Available online at www.sciencedirect.

com

Acta Materialia 56 (2008) 3338–3343

www.elsevier.com/locate/actamat

Evaluating plastic flow properties by characterizing indentation

size effect using a sharp indenter
Ju-Young Kim a,*, Seung-Kyun Kang b, Julia R. Greer a, Dongil Kwon b
a
Materials Science, California Institute of Technology, Pasadena, CA 91125, USA
b
Department of Materials Science and Engineering, Seoul National University, Seoul 151-744, Republic of Korea

Received 6 December 2007; received in revised form 28 January 2008; accepted 5 February 2008
Available online 21 April 2008

Abstract

The strain-hardening exponent, used in describing the plastic flow properties of materials, is evaluated from the characteristic length
in the indentation size effect (ISE). A linear relationship is found between the strain-hardening exponent and the log of the ISE charac-
teristic length for Ni and SCM21 (structural steel) samples with different plastic pre-strain values. These results are explained through the
Taylor dislocation hardening model and a representative stress–strain approach. A dimensionless function characterizing the plastic
deformation using only parameters generally measurable by nanoindentation testing is also proposed. The feasibility of developing a
unique dimensionless function is studied for 22 metals.
Ó 2008 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Indentation size effect; Nanoindentation; Strain-hardening exponent; Plastic flow properties

1. Introduction properties are measured. Since the experimental procedure

for nanoindentation is simple and nondestructive, and
Nanoindentation testing has been used widely to mea- places no limits on the sample size and geometry, using
sure the mechanical properties of materials at small scales, nanoindentation to measure plastic flow properties has
by which hardness and elastic modulus of materials are attracted much research interest [11–22].
generally evaluated [1–8]. Hardness is a measure of the A representative stress–strain approach has been pro-
resistance of a material to plastic deformation, which is posed to measure the yield stress, ultimate tensile strength
sometimes related to its flow stress. Tabor demonstrated and strain-hardening exponent of metals by defining a rep-
the relation between hardness and yield strength [9], and resentative stress and strain beneath a spherical indenter
the relation between hardness and flow stress at a repre- [11,12]. This approach exploited the idea that the various
sentative strain beneath the indenter has been established representative strain states induced in a material beneath
by many experimental and finite-element method (FEM) a spherical indenter depend on the indentation depth.
simulation results [10–22]. The elastic modulus is also However, the value of using a spherical indenter at the
measured by nanoindentation techniques [2], and thus nanometer scale is still a subject of some debate, since it
the mechanical properties of a material usually derived is difficult to correlate the indentation size effect (ISE) with
from the uniaxial tensile stress–strain curve can be evalu- an indenter radius as well as indentation depth [23,24].
ated by nanoindentation testing alone if the plastic flow Jayaraman et al. proposed a way to measure plastic prop-
erties by fitting a constitutive equation to two different rep-
resentative stress–strain points obtained by two sharp
*
Corresponding author. indenter geometries: Berkovich and cube-corner indenters
E-mail address: [email protected] (J.-Y. Kim). [16]. While better results are expected when the plastic flow

1359-6454/$34.00 Ó 2008 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.actamat.2008.02.049
 J.-Y. Kim et al. / Acta Materialia 56 (2008) 3338–3343 3339

curve is fitted to more representative stress–strain data length is analyzed via a Taylor dislocation hardening
measured by the various sharp indenters with different model and the representative stress–strain approach. We
angles, the effectiveness of this technique is limited due to report this relationship for 22 different metals.
the many experiments required. Algorithms using FEM
have also been proposed [13,14,17–22], and inverse analysis 2. Experiments
[18,19] shows that universal dimensionless functions are set
up by many forward analyses, by which the indentation The strain-hardening exponents of samples were inten-
force–displacement curve is determined from the stress– tionally controlled by interrupted tensile tests. Uniaxial
strain curve using FEM simulations, and a material’s tensile tests were performed on pure polycrystalline Ni
stress–strain curve is determined inversely by inputting and SCM21 at a displacement rate of 1 mm min1 until
the parameters of the indentation force–displacement curve necking occurred. On the basis of the stress–strain curve,
and indenter geometry into these dimensionless functions. tensile tests were interrupted at true strain values of 5%,
These FEM simulation-based methods are, however, 10% and 15% for Ni, and 2% and 5% for SCM21. The
strongly dependent on the particular FEM simulation uniformly deformed gauge sections were cut and gently
and dimensionless functions used. The experimental evalu- polished up to 0.3 lm alumina powder. Nanoindentation
ation of plastic flow properties by nanoindentation testing tests were conducted on undeformed and 5%, 10% and
using a single sharp indenter, as discussed here, thus offers 15% pre-deformed Ni samples, as well as on undeformed
clear advantages over existing techniques. and 2% and 5% pre-deformed SCM21 samples. The con-
One of the key parameters representing plastic flow tinuous stiffness measurement (CSM) technique of the
properties is the strain-hardening exponent, n, for those Nanoindenter XP (MTS Inc., TN, USA) using a three-
materials that obey the Hollomon equation, r = Ken, where sided pyramidal Berkovich indenter was used in all nan-
r is the true stress, e is the true strain and K is the strength oindentation tests. The loading/unloading rate was con-
coefficient [25]. Most materials show a linear dependence stant at 0.2 mN s1; the maximum load was 100 mN,
with the slope corresponding to the strain-hardening expo- and a drift rate below 0.05 nm s1 was allowed. To estab-
nent when true tensile stress is plotted vs. strain on a log– lish the relationship between the strain-hardening expo-
log scale. The present work began by considering the pos- nent and the ISE characteristic length, 22 metals were
sibility of a relationship between the strain-hardening subjected to uniaxial tensile tests and nanoindentation
exponent, n, and the ISE characteristic length, h*. The tests under the conditions described above. The samples
ISE is an increase in hardness with decreasing indentation were general structural steels (S45C, SKD11, SKH9,
depth observed in numerous nanoindentation experiments SKS3, SKD61, SCM4), cast irons (CI300, CI600), stain-
[23,24,26–45]. It is believed to be associated with the geo- less steels (STS 303, STS 304, STS 316L), API X-grade
metrically necessary dislocations (GNDs) induced by an steels (API X65, X70, X80), Ti alloys (pure Ti, Ti–6Al–
inhomogeneous plastic deformation. Nix and Gao ana- 6V, Ti–5Al–2.5Sn, Ti–10V–2Fe–3Al), Al alloys (Al2011,
lyzed the ISE in crystalline materials by calculating the Al6061, Al7076) and pure polycrystalline Cu. Their
density of GNDs around a sharp indenter based on a elastic moduli, Poisson’s ratios and shear moduli were
Taylor dislocation hardening model [27]; Gao et al. and measured by an ultrasonic pulse-echo technique using a
Huang et al. reformulated the mechanism-based strain gra- two-channel digital real-time oscilloscope for the theoret-
dient plasticity [29,30], and Swadener et al. extended the ical calculation of the representative stress induced by
Nix–Gao model to a spherical indenter, taking the inden- nanoindentation.
ter’s geometry as a parabola [23]. Although the presence
of local strain gradients producing a higher GND density 3. Results and discussion
is the key explanation for the ISE, while any increase
in strength during uniaxial deformation is caused by homo- Fig. 1 shows the tensile true stress–strain curves for
geneous hardening from an increase in the density of statis- undeformed Ni and SCM21; the strain-hardening expo-
tically stored dislocations (SSDs), Elmustafa et al. verified, nents n and strength coefficients K of undeformed and
by analyzing trends in the activation volume with hardness, pre-deformed samples are presented in Table 1. These val-
that the ISE is driven by the same dislocation mechanisms ues were calculated from the uniaxial stress–strain curves
as strain hardening [36]. of undeformed samples by assuming that the stress–strain
This work examines the relationship between the strain- curves of pre-deformed samples have constant elastic mod-
hardening exponent and the ISE characteristic length. We ulus during unloading and reloading. The true stress–strain
perform interrupted uniaxial tensile tests on Ni and curves of undeformed Ni and SCM21 samples were found
SCM21 (structural steel), and thus obtain different values to be described well by the Hollomon equation. The strain-
of the strain-hardening exponent by intentional plastic hardening exponent n was found to decrease and the
pre-strain. The values of ISE characteristic lengths were strength coefficient K to increase with increasing plastic
measured by nanoindentation using a Berkovich indenter. pre-strain. Fig. 2 shows the hardness variation with inden-
The experimentally determined relationship between the tation depth. ISE characteristic lengths were calculated by
strain-hardening exponent and the ISE characteristic using the Nix–Gao model:
3340 J.-Y. Kim et al. / Acta Materialia 56 (2008) 3338–3343

a 5

500 Undeformed
5% pre-strain
10% pre-strain
400 4

Hardness (GPa)
True stress (MPa)

15% pre-strain

300

3
200

100 2

0
0 5 10 15 20 25 250 500 750 1000

True strain (%) Indentation depth (nm)

b 600 5
Undeformed
2% pre-strain
5% pre-strain
True stress (MPa)

450

Hardness (GPa)
4

300

150

2
0
0 3 6 9 12 15 250 500 750 1000
True strain (%) Indentation depth (nm)
Fig. 1. Uniaxial tensile stress–strain curves of undeformed (a) Ni and (b) Fig. 2. Hardness vs. indentation depth for (a) Ni and (b) SCM21.
SCM21.

featuring in other studies [27,28,44]. Hardness values mea-

rffiffiffiffiffiffiffiffiffiffiffiffiffi sured at indentation depths greater than 200 nm were used
H h
¼ 1þ ð1Þ to characterize the indentation size effect to alleviate the ef-
H0 h fect of the work-hardened layer formed by mechanical pol-
where H is the hardness, H0 is the macroscopic hardness ishing [31]. The macroscopic hardness H0 was found to
and h is the indentation depth (also shown in Table 1) increase and the ISE characteristic length h* to decrease
[27]. Note that ISE can be seen to be stronger in Fig. 2 than with increasing plastic pre-strain. The macroscopic hard-
the measured values in Table 1 simply because the sample ness and ISE characteristic length in the Nix–Gao model
used in Fig. 2 had a shallower indentation depth than that are given by

Table 1
Parameters measured from uniaxial tensile stress–strain curves and characterized for ISE, and theoretically calculated flow stress and strain-hardening
exponent values
Material Ni SCM21
Plastic pre-strain (%) 0 5 10 15 0 2 5
n 0.452 0.211 0.117 0.062 0.151 0.072 0.035
K (MPa) 864.14 601.84 541.68 511.53 902.66 751.36 702.97
h* (nm) 812.29 628.25 549.76 516.41 811.95 538.47 454.28
H0 (GPa) 1.339 1.506 1.709 1.819 1.468 1.719 1.873
(± 0.039) (±0.062) (±0.072) (±0.065) (±0.079) (±0.094) (±0.117)
Flow stress at 7% true strain (MPa) 252.48 331.43 392.48 440.72 563.47 570.82 576.46
Representative stress calculated 506.35 575.76 615.49 635.05 550.89 676.46 736.49
from h* (MPa)
Calculated n using Eq. (5) 0.203 0.017 0.048 0.082 0.206 0.059 0.001
 J.-Y. Kim et al. / Acta Materialia 56 (2008) 3338–3343 3341
pffiffiffi pffiffiffiffiffi
H 0 ¼ 3 3alb qs ð2aÞ to 0.07 for the Berkovich indenter [10]. Therefore, the flow
 2 stress values at 7% true strain were measured from the uni-
81 l
h ¼ ba2 tan2 h ð2bÞ axial tensile stress–strain curves, and the representative
2 H0 stress values in nanoindentation by the Berkovich indenter
where a is a geometric constant, l is the shear modulus, b is were subsequently calculated from Eq. (4) by using the
the magnitude of the Burgers vector, qs is the density of experimentally measured shear modulus, ISE characteristic
SSDs and h is the complementary angle of the conical in- length, and a = 0.5. Table 1 shows the relatively good
denter half-angle (19.7° for the Berkovich indenter). In agreement between the measured and calculated flow stress
the Nix–Gao model, the von Mises flow rule applies and values when taking into consideration the geometric con-
a Tabor factor of 3 is used to convert the flow stress to stant a, which is known to be a constant with value from
hardness. An increase in the plastic pre-strain produces a 0.3 to 0.6. Equating Eqs. (3) and (4) and converting them
higher SSD density, which is manifested by the strain-hard- to dimensionless format by using Eqs. (2a) and (2b) yields
ening behavior and leads, via Eq. (2a), to an increase in the the following expression for the strain-hardening exponent
macroscopic hardness. Since all the parameters in Eq. (2b)    
1 h 1 K 1
except for the macroscopic hardness are insensitive to the n¼ ln  ln þ
2 ln er b ln er l ln er
plastic pre-strain, the ISE characteristic length is the only  
factor that must decrease in response to an increase in 3
 ln pffiffiffi a cot h ð5Þ
the plastic pre-strain. 2
The strain-hardening exponent and log value of the ISE Here the representative strain er is a function of indenter
characteristic length normalized by the Burgers vector geometry, i.e. it is constant for a self-similar sharp indenter.
(0.25 nm for Ni and 0.248 nm for SCM21) are remarkably Therefore, the relationship between n and ln(h*/b) is linear
linear for each material, as shown in Fig. 3. The deforma- unless the other two terms interrupt the linear relationship.
tion state of the material underneath the sharp indenter is Eq. (5) also suggests that n can be evaluated from h* when
defined by the representative stress and strain, which is the material parameters, b, K, l, and a, and the indenter
identified with the point on the true stress–strain curve geometry h are known. Since the undeformed and pre-de-
obtained via a uniaxial tensile test. The representative formed samples have the same values of b, l, a and h; only
stress rr can also be described by the Hollomon equation K is changed by the plastic pre-strain, as shown in Table 1.
rr ¼ Kenr ð3Þ Consequently, the third term in Eq. (5) does not vary with
the strain-hardening exponent, and the second term can al-
where er is the representative strain induced by nanoinden- ter the linear relationship between n and ln(h*/b).
tation. In particular, assuming that the Taylor dislocation Fig. 4 shows the relationship between the dimensionless
hardening model and the von Mises flow rule apply, as as- terms corresponding to the ISE characteristic length h*
sumed by the Nix–Gao model, we also have (first dimensionless term in Eq. (5)) and the strength coef-
pffiffiffi pffiffiffiffiffi
rr ¼ 3alb qs ð4Þ ficient K (second term in Eq. (5)). The linear relationship
between these two dimensionless terms indicates that the
where rr is the representative flow stress at a representative strain-hardening exponent n can be described by a linear
strain er, a function of indenter geometry. Representative function of ln(h*/b) in Eq. (5). The ISE characteristic
strain is defined by Johnson as 0.2  coth, corresponding

-1.70 Ni
0.45 SCM21
Ni
Dimensionless K term

SCM21 -1.75

0.30 -1.80
n

-1.85
0.15

-1.90

0.00 1.41 1.44 1.47 1.50 1.53

7.4 7.6 7.8 8.0 8.2 Dimensionless h* term
ln (h*/b)
Fig. 4. Dimensionless term corresponding to the strength coefficient ln(K/
Fig. 3. Strain-hardening exponent n vs. log of ISE characteristic length l) vs. the term corresponding to ISE characteristic length ln(h*/b) in Eq.
normalized by Burgers vector (log (h*/b)) for Ni and SCM21. (5).
3342 J.-Y. Kim et al. / Acta Materialia 56 (2008) 3338–3343

length h* is proportional to the inverse square of the mac- Cu

roscopic hardness H0 in Eq. (2b), since the other parame- 0.4
ters in Eq. (2b) are insensitive to the plastic pre-strain. STS304
The macroscopic hardness can be converted to a represen- STS303
tative stress rr by using Eq. (4). Equating the macroscopic 0.3
CI 300 CI 600
hardness and representative stress H 0 =3 ¼ Kenr and rear- S45C STS316L
SKD61 SKD11
ranging terms by using the log and linear relationship

n
0.2 SKS3 SKH9
between n and ln(h*/b) leads easily to the linear relation- Al2011 API X65
ship between ln(h*/b) and ln(K/l). In addition, the revision SCM 4
API X70 Ti
of the Nix–Gao model in Ref. [44], in which storage vol- 0.1 API X80
Al7076 Al6061
ume of GNDs was corrected by multiplying proportional- Ti-6Al-6V
Ti-5Al-2.5Sn
ity factor, leading to smaller h*, does not change the basic Ti-10V-2Fe-3Al
linear relationship in Eq. (5). 0.0
6.0 6.5 7.0 7.5 8.0 8.5
If the strain-hardening exponent n can be described by
ln (h*/b)
ln(h*/b), as shown above, it would be very useful in analyz-
ing the plastic deformation induced by nanoindentation, as Fig. 5. Strain-hardening exponent n vs. log values of ISE characteristic
well as in evaluating the strain-hardening exponent by a length normalized by magnitude of Burgers vector ln(h*/b) for various
general nanoindentation test. The true stress–strain behav- metals.
ior of a material is assumed to be
 less steels, API X-grade steels, titanium alloys, aluminum
Ee; e 6 ey
r¼ ð6Þ alloys and pure polycrystalline copper. While some scatter
Ken ; e P ey is present in the data, it is clear that most metals show a lin-
where ey is the yield strain. In trying to obtain a quantita- ear dependence of the log of ISE characteristic length nor-
tive mechanical response, the deformation induced by malized by the Burgers vector and the strain-hardening
introducing a sharp indenter into a material can be de- exponent. Note that a linear relation between n and
scribed by a dimensionless function P ln(h*/b) is obtained only when ln(h*/b) and ln(K/l) have
P ¼ PðEr ; ey ; n; hÞ ð7Þ a linear relationship, as in the controlled plastic pre-strain
situation. A roughly linear dependence between n and
where Er is the reduced elastic modulus given by ln(h*/b) for the randomly selected 22 metals might arise
 1 from the generally lower strain hardening exponent and
1  m2 1  m2i
Er ¼ þ ð8Þ lower ISE characteristic length of higher-strength materi-
E Ei
als. This result shows the feasibility of developing a unique
where Ei and mi are the elastic modulus and Poisson’s ratio dimensionless function describing the relationship between
of the indenter. Ei, mi and m are generally known, and the n and ln(h*/b) for specific materials.
elastic modulus of a material can be evaluated from the
measured reduced elastic modulus. In nanoindentation, 4. Conclusions
the yield strain (second parameter in Eq. (7)) is related to
the portion of elastic strain at a given total (representative) Strain-hardening exponents and ISE characteristic
strain that is independent of indentation depth for a sharp lengths of Ni and SCM21 with different plastic pre-strain
indenter because of its self-similarity. This parameter can were measured and analyzed. A linear relationship between
be inferred from a variety of sources, for instance the ratio the strain-hardening exponent and the log of the ISE char-
of yield strength to elastic modulus ry/E, the ratio of hard- acteristic length normalized by the Burgers vector was
ness to elastics modulus H/E, the ratio of final indentation found. This is analyzed by using the Taylor dislocation
depth to maximum indentation depth hf/hmax and the ratio hardening model and the representative stress–strain
of work recovered during elastic unloading to total work approach. We showed that a linear relationship between
input during loading Welastic/Wtotal [13,14,16–22,46–52]. If n and ln(h*/b) is achieved when the relationship between
the strain-hardening exponent is evaluated from the ISE ln(h*/b) and ln(K/l) is linear, as is the case in the samples
characteristic length, as described above, and the angle of with different plastic pre-strain. On the basis of results from
the sharp indenter is given, the plastic deformation induced 22 different metals and alloys, a dimensionless function
by nanoindentation can be determined by relying only on using the relationship between the strain-hardening expo-
those parameters measurable in nanoindentation tests as nent and ISE characteristic length was proposed to
    describe plastic deformation by nanoindentation with a
W elastic h
P ¼ PðEr ; ey ; n; hÞ ¼ Ph Er ; ; ln ð9Þ sharp indenter. A clear correlation between the strain-
W total b
hardening exponent and the ISE characteristic length
Fig. 5 shows the relationship between the ISE characteristic means that the material deformation induced by nanoin-
length and strain-hardening exponent for the 22 different dentation with a sharp indenter can be fully described by
metals, including general structural steels, cast irons, stain- using only parameters measurable through nanoindentation
 J.-Y. Kim et al. / Acta Materialia 56 (2008) 3338–3343 3343

tests. This result is significant in the determination of nan- [20] Venkatesh TA, Van Vliet KJ, Giannakopoulos AE, Suresh S. Scripta
oindentation related plastic flow properties, such as pile- Mater 2000;42:833.
[21] Bucaille JL, Stauss S, Felder E, Michler J. Acta Mater 2003;51:1663.
up/sink-in and plastic zone size. As an application of this [22] Bouzakis KD, Michailidis N. Thin Solid Films 2004;469–470:227.
result, we are currently working on correlating the nanoin- [23] Swadener JG, George EP, Pharr GM. J Mech Phys Sol 2002;50:681.
dentation derived and conventional hardness values by [24] Qu S, Huang Y, Pharr GM, Hwangg KC. Int J Plasticity
measuring the contact depth in the loaded state from the 2006;22:1265.
ISE characteristic length. [25] Hollomon JH. Trans AIME 1945;162:268.
[26] Ma Q, Clarke DR. J Mater Res 1995;10:853.
[27] Nix WD, Gao H. J Mech Phys Sol 1998;46:411.
Acknowledgement [28] McElhaney KW, Vlassak JJ, Nix WD. J Mater Res 1998;13:1300.
[29] Gao H, Huang Y, Nix WD, Hutchinson JW. J Mech Phys Sol
This work was supported by the Korea Research Foun- 1999;47:1239.
dation Grant (KRF-2007-357-D00138) funded by the Kor- [30] Huang Y, Gao H, Nix WD, Hutchinson JW. J Mech Phys Sol
2000;48:99.
ean Government (MOEHRD). [31] Liu Y, Ngan AHW. Scripta Mater 2001;44:237.
[32] Qiu X, Huang Y, Nix WD, Hwang KC, Gao H. Acta Mater
References 2001;49:3949.
[33] Tymiak NI, Kramer DE, Bhar DF, Wyrobek TJ, Gerberich WW.
[1] Doerner MF, Nix WD. J Mater Res 1986;1:601. Acta Mater 2001;49:1021.
[2] Oliver WC, Pharr GM. J Mater Res 1992;7:1564. [34] Swadener JG, George EP, Pharr GM. J Mech Phys Sol 2002;50:681.
[3] Oliver WC, Pharr GM. J Mater Res 2004;19:3. [35] Gao H, Huang Y. Scripta Mater 2003;48:113.
[4] Cheng YT, Cheng CM. Mater Sci Eng R 2004;44:91. [36] Elmustafa AA, Stone DS. J Mech Phys Sol 2003;51:357.
[5] Gouldstone A, Chollacoop N, Dao M, Li J, Minor AM, Shen YL. [37] Qu S, Huang Y, Nix WD, Jiang H, Zhang F, Hwang KC. J Mater
Acta Mater 2007;55:4015. Res 2004;19:11.
[6] Fischer-Cripps AC. Vacuum 2000;58:569. [38] Feng G, Nix WD. Scripta Mater 2004;51:599.
[7] Mukhopadhyay NK, Paufler P. Int Mater Rev 2006;51:209. [39] Zhang TY, Xu WH, Zhao MH. Acta Mater 2004;52:57.
[8] Schuh CA. Mater Today 2006;9(5):32. [40] Wei Y, Wang X, Zhao M. J Mater Res 2004;19:208.
[9] Tabor D. Hardness of metals. OX: Clarendon Press; 1951. [41] Kim JY, Lee BW, Read DT, Kwon D. Scripta Mater 2005;52:353.
[10] Johnson KL. Contact mechanics. CA: Cambridge University Press; [42] Kim JY, Lee JJ, Lee YH, Jang JI, Kwon D. J Mater Res
1985. 2006;21:2975.
[11] Ahn JH, Kwon D. J Mater Res 2001;16:3170. [43] Kim JY, Kang SK, Lee JJ, Jang JI, Lee YH, Kwon D. Acta Mater
[12] Kim JY, Lee KW, Lee JS, Kwon D. Surf Coat Tech 2006;20:4278. 2007;55:3555.
[13] Taljat B, Zacharia T, Kosel F. Int J Solids Struct 1998;35:4411. [44] Durst K, Backes B, Franke O, Göken M. Acta Mater 2006;54:2547.
[14] Dao M, Chollacoop N, Van Vliet KJ, Venkatech TA, Suresh S. Acta [45] Huang Y, Zhang F, Hwang KC, Nix WD, Pharr GM, Feng G. J
Mater 2001;49:3899. Mech Phys Sol 2006;54:1668.
[15] Herbert EG, Pharr GM, Oliver WC, Lucas BN, Hay JL. Thin Solid [46] Cheng YT, Li Z. Philos Mag A 2002;82:1821.
Films 2001;398–399:331. [47] Chnge YT, Cheng CM. Appl Phys Lett 1998;73:614.
[16] Jayaraman S, Hahn GT, Oliver WC, Rubin CA, Bastias PC. Int J [48] Malzbender J, De With G. J Mater Res 2002;17:502.
Solids Struct 1998;35:365. [49] Pharr GM, Bolshakov. J Mater Res 2002;17:2660.
[17] Cheng YT, Cheng CM. Int J Solids Struct 1999;36:1231. [50] Bolshakov A, Pharr GM. J Mater Res 1998;13:1049.
[18] Chollacoop N, Dao M, Suresh S. Acta Mater 2003;51:3713. [51] Hay JC, Bolshakov A, Pharr GM. J Mater Res 1999;14:2296.
[19] Giannakipoulos AE, Suresh S. Scripta Mater 1999;40:1191. [52] Joslin DL, Oliver WC. J Mater Res 1990;5:123.