Surface & Coatings Technology 200 (2006) 4153 – 4165

www.elsevier.com/locate/surfcoat

Review

Critical review of analysis and interpretation of nanoindentation test data

A.C. Fischer-Cripps*
CSIRO Industrial Physics, PO Box 218, Lindfield NSW 2070, Australia

Received 5 January 2005; accepted in revised form 18 March 2005

Available online 30 April 2005

Abstract

This paper reviews commonly used methods of analysing and interpreting nanoindentation test data, with a particular emphasis on the
testing of thin films. The popularity of nanoindentation testing is evidenced by the large number of papers that report such measurements in
recent years. Unfortunately, there appear to be several issues that are emerging as common sources of error in using this technique. The
present paper is aimed at highlighting these errors for the benefit of those practitioners who wish to use the technique but are not fully
conversant with the field.
D 2005 Elsevier B.V. All rights reserved.

Keyword: Nano-indentation

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4153
2. Indenter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4154
3. Contact mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4155
3.1. Loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4155
3.2. Maximum load. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4156
3.3. Unloading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4156
4. Fitting the load – displacement curve . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4158
5. Corrections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4159
5.1. Initial penetration or zero point determination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4159
5.2. Frame compliance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4159
5.3. Area function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4161
5.4. Materials issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4161
6. Interpretation of the results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4162
6.1. Elastic modulus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4162
6.2. Hardness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4163
7. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4164
Acknowledgement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4165
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4165

1. Introduction

Nanoindentation testing is a fairly mature technique

* Tel.: +61 2 9413 7544; fax: +61 2 9413 7547. which uses the recorded depth of penetration of an
E-mail address: [email protected]. indenter into the specimen along with the measured
0257-8972/$ - see front matter D 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.surfcoat.2005.03.018
4154 A.C. Fischer-Cripps / Surface & Coatings Technology 200 (2006) 4153 – 4165

applied load to determine the area of contact and hence area is that which does take into account the sloped sides
the hardness of the test specimen. Many other mechanical of the impression. Initially, one might expect that it would
properties can also be obtained from the experimental be more correct to always work with the actual surface
load –displacement curve, the most straight-forward being area of the residual impression and this is indeed what the
the elastic modulus. Other properties such as the strain- standard Vickers hardness number is based upon. How-
hardening index, fracture toughness, yield strength and ever, the indentation load, divided by this actual area does
residual stress can also be obtained in some circum- not have a real physical significance other than it has
stances. The method relies on there being an accurate been found to provide a useful measure of hardness in an
determination of the initial contact of the indenter with engineering context. When one divides the indenter load
the specimen surface, corrections for any penetration that by the projected area of the impression, one obtains the
arises during this initial contact, corrections for compli- mean contact pressure which is a physical quantity. The
ance of the loading column, corrections for the departure connection between the hardness of a material and the
of ideal shape of the indenter, and corrections for mean contact pressure has some particular importance,
materials-related issues such as piling-up and indentation especially for thin film specimens and this will be
size-effect, residual stress, etc. Review articles and books discussed in a later section.
provide a good starting point for those entering the field Experience shows that the face angles of a Berkovich
[1 –5]. However, it is the present author’s experience that indenter can be made with high accuracy. Errors in the
there is emerging a common set of misconceptions and final results from variations in face angles of indenters
errors in the application of the technique. The present from a reputable supplier are usually negligible. The
paper seeks to highlight these issues for the benefit of reason why a three-sided Berkovich indenter is used in
those who are not conversant with the method but wish to nanoindentation work over the more familiar Vickers
report their work using the best available practices. indenter is that a sharper tip is possible when the indenter
is ground. With a four-sided Vickers pyramid, there is
always an undesirable line of conjunction at the tip. With
2. Indenter a three-sided Berkovich indenter, it is much easier to
grind the faces of the indenter to meet at a single point
The three-sided Berkovich indenter is the most popular than a line. An ideal Berkovich indenter tip has an
geometry for nanoindentation testing. A typical Berkovich infinitely sharp radius, but this is impossible to achieve in
indenter tip is shown in Fig. 1. The face half-angle of a practice. Usually, the tip radius of a new Berkovich
Berkovich indenter used for nanoindentation work is indenter is in the order of 50 –150 nm with an error of
65.27- å 65.3-. This gives the same projected area to approximately 50% in the estimation. Very small radius
depth ratio as the more commonly used four-sided Vickers tips are usually obtained from sorting a production batch.
indenter (face angle 68-) used in microhardness testing. It As a result of the finite radius of the indenter tip, it is
is important to distinguish between the projected area of necessary to employ contact mechanics equations appro-
impression and the actual area of impression. The priate to a spherical indenter for the initial part of the
projected area is that which is observed when looking loading/unloading cycle, and equations for an equivalent
face-on to a residual impression in a specimen surface at conical indenter for the higher loading and unloading
normal incidence to the surface without regard to the portions. To proceed, we first examine the equations of
actual sloping sides of the impression. The actual contact elastic contact for a spherical indenter.

(a) (b)

θ
hc

Fig. 1. (a) High-magnification SEM scan of the tip of a Berkovich diamond indenter and (b) schematic of indenter geometry. A is the projected contact area, h
is the face angle, and h c is the depth of the circle of contact measured from the apex of the indenter.
 A.C. Fischer-Cripps / Surface & Coatings Technology 200 (2006) 4153 – 4165 4155

3. Contact mechanics (a) (b)

Elastic
1. R contact
3.1. Loading a

Mean contact pressure pm

Hertz formulated elastic equations of contact for con- Transition
2. region
tacting spherical surfaces in the latter part of the 19th
century [6]. The most significant equation from Hertzian 2. 3.
H
contact mechanics relates the contact radius a to the
Fully-developed
combined radius R, the applied load P, and the elastic
1. plastic zone
properties of the contacting bodies. For the condition 3.
R H a, the Hertz formula is:
3 PR Indentation strain a/R
a3 ¼ : ð1Þ Plastic
4 E4 zone
In most cases of interest in nanoindentation testing, the
indenter is in contact with what is essentially a semi-
infinite half space and so R is the radius of the indenter Fig. 2. (a) Indentation stress vs. indentation strain for spherical indenter and
alone. In like manner, the elastic modulus E* is the (b) schematic of evolution of the plastic zone.
combined elastic modulus of the contacting bodies. In
practice, the indenter is usually made from diamond and
so, for most materials, E* is dominated by the elastic comes a point where any increase in load results in a
properties of the specimen. The relationship between the proportional increase in the contact radius and so the
two is given by: mean contact pressure becomes constant (ignoring any
 2 strain hardening effects). Under these conditions of a fully
1 ð 1  v2 Þ 1  v¶ Þ
developed plastic zone, we call the mean contact pressure
¼ þ ð2Þ
E4 E E¶ the hardness H of the specimen.
where E¶ and v¶, and E and v, describe the elastic modulus It is important to understand that there is a transition
and Poisson’s ratio of the indenter and the specimen, from purely elastic contact to a plastic contact even for a
respectively. The significance of Eq. (1) is that it may be Berkovich indenter. The relative contribution of the elastic
expressed in terms of the mean contact pressure, p m, as a portion to the shape of the overall load displacement
function of the indentation strain, defined as the ratio of curve depends upon the indenter tip radius and the yield
the contact radius a over the indenter radius R: strength of the specimen. The sharper the tip, the less the
  contribution of the initial elastic response. For an initial
P 4 E4 a elastic contact, the mean contact pressure increases with
pm ¼ 2 ¼ : ð3Þ
pa 3 p R increasing load. At the condition of a fully developed
plastic zone, the mean contact pressure levels off to a
The square bracketed terms in Eq. (3) are constants constant value with increasing load and this value of
and so for a completely elastic contact, the mean contact mean contact pressure is called the hardness. In fact,
pressure, or indentation stress, is linearly proportional to hardness, as a material property, is very difficult to define
the indentation strain as shown by the solid line in Fig. 2 since it depends on how it is measured. Indentation tests
where the slope is proportional to E*. Consider the are very commonly used and so the hardness so obtained
magnitude of the contact stresses, in particular, the shear is often called the ‘‘indentation hardness’’ to distinguish it
stress in the specimen material as load is applied to a from other methods that might employ, say, a scratch
spherical indenter. For a fully elastic response, the technique.
principal shear stress for indentation with a spherical Once the penetration depth becomes larger than the tip
indenter is a maximum at å 0.47p m at a depth of å 0.5a radius, the pyramidal shape of the indenter becomes the
beneath the specimen surface directly beneath the indenter dominant geometrical feature of the indentation. The
[7]. We may employ either the Tresca or von Mises shear contact then usually involves an appreciable amount of
stress criteria, where plastic flow occurs at s å 0.5Y, where plastic deformation within the specimen. As we shall see,
Y is the material yield stress, to show that plastic it is the data taken on the unloading that is used to
deformation in the specimen beneath a spherical indenter determine the area of contact at the maximum load. This
can be first expected to occur when p m å 1.1Y [8,9]. As is done by using equations of contact for a conical
load is further applied, the plastic zone grows in size and indenter where we mathematically transform the actual
this results in a leveling off of the indentation stress as pyramidal geometry into an equivalent cone. That is, we
shown by the dashed line in Fig. 2. This indicates a find a cone angle that gives the same area-to-depth ratio
departure from the linear elastic response (Eq. (3)). There as the actual pyramidal indenter. This can be easily
4156 A.C. Fischer-Cripps / Surface & Coatings Technology 200 (2006) 4153 – 4165

calculated from geometry where for a Berkovich indenter 3.2. Maximum load
of face angle h, we have:
pffiffiffi During an indentation test, as the indenter is loaded into
A ¼ 3 3h2c tan2 h ð4Þ the specimen, the triangular geometry of the Berkovich
indenter eventually dominates any tip-rounding effect and
and for a conical indenter of half-apical angle a we have: the deformation has the property of geometrical similarity.
This means that unless one has an external reference, it is
A ¼ ph2c tan2 a ð5Þ
not possible to set a scale of the deformation since the load,
where h c is the distance as measured vertically from the depth and size of the plastic zone all scale at the same rate.
indenter tip as shown in Fig. 1(b). Equating Eqs. (4) and Contact with a spherical indenter does not have the property
(5) for h = 65.27-, we have a = 70.296- å 70.3-. of geometrical similarity unless it is expressed in terms of
Using the equations of contact for a conical indenter, it is the ratio a/R. Usually, the maximum load is selected so as to
possible, using a reverse analysis of the methods to be be certain that a fully developed plastic zone in the
described below, to predict the shape of the load – displace- specimen material has been achieved.
ment curve even when the contact is fully ‘‘plastic’’ in the At full load, a hold period is often applied to account for
sense of a fully developed plastic zone. It can be shown that, creep effects before the indenter is unloaded. Creep and
assuming a sharp transition from elastic to plastic behaviour, thermal drift in nanoindentation testing are very important
the load –displacement curve, where is the total depth of sources of error and are difficult to distinguish. In both
penetration from the specimen free surface, can be predicted cases, for a constant load hold period, the depth reading may
according to [1]: change. In the case of thermal drift, the depth reading may
" increase (deeper into the specimen) with time or decrease.
rffiffiffiffiffiffi  rffiffiffiffirffiffiffiffiffiffi#2 When creep occurs, the depth reading usually increases.
1 E4 2ð p  2Þ p H
P ¼ E pffiffiffi
4
þ h2 : Thermal drift is not drifting of the electronic signals
ptana H p 4 E4
(although this can happen) but is usually brought about by
ð6Þ changes in the dimension of the contact (indenter, indenter
shaft, and specimen) from thermal expansion or contraction
In this equation, E* is the combined modulus (Eq. (2)),
due to temperature changes. Thermal drift can usually be
H is the hardness (assumed to be constant—i.e. no strain-
minimized by enclosing the apparatus in an insulating
hardening) and a is the equivalent cone angle (70.3- for a
enclosure or waiting a sufficiently long time after handling
Berkovich indenter). The important feature of this equation
to allow thermal equilibrium to be established before
is that the load is related to the square of the displacement.
beginning the test. Creep occurs due to the movement of
The overall shape of the load –displacement curve is shown
material within the specimen under high pressure in the
in Fig. 3.
indentation stress field. In some materials, creep effects may
50 Power law fitting:
slowly dissipate while in others, they may increase as long
B: 193.1034 as the load is applied. The amount of creep can also depend
hr: 0.3085 on the rate of application of load. For materials with a
m: 1.2784 significant amount of creep, it can sometimes be impossible
40 Unloading analysis:
hi : 0.00803 μm to conduct a nanoindentation test in the usual way, and
Pt : 49.73 mN instead, a visco-elastic creep test may be required. For our
ht : 0.65575 μm present purposes, we shall assume that creep and thermal
30 hc : 0.44086 μm
drift issues are not significant and that load is applied quasi-
P (mN)

dP/dh : 183.90 mN/μm

a : 1.2327 μm statically, i.e. the deformation is time-independent.
A : 4.7738 μm2
20 H : 10.4172 GPa
3.3. Unloading
E*(s+i): 72.2778 GPa
E (s) : 75.3504 GPa
After maximum load, or the optional hold period, the
10 applied load is reduced and the resulting penetration depth
recorded. During unloading, there is usually some degree of
elastic recovery of the specimen material as the elastically
0 strained material outside of the plastic zone relaxes and tries
0 0.2 0.4 0.6 0.8 to resume its original shape. This is usually prevented from
h (μm) happening entirely by the presence of the plastic zone which
can be now considered ‘‘dead’’ material. There remains in
Fig. 3. Typical load – displacement curve with a Berkovich indenter on
fused silica at 50 mN. The inset shows values of various quantities of the specimen surface a residual impression. On the load –
interest. (Note: no area correction was applied for the analysis shown in the displacement curve, elastic recovery is evidenced by a
inset). reduction in displacement with decreasing load. For a fully
 A.C. Fischer-Cripps / Surface & Coatings Technology 200 (2006) 4153 – 4165 4157

elastic contact, the unloading curve lies on top of the where a is the radius of the indenter equal to the radius of the
loading curve. For an elastic –plastic contact, the unloading circle of contact and where it should be noted that the depth of
curve is different to the loading curve and the area enclosed penetration is linearly dependent on the load. For this reason,
between them represents the energy lost (as heat) during some instruments prefer to use Eq. (10) rather than Eq. (9) and
plastic deformation. The unloading behaviour is assumed to we shall discuss the significance of this later.
be an entirely elastic affair. That is, even though there is a Referring now to Eq. (9), the derivative of the load P
plastic zone formed, during unloading, this remains ‘‘inert’’ with respect to the displacement h (the contact stiffness) is
and the unloading deformations that do occur are all elastic. given by:
It is assumed that there is no ‘‘reverse’’ plasticity (i.e. no
dP 2E 4 tana
plastic deformation occurring during unloading). It is the ¼2 h: ð11Þ
unloading data that is ultimately used to determine the dh p
radius of the circle of contact and hence the contact area. Substituting into Eq. (9), we have:
The slope of the unloading curve at any point is called the
1 dP
contact stiffness. As will be shown below, the contact P¼ h: ð12Þ
stiffness can be used in conjunction with the calculated 2 dh
contact area to provide a measurement of the combined If the total depth of penetration is h t, at load P t, then as the
elastic modulus of the system. load is removed, the indenter moves through a distance h e as
Treating the contact as an axis-symmetric cone, the shown in Fig. 4.
contact between a rigid conical indenter and an elastic half At P = P t the displacements h r=0 = h e and h r=a = h a . From
space is found from [10]: Eq. (8) at r = a, the plastic (or contact) depth h c is found
pa 4 from:
P¼ E acota; ð7Þ  
2 2ð p  2Þ Pt
hc ¼ ht  ; ð13Þ
where a is the cone semi-angle. The quantity acota is the p dP=dh
depth of penetration h c measured at the circle of contact. where P t and dP/dh are measured during an experiment.
The depth beneath the specimen free surface within the The square-bracketed term in Eq. (13) is often given the
circle of contact is given by: symbol e and evaluates to 0.72 but it is common practice to
p r use a value of 0.75 since this has been shown to account for
h¼  acota r V a: ð8Þ non-uniformities in the material response as the load is
2 a
withdrawn. Once a value for h c has been determined, the
With r = 0, and substituting into Eq. (7), we obtain: area of contact is found from Eq. (5) where, for a Berkovich
4  indenter (a = 70.3-), Eq. (5) evaluates to:
2E
P¼ tana h2 ; ð9Þ
p A ¼ 24:5h2c : ð14Þ
where the load is proportional to the square of the The reduced elastic modulus is found from [11]:
penetration depth. Eq. (9) should be compared with Eq. pffiffiffi
(6) where it can be seen that for the case of a conical dP 1 p
E ¼
4
p ffiffiffi : ð15Þ
indenter, the square relationship between the load and the dh 2 A
depth features in both cases, although with different Experiments and finite element analysis shows that a
constants of proportionality. correction factor is needed for Eq. (15). The correction
For the case of the unloading data in a nanoindentation
test, we consider the contact conditions at this point to be
entirely elastic. That is, we can unload and reload as many a
times as we like and travel down and up the unloading curve
ha
without any further plastic deformation taking place. unloaded
Experiments show that in nearly all cases, reloading the
residual impression to the same maximum load does indeed
take place up the unloading curve implying that the hf
unloading and the reloading is elastic. So, the unloading
ht
curve can be treated as if it were an elastic loading hc
according to Eq. (6). Historically, it was noted that the he α
initial unloading response for many materials was linear.
This implied that the contact conditions were those of a loaded
cylindrical punch instead of a cone where:
Fig. 4. Cross-section of profile of specimen surface at full load, and full
P ¼ 2aE 4 h; ð10Þ unload for an elastic – plastic indentation. Symbols are defined in the text.
4158 A.C. Fischer-Cripps / Surface & Coatings Technology 200 (2006) 4153 – 4165

factor is given the symbol b and is applied as the factor 1/b In the derivation given in Section 3, it can be seen that to
to the measured value of dP/dh. Accordingly we have: determine the contact depth h c, we need to evaluate the
pffiffiffi contact stiffness dP/dh at full load P t. This is usually done
1 dP 1 p by fitting an equation to the unloading data, then finding the
E ¼
4
pffiffiffi ð16Þ
b dh 2 A derivative of that equation which gives the contact stiffness
dP/dh. If the data were to satisfy Eq. (6), then all we would
and need to do is to fit a polynomial expression of the 2nd
Pt degree to the data, perhaps using a least squares approach,
hc ¼ ht  eb ð17Þ and then find the derivative and the slope.
dP=dh
Here is where we return to the significance of Eq. (10).
where dP/dh in Eqs. (16) and (17) is that as actually Initially, it was noticed that the unloading response, at
measured from the experimental data. King [12] performed least for the initial part, was fairly linear. This means that
finite element calculations for elastic indentations formed instead of a 2nd order polynomial (Eq. (9)), the contact
with flat-ended punches of triangular and circular cross- appeared to have a linear dependence on the load as if the
sections and found b = 1.034. Dao et al. [13] determined indenter were a cylindrical punch (as per Eq. (10)). For
b = 1.096 for a Berkovich indenter and b = 1.0595 for a some materials, the initial unloading data is indeed almost
70.3- conical indenter from a 3D elastic – plastic finite linear and a linear fit to the upper portion of the
element analysis (note that 1.096/1.0595 = 1.034). Cheng unloading data is entirely reasonable. However, for other
and Cheng [14] determined, on the basis of elastic – plastic materials, particularly those with a relatively large amount
finite element calculations with a 68- conical indenter at of elastic recovery (low value of the ratio E/H), the
v = 0.3, that b = 1.05 for non-work-hardened materials. unloading data was observed to be curved, but more than
Larsson et al. [15] estimated b = 1.14 on the basis of a 3D would be expected on the basis of a conical indenter (Eq.
elastic solution for a Berkovich indenter using the finite (9)). In most cases, a 2nd degree polynomial fit to the
element method. Hay et al. [16] calculated b = 1.067 on the unloading data provides a reasonable fit to the data, but
basis of radial displacements within the circle of contact for for best results, a power law fit is appropriate. For a
an elastic contact with a conical indenter (a = 70.3- and power law fit, we fit the data using:
v = 0.3) from finite element results. Borodich and Keer [17] P ¼ Bðh  hf Þm ð18Þ
proposed b = 1.058 (at v = 0.17) on theoretical grounds for
full adhesive (non-slip) contact for a conical indenter. where m is the power law index, B is a constant, and h f
Martin and Troyon [18] measured b = 1.063 experimentally is the final residual depth measured from the original
with a Berkovich indenter. Gong et al. [19] evaluated specimen free surface, all of which are unknown
b = 1.032 from experimental results involving a consider- quantities. Fitting can only be done using an iterative
ation of the presence of the residual stress field on the procedure (see Appendix 4 in Ref. [1]). The fitting
unloading response. In carefully considering the above, it procedure relies on there being an initial guess for B, m
appears that the factor 1.034 reasonably represents the and h f, all of which can be obtained by an initial 2nd
change from axis-symmetric to a pyramidal cross-section degree fit. The iterative procedure then provides the best
while an additional factor å 1.06 is required on the basis of fit values for these unknowns based upon a least squares
a variety of causes ranging from radial displacements to approach.
interfacial friction. The product of the two gives 1.09. A When a power law fitting is done, it is found that the
similar survey was conducted by Oliver and Pharr [5] who index m ranges from 1.1 to 1.8 depending on the specimen
suggested a value of 1.05 as a reasonable estimate. material. The reason for the index not being precisely equal
Evidently, the factor b should be greater than the commonly to 2 appears to be due to that fact that the contact described
used 1.034 but at present there appears to be no consensus by Eq. (9) assumes that the notional conical indenter is in
as to why the factor arises and what value it should take. contact with a conical residual impression. The presumption
is, and this is not often appreciated, that the specimen
material elastically recovers so that the sides of the residual
4. Fitting the load – displacement curve impression remain straight. This is essentially the origin of
the factor 2 in Eq. (13). Careful measurement with an AFM
We are now in a position to analyse the experimental and finite element analysis shows that the sides of the
data in the form of a load – displacement curve. In its most residual impression are usually curved upwards. According
basic form, the curve consists of a quasi-static loading to to Pharr and Bolshakov [20], this results in contact actually
full load followed by an unloading to zero load. From the occurring between a notional elastic half space and an
point of view of the analysis equations given above, it is indenter with an arbitrary shape that provides the necessary
the initial unloading data that is the most significant observed unloading response. Thus, we need not be too
although we shall see that the other data also is of concerned about the actual shape of the effective indenter
considerable importance. because it is embodied in the power law index m in Eq. (18).
 A.C. Fischer-Cripps / Surface & Coatings Technology 200 (2006) 4153 – 4165 4159

The notion of the effective shape of the indenter also Afterwards, a smooth curve is fitted to the initial data and
accounts for the commonly used choice of e = 0.75 [20]. then extrapolated back to zero force. This provides a
The analysis procedure given above is often called the measure of the initial contact depth which is then added
‘‘Oliver and Pharr’’ method after the authors who gave it on to all the subsequent depth readings. The initial
particular study [21]. Some practitioners often misapply the penetration correction has the effect of shifting the load –
Oliver and Pharr method by using a linear fit to the displacement curve to the right as shown in Fig. 5.
unloading data, using a polynomial fit (2nd degree) without The initial penetration h i is applied to all the displace-
actually checking to see if the fitted line corresponds to the ment readings to give a corrected depth of penetration.
actual experimental data or using a power law fit but instead
h¶ ¼ h þ hi ð19Þ
of finding values of B, h f and m from the data, use values
verbatim from Oliver and Pharr’s paper. In most cases, the where h¶ is the corrected value of the penetration depth and
resulting errors are not large because it is the initial h is the depth recorded by the instrument. The significance
unloading data that is usually used for the fitting. However, of the initial penetration correction depends upon the
the recent ISO standard on instrumented indentation testing magnitude of the maximum load applied to the indenter
[22] recommends using 80% of the unloading data in which relative to the magnitude of the initial contact force. The
case the fitting procedure becomes very important. The lowest force that can be detected or set in a typical
reason for using the initial part of the unloading data stems nanoindenter is in the order of 2– 5 AN. Typically, the
from Doerner and Nix [23] who wished to only use the initial penetration is a few nanometers. For a typical
observed linear portion of the data and Eq. (10). If a power maximum load of over 10 mN in a moderately hard and
law fit is used, it is perhaps more appropriate to use a greater stiff material, the initial penetration correction will make a
proportion of the unloading data so as to obtain a more few percent difference to the computed values of E and H.
precise fit as suggested in ISO 14577. Because the contact
stiffness is evaluated at the maximum load, there is a large 5.2. Frame compliance
‘‘leverage’’ effect which magnifies errors in determining the
contact depth from Eq. (13). Errors in the quantity dP/dh Frame compliance is the contribution to the depth
affect the computed value of h c. Since the area of contact readings arising from deflections of the load frame instead
depends on h c2 (Eq. (14)), the error in the calculated area will of displacement into the specimen material. When load is
be twice the percentage error in h c. applied, the reaction force is taken up by deflection of the
load frame and it is this deflection that is typically added to
the depth readings. The deflection is usually linearly
5. Corrections proportional to the applied load in the manner of a linear
spring.
As mentioned previously, it is necessary in most nano- The correction takes the form of the product of the
indentation testing to account for various sources of error in known instrument compliance C f and the force P and is
the instrumented data before arriving at a final value for subtracted from the depth reading as recorded by the
modulus and hardness of the test specimen. Three of the
most common, and necessary corrections are the initial
penetration, the instrument frame compliance, and the 4
indenter area function. There are also some materials-related
issues which are also important but more difficult to correct
for, the main ones being the presence of indentation size 3
effects, and piling-up of material around the indentation.
P (mN)

5.1. Initial penetration or zero point determination 2

A very important procedure in nanoindentation testing is

the determination of the initial contact between the indenter
1
and the specimen. This is important because the contact
point determines the datum of the displacement measure-
ment. There are a variety of ways in which the contact point 0
is estimated. In one method, the force and depth signals are
-0.05 0 0.05 0.1 0.15 0.2
constantly monitored at a rapid rate and the software or the
operator determines the point of contact from a sharp rise in h (μm)
hi
the force data. In another method, an initial contact force, set
by the user, is commanded, and once achieved, the depth Fig. 5. Initial penetration h i is estimated by extrapolating the recorded
sensor is then set to zero and the loading then proceeds. load – displacement data back to zero load.
4160 A.C. Fischer-Cripps / Surface & Coatings Technology 200 (2006) 4153 – 4165

instrument (note, the initial penetration correction should be 0.003

applied first):

h¶ ¼ h  PCf : ð20Þ
0.002

The effect of the compliance correction is to shift the

dh/dP
load –displacement curve to the left, somewhat negating
the effect of the initial penetration correction. However, the
compliance correction is dependent on the load and is not 0.001
a constant (as is the initial penetration correction) so it is
not possible to completely balance the two effects over the
whole load range of the test. The compliance correction is
particularly important when testing very stiff materials. In 0
this case, the contribution to the depth signals from the 0 2 4 6 8
frame compliance is a greater proportion of the overall 1/hc
signal. If not correctly applied, the compliance correction
can make up to a 50% difference in the resulting Fig. 6. Plot of dP/dh vs. 1/h c for data taken on sapphire over a load
estimations for modulus and hardness. range of 20 mN to 100 mN A linear fit to the data gave an intercept
C f = 0.00014 Am/mN.
The compliance correction obviously requires a knowl-
edge of the value of C f. Often, this is measured by the
instrument manufacturer and is set internally within the
software. While this might provide some measure of to give more weight to the data from the higher end of
correction, trusting this value can lead to significant the load range in determining the value of C f.
problems. In particular, it should be noted that the factor A common mistake in operating or calibrating a
C f contains all the compliances of the system which are nanoindentation instrument is to determine the compli-
linearly dependent upon the applied force. These com- ance C f using only one reference standard. For example,
pliances are those of the load frame, the indenter shaft it is common practice to use fused silica to determine the
and (and what is often overlooked) the specimen area function (see Section 5.3) of the indenter and to also
mounting. A common mistake is to mount the specimen estimate the compliance correction. What is not often
using a thick layer of mounting resin which itself is quite appreciated is that the two procedures are inextricably
compliant and may not be accounted for by the linked so that one may indeed arrive at a value of C f and
manufacturer’s figure for C f. The best way to measure an area function which gives perfectly good results for
and verify the value of C f is to conduct a series of tests at tests on materials with a similar value of modulus and
different loads on a fairly stiff specimen (e.g. sapphire). hardness as the reference specimen, but may give very
The contact stiffness dP/dh as measured by the instrument poor results for harder and stiffer materials. For best
actually consists of the actual stiffness of the contact at results, one should determine the C f on a variety of
the indenter, given the symbol S, and the reciprocal of the standard specimens (fused silica, silicon and sapphire
instrument compliance C f. Expressed in terms of com- provide a very good range) and make sure that the value
pliances, we have: of C f is consistent for each. The area function should
also be calculated using the new value of C f and then C f
dh 1 re-evaluated until there is a convergence to a fairly
¼ þ Cf : ð21Þ constant value. This does not mean that tests on the
dP S
standard specimen should be repeated until convergence
Interpreting Eq. (21) in terms of the analysis method is obtained, but simply re-analysed.
given above, we have: In some instruments, compliance corrections are
"rffiffiffiffiffiffiffiffiffi # eliminated, or nearly so, by using a differential method
dh p 1 1 of measuring indenter penetration. In this method, the
¼ þ Cf ð22Þ depth sensor datum is located on a reference ring which
dP 24:5 2E4 hc
typically advances ahead of the indenter and makes
where we can see that a plot of dh/dP vs. 1/h c gives a contact with the specimen surface outside the prospective
straight line with an intercept C f. Fig. 6 shows such a plot indentation site. This arrangement also serves to minimize
with real data taken on sapphire with a Berkovich thermal drift errors. The disadvantage is that the method
indenter. Note that the data at the higher end of the is restricted to relatively hard materials (lest the reference
range arises from load – displacement data at the low end ring sink into the specimen or the force applied to the
of the depth measurement and has a greater error reference ring causes unwanted local deformations in the
associated with it. In some cases, it may be necessary specimen surface).
 A.C. Fischer-Cripps / Surface & Coatings Technology 200 (2006) 4153 – 4165 4161

10
(23) (e.g. terms that correspond to tip rounding and indenter
face angle), such attempts tend to ignore local variations in
8 geometry. Eq. (23), flexible as it might seem, ignores the
fact that such local irregularities are unlikely to present
themselves at regular harmonic intervals in the indenter
6 geometry, or conversely, use of such an equation might
A/Ai

introduce irregularities of a harmonic nature where none are

justified. That is, it depends on whether the scatter in Fig. 7
4
represents the real irregular shape of the indenter or is a
result of some other cause not associated with the actual
2 indenter shape. The issue is very important since even small
errors in the area function make a significant difference to
the final results obtained with the instrument. It is the
0 present author’s experience that fitting equations such as
0 0.4 0.8 1.2 1.6 that above lead to systematic errors over the load range of
hc (μm) the instrument since the fitting is never perfect over this
range. It is far better to smooth out some of the data using a
Fig. 7. Area function for a Berkovich indenter expressed in terms of the
running average which retains any short to medium length
ratio A/A i vs. h c where A i is the ideal area of contact for a perfect indenter.
Convergence to A/A i = 1 at large values of h c indicates that the face angles scale deviations in the shape of the indenter while
of the indenter are correct. The sharp rise in the value of A/A i at low values smoothing out small length scale irregularities from surface
of h c indicates tip rounding. roughness of the standard reference specimen.
When establishing an area function, it is important to do
so over the whole load range of the instrument. If
5.3. Area function measurements are to be taken at relatively low loads, then
more data points should be taken at the low end of the load
The area function of the indenter is a function or a table range, giving particular attention to the point at which the
of values that provides the best estimate of the area of area function ratio shows a sharp increase. In Fig. 7, the
contact as a function of the contact depth for the indenter results from four separate indentation tests were combined
being used. As mentioned previously, indenters are usually to form a single area function for the indenter which covered
made with high precision and accuracy with regard to the penetration depths up to about 2 Am.
face angles, but the ideal geometry often breaks down near
the very tip where tip rounding has occurred (either through 5.4. Materials issues
use of the indenter or just from the manufacturing process).
The area function can be found using a variety of methods. Even if the corrections listed above are attended to,
An independent measure using an AFM or similar profiling there remain various materials-related effects which can
instrument may be beneficial but is usually very inconven- cause errors in the estimation of modulus and hardness of
ient. It is more common to test a material of known modulus the test specimen. The most significant of these materials
(or hardness) and to use Eq. (15) to determine the value of A issues is that of piling-up. To understand what piling-up
as a function of h c over the load range of the instrument. is, it is necessary to be aware of the opposite effect of
This procedure usually results in a range of data points to sinking-in. Fig. 4 shows a cross-section of the profile of
which a smoothing equation can be applied. the specimen surface under full load. It is evident that the
In many cases, the area function equation takes the form: circle of contact lies below the specimen free surface. A
certain amount of sinking-in of the surface is expected,
A ¼ C1 h2c þ C2 hc þ C3 h1=2 1=4
c þ C 4 hc þ N ð23Þ
even for a purely elastic contact. For a contact involving
where C 1 is usually a number close to 24.5 and the
remaining fitting constants account for the tip rounding and
Sinking-in Piling-up
other departures from the ideal shape. For example, Fig. 7
a
shows a typical area function for a new Berkovich indenter. a
There is evidently some scatter in the data, particularly (and
expectedly) at the lower end of the load range. This scatter
arises from instrument errors (particularly near the low end
hc
of the instrument range), surface roughness, and irregu- ht
hc
larities in the indenter. Should one then smooth out this
scatter using an equation such as Eq. (23) above? While it is
tempting to express area function equations in terms of
physically meaningful parameters in equations such as Eq. Fig. 8. Schematic of piling-up compared to sinking-in.
4162 A.C. Fischer-Cripps / Surface & Coatings Technology 200 (2006) 4153 – 4165

plastic deformation, there is the possibility of the surface 6. Interpretation of the results
actually being raised up at the edge of the indentation as
illustrated in Fig. 8. Assuming now we have made the necessary corrections
When piling-up does occur (as if often the case with and obtained values of the reduced modulus and hardness, it
indentation testing of metals), the material which has is now required that we correctly interpret the data. The
piled-up supports the load applied to the indenter and so following discussion is directed towards the testing of thin
the indenter penetrates less compared to the case where films but some of the issues are applicable to tests on bulk
there is no piling-up. However, the analysis method specimens. Of particular interest in the testing of thin films
discussed in Section 3.3 does not take this into account is the well-known 10% rule. This rule states that in order to
and so the material appears to be stiffer and harder. The minimize the effect of the substrate properties of the
area of contact A, computed from the combination of Eqs. measurement of the film properties, the indentation depth
(15) and (18), is less than the actual area and so both E should be no more than 10% of the film thickness. This
and H are over-estimated. These over-estimates can be widely quoted and often misused rule has its basis in the
quite substantial, up to 50% in some cases. The only determination of hardness testing where the intention was to
satisfactory method to eliminate this error is to measure restrict the extent of the plastic zone to the film material.
the area of contact directly. An AFM image is probably Common practice has transferred the rule to the determi-
the most convenient, but even so, requires considerable nation of modulus although there is no real physical basis to
capital outlay and investment in time. There are some support its use for that purpose.
analytical [24,25] treatments which deal with piling-up but
these generally apply to particular materials system. 6.1. Elastic modulus
What is not often appreciated is that piling-up need not
actually be described by h c > h t. Any increase in the value When testing a thin film deposited on a substrate using
of h c over that which is expected by the equations of an indentation technique, most of the deformation is
contact used in the unloading analysis represents a piling- localized to the vicinity of the indenter and so we can
up even if h c < h t. Further, the amount of piling-up for ignore any large-scale bending or compression of the
even nominally brittle materials may take place if the specimen (or hope that these deformations are accommo-
loading is high enough. For example, Fig. 9 shows piling- dated by the compliance correction). The indentation stress
up in an impression in fused silica with a 70.3- conical field is a complex mixture of biaxial stresses, there being
indenter under a load of 175 mN. Piling-up is perhaps the elements of tension, compression and shear. In the simplest
most important unresolved issue in nanoindentation test- treatment, we can imagine that the indenter bears down on
ing. There are two ways to approach the problem: (i) an elastic film that in turn is connected to an elastic
perform an area function calibration on a reference substrate. The reaction force is thus supplied by a linear
specimen which has a similar ratio of E/H as the test combination of two springs in series. Of course, in reality,
specimen (such that piling-up is accommodated within the the reaction force is supplied by material within the film
area function correction) or (ii) ignore the effect and treat acting radially inwards from many directions and not simply
the resulting values of E and H as comparative values. in the vertical direction, but the simple linear model does
Although (ii) may sound somewhat unsatisfactory, that is reinforce the important result that the elastic modulus so
precisely what is implicitly recommended by ISO 14577 measured, no matter how small the penetration depth, is a
in the labeling of the quantities as the ‘‘indentation combined modulus of the film and substrate system. The
modulus’’ and the ‘‘indentation hardness’’. task then is to separate out the substrate modulus (or
compliance) so that we are left with a measurement of the
modulus of the film. There are several methods by which
this is done.
The simplest method involves taking a series of
indentations over a range of loads and then plotting the
elastic modulus as a function of penetration depth. It is
necessary to obtain a sufficient number of data points at low
loads so that the data near the lower end of the depth axis
captures, as much as possible, the film properties. Once
obtained and plotted, a smooth curve is fitted to the data and
8 μm extrapolated back to zero depth (where there is of course no
data point) as shown in Fig. 10. The vertical axis intercept is
then taken as the best estimate of the film modulus. Simple
Fig. 9. AFM scan of a residual impression in fused silica after indentation as it is, the method can provide inaccurate results for the
with a 70.3- conical indenter at 175 mN. Piling-up is in evidence around the unwary because it assumes that the error in the area function
edge of the impression. used to analyse the data is small, even at very small
 A.C. Fischer-Cripps / Surface & Coatings Technology 200 (2006) 4153 – 4165 4163

600 I o is in effect a weighting function that equals zero as the

film thickness approaches zero, and approaches unity for
Ef
large values of film thickness. A simpler approach is given
500 by Jung et al. [27] who propose an empirical power law
relationship:
E (GPa)

L
400 Ef
Ec ¼ Es þ ; ð27Þ
Es
ES
300 where
1
L¼ ; ð28Þ
200
1 þ Aðh=t ÞC
0 0.2 0.4 0.6 0.8 1 where the constants A and C are found a best-fit analysis of
ht (μm) the data. More formal treatments involving numerical
solutions of complex exact analytical functions are given
Fig. 10. Estimation of elastic modulus of super-hard site nc-TiN/a-BN film
by Stone [28], Yoder et al. [29], Schwarzer [30], Chuboba et
(5 Am thick) on a steel substrate using the extrapolation technique. At large
indentation depths, the combined modulus tends towards the substrate al. [31], and also Perriot and Barthel [32].
modulus E S. Extrapolation of the data to zero depth provides an estimate of As a first order approximation, the extrapolation techni-
the film modulus E f. que provides a readily accessible method for determining
the film-only modulus. Where data cannot be easily treated
penetrations. Consider the area function shown in Fig. 7. in this way, a simple empirical approach such as that given
Evidently, any estimates of modulus at depths of penetration in Eq. (27) is probably the next best method. Where the
below about 150 nm will result in the area function ratio situation is complicated by a large mismatch in properties
being taken on the steeply rising part of the curve where the between film and substrate, or the film is very thin, then the
error in the estimation of the ratio will be very high. One more formal treatments are appropriate. It should be noted
should then treat the modulus data at these depths with some in closing this section that a sharp tip indenter is not
caution. It is easy for the plot of modulus vs. depth to be just required for a measurement of elastic modulus. Indeed, a
reflecting the shape of errors in the area function rather than spherical indenter may be more beneficial since the area
any real trends in the film properties. A better approach function data at small penetration depths will not be so
would be to perform this estimation for a range of sharply rising (as in Fig. 7) and so the associated errors in
specimens of different thickness and perhaps even on the estimation of the contact areas will be likely to be much
different substrates. smaller.
As an alternative to the above procedure, we may
recognize from the outset the combined nature of the elastic 6.2. Hardness
modulus values and attempt to fit a suitable constitutive
equation to determine the film modulus on its own. There Assuming then that we have a very sharp Berkovich tip,
are a variety of treatments available. For example, King [12] how then is the hardness of the film determined? The 10%
used the finite element method to obtain the combined rule provides a reasonable starting point. The assumption
modulus of the film of thickness t and substrate E c: behind the rule is that if the total depth of penetration of the
 
1 1  v2f Þ  pffiffiffi 
at= A 1  v2s Þ  at=pffiffiAffi  indenter is less than 10% of the film thickness, then the
¼ 1e þ e plastic zone associated with the indentation will be
Ec Ef Es
contained entirely within the film and there will be a
 2
1  v¶ Þ negligible contribution from any elastic deformation of the
þ ; ð24Þ substrate. This is a somewhat simplified assumption but one
E¶
where a is an empirical constant. Gao et al. [26] proposed: that may provide reasonable values in many cases,
especially when the film is not too thin and the indenter
Ec ¼ Es þ ðEf  Es ÞIo ; ð25Þ not too blunt.
where I o is a function of t/a (where a is the contact radius) A slightly better approach is to perform a series of
given by: indentations at increasing depths of penetration and then
"  plotting the calculated hardness against depth. Much like the
2 1 t t ðt=aÞ2 extrapolation technique for modulus in the previous section,
Io ¼ tan þ ð1  2vÞ ln 1 þ
p a a ðt=aÞ2 we draw a smooth curve through the data. Fig. 11 shows
  data obtained on a super-hard nc-TiN/a-BN film.
t=a 1
 2
: ð26Þ There are several points of interest to be noted in Fig. 11.
1 þ ðt=aÞ 2p ð 1  vÞ The initial rise in hardness at low penetration depths is
4164 A.C. Fischer-Cripps / Surface & Coatings Technology 200 (2006) 4153 – 4165

60
More formal methods of estimating the film-only hard-
ness rely on mathematical modeling of the film – substrate
system. As far back as 1973, Bückle [33] proposed that the
50 composite hardness of a thin film/substrate system H c could
be found from:
H (GPa)

Hc ¼ Hs þ aðHf  Hs Þ; ð29Þ
40
where H s is the hardness of the substrate and H f is the
hardness of the film. In this equation, a is an empirically
derived parameter. Jonsson and Hogmark [34] propose an
30 area law of mixtures such that:
Af As
Hc ¼ Hf þ Hs ; ð30Þ
A A
20
0 0.2 0.4 0.6 0.8 1 where A f and A s are the relative parts of the contact area
ht (μm) carried by the film and the substrate, respectively, and A is
the total contact area. A similar treatment, but based on a
Fig. 11. Measured hardness of a super-hard nano-composite nc-TiN/a-BN volume of deformation law of mixtures, was proposed by
film (5 Am thick) on a steel substrate. Three regimes, initial rise, plateau and Burnett and Rickerby [35,36]:
fall-off can be identified.
Vf Vs
Hc ¼ Hf þ Hs : ð31Þ
V V
usually, but mistakenly, ascribed on an indentation size
effect (ISE), the details of which are usually not divulged. In For a soft film on a hard substrate, Bhattacharya and Nix
fact, an initial rise in hardness, in most cases and especially [37] propose that the combined hardness is determined by:
in very hard materials, is entirely expected on the grounds  !
Yf Es h 2
that the indenter tip, at this scale of testing, is initially round Hc ¼ Hs þ ðHf  Hs Þexp  ; ð32Þ
and so the contact is elastic. The value of hardness Ys Ef t
calculated under these conditions is the mean contact where Y f and Y s are the material yield stresses for the film
pressure, which, for a given load, will be less than the and substrate, respectively, and the exponential is a
hardness of the specimen material. Once we have a fully weighting function. For hard films on a soft substrate, the
developed plastic zone, the hardness of the film should expression becomes:
reach a constant or plateau. At larger loads, or greater rffiffiffiffiffiffi 
depths, the influence of the substrate will then be felt and Hf Ys Es h
Hc ¼ Hs þ ðHf  Hs Þexp  : ð33Þ
the hardness may increase or decrease according to the Hs Yf Ef t
relative hardness of the film and substrate. It is important to
use the sharpest possible tip when one desires to measure Many other models attempt to correlate the combined or
the hardness of a very thin film. If the indenter is too blunt, effective hardness with the hardness of the individual
then most of the deformation may be elastic and the mean components [27,38 – 40]. Given so many treatments, which
contact pressure so measured will not reflect the film should be chosen? In most practical situations, if the film is
hardness. thick enough and the indenter sharp enough, the simple
The identification of these three regimes of hardness extrapolation technique is probably the most straight-
(initial rise, plateau, and falling-off or increase) are forward and accessible. The analytical treatments require a
important validity checks on the data so obtained and certain sense of ‘‘calibration’’ on specimens of known
quoted for the film-only hardness. (a) Should there be no hardness. However, the advantage of these techniques is that
initial rise, the cause should be investigated. It is unlikely the hardness of films on substrates whose properties are not
that the indenter is infinitely sharp. More likely, the load known may be measured.
applied is too large and the indenter has gone through to
the substrate without providing any data points for the
film. (b) Should there be no plateau region, then it is 7. Conclusion
likely that the indenter is too blunt and so a fully
developed plastic zone could not be established before it The present paper seeks to clarify the issues involved in
extended into the substrate. (c) Should there be no the analysis of nanoindentation test data so that practitioners
obvious fall-off or increase in hardness at large pene- may avoid common mistakes made in the analysis and
tration depths, then perhaps the film and substrate have presentation of their data. Common mistakes are the failure
similar mechanical properties or the film is very thick to recognize the initial rise in hardness as resulting from
relative to the maximum indentation depth. elastic contact from the rounded tip of the indenter,
 A.C. Fischer-Cripps / Surface & Coatings Technology 200 (2006) 4153 – 4165 4165

improper fitting of the area function data, failure to account [13] M. Dao, N. Chollacoop, K.J. Van Vliet, T.A. Venkatesh, S. Suresh,
for instrument compliance or check that the instrument Acta Mater. 49 (2001) 3899.
[14] Y.-T. Cheng, C.-M. Cheng, J. Appl. Phys. 84 (3) (1998) 1284.
compliance is independent of the sample being tested, not [15] P.-L. Larsson, A.E. Giannakopoulos, E. Soderlund, D.J. Rowcliffe, R.
fitting the unloading data with an iterative technique or Vestergaard, Int. J. Solids Struct. 33 (2) (1996) 221.
selecting too many data points when using a polynomial or [16] J.C. Hay, A. Bolshakov, G.M. Pharr, J. Mater. Res. 14 (6) (1999)
linear fitting, failure to account for piling-up, and failure to 2296.
[17] F.M. Borodich, L.M. Keer, Proc. R. Soc. Lond., A 460 (2004) 507.
correctly determine the film-only modulus and hardness by
[18] M. Martin, M. Troyon, J. Mater. Res. 17 (9) (2002) 2227.
insufficiently spaced data. All these issues can have a [19] J. Gong, H. Miao, Z. Peng, Acta Mater. 52 (2004) 785.
significant effect on the quality of the data presented and [20] G.M. Pharr, A. Bolshakov, J. Mater. Res. 17 (10) (2002) 2660.
also their interpretation. [21] W.C. Oliver, G.M. Pharr, J. Mater. Res. 7 (4) (1992) 1564.
[22] ISO 14577, Metallic materials—Instrumented indentation test for
hardness and materials parameters, ISO Central Secretariat, 1 rue de
Varembé, 1211 Geneva 20 Switzerland.
Acknowledgement [23] M.F. Doerner, W.D. Nix, J. Mater. Res. 1 (4) (1986) 601.
[24] K.W. McElhaney, J.J. Vlassak, W.D. Nix, J. Mater. Res. 13 (5) (1998)
The author is indebted to Martin Soh for interesting 1300.
discussions on the issues discussed in this paper. [25] A. Bolshakov, G.M. Pharr, J. Mater. Res. 13 (4) (1998) 1049.
[26] H. Gao, C.-H. Chiu, J. Lee, Int. J. Solids Struct. 29 (20) (1992) 2471.
[27] Y.-G. Jung, B.R. Lawn, M. Martyniuk, H. Huang, X.Z. Hu, J. Mater.
Res. 19 (10) (2004) 1.
References [28] D.S. Stone, J. Mater. Res. 13 (11) (1998) 3207.
[29] K.B. Yoder, D.S. Stone, R.A. Hoffman, J.C. Lin, J. Mater. Res. 13 (11)
[1] A.C. Fischer-Cripps, Nanoindentation, 2nd edR, Springer-Verlag, N-Y, (1998) 3214.
2004. [30] N. Schwarzer, ASME J. Tribol. 122 (4) (2000) 672.
[2] A.C. Fischer-Cripps, Vacuum 58 (4) (2000) 569. [31] T. Chudoba, N. Schwarzer, F. Richter, Thin Solid Films 355 – 356
[3] J.L. Hay, G.M. Pharr, ASM Handbook, Materials Testing and (1999) 284.
Evaluation, vol. 8, 2000, p. 232. [32] A. Perriot, E. Barthel, J. Mater. Res. 10 (2) (2004) 600.
[4] Y. Gogotsi, V. Domnich (Eds.), High Pressure Surface Science and [33] H. Bückle, in: J.W. Westbrook, H. Conrad (Eds.), The Science of
Engineering, Institute of Physics, Bristol, UK, 2004. Hardness Testing and its Applications, American Society for Metals,
[5] W.C. Oliver, G.M. Pharr, J. Mater. Res. 19 (1) (2004) 3. Metals Park, OH, 1973, p. 453.
[6] H. Hertz, J. Reine Angew. Math. 92 (1881) 156 (Translated and [34] B. Jonsson, S. Hogmark, Thin Solid Films 114 (1984) 257.
reprinted in English in Hertz’s Miscellaneous Papers, Macmillan and [35] P.J. Burnett, D.S. Rickerby, Thin Solid Films 148 (1987) 41.
Co., London (1896) Ch. 5). [36] P.J. Burnett, D.S. Rickerby, Thin Solid Films 148 (1987) 51.
[7] M.T. Huber, Ann. Phys. 14 (1) (1904) 153. [37] A.K. Bhattacharya, W.D. Nix, Int. J. Solids Struct. 24 (12) (1988)
[8] D. Tabor, The Hardness of Metals, Clarendon Press, Oxford, 1951. 1287.
[9] K.L. Johnson, Contact Mechanics, Cambridge University Press, [38] S.J. Bull, Mater. Res. Symp. Proc. 750 (2003) Y6.1.1.
Cambridge, 1985. [39] A.M. Korunsky, M.R. McGurk, S.J. Bull, T.F. Page, Surf. Coat.
[10] I.N. Sneddon, Proc. Camb. Philos. Soc. 44 (1948) 492. Technol. 99 (1998) 171.
[11] G.M. Pharr, W.C. Oliver, F.R. Brotzen, J. Mater. Res. 7 (3) (1992) 613. [40] E.S. Puchi-Cabrera, J.A. Berrios, D.G. Teer, Surf. Coat. Technol. 157
[12] R.B. King, Int. J. Solids Struct. 23 (12) (1987) 1657. (2002) 185.