s = 0, the deflection angle u(0) = c, therefore:
4.1. Uniaxial compression in the z direction
Z uL=2 Z L rffiffiffiffiffiffi rffiffiffiffiffiffi
1 2 2P L 2P
ðsin uL=2  sin uÞ 2 @ u ¼ ds ¼ ð11Þ Consider a remote compressive stress r that is applied to the
c 0 EI 2 EI infinite structure along the z direction. Under remote compressive
Define k and a as: stress, the force components on a unit cell are shown in Fig. 4(a).
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Due to the symmetry, there exists no effective bending, torsional
1 þ sin uL=2 moments or shear forces at the boundaries of the unit cell. Also,
k¼ ð12:1Þ due to the force equilibrium, there are no effective normal forces
2
acting on the boundaries of the unit cell. Because each vertical
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi strut at the corners is shared by four adjacent unit cells, the force
1 þ sin u exerted on each of them is divided evenly between the four unit
sin a ¼ ð12:2Þ cells as shown in Fig. 4(a). From the force equilibrium, the force
1 þ sin uL=2
F is readily obtained as
And Eq. (11) could then be rewritten as:
1
Z rffiffiffiffiffi F¼ rA ¼ 2rL2 sin2 h ð16Þ
p
2 1 L P 2
FðkÞ  Fða0 ; kÞ : pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi da ¼ ð13Þ
2 2 2 EI The unit cell shown in Fig. 4(a) can be further simplified into the
a0 1  k sin a
pffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffi structure shown in Fig. 4(b) due to the high degree of symmetry.
1þsin c
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where sin a0 ¼ p1þsin ¼ 1k 1þsin c
according to Eq. (12). Further decomposition of the force components are shown in
u 2
L=2
Fig. 4(c). From Fig. 4(c) it can be seen that the vertical struts are
Eq. (13) is a variation of an elliptic integral of the first kind in only subject to normal compressive stresses, while the re-entrant
Legendre Form (Good, 2001). Once module angle a and elliptic struts are subject to a combination of normal and shear stresses.
modulus k are determined, the integration results can be readily From equilibrium, the force components could be obtained as
obtained. Eq. (13) only has one independent parameter, which
was selected to be k. The values of k and a0 can be determined F 1 3
P¼ sin h ¼ rL2 sin h ð17Þ
through numerical calculation for given geometrical design param- 4 2
eters. After k is determined, uL=2 can be readily obtained by
Eq. (12), which yields maximum deflection angle values that are F 1 2
T¼ cos h ¼ rL2 sin h cos h ð18Þ
fully determined by the stress level and geometrical parameters. 4 2

(a) Loading condition of a unit (b) Loading of the simplified (c) Force decomposition of
cell structure the simplified structure
Fig. 4. Loading analysis of the unit cell under compressive stress in the z direction.
L. Yang et al. / International Journal of Solids and Structures 69–70 (2015) 475–490 479

used as the material for experimental verification. The comparison


clearly illustrates the effects of the two different parameters on the
Poisson’s ratio values mzx. With the increase of the H/L ratio a and
the decrease of the re-entrant angle h, the negative Poisson’s ratio
becomes greater. It could be observed that the Poisson’s ratio val-
ues become more sensitive to the re-entrant angle h when it is rel-
atively small, which could potentially be taken advantage of during
engineering designs.
From Eq. (14) and (15), the expressions of Du and Dv appear to
be dependent on both the stress r and the modulus of the material
E, therefore their effect on the Poisson’s ratio value must be iden-
tified. From previous research, it was suggested that the
stress/strain level have a significant effect on the Poisson’s ratio
of the re-entrant auxetic structures (Levy and Goldfarb, 2006;
Fig. 5. Relationship between coordinate systems of re-entrant struts under Wan et al., 2004), although the conclusions were drawn with unre-
compressive stress in the z direction. alistic deflection angles. On the other hand, no previous literature
is available to clarify whether the solid material property (modulus
The relationship between the local coordinate system and the E) would have any non-linear effect on the performance of the cel-
global coordinate system for the re-entrant struts is readily shown lular structures. Although not shown in this paper, further investi-
in Fig. 5, which can be written as gations showed that higher stress levels have a more pronounced
effect on the Poisson’s ratios for structures with more slender
Dx ¼ Du sin h þ Dv cos h ð19Þ struts. However, for structures with slender struts, the maximum
achievable stress levels are limited as the overall relative density
Dz1 ¼ Dv sin h  Du cos h ð20Þ of the structure becomes smaller. For practical design purposes,
the stress dependency of the Poisson’s ratio values is therefore
Note that Dx and Dz1 only represent the dimensional change of
much smaller (<10%), and Eq. (23) can be considered to be accu-
half of the re-entrant strut.
rate. Further evaluation also showed that the modulus of the solid
The compression of the vertical struts also contributes to the
material does not have any effect on the structure’s Poisson’s ratio,
dimensional changes in the z direction, and can be written as
which was fully expected as the Poisson’s ratios of a structure
2
2rHL2 sin h should only be a geometrical behavior.
Dz 2 ¼ 2
ð21Þ The Poisson’s ratio value of the re-entrant honeycomb structure
Et
in the z direction under compression was also obtained in a previ-
Note that Dz2 represents the axial compression of one whole ous study based on the small deflection Timoshenko model with
vertical strut. Therefore, the total dimensional reduction in the z shear induced deflection included (Yang et al., 2012b). In the
direction for the simplified structure shown in Fig. 4(b) is then Timoshenko model, the deflection angle is assumed to be suffi-
Dz ¼ 2Dz1 þ Dz2 ð22Þ ciently small so that the deflection induced axial length reduction
can be ignored. The simplified expression of the Poisson’s ratio
under z-axis compression was determined as (Yang et al., 2012b):
4.1.1. Poisson’s ratios  
L2 2
Poisson’s ratio in the z direction under compressive stress can DxðH  L cos hÞ E
þ 6t
5G
cos hða  cos hÞ
be readily obtained as mzx ¼  ¼
L2 sin2 h sin2 h
ð24Þ
DzðL sin hÞ þ 6t
2
þ 4aEt
2
E 5G
2Dx
ex 2ðH  L cos hÞDx 2ða  cos hÞDx
mzx ¼  ¼  L sin h
¼ ¼ Comparing Eq. (24) with Eq. (23), it was found that the large
ez Dz
HL cos h
L sin hð2D z 1 þ D z 2 Þ sin hð2Dz1 þ Dz2 Þ
deflection model and the small deflection Timoshenko model yield
ð23Þ very similar results for Poisson’s ratio mzx, with the magnitudes of
difference being smaller than 102. Although not shown here, it
where a = H/L. Fig. 6 shows a plot of the relationship between
is also found that the large deflection model shows considerable
the design parameters a, h, and the Poisson’s ratio values in the z
discrepancies with the Euler–Bernoulli beam model in the predic-
direction at 10 kPa stress. The solid material property values used
tions of mzx. This could be readily explained by the fact that the
for the calculation are adopted from Ti–6Al–4V, which was later
shear induced deflection has more significant effects on the
Poisson’s ratios of the re-entrant auxetic structure compared to
the axial length reduction.

4.1.2. Effective modulus


During the calculation of the effective modulus, the unit cell
structure was considered a ‘‘block’’, and the modulus values were
calculated for the geometrical bounding box of the unit cell, which
as a dimension of 2(H  Lcosh)  2Lsinh  2Lsinh. The effective
modulus is therefore always lower than the modulus of a solid
block of material having the same overall dimensions. From the
results obtained by Eq. (22), the modulus of the re-entrant honey-
comb structure can be readily determined as
r r rðH  L cos hÞ
Ez ¼ ¼ ¼ ð25Þ
Fig. 6. Poisson’s ratios in the z direction under compression with different designs e HLDcos
z
h
ð2Dz1 þ Dz2 Þ
L = 5 mm, t = 1 mm, r = 10 kPa, E = 114 GPa, G = 43 GPa.
480 L. Yang et al. / International Journal of Solids and Structures 69–70 (2015) 475–490

In order to evaluate the effect of geometry on the effective mod-


ulus, Ez/Es is used in which Ez and Es represent the effective modu-
lus of the auxetic structure in the z direction and the modulus of
the solid material, respectively. The effect of the design parameters
on the effective modulus Ez is shown in Fig. 7. It is clear that the
effective modulus of the structure increases as the H/L ratio
increases and the re-entrant angle h decreases. This conclusion is
intuitive since the reduced bending moment of the re-entrant
struts and the increase of the total dimension in z will result in a
decrease of the strain values.
It is also interesting to observe that the stress level does not
have a significant effect on the effective modulus over the entire
design space. This is slightly different with the observation for
Fig. 8. Loading of a re-entrant strut.
Poisson’s ratios and could be explained straightforwardly, as the
non-linearity of deflection is sufficiently small at realistic stress
levels. Following the calculations presented in the previous work
From the previous work that used the small deflection (Yang et al., 2012b, 2013), the maximum compressive force Fz for
Timoshenko model, the effective modulus of the re-entrant honey- the structure under compression in the z direction is:
comb under compression in the z direction can be determined by  
2 2
Eq. (26) (Yang et al., 2013), which was also verified to result in less r2Y  9F z64tsin4 h
t3 F 2z cos2 hrY F L
than 1% error compared to Eq. (25) from the large deflection model.   ¼ z sin h ð28Þ
4r Y 2 2
64r2Y  9F z tsin h
t 8
4

ða  cos hÞ
Ez ¼  4  ð26Þ The value of Fz could then be obtained by solving for Eq. (28)
2aL2 sin2 h L 3L2 4
Et2
þ 2Et 4
þ 5Gt 2 sin h numerically. The resulting compressive strength rz in the z direc-
tion can be determined by Eq. (29).
Fz
rz ¼ 2
ð29Þ
4.1.3. Compressive strength 2L2 sin h
The strength of the cellular structure is closely related to its
Note that Eq. (29) only considers the failure of the re-entrant
thermal history during the manufacturing process. In previous
struts, while under compression the vertical struts could also
studies by the authors, it was found that the Ti6Al4V auxetic struc-
potentially be subject to elastic failure via buckling. After taking
tures fabricated via electron beam melting (EBM) process exhibit
elastic failure mode into consideration, the compressive failure
rather ‘‘brittle’’ structural behavior (Yang et al., 2012a,b). Due to
strength of the re-entrant auxetic cellular structures in the z direc-
the relatively low ductility of the titanium materials fabricated
tion is shown in Fig. 9 (Yang et al., 2013). As the re-entrant angle h
via EBM process and the catastrophic failure mode during the com-
becomes larger, the strength of the structure also decreases. Also,
pression, a rather conservative yield analysis was used for the pre-
at small h levels, the structure is more prone to elastic buckling
diction of the compressive strength for the re-entrant auxetic
failure when H/L ratio becomes large.
cellular structure. Following the analysis method used by the
authors in the previous work (Yang et al., 2013), under remote
4.2. Uniaxial compression in the y direction
stress r the re-entrant struts are subject to more critical loading
conditions and were therefore analyzed. Under the combined load-
A similar approach was taken for the modeling of uniaxial com-
ing of shear and normal forces as shown in Fig. 8, the principal Von
pression properties of the re-entrant auxetic structure in the y
Mises stress on the strut can be expressed as (Yang et al., 2013):
direction. Define a remote compressive stress r on the re-entrant
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi auxetic structure in the x direction as shown in Fig. 10(a). For the
r r2 2
r1 ¼ þ þs ð27Þ
2 4
where r and s are the normal and shear stresses on the strut,
and r1 is the principal stress, which in turn needs to meet the cri-
teria that r1 < rY, where rY is the yield stress of the solid material.

Fig. 7. Effective modulus Ez/Es under compression with different designs L = 5 mm, Fig. 9. Failure strength in z direction with different designs t = 1 mm, L = 5 mm,
t = 1 mm, E = 114 GPa, G = 43 GPa, r = 10 kPa. E = 114 GPa, G = 43 GPa.
L. Yang et al. / International Journal of Solids and Structures 69–70 (2015) 475–490 481

T2  T1 ¼ 0 ð34Þ
From Eqs. (30)–(34), the force components can be readily solved
as
F
T1 ¼ T2 ¼ cot h ð35Þ
8

F
P1 ¼ P2 ¼ cos h ð36Þ
8
Subsequently, the deflections Du and Dv could be solved in the
same way using Eqs. (14) and (15), and the dimensional changes
for both types of re-entrant struts in the principal directions of
the structure can be obtained as illustrated in Fig. 12:
(a) A max-symmetry unit cell (b) Simplified structure
Dz ¼ Dv sin h  Du cos h
Fig. 10. Compressive loading of re-entrant auxetic structure in the y direction. ð37Þ
Dx ¼ Dy ¼ Du sin h þ Dv cos h

unit cell shown in Fig. 10(a), each re-entrant strut at the edges in
the y direction is shared by two unit cells. Since the stress is 4.2.1. Poisson’s ratios
remote, the compressive force applied on each whole re-entrant For compression in the x direction, there exist two different
strut is F/4 as shown in Fig.10(a). The relationship between the Poisson’s ratios in the two directions. From Eq. (37) the Poisson’s
force and the stress is: ratios of the re-entrant auxetic structure compressed in the x direc-
tions are:
F ¼ 4rðH  L cos hÞðL sin hÞ ð30Þ
ez 2DzðL sin hÞ Dz sin h
Due to the geometrical symmetry, the unit cell could be further myz ¼  ¼ ¼ ð38Þ
ey 2DyðH  L cos hÞ Dyða  cos hÞ
simplified into the structure shown in Fig. 10(b). Due to the equi-
librium and symmetry, the vertical struts, which now appear hor- ex
izontally, do not experience any effective force or moment myx ¼  ¼ 1 ð39Þ
ey
component. It is also noted that re-entrant struts 1 and 2 are not
mechanically symmetrical and, therefore should be considered where a = H/L. The relationship between the Poisson’s ratio
separately. Further decomposition of the simplified structure is myz and the design parameters is shown in Fig. 13. It can be seen
shown in Fig. 11. from Fig. 13 that with an increasing re-entrant angle h and a
The shear forces P1 and P2 (not shown in figures) of each decreasing H/L ratio, the negative Poisson’s ratio in the y direction
re-entrant strut can be determined from force equilibrium as:
F
P1 ¼ cos h  T 1 sin h ð31Þ
4

P2 ¼ T 2 sin h ð32Þ
It is noted that for both types of struts, the boundary conditions
are identical. Since type 1 and type 2 struts need to accommodate
the deflection so that no global torsional effects appear in the sim-
plified structure as is required by the structure’s symmetry, their
deflections need to be identical. Therefore,
F
M1 ¼ M2 ) cos h  T 1 sin h ¼ T 2 sin h ð33Þ
4
Furthermore, the equilibrium of the vertical struts requires that Fig. 12. Dimensional changes of the re-entrant half strut under compressive stress
the effective axial force equals zero: in y direction.

(a) Re-entrant struts type 1 (b) Re-entrant struts type 2


Fig. 13. Poisson’s ratio myz with different designs L = 5 mm, t = 1 mm, G = 114 GPa,
Fig. 11. Decomposition and analysis of the strut components. G = 43 GPa, r = 10 kPa.
482 L. Yang et al. / International Journal of Solids and Structures 69–70 (2015) 475–490

becomes greater. This shows the opposite trend to the Poisson’s r rL sin h
Ey ¼ ¼ ð41Þ
ratios in the z direction as shown in Fig. 6. Also, from the compar- ey 2 Dy
ison of the Poisson’s ratio values shown in Fig. 6, the negative
From Eq. (41), the relationship between the design parameters
Poisson’s ratio values myz spread over a smaller range compared
and the effective modulus is shown in Fig. 14. Again, the dimen-
to the z direction. On the other hand, myx shows a constant value
sionless effective modulus Ey/Es is used to evaluate the effect of
of 1, which is contributed by the symmetric deflections in both
the structure, where Ey and Es stand for the effective modulus of
x and y directions.
the structure and the solid modulus of the material, respectively.
The Poisson’s ratio myz also varies with the change of stress
From Fig. 14, the effective modulus value in the y direction
levels. However, the variation is significantly smaller (not shown)
increases with an increase of the re-entrant angle h, or a decrease
throughout the range of physically feasible stress levels
of the H/L ratio. This again shows an opposite trend to the effective
(<100 MPa). The same conclusion was observed for the calculation
modulus in the z direction. The range of effective modulus values
of effective modulus in the y direction. Combining the results from
in the y direction is considerably smaller than that in the z direc-
both directions, for most engineering design purposes, stress level
tion. Also, from Figs. 6, 7, 13 and 14, an apparent relationship
should not become a concern in the design of the re-entrant aux-
can be shown between the Poisson’s ratio values and the effective
etic structures.
modulus. With an increased value of the negative Poisson’s ratios,
On the other hand, using the small deflection Timoshenko
the effective modulus also increases.
model, the Poisson’s ratio value myz under compression could be
From previous results, it was not surprising that the effective
determined as (Yang et al., 2012b):
modulus in the y direction determined by the Timoshenko model
2 also yielded satisfactory accuracies, which are shown in Eq. (42)
sin h
myz ¼  ð40Þ (Yang et al., 2012). The results give rise to a general conclusion that
cos hða  cos hÞ for metallic re-entrant auxetic structures, the small deflection
model that incorporates bending and shear mechanisms should
A comparison shows that Eq. (40) yields minimum errors com-
be adequate for the design purposes.
pared to the results from large deflection model. It’s worth noting
that Eq. (40) has the same form as the 2D equation obtained by 1 1
Ey ¼   ð42Þ
Wan et al. (2004). This can be understood readily once one consid- L2 cos2 hða  cos hÞ L2 6
Et 4
þ 5Gt 2
ers that under the small deflection assumption, the Poisson’s ratio
values are only dependent on the structural geometry. Since the 3D
structure and the 2D structure have the same 2D unit cell under 4.2.3. Compressive strength
this specified loading condition, their geometrical behaviors are Using the same approach as described previously (Yang et al.,
also expected to be the same. 2013), the maximum compressive force Fy and compressive stress
Gibson and Ashby have previously derived the Poisson’s ratios ry of the re-entrant auxetic structure in the y direction can be
equations for a regular honeycomb structure, which also predicted determined by:
the existence of negative Poisson’s ratios with re-entrant angles.    
2 2
Their equation has a similar form as Eq. (40) except that an addi- r2Y  9F256t
y cos h
4 t3 cos4 h
sin2 h
þ 4 F 2y rY FyL
tional term for axial compression was also included (Gibson and   ¼ cos h ð43Þ
4rY 256rY  t4
2 9F 2y cos2 h
t 16
Ashby, 1997). Due to this additional term, the Poisson’s ratio would
exhibit a non-monotonic trend in relation to the strut angles. The
Poisson’s ratio would achieve maximum absolute value at a small Fy
strut angle (10°). However, considering that at such a low strut ry ¼ ð44Þ
4ðH  L cos hÞðL sin hÞ
angle the auxetic effect in the z direction would also be minimized,
this additional factor was not considered critical in the current Note that since the vertical struts are not subject to effective
study. loading, the structure exhibits only plastic failure when com-
pressed in the y direction. The relationship between the compres-
sive strength ry and the geometrical design is shown in Fig. 15. The
4.2.2. Effective modulus compressive strength ry exhibits a non-monotonic trend in rela-
Since the vertical struts do not contribute to the cell deforma- tion to the re-entrant angle h, and high strength is achieved at both
tion, the effective modulus of the 3D re-entrant auxetic structure very large and very small h values. This is caused by the compound
under compression in the y direction is therefore solely deter- effect of both the effective cross sectional area and the alignment
mined by the re-entrant struts as: of the re-entrant struts along the loading direction. Also, as smaller

Fig. 14. Effective modulus Ey/Es under compression in y with different designs Fig. 15. Failure strength in the y direction t = 1 mm, L = 5 mm, E = 114 GPa,
L = 5 mm, t = 1 mm, E = 114 GPa, r = 1 kPa. G = 43 GPa.
L. Yang et al. / International Journal of Solids and Structures 69–70 (2015) 475–490 483

H/L ratio and h values result in large failure strength in both y and z Table 2
directions, this allows the designer to achieve high strengths for all Design parameters for compression in the y direction.

three principal directions for the re-entrant auxetic structures. Design h H L H/L t Relative density myz
(°) (mm) (mm) (mm) (%)
B1 70 5.5 4.25 1.29 0.9 13.41 2.71
4.3. Re-entrant auxetic cellular structures under tension
B2 70 4.14 3.2 1.29 0.9 22.06 2.71
B3 45 7.8 3.77 2.07 0.9 22.45 0.52
For a beam network structure that is subject to elastic deforma- B4 45 6.9 3.3 2.09 0.9 27.18 0.51
tion, it is commonly known that the elastic properties of the struc-
ture including effective modulus and Poisson’s ratios remain
constant regardless of the loading direction. This allows for the
homogenization of the cellular structures into an equivalent con-
tinuum solid (Gonella and Ruzzene, 2008). In this study, the
mechanical properties of the re-entrant auxetic structure under
tensile stress was also modeled to verify the validity of the results
for compression. Although the equations that describes tension
appeared mathematically different, they yield the same results
for all mechanical properties discussed. This indicated that the
modeling results from compressive study were valid, and that
the re-entrant auxetic structure could be treated as a homogenous
solid when boundary effects are able to be ignored.

5. Model verification
Fig. 16. Relationships between design configurations for compressive testing.
5.1. Experimental design

During the modeling of the re-entrant auxetic structures, it was parameters for each configuration, and Fig. 16 further illustrates
assumed that the sample size was sufficiently large to neglect their relationships. Experimental design sets A and B correspond
boundary effects. This assisted in the simplification of the model- to designs for compression in the z and y directions, respectively.
ing. However in many engineering applications, this assumption The configurations were designed in such a way that the effect of
may not be realistic since the structures always have a rather lim- the Poisson’s ratio and the relative density could be investigated
ited size. With a smaller number of unit cells, the asymmetrical separately. For example, design A1 and A2 have the same
boundary conditions will exert a more significant effect on the per- Poisson’s ratios but different relative density, while A2 and A3 have
formance of the structure by introducing local stress concentra- similar relative densities but very different Poisson’s ratios.
tions, which is often referred to as the size effect. For regular Designs A and B used essentially the same parameter sets for geo-
extruded honeycomb structures, the suggested number of repeti- metrical designs, and the slight differences between them were
tions of the unit cell to avoid a significant size effect is 8 mostly due to the machine limitations at the time of the experi-
(Andrews et al., 2001). It was predicted that for auxetic structures, ments. Therefore for qualitative discussion purposes, A and B can
the size effect becomes less significant when the Poisson’s ratio be treated as the same designs loaded in different orientations.
approaches 1 (Lakes, 1993). In the current work, the size effect All of the samples for the compressive testing were fabricated
of the re-entrant auxetic structure was also investigated using using an Arcam A2 Electron Beam Melting (EBM) system. For
finite element analysis (FEA). designs A1/B1 and A2/B2, the unit cells were patterned into a
During the modeling process, the influence of the manufactur- 4  4  4 cube. For designs A3/B3 and A4/B4, the unit cells were
ing method on lattice material properties was not considered. patterned into a 4  4  3 cube due to the large H/L ratio. The sam-
However, manufacturing processes can have a significant effect ple sizes were restricted so that all samples could be built in the
on the properties of structures (Gibson et al., 2010; Yang et al., EBM in the same batch to minimize the potential manufacturing
2012a,b). With AM processes, manufacturing related issues such variability. Six samples of each design were produced using
as anisotropy, accuracy and surface defect generation could have 100/+325 mesh spherical Ti–6Al–4V powder produced by the
significant effects on mechanical properties, and therefore must plasma rotating electrode process (PREP). Since the EBM process
be incorporated into the design process. Very limited research on produces relatively good isotropic properties (Hrabe et al., 2012),
the design of cellular structures with AM processes has been con- no special treatment was used in the model to accommodate this
ducted. In this study, experimental based studies were performed issue.
in order to provide integral information for the process related In order to observe the auxetic behavior effectively, a different
characteristics of such structures. set of samples were designed to be fabricated with low-modulus
Four configurations were designed for compressive testing in polymer materials as shown in Table 3. For design P1, the unit cells
the z and the x directions. Tables 1 and 2 show the design were patterned into an 8  8  7 cube. For design P2, the unit cells
were patterned into an 8  8  5 cube. These samples were

Table 1
Design parameters for compression in the z direction.
Table 3
Design h H L H/L t Relative density mzx Design parameters for Poisson’s ratio measurement.
(°) (mm) (mm) (mm) (%)
Design h H L H/ t Relative mzx mxz
A1 70 5.50 4.26 1.29 0.8 12.77 0.37
(°) (mm) (mm) L (mm) density (%)
A2 70 4.13 3.19 1.29 0.8 19.96 0.37
A3 45 7.74 3.78 2.05 0.8 23.30 1.90 P1 60 6 4 1.5 1 11.27 0.47 1.50
A4 45 6.46 3.15 2.05 0.8 31.09 1.90 P2 45 9 4.5 2 1 19.38 1.09 0.55
484 L. Yang et al. / International Journal of Solids and Structures 69–70 (2015) 475–490

fabricated using an Objet 30 Polyjet system with the In order to evaluate the size effect for the re-entrant auxetic
VeroWhitePlus™ material, which has low modulus (2–3 GPa) and structure, FEA analysis was employed. Different designs of the
is capable of undergoing relatively large amounts of deformation structure with the same design parameters were modeled with
at low loading levels. The Objet process also produces relatively 3  11 unit cell repetitions in two directions, and 1–10 repetitions
good isotropic properties (Barclift and Williams, 2012), therefore in the third direction. The design parameters for the size effect sim-
enables direct verification of the theories without modifications. ulation are H = 6 mm, L = 3 mm, h = 60°, t = 1 mm. As demonstrated
It was expected that due to the boundary effect, the actual values in Fig. 17, the number of unit cells in the loading direction was 11
of the measured Poisson’s ratios would not match the theoretical to ensure that the size effect and boundary effect in this direction
predictions, however, it was expected that the experiments could could be limited. The three repetitions on one side of the structure
qualitatively verify the auxetic behavior of different designs. ensured that the center layer had the maximum geometrical sym-
metry, therefore minimizing the size effect at this layer. Upon load-
ing, the modulus values were obtained for the unit cell located at
the center of the structure, and the Poisson’s ratio values were
obtained for the middle layer parallel to the loading direction.

5.2. Experimental procedures

The samples for compressive testing were produced using an


Arcam Electron Beam Melting (EBM) machine. The standard pro-
cess parameters for mesh structures embedded with the system
were used. The process layer thickness was 0.07 mm for all the
samples. The dimensions of the EBM fabricated samples were mea-
sured with calipers, and the masses were determined using a dig-
ital scale with a resolution of 0.1 g. Due to the stair stepping effect
and accuracy of the process, the cross section of the struts in the
resulting samples were not strictly square shape. To assess strut
size variability between the different designs, the strut sizes for
both the re-entrant struts (t1) and the vertical struts (t2) were mea-
sured using a digital microscope on randomly chosen samples from
each batch. The values for the vertical struts, t3 and t4, were not
measured separately because of their symmetry in relation to the
EBM build direction. The density of each lattice sample was calcu-
lated by dividing the measured mass by the bounding box volume.
The relative densities were obtained by dividing the lattice block
density by the density of bulk Ti–6Al–4V.
Compressive testing was performed using an Applied Test
Systems 1620C (max. capacity 100 kN) at a constant strain rate
of 1.27 mm/min. A calibrated load cell with 100 kN was used to
measure the compressive force. The samples were placed between
two tool steel plates and compressed until the total strain
exceeded 50%. The compressive strain was obtained from an exten-
Fig. 17. Finite element analysis of the size effect.
someter attached to the two platens, and the stress was obtained

Fig. 18. The setup for FEA simulation of structures.


L. Yang et al. / International Journal of Solids and Structures 69–70 (2015) 475–490 485

Table 5
Strut sizes of the samples.

Batch t1 (mm) t2 (mm)


Compression in z (A) 0.79 ± 0.1 0.87 ± 0.06
Compression in x (B) 1.0 ± 0.07 0.95 ± 0.05
Objet 1.02 ± 0.03 1.02 ± 0.03

Fig. 19. Re-entrant auxetic samples by EBM.

Fig. 20. Reduction of effective length of the re-entrant struts.

by dividing the applied force by the lattice block’s area (i.e.


width  depth).
For all simulations, the Poisson’s ratios were obtained using the
Samples P1 and P2 were compressed using the same setup in
displacement values at the center layer. The effective modulus of
both the x and the z direction in order to evaluate the Poisson’s
each structure was obtained by dividing the stress by the total
ratios in all three directions (i.e. behavior in x and y is identical
strain of the model. Lastly, the strength was obtained by determin-
due to symmetry). For measurements of the Poisson’s ratios, the
ing the minimum stress level under which the yield strength of Ti–
deflection of the center of one side of the sample was measured
6Al–4V (=1050 MPa) is obtained in any strut.
by a deflectometer, and the resulting value was multiplied by 2
to reflect the total lateral dimensional change of the sample.
The FEA simulations were performed with SolidWorks COSMOS. 5.3. Results and discussion
The integration of the CAD software with simulation module
enables the cellular designs in SolidWorks to be analyzed directly. Fig. 19 shows some of the samples produced in the Arcam EBM
Static elastic mechanical analysis was performed for the elastic machine, and Table 4 shows the dimensions of the samples as well
response of the structures, therefore no other special meshing as the equivalent actual design parameters. One polymer sample
treatment of boundary conditions were required. All meshing for each design was fabricated by the Objet30 for the observation
operations were performed with default automated meshing using of auxetic behaviors. Due to the sintering effect on the surface of
tetrahedral units. Preliminary comparison between the experimen- the EBM samples, the strut diameters were expected to have some
tal results and the FEA simulations confirmed the validity of such variability. From Table 4, it can be seen that the dimensional vari-
simplified treatment with simulation setup. Since the actual fabri- ation between the samples was quite small, which indicates good
cated samples had slightly different actual dimensions, the average process stability and thus consistent part quality. The masses of
measured dimensions of fabricated samples were used to update samples P1 and P2 were not measured since that they were not
the CAD model dimensions during simulations in order to improve needed for the Poisson’s ratio measurement. The diameters of the
simulation accuracy. The simulation results, along with the exper- re-entrant (t1) and vertical (t2) struts for the samples are shown
imental results, were then compared with the theoretical calcula- in Table 5.
tions. Fig. 18 shows the boundary conditions for the simulation, One manufacturing related issue was the finite thickness of the
where the structure was fixed at one end, and normal loading actual structures. Due to the thickness of the struts, the effective
was applied to the other end. The two platens were bonded to length of the fabricated re-entrant struts that are subject to bend-
the structure and were set to be rigid in order to simulate the ing is shorter than the designed length, as illustrated in Fig. 20. As
actual experiment. the re-entrant angle becomes smaller, the effective length becomes

Table 4
Parameters of the samples.

Design L1 (mm) L2 (mm) L3 (mm) H (mm) L (mm) Mass (g) Relative density (%)
A1 32.08 ± 0.02 32.08 ± 0.03 31.71 ± 0.06 5.42 ± 0.01 4.27 ± 0.00 19.7 ± 0.3 13.7 ± 0.2
A2 24.18 ± 0.03 24.16 ± 0.04 23.62 ± 0.04 4.05 ± 0.01 3.21 ± 0.00 14.0 ± 0.0 22.9 ± 0.1
A3 21.54 ± 0.03 21.54 ± 0.06 29.57 ± 0.06 7.62 ± 0.01 3.81 ± 0.00 14.3 ± 0.3 21.6 ± 0.4
A4 18.04 ± 0.03 18.02 ± 0.04 24.77 ± 0.20 6.38 ± 0.04 3.19 ± 0.01 11.1 ± 0.2 31.1 ± 0.5
B1 32.86 ± 0.03 32.89 ± 0.00 31.99 ± 0.05 5.49 ± 0.01 4.37 ± 0.00 23.0 ± 0.2 15.1 ± 0.1
B2 25.11 ± 0.06 25.06 ± 0.03 24.07 ± 0.06 4.15 ± 0.01 3.34 ± 0.01 16.3 ± 0.2 24.4 ± 0.3
B3 22.28 ± 0.03 22.31 ± 0.03 30.07 ± 0.02 7.80 ± 0.00 3.94 ± 0.00 15.7 ± 0.1 23.7 ± 0.1
B4 19.68 ± 0.05 19.70 ± 0.04 26.98 ± 0.07 6.96 ± 0.01 3.48 ± 0.00 13.1 ± 0.1 28.2 ± 0.3
P1 56.64 56.85 55.55 6.02 4.10 – –
P2 51.56 51.71 57.84 9.01 4.56 – –
486 L. Yang et al. / International Journal of Solids and Structures 69–70 (2015) 475–490

Table 6 different designs exhibit significantly different Poisson’s ratio val-


Poisson’s ratio values of two designs. ues, as originally estimated. Design P1 exhibits greater negative
Design Measured values mzy values, while design P2 exhibits greater negative myz values.
mzy myz myx It is known that due to the termination of the structures at the
boundaries, the force and moment components on the boundary
P1 0.14 0.57 0.23
P2 0.37 0.09 0.31
struts of a cellular structure cannot be balanced, therefore might
cause additional deformation that affects the measured Poisson’s
ratio values (Andrews et al., 2001; Onck et al., 2001). On the other
hand, the trend of experimental auxetic behavior agrees well with
shorter. Because of the tilted geometry, it is not easy to estimate the theoretical predictions. In many actual applications, the
the effect of the length reduction. In addition, the fillets at the joint Poisson’s ratio values are not of direct concern. Therefore, in the
corners formed by sintering further complicate the issue. following sections, Eqs. (23), (34) and (35) were used to simply
In the current study, the effective length change was addressed help demonstrating the effect of auxetic behavior on mechanical
by assuming that the reduction of the effective length was deter- properties without the loss of validity.
mined by the thickness (t) of the re-entrant strut and the
re-entrant angle (h), as expressed in Eq. (45).
5.3.2. Modulus
t
DL ¼ ð45Þ Fig. 21 compares experimental, FEA, and analytical results for
2 sin h
the effective modulus in the z direction Ez. Again the H/L ratios
In the case of the current batch of samples, t = t2. The treatment used in the calculation took the length reduction of L into account.
of effective length is demonstrated in Fig. 20. The calculated modulus, the FEA modulus and the experimental
results showed reasonably good agreement. Note that due to the
5.3.1. Poisson’s ratio idealized loading conditions assumed in the FEA and the analytical
The Poisson’s ratio values of the samples from experiment were models, some deviations between predicted and experimental
listed in Table 6. From the experimental results, it is clear that the results are to be expected.

Fig. 21. Comparison of effective modulus Ez.

Fig. 22. Comparison of effective modulus Ey.


L. Yang et al. / International Journal of Solids and Structures 69–70 (2015) 475–490 487

Since geometries A2 and A3 have similar relative densities, it experiments and the analytical model can be attributed to the
could be expected that the difference between the modulus values inaccurate estimation of the effective strut length L. From Eq. (4),
stems largely from the Poisson’s ratio values. From Fig. 21, it can be it is apparent that the compressive strength in the x direction is
seen that the effective modulus value Ez of A3 is almost 3 times highly sensitive to the variation of L values. In the case of compres-
that of A2, indicating that by controlling the Poisson’s ratio values, sive strength in the y direction, the error in the estimation of L
it is possible to achieve control of modulus values over a wide becomes significant without the contribution of the deflection
range, as predicted from the analytical models. from the vertical struts.
The good agreement between theory and experimentation Another possible cause is the warp locking effect introduced by
helps to verify the validity of Eq. (26). For preliminary design the torsion. The structure does not have the same degree of sym-
screening purposes, the analytical model allows engineers to much metry in the y direction as it has in the z direction. When the struc-
more quickly select geometric design parameters that will produce ture is not infinite, it was therefore expected that torsional
a desired modulus than running dozens of time consuming FEA moments would exist in the re-entrant struts. It is known that
studies. under a torsional moment, beams with a rectangular
Fig. 22 shows the comparison of the effective modulus Ey. The cross-section will exhibit warping if not restricted. When the
FEA results and the experimental results were quite close, how- warping is locked, the resulting internal stress might significantly
ever, the analytical model deviated more significantly from the affect the stiffness of the beam (Dufort et al., 2001).
experimental results in this case. For designs with a re-entrant Comparing geometries B2 and B3 which have similar relative
angle of 70°, the analytical model overestimated Ey by roughly densities, it is verified that the same conclusion holds for the rela-
50%. For designs with a re-entrant angle of 45°, it underestimated tionship between the Poisson’s ratio values and the effective mod-
Ey by roughly 50%. ulus. With a large auxetic behavior, the effective modulus also
The good agreement between the FEA results and the experi- increases. Furthermore, comparing the effective modulus values
ments suggested that the variation of properties contributed by between the two directions, the modulus Ey is generally larger than
the manufacturing quality could be reasonably neglected for Ey. Ez, which is partly caused by the lack of deflection of the vertical
One possible cause for the large discrepancies between the struts, which acts as a softening mechanism on Ez.

Fig. 23. Comparison of compressive strength rz.

Fig. 24. Comparison of compressive strength ry.


488 L. Yang et al. / International Journal of Solids and Structures 69–70 (2015) 475–490

5.3.3. Compressive strength Comparisons between geometries B2 and B3 also agree with the
A comparison of compressive strength in the z direction is previous discussion. While having similar relative densities, the
shown in Fig. 23. The results from the FEA and the calculations compressive strength of B2 is twice that of B3, corresponding to
agree well with the experimental results. The comparison between the greater negative Poisson’s ratio value. Comparing the compres-
geometries A2 and A3 shows a difference of about 60% between the sive strength of the two directions, the x direction generally exhi-
compressive strengths, again indicated the significant effect of the bits larger compressive strength compared to the z direction,
Poisson’s ratio values on the mechanical properties. As predicted especially when the re-entrant angle h is large. The combination
by the theory, larger Poisson’s ratio values generally correspond of high modulus and high strength makes the designs A1, B1, A2,
to larger strength values. and B2 ideal candidates for applications where high uniaxial
Fig. 24 shows a comparison of compressive strength in the y mechanical performance is required. Designs A3, B3, A4, and B4
direction. The FEA values are slightly different compared to the exhibit more homogeneous properties in all principal directions,
experimental results. While exhibiting the same trend as the and are therefore more suitable for applications where the struc-
experimental results, the theoretical predictions show some devi- tures are subject to a combination of axial and shear loadings.
ations. Again, the error introduced by Eq. (45), and especially the
locked warping on the re-entrant struts were expected to con-
5.3.4. Size effect
tribute to the deviation. As illustrated in Fig. 25, when a beam with
The experiments were carried out on designs with a repetition
a rectangular cross section is subject to a torsional moment, the
of four unit cells in order to target cellular material samples that
corners of the cross section tend to warp outwards along the axial
would have an overall thicknesses in the range of 25 mm. While
direction. In the cases where the warping is locked, it is expected
this target size was chosen based upon personal experience with
that the corner sections will be subject to additional compressive
products designed to include cellular materials, the four unit cell
stresses, while the rest of the peripheral sections will be subject
repetitions are fewer than the recommended value for regular cel-
to additional tensile stresses. It is known that for rectangular
lular structures (Andrews et al., 2001; Dai and Zhang, 2009) and
beams in bending, the maximum shear stress occurs at the middle
honeycomb lattices. Therefore it was necessary to evaluate the
plane of the beam, where the lock-introduced stress is tensile
effect of the number of unit cells on the structural properties.
stress, in opposite direction to the primary normal stress.
From the FEA analysis, the effective modulus values and the
Therefore, the locking of the warping could act as a strengthening
Poisson’s ratio values as a function of the numbers of unit cell rep-
mechanism in the case of compression.
etitions are shown in Figs. 26 and 27 respectively. From both fig-
ures, it can be clearly seen that for this re-entrant lattice auxetic
structure, the size effect is rather small. The modulus values Ey
and Ez are nearly constant once the number of repetitions is greater
than or equal to 3, and the Poisson’s ratio values myz are nearly con-
stant when the number of repetitions is greater than or equal to 2.
The Poisson’s ratio in the z direction mzy is slightly more sensitive to
the size effect, and the values level off when the number of unit
cells reaches 4. In all cases, the size effect on the samples evaluated
in this study could be reasonably neglected. This conclusion is also
supported by the theoretical prediction for auxetic structures
which indicate that the size effect is expected to be much smaller
compared to regular cellular structures. This conclusion has signif-
icant engineering implications, since the number of unit cells in
many practical applications is restricted as mentioned above. For
instance, the number of unit cells that will fit in the thickness of
a body panel or fuselage structure is limited, and smaller size effect
could enable more accurate design of such structures with cellular
Fig. 25. Internal stress from the locking of torsion warping in a rectangular beam.
cores.

Fig. 26. Size effect for effective modulus. Fig. 27. Size effect for Poisson’s ratio values.
L. Yang et al. / International Journal of Solids and Structures 69–70 (2015) 475–490 489

6. Conclusions Barclift, M.W., Williams, C.B., 2012. Examining variability in the mechanical
proeprties of parts manufactured via polyjet direct 3D printing. In:
Proceedings of 23th Solid Freeform Fabrication Symposium, Austin, TX, USA.
Through analytical modeling utilizing large deflection theory, Belendez, T., Neipp, C., Belendez, A., 2002. Large and small deflections of a cantilever
various mechanical properties of the 3D re-entrant honeycomb beam. Euro. J. Phys. 23, 371–379.
Bertoldi, K., Reis, P.M., Willshaw, S., Mullin, T., 2009. Negative Poisson’s ratio
auxetic structure were established as functions of material proper-
behavior induced by an elastic instability. Adv. Mat. 21, 1–6.
ties and geometrical design parameters. The characteristic strut Bezazi, A., Scarpa, F., 2007. Mechanical behaviour of conventional and negative
ratio H/L and re-entrant angle h both have significant effects on Poisson’s ratio thermoplastic polyurethane foams under compressive cyclic
loading. Int. J. Fatigue 29, 922–930.
the mechanical properties of the re-entrant honeycomb auxetic
Bezazi, A., Scarpa, F., 2009. Tensile fatigue of conventional and negative Poisson’s
structure, and the relationships between these properties and the ratio open-cell PU foams. Int. J. Fatigue 31, 488–494.
geometrical parameters could be fully determined by design Bianchi, M., Scarpa, F.L., Smith, C.W., 2008. Stiffness and energy dissipation in
equations. polyurethane auxetic foams. J. Mater. Sci. 43, 5851–5860.
Caddock, B.D., Evans, K.E., 1989. Microporous materials with negative Poisson’s
The small deflection Timoshenko model yields results with ade- ratios: I. Microstructure and mechanical properties. J. Phys. D Appl. Phys. 22,
quate accuracy compared to the large deflection model for bending 1877–1882.
dominated auxetic structures under physically realistic stress Choi, J.B., Lakes, R.S., 1995. Analysis of elastic modulus of conventional foam and of
re-entrant foam materials with a negative Poisson’s ratio. Int. J. Mech. Sci. 37,
levels, which indicates that the effect of axial extension/contrac- 51–59.
tion due to bending is negligible compared to the shear and bend- Dai, G., Zhang, W., 2009. Size effects of effective Young’s modulus for periodic
ing effects. Therefore, for many future application purposes, the cellular materials. Sci. China Ser. G: Phys. Mech. Astron. 52, 1262–1270.
Deshpande, V.S., Fleck, N.A., Ashby, M.F., 2001. Effective properties of octet-truss
Timoshenko small deflection model could be conveniently used lattice material. J. Mech. Struct. 49, 1747–1769.
for the computationally efficient determination of the structural Dufort, L., Grediac, M., Surrel, Y., 2001. Experimental evidence of the cross-section
properties. warping in short composite beams under three point bending. Compos. Struct.
51, 37–47.
Using both experiments and simulations, the mechanical prop-
Evans, K.E., Caddock, B.D., 1989. Microporous materials with negative Poisson’s
erties of the re-entrant honeycomb auxetic cellular structure were ratios: II. Mechanisms and interpretation. J. Phys. D Appl. Phys. 22, 1883–1887.
verified. Four configurations for testing the compressive properties Evans, K.E., Nkansah, M.A., Hutchinson, I.J., 1994. Auxetic foams: modelling negative
Poisson’s ratios. Acta Metall. Mater. 42, 1289–1294.
in the x and the z directions were designed and manufactured
Gibson, L.J., Ashby, M.F., 1997. Cellular Solids: Structure and Properties, second ed.
using Arcam’s EBM process, and two configurations for the Cambridge University Press, New York, USA.
Poisson’s ratio measurements were designed and manufactured Gibson, I., Rosen, D.W., Stucker, B., 2010. Additive manufacturing technologies:
using Objet’s Polyjet process. The Poisson’s ratios of the structures rapid prototyping to direct digital manufacturing. Springer, New York, USA.
Gonella, S., Ruzzene, M., 2008. Homogenization and equivalent in-plane properties
showed the same trends as predicted. For qualitative guidance, the of two-dimensional periodic lattices. Int. J. Solids Struct. 45, 2897–2915.
theoretical calculations can help determine the trend of the Good, R.H., 2001. Elliptic integrals, the forgotten functions. Euro. J. Phys. 22, 119–126.
Poisson’s ratio values as a quick design guideline. A simple equa- Grima, J.N., 2006. Auxetic behavior from rotating triangles. J. Mater. Sci. 41, 3193–
3196.
tion provides an estimation of the length reduction of the Grima, J.N., Gatt, R., Ravirala, N., Alderson, Andrew, Evans, K.E., 2006. Negative
re-entrant struts due to the thickness. The estimation is relatively Poisson’s ratios in cellular foam materials. Mater. Sci. Eng. A 423, 214–218.
accurate when h is relatively large, however, it becomes less accu- Herakovich, C.T., 1984. Composite laminates with negative through-the-thickness
Poisson’s ratios. J. Compos. Mater. 18, 447–455.
rate at small re-entrant angles. With length reduction adjustment, Howell, B., Prendergast, P., Hansen, L., 1996. Examination of acoustic behavior of
the theoretical results of the strength and modulus in the z direc- negative Poisson’s ratio materials. Appl. Acoust. 43, 141–148.
tion agree very well with the experimental results as well as the Hrabe, N., Kircher, R., Quinn, T., 2012. Effects of processing on microstructure and
mechanical properties of Ti–6Al–4V fabricated using electron beam melting
FEA simulations, while the theoretical results in the x direction
(EBM): orientation and location. In: Proceedings of 23th Solid Freeform
showed deviations from the experiments and FEA simulations. Fabrication Symposium, Austin, TX, USA.
This was suggested to be due to the error in the estimation of Hutchinson, J.R., 2001. Shear coefficient for timoshenko beam theory. J. Appl. Mech.
68, 87–92.
the length reduction of the re-entrant struts, as well as the tor-
Hutchinson, R.G., Fleck, N.A., 2006. The structural performance of the periodic truss.
sional moments introduced by the asymmetrical geometry of the J. Mech. Phys. Solids 5, 756–782.
loading. The warping of the struts caused by the torsional Lakes, R., 1987. Foam structures with a negative Poisson’s ratios. Science 235, 1038–
moments can introduce additional deformation as well as internal 1040.
Lakes, R.S., 1993. Design considerations for materials with negative Poisson’s ratios.
stresses on the structure, and therefore significantly affect the J. Mech. Des. 115, 696–700.
mechanical properties. These issues are associated with the manu- Lakes, R.S., Elms, K., 1993. Indentability of conventional and negative Poisson’s ratio
facturing process as well as the difference between ideal mathe- foams. J. Compos. Mater. 27, 1193–1202.
Larsen, U.D., Sigmund, O., Bouwstra, S., 1997. Design and fabrication of compliant
matical model and physical reality, and must be incorporated micromechanisms and structures with negative Poisson’s ratio. Int. J. Mech. Sci.
into the design model. More works are required in this area in 6, 99–106.
the future in order to provide improved design predictions. Lee, J., Choi, J.B., Choi, K., 1996. Application of homogenization FEM analysis to
regular and re-entrant honeycomb structures. J. Mater. Sci. 31, 4105–4110.
As expected, this re-entrant auxetic structure has relatively lit- Levy, O., Goldfarb, I., 2006. Design considerations for negative Poisson ratio
tle sensitivity to the number of unit cells. When the number of unit structures under large deflection for MEMS applications. Smart Mater. Struct.
cells is greater than or equal to 3, the size effect can effectively be 15, 1459–1466.
Lira, C., Innocenti, P., Scarpa, F., 2009. Transverse elastic shear of auxetic multi re-
neglected. This is particularly useful for many practical design entrant honeycombs. Compos. Struct. 90, 314–322.
applications in which the number of unit cells is restricted due Onck, P.R., Andrews, E.W., Gibson, L.J., 2001. Size effects in ductile cellular solids.
to the overall component size. Part I: modeling. Int. J. Mech. Sci. 43, 681–699.
Prall, D., Lakes, R.S., 1996. Properties of a chiral honeycomb with a Poisson’s ratio -1.
Int. J. Mech. Sci. 39, 305–314.
Salit, V., Weller, T., 2009. On the feasibility of introducing auxetic behavior into
References thin-walled structures. Acta. Mater. 57, 125–135.
Scarpa, F., Bullough, W.A., Lumley, P., 2004a. Trends in acoustic properties of iron
particles seeded auxetic polyurethane foam. Proc. Inst. Mech. Eng. Part C J.
Alderson, A., Rasburn, J., Evans, K.E., 2005. Mass transport properties of auxetic
Mech. Eng. Sci. 218, 241–244.
(negative Poisson’s ratio) foams. Phys. Status Solidi (B) 242, 509–518.
Scarpa, F., Ciffo, L.G., Yates, J.R., 2004b. Dynamic properties of high structural
Alderson, K.L., Simkins, V.R., Coenen, V.L., Davies, P.J., Alderson, A., Evans, K.E., 2009.
integrity auxetic open-cell foam. Smart Mater. Struct. 13, 49–56.
How to make auxetic fibre reinforced composites. Phys. Status. Solidi (B) 57,
Scarpa, F., Panayiotou, P., Tomlinson, G., 2000. Numerical and experimental uniaxial
1865–1874.
loading on in-plane auxetic honeycombs. J. Strain Anal. 35, 383–388.
Almgren, R.F., 1985. An isotropic three-dimensional structure with Poisson’s ratio =
Scarpa, F., Pastorino, P., Garelli, A., Patsias, S., Ruzzene, M., 2005. Auxetic compliant
-1. J. Elast. 15, 427–430.
flexible PU foams: static and dynamic properties. Phys. Status. Solidi (B) 242,
Andrews, E.W., Gioux, G., Onck, P., Gibson, L.J., 2001. Size effects in ductile cellular
681–694.
solids. Part II: experimental results. Int. J. Mech. Sci. 43, 701–713.
490 L. Yang et al. / International Journal of Solids and Structures 69–70 (2015) 475–490

Scarpa, F., Smith, F.C., 2004. Passive and MR fluid-coated auxetic PU foam – Wan, H., Ohtaki, H., Kotosaka, S., Hu, G., 2004. A study of negative Poisson’s ratios in
mechanical, acoustic, and electromagnetic properties. J. Int. Mater. Syst. Struct. auxetic honeycombs based on a large deflection model. Euro. J. Mech. A/Solids
15, 973–979. 23, 95–106.
Scarpa, F., Tomlin, P.J., 2000. On the transverse shear modulus of negative Poisson’s Williams, J.J., Smith, C.W., Evans, K.E., Lethbridge, Z.A.D., Walton, R.I., 2007. An
ratio honeycomb structures. Fatigue Fract. Eng. Mater. Struct. 23, 717–720. analytical model for producing negative Poisson’s ratios and its application in
Scarpa, F., Tomlinson, G., 2000. Theoretical characteristics of the vibration of explaining off-axis elastic properties of the NAT-type zeolites. Acta. Mater. 55,
sandwich plates with in-plane negative Poisson’s ratio values. J. Sound Vib. 230, 5697–5707.
45–67. Yang, D.U., Lee, S., Huang, F.Y., 2003. Geometric effects on micropolar elastic
Schwerdtfeger, J., Heinl, P., Singer, R.F., Korner, C., 2010. Auxetic cellular structures honeycomb structure with negative Poisson’s ratio using the finite element
through selective electron-beam melting. Phys. Status Solidi (B) 247, 269–272. method. Finite Element Anal. Des. 39, 187–205.
Smith, F.C., Scarpa, F., Burriesci, G., 2002. Simultaneous optimization of the Yang, L., Cormier, D., West, H., Harrysson, O., Knowlson, K., 2012a. Non-stochastic
electromagnetic and mechanical properties of honeycomb materials. Proc. Ti–6Al–4V foam structures with negative Poisson’s ratio. Mater. Sci. Eng. A 558,
SPIE Int. Soc. Opt. Eng. 4701, 582–591. 579–585.
Smith, F.C., Scarpa, F., Chambers, B., 2000. The electromagnetic properties of re- Yang, L., Harrysson, O., Cormier, D., West, H., 2013. Modeling of the uniaxial
entrant dielectric honeycombs. Smart Mater. Struct. 10, 451–453. compression of a 3D periodic re-entrant honeycomb structure. J. Mater. Sci. 48,
Spadoni, A., Ruzzene, M., 2012. Elasto-static micropolar behavior of a chiral auxetic 1413–1422.
lattice. J. Mech. Phys. Solids 60, 156–171. Yang, L., Harrysson, O., West, H., Cormier, D., 2012b. Compressive properties of Ti–
Theocaris, P.S., Stavroulakis, G.E., Panagiotopoulos, P.D., 1997. Negative Poisson’s 6Al–4V auxetic mesh structures made by electron beam melting. Acta. Mater.
ratios in composites with star-shaped inclusions: a numerical homogenization 60, 3370–3379.
approach. Arch. Appl. Mech. 67, 274–286.

You might also like