You are on page 1of 14

Modelling and Simulation in Materials Science and Engineering

Related content
- Empirical Relationship between Relative
A multi-cell FE-model for compressive behaviour Electrical Conductivity and Relative
Density of the Al-Foam Fabricated through
analysis of heterogeneous Al-alloy foam Pressure Assisted Sintering/Dissolution
Process
Mazli Mustapha, Fauzi Ismail and Othman
Mamat
To cite this article: A Kim et al 2006 Modelling Simul. Mater. Sci. Eng. 14 933
- Geometry and bluntness tip effects on
elastic–plastic behaviour during
nanoindentation of fused silica:
experimental and FE simulation
D Torres-Torres, J Muñoz-Saldaña, L A
View the article online for updates and enhancements. Gutierrez-Ladron-de Guevara et al.

- Fracture dynamics of polymeric foams


S Deschanel, L Vanel, N Godin et al.

Recent citations
- Mechanical properties of regular porous
biomaterials made from truncated cube
repeating unit cells: Analytical solutions
and computational models
R. Hedayati et al

- Study on Convergence Technique through


Strength Analysis of Stabilizer Link by
Type
Jae-Ung Cho

- Study on Convergence Technique through


Structural Analysis on the Axle of Railway
Vehicle
Jae-Ung Cho

This content was downloaded from IP address 103.27.9.50 on 10/11/2017 at 15:18


INSTITUTE OF PHYSICS PUBLISHING MODELLING AND SIMULATION IN MATERIALS SCIENCE AND ENGINEERING
Modelling Simul. Mater. Sci. Eng. 14 (2006) 933–945 doi:10.1088/0965-0393/14/6/004

A multi-cell FE-model for compressive behaviour


analysis of heterogeneous Al-alloy foam
A Kim1,2 , K Tunvir1 , G D Jeong1 and S S Cheon1
1 Division of Mechanical and Automotive Engineering, Kongju National University, 182

Shinkwan-dong, Kongju, Chungnam 314-701, Korea

E-mail: amkee@kongju.ac.kr (Kim), kazitun@kongju.ac.kr (Tunvir), kvskddo@kongju.ac.kr


(Jeong) and sscheon@kongju.ac.kr (Cheon)

Received 20 January 2006


Published 12 June 2006
Online at stacks.iop.org/MSMSE/14/933

Abstract
Compressive mechanical behaviours of closed cell Al-alloy foams produced
by melt based and powder metallurgical methods were investigated by a finite
element model composed of multiple unit lattices. The unit lattice consists
of spherical and cubic sections possessing a thickness ratio between them. A
Gaussian distribution of the relative density among the lattices and the random
allocation of lattices were implemented in the model to address the structural
heterogeneity. The constitutive relation for the lattice material was determined
by a nondestructive instrumented sharp indentation test on the cell wall. The
simulated compressive stress–strain curves for the ductile Al–Si–Ca foam were
found to be in good agreement with the experimental ones over the entire strain
range while the resistance of brittle Al–Si–Cu–Mg foam to the deformation
turned out to be lower than that predicted by the model after about 0.50 strain.
The discrepancy between experimental and simulation results for Al–Si–Cu–
Mg foam in high strain range was associated with the disintegration of cell
walls broken by the large deformation.
(Some figures in this article are in colour only in the electronic version)

1. Introduction

Metal foams have attracted a growing research interest in automotive and aerospace industries
due to their ultra-light weight and other attractive mechanical characteristics such as high
specific strength, excellent impact energy absorption, good damping and sound absorption.
In particular, their crashworthiness in structural application is mainly due to the deformation
characteristics to absorb the impact energy in a controlled manner. Al foam-filled composite
panels are widely being considered these days for structural use in aircrafts and satellites,

2 Author to whom any correspondence should be addressed.

0965-0393/06/060933+13$30.00 © 2006 IOP Publishing Ltd Printed in the UK 933


934 A Kim et al

and the functional use in automotive bumpers, roofs and door beams for improving the
crashworthiness of vehicles which will ultimately reduce the damage of vehicles and increase
the safety of the passengers during crushing. Thus, the characterization of compressive
behaviour is one of the main tasks for the industrial application of Al foams in various fields.
It can be achieved by the model and the related numerical tools that can reliably account
various mechanical aspects of foams. A great exertion has been accomplished towards the
development of numerical models for mechanical behaviour analysis of Al foams.
Gibson and Ashby [1] predicted the initial bending collapse of the cubic skeleton model
with the theoretical prediction on the crushing behaviour of closed cell foam structure. A
tetrakaidecahedral unit cell model was introduced by Simone and Gibson [2], and the effects
of cell face curvature and corrugation of closed cell foams on the elastic modulus and plastic
collapse stress were investigated through finite element (FE) analysis. Overaker et al [3]
assessed some conclusions regarding the elastic behaviour and the effect of morphology of
cell on elastic behaviour by constructing a 2D hexagon foam. To represent the cellular material
a visco-elastic FE model has been introduced by Hučko and Faria [4]. Extensive work has been
done by Santosa and Weirzbicki [5] in approximating the closed cell Al-alloy foam structure
as an assembly of large and small closely packed cells. The analytical formulation was derived
based on energy consideration, and the numerical simulation was performed to validate the
kinematics of model with the truncated cube-pyramidal closely packed model. In succession
to Santosa and Weirzbicki’s [5] work, Meguid et al [6] proposed the cruciform-hemisphere
model with a hemispherical section instead of a pyramidal section considering that closed
cells of Al foam are spherical. Kenesei et al [7] investigated the changes in the plateau
stress and energy absorption of metal foam during compression with a simple model using
Gibson and Ashby [1] cubic arrangement in which the cell size distribution was assumed to
be Gaussian. The cell size distribution in their study was found to be not so effective on the
plateau stress and energy absorption capacity while the influence of relative density and ratio
of maximum to minimum cell size were found to be significant. Recently, real 3D structures
of the open cell and closed cell foams using CT (computer tomography) technology were well
accomplished [8–10]. Benouali et al [10] found the hydro foam to be anisotropic in cell size,
shape and orientation, but the LKR foam to be isotropic. Both foams were found to possess
heterogeneities and imperfections in their structures. Indeed the detailed foam structures would
provide a decent structure for modelling the heterogeneous Al-alloy foams.
In this paper, a multiple lattice FE model with a thickness ratio between sections in the
lattice was proposed for the compressive behaviour analysis of globally isotropic but locally
heterogeneous closed cell Al-alloy foam. The merit of the lattice model is that geometrical
parameters such as wall thickness, ratio of thickness to lattice width and the size ratio between
sections associated with the relative density easily allow one to characterize the mechanical
properties of materials. A Gaussian distribution and the random allocation of lattice density
in the foam were adopted to realize the local structural heterogeneities as well as to retain the
global isotropy of the foam. In addition, the simulated compressive stress–strain behaviours
along with the localization pattern of deformation were validated by the experimental
results.

2. Modelling

2.1. Unit lattice


Figure 1 shows a typical optical image of closed cell Al-alloy foam. It reveals that the foam
is an array of cells having a wide spectrum of size. Thus, a structure consisting of a number
A multi-cell FE-model for compressive behaviour analysis of Al-alloy foam 935

Figure 1. A typical image of closed cell Al-alloy foam.

Figure 2. Structure of foam model: (i) two-dimensional morphology and (ii) isometric view of
unit cubic lattice.

of densely packed cells can be considered to represent the foam. The small cells are thicker
than the big cells, indicating that more material has flown into the small cell surrounded by the
large cells during foaming. Based on this observation, a unit lattice is proposed as shown in
figure 2(b). The unit lattice is a cube truncated by a 1/8 sphere at all corners. The thicknesses
of cubic and spherical sections are t1 and t2 , respectively.
From the geometrical relation, the relative density (ρf /ρs ) of a cubic–spherical lattice can
be calculated by the volume fraction of skeleton in a cube as follows:
      2  
ρf π D 2 t1 D t2
=3 1− +π . (1)
ρs 4 W W W W
Here ρf and ρs are densities of foam and solid, respectively, from which the foam was
made. W is the width of lattice and D is the diameter of spherical section. Using (1), the
thickness ratio (t2 /t1 ) can be calculated by the following equation:
          2 −1
t2 ρf W π D 2 D
= −3 1− π . (2)
t1 ρs t1 4 W W
936 A Kim et al

Figure 3. FE foam model.

If the thickness ratio of the spherical section to the cubic section is 1 with a constant cell
size ratio (D/W ) of 0.5, the model becomes similarly to the cruciform-hemispherical lattice
previously proposed by Meguid et al [6]. The scaling law for densification of closed cell
Al-alloy foam has been suggested by Ashby et al [11] as the following:
    3 
ρf ρf
εd = 0.9 1 − 1.4 + 0.4 . (3)
ρs ρs
The densification strain (εd ) of the cubic-spherical model can also be calculated with
D/W using the following equation (see appendix A):
D
εd = 1 − 0.54 . (4)
W
Thus, substituting (3) into (4), the D/W of lattice is determined by the following equation:
   3
D ρf ρf
= 0.185 + 2.33 − 0.67 . (5)
W ρs ρs

2.2. Finite element model for the foam


The complete FE model used for the analysis of compressive behaviour of Al-alloy foam is
shown in figure 3. The FE model was designed and analysed by the nonlinear explicit FE
A multi-cell FE-model for compressive behaviour analysis of Al-alloy foam 937

Figure 4. Discrete Gaussian distribution for a mean relative density of 0.12 with various standard
deviations.

code LS-DYNA 3D. The model consists of 80 unit lattices (4 units × 4 units × 5 units). The
four-noded Belytschko–Tsay Lin shell element having six degrees of freedom per node (three
translations and three rotations) with reduced integration and hourglass control was used. The
number of elements used in each lattice was 1164. A piecewise linear plastic material model in
LS-DYNA 3D material library was used [12]. The contact between the upper surface of the FE
model and the compression plate was modelled with surface-to-surface sliding contact. During
progressive deformation and densification, the interpenetration of sections was prevented by
using self-contact interface. One translational movement of the top surface was allowed with
five degree constraints while six degree constraints were applied to the bottom of the model.
Quasi-static solution by an explicit time integration method sometimes accumulates errors
(in stress distribution and contact phenomena) because of the dynamic effects. In order to avoid
these effects, the loading velocity in simulation was ramped from 0 to 0.02 mm s−1 during the
first 5 ms time period and then kept constant. The material mass density was scaled up with a
factor of 102 to reduce the computation time [13].

2.3. Density variation in heterogeneous foam

The foams possess morphological imperfections such as curved and wrinkled cell walls
(waviness and corrugation of the cell walls), non-uniform wall thickness, non-uniform cell
size distribution, fractured cell walls, missing cell walls, etc [14–16]. All those heterogeneities
were assumed to be a density variation in the foam and were modelled simply by varying the
density of lattice statistically in the model. The Gaussian distribution of the lattice density
was accomplished using a mean value and the standard deviation. Figure 4 shows the discrete
Gaussian distribution with a mean relative density (ρf /ρs = 0.12) of foam and various standard
deviations (0.024, 0.048 and 0.065). As an example, table 1 represents the numbers of lattices
assigned to various density levels in case of relative density being 0.12 and standard deviation
0.024. Finally the equal probability random process was utilized for the allocation of lattices
in the foam model without any bias [17].
938 A Kim et al

Table 1. Number of cells corresponding to each density level in the multiple cell model calculated
with standard deviation of 0.024.
Relative Density Level Probability Number of cells
<0.06 0.0031 —
0.06 0.0073 1
0.07 0.0190 2
0.08 0.0415 3
0.09 0.0761 6
0.10 0.1175 9
0.11 0.1524 12
0.12 0.1663 14
0.13 0.1524 12
0.14 0.1175 9
0.15 0.0761 6
0.16 0.0415 3
0.17 0.0190 2
0.18 0.0073 1
>0.18 0.0031 —
Total 1.0000 80

3. Experimental

3.1. Instrumented sharp indentation test on cell wall


The constitutive relationship for the lattice material was evaluated by the nondestructive
instrumented sharp indentation test on the thin cell wall of foam. Melt based Al–Si–Ca
foam and powder metallurgically made Al–Si–Cu–Mg foam were studied. Al–Si–Ca foam
was provided by the commercial manufacturer, FoamTech Ltd in Korea. The details of Al–
Si–Cu–Mg foams production can be found in [18]. The instrumented sharp indentation test
was performed on the cell walls using a MTS nano-indenter XP with a Berkovich tip indenter.

3.2. Uni-axial compression test of Al-alloy foam


Specimens for compression test were cut to the dimensions of 35 mm × 35 mm × 40 mm using
an electrical discharge cutting machine. Such a dimension was chosen so that the edge length
in all cases was at least seven times the cell size. This is required to avoid the edge effect that
may reduce the measured elastic modulus and compressive strength of the foam. The relative
density of Al–Si–Ca foam specimen was 0.12 while that of Al–Si–Cu–Mg foam specimen was
0.16. The MTS 810 machine was used to perform the uni-axial compression test. The load and
displacement were monitored by a computer equipped with a data acquisition system. In the
test, the load was applied at a constant displacement speed of 0.02 mm s−1 , and the specimens
were compressed between parallel steel plates to ensure the axial loading. The compression
was stopped when about 0.85 strain was reached.

4. Result and discussion

4.1. Determination of stress–strain relation of lattice wall material


The initial portion of the unloading curve and the loading curvature of load-indentation depth
curve were used to evaluate the elasto-plastic parameters of the cell wall material such as elastic
A multi-cell FE-model for compressive behaviour analysis of Al-alloy foam 939

Figure 5. Experimental and simulated load-indentation depth curves for Al–Si–Cu–Mg and
Al–Si–Ca foam cell wall materials.

modulus E, yield stress σy and strain hardening exponent n (assuming that the elasto-plastic
properties follow Hook’s law, σ = Eε, before the yield point and power law, σ = Kε n , after
that). In order to evaluate these parameters from the load-indentation depth curve, the reverse
algorithm based on dimensionless functions which relate the load-indentation depth curve to
the elasto-plastic parameters of the indented material proposed by Dao et al [19] was used.
To verify the validity of extracted material parameters, FE analysis of the indentation was
carried out by an axisymmetric semi-infinite 2D model through the FE code ABAQUS using
the average of extracted material parameters. The average material parameters were extracted
from five different load-indentation depth curves for each of the foams. The simulated and
experimental load-indentation depth curves for the cell wall materials are shown in figure 5.
The experimental load-indentation depth curves for both foams scattered, which might be
caused by the micro-structural heterogeneity in the foams. Nevertheless, the simulated curves
were within the scatter range of the experimental curves. The extracted average material
parameters for Al–Si–Cu–Mg foam cell wall were E = 72 GPa, σy = 242 MPa and n = 0.14,
while those for Al–Si–Ca foam cell wall were E = 78 GPa, σy = 85 MPa and n = 0.12. The
power law true stress–strain curves based on the average extracted material parameters are
shown in figure 6.

4.2. Determination of thickness ratio


The cell wall thickness of Al-alloy foam is varied with the spatial position in the foam. Since
a unit lattice of the foam model consists of a truncated cube and a sphere, one may assume
that the big cells play a role of cube and the rest of them represent a sphere in lattice. Thus,
the width of cube (W ) and the diameter of sphere (D) can be approximated by the average
diameters of upper 50% cells in size and the rest of them, respectively. The diameter and wall
thickness were measured from the 2D image of foams using an image analysing software,
BMI plus, version 4.0 (Winatech Co. Ltd., Korea), installed in computer equipped with a
digital microscope. The binarization of image was performed in such a way that the cell
walls are black while the rest of the image, i.e. the pores, is white as shown in figure 7. The
940 A Kim et al

Figure 6. Constitutive relations for Al–Si–Ca and Al–Si–Cu–Mg foam cell wall materials.

Figure 7. Binarized image used for cell size and wall thickness measurement: (a) Al–Si–Ca foam
and (b) Al–Si–Cu–Mg foam.

average diameters of white islands in the figure surrounded by black region were measured.
The average wall thickness of a big cell for each foam was also measured along the cell
wall to represent t1 as shown in figure 7. W , t1 , D/W and t2 /t1 by (2) are contained in
table 2. D/W by (5) was used instead of the measured one in this study because the values
A multi-cell FE-model for compressive behaviour analysis of Al-alloy foam 941

Table 2. Cell morphology measurement for Al–Si–Ca and Al–Si–Cu–Mg foams.

D/W t2 /t1
Foams ρf /ρs W (mm) t1 (mm) t1 /W D/W equation (5) equation (2)
Al–Si–Ca 0.12 4.93 0.150 0.030 0.45 0.46 2.3
Al–Si–Cu–Mg 0.16 3.88 0.137 0.035 0.48 0.56 2.9

by the equation were easily obtained only with the relative density and also close to the
measured ones.

4.3. Compressive behaviour of homogeneous Al-alloy foam model


The homogenous Al-alloy foam was realized by allocating the same density to all lattices in the
model. Figure 8 represents the compressive stress–strain curves obtained by the homogeneous
model as well as the experimental ones. The simulations were performed for three different
thickness ratios for Al–Si–Ca and Al–Si–Cu–Mg foams. After the yield, the oscillating stress–
strain curves were maintained by the successive buckling of the cell sections over the wide
compressive strain range. It was manifest that the model with t2 /t1 = 1 (Meguid’s lattice
in [6]) over-estimated the yield stress and the plateau stress approximately 2–3 times higher
than the experimental ones. The foam resistance against the deformation (plateau stress level)
was lowered by the increment in thickness ratio. In particular, the thickness ratios (t2 /t1 = 2.3
for Al–Si–Ca foam and t2 /t1 = 2.9 for Al–Si–Cu–Mg foam) in table 2 provided the good
agreement of plateau stress level between the experiment and the simulation. However,
although the FE simulation provided the plateau stress levels close to experimental results,
the differences in the initial yield point, the slope and oscillation of stress–strain curve in
the plateau region and densification between the simulation and the experiment were still
significantly recognized.

4.4. Compressive behaviour of heterogeneous Al-alloy foam model


Three different standard deviations of 0.024, 0.048 and 0.065 for each mean relative density
were utilized for the simulation (the mean relative densities of 0.12 for Al–Si–Ca foam and
0.16 for Al–Si–Cu–Mg foam). The higher standard deviation reflects the higher level of
heterogeneity. Figure 9 represents the compressive stress–strain curves simulated by the FE
foam model and the experimental curves. The yield stress decreased as the standard deviation
increased. Moreover, the higher variation of lattice density provided the steeper positive slope
and the less oscillation in the plateau region of curve, which allows the model to better imitate
the experimental curves. The standard deviation of 0.065 for Al–Si–Ca foam provided the
best match in comparison with the experiment curve in the entire strain range. The standard
deviation of 0.048 for the Al–Si–Cu–Mg foam also provided the well matched result except
for the departure from the experimental curve after about 0.50 strain (figure 9(b)). The easy
disintegration in the brittle cell wall of the Al–Si–Cu–Mg foam caused by the high strain might
keep the plateau stress less than that simulated by the model because the FE simulation could
not take into account the fracture in the deformed solid.

4.5. Deformation pattern of foam


Figures 10 and 11 show the deformation patterns of ductile Al–Si–Ca foam and brittle Al–Si–
Cu–Mg foam obtained by the experiment and the simulation by heterogeneous FE foam model.
942 A Kim et al

Figure 8. Stress–strain curves obtained by the homogeneous FE foam model: (a) Al–Si–Ca foam
and (b) Al–Si–Cu–Mg foam.

The collapse in the model and the real foam was initiated at the weakest spot, i.e. the lowest
density lattice in the model. The weakest lattice was locked up after being fully crushed, and
the deformation progressed into the second weakest lattice in a sequence. Thus, the localized
deformation pattern due to randomly distributed weak lattices in the model was manifested.
In particular, the edge of the deformed real Al–Si–Cu–Mg foam (at 0.55 strain in figure 11(a))
showed much disintegration of cells fractured by the concentration of deformation indicating
that the reduction in the compressive resistance of the Al–Si–Cu–Mg foam was caused by the
disintegration in the cell walls as discussed in the previous section.

5. Conclusion

The heterogeneous closed cell Al-alloy foam was realized by varying the lattice density in
the cubic-spherical FE foam model with a thickness ratio. It turned out to be very effective
to predict the compressive stress–strain curves for the closed cell Al-alloy foams. The local
A multi-cell FE-model for compressive behaviour analysis of Al-alloy foam 943

Figure 9. Stress–strain curves obtained by the heterogeneous FE foam model: (a) Al–Si–Ca foam
and (b) Al–Si–Cu–Mg foam.

density variation in the model allowed the model to imitate the deformation localization of the
closed cell Al-alloy foam precisely. However, the statistical aspects of density distribution in
the foam need to be characterized for the proper use of the heterogeneous foam model.

Acknowledgments

This research was sponsored by the Korea Science and Engineering Foundation (KOSEF)
under Grant No R01-2002-000-00093-0(2002) from its basic research program. The authors
of this paper wish to acknowledge the financial support of KOSEF.

Appendix A

It is assumed that during the crushing of the spherical section, ABCE hinge line (figure A1)
is created along the circumference of the spherical section, and the length of arc OA (OB,
944 A Kim et al

Figure 10. Deformation patterns at 0.10, 0.25, 0.40 and 0.55 strain for Al–Si–Ca foam: (a)
experiment and (b) simulation of heterogeneous model with relative density of 0.12, st. dev. of
0.065 and t2 /t1 = 2.3.

Figure 11. Deformation patterns at 0.10, 0.25, 0.40 and 0.55 strain for Al–Si–Cu–Mg foam: (a)
experiment and (b) simulation of heterogeneous model with relative density of 0.16, st. dev. of
0.048 and t2 /t1 = 2.9.

OC, OE) is equal to D/2. The portion of the spherical section above the hinge line starts to
rotate about the hinge line, and for the compatibility the adjacent cubic cell wall (AOXRQK)
continues to fold, making hinge lines OAK, OX, AR and KQ. Finally, the spherical section
takes the shape of a cylinder of diameter D at the onset of densification making a plane ABCE
parallel to KLMN (figures A1(b) and (c)). Thus the height of the cylinder at the onset of
densification is as follows:

CM = 0.27 · D. (A.1)
A multi-cell FE-model for compressive behaviour analysis of Al-alloy foam 945

Figure A1. Geometry of the spherical section: (a) before the crushing, (b) at the onset of the
densification and (c) by FE simulation at the onset of densification.

Since the lattice possesses two hemi-spherical sections along the width (W ), the shortening of
the lattice at onset of densification is as follows:
Shortening = W − 2 · 0.27 · D = W − 0.54 D. (A.2)
Hence, the densification εd defined as a ratio of the shortening to the lattice width at the onset
of densification becomes
D
εd = 1 − 0.54 . (A.3)
W

References

[1] Gibson L J and Ashby M F 1997 Cellular Solids: Structure and Properties (Cambridge, UK: Cambridge
University Press)
[2] Simone A E and Gibson L J 1998 Acta Mater. 46 3929–35
[3] Overaker D W, Cuitiño A M and Langrana N A 1998 Mech. Mater. 29 43–52
[4] Hučko B and Faria L 1997 Comput. Struct. 62 1049–57
[5] Santosa S and Wierzbicki T 1998 J. Mech. Phys. Solids 46 645–69
[6] Meguid S A, Cheon S S and Abbasi N EI 2002 Finite Elem. Anal. Des. 38 631–43
[7] Kenesei P, Kádár C, Rajkovits Z and Lendvai J 2004 Scr. Mater. 50 295–300
[8] Wicklein M and Thoma K 2005 Mater. Sci. Eng. A 397 391–9
[9] Dillard T, N’guyen F, Maire E, Salvo L, Forest S, Bienvenu Y, Bartout J D, Croset M, Dendievel R and Cloetens
P 2005 Phil. Mag. 85 2147–75
[10] Benouali A H, Froyen L, Dillard T, Forest S and N’guyen F 2005 J. Mater. Sci. 40 5801–11
[11] Ashby M F, Evans A G, Gibson L J, Hutchinson J W, Fleck N A and Wadley H G N 1998 Metal Foams: A
Design Guide (Burlington, MA: Heinemann, Butterworth) p 53
[12] 2001 LS-DYNA Keywords User’s Manual vol 1, version 960, Livermore Software Technology Corporation
[13] Santosa S, Wierzbicki T, Hanssen A G and Langseth M 2000 Int. J. Imp. Eng. 24 509–34
[14] Markaki A E and Clyne T W 2001 Acta Mater. 49 1677–86
[15] Koza E, Leonowicz M, Wojciechowski S and Simancik F 2003 Mater. Lett. 58 132–5
[16] Chen C, Lu T J and Fleck N A 1999 J. Mech. Phys. Solids. 47 2235–72
[17] Lyman O 1988 An Introduction to Statistical Methods and Data Analysis ed M Payne (Boston: PWS-KENT
Publishing Company) p A 23
[18] Kim A, Cho S S and Lee H J 2004 Mater. Sci. Technol. 20 1615–20
[19] Dao M, Chollacoop N, Van Vliet K J, Venkatesh T A and Suresh S 2001 Acta Mater. 49 3899–918

You might also like