Вы находитесь на странице: 1из 26

International Journal of Mechanical Sciences 43 (2001) 895}920

Investigation of sharp contact at rigid}plastic conditions


Per-Lennart Larsson*
Department of Solid Mechanics, Royal Institute of Technology, S-100 44, Stockholm, Sweden
Received 1 November 1999; received in revised form 17 February 2000

Abstract

Sharp contact problems are examined theoretically and numerically. The analysis is focused on elas-
tic}plastic material behaviour and in particular the case when the local plastic zone arising at contact is so
large that elastic e!ects on the mean contact pressure will be small or negligible. It is shown that, save for the
particular case of a rigid}plastic power-law material, at such conditions, there is no single representative
value on the uniaxial stress-strain curve that can be used in order to evaluate the global parameters at
contact. However, the present numerical results indicate that good accuracy predictions for the mean contact
pressure can be achieved when this variable is described by two parameters corresponding to the stress levels
at, approximately, 2 and 35% plastic strain. Regarding the size of the contact area, it is shown that this
quantity is very sensitive to elastic e!ects and any general correlation with material properties is complicated
at best. The numerical analysis is performed by using the "nite element method and the theoretical as well as
the numerical results are compared with relevant experimental ones taken from the literature. From
a practical point of view, the presented results are directly applicable to material characterization or
measurements of residual mechanical "elds by sharp indentation tests, but also for situations such as contact
in gears or in electronic devices.  2001 Elsevier Science Ltd. All rights reserved.

1. Introduction

Sharp indentation, or hardness, tests, associated with names like Knoop, Vickers and Berkovich,
have for a long time been used for characterization of conventional engineering materials such as
metals and alloys. In recent years, such tests have received increasing attention and appreciation
due to the development of new experimental devices like the nanoindenter, Pethica et al. [1],
enabling an experimentalist to determine the material properties from extremely small samples of

* Tel.: 0046-8-790-7540; fax: 0046-8-411-2418.


E-mail address: pelle@hallf.kth.se (P.-L. Larsson).

0020-7403/01/$ - see front matter  2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 2 0 - 7 4 0 3 ( 0 0 ) 0 0 0 5 6 - 4
896 P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920

Fig. 1. Normalized hardness, H/p , as function of !"E tan b/((1!l)p ). Schematics of the correlation of sharp
P W
indentation testing of elastic}plastic materials as originally suggested by Johnson [3]. The three levels I}III, pertinent to
di!erent indentation response, are also indicated in the "gure.

the material. Additionally, for many new engineering materials like ceramics, a standard uniaxial
test often fails to deliver reliable results and, accordingly, indentation is the only alternative for
material characterization. Indentation is also a convenient tool for determining the material
properties of thin "lms or strings in ready-to-use engineering devices.
The most important quantities given by an indentation test are the hardness H, here interpreted
as the mean contact pressure, the contact area A or alternatively, the relation between indentation
load, P, and indentation depth, h, during loading and unloading. These quantities can then be used
to determine the constitutive properties of the material by taking advantage of results from earlier
theoretical and experimental analyses by Tabor [2], Johnson [3,4], and Doerner and Nix [5].
Johnson [3,4] also suggested that the outcome of a sharp indentation test on elastic-plastic
materials will fall into one out of three levels, depending on the material properties and the type of
indenter used, as shown schematically in Fig. 1 where the normalized hardness, HM "H/p , is
P
depicted as function of the parameter
E tan b
!" . (1)
(1!l)p
W
P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920 897

In Fig. 1, E is Young's modulus, l is Poisson's ratio, p is the #ow stress at "rst yield, p is the #ow
W P
stress at a representative value of the plastic strain, a concept "rst suggested by Tabor [2] to be
discussed in detail below, and b is the angle between the sharp indenter and the undeformed surface
of the material. It should be mentioned that for materials with appreciable strain-hardening
a better correlation with the schematic curve in Fig. 1 is given if p in Eq. (1) is replaced by some
W
stress level, most frequently p , representing the stress}strain characteristics at higher values of the
P
plastic strain. This parameter will here and in the sequel be denoted by ! and explicitly de"ned as
F
E tan b
! " . (2)
F (1!l)p
P
In short, the three levels shown in Fig. 1 can be characterized as follows; In level I, !, ! (3, very
F
little plastic deformation occurs during the indentation test and all global properties can be derived
from an elastic analysis. In level II, 3(!, ! (40, an increasing amount of plastic deformation is
F
present and both the elastic and the plastic properties of the material will in#uence the outcome of
a hardness test. Johnson [3,4] suggested, based on the fact that in such a situation the stress "eld
just beneath the indenter is almost hydrostatic, that the process was very similar to the case of
expansion of a spherical cavity in a large solid due to an internal pressure and derived the formula

 
2 E tan b
H" p 1#ln (3)
3 P 3(1!l)p
W
for the hardness. Finally then, in level III, !, ! '40, plastic deformation is present all over the
F
contact area and elasticity no longer in#uences the hardness value for the material. This is also the
region pertinent to most standard engineering materials, such as steel, many aluminiums and
copper just to mention a few. By analysing the results from a number of experiments Tabor [2]
concluded that the hardness in level III could be derived from the simple formula
H"Cp , (4)
P
where C is a constant that only depends on the geometry of the sharp indenter while, as mentioned
above, p is the #ow stress at a representative value of the plastic strain, e . Based on experimental
P P
results Tabor [2] also suggested the values C+3 and e +0.08 for a Vickers indenter while Atkins
P
and Tabor [6] found C+2.54 and e +0.11 for a cone indenter with an angle of 223 between the
P
indenter and the undeformed surface. In this context it should also be mentioned that the
representative stress used by Johnson [3,4] in Eq. (3) is essentially de"ned in the same way as
proposed by Tabor [2]. In short, Eqs. (3) and (4) have been used extensively in order to determine
the elastic}plastic material properties of a material from a simple hardness test yielding results of,
for many applications, su$cient accuracy. However, with the development of modern computers
and new numerical methods it has been possible to investigate the region of validity for the
above-mentioned formulae in some more detail by computational simulations of indentation of
materials with tailored constitutive properties. In doing so, Giannakopoulos et al. [7] and Larsson
et al. [8] analysed Vickers and Berkovich indentation of elastic}plastic materials in level II, by
using "nite element methods, and found that high accuracy could only be achieved, when using an
equation with the same form as Eq. (3), by describing the hardness with two parameters. In short,
leaving out details about the deformation at the contact boundary, very good agreement with
898 P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920

numerical and experimental results were found when the #ow stress p (e "0.08) in Eq. (3) were
P P
replaced by a two-parameter characterization (p #p (e "0.3)), p (e"0.3) being the #ow stress
W P P P
at 30% plastic strain. In addition, according to Giannakopolos et al. [7] and Larson et al. [8], the
constant 2/3 should be replaced by the values 0.29, for Vickers, and 0.27, for Berkovich, as
determined from the numerical calculations. Numerical investigations of cone indentation have
also should be performed (cf. Refs. [9,10] and [11], but these studies have not addressed the issue of
the accuracy of Eq. (3) and (4) save for the case of level II indentation of elastic}perfectly plastic
materials [9].
A corresponding analysis for sharp indentation tests in the rigid}plastic regime, level III, has not
yet been performed. This is somewhat surprising remembering that substantial experimental e!orts
are devoted nowadays in order to determine the constitutive properties of a material by using
indentation testing and the accuracy of this procedure is very much determined by the reliability of
the formulae involved when transforming indentation parameters to material properties. In this
context, it should also be remembered that Eq. (4) is derived from indentation experiments with
Vickers, Tabor [2], and cone, Atkins and Tabor [6], indenters on basically two materials, namely
mild steel and annealed copper and other experimental studies (cf. e.g. Ref. [12]), have reported that
hardness values for many metals are in#uenced by plastic strain values much higher than 0.08. Such
a result is also con"rmed by a recent investigation by Chaudri [13] who performed a very careful
experimental study of annealed copper and found substantial deviations between experimentally
determined Vickers hardness values and hardness values given by Eq. (4). As a complement to this
discussion, it must be clearly stated that the hardness value alone is in most cases of practical
interest not su$cient in order to completely determine the plastic properties of the material.
Additional information is, however, given from the size of the contact area at indentation, a matter
discussed in detail in connection with material characterization by Brinell indentation (cf. e.g.
Ref. [14]), but in the context of sharp indentation this feature has not been taken advantage of by
experimentalists.
With the above-mentioned in mind it seems clear that a careful study of sharp contact problems
in the rigid plastic regime is warranted. In doing so there is of course a substantial amount of
experimental results to be utilized and what is needed in order to be able to draw some more
detailed conclusions is a systematic and comprehensive numerical study of sharp indentation tests.
The reason for such a numerical analysis is the possibility to tailor the mechanical behaviour of the
indented materials in order to study some delicate questions arising when interpreting present and
previous experimental and numerical results from sharp indentation testing. Accordingly, it is then
the aim presently to perform such an analysis of sharp indentation tests with cone, Vickers as well
as Berkovich indenters, for materials pertinent to level III. The analysis is of course motivated by
the fact that indentation is a simple experimental tool for determining the constitutive properties of
a material and that the outcome of an analysis of such test can be used in order to make predictions
about the mechanical behaviour relevant for other problems involving sharp contacts, for example
in gears and electronic devices. Another application where the accuracy of Eq. (4), and related
results for other quantities given by an indentation test, is of importance concerns the measurement
of residual "elds by indentation. It has recently been shown by Suresh and Giannakopoulos [15]
and by Carlsson and Larsson [16] that indentation can serve as a very useful alternative to, for
example, X-ray or hole-drilling when the measurement of residual stresses or strains is at issue. In
such a situation, the number of experiments required in order to determine the residual "eld values
P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920 899

involved can be substantially reduced if, for example, the plastic #ow stress can be related to the
hardness value by a general and accurate formula in the spirit of Eq. (4).
The boundary value problems arising when analysing sharp contacts are unavoidably very
complicated as, for one thing, nonlinear kinematics have to be accounted for and "nite element
techniques then appear to be the natural choice of numerical tool. Presently, the general purpose
program ABAQUS [17] will be relied upon.

2. Governing equations and analysis

The mechanical problem discussed above and under consideration presently concerns a homo-
geneous, isotropic, elastic}plastic half-space indented in the normal direction to the surface by
a rigid sharp indenter. It is assumed that the resulting deformation takes place under quasi-static
and isothermal conditions. Furthermore, only frictionless contact is considered. The latter assump-
tion is acceptable in the present situation as mainly global properties are at issue and such variables
are relatively insensitive to frictional e!ects (cf. e.g. Refs. [18,19]). Three di!erent types of indenters
will be analysed as depicted in Fig. 2a (a cone indenter with an angle b"223), in Fig. 2b (a Vickers
indenter), and in Fig. 2c (a Berkovich indenter). In Fig. 2, P is the indentation load and h is the
indentation depth. It should be mentioned that a cone angle of 223 was chosen as this corresponds
to the angle between the indenter and the material at Vickers indentation. Finally then, it should be
clearly stated that due to the large rotations imposed by the rigid indenter nonlinear kinematics has
to be accounted for in the analysis (cf. e.g. Ref. [8]).
The formulation of the resulting boundary value problem has been made earlier in some detail,
Giannakopoulos et al. [7] and Larsson et al. [8], and for this reason, only the basic aspects of the
problem will be stated presently. Starting then with the constitutive speci"cation the incremental,
rate-independent Prandtl}Reuss equations for classical Mises plasticity with isotropic hardening,
remembering that the loading part of an indentation test is of most immediate interest presently,
read, by aid of the Kronecker identity tensor d ,
GH

 
E l 3q q (E/(1#l))
q( " d d # d d ! GH IJ e (5)
GH (1#l) GI HJ (1!2l) GH IJ 2q((2/3)K#E/(1#l)) IJ
C
in a large strain formulation. In Eq. (5), e is the rate of deformation connected to the material
GH
velocity u as
G

 
*u *u
e " G # H 2, (6)
GH *x *x
H G
where x is the current position of a material point initially at X , and
G G
q( "q !(u q #q u ) (7)
GH GH GI IH GJ HJ
is the Jaumann rate of the Kirchho! stress, q , u being the material spin. The Kirchho! stress is
GH GH
related to the Cauchy stress, p , as q "Jp , J being the ratio of volume in current state to volume
GH GH GH
in previous state, and q and q are Mises e!ective stress and deviatoric stress respectively. Finally,
C GH
900 P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920

Fig. 2. Schematics of the geometry of the indentation tests: (a) Cone indentation, top and side view; (b) Vickers
indentation, top and side view; (c) Berkovich indentation, top and side view.
P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920 901

in Eq. (5), K is the instantaneous slope of the uniaxial Kirchho! stress, q, versus the logarithmic
accumulated plastic strain, e , de"ned in a multiaxial situation as
N

 
R 2 
e " (e ) (e ) dt (8)
N 3 N GH N GH

(e ) being the plastic part of the rate of deformation. It should be mentioned here that the
N GH
elastic}plastic constitutive speci"cation in Eq. (5) is only valid at plastic loading when q "q(e ),
C N
the initial yield stress being given by q "q(0), and at elastic loading, or unloading, a hypoelastic
W
formulation of Hooke's law, the "rst part of Eq. (5), is relied upon.
Presently, with plastic strains being substantially larger than the elastic ones, the response of the
half-space at indentation is to a large extent governed by the function q(e ), or equivalently
N
represented by the Cauchy stress p(e ) which is more suitable for a situation where the material
N
response is determined from uniaxial tests. For simplicity, this relation is often approximated by
power-law functions according to
p"p #p eK, (9)
W  N
p and m being material constants. Such an approximation will also be utilized presently when

appropriate even though a substantial part of the numerical calculations will be performed with
a more irregular relation between p and e .
N
To complete the boundary value problem it remains to state that throughout the analysis,
equilibrium equations in absence of body forces have to be satis"ed and that outside the contact
area the surface of the half-space is assumed traction free resulting in the boundary conditions
p n "p n "0, (10a)
H H H H
p n "0, (10b)
H H
n being the outward unit normal vector of the half-space de"ned in the deformed con"guration.
H
Furthermore, as presently no friction is assumed between the indenter and the material, Eq. (10a)
still holds within the area of contact while Eq. (10b) is replaced by unilateral kinematic constraints
given by the shape of the indenters as depicted in Fig. 2.
Formally then, the indentation load is given by

P"!
A
p N dA
H H A
(11)

where A is the actual area of contact and N is the inward unit normal vector to the rigid surface of
A H
the indenter, as de"ned in Fig. 2. The material hardness at indentation, which in this study is
de"ned in the same way as the average contact pressure, is then calculated directly, using the
projected contact area A as
P P
H" " . (12)
A A cos b
A
It should be pointed out that within this setting the problem is self-similar with no characteristic
length present neither in the governing equations nor in the boundary conditions. Consequently,
the hardness de"ned in Eq. (12), as well as the ratio h/(A, will be constant during the loading
902 P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920

sequence of an indentation test and stresses and strains will be functions of the dimensionless
variable x /(A, and material properties, alone. Here, (A should be interpreted as a characteristic
G
contact length. As a result, the indentation load and the contact area are directly proportional to
the square of the indentation depth according to


P"g h,
 (13)
A"g h

with the functions g and g depending only on the geometry of the indenter and the constitutive
 
properties of the indented material. With Eq. (13) in mind it then proves convenient, to be utilized
below, to introduce the indentation parameter
A
c" , (14)
A
LMK
A being the nominal contact area resulting if the material neither sinks-in nor piles-up along the
LMK
contact boundary. Accordingly, if c(1 the resulting contact area is smaller than what could
be expected from purely geometrical considerations and the other way around if c'1. From
Eqs. (12)}(14) it can now be concluded that for a particular indenter geometry, the global
indentation variables H and c are functions of the material properties alone and partly for this
reason, these parameters will be used extensively in the presentation of the numerical results below.
Within the formulation of the boundary value problem above, the so-called indentation size e!ects
are not accounted for. However, such e!ects, frequently modelled by strain gradient plasticity
(cf. e.g. Refs. [20}22]), will only be of interest at very small values of the indentation depth, 1 lm
or less say, and as the main objective here is to study the relationship between macroscopic
material properties and indentation parameters this feature will not be accounted for in the present
analysis.
From a mathematical point of view, the boundary value problem just outlined is a very
complicated one with both geometric and material nonlinearities present and as a result, numerical
methods have to be used. It then proved advantageous to use the commercial "nite element
program ABAQUS [17] in order to draw upon the experience gained from earlier and similar
analyses of sharp indentation tests [7,8,18,23]. The only new feature presently concerns the fact
that elastic strains will be essentially negligible in the highly deformed region around the contact
area. However, this did not introduce any numerical problems of fundamental importance.
Accordingly, no extensive testing of the numerical scheme was performed as features such as
convergence, comparison with experimental hardness values and P}h relations and far-"eld
properties has been undertaken in the earlier investigations mentioned above. Instead, it seems
appropriate to directly discuss the "nite element meshes used in the numerical analysis. In
a two-dimensional situation then, remembering that axisymmetry prevails at cone indentation, the
mesh depicted in Fig. 3a and b was used. In order to avoid any far-"eld boundary e!ects the mesh
had to be extended to, at least, 30 times the radius of contact, both horizontally and vertically.
Normally, convergence of, de"ned here as the numerical scatter being less than $2.5%, numer-
ically determined indentation parameters required at least 15 elements within the contact area. In
general, the calculations were continued until the contact area included 20}30 elements. The
two-dimensional "nite element mesh consisted of 4798 four-noded axisymmetric elements and 5094
P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920 903

Fig. 3. Finite element meshes used for the numerical calculations. In all cases, loading was applied in the negative
X -direction: (a) Complete mesh used in the analysis of the cone indentation test; (b) Details close to the region of contact

of the mesh used in the analysis of the cone indentation test; (c) Details close to the region of contact of the mesh used in
the analysis of the Vickers test.

nodes. The elements were of hybrid type, i.e. the displacements were approximated with bilinear
shape functions while the hydrostatic pressure attained a constant value in each element, in order
to improve convergence at almost incompressible deformation. The mesh used for Vickers
904 P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920

indentation is depicted in Fig. 3c and had essentially the same features as the one discussed above
for cone indentation but, of course, extended to a three-dimensional situation. Accordingly, only
details close to the region of contact is depicted in Fig. 3c. Due to the symmetries involved, only 1/8
of the half-space had to be modelled resulting in a mesh with 14055 eight-noded block elements and
16527 nodes and with the displacements approximated by trilinear shape functions while the
hydrostatic pressure was assumed constant in each element. The mesh used for Berkovich
indentation was essentially the same as the one depicted in Fig. 3b and will not be shown here.
However, due to the triangular base of the Berkovich indenter, 1/6 of the half-space had to be
modelled resulting in a mesh with 17,199 eight-noded block elements of hybrid type and 19,901
nodes. At three-dimensional indentation, approximately 50 elements had to be in contact in order
to achieve convergence of important parameters. However, all three-dimensional calculations were
continued until the contact region included at least 60}70 elements.

3. Results and discussion

Before starting to discuss the outcome of the numerical analysis presently undertaken it certainly
deserves to state once again that of interest here is only the behaviour of sharp indentation tests in
the rigid}plastic regime, i.e. level III in Fig. 1. Accordingly, all the materials investigated will have
a combination of elastic and plastic properties that gives a response at indentation pertinent to this
regime. In this context though, it seems sensible, remembering that the value !"40 suggested by
Johnson [4] as the onset of rigid}plastic behaviour is not a very sharp limit, to perform some
numerical calculations in order to determine the range of !-values being pertinent to rigid}plastic
contact behaviour. In doing so, cone indentation of a large number of elastic}perfectly plastic
materials, with many di!erent combinations of yield stress, Young's modulus and Poisson's ratio,
were investigated and the result, represented by the normalized hardness as function of !, is
depicted in Fig. 4. Perfect plasticity was chosen as this will introduce only one possible representa-
tive stress, the yield stress, at normalization and it seemed appropriate at this stage not to introduce
any premature discussion about representative stresses at strain-hardening plasticity, which identi-
"cation, indeed, is the main issue of this investigation. With this in mind, it is obvious from the
results in Fig. 4 that at least levels II and III indentation can be clearly identi"ed, no attempt was
made presently to accurately determine the range of level I, i.e. elastic, indentation, and that at,
approximately, !"20 rigid}plastic behaviour starts to dominate the indentation process. The
e!ect of strain-hardening is, however, not obvious and for this reason, it seemed appropriate to use
!, ! "50, with the stress at a plastic strain of 8% somewhat tentatively used as the representative
F
stress p in Eq. (3), as a lower limit for all the materials investigated in this study. Indeed, in most
P
cases !, ! was chosen to be larger than 100. Contemplating a bit more over the results presented
F
in Fig. 4, it is also interesting to note that the results in Fig. 4 suggest a relation (as also indicated in
Fig. 4),
H"2.55p , (15)
W
at cone indentation in the rigid plastic regime. This is in contrast to the situation at, for example,
Vickers indentation where a factor of approximately 3 is found (cf. e.g. Ref. [2]), but the result in
Eq. (15) is nevertheless con"rmed by experiments by Atkins and Tabor [6] who suggested the value
P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920 905

Fig. 4. Normalized hardness, H/p , as function of !"E tan b/((1!l)p ) at cone indentation of elastic}perfectly plastic
W W
materials. (*), (H/p )"2.55. (O) present numerical results.
W

2.54 in case of cone indentation, with an angle of 223 between the cone and the surface of the
material, of highly work-hardened steel and copper. The result in Eq. (15) is also in good agreement
with theoretical ones by Lockett [24] who, based on an axisymmetric slip-line analysis, suggested
a factor of approximately 2.69 at cone indentation with an angle of 203 between the cone and the
surface of the material.
With these preliminaries now taken care of it seems appropriate to focus on the main issue of this
study, i.e. the possibility of the presence of a representative strain, or stress, at sharp indentation of
strain-hardening plastic materials. It should then be mentioned once again that for a long time now
the results by Tabor [2] with C+3 and e +0.08 in Eq. (4) have been accepted and used
P
for material characterization by Vickers, and lately also, in particular at nanoindentation, by
Berkovich indentation. However, Tabor's [2] results were derived from experiments on essentially
two materials, mild steel and copper, and it was found by Dugdale [12] soon afterwards that the
material hardness was dependent on the shape of the stress}strain curve at much higher strain
values than 0.08. This result has been con"rmed in a comprehensive and very interesting experi-
mental study by Chaudri [13] concerning Vickers indentation on annealed polycrystalline copper.
Chaudri [13] concluded that in order to "nd a universal relation according to Eq. (4), it was
necessary to use a representative value of the #ow stress corresponding to a strain of, approxim-
ately, 0.2, or even higher. This result is in accordance with numerical ones by Giannakopoulos et al.
[7] and Larsson et al. [8] concerning Vickers and Berkovich indentation respectively, even though
the main interest then was directed towards material behaviour pertinent to level II in Fig. 1. These
906 P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920

authors suggested a representative stress value evaluated at a plastic strain of 0.3, as mentioned
earlier though this stress value had to be supplemented by the yield stress in order to get results of
high accuracy, within a small strain formulation of the problem which at compressive loading
corresponds to a logarithmic strain value of 0.35 in a more appropriate large strain formulation of
the problem.
In short, the numerical investigation can be divided into three di!erent steps. First of all,
indentation of a rigid}plastic power-law material de"ned as
p"p eK, (16)
 N
i.e. a material according to Eq. (9) with zero yield stress, was analysed in order to determine
whether or not the concept of a representative strain holds in such a situation with a limited
number of constitutive parameters to account for. Secondly, indentation of materials with more
irregular stress}strain relations is investigated aiming at a general description of hardness at sharp
indentation. Finally, certain aspects pertinent to the contact area at indentation are discussed with,
in particular, characterization of constitutive properties and material states in mind. In this
context, it should also be mentioned that no results are presented related to the material response
at unloading even though this is of substantial importance when, for example, determination of the
elastic material properties by indentation, cracking or fatigue are at issue. However, as the
conclusions drawn from the present results pertinent to unloading are in, essentially, total
agreement with earlier "ndings for level II indentation (cf. Refs. [7,8]), this matter was not dwelled
upon further here.

3.1. The hardness at indentation of a rigid}plastic power-law material

Accordingly, the analysis will initially focus on indentation of a rigid}plastic power-law material
(Eq. (16)). Such a material is of particular interest due to simplicity as, for one thing, the in#uence
from strain-hardening on the material hardness can be represented by only two parameters, p

and m. In this case, as self-similarity prevails and no characteristic length is present in the boundary
value problem, it is easily shown that

   
h K x
p "p p G (17)
GH  (A GH (A

(cf. e.g. Ref. [14]). Furthermore, as a result of Eqs. (11)}(13) and (17), the hardness is related to the
constitutive parameters as
H"B (m)p [B (m)]K, (18)
  
regardless of the shape of the sharp indenter, where the parameters B and B are, as indicated,
 
functions of the power-law exponent m. Indeed, the parameter B can be interpreted as the

representative strain of the material at a particular value on m. In this context, it should be
mentioned that in a rigorous numerical analysis of Brinell indentation of rigid}plastic power-law
materials by Biwa and Stora kers [25] it has been shown that the corresponding parameters in such
a case are relatively constant for a wide range of m-values. Accordingly, it is possible to accurately
correlate the Brinell hardness value to the uniaxial stress}strain curve of the indented material
P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920 907

Fig. 5. Normalized hardness, HI "H/p , as function of 1/m at indentation of a rigid}plastic power-law material

according to Eq. (19). (a) Cone indentation. (O) present numerical results. (*), Eq. (19). (- - -) experimental results by
Atkins and Tabor [6] with the constant values B "2.54 and B "0.11 in Eq. (18). (b) Vickers indentation. (O) present
 
numerical results. (*), Eq. (20). (- - -) experimental results by Tabor [2] with the constant values B "3 and B "0.08 in
 
Eq. (18). (- ) -) experimental results by Chaudri [13] with the constant values B "3.1 and B "0.20 in Eq. (18).
 

yielding results of substantial practical importance. Presently though, with sharp indentations at
issue, the situation is more involved as an analysis based on linear kinematics is not su$cient and
the numerical methods utilized by Biwa and Stora kers [25] can not be used to advantage in a large
strain formulation of the problem. Instead, brute force "nite element calculations must be adhered
to. The outcome of these calculations is shown in Fig. 5a, cone indentation, and in Fig. 5b, Vickers
indentation, where the normalized hardness, HI "H/p , is depicted as function of 1/m. Regarding

cone indentation, judging from the results in Fig. 5a, it is evidently possible to, approximately, "t
these results to a straight line resulting in a representative strain value of 18%. Accordingly, by Eqs.
(15) and (18),
HI "2.55(0.18)K. (19)
This result being in fairly good agreement with the experimental "ndings by Atkins and Tabor [6],
yielding B "2.54 and B "0.11 in Eq. (18) and also shown in Fig. 5a. In this context, it must,
 
however, be immediately emphasized that no numerical results were presented presently for
(1/m)'0.7. The reason for this being that close adherence to Eq. (16) was then, in practice,
impossible when using straightforward "nite element calculations as convergence could not be
achieved for su$ciently high values on ! when linear, or close to linear, strain hardening was at
F
908 P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920

issue. Accordingly, a reliable rigid plastic solution could not be derived in this case. Such
a drawback is, however, of less practical importance as the rigid plastic power law relation in
Eq. (16) will have a decreasing physical relevance as mP1. Essentially then, the important
ingredient of sharp indentation of rigid plastic power law materials are well captured by the
numerical results presented in Figs. 5a and b as discussed below.
Results for Vickers indentation of rigid}plastic power-law materials are shown in Fig. 5b
together with the experimental "ndings by Tabor [2], B "3 and B "0.08 in Eq. (18), and by
 
Chaudri [13], B "3.1 and B "0.20 in Eq. (18). It should then be immediately stated that the B -
  
and B -values presented above as pertinent to the investigation by Chaudri [13] was not adopted

by the author as a suggestion for a general description of Vickers indentation of power-law
materials but was only used in order to clearly identify the in#uence from high plastic strain levels
in such a situation. Indeed, the B -value was somewhat tentatively chosen here based on the results

by Chaudri [13] for heavily work-hardened copper. Despite of this though, it is obvious from the
results shown in Fig. 5b that there are signi"cant di!erences between the three sets of results with
the numerical ones falling almost right between the results from the two experimental studies. Such
a di!erence is perhaps not altogether surprising remembering that, "rst of all, the experimental
studies include a limited amount of di!erent materials and that, secondly, although the materials
studied by Tabor [2] and Chaudri [13] are initially well "tted to a power-law function according to
Eq. (16), the di!erent degrees of strain-hardening are achieved by cold work-hardening which will
result in stress}strain curves not accurately described by a rigid}plastic power-law relation, at least
not for heavily work-hardened specimens. Still, the numerical "ndings are in excellent agreement
with the concept of a single representative strain as the "nite element results are well "tted to
a curve
HI "2.8(0.15)K (20)
as also depicted in Fig. 5b. It could be noted that the result for m"0 suggests a relation
H"2.8p , (21)
W
being in good agreement with experimental "ndings by, for example, Tabor [2] and Chaudri [13]
pertinent to heavily work-hardened materials, in particular so when remembering that frictional
e!ects were not accounted for in the present analysis.
From the results above, it can be concluded that the concept of a representative strain holds
within good accuracy at indentation of a rigid}plastic power-law material. Accordingly, Eqs. (19)
and (20) can be used in a straightforward manner for material characterization of materials for
which adherence to Eq. (16) can be assumed a priori. A complete description of the material
properties, however, requires more information to be discussed in some detail below.

3.2. The hardness at indentation of a material with an irregular stress}strain relation

Before doing this though, it seems advisable to "rst investigate whether or not the results in Eqs.
(19) and (20) are valid also in a situation where the stress}strain characteristics of the material is not
accurately described by a rigid}plastic power-law relation. Remembering the experimental, and
numerical, results quoted above it could be argued that the stress value at a plastic strain of 0.15 in
Eq. (20) is essentially an average value of the e!ective stress under the indenter and that a more
P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920 909

Fig. 6. Finite element calculations performed in order to determine whether or not there exists a value of the
accumulated plastic strain above which the stress}strain characteristics of the material no longer in#uences the material
hardness: (a) Stress}strain curves used in the numerical simulations of Vickers indentation; (b) Vickers hardness as
function of the ultimate strain, e . (O) results for the materials 1}5 in Fig. 6a. (*), result for material 6, e "R. (- - -),
F F
result, according to Eq. (21), for an elastic}perfectly plastic material with p "175 MPa.
W

irregular stress}strain curve would produce results that deviates substantially from Eq. (20). As
a "rst step then, a numerical study was performed in order to determine whether or not stress levels
at plastic strains above 8}20% will in#uence the outcome of an indentation test. In doing so, cone
and Vickers indentation of the materials with uniaxial stress}strain curves according to Fig. 6a
were investigated numerically. Regarding the di!erent material behaviours, the materials all had
the same stress}strain characteristics, with E"70 GPa, l"0.3, p "175 MPa, except that hard-
W
ening was terminated at di!erent values of the uniaxial #ow stress aiming at a clear picture of the
in#uence from high stress levels on the material hardness. Furthermore, all curves, except the one
for material 1, had the same stress value at e +0.08, i.e. they should have the same material
N
hardness according to Tabor [2]. The results from the calculations are shown in Fig. 6b where the
numerically determined hardness is depicted as a function of the ultimate strain e , i.e. the strain at
F
which hardening is terminated. Basically, two conclusions of importance can be drawn from these
results. First of all, it is clear that stresses at strain levels much higher than 0.08 and 0.20, will a!ect
the hardness while secondly, the stress}strain characteristics at higher strain values than approxim-
ately 0.35 will not at all in#uence the outcome of a Vickers test, despite of the fact that for material
6, with an ultimate strain formally equal to in"nity, substantial strain-hardening will be present
also above this strain level. Some generality of results was also found as the conclusions drawn
910 P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920

above for the Vickers test were con"rmed by cone indentation calculations. It should be mentioned
that the strain level 0.35 agrees very well with the corresponding result for level II indentation
presented by Giannakopoulos et al. [7] and Larsson et al. [8].
The results presented above give clear evidence that a single representative strain value is not
su$cient in order to fully describe sharp indentation in level III. From a practical point of view
then, one of the aims of the present paper is to determine whether or not a more truthful description
of the relation between material hardness and stress}strain characteristics can be determined. Such
a relation must, of course, according to the results presented in Fig. 6b, include the #ow stress value
at 35% plastic strain but additional parameters are left to be determined. As a "rst attempt in this
direction it seems sensible to determine whether or not there exist also a lower limit on the uniaxial
stress}strain curve below which the strain-hardening of the material does not in#uence the
hardness. Obviously, no conclusions in this matter can be drawn from the results in Fig. 6 as the
materials investigated had basically the same hardening at low values on the e!ective plastic strain.
In order then to settle this matter numerical simulations of cone indentation were performed on the
materials presented in Fig. 7a. These materials all have the same hardening except that there is an
initial regime of perfectly plastic behaviour, with a yield stress p , up to a strain value e where the
W J
material hardening initiates. Such a material behaviour is, indeed, of some practical importance as
being pertinent to soft, carbon steel and was found appropriate to analyse in this context with the

Fig. 7. Finite element calculations performed in order to determine whether or not there exists a value of the
accumulated plastic strain below which the stress}strain characteristics of the material no longer in#uences the material
hardness: (a) Stress}strain curves used in the numerical simulations of cone indentation; (b) Cone hardness as function of
the strain value where hardening initiates, e . (O) results for the materials 2}7 in (a). (*) result for material 1, e "0.
J J
P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920 911

existence of a low representative stress value at issue. Regarding details about the stress}strain
curves in Fig. 7a, a reference material, material 1, adhering to a power law relation between #ow
stress and plastic strain according to Eq. (9), and with the constitutive properties
E"200 GPa, l"0.3, p "50 MPa, p "1000 MPa, m"3.3, (22)
W 
was used. At cone indentation, the hardness for this material was determined to be 1.56 GPa.
Only numerical simulations of cone indentation was performed in this case remembering
the good agreement between cone and Vickers indentation for the results presented in
Fig. 6b.
The outcome of these calculations are depicted in Fig. 7b. Judging from the results there exists
also a lower limit below which the particular shape of the stress}strain curve no longer will produce
any signi"cant changes in the hardness value. According to Fig. 7b this limit is, remembering the
scatter in numerical results discussed earlier, the stress value at a logarithmic plastic strain of 0.02,
approximately. It is interesting to compare the results from these calculations with a numerical
study by Bolshakov et al. [26], supplemented by nanoindentation experiments by Tsui et al. [27],
dealing with cone indentation of aluminium alloy 8009. The stress}strain curve for this particular
material show some hardening at small strain values but at approximately 2% plastic strain the
stress peaks at a constant value. Now, if, as the calculations show, stresses evaluated at strains
smaller than 0.02 have no substantial in#uence on the material hardness, this quantity should be
given directly from Eq. (15) with the yield stress replaced by the peak stress for the material
analysed by Bolshakov et al. [26]. Indeed, this is also the case, within very good accuracy, giving
further con"dence to the conclusion discussed above.
So far then, what has been determined is the boundaries of the in#uence from the stress}strain
curve on the hardness of an elastic}plastic material in a limited number of cases. It could be
expected that also the hardening characteristics of the material between these two boundaries
would in#uence the hardness. However, this it not at all obvious, remembering that also a one-
parameter description of sharp indentation tests, cf. e.g. [2] and [13], have been found to give
results of good accuracy for particular situations and a numerical study of the in#uence from
stresses at intermediate values of strains is certainly warranted. This was also performed presently
and in doing so, the material according to Eq. (22) was, again, chosen as a reference material.
Secondly, four other materials with stress}strain curves depicted in Fig. 8a, were analysed and their
hardness values determined. It should be noted in Fig. 8a that the materials 1}4 all have completely
di!erent strain-hardening characteristics except that the values on the stresses evaluated at the
plastic strain values 0.02 and 0.35 coincide. Additionally, material 5 has the same constitutive
characteristics as in Eq. (22) except for the elastic modulus that attains the value 70 GPa. This
material was included in order to also study the role of the elastic modulus. The outcome of this
investigation is depicted in Fig. 8b, where the parameter ! de"ned in Eq. (1) is used in order to
distinguish between the "ve di!erent materials, and as can be seen, the hardness values are
practically the same. Accordingly, judging from the results in Fig. 8b, the shape of the stress}strain
curve between the two boundaries, de"ned by the plastic strains 0.02 and 0.35, has a very limited
in#uence on the hardness. From a practical point of view, these results are encouraging as
they suggest that the hardness can be determined with some accuracy by a two-parameter
description where the parameters involved correspond to two stress levels on the uniaxial
stress}strain curve.
912 P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920

Fig. 8. Finite element calculations performed in order to determine whether or not stresses at intermediate values of
strains, i.e. strain values between 0.02 and 0.35, will in#uence the material hardness: (a) Stress}strain curves used in the
numerical simulations of cone indentation; (b) Cone hardness as function of ! "E tan b/((1!l)p(e "0.08)). (O)
F N
results for the materials 1}5 in (a). (*) average hardness value for all "ve materials.

In short then, the numerical simulations performed so far indicates that it is plausible to suggest
an approximate formula
H"C(b) f (p , p , b), (23)
J F
relating the stress}strain curve for an elastic}plastic material to the hardness. For clarity, it should
be stated that p and p in Eq. (23) are the stresses corresponding to the plastic strain values 0.02
J F
and 0.35, respectively, while C, as indicated, is a function of the indenter geometry alone. However,
this suggestion rests on results from a limited number of numerical simulations and it is therefore
the intention below to test the accuracy of Eq. (23) and, of course, also to determine the actual form
of the function f (p , p , b), as well as C(b). Accordingly, a large number of numerical simulations, of
J F
cone, Vickers and Berkovich indentations of elastic}plastic materials with a wide variety in the
shape of the stress}strain curves, were performed.
Before presenting the outcome of this numerical investigation though, it should be stated that for
reasons of applicability, simplicity must not be left out of consideration when aiming at an
approximative description of the mean contact pressure. Accordingly, a suggestion
H"C p #C p "C (p !p )#Kp , (24)
 J  F  J F F
P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920 913

C and C "K!C being constants depending only on the shape of the indenter, appears to be
  
a sensible "rst attempt. Furthermore, remembering the high accuracy results for ideal plasticity in
Eqs. (15) and (21), it seems advisable to enforce the values K"2.55 for a cone and K"2.8 for
a Vickers indenter. In this context, it also deserves to mention that it was found from the present
"nite element calculations that Berkovich hardness values are very close to the corresponding
Vickers ones in the rigid}plastic regime. Indeed, for the material de"ned in Eq. (22), the two
hardness values di!ered with less than 0.3%. This comes as no surprise though, as the Berkovich
indenter is designed in such a way that it will, at least nominally, displace the same volume as the
Vickers indenter at the same indentation depth. Accordingly, no distinction will be made between
Vickers and Berkovich indentation in the discussion below.
What remains then, in order to arrive at a complete description of the mean contact pressure by
Eq. (24), is to determine an appropriate value of C . As the results presented in Fig. 5 are pertinent

to a fairly general class of constitutive behaviour, rigid power-law plasticity, it seems advisable to
determine the C -values aiming at close adherence to these results, while not neglecting the need

for simplicity. In doing so, good agreement was found by choosing


1.0 (cone),
C " (25)
 1.4 (Vickers and Berkovich).

As an alternative, the constants C and C (or K) could have been determined directly from
 
Eq. (24), together with the results in Fig. 5, by using the least-squares method. Presently though,
with in particular high numerical accuracy in mind, it was thought advisable to enforce agreement
with the perfectly plastic solution. However, as can be seen in Fig. 9a (cone indentation), and in Fig.
9b (Vickers indentation), good agreement was found between numerical results and the approxi-
mative formulation according to Eq. (24), together with Eq. (25), and this matter was not dwelled
upon further presently. Regarding the results presented in Fig. 9, it should be immediately
emphasized once again that Eqs. (19) and (20) are based upon numerical results for (1/m)(0.7. For
this reason, the deviations between predicted hardness values given by these equations and the
corresponding ones given by Eq. (24), when m is close to 1, does not necessarily indicate that the
predictive capability of Eq. (24) is substantially reduced for these m-values. Indeed, it has been
shown by Biwa and Stora kers [25], in their analysis of Brinell indentation of rigid}plastic
power-law materials that numerically determined hardness values for m+1 are somewhat higher
than what is predicted by a single representative strain formula derived from numerical results at
high m-values.
Contemplating a bit further on the C -values in Eq. (25) it is perhaps surprising that there is

a substantial di!erence between the results for cone and Vickers indentation. The explanation for
this can be found when investigating the e!ective plastic strain trajectories shown in Fig. 10 for
cone and Vickers indentation of the material in Eq. (22). Clearly, the plastic strains are substantially
higher beneath the cone indenter, and in particular, close to the tip of the indenter, suggesting that
higher strain values will have a more signi"cant e!ect at cone indentation as also indicated by the
C -values presented in Eq. (25). It should be mentioned that this feature indicates a strong

relationship between the plastic strain, or the #ow stress, levels just beneath the indenter and the
material hardness. A great deal of e!ort was devoted to shed some light on this matter and to give
a more quantitative physical interpretation of the results presented above or to "nd an alternative
914 P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920

Fig. 9. Normalized hardness, HI "H/p , as function of 1/m at indentation of a rigid}plastic power-law material

according to Eq. (16): (a) Cone indentation (*), Eq. (24) together with the C -value in Eq. (25) pertinent to cone

indentation. (- - -), Eq. (19). (O) present numerical results; (b) Vickers indentation (*), Eq. (24) together with the C -value

in Eq. (25) pertinent to Vickers indentation (- - -), Eq. (20). (O) present numerical results.

expression with a more solid physical background. Presently though, it was not possible to, for
example, relate the variation in #ow stress at the contact surface to the contact pressure, and
thereby to the hardness through Eqs. (11) and (12), and this matter is still open for further
investigation.
With a suggestion for a general, although admittedly approximate, formula relating the material
hardness to stress}strain characteristics it now seems appropriate to investigate the accuracy of the
proposed relation. In doing so, in addition to the earlier results in Figs. 5}8, numerical calculations
were performed determining the material hardness at cone, Vickers and Berkovich indentation for
a large number of materials with a wide variety of di!erent elastic}plastic material characteristics
ranging from linear strain hardening and power-law hardening as de"ned in Eqs. (9) and (16) to
more irregular stress}strain relations. The numerical results, for strain-hardening materials exclus-
ively, were then compared with predictions given by Eq. (24), together with C -values from Eq. (25),

and the outcome is shown in Fig. 11a with normalized hardness values depicted as function of ! .
F
Clearly, there is good agreement between numerical results and predictions and the two sets of
results di!er with at most 10% save for a particular case discussed below. Indeed, Eq. (24) seems to
sometimes give results of high accuracy also when ! is smaller than 50, adopted here as the lower
F
boundary for rigid}plastic indentation response, even though reliability could not be expected
then. This is also shown in Fig. 11a as the ratio between predicted and numerically determined
P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920 915

Fig. 10. Trajectories, in the X }X plane, of the accumulated plastic strain, e , at indentation of the material de"ned in
  N
Eq. (22). The di!erence in strain value between two adjacent trajectories is 0.05 with the lowest value, e "0.05, indicated
N
in the "gures: (a) Cone indentation; (b) Vickers indentation.

hardness takes on the value 1.58 when !"16.7, clearly indicating the in#uence from elasticity.
Regarding the case mentioned brie#y above for which numerical results and predictions di!er with
more than 10%, this concerns two of the materials with a perfectly linear relation between #ow
stress and plastic strain at ! "56.4 and 493.3. While it could be argued that the deviation in one
F
of these cases is due to elastic e!ects at low ! -values this does not apply for ! "493.3.
F F
Accordingly, the conclusion must be that Eq. (24) can not be expected to give reliable results for
perfectly linear hardening materials, at least not in a consistent manner. However, such a material
is, in practice, not a realistic description of the material behaviour at high plastic strains.
A corresponding comparison as the one performed above, but now with experimental sharp
indentation results taken from the literature, is certainly also warranted. When performing such
a comparison, there exists an exhaustive amount of results from experimental investigations to
explore. However, this number is substantially reduced when demanding that, in order to be
included in the comparison, the materials investigated in the experimental studies should have
constitutive properties yielding an indentation response pertinent to the rigid}plastic regime and
that hardness values as well as uniaxial stress}strain characteristics must be clearly stated. With
these requirements in mind, it was found necessary to limit the number of investigations used in the
comparison to Tabor [2], Atkins and Tabor [6], Dugdale [12], Chaudri [13] and Tsui et al. [27].
The outcome is shown in Fig. 11b and obviously, there is good agreement between experimental
results and the predictions based on Eq. (24), together with C -values reported in Eq. (25), and

there are only two experimental results that deviate with more than 10% from the predicted value.
These results concern Vickers indentation of steel and annealed copper as reported by Dugdale
[12]. In both cases though, the di!erence between experiments and predictions are moderate,
H /H "0.89 and 1.14, respectively, and these deviations can most probably be explained by
NPCB CV
the fact that it was very di$cult to determine, in an accurate manner, the explicit stress values at
2 and 35% plastic strain from the stress-strain relations presented by Dugdale [12]. In average, the
916 P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920

Fig. 11. (a) Ratio between the hardness value predicted from Eq. (24), H , and the corresponding numerically
NPCB
determined hardness value, H , as function of ! "E tan b/((1!l)p(e "0.08)). (*) (H /H )"1. (- - -) approx-
LSK F N NPCB LSK
imate lower boundary for the region of rigid}plastic indentation response. (O) present results for cone indentation of
power-law materials according to Eqs. (9) or (16). () present results for cone indentation of linear strain-hardening
materials. () present results for cone indentation of materials with irregular strain hardening. (#) present results for
Vickers and Berkovich indentation of power-law materials according to Eqs. (9) or (16). (;) present results for Vickers
and Berkovich indentation of linear strain-hardening materials. (*) present results for Vickers and Berkovich indentation
of materials with irregular strain hardening. (b) Ratio between the hardness value predicted from Eq. (24), H , and the
NPCB
corresponding experimentally determined hardness value, H , as function of ! "E tan b/((1!l)p(e "0.08)). In some
CV F N
of the investigations quoted below, the elastic properties of the indented materials are not reported. In order to determine
an approximate value on ! in such a situation, E"210 GPa,l"0.3 and E"110 GPa,l"0.33 is used for steel and
F
copper, respectively (*), (H /H )"1. (O) results for cone indentation of copper and mild steel [6]. () results for
NPCB CV
Vickers indentation of copper and mild steel, Tabor [2]. () results for Vickers indentation of copper and alloy steels
[12]. (
) results for Vickers indentation of copper [13]. (*), results for Berkovich indentation of aluminium alloy 8009
[27]. Results for pre-stressed specimens are not shown.

ratio H /H in Fig. 11b is slightly below 1. However, frictional e!ects were not accounted for in
NPCB CV
the "nite element calculations and this feature would (cf. e.g. Ref. [16]), result in a small increase of
the numerically calculated hardness values, i.e. the ratio would increase slightly.
In short then, the outcome of the comparisons made in Fig. 11 is encouraging as it shows that,
for many materials of practical interest, Eq. (24) together with Eq. (25) will give high accuracy
predictions for the material hardness. Having said this though, it should once more be clearly
stated that this formula is, indeed, an approximate one as, for one thing, indicated by the fact that
the functional form of Eq. (24) is not in accordance with the result from self-similarity consideration
P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920 917

as reported in Eq. (18). Even though Eq. (18) is strictly valid only for a rigid plastic power-law
material, the deviation in form between Eqs. (18) and (24) clearly underlines the approximative
nature of the two-parameter description presented presently. A further drawback of the present
approach, in contrast to the situation with a single representative stress value, concerns the fact
that at material characterization by indentation a two-parameter description will not give the
absolute values of the #ow stress at a particular strain value without some prior knowledge of the
stress}strain characteristics. On the other hand, when determination of residual stress or strain
levels are at issue this matter constitutes no problem as in general in such a situation the
constitutive behaviour of the material is known and direct advantage can be taken of the
predicative capability of Eq. (24) (cf. [16]).

3.3. Relationship between contact area and indentation depth

Having focused, so far, almost exclusively on the behaviour of the mean contact pressure it now
seems appropriate to investigate in some detail features related to the contact area resulting at
sharp contact problems. This is of substantial importance at material characterization by indenta-
tion as the hardness, at best, can give valuable, but incomplete, information about the plastic
properties of the material and a full description of the constitutive behaviour requires further
information from other variables such as the size of the contact area between the indenter and the
material. This feature being even more pronounced when a two-parameter description of the mean
contact pressure is utilized as in such a situation the absolute value of the #ow stress at a given
strain level cannot be determined from the material hardness alone.
Starting then by "rst recognizing once again that, due to the self-similarity of the boundary value
problem, the invariant c, de"ned in Eq. (14), is a convenient parameter often used to characterize
the contact area at indentation. It then seems appropriate to also study the fundamental behaviour
of this parameter at sharp indentation and cone indentation of elastic-perfectly plastic materials
then appears as the obvious choice for such an investigation. The outcome of the numerical
analysis is depicted in Fig. 12a and, surprisingly, it was found that c is substantially more sensitive
than the hardness (cf. Fig. 4), to elastic e!ects and for a wide variety of material properties c&ln !.
Indeed, this result is valid for approximately 5(!(400, i.e. for many standard engineering
materials of practical importance and, furthermore, the present calculations show that the
strong in#uence from elasticity on c is found also at indentation of strain-hardening materials.
When material characterization by sharp indentation is at issue, this feature must be considered
as a drawback as it indicates that the invariant c can not be used, in general, as a measure
of the strain-hardening properties of the material in the entire rigid plastic region, without
properly accounting for elastic e!ects. Judging from the experience from the corresponding
procedure for the material hardness, and in particular, then in a situation with an irregular relation
between #ow stress and plastic strain, any attempt to correlate in an approximate manner the value
of c with the stress}strain characteristics of the material in the spirit of Eq. (24) will be speculative
at best and high accuracy could de"nitely not be expected. Accordingly, this was not aimed at
presently.
Despite the discussion above though, a truly rigid}plastic solution including both the hardness
and the size of the contact area is still valid for many materials of practical interest and in particular
so for power-law materials accurately described by Eq. (16), for which by de"nition elastic e!ects
918 P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920

Fig. 12. (a) Indentation invariant, c, as function of !"E tan b/((1!l)p ) at cone indentation of elastic}perfectly
W
plastic materials. (O) present numerical results. (b) Indentation invariant, c, as function of 1/m at indentation of
a rigid}plastic power-law material according to Eq. (16). (O, - - -) present numerical results, and a "t to the results, for
cone indentation. (, - ) -) present numerical results, and a "t to the results, for Vickers indentation.

are negligible. Indentation of such materials are, of course, of substantial interest presently and this
matter was studied in some detail in order to arrive at results of some generality. The outcome of
the investigation is presented in Fig. 12b where c is depicted as function of the power-law exponent
m at cone and Vickers indentation of such materials. In this context it should be stated (cf. e.g. Ref.
[14]), that in a rigid power law plastic situation c is a function of m alone. Regarding the results for
Vickers indentation shown in Fig. 12b it is obvious that the indentation invariant c is sensitive to
three-dimensional e!ects which must be properly accounted for at material characterization.

4. Conclusions

Sharp indentation of elastic}plastic materials in the rigid}plastic regime has been investigated
numerically by aid of the "nite element method and the most relevant "ndings can be summarized
as follows.

E The global properties given by an indentation test show fundamentally di!erent behaviour as
the size of the contact area proved to be very sensitive to elastic e!ects. This is not so for the
P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920 919

mean contact pressure, or the material hardness, as in such a situation, a rigid}plastic solution is
su$cient in most cases of practical interest.
E At indentation of rigid}plastic power-law materials the hardness is well-described by a single
representative strain level in the spirit of Tabor [2].
E In a general situation, i.e. at indentation of materials with more irregular stress}strain relations,
the concept of a single representative strain is no longer valid. Presently for this case, an
alternative two-parameter description of the mean contact pressure is suggested with the two
parameters corresponding to the stress levels at 2 and 35% plastic strain. The practical
usefulness of this two-parameter description is mainly limited to a situation where substantial
information about the stress}strain characteristics of the material is known prior to indentation.
As a "nal comment, it should be emphasized that the results presented are also valid for more
general problems involving sharp contacts. As examples of such applications, typically contacts in
gears or in electronic devices could be mentioned even though a complete picture of these problems
also must include detailed information about "eld variables.

References

[1] Pethica JB, Hutchings R, Oliver WC. Hardness measurements at penetration depths as small as 20 nm. Philosophi-
cal Magazine 1983;A48:593}606.
[2] Tabor D. Hardness of metals. Oxford: Oxford University Press, 1951.
[3] Johnson KL. The correlation of indentation experiments. Journal of the Mechanics and Physics of Solids
1970;18:115}26.
[4] Johnson KL. Contact mechanics. Cambridge: Cambridge University Press, 1985.
[5] Doerner MF, Nix WD. A method for interpreting the data from depth-sensing indentation instruments. Journal of
Materials Research 1986;1:601}9.
[6] Atkins AG, Tabor D. Plastic indentation in metals with cones. Journal of the Mechanics and Physics of Solids
1965;13:149}64.
[7] Giannakopoulos AE, Larsson P-L, Vestergaard R. Analysis of Vickers indentation. International Journal of solids
and Structures 1994;31:2679}708.
[8] Larsson P-L, SoK derlund E, Giannakopoulos AE, Rowcli!e DJ, Vestergaard R. Analysis of Berkovich indentation.
International Journal of Solids and Structures 1996;33:221}48.
[9] Bhattacharya AK, Nix WD. Finite element simulation of indentation experiments. International Journal of Solids
and Structures 1988;24:881}91.
[10] Bhattacharya AK, Nix WD. Analysis of elastic and plastic deformation associated with indentation testing of thin
"lms on substrates. International Journal of Solids and Structures 1988;24:1287}98.
[11] Laursen TA, Simo JC. A study of microindentation using "nite elements. Journal of Materials Research
1992;7:618}26.
[12] Dugdale DS. Vickers hardness and compressive strength. Journal of the Mechanics and Physics of Solids
1958;6:85}91.
[13] Chaudri MM. Subsurface strain distribution around Vickers hardness indentations in annealed polycrystalline
copper. Acta Materialia 1998;46:3047}56.
[14] Stora kers B, Biwa S, Larsson P-L. Similarity analysis of inelastic contact. International Journal of Solids and
Structures 1997;34:3061}83.
[15] Suresh S, Giannakopoulos AE. A new method for estimating residual stresses by instrumented sharp indentation.
Acta Materialia 1998;46:5755}67.
[16] Carlsson S, Larsson P-L. 2000, in preparation.
[17] ABAQUS. Users manual, version 5.6. Pawtucket, RI: Hibbitt, Karlsson and Sorensen, Inc., 1996.
920 P.-L. Larsson / International Journal of Mechanical Sciences 43 (2001) 895}920

[18] Giannakopoulos AE, Larsson P-L. Analysis of pyramid indentation of pressure-sensitive hard metals and ceramics.
Mechanics of Materials 1997;25:1}35.
[19] Carlsson S, Biwa S, Larsson P-L. On frictional e!ects at inelastic contact between spherical bodies. International
Journal of Mechanical Sciences 1999;42:107}28.
[20] Fleck NA, Hutchinson JW. A phenomenological theory for strain gradient e!ects in plasticity. Journal of the
Mechanics and Physics of Solids 1993;41:1825}57.
[21] Fleck NA, Muller GM, Ashby MF, Hutchinson JW. Strain gradient plasticity: theory and experiment. Acta
Metallallurgia Materialia 1994;42:475}87.
[22] Shu JY, Fleck NA. The prediction of a size e!ect in microindentation. International Journal of Solids and
Structures 1998;35:1363}83.
[23] Larsson P-L, Giannakopoulos AE. Tensile stresses and their implication to cracking at pyramid indentation of
pressure-sensitive hard metals and ceramics. Materials Science and Engineering 1998;A254:268}81.
[24] Lockett FJ. Indentation of a rigid/plastic material by a conical indenter. Journal of the Mechanics and Physics of
Solids 1963;11:345}55.
[25] Biwa S, Stora kers B. An analysis of fully plastic Brinell indentation. Journal of the Mechanics and Physics of Solids
1995;43:1303}33.
[26] Bolshakov A, Oliver WC, Pharr GM. In#uences of stress on the measurement of mechanical properties using
nanoindentation: Part II. Finite element simulations. Journal of Materials Research 1996;11:760}8.
[27] Tsui TY, Oliver WC, Pharr GM. In#uences of stress on the measurement of mechanical properties using
nanoindentation: Part I. Experimental studies in an aluminium alloy. Journal of Materials Research 1996;11:752}9.

Вам также может понравиться