You are on page 1of 7

Materials Science & Engineering A 558 (2012) 579–585

Contents lists available at SciVerse ScienceDirect

Materials Science & Engineering A


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

Non-stochastic Ti–6Al–4V foam structures with negative Poisson’s ratio


Li Yang a, Denis Cormier b,n, Harvey West a, Ola Harrysson a, Kyle Knowlson a
a
Edward P. Fitts Department of Industrial & Systems Engineering, North Carolina State University, 400 Daniels Hall, 111 Lampe Drive, Raleigh, NC 27695, USA
b
Department of Industrial Systems Engineering, Rochester Institute of Technology, 81 Lomb Memorial Drive, Rochester, NY 14623-5603, USA

a r t i c l e i n f o abstract

Article history: This paper details the design, fabrication, and testing of non-stochastic auxetic lattice lattice structures.
Received 19 March 2012 All Ti–6Al–4V samples were created via the Electron Beam Melting (EBM) additive manufacturing
Received in revised form process. It was found that the Poisson’s ratio values significantly influence the mechanical properties of
13 August 2012
the structures. The bending properties of the auxetic samples were significantly higher than those of
Accepted 13 August 2012
currently commercialized metal foams. The compressive strength was moderately higher than available
Available online 19 August 2012
metal foams. These results suggest that metallic auxetic structures have considerable promise for use in
Keywords: a variety of applications in which tradeoffs between mass and mechanical properties are crucial.
Cellular materials & 2012 Elsevier B.V. All rights reserved.
Porous materials
Titanium alloys
Failure

1. Introduction Eqs. (1) and (2):


E
For a specimen subjected to uniaxial stress in the elastic G¼ ð1Þ
2ð1 þ nÞ
region, Poisson’s ratio is defined as the negative ratio of the
transverse strain to the axial strain, and can theoretically range E
K¼ ð2Þ
from  1.0 to 0.5 for isotropic structures and materials. Although 3ð12nÞ
materials with negative Poisson’s ratios are not commonplace, the
where n is the Poisson’s ratio of the material.
emergence of additive manufacturing processes with few geo-
From Eqs. (1) and (2), it is clear that when Poisson’s ratio is
metric limitations has generated considerable interest in this
negative, the material will have G*K which indicates significantly
class of materials.
higher shear modulus than bulk modulus. Materials that possess
Materials with negative Poisson’s ratios (auxetic materials)
superior shear modulus are expected to possess high indentation
were first reported by Lakes in 1987 [1]. He described a straight-
resistivity [6], high torsional rigidity [7], high bending stiffness;
forward method to produce polymer foam structures with nega-
shear resistance [8], and high energy absorption efficiency [4].
tive Poisson’s ratios. The technique involves first compressing a
Furthermore, auxetic materials are also appealing candidates for
regular foam structure in three orthogonal directions and then
use as cores in sandwich panels [9]. During the bending of a
holding it there at an elevated temperature for a period of time.
sandwich panel, the inner half of the bending section is subjected
After allowing it to cool to room temperature, it is then decom-
to compressive stress while the outer half is subjected to tensile
pressed. The technique typically starts with a low relative density
stress. For regular materials with a positive Poisson’s ratio, the
foam in order to accommodate the volume of materials com-
compressed section’s cross sectional area tends to expand while
pressed [2], and it is primarily used with polymer materials.
the elongated section’s area tends to shrink. This produces
A number of researchers have successfully employed this process
internal shear stress and buckling of the cellular structure. For
to produce structures with negative Poisson’s ratio [3–5].
auxetic materials, the compressed section exhibits transverse
According to the theory of elasticity for isotropic materials,
shrinkage, therefore these parts are expected to have better
the shear modulus G and bulk moduli K can be determined by
overall mechanical properties during bending.
Although there have been theoretical analyses of auxetic
structures, challenges with manufacturing these geometries
n
have limited their use in practical applications until very
Corresponding author. Tel./fax: þ 1 585 475 2713.
E-mail addresses: lyang5@ncsu.edu (L. Yang), drceie@rit.edu (D. Cormier),
recently [10]. Furthermore, the majority of research dealing with
hawest@ncsu.edu (H. West), harrysson@ncsu.edu (O. Harrysson), auxetic materials has been focused on polymer materials. As new
kyle.knowlson@gmail.com (K. Knowlson). techniques for synthesis of these materials became available,

0921-5093/$ - see front matter & 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.msea.2012.08.053
580 L. Yang et al. / Materials Science & Engineering A 558 (2012) 579–585

applications in areas such as medical implants and energy a less negative ny. On the other hand, design variation 2 (DV2) has
absorbers will be explored. A prerequisite for actual implementa- large H/L ratio and y and is therefore expected to exhibit a greater
tion, however, is that the mechanical properties must be consis- negative ny. Using Eqs. (3) and (4), the theoretical Poisson’s ratios
tent and predictable. In this paper, the Electron Beam Melting ny for DV1 and DV2 are listed in Table 1. This assumes a
(EBM) additive manufacturing process is employed to produce sufficiently large number of unit cells so that the edge effects
foam structures with predetermined auxetic behavior using Ti– could be ignored. Due to the finite number of unit cells in the
6Al–4V. The EBM process is capable of fabricating complicated actual samples, it is to be expected that the actual Poisson’s ratios
open cell metal foam structures from CAD models whose mechan- might be influenced by the edge effects.
ical properties are comparable to those obtained from theoretical The cross section of the strut for both design variations is a
calculations [11]. square with width of 1 mm. Two different types of specimens were
fabricated for bending and compressive testing respectively. Fig. 2
shows the actual parts built in Ti–6Al–4V via the EBM process. For
2. Auxetic structure design both the bending and compressive test specimens, the repetition of
unit cell are 2  2  32 and 6  6  4, respectively.
A number of auxetic structures have been proposed in the
literature [1,2,12–15], the majority of which are two-dimensional.
A number of these geometries are easily interpreted as 3D 3. Experimental approach
structure, while others show significant chirality and therefore
cannot be readily translated into 3D with the same symmetry Six bending specimens for each design variation were built in a
[16]. In this study, the re-entrant lattice auxetic structure single batch in an Arcam A-2 EBM system. Two DV1 and four DV2
designed by Warren [12] is adopted and expanded into a 3D specimens were built in a second batch using the same processing
geometry. parameters. Following fabrication, the specimens were left in the
For materials with high elastic modulus such as Ti–6Al–4V, chamber to cool down overnight in vacuum.
deformation of the re-entrant lattice structure is mostly attrib- Bending and compression tests were performed on an Applied
uted to the deflection of ribs in the structure. Wan et al. [17] Test Systems model 1620C which has a maximum loading force of
showed that for the 2D re-entrant lattice structure, the Poisson’s 100 kN. For the bending tests, the support span was 400 (101.6 mm),
ratios are largely determined by the size of the struts as well as the loading nose span was 200 (50.8 mm), and the support and load
the re-entrant angle. For structures whose struts have a square rollers had a diameter of 12.7 mm.
cross section, the unit cell design can be described in terms of For the compression tests, two parallel plates were used to
length of the diagonal ribs (L), the length of the vertical ribs (H), avoid stress concentration on the structures, and the loading
and the angle (y) between the diagonal ribs and the horizontal speed was 1.27 mm/min. Poisson’s ratio for each sample was
planes. Their work shows that for 2D re-entrant lattice, the H/L obtained by pausing the ATS machine and measuring the sample
ratio and y have an opposite effect on the Poisson’s ratios of the size in the transverse directions using digital calipers. Due to the
two principal directions. Under a small deflection assumption, the relatively high modulus of the specimens, the transverse dimen-
relationship between Poisson’s ratio and the parameters is sion contraction was on the order of 0.01 mm scale which was
approximated by Eqs. (3) and (4) [17]. close to the resolution of the caliper. The measurement of
cos2 y Poisson’s ratio was therefore only an approximation. Provided,
nx ¼  ð3Þ Poisson’s ratio is negative and there are significant differences
ðH=LsinyÞsiny
between two design variations, the discussion can be qualitatively
ðH=LsinyÞsiny justified.
ny ¼  ð4Þ Although the EBM process fully melts the titanium powder,
cos2 y
there is always a boundary between powder that is melted during
The expanded 3D structure studied by the current research is
fabrication and the surrounding powder that is not melted. At this
shown in Fig. 1. Since the 2D structure is anisotropic, the
interface, some particles will sinter and become partially attached
corresponding expanded 3D structure is also anisotropic. The
to the struts. Due to this sintering effect, the actual dimensions of
design parameters are shown in Fig. 1 in the 2D demonstration.
each given EBM part slightly differs from the specified CAD model
Assuming an ideal situation in which the number of unit cell
repetitions is infinite in all three principal directions, the relation-
Table 1
ship between the Poisson’s ratio and the parameters also happens Design parameters and theoretical Poisson’s ratio values in y.
to be expressed as indicated by Eqs. (3) and (4). In order to
estimate the effect of Poisson’s ratio on the mechanical properties Design H (mm) L (mm) y (deg) Theoretical PR in y
of the structure, two different designs are used in this experiment
DV1 2.38 2.714 20  0.351
as indicated in Table 1. Design variation 1 (DV1) has a small H/L
DV2 5.685 1.591 45  1.970
ratio and relatively small y and therefore is expected to exhibit

Fig. 1. 3D unit cell for auxetic lattice structure.


L. Yang et al. / Materials Science & Engineering A 558 (2012) 579–585 581

Fig. 2. Specimens built by EBM (a) bending specimens; (b) compressive specimens.

Table 2
Parameters for bending specimens.

Design L1 ðmmÞ L2 ðmmÞ L3 ðmmÞ TRD (%) ARD (%)

DV1 143.617 0.05 10.097 0.01 8.177 0.04 34.47 36.40


DV2 143.717 0.06 10.12 7 0.03 16.78 7 0.03 30.22 27.24

Fig. 3. Calculation of modulus values.


Table 3
Parameters for compressive specimens.
dimensions. The actual dimensions and weights of the specimens
were therefore measured and compared with the theoretical Design L1 ðmmÞ L2 ðmmÞ L3 ðmmÞ TRD (%) ARD (%)
designs using digital calipers and an electronic balance. The
DV1 26.75 7 0.04 26.84 7 0.05 15.71 70.01 34.47 36.04
measurement indicated that the dimension and density variation
DV2 26.74 7 0.03 26.79 7 0.08 33.29 70.06 30.22 26.51
was negligible for this experiment. For both bending and com-
pression samples, the standard deviation of the dimension mea-
surements were o0.5%. as the value of compressive stress up to the machine’s limit.
The estimated modulus of the structures during bending (Eb) and During compressive testing, the strain values reached approxi-
compression (Ec) were determined using Eq. (5) [18] and Eq. (6). mately 17% and 46% for DV1 and DV2, respectively. The total
energy absorption (W) values at 50% strain for compressive
F b D3 F D3
o¼ ) Eb ¼ b ð5Þ testing were interpolated based on the testing curve patterns. In
96 Eb I 96 Io
addition, the specific energy absorption (w) was also obtained by
Dh s sh Fc h calcualting the energy absorption per unit weight of the structure.
e¼ ¼ ) Ec ¼ ¼ ð6Þ
h Ec Dh A Dh
where o is the deflection of the beam at the contacting nose 4. Results and analysis
position along the beam, and s and e are the stress and strain of the
compressive specimen, respectively. Fb is the corresponding total Tables 2 and 3 show the dimensions of the actual samples, and
force, D is the span between two supports, Fc is the compressive compare the actual and predicted relative densities. In the tables
force, h is the total length of the specimen along the compressive L1, L2 and L3 are the average length, width and height of the
direction, A is the cross sectional area of the geometry bounding box samples, respectively. The dimensional variations between speci-
of the compressive specimens, and I is the second moment of area of mens are quite small ( r0.5 mm), and are therefore neglected for
the geometry bounding box for the bending specimens. The dimen- the future discussion. In Tables 2 and 3 the TRD and ARD stand for
sions and force components are futher demonstrated in Fig. 3 for the theoretical relative denstiy and the actual relative density,
clarification. respectively.
Structures having negative Poisson’s ratios were expected to Not surprisingly, there exist slight differences between theo-
possess higher compressive strength (sC) and flexural strength retical and actual relative densities. This is to be expected based
(sF). In order to estimate the strength values, Eqs. (7) and (8) were on the aforementioned sintering effect. For both bending and
used. compressive specimen sets, the difference between actual and
My y predicted density was less than 10%.
sF ¼ ð7Þ
The force–displacement curves for the bending test are shown
I
in Fig. 4. In some of the curves, a plateau stage near the start of
F elastic deformation can be seen which deviates from the regular
sC ¼ ð8Þ
A pattern. This phenomenon was likely to be caused by the slack of
The results obtained in this study were compared with the the crosshead at the start of compression. Therefore, for these
corresponding mechanical properties of other metal foams. samples, the maximum displacement values were offset by the
Values were normalized by the properties of the bulk materials slack values. The lack of significant yielding prior to failure
in order to estimate the effect of the lattice geometry versus the indicates that the structures underwent sudden and catastrophic
bulk material. Due to the loading limit of the testing machine, the fracture which is typical for this type of metal foam.
compressive strength for the specimens of DV1 was not reached. From Eq. (7), the modulus of elasticity E and the flexural
The compressive stress value was therefore conservatively taken strength BS of DV1 and DV2 can be determined as shown in
582 L. Yang et al. / Materials Science & Engineering A 558 (2012) 579–585

Table 4. In Table 4 W and w represent the total energy absorption


and specific energy absorption during the bending, respectively.
Again the values are averaged over the 6 specimens. The standard
deviation is indicated in parentheses. With a target 75% margin
of error for modulus and flexural strength, a sample size of six
specimens, and the sample standard deviations shown in Table 4,
the computed confidence intervals for the modulus of DV1 and
DV2 are 99.8% and 99.6%, respectively. Likewise, the computed
confidence intervals for the flexural strength of DV1 and DV2 are
99.9% and 92.0%, respectively. For purposes of preliminary scien-
tific experimentation using a relatively new fabrication technol-
ogy, these confidence values were deemed to be quite reasonable.
From the results, DV1 shows significantly higher average
modulus and average flexural strength than DV2. DV2 has higher
total energy absorption before fracture, which is attributable to
the greater strut thickness.
The strain–stress curves for the compressive tests are shown in
Fig. 5, and the corresponding data is shown in Table 5. The test
results for DV2 show a cyclic pattern throughout the compression
which was also observed by Cansizoglu during testing of hex-
agonal lattice structures made by EBM [19]. The tests ended at
strain values of approximately 47% for DV2. The energy absorp-
tion values up to that point were taken to be the value for 50%
strain for comparison with other published work. The DV1 speci-
mens did not crush under the machine’s limit of 100 kN during
Fig. 5. Strain–stress curve for compression experiments (a) design variation 1; (b)
the test, therefore the compressive strength for these samples
design variation 2.

Table 5
Mechanical properties from compressive tests.

Design SðMPaÞ EðGPaÞ W 50% ðJÞ w 50% ðJ=kgÞ

DV1 130.16* 1.21 300 16.67


DV2 58.61( 7 5.38) 1.63( 7 0.13) 285.13( 7 32.63) 10.18( 71.17)

n
Reaches maximum loading capacity of the tester.

exceeds 130 MPa. Also, it was expected that DV1 specimens


would have the same cyclic loading pattern. The energy absorbed
is therefore estimated by assuming that the curve shown in
Fig. 5(a) represents one complete cycle and can be repeated up
to 50% strain.
Poisson’s ratios for the DV1 and DV2 specimens were about
 0.052 and 0.097, repsectively. The measured values for both
design configurations were negative, with DV2 being more
negative than DV1, as expected due to the larger re-entrant angle
as well as the H/L ratio for DV2.

5. Discussion

5.1. Flexural properties

To date, the majority of research involving flexural perfor-


mance of foam materials has involved foam core sandwich panels
Fig. 4. Force–displacement curves for four-point bending tests (a) design variation with face skins [20–23]. Auxetic materials show promise for
1; (b) design variation 2. improved flexural performance with respect to conventional foam
structures. Fig. 6(a) and (b) compare flexural performance of
auxetic structures with other metal foam sandwich panels
Table 4
reported in the literature [19–26]. In order to allow comparison
Mechanical properties from bending test. of properties for lattices built in different materials, the lattice
strength and modulus values were normalized by dividing them
Design EðGPaÞ BSðMPaÞ wðJÞ wðJ=kgÞ by the bulk material strength and modulus, respecitvely.
Note that the foam materials used for comparison have solid
DV1 11.25(7 0.43) 143.54( 7 7.30) 1.71(7 0.18) 126.54( 712.99)
DV2 1.88(7 0.08) 54.30( 7 3.80) 3.05(7 0.53) 146.72( 725.41)
face skins, whereas the auxetic lattice materials presented here do
not. Although sandwich structures with face skins are expected to
L. Yang et al. / Materials Science & Engineering A 558 (2012) 579–585 583

outperform structures without face skins, the auxetic lattice unit cell means that nodes actually come into contact with each
structures presented here compared very favorably with non- other during compression and tend to ‘‘brace’’ the structure.
auxetic foams having face skins. In relation to published results, Greater force is therefore needed to achieve further compression,
DV1 had relatively high bending modulus as well as flexural thus the significant increase in apparent strength relative to the
strength. more open DV2 unit cell geometry created by its 451 re-entrant
According to Eq. (1), however, auxetic structures will exhibit angle. The cyclic stress–strain pattern shown in Fig. 5 is due
higher shear modulus as the Poisson’s ratio becomes increasingly to the fact that failure always occured on the ribs within the same
negative. As a consequence, the auxetic structures can be layer normal to the compressive direction. In other words, the
expected to exhibit higher flexural strength compared to regular part is compressed until one entire layer collapses. Then the
foam structures. During bending, the normal stress direction next layer starts to take up the load until it collapses. Unlike
is along the x axis. DV1 would be expected to exhibit higher most metal foams, non-stochastic auxetic lattice materials have a
flexural strength than DV2. The experimental results supported regular layered structure. When an individual strut fails, the
this hypothosis. load supported by that layer is carried by a smaller number of
Not surprisingly, both design variations exhibited relatively struts. This places stress concentration on other struts in the same
low ductility during flexural testing. This has also been observed layer, and collapse of the entire layer quickly follows. Not
in other EBM fabricated Ti6–Al–4V lattice structures and can be surprisingly, the total number of cycles seen in the stress–strain
attributed to two factors. First, Ti–6Al–4V does not have high curve is equal to the number of layers in the specimen. Fig. 7
plastic deformability after yield. Second, lattice structures with shows a sample during testing in which two distinct layers have
small diameter sloped struts inevitably have some fabrication collapsed.
defects resulting from the well known stair stepping effect The compressive modulus and strength of the specimens are
common to all additive processes [11]. The stair stepping effect shown in Fig. 8(a) and (b) respectively [19,24,26–37]. Again, the
reduces the mechanical strength by reducing the effective strut normalized properties were compared in order to take the solid
diameter. In this case, the strut diameter reduction is estimated material properties into account. The auxetic structures exhibited
as 19% and 28% for DV1 and DV2 respectively. Therefore, it is higher compressive strength and average modulus compared
reasonable to predict that under improved fabrication conditions, with most other metal foams. Although DV1 has higher relative
the bending modulus and strength will significantly improve. density than DV2, it exhibited smaller modulus. This conflicts
with Eq. (12). This could be partly contributed to the Poisson’s
5.2. Compressive properties ratio values. Since the axis of compression lies along the part’s y
direction, DV2 is expected to exhibit greater negative Poisson’s
Due to the loading limit of the testing machine, the maximum ratio compared to DV1. During the compression tests, the upper
compressive strength and energy absorption of DV1 could not be and lower surface of the lateral movement of specimens
determined. This result was partly associated with the fact that were restricted by contact with the parallel plates. On the other
the re-entrant joints move inwards during compression. The hand, the center section of the specimens tended to shrink
small distance between the re-entrant nodes within each DV1 transversely. Consequently, additional shear stress is generated
in the cross-sectional plane normal to the compressive direction.
As previously discussed, auxetic structures with greater negative
Poisson’s ratio are expected to be more resistant to shear
deformation.
Energy absorption is another property of interest for many
applications of metal foams. Energy absorption during the compres-
sion test was compared with regular foam structures as shown in
Fig. 9 [26,27,29,38]. In order to take the solid material properties into
consideration, the energy absorption per unit weight (kJ/kg) was
used. From Fig. 9, it is clear that the energy absorption per unit weight
of both DV1 and DV2 were significantly greater than most other
metal foams. It is also seen that with greater negative Poisson’s ratio,
the energy absorption ability of the auxetic lattice structure also
increases. Due to the stair stepping effect common to layer-based
manufacturing processes, the tested structures were expected to be
actually weaker than the ideal case, which suggested that the auxetic

Fig. 6. Comparison of bending properties between auxetic lattice and regular


sandwich panels (a) normalized bending modulus; (b) normalized flexural
strength. Fig. 7. Layer collapse during compressive test.
584 L. Yang et al. / Materials Science & Engineering A 558 (2012) 579–585

Fig. 8. Comparison of compressive properties of various structures (a) normalized modulus; (b) normalized strength.

lattice would exhibit even better mechanical properties. Overall, the


discussion showed that the auxetic lattice structure holds great
potential for future energy absorption applications.

6. Conclusions

Non-stochastic auxetic lattice structures were manufactured


with two different geometric parameter variations. Bending and
compression tests were conducted for both design variations. The
improvement of bending properties was so significant that
auxetic structures without solid skins on the upper and lower
faces were able to match the strength of the regular sandwich
panel structures. On the other hand, the compressive test showed
Fig. 9. Comparison of energy absorption per unit weight of various metal foams. less significant, though still promising, improvement. Due to the
L. Yang et al. / Materials Science & Engineering A 558 (2012) 579–585 585

stair stepping effect and other fabrication induced defects, the [17] H. Wan, H. Ohtaki, S. Kotosaka, G. Hu, Eur. J. Mech. A—Solid. 23 (2004)
actual structure did not exhibit as high compressive strength and 95–106.
[18] D. Roylance, Beam displacements. Massachussetts institute of technology
energy absorption as expected. However, structures with negative open courseware, Mech. Mater. 10 (2000).
Poisson’s ratios were experimentally proven to possess excellent [19] O. Cansizoglu, Mesh Structures with Tailored Properties and Applications in
mechanical properties that warrant further study and process Hip Stems, Ph.D. Dissertation, North Carolina State University, Raleigh, 2008.
development. [20] V. Crupi, R. Montanini, Int. J. Impact Eng. 34 (2007) 509–521.
[21] C. Chen, A.-M. Harte, N.A. Fleck, Int. J. Mech. Sci. 43 (2001) 1483–1506.
[22] M. Styles, P. Compston, S. Kalyanasundaram, Compos. Struct. 80 (2007)
References 532–538.
[23] K.R. Kabir, T. Vodenitcharova, M. Hoffman, Int. J. Mod. Phys. B 23 (2009)
1733–1738.
[1] R. Lakes, Sci. 235 (1987) 1038–1040. [24] C.-H. Lim, I. Jeon, K.-J. Kang., Mater. Des. 30 (2009) 3082–3093.
[2] R.S. Lakes, R. Witt, Int. J.Mech. Eng. Educ. 30 (2002) 50–58. [25] P. Schaffler, G. Hanko, H. Mitterer, P. Zach, In: Proceedings of 5th Inter-
[3] F. Scarpa, F.C. Smith, J. Intel. Mat. Syst. Str. 15 (2004) 973–979. national Conference on Porous Metal and Metallic Foams, Montreal, Canada,
[4] A. Bezazi, F. Scarpa, Int. J. Fatigue 29 (2007) 922–930.
September 2007, pp. 7–10.
[5] F. Scarpa, P. Pastorino, A. Garelli, S. Patsias, M. Ruzzene, Phys. Status Solidi B
[26] T. Miyoshi, M. Itoh, S. Akiyama, A. Kitahara, Adv. Eng. Mater. 2 (2000) 179.
242 (2005) 681–694.
[27] L.J. Vendra, A. Rabiei, Mat. Sci. Eng. A 465 (2007) 59–67.
[6] R.S. Lakes, K. Elms, J. Comput. Math. 27 (1993) 1193–1202.
[28] Duocel Aluminum Foam, ERG Materials and Aerospace Corporation Product
[7] K.E. Evans, J. Mater. Sci. 30 (1995) 3319–3332.
Specification Data Sheet, 2011.
[8] R.S. Lakes, ASME J. Mech. Des. 115 (1993) 696–700.
[29] E. Andrews, W. Sanders, L.J. Gibson, Mat. Sci. Eng. A 270 (1999) 113–124.
[9] F. Scarpa, P.J. Tomlin, Fatigue Fract. Eng. Mater. Struct. 23 (2000) 717–720.
[30] J. Zhou, P. Shrotriya, W.O. Soboyejo, Mech. Mater. 36 (2004) 781–797.
[10] W. Yang, Z.-M. Li, W. Shi, B.-H. Xie, M.-B. Yang, J. Mater. Sci. 39 (2004)
[31] W.-Y. Jang, S. Kyriakides, Int. J. Solids Struct. 46 (2009) 617–634.
3269–3279.
[32] D. Ruan, G. Lu, F.L. Chen, E. Siores, Compos. Struct. 57 (2002) 331–336.
[11] O. Cansizoglu, O. Harrysson, D. Cormier, H. West, T. Mahale, Mater. Sci. Eng. A
[33] T.G. Nieh, K. Higashi, J. Wadsworth, Mat. Sci. Eng. A 283 (2002) 105–110.
492 (2008) 468–474.
[34] A.E. Simone, L.J. Gibson, Acta Mater. 46 (1998) 3109–3123.
[12] T.L. Warren, J. Appl. Phys. 67 (1990) 7591–7594.
[35] V.S. Dashpande, N.A. Fleck, M.F. Ashby, J. Mech. Phys. Solids 49 (2001)
[13] U.D. Larsen, O. Sigmund, S. Bouwstra, J. Microelectromech. S. 6 (1997)
1747–1769.
99–106.
[36] J. Wang, K.E. Evans, K. Dharmasena, H.N.G. Wadley, Int. J. Solids Struct. 40
[14] C.W. Smith, J.N. Grima, K.E. Evans, Acta Mater. 48 (2000) 4349–4356.
[15] N. Gaspar, X.J. Ren, C.W. Smith, J.N. Grima, K.E. Evans, Acta Mater. 53 (2005) (2003) 6981–6988.
2439–2445. [37] Y.-H. Lee, B.-K. Lee, I. Jeon, K.-J. Kang., Acta Mater. 55 (2007) 6084–6094.
[16] D. Prall, R.S. Lakes, Int. J. Mech. Sci. 39 (1996) 305–314. [38] R. Montanini, Int. J. Mech. Sci. 47 (2005) 26–42.

You might also like