International Journal of Impact Engineering 169 (2022) 104315

Contents lists available at ScienceDirect

International Journal of Impact Engineering

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

Effects of geometry and boundary constraint on the stiffness and negative

Poisson’s ratio behaviour of auxetic metamaterials under quasi-static and
impact loading
Christopher Mercer a, *, Thomas Speck b, d, Junyi Lee c, Daniel S. Balint c, Marc Thielen d, *
a
Surface and Adhesion Science Group, Research Center for Structural Materials, National Institute for Materials Science, Tsukuba, Japan
b
Cluster of Excellence livMatS @ FIT – Freiburg Center for Interactive Materials and Bioinspired Technologies, Freiburg, Germany
c
Department of Mechanical Engineering, Imperial College London, London, United Kingdom
d
Plant Biomechanics Group, @ Botanic Garden, Faculty of Biology, University of Freiburg, Freiburg, Germany

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

Keywords: Auxetic metamaterials exhibit a negative Poisson’s ratio and superior energy dissipation characteristics under
Auxetics dynamic (impact) loading. This makes them attractive candidates for components that provide crash or impact
Metamaterial protection (such as crumple zones) in automotive and aerospace applications. However, little prior work on the
Impact testing
effect of boundary constraint on the behaviour of these structures exists. Therefore, the primary goal of this
Selective laser melting
Re-entrant unit cell
investigation is to assess the effect of boundary constraint on the energy-absorbing properties of auxetic meta­
3-D printing materials. Specifically, auxetic metamaterials (lattice structures) were fabricated from stainless steel powders via
3-D laser printing (selective laser melting). Additional structures were printed with constraining walls attached
in order to evaluate the effects of boundary constraint on the auxetic behaviour. A comprehensive investigation
of the mechanical performance of the structures was then carried out under both quasi-static and dynamic
(impact) loading conditions. Poisson’s ratio and elastic modulus values were determined from the experimental
results and compared with theoretical calculations. Dissipation of vertical forces into the lateral boundaries was
confirmed by about 20% less negative Poisson’s ratios, for otherwise identical auxetic structures. In addition,
dynamic loading simulations have shown that with optimized geometric parameters, auxetic metamaterials are
capable absorbing over 90% of the energy of an impact. However, critical overestimation of the energy absorbing
abilities of auxetic crash protectors could occur if the effects of lateral constraint are not carefully considered.

1. Introduction compressing and then heating them [4,5]. Nowadays auxetic structures
are often produced as lattices using additive manufacturing techniques.
Auxetic structures and materials are characterised by the fact that Lattice structures from all kinds of unit cells have been realized, the one
they have a negative Poisson ratio. This means that when compressed, most frequently used (and simplest) is probably the so-called re-entrant
they contract perpendicularly to the direction of compression which is in honeycomb cell, as described by Almgren [6] in 2D, as seen in Fig. 1.
contrast to most other materials that either expand laterally under Grima et al. [7] successfully demonstrated via an analytical model that a
pressure or do not deform laterally, i.e. have a Poisson’s ratio of (almost) similar 3D structure always exhibits negative Poisson’s ratio behaviour
0, such as cork. Only a very few natural auxetic structures/materials are when the geometry in that plane is re-entrant. “Re-entrant” simply
known, for instance silicon dioxide (SiO2) in the α-cristobalite structure means that the angles formed by the diagonal struts in the unit cell point
[1] or the peel of the pomelo (Citrus maxima) [2,3]. Hence, most existing inwards, as shown in Fig. 1 (as opposed to outwards in a conventional
auxetic structures are so-called mechanical metamaterials, i.e. artificial hexagonal unit cell).
structures or materials that have been designed and engineered. Prob­ A particularly interesting and promising area of application is
ably the first method developed to produce auxetic structures consists in penetration and impact protection, for which such structures with a
taking conventional open-pored foams as precursor, triaxially negative Poisson’s ratio offer some advantages [8–10]: whereas in

* Corresponding authors.
E-mail addresses: [email protected] (C. Mercer), [email protected] (M. Thielen).

https://doi.org/10.1016/j.ijimpeng.2022.104315
Received 6 October 2021; Received in revised form 2 June 2022; Accepted 29 June 2022
Available online 2 July 2022
0734-743X/© 2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
C. Mercer et al. International Journal of Impact Engineering 169 (2022) 104315

conventional structures (e.g. open pored foams) a uniaxial densification the lateral forces. In order to establish a fundamental understanding of
occurs during compression, auxetic structures are densified triaxially the relative effects and importance of these factors, it was necessary to
[11]. This results in a densified region of material directly beneath an employ a well-known auxetic geometry in this study.
indenter, providing resistance to further penetration or compressive In addition to the experimental study, a numerical model was
deformation. General disadvantages of auxetic foams are that they are developed to determine theoretical values of modulus and Poisson’s
less stiff [11,12] and have a higher density [11] and thus higher weight ratio for the different lattice geometries and loading conditions
as compared to their non-auxetic counterparts. The exact properties employed in the study. The calculated and experimentally measured
however strongly depend on the unit cell’s topology [13], for example values were then compared. Also, finite element simulations of the
Young’s modulus decreases as the angle between the horizontal and impact experiments were carried out to determine the amount of energy
vertical struts (the re-entrant angle) is decreased (Fig. 1) [14]. The the structures are theoretically capable of dissipating during impact, and
comparatively low stiffness can be addressed by combing auxetic lattices again, compared with experimental results. Finally, homogenization/
with hardly compressible soft materials into a composite [15], which optimization studies, similar to those performed previously on 3-D low
however adds weight and increases the density even further. The thermal expansion lattices [25], were performed to assess the effect of
mentioned re-entrant honeycomb structures have often been investi­ varying the strut/node thickness and re-entrant angle on the modulus,
gated as 2D [16,17] or pseudo 3D [15,18,19] structures so far and/or Poisson’s ratio and energy dissipation ability of these auxetic structures.
whole area compression has often been applied for testing their me­ The energy dissipation characteristics were determined by calculating
chanical properties [20], which is well able to reveal the disadvantages the kinetic energy of the impactor before and after the simulated im­
of auxetic structures, however might obfuscate some of their advan­ pacts. From the results, an optimized geometry with maximized aux­
tages. Less frequently, real 3D structures were examined [19–23]. When etic/energy dissipating performance can be realized.
the whole sample is being compressed, no triaxial densification can take
place, since the overall material density (defined as mass of material per 2. Materials and methods
projected area of the compression plate) will not change, and thus only a
reduced stiffness is assessed. This means that one of the main advantages 2.1. Design and fabrication of auxetic metamaterials
of auxetic materials, which is to shunt away impact forces and energy
laterally into surrounding areas [24], cannot be fully realized in this way Inspired by some of the first descriptions of auxetic lattices and unit
due to a lack of non-densified neighbouring material. cells [4–6], 3-dimensional re-entrant unit cells were designed using the
In addition, the energy dissipating properties of these types of CAD software Rhinoceros 5 (Robert McNeel & Associates, Seattle, WA,
structures may be severely compromised by the presence of boundary USA). The structures feature a 3 × 3 × 3 array of auxetic unit cells. It
constraint. In service, the structures will almost certainly be constrained should be noted that with such a relatively small number of cells
in some way by the surrounding component to which they are attached. size/boundary effects may be a factor when boundary constraint is
Yet very little work has been reported in the literature regarding the present. However, the number of cells was primarily limited by the size
effect of boundary constraint on auxetic metamaterials. scale of the impact testing apparatus. If more cells had been used, they
Therefore, the overarching purpose of this investigation is to study would of necessity have been smaller. The consequences of this would be
the effect of thick constraining walls on the mechanical performance and that (i) the struts may have been too thin or (ii) the loading configura­
energy-absorbing properties of a typical 3-D printed, metallic, re-entrant tion may not have been bending dominated. Two versions were
metamaterial under both whole area compression and localized densi­ designed: An auxetic unit-cell similar to those in e.g. [26], from which
fication utilizing quasi-static and dynamic (impact) loading conditions. lattices in the strict definition of the term can be assembled (Fig. 2a), and
Differences in elastic modulus and Poisson’s ratio measured for the two a unit-cell with the central strut removed (not to be confused with the
loading cases are then critically compared. This work shows for the first missing rib foam as described by Smith et al. (2000) [27]) in order to
time: (i) a detailed comparison of whole area versus localized save weight and ease the removal of support material after printing
compression, (ii) the effect of removing a central strut from the auxetic (Fig. 2b). Drawings were also produced with 5 mm thick walls on the
lattice unit cell (in order to reduce weight and facilitate easier support bottom and the lateral sides of the structure to simulate the effects of
removal following 3-D printing), and most importantly (iii) how the constraining components and structures to which the energy-absorbing
thick constraining walls, which were added to the bottom and sides of auxetic would be attached in service (Fig. 2c-e). The geometric param­
some of the auxetic specimens, affect the auxetic response by absorbing eters used in the design are shown in Fig. 1. In this study, H = 16.2 mm,

Fig. 1. Schematic illustration of a re-entrant unit cell showing the parameters used in the design of the three-dimensional unit cell.

2
C. Mercer et al. International Journal of Impact Engineering 169 (2022) 104315

Fig. 2. CAD drawing of an auxetic – re-entrant – unit cell with a central strut highlighted in red (a) and without central strut (b). CAD drawing of the final auxetic
structure with constraining walls. The outer horizontal struts are shared by adjacent cells (highlighted in green) (c). Laser printed auxetic metamaterial with 5 mm
thick constraining walls prior (d) and after (e) removing of the support material.

W = 11.6 mm, t = 1.2 mm and θ = 47◦ . testing machine (Shimadzu Corporation, Kyoto, Japan). The experi­
A Concept Laser Mlab cusing R 3-D laser printer (Concept Laser ments were performed under displacement control using a rate of 0.2
GmbH, Lichtenfels, Germany) was then used to print the metamaterials mm/min and the load was measured. A Nikon D5300 digital camera
via selective laser melting from the stainless steel powder CL 92PH (GE (Nikon Corporation, Tokyo, Japan) was used to record high definition
Additive, Cincinnati, OH, USA). Stainless steel was chosen due to its ease movies of the tests. Images from these movies were then used to measure
of fabrication using additive manufacturing technology and its low Poisson’s ratio values of the structures. Experiments were conducted in
susceptibility to warping and cracking caused by residual stress, as is the both the x/y (short) axis and the z (long) axis orientations under (i)
case with some other materials. However, if materials susceptible to “whole area” compression, for unconstrained specimens, (wherein the
residual stress are used, then post-fabrication heat treatments can be upper loading platen was larger than the specimen so that the whole
performed to relieve the stress. The 3-D printing parameters are sum­ specimen was compressed), and (ii) “central area” compression which
marized in Table 1. It is important to note that the printing parameters used a smaller upper loading platen with a diameter equal to the width
were kept constant for all printed samples. The support struts were then (short axis) of the specimen (35.8 mm).
removed manually from the printed structures using wire cutters and the Since auxetic materials are considered to have great potential for
parts were grit blasted and shot peened to improve the surface finish. applications in the field of impact protection, they have also been sub­
jected to dynamic tests. Impact testing was done using a custom build
2.2. Quasi-static and dynamic (impact) experiments setup as described in [28,29], consisting of an instrumented anvil, on
which the samples were placed, a guide tube, and a 0.5 kg cylindrical,
In order to obtain an understanding of the stiffness and Poisson’s flat-ended brass drop weight (Fig. 3). The latter was dropped onto the
ratio values of the printed metamaterials, quasi-static compression ex­ samples from 2.0 m height. The sampling rate of the 20 kN force sensor
periments were performed using a Shimadzu AG-X series servo-electric (model 8402–6020, Burster Präzisionsmesstechnik GmbH & Co KG,
Gernsbach, Germany) mounted under the anvil in order to measure the
Table 1 force transmitted through the sample was set to 100 kHz. In order to be
Laser printing parameters. able to compare unconstrained and constrained samples, all samples
Parameter Value
were fixed to the anvil using Loctite 454 cyanoacrylate adhesive (Henkel
AG & Co. KGaA, Düsseldorf, Germany). All samples were submitted to
Laser power 95 W
multiple subsequent impacts, which were recorded using a high-speed
Laser movement speed 90 mm/s
Spot size 30 μm camera (model MotionPro Y4, Integrated Design Tools, Inc., Talla­
Slice thickness 25 μm hassee, FL, USA). In order to avoid glare, samples were sprayed with an

3
C. Mercer et al. International Journal of Impact Engineering 169 (2022) 104315

Fig. 3. Drop weight test rig (a). Images from high-speed videos of the constrained sample consisting of unit cells without central struts after the 4th, 9th, 14th and
19th impact respectively (b – f).

anti-glare-spray (3-D Laser Scanning Antireflection Spray, Helling metamaterial. However, due to the repeating nature of the lattice
GmbH, Heidgraben, Germany). structure, when sufficient amounts of unit cells are present, the entire
Poisson’s ratio values were determined from both quasi-static and structure can behave similarly to bulk material. Thus, many homoge­
dynamic experiments by measuring the maximum displacements in the nization methods have been developed in order to determine Young’s
y- and z-directions at the mid-points of the specimen. The axial and modulus and Poisson’s ratio of lattice structures, for example [30–33].
lateral strains (εy and εz), and hence the Poisson’s ratio νyz could then be In this paper, the method by Vigliotti and Pasini [33] was used to
calculated. determine the stiffness matrix, K, which was then used to calculate
Young’s modulus.
3. Theoretical modelling This method is a multiscale method. Firstly, the unit cell of the
auxetic metamaterial is generated, and the boundary struts and nodes
3.1. Analytical and numerical modelling are identified, as seen in Fig. 4. The boundary nodes are then set to have
periodic displacements, this gives rise to a corresponding macroscale
Analytical calculations and finite element simulations were used to displacement of the entire lattice. Then, the microscale stresses that are
study the effects of varying the parameters on the mechanical properties a result of the macroscale displacements are calculated. The microscale
of the metamaterial. In this study, the mechanical properties that were deformation of each strut is then determined by solving the analytical
investigated are Young’s modulus, Poisson’s ratio and the energy Timoshenko beam solution for each individual beam in the unit cell.
absorbed by the lattice structure under impact. In this work, the After that, the macroscale stresses are then calculated by finding the
analytical and numerical methods will first be validated by comparing gradient of the strain energy density with respect to the macroscopic
the results from the numerical studies to the experimental results. After strains. This allows the stiffness matrix to be determined. Lastly, the
validating the analytical and numerical solutions, parametric studies macroscopic forces are then calculated by applying the virtual work
were then conducted by changing the strut thickness and the re-entrant principle. The material properties that were used in this study are shown
angle θ (Fig. 1) to study the effects of these geometrical parameters on in Table 2.
Young’s modulus, Poisson’s ratio and energy absorbed. The methodol­ The analytical model uses a number of Johnson-Cook plastic model
ogy of the analytical and numerical studies will be described in the next and damage model parameters. These parameters are material depen­
subsections. dent and unknown for the printed material used in this study. Therefore,
in the simulations, the properties for stainless steel 316L were used
3.2. Analytical determination of Young’s modulus and Poisson’s ratio instead since the Johnson-Cook parameters are known for this material.
However, we consider the properties of stainless steel 316L to be suffi­
The mechanical properties of bulk material, such as Young’s ciently close to the material used in the study. Despite the potential
modulus and Poisson’s ratio, are not apparent for the auxetic variation in material properties between conventionally machined

4
C. Mercer et al. International Journal of Impact Engineering 169 (2022) 104315

Fig. 4. Unit cell of the auxetic metamaterial.

materials and additive manufactured materials, the predictions made ⎡ ⎤

using the parameters selected appear to be fairly accurate as seen in 1
−
vyx vzx
0 0 0 ⎥
⎢E
Table 3. As can be seen in Table 2, simulations were also performed for ⎢ x
⎢
Ey Ez ⎥
⎥
Ti-6Al-4V to give the reader an idea of how the results can vary with a ⎢ vxy
⎢− 1 vzy ⎥
− 0 0 0 ⎥
change of base material. Ti-6Al-4V was chosen since it is a well-known ⎢
⎢ Ex Ey Ez
⎥
⎥
⎢ ⎥
aerospace alloy. ⎢ v vyz 1 ⎥
⎢ − xz 0 ⎥
The homogenised stiffness matrix, K, for the auxetic metamaterial ⎢
⎢ Ex
−
Ey Ez
0 0 ⎥
⎥
will be in the form shown in Eq. (1) using the Voigt notation. K=⎢
⎢
⎥
⎥ (2)
⎡ ⎤ ⎢ 1 ⎥
Cxx Cxy Cxz 0 0 0 ⎢0 0 0 0 0 ⎥
⎢ Gxy ⎥
⎢ Cyx Cyy Cyz 0 0 0 ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥
⎢ Czx Czy Czz ⎢ 1 ⎥
0 0 0 ⎥ ⎢0 0 0 0 0 ⎥
K=⎢ ⎢ 0
⎥ (1) ⎢ Gxz ⎥
⎢ 0 0 Gxy 0 0 ⎥⎥ ⎢ ⎥
⎣ 0 ⎦ ⎢ ⎥
0 0 0 Gxz 0 ⎣ 1 ⎦
0 0 0 0 0
0 0 0 0 0 Gyz Gyz

Where Cxy = Cyx , Cxz = Czx = Cyz = Czx and Gxz = Gyz Where Ex = Ey , νxy = νyx , νxz = νyz and νzx = νzy
The equivalent Young’s modulus can then be found from the inverse Therefore, based on Eq (2), the full form of the Hooke’s law can be
of the stiffness matrix as follows. written as follows.

⎡ ⎤
⎢ ( ) ⎥
⎢ Ex (1 − νxx ) Ey νxy Ez (νxz ) ⎥
⎢ ( )( ) 0 0 0 ⎥
⎢ ⎥
⎡ ⎤ ⎢ (1 + νxx )(1 − 2νxx ) 1 + νxy 1 − 2νxy (1 + νxz )(1 − 2νxz ) ⎥⎡ ⎤
σxx ⎢ ⎥ εxx
⎢ ( ) ( ) ( ) ⎥
⎢ ⎥ ⎢ Ex νyx Ey 1 − νyy Ez νyz ⎥⎢ ⎥
⎢ σyy ⎥ ⎢ ( 0 ⎥ ⎢ ⎥
⎢ ⎥ ⎢ 1 + ν )( 1 − 2ν ) ( )(
1 + νyy 1 − 2νyy
) ( )(
1 + νyz 1 − 2νyz
) 0 0 ⎥⎢ εyy ⎥
⎢ ⎥ ⎢ yx yx ⎥⎢ ⎥
⎢ σzz ⎥ ⎢ ( ) ⎥⎢ εzz ⎥
⎢ ⎥ ⎢ ⎥⎢ ⎥
⎢ ⎥=⎢ Ex (νzx ) Ey νzy Ez (1 − νzz ) ⎥⎢ ⎥ (3)
⎢ σxy ⎥ ⎢ ( )( ) 0 0 0 ⎥⎢ εxy ⎥
⎢ ⎥ ⎢ (1 + νzx )(1 − 2νzx ) 1 + νzy 1 − 2νzy (1 + νzz )(1 − 2νzz ) ⎥⎢ ⎥
⎢ ⎥ ⎢ ⎥⎢ ⎥
⎢ σ xz ⎥ ⎢ ⎥⎢ εxz ⎥
⎣ ⎦ ⎢ 0 0 0 Gxy 0 0 ⎥ ⎣ ⎦
⎢ ⎥
σ yz ⎢ ⎥ εyz
⎢ 0 0 0 0 Gxz 0 ⎥
⎢ ⎥
⎢ ⎥
⎢ 0 0 0 0 0 Gyz ⎥
⎣ ⎦

5
C. Mercer et al. International Journal of Impact Engineering 169 (2022) 104315

Table 2 ⎡ ⎤
Material properties [34,35].
1 νyx
⎢ Ex − 0 ⎥
⎢ Ey ⎥
Ti- Stainless steel 316L ⎢ ⎥
6Al–4V ⎢ νxy 1 ⎥
− 1
Kplane stress =⎢
⎢ − Ex 0 ⎥
⎥ (6)
Density (kg m− 3) 4430 7850 ⎢ Ey ⎥
⎢ ⎥
Young’s Modulus (GPa) 110 207 ⎣ 1 ⎦
Poisson’s Ratio 0.33 0.29 0 0
Gxy
Johnson-Cook Plastic Model Parameter, A 1098 792
(MPa)
Johnson-Cook Plastic Model Parameter, B 1092 510
(MPa) 3.3. Finite element simulations
Johnson-Cook Plastic Model Parameter, C 0.014 0.014
Johnson-Cook Plastic Model Parameter, n 0.93 0.26 In addition to the analytical solutions discussed previously, finite
Johnson-Cook Damage Model Parameter, D1 -0.09 0.05
element simulations were also performed in order to study the energy
Johnson-Cook Damage Model Parameter, D2 0.25 3.44
Johnson-Cook Damage Model Parameter, D3 -0.5 -2.12
absorption capabilities of the metamaterial during impact loading. The
Johnson-Cook Damage Model Parameter, D4 0.014 0.61 finite element simulations were performed using Abaqus/Explicit finite
element solver.
The lattice structure used in the finite element simulations is shown
Table 3 in Fig. 5, where each strut is treated by as a Timoshenko beam. In this
Comparison of Young’s modulus from the analytical calculations, finite element
figure, the horizontal struts and diagonal struts have different strut
analysis and experiments.
thickness as shown in Fig. 1. The strut thickness of the horizontal struts,
Analytical Experiments which is equal to the thickness of the node will be used as the reference
Ey (MPa) 157 161 in this study. The structure has the same number of cells as the structure
Ey extra strut (MPa) 157 165 used in the experiments. The struts in the finite element solutions were
Ez (MPa) 472 296 modelled using Timoshenko beam models, which is equivalent to Beam
Ez extra strut 546 536
Ey contrained (MPa) 196 255
32 Element in Abaqus/Explicit. Each element has a size of 0.8 mm.
Ey constrained extra strut (MPa) 196 287 The same two materials were used in the finite element simulations
as in the analytical model, namely Ti-6Al-4 V and 316 L steel. These
materials were simulated using the Johnson-Cook plasticity models
Under plane strain conditions, for example when εzz = εxz = εyz = 0, described by Eq. (7) and the Johnson-Cook ductile damage model Eq. (8)
Eq. (3) can be used directly without modification by just setting the [35]. The material parameters used in the finite element models are the
relevant strains in the equation to zero. Therefore, the effective Pois­ same as those used in the analytical solutions and are shown in Table 2.
son’s ratio is the same for the plane strain case as the full 3D-case. Since
the analytical method introduced here assumes the material is homog­ σ = (A + Bεn )⋅(1 + Cln(ε̇∗ )) (7)
enous and the boundary effects near the boundaries in a plane strain ( ( ))
condition may be significant near the boundaries, this method may be σvon Mises
εf = D1 + D2 exp D3 ⋅ ⋅(1 + D4 ln(ε̇∗ )) (8)
inaccurate for the plane strain condition if the lattice structure is thin. σmean
This is because for a thin material the boundary effects are more sig­ In the finite element simulations, the bottom of the lattice is set to be
nificant, and a larger proportion of the lattice structure or material is fixed. A cylinder made of hexahedral elements was used to simulate the
affected by it. This effect is similar to that observed in a bulk homoge­ pellet and the material properties as shown in Table 2 were used to
nous material and the reader should be careful to ensure that the ma­ model the cylinder. The cylinder in the model is set to be initially in
terial is sufficiently thick when assuming plane strain conditions. contact with the top of the lattice structure as shown in Fig. 5. However,
On the other hand, under plane stress conditions, for example σzz = the initial velocity of the cylinder is set to be 6.26 m s− 1, which is
σxz = σyz = 0, Eq. (3) cannot be used directly but can be simplified as equivalent to the velocity of an object after falling 2 m in free fall. The
follows. velocity of the cylinder before and after impact was calculated and was
⎡
Ex Ex νyx
⎤ used to determine the energy absorbed. Once the simulations were
⎡ ⎤ ⎢ 1 − νxy νyx 1 − νxy νyx 0 ⎥⎡ ⎤ validated with the experimental results, they were repeated with
σ xx ⎢ ⎥ εxx different geometries to study the effects of geometry on the energy
⎢ ⎥
⎣ σ yy ⎦ = ⎢ Ey νxy
⎢ Ey ⎥⎣ εyy ⎦
⎥ (4) absorbed. The geometries studied include (i) a fixed node thickness of
⎢ 1 − νxy νyx 1 − νxy νyx 0 ⎥ εxy
σ xy ⎣ ⎦ 1.2 mm but with a varying θ ranging from 30∘ to 75∘ and (ii) a fixed θ as
0 0 Gxy used in the experiments but a varying node thickness ranging from 0.5
mm to 2.5 mm.
In order to obtain the Poisson’s ratio for the plane stress case, the
boundary conditions σ zz = σ xz = σyz = 0, can be substituted into Eq. (1)
4. Experimental results
to give the following equation.
⎡ ( ) ( ) ⎤
Czx Czy 4.1. Quasi-static mechanical behaviour
⎢ Cxx 1 − Czz Cxy 1 −
Czz
0 ⎥
⎢ ⎥
⎢ ( ) ( ) ⎥ Stress-strain curves of the auxetic metamaterials subjected to whole
Kplane stress = ⎢ C C ⎥ (5)
⎢ C 1 − zx 0 ⎥ area compression in both the y- and z-axes, with and without the extra
zy
⎢ yx Cyy 1 − ⎥
⎣ Czz Czz ⎦
central strut, are presented in Fig. 6. In the y-direction, the stiffness
0 0 Gxy values are approximately equal, indicating the removal of the central
The equivalent Poisson’s ratio under plane stress conditions can then strut does not markedly affect the stiffness in this direction (Fig. 6a).
be found by finding the inverse of Kplane stress . However, the stiffness values measured in the z-direction do show a
significant difference depending on whether the central strut is present
(Fig. 6b). In fact, the measured modulus of the specimen with the struts
removed is almost half that of the structures with the central struts
present.

6
C. Mercer et al. International Journal of Impact Engineering 169 (2022) 104315

Fig. 5. Finite element analysis setup for impact test.

Fig. 6. Quasi-static whole area compression test results. (a) y-axis; (b) z-axis.

Fig. 7. Quasi-static central area compression test results.

7
C. Mercer et al. International Journal of Impact Engineering 169 (2022) 104315

The stress-strain curves corresponding to the tests conducted under deformation, which in all four auxetic structures was mainly restricted
central area compression, for both unconstrained and constrained to the top layer of unit cells. In Fig. 3 (b –f) damage pattern for the
structures, in the y-direction are shown in Fig. 7. It is immediately constrained sample without central struts are shown after the 4th, 9th,
apparent that the stiffness values are substantially higher than those 14th and 19th impact, respectively. Poisson’s ratios were determined for
measured under whole area compression. Also, it can be seen from Fig. 7 the first impact test of each structure. All samples exhibited a negative
that the addition of constraining walls results in a (slightly) higher Poisson’s ratio, i.e. auxetic behaviour with the constrained version of
modulus than in the unconstrained cases. both samples being less auxetic than their counterpart without con­
It is important to note that in the cases of added boundary constraint, straining walls, indicating an elevated stiffness of these structures. The
the deformation of the lattice structure is still bending dominated. No walls bent slightly inwards during the impact (see supplementary video:
stretch dominated deformation occurs. This can be demonstrated by the “Impact Test Constrained.avi”) and thus did not completely inhibit
Maxwell rule in 3D given by Eq (9) [36]. lateral contraction.
b − 3j + 6 = s − m (9)
4.3. Numerical modelling results
Where b is the number of struts, j is the number of free joints, s is the
number of states of self stress and m is the number of mechanisms. 4.3.1. Compression test
According to Eq (9), the lattice structure is bending dominated if the Young’s modulus values calculated from the experiments are
left-hand side is less than 0. For the lattice structure studied, the number compared along with the analytical solutions in Table 3. In this table, we
of struts is 771 while the number of free nodes is 417 (after subtracting compared the engineering strains of the experimental and analytical
the nodes that are constrained). Therefore, the left-hand side of Eq (9) is data. We have selected to present this data using engineering strains as
-480 which is significantly lower than 0. Therefore, the deformation of the comparison was done in the elastic region of the stress-strain curve.
the structures even when constrained is bending dominated. As seen in the table, Young’s modulus values from the analytical ana­
lyses agree well with the experiments except for Young’s modulus in the
vertical direction, Ez , without the additional strut and the constrained
4.2. Dynamic mechanical behaviour cases.
The inaccuracies in the prediction of Young’s modulus in the vertical
Force-time curves of the first impact of the four impact-tested auxetic direction without the additional strut can be attributed to the struts in
metamaterials, are presented in Fig. 8. The peak force ranges between the lattice buckling or the lattice slipping that is now accounted for in
9.2 and 9.8 kN, except for the “conventional” structure - i.e. for the the analytical model. The inaccuracies in predicting Young’s modulus of
unconstrained auxetic structure with additional strut – for which the the lattice constrained at both ends is likely to be caused by the
force peak reaches 16.4 kN. In Fig. 9, which shows the curves of repeated analytical model accounting for the outermost nodes being constrained
impacts, it can be noted though, that the first impact not necessarily in the z-direction only, while in the experiments the thick walls will also
entails the highest peak force. No clear trend can be identified with constrain the outermost nodes in the x- and y- directions. Therefore, the
regard to changes in impact duration or force peak with increasing analytical method will underpredict Young’s modulus of the meta­
number of impacts, and there is considerable variation in the height of material in this situation. However, the boundary conditions of the
the peak force. Every subsequent impact leads to increasing plastic constrained case being free to move in the x- and y- directions have been

Fig. 8. Curves representing the force that transmitted through the sample and measured at the bottom of the anvil vs. time resulting from the respective first impact
test performed with each of the samples.

8
C. Mercer et al. International Journal of Impact Engineering 169 (2022) 104315

Fig. 9. Curves representing the force that transmitted through the sample and measured at the bottom of the anvil vs. time curves resulting from repeated impact
tests. (a) Constrained sample with central struts. (b) Constrained sample without central struts. (c) Unconstrained sample with central struts. (d) Unconstrained
sample without central struts.

kept in the further analyses because this result is more useful for the predict the effects of the geometry on Young’s modulus and Poisson’s
design of these metamaterials that will scale to more than the 3 × 3 × 3 ratio. Additionally, the analytical analysis used to predict the properties
unit cells in this study, where the boundary conditions will be closer to in Table 3 assumes that the lattice structure extends infinitely in all
the analytical solution. Hence, based on these results, the models can directions. However, the experimental results of the 3 × 3 × 3 unit cell
successfully predict Young’s modulus accurately. The models can be structure being close to the analytical predictions, as shown in Table 3,
used to provide a qualitative indication of the material behaviour and indicates that even with a relatively small number of unit cells, me­
chanical behaviour of the actual lattice structure is fairly similar to the
ideal behaviour despite the size effects and only 27 unit cells in the
Table 4 experimental samples.
Theoretical and experimentally measured Poisson’s ratio values. The Poisson’s ratio υyz (corresponding to a lateral strain in the z-
Specimen νyz (Theory) νyz (Quasi-Static) νyz (Dynamic) direction as a result of an applied stress in the y-direction) determined
No extra strut, unconstrained -0.45 -0.34 -0.37 by the numerical modelling are compared with the experimentally
No extra strut, constrained -0.45 -0.31 -0.18 measured values in Table 4. As can be seen from the table, all values are
Extra strut, unconstrained -0.45 -0.36 -0.28 negative, indicating auxetic behaviour under both quasi-static and dy­
Extra strut, constrained -0.45 -0.28 -0.22 namic loading in all conditions. Additionally, the experimentally

9
C. Mercer et al. International Journal of Impact Engineering 169 (2022) 104315

Table 5 addition of the extra strut will increase the auxetic effect in the z-di­
Young’s modulus and Poisson’s ratio for the structure used in the experiments rection. Lastly, Young’s modulus of a lattice structure composed of
based on the measured values in the figures above. stainless steel is larger than one made from Ti-6Al-4 V. This result is
Steel no Steel extra Ti-6Al-4 V no Ti-6Al-4 V extra expected as the stiffness of bulk stainless steel is significantly higher than
strut strut strut strut that of Ti-6Al-4 V.
Ex = Ey 157.3 157.3 83.54 83.54 A detailed parametric study of the effects of node thickness and
(MPa) angle, θ, on Young’s modulus and Poisson’s ratio is included in the
Ez (MPa) 472.0 546.6 250.8 290.4 supplementary material.
νxz = νyz -0.4464 -0.4464 -0.4464 -0.4464
νzx = νzy -1.338 -1.552 -1.339 -1.5517
νxy = νyx -0.8577 -0.8577 -0.8578 -0.8578
4.3.3. Effects of geometry on energy absorption
In order to study the effects of node thickness (and thus strut thick­
ness) and re-entrant angle of the auxetic unit cell, θ, on energy absorp­
measured values and theoretical values are all in reasonably good tion abilities, finite element simulations of the material impact test are
agreement. In auxetic structures, Poisson’s ratio will generally vary as a performed for different geometries. The reduction of the kinetic energy
function of strain. However, a detailed assessment of the variation of of the impactor during the impact event is calculated to determine the
Poisson’s ratio with strain is beyond the scope of the current study. energy absorbed by the auxetic lattice structure.
Establishing that the specimens exhibit negative Poisson’s ratio under The effects of the node thickness on the energy absorbed by the
all unconstrained and constrained conditions is deemed to be to be lattice metamaterial are shown in Fig. 10. As seen in the figure, the
sufficient for assessing the effect of constraint. The theoretical Poisson’s energy absorbed by the metamaterial decreases with increasing node
ratio for the constrained and unconstrained cases are the same mainly thickness. Based on Fig. 10, an auxetic lattice with a node thickness of
because the constrained case is under the plane strain condition and as less than 0.5 mm is capable of absorbing over 90% of the kinetic energy
explained in the methods section, Poisson’s ratio in the stiffness matrix is of the impactor. However, this reduces to less than 40% for node
unchanged. thicknesses above 2.0 mm. Therefore, in order to increase the energy
absorption capability, the node thickness should be reduced as much as
4.3.2. Effects of geometry on Young’s modulus and Poisson’s ratio of the possible. A smaller node thickness (and therefore thinner diagonal
metamaterials struts) means that, in principle, the hinge effect that enables the negative
The analytical solutions proposed in the earlier section were used to Poisson’s ratio behaviour operates at a lower stress. This will, in turn,
estimate Young’s modulus and Poisson’s ratio of the metamaterials. The lead to a greater auxetic effect. In practice, however, the lower limit of
values of these properties were calculated for the lattice design used in node thickness will be determined by the strength of the material
the experiments and are summarized in Table 5. Several observations employed. The data shown in Fig. 10 can be fitted with the following
can be made from the table. Firstly, Young’s modulus values of the equation.
lattice structure in the x- and y- directions are identical, but the values in
53.358
the z-direction is significantly higher. This result is expected as the lat­ Absorption = (10)
Thickness0.925
tice is symmetrical about the z-axis. Secondly, Poisson’s ratio of the
structure is negative for all orientations, which is generally a charac­ Based on this equation, the lattice can only absorb up to 6.34% of the
teristic of auxetic metamaterials. Another interesting observation that energy with a node thickness of 10 mm. As the energy absorption ca­
can be made from Table 5 is that the addition of the extra strut will only pabilities reduce greatly with increasing node thickness, large node
increase the stiffness in the z-direction, while the stiffness in the x- and y- thicknesses should be avoided whenever possible.
directions remains constant. The experimental results, as seen in Fig. 11 shows the effects of varying the re-entrant angle, θ, on the
Table 3, also show similar behaviour. For example, there is a significant energy absorbed. As seen in the figure, the energy absorbed by the
increase in stiffness for the lattice structure in the z-direction when the structure increases steadily from 50% at 35◦ to almost 60% around 45◦
extra strut is added, as seen in Table 3. However, only a small increase in However, the energy absorbed then decreases to a minimum value of
the stiffness in the y-direction is observed when the extra strut is approximately 50% at θ = 50◦ After this minimum value, the energy
included as seen in Table 3. Similarly, the addition of the extra strut has absorbed increases steadily again to 60% at approximately 70 before
◦

only increased the magnitude of Poisson’s ratio in the zx- and zy- decreasing slightly. This result suggests that there is a small range of
orientation, νzx and νzy , as seen in Table 5. This result suggests that the values of θ that should be avoided as this will lead to a markedly reduced

Fig. 10. Energy absorbed vs node thickness. Fig. 11. Energy absorbed vs re-entrant angle.

10
C. Mercer et al. International Journal of Impact Engineering 169 (2022) 104315

value of energy absorption. The reason for minimum value of energy reducing the ability to absorb impact energy under otherwise identical
absorption at a re-entrant angle of 50 is not fully understood. However, conditions. In the worst case, this underestimation of the stiffness could
◦

since the deviation from monotonic behaviour occurs at approximately push accelerations and forces occurring during an impact beyond critical
the mid-point of the evaluated range of θ, it is surmised that it may be a levels.
consequence of competing factors such as lattice stiffness (which de­ When taking into account the boundary constraint and its influence
creases with decreasing re-entrant angle) and Poisson’s ratio (which on the stiffness, impact forces and energy are still successfully shunted to
becomes less negative with decreasing re-entrant angle). Ideally, to surrounding regions.
maximize the energy absorbed, θ should be set at approximately 70◦ The removal of the central strut of the unit cell to save weight and
facilitate easier support removal has no effect on νyz but will reduce the
5. Discussion νzx/νzy value by approximately 15%. Also, since the central strut runs in
the z-direction, it will have little influence on the stress-strain behaviour
The experiments and theoretical modelling work performed in this in the y-direction, particularly when the sample is unconstrained (Fig. 6
investigation have confirmed that all of the printed metamaterials (a)). However, removal of the central strut does significantly affect the
exhibit auxetic behaviour under both quasi-static and dynamic (impact) stiffness of the structure in the long axis (z)-direction (Fig. 6(b). Again,
loading conditions (Table 4). While the additional struts have almost no this needs to be taken in account in future applications of the structural
influence on the mechanical properties for quasi-static loading in the y- metamaterials.
direction (Fig. 6(a)), it can be seen that for dynamic impact loading in The investigation has also shown that the elastic modulus values are
the same direction (at least for the first impact) the additional struts lead substantially higher in the cases where the structures were subjected to
to a significant increase in the transmitted force in case of the uncon­ central area compression. This is an important result and demonstrates
strained specimen (Fig. 8). In the case of the constrained specimens, the than standardized whole area compression tests (typically used in the
effect of added struts may be masked by the overall influence of the literature) are not adequate to fully characterize the mechanical
constraining walls. However, while the repeatability of the force-time behaviour of these types of metamaterials. When only part of the spec­
curves (Fig. 9) indicates that all structures are mechanically robust imen is compressed, as will often be the case when such structures are
and stable over repeated impacts with no incidence of catastrophic employed in practical applications, the loading configuration and hence
failure, it also shows that the measured force-time curves vary consid­ mechanical response of the structure will be markedly different.
erably. It is not that the force increases with the number of repetitions, as Based on the simulation results shown in this study, the calculated
one might have expected due to the increasing plastic pre-damage energy absorption was 48%. The simulation result is in good agreement
(Fig. 3b-f). It rather seems to be due to the fact that the impactor does with measurements made from the movies produced from the high-
not always hit the specimen in exactly the same way. This may be speed camera images captured during the impact tests, which showed
because the top row of auxetic unit cells is already deformed (Fig. 9) as a that the energy absorption of the printed metallic structures varies
result of prior impacts, or because in practice the impactor hits the around 50% to 60%. However, the repeated tests also show that such
specimen at a very slight angle, as is presumably the case during each regular lattice structures are sensitive to how exactly they are loaded, e.
respective first impact (Fig. 8). Since in the present experimental case g. whether the impact hits slightly offset or at a slight angle. It is also
the lateral constraints are directly connected to the bottom plate, the possible that the energy absorption of this metamaterial may be
occurring forces are ultimately also transmitted to the force sensor un­ improved further by the addition of magnetic inclusions, as reported by
derneath. Nevertheless, no positive lateral strains were observed in any Dudek et al. [37]. In this work, the authors show that appropriately
of the specimens. Even with the addition of lateral constraint, the distributed magnetic inclusions can minimize the force that is trans­
structure still exhibits auxetic behaviour, albeit at a slightly reduced ferred to the structure during impact.
level. This demonstrates quite impressively that sufficient forces are In addition, the simulations have shown that greater energy ab­
deflected laterally to deflect even the massive walls. Poisson’s ratios sorption can be achieved with thinner struts. The reason for this is that a
were determined to be of the order of 20% lower (less negative) when reduced strut thickness will allow the hinge effect that enables negative
constraining walls were present. In fact, it’ s the constraining walls that Poisson’s ratio behaviour to operate at a lower stress. This will in turn
absorb the forces that are diverted by the auxetic structure and dissipate result in a greater auxetic effect.
the energy introduced by an impact by being deformed themselves. This
mutual interaction of all components involved should be considered 6. Conclusions
when designing such metamaterials for future crash protection
applications. The quasi-static and dynamic (impact) mechanical response of 3-D
In should be noted that constraint could not be added to the front and printed, metallic, auxetic metamaterials has been investigated. Several
back of the specimen, otherwise the auxetic structure would have been important findings have emerged from the study, which are summarized
completely enclosed and observation of the deformation would not have below.
been possible. However, front and rear constraint would have an addi­
tional effect on auxetic response. But it is unlikely that an auxetic crash 1. All structures exhibited auxetic behaviour (negative Poisson’s ratio)
protection structure would be completely constrained on all sides during under both quasi-static and impact loading even when thick con­
service and so the constraint design chosen is considered to be reason­ straining walls were added to the sides and bottom of the structure.
able. The thickness of the constraining walls was chosen based on what However, the addition of constraint does reduce the auxetic effect
was considered to be realistic and sensible for the relevant applications. somewhat (typically around 20%), and thus shows that axial forces
However, it should also be noted that the wall thickness will have an are successfully diverted into the lateral constraining walls. Failure
effect on the auxetic behaviour of the metamaterials, i.e. the auxetic to take into account such effects of boundary constraint could lead to
effect will decrease as wall thickness increases. critical overestimation of the energy-absorbing abilities of auxetic
Our experiments show that for auxetic samples it is important to be crash protectors in service.
able to predict their behaviour in situ, i.e. at the site of their application 2. The exclusion of the central strut in the unit cell of the structure (to
in service. Since their main objective is the lateral diversion of forces reduce weight and facilitate easier removal of support following 3D
into neighbouring structures, it is vital to factor in these as well. The printing) has only a very minor influence on the auxetic behaviour
installation and thus lateral constraining can significantly increase the but does reduce the stiffness of the structure in the long axis direction
stiffness of auxetic impact protectors by blocking the auxetic effect and by almost a factor of two.

11
C. Mercer et al. International Journal of Impact Engineering 169 (2022) 104315

3. The work has shown that the structures perform very differently [8] Evans KE, Alderson A. Auxetic materials: functional materials and structures from
lateral thinking! Adv Mater 2000;12:617–28.
when compressed in the central region only, as compared with
[9] Allen T, Shepherd J, Hewage TAM, et al. Low-kinetic energy impact response of
compression of the whole specimen. In fact, elastic modulus values auxetic and conventional open-cell polyurethane foams. Phys Status Solidi B 2015;
can be up to 50% higher when compressed in the central region only. 252:1631–9.
Therefore, it is believed that the “whole area” compression experi­ [10] Duncan O, Shepherd T, Moroney C, et al. Review of auxetic materials for sports
applications: expanding options in comfort and protection. Appl Sci. 2018;8:941.
ments generally performed as standard in the literature, are in fact, [11] Chan N, Evans KE. The mechanical properties of conventional and auxetic foams.
inadequate to fully characterize the mechanical behaviour of these part I: compression and tension. J Cell Plast 1999;35:130–65.
metamaterials. [12] Lakes RS, Elms K. Indentability of conventional and negative Poisson’s ratio foams.
J Compos Mater 1993;27:1193–202.
4. The dynamic loading simulations indicate that these structures are [13] Yang L, Harrysson O, West H, et al. Modeling of uniaxial compression in a 3D
capable of absorbing over 90% of the energy of an impact. This is periodic re-entrant lattice structure. J Mater Sci 2012;48:1413–22.
similar to the energy dissipation ability of the pomelo peel that [14] Lee J, Choi JB, Choi K. Application of homogenization FEM analysis to regular and
re-entrant honeycomb structures. J Mater Sci 1996;31:4105–10.
provided the inspiration for this study. However, although the en­ [15] Li T, Liu F, Wang L. Enhancing indentation and impact resistance in auxetic
ergy dissipation characteristics of the printed metallic structures are composite materials. Compos Part B-Eng 2020;198:108229.
in practice, inferior to this, they are still able to absorb at least 50% of [16] Masters IG, Evans KE. Models for the elastic deformation of honeycombs. Compos
Struct 1996;35:403–22.
the energy of an impact, provided that a ductile material is [17] Zhang X, Yang D, Zhang X, et al. Mechanical properties of auxetic cellular material
employed. consisting of Re-entrant hexagonal honeycombs. Materials (Basel) 2016;9:900.
5. The theoretical simulations of geometrical effects on parameters [18] Yang L, Harrysson O, West H, et al. Mechanical properties of 3D re-entrant
honeycomb auxetic structures realized via additive manufacturing. Int J Solids
such as elastic modulus, Poisson’s ratio and energy absorption
Struct 2015;69–70:475–90.
capability performed in this study provide useful tools for optimizing [19] Wang N, Liu W, Tang A, et al. Strain isolation: a simple mechanism for
the design of such lattice metamaterials for practical engineering understanding and detecting structures of zero Poisson’s ratio. Phys Status Solidi B
applications. 2014;251:2239–46.
[20] Xue Y, Wang X, Wang W, et al. Compressive property of Al-based auxetic lattice
structures fabricated by 3-D printing combined with investment casting. Mater Sci
Declaration of Competing Interest Eng A 2018;722:255–62.
[21] Schwerdtfeger J, Heinl P, Singer RF, et al. Auxetic cellular structures through
selective electron-beam melting. Phys Status Solidi B 2010;247:269–72.
The authors declare that they have no known competing financial [22] Yang C, Vora HD, Chang Y. Behavior of auxetic structures under compression and
interests or personal relationships that could have appeared to influence impact forces. Smart Mater Struct 2018;27:025012.
the work reported in this paper. [23] Li P, Yue J, Li X, Wan W. Axial compression and collapse properties of 3D Re-
entrant hexagonal auxetic structures. In: ASME 2020 39th International
Conference on Ocean, Offshore and Artic Engineering; 2020. 2A. V02AT02A011.
Acknowledgments [24] Wang X, Chen Y, Ma L. The manufacture and characterization of composite three-
dimensional re-entrant auxetic cellular structures made from carbon fiber
reinforced polymer. J Composite Materials 2018;52:3265–73.
This investigation was partly supported by the National Institute for [25] Mercer C, Lee J, Balint DS. An investigation of the mechanical behavior of three-
Materials Science, Tsukuba, Japan (Project Number PD3020). TS ac­ dimensional low expansion lattice structures fabricated via laser printing. Compos
knowledges funding by the Deutsche Forschungsgemeinschaft (DFG, Struct 2018;206:80–94.
[26] Yang L, Harrysson O, West H, et al. Compressive properties of Ti–6Al–4V auxetic
German Research Foundation) under Germany’s Excellence Strategy –
mesh structures made by electron beam melting. Acta Mater 2012;60:3370–9.
EXC-2193/1 – 390951807. [27] Smith CW, Grima JN, Evans K. A novel mechanism for generating auxetic
behaviour in reticulated foams: missing rib foam model. Acta Mater 2000;48:
Supplementary materials 4349–56.
[28] Speck T., Bold G., Masselter T., et al. Biomechanics and functional morphology of
plants—inspiration for biomimetic materials and structures. In: Geitmann A, Gril J,
Supplementary material associated with this article can be found, in editors. Plant biomechanics. Cham: Springer International Publishing; 2018. p.
the online version, at doi:10.1016/j.ijimpeng.2022.104315. 399–433.
[29] Thielen M, Speck T, Seidel R. Impact behaviour of freeze-dried and fresh pomelo
(Citrus maxima) peel: influence of the hydration state. R Soc open sci 2015;2:
References 140322.
[30] Deshpande V, Fleck N, Ashby M. Effective properties of the octet-truss lattice
[1] Yeganeh-Haeri A, Weidner DJ, Parise JB. Elasticity of α-cristobalite: a silicon material. J Mech Phys Solids 2001;49:1747–69.
dioxide with a negative Poisson’s ratio. Science 1992;257:650–2. [31] Lee BK, Kang KJ. A parametric study on compressive characteristics of Wire-woven
[2] Masselter T., Bold G., Thielen M., et al. Bioinspired Materials and Structures: a Case bulk Kagome truss cores. Compos Struct 2010;92:445–53.
Study Based on Selected Examples. In: Yang T, Xiao L, Lamboni L, editors. [32] Ptochos E, Labeas G. Elastic modulus and Poisson’s ratio determination of micro-
Bioinspired materials science and engineering. Wuhan: Wiley; 2018. [p. 253–266. lattice cellular structures by analytical, numerical and homogenisation methods.
[3] Bührig-Polaczek A, Fleck C, Speck T, et al. Biomimetic Cellular metals - using J Sandw Struct Mater 2012;14:597–626.
hierarchical structuring for energy absorption. Bioinspir Biomim 2016;11(4): [33] Vigliotti A, Pasini D. Stiffness and strength of tridimensional periodic lattices.
045002. Comput Methods Appl Mech Eng 2012;229:27–43.
[4] Lakes R. Foam structures with a negative Poisson’s ratio. Science 1987;235: [34] Zhang Y, Outeiro JC, Mabrouki T. On the selection of Johnson-Cook constitutive
1038–40. model parameters for Ti-6Al-4V using three types of numerical models of
[5] Lakes R.S., inventor; University of Iowa Research Foundation UIRF, assignee. orthogonal cutting. Procedia CIRP 2015;31:112–7.
Polyhedron cell structure and method of making same. United States patent US [35] Johnson GR, Cook WA. Fracture characteristic of three metals subjected to various
4,668,557A. 1987 May 26. strains, strain rates, temperatures and pressures. Eng Fract Mech 1985;21:31–48.
[6] Almgren RF. An isotropic three-dimensional structure with Poisson’s ratio =− 1. [36] Deshpande VS, Ashby MF, Fleck NA. Foam Topology: bending versus Stretching
J Elasticity 1985;15:427–30. Dominated Architectures. Acta Mater 2001;49(6):1035–40.
[7] Grima JN, Caruana-Gauci R, Attard D, Gatt R. Three-dimensional cellular [37] Dudek KK, Wolack W, Gatt R, Grima JN. Impact resistance of composite magnetic
structures with negative Poisson’s ratio and negative compressibility properties. metamaterials. Sci Rep 2019;9:3963–72.
Proc. Roy. Soc. A. 2012;468:3121–38.

12