Nanomechanical Characterization of Solid Surfaces and Thin Films PDF

Nanomechanical characterisation of solid surfaces
and thin films

Bharat Bhushan and Xiaodong Li
Depth-sensing nanoindentation measurement

techniques are commonly used to measure
nanomechanical properties of surface layers of
bulk materials and of ultrathin coatings. The
nanoindentation apparatus continuously monitors
the load and the position of the indenter relative to
the surface of the specimen (depth of an indent)
during the indentation process. Mechanical
property measurements can be made at a
minimum penetration depth of about 20 nm (or a
plastic depth of about 15 nm). In this paper,
mechanical property measurements using
nanoindentation techniques are reviewed and
discussed. Emphasis is given to the use of sharp
indenters and how they can be used to measure
hardness, elastic modulus, continuous stiffness,
scratch resistance, filmsubstrate adhesion,
residual stresses, time dependent creep and
relaxation properties, fracture toughness, and
fatigue. Applications of the continuous stiffness
measurement technique to the characterisation of
multilayered structures and graded materials are
presented. Recent progress in the development of
the nanofatigue measurement technique is briefly
reviewed. Directions for future research in
nanoindentation are discussed.
IMR/398
The authors are in the Nanotribology Laboratory for
Information Storage and MEMS/NEMS, The Ohio State
University, 206 West 18th Avenue, Columbus, OH
432101107, USA (bhushan.2@osu.edu).
2003 IoM Communications Ltd and ASM International.
Published by Maney for the Institute of Materials, Minerals
and Mining and ASM International.
Introduction
Mechanical properties of solid surfaces and thin lms
are of interest as the mechanical properties aVect the
tribological performance of surfaces (Bhushan, 1999a,
2001, 2002). Among the mechanical properties of
interest, one or more of which can be obtained using
commercial and specialised indentation testers,
are elasticplastic deformation behaviour, hardness,
Youngs modulus of elasticity, scratch resistance, lm
substrate adhesion, residual stresses, time dependent
creep and relaxation properties, fracture toughness,
and fatigue. Indentation measurements can assess
structural heterogeneities on and underneath the surface, such as diVusion gradients, precipitates, presence
of buried layers, grain boundaries, and modi cation
of surface composition. Physical contacts at sliding
interfaces in magnetic storage devices and micro/
nanoelectromechanical systems (MEMS/NEMS)
occur at very low loads, thus, friction and wear of
sliding surfaces is primarily controlled by the physical
and chemical properties of a few surface atomic layers.
Ultrathin lms, as thin as a couple of nanometres,
DOI 10.1179/095066003225010227
are used in many applications including magnetic

storage devices and MEMS/NEMS (Bhushan 1996,
2001). Measurements of surface layers and ultrathin
lms require specialised instrumentation. Some of the
common apparatuses are the depth-sensing nanoindenter and the atomic force microscope (AFM)
(Bhushan, 1999b).
Since the early 1980s, the study of mechanical
properties of materials on the nanoscale has received
much attention, as these properties are size dependent
(Bhushan et al., 1996b). These studies have been
motivated partly by the development of nanocomposites and the application of nanometre thick lms
for miniaturisation of engineering and electronic components, and partly by newly available methods of
probing mechanical properties in small volumes. The
nanoindenter is maturing as an important tool for
probing the mechanical properties of small volumes
of material (Bhushan, 1999b). Indentation load
displacement data contain a wealth of information.
From the loaddisplacement data, many mechanical
properties such as hardness and elastic modulus can
be determined without imaging the indentation
impression. The nanoindenter has also been used to
measure the fracture toughness and fatigue properties
of ultrathin lms, which cannot be measured by
conventional indentation tests. With a tangential force
sensor, nanoscratch and wear tests can be performed
at ramping loads. Atomic force microscopes are ideal
for imaging of nanometre scale indentations, providing useful information about nanoindentation
deformation and cracking (Bhushan, 1999b). When
an indentation system is used in conjunction with an
atomic force microscope, in situ imaging can be
obtained. For the development of nanoindentation
techniques during the past decade, see Materials
Research Society Symposium Proceedings (Baker
et al., 1994, 2000; Drory et al., 1995; Gerberich et al.,
1996; Cammarata et al., 1997, Moody et al., 1998;
Vinci et al., 1999; Ozkan et al., 2001).
In this review, rst a commercial depth-sensing
nanoindentation test apparatus will be discussed followed by data analysis and the use of nanohardness
apparatuses for the determination of various mechanical properties of interest. The emphasis here is on
the measurement of ultrathin lms.
Measurement of mechanical
properties
Hardness implies resistance to local deformation. The
most commonly used hardness measurements are:
scratch hardness and static indentation hardness
(Tabor, 1951). Scratch hardness is the oldest form of
hardness measurement. It depends on the ability of
one material to scratch another or to be scratched
by another solid. Solid and thin- lm surfaces are
International Materials Reviews 2003
Vol. 48
No. 3
125
126
Bhushan and Li
Nanomechanical characterisation of solid surfaces and thin films
scratched by a sharp stylus made of hard material,

typically diamond, and either the loads required to
scratch or fracture the surface or delaminate the lm
or the normal/tangential loadscratch size relationships are used as a measure of scratch hardness
and/or interfacial adhesion (Heavens, 1950; Tabor,
1951, 1970; Benjamin and Weaver, 1960; Campbell,
1970; Ahn et al., 1978; Bhushan, 1987; Wu, 1991;
Bhushan et al., 1995, 1996a; Bhushan and Gupta,
1995; Gupta and Bhushan, 1995a,b; Blau et al., 1997).
The methods most widely used in determining the
hardness of materials are (quasi ) static indentation
methods. Indentation hardness is essentially a measure of their plastic deformation properties and only
to a secondary extent of their elastic properties. There
is a large hydrostatic component of stress around the
indentation, and since this plays no part in plastic
ow the indentation pressure is appreciably higher
than the uniaxial ow stress of the materials. For
many materials, it is about three times as large, but
if the material shows appreciable elasticity, the yielding of the elastic hinterland imposes less constraint
on plastic ow and the factor of proportionality
may be considered less than 3. Indentation hardness
depends on the time of loading and on the temperature and other operating environmental conditions.
In the indentation tests, a spherical, conical, or pyramidal indenter is forced into the surface of the
material and forms a permanent (plastic) indentation
in the surface of the material to be examined. The
hardness number (GPa or kg mm 2), equivalent to
the average pressure under the indenter, is calculated
as the applied normal load divided by either the
curved (surface) area (Brinell, Rockwell, and Vickers
hardness numbers) or the projected area (Knoop and
Berkovich hardness numbers) of contact between the
indenter and the material being tested under load
(Lysaght, 1949; Berkovich, 1951; Tabor, 1951, 1970;
Mott, 1957; ONeill, 1967; Westbrook and Conrad,
1973; ASME, 1979; Blau and Lawn, 1986; Bhushan
and Gupta, 1997; VanLandingham et al., 2001; FisherCripps, 2002).
In a conventional indentation hardness test, the
contact area is determined by measuring the indentation size using a microscope after the sample is
unloaded. At least for metals, there is little change in
the size of the indentation on unloading so that the
conventional hardness test is essentially a test of
hardness under load, although it is subject to some
error due to varying elastic contraction of the indentation (Stilwell and Tabor, 1961). More recently, in
depth-sensing indentation hardness tests, the contact
area has been determined by measuring the indentation depth during the loading/unloading cycle
(Pethica et al., 1983; Blau and Lawn, 1986; Bravman
et al., 1989; Doerner et al., 1990; Nix et al., 1992;
Oliver and Pharr, 1992; Nastasi et al., 1993; Townsend
et al., 1993; Bhushan et al., 1995, 1996a; Bhushan and
Gupta, 1995; Gupta and Bhushan, 1995a,b; Bhushan,
1996, 1999b; Hainsworth et al., 1996; Hay et al.,
1999). Hardness data can be obtained from depthsensing instruments without imaging the indentations
with high reproducibility. This is particularly useful
for the small indentations required for hardness
measurements of extremely thin lms. Depth measureInternational Materials Reviews 2003
Vol. 48
No. 3
Extended load range of static indentation

hardness testing
ments have, however, a major weakness arising from

piling up and sinking in of material around the
indentation. The measured indentation depth needs
to be corrected for the depression (or the hump) of
the sample around the indentation, before it can be
used for calculation of the hardness (Doerner and
Nix, 1986; Doerner et al., 1986; Nix, 1989; Oliver and
Pharr, 1992). Youngs modulus of elasticity is the
slope of the stressstrain curve in the elastic regime.
It can obtained from the slope of the unloading curve
(Oliver and Pharr, 1992).
In addition to measurements of hardness and
Youngs modulus of elasticity, static indentation tests
have been used for measurements of a wide variety
of material properties such as elasticplastic deformation behaviour (Pethica et al., 1983; Doerner and Nix,
1986; Stone et al., 1988; Fabes et al., 1992; Oliver and
Pharr, 1992), ow stress (Tabor, 1951), scratch resistance and lmsubstrate adhesion (Heavens, 1950;
Tabor, 1951; Benjamin and Weaver, 1960; Campbell,
1970; Ahn et al., 1978; Bhushan, 1987; Wu, 1991;
Bhushan et al., 1995, 1996a; Bhushan and Gupta,
1995; Gupta and Bhushan, 1995a,b; Blau et al., 1997;
Huang et al., 2002), residual stresses (Swain et al.,
1977; Marshall and Lawn, 1979; LaFontaine et al.,
1991), creep (Westbrook, 1957; Mulhearn and Tabor,
19601961; Atkins et al., 1966; Walker, 1973; Chu
and Li, 1977; Hooper and Brookes, 1984; Li et al.,
1991; Bhushan et al., 1996b), stress relaxation (Hart
and Solomon, 1973; Chu and Li, 1980; Hannula et al.,
1985; Mayo and Nix, 1988; Mayo et al., 1990;
LaFontaine et al., 1990a,b; Raman and Berriche, 1992;
Wu, 1991; Nastasi et al., 1993; Bhushan et al., 1996b),
fracture toughness and brittleness (Palmquist, 1957;
Lawn et al., 1980; Chantikul et al., 1981; Mecholsky
et al., 1992; Lawn, 1993; Bhushan et al., 1996a; Li
et al., 1997; Hainsworth et al., 1998; Li and Bhushan,
1998a), and fatigue (Li and Chu, 1979; Wu et al.,
1991; Li and Bhushan, 2002b,c).
The extended load range of static indentation testing is shown schematically in Fig. 1. It should be
noted that only the lower micro- and nanohardness
load range can be employed successfully for measurements of extremely thin (submicrometre thick) lms.
The intrinsic hardness of surface layers or thin lms
becomes meaningful only if the in uence of the substrate material can be eliminated. It is therefore
generally accepted that the depth of indentation
should never exceed 30% of the lm thickness
(ASME, 1979). Finite element analysis of layered
solid suggests that the limit of depth of indentation
can be as low as about 10% (Bhushan, 1996). The
minimum load for most commercial microindentation
Bhushan and Li
127
testers available is about 10 mN. Loads on the order

of 1 mN to 1 mN are desirable if the indentation
depths are to remain as a few tens of nanometres. In
this case, the indentation size sometimes reaches the
resolution limit of a light microscope, and it is almost
impossible to nd such a small imprint if the measurement is made with a microscope after the indentation
load has been removed. Hence, either the indentation
apparatuses are placed in a scanning electron microscope (SEM) or in situ indentation depth measurements are made. The latter measurements, in addition,
would oVer the advantages of observing the penetration process itself. In viscoelastic/viscoplastic materials, since indentation size changes with time, in situ
measurements of the indentation size are particularly
useful, and can, in addition, provide more complete
creep and relaxation data of the materials.
In this section, a commercial depth-sensing nanoindentation test apparatus will be reviewed followed
by data analysis.
force microscope, and an xyz motorised precision

table for positioning and transporting the sample
between the optical microscope and indenter (Fig. 2a).
The load on the indenter is generated using a voice
coil in permanent magnet assembly, attached to the
top of the indenter (loading) column. The generated
load is simply the vector product of the current
through the coil and the magnetic eld strength of
the permanent magnet. This type of load application
is very simple in its application as well as very linear
in its calibration, unlike many other load application
schemes found in other instruments, e.g. electrostatic
application. This type of load application allows
for very fast, closed loop, feedback control over the
displacement as it completely separates the load application system and the displacement measuring system.
Nanoindentation apparatuses
Earlier work by Alekhin et al. (1972), Ternovskii et al.
(1973), and Bulychev et al. (1975, 1979) led to the
development of depth-sensing apparatuses. In the
depth-sensing indentation apparatuses, the load
indentation depth is continuously monitored during
the loading and unloading processes (Newey et al.,
1982; Pethica et al., 1983; Wierenga and Franken,
1984; Bhushan et al., 1985, 1988; Hannula et al., 1986;
Wierenga and van der Linden, 1986; Tsukamoto et al.,
1987; Williams et al., 1988; Yanagisawa and
Motamura, 1987; Wu et al., 1988; Loubet et al., 1993;
Randall et al., 1996). For a detailed review, see
Bhushan (1999b).
Options
General description and principle of operation

Commercially available nanoindenters are made by
MTS, Oak Ridge, TN; Hysitron, Minneapolis, MN;
CSIRO, Lind eld, Australia; CSM Instruments SA,
Neucha tel, Switzerland, and Micro Materials,
Wrexham, UK. For a summary of details of these
instruments, see Table 1.
A commonly used commercial depth-sensing nanoindentation apparatus is manufactured by MTS Nano
Instruments Innovation Center, Oak Ridge, TN.
Ongoing development of this apparatus has been
described by Pethica et al. (1983), Oliver et al. (1986),
Oliver and Pethica (1989), Oliver and Pharr (1992).
This instrument is called the Nano Indenter. The
most recent model is Nano Indenter XP (MTS, 2001).
The apparatus continuously monitors the load and
the position of the indenter relative to the surface of
the specimen (depth of an indentation) during the
indentation process. The area of the indentation is
then calculated from a knowledge of the geometry of
the tip of the diamond indenter. Mechanical property
measurements can be made at a minimum penetration
depth of about 20 nm (or a plastic depth of about
15 nm) (Oliver et al., 1986). Speci cations and commonly used operating parameters for commercial
nanoindenters are given in Table 2. A description of
Nano Indenter XP follows.
The Nano Indenter XP consists of three major
components: the indenter head, an optical/atomic
Table 1
Selected details of commercial nanoindenters* according to model (and

vendor): for updated data, please see
vendors web pages
Specifications
Nano Indenter XP (MTS, Nano Instruments Innovation

Center, 1001 Larson Drive, Oak Ridge, TN 37830, USA)
Scratch measurement
Displacement resolution:
Lateral force measurement
005 nm
Contact stiffness during
Load resolution: 50 nN
loading
Min. depth: 15 nm
AFM objective
Max. depth: 500 mm
Min. load: 01 mN
Max. load: 500 mN, 10 N
TriboScope/TriboIndenter (Hysitron Inc., 5251 W. 73rd St.,
Suite A2, Minneapolis, MN 55439, USA)
Scratch measurement
Acoustic emission monitoring
005 nm
loading
Min. depth: 5 nm
AFM
Max. depth: 5 mm, 50 mm
Min. load: 01 mN
Max. load: 30 mN, 500 mN
UMIS II (CSIRO Telecommunication & Industrial Physics,
Industrial Physics, PO Box 218, Lindfield, NSW 2070,
Australia)
Scratch measurement
005 nm
loading
AFM objective
Min. depth: 20 nm
Max. depth: 2 mm, 20 mm
Min. load: 05 mN
Max. load: 10 mN, 100 mN
Nano Hardness Tester (CSM Instruments SA Jaquet-Droz 1,
PO Box 41, CH2007 Neuchatel, Switzerland)
Scratch measurement
005 nm
loading
Load resolution: 1 mN
AFM objective
Min. depth: 20 nm
Max. depth: 20 mm
Min. load: 05 mN
Max. load: 300 mN
NanoTest System (Micro Materials Ltd, Unit 3, The Byre,
Wrexham Technology Park, Wrexham LL13 7YP, UK)
Scratch measurement
Acoustic emission monitoring
005 nm
loading
Min. depth: 20 nm
Impact
Max. depth: 50 mm
AFM objective
Min. load: 05 mN
High temperature stage
Max. load: 500 mN, 20 N
Humidity control
* Where two values of maximum load (and corresponding
maximum depths) are given, lower value is for standard heads
and higher value is for high load heads.
Atomic force microscope.
Vol. 48
No. 3
128
Bhushan and Li
Two interchangeable indenter heads are available

with a load range of 0500 mN. The displacement of
the indenter is measured using a three plate capacitive
displacement sensor. All three plates are circular discs
approximately 15 mm thick. The two outer plates
have a diameter of 50 mm, and the inner, moving
plate, is half that size. The indenter column is attached
to the moving plate. Two outer plates are maintained
at equal and opposite drive voltages (2 V at 125 kHz).
The output voltage of the centre pick-up plate is
uniquely related to the position of that plate in the
capacitive gap. The pick-up voltage is measured using
a 16 bit analog to digital converter running at a
typical rate of 4 kHz. The displacement system is
calibrated using a laser interferometric calibration
system. This plate-and-indenter assembly is supported
by two leaf springs cut in such a fashion as to have
very low stiVness. The motion is damped by air ow
around the central plate of the capacitor, which is
attached to the loading column.
At the bottom of the indenter rod, a three-sided
pyramidal diamond tip (Berkovich indenter, to be
discussed below) is generally attached. The indenter
head assembly is rigidly attached to the U beam
below which the xyz table rides (Fig. 2a). The
optical microscope is also attached to the beam. As
an option to the Nano Indenter XP system, the
optical microscope may be replaced with a microscope objective that combines the convenience of an
optical objective and the power of atomic force
Table 2
Specification
and
commonly
used
operating parameters for commercial
nanoindenter
Load range
Standard heads
0500 mN
High load head
010 N
Load resolution
Standard heads
50 nN
High load head
50 nN
Vertical displacement range
0500 mm
Vertical displacement
resolution
005 nm
Typical approach rate
10 nm s 1
Typical indentation load rate
10% of peak load/s
Typical indentation
displacement rate
10% peak diplacement/s
Optical microscope
magnification
Up to 1500
AFM objective magnification
Up to 106
Spatial resolution of xyz
400 nm in x and y
table
directions
Area examined in single
series of indentations
150150 mm
Minimum penetration depth
~20 nm
Continuous stiffness option
Frequency range
10150 Hz
Time constant
033 s
Smallest measurable
distance
01 nm
Scratch and tangential force option
Scratch velocity
Max. 100 mm s 1 with
20 points/mm
Tangential displacement
range
2 mm
Tangential displacement
resolution
100 nm
Tangential load resolution
10 mN
Minimum measurable
tangential load
10 mN
Vol. 48
No. 3
microscopy. The position of an indentation on a

specimen is selected using either the optical microscope (maximum magni cation of 1500) or the
atomic force microscope. The specimens are held on
an xyz table the position of which, relative to the
microscope or the indenter, is controlled with a
joystick. The spatial resolution of the position of the
table in the xy plane is 400 nm and its position is
observed on the computer screen. The three components just described are enclosed in a heavy wooden
cabinet to ensure the thermal stability of the samples.
The entire apparatus is placed on a vibration isolation table.
The nanoindenter also comes with a continuous
stiVness measurement device (Oliver and Pethica,
1989; Pethica and Oliver, 1989). This device makes
possible the continuous measurement of the stiVness
of a sample, which allows the elastic modulus to be
calculated as a continuous function of time (or indentation depth). Useful data can be obtained from
indents with depths as small as 15 nm. Because of the
relatively small time constant of the measurements,
the device is particularly useful in studies of time
dependent properties of materials.
Indenters
The main requirements for the indenter are high
elastic modulus, no plastic deformation, low friction,
smooth surface, and a well de ned geometry that
is capable of making a well de ned indentation
impression. The rst four requirements are satis ed
by choosing diamond material for the tip. A well
de ned perfect tip shape is diYcult to achieve. The
Berkovich indenter is a three-sided pyramid and
provides a sharply pointed tip compared with the
Vickers or Knoop indenters, which are four-sided
a major components: indenter head, xyz motorised

precision table, and tangential force option; b detailed
schematic of tangential force option hardware (not to scale
and front and rear prongs not shown)
Schematics of Nano Indenter XP
Bhushan and Li
pyramids and have a slight oVset (051 mm) (Tabor,

1970; Bhushan, 1999b). Because any three non-parallel planes intersect at a single point, it is relatively
easy to grind a sharp tip on an indenter if Berkovich
geometry is used. However, an indenter with a sharp
tip suVers from a nite, but exceptionally diYcult to
measure, tip bluntness. In addition, pointed indenters
produce a virtually constant plastic strain impression
and there is the additional problem of assessing the
elastic modulus from the continuously varying
unloading slope. Spherical indentation overcomes
many of the problems associated with pointed
indenters. With a spherical indenter, one is able to
follow the transition from elastic to plastic behaviour
and thereby de ne the yield stress (Bell et al., 1992).
However, a sharper tip is desirable, especially for
extremely thin lms requiring shallow indentation.
Therefore, the Berkovich indenter is most commonly
used for measurements of nanomechanical properties.
Experimental procedures have been developed to
correct for the tip shape, to be described below.
The Berkovich indenter, directly brazed to a 304
stainless steel holder, is a three-sided (triangular
based) pyramidal diamond, with a nominal angle of
653 between the (side) face and the normal to the
base at apex, an angle of 769 between edge and
normal, and with a tip radius of less than 01 mm
(Fig. 3) (Berkovich, 1951). The typical indenter is
shaped to be used for indentation (penetration) depths
of 10 to 20 mm. The indentations appear as equilateral
triangles (Fig. 3b) and the height of triangular indent
l is related to the depth h as
h/l=(1/2) cot 769=1/859
. (1a)
The relationship h(l) is dependent on the shape of

the indenter. The height of the triangular indent l is
related to the length of one side of the triangle a as
l=0866a
. (1b)
. (1c)
and
h/a =1/7407
The projected contact area A for the assumed

geometry is given as
A =0433a2=2376h2
. (2)
Another three-sided pyramidal indenter, the cube

corner indenter, can displace more than three times
the volume of the Berkovich indenter at the same
load, thereby producing much higher stresses and
strains in the vicinity of the contact and reducing the
cracking threshold. This makes the cube corner
indenter ideal for the estimation of fracture toughness
at relatively small scales (Li et al., 1997). The spherical
indenter initiates elastic contact and then causes
elasticplastic contact at higher loads. This indenter
is well suited for the examination of yielding and
work hardening. However, it is very diYcult to obtain
a precise sphere with a diameter of less than 100 mm
made of diamond. This fact limits its application in
nanoindentation testing.
The exact shape of the indenter tip needs to be
measured for determination of hardness and Youngs
modulus of elasticity. Since the indenter is quite blunt,
direct imaging of indentations of small size in the
129
a schematic; b indent impression; c photograph of

indenter shape
Berkovich indenter
SEM is diYcult. Determination of tip area function

will be discussed below.
Indentation procedure
The indentation procedure in this section is based on
directions produced by the MTS Nano Instruments
Innovation Center (2001). An indentation test
involves moving the indenter to the surface of the
material and measuring the forces and displacements
associated during indentation. The surface is located
for each indentation by lowering the indenter at a
constant loading rate against the suspending springs
and detecting a change in velocity on contact with
the surface. In the testing mode, the load is
incremented in order to maintain a constant loading
rate or constant displacement rate. The load and
indentation depths are measured during indentation
both in the loading and unloading cycles. The force
contribution of the suspending springs and the displacements associated with the measured compliance
of the instrument are removed.
In practical indentation tests, a Berkovich tip with
a radius of about 50 nm is used. Multiple loading
and unloading steps are performed to examine the
reversibility of the deformation, ensuring that the
unloading data used for analysis purposes are mostly
elastic. A typical indentation experiment consists of a
combination of several segments, e.g. approach, load,
hold, and unload. Either constant loading or constant
displacement experiments can be performed (Oliver
and Pharr, 1992; Bhushan, 1999b). A typical constant
loading indentation experiment consists of eight steps:
approaching the surface at 10 nm s 1; loading to
peak load at a constant loading rate (10% of peak
load/s); unloading 90% of peak load at a constant
unloading rate (10% of peak load/s); reloading to
peak load; holding the indenter at peak load for 10 s;
Vol. 48
No. 3
130
Bhushan and Li
Representative loaddisplacement curve during

loading/unloading
sequence
in
constant
loading experiment
unloading 90% of peak load; holding the indenter

after 90% unloading; nally unloading completely.
The rst hold step is included to incorporate the
corrections due to thermal drift. Figure 4 shows a
representative loaddisplacement curve of an indentation made at 15 mN peak indentation load for
silicon (100).
Acoustic emission measurements during
indentation
Acoustic emission (AE) measurement is a very sensitive technique to monitor cracking of surfaces and
subsurfaces. The nucleation and growth of cracks
result in a sudden release of energy within a solid,
then some of the energy is dissipated in the form of
elastic waves. These waves are generated by sudden
changes in stress and in displacement that accompany
the deformation. If the release of energy is suYciently
large and rapid, then elastic waves in the ultrasonic
frequency regime (AE) will be generated and these
can be detected using PZTs (lead zirconate titanate)
via expansion and compression of the PZT crystals
(Yeack-Scranton, 1986; Scruby, 1987; Bhushan, 1996).
Weihs et al. (1992) used an AE sensor to detect
cracking during indentation tests using a nanoindenter. The energy dissipated during crack growth can
be estimated by the rise time of the AE signal. They
mounted a commercial transducer with tungsten
impregnated epoxy backing for damping underneath
the sample. The transducer converts the AE signal
into a voltage that is ampli ed by oscilloscopes and
used for continuous display of the AE signal. Any
correlation between the AE signal and the load
displacement curves can be observed. (Also see Wu,
1991.)
Nanoscratch and lateral force measurements
Several micro- and nanoscratch testers are commercially available, such as the Taber shear/scratch tester
model 502 with a no. 13958 diamond cutting tool
(manufactured by Teledyne Taber, North Tonawanda,
NY) for thick lms; Revetest automatic scratch tester
(manufactured by CSM Instruments SA, Neucha tel,
Switzerland) for thin lms (Perry, 1981, 1983; Sekler
et al., 1988), and Nano Indenter for ultrathin lms
(Wu, 1991; MTS, 2001; Bhushan et al., 1995, 1996a;
Vol. 48
No. 3
Bhushan and Gupta, 1995; Gupta and Bhushan,

1995a,b; Huang et al., 2002).
Described here is the nanoscratch and lateral force
option which allows the making of scratches of various lengths at programmable loads. Lateral (friction)
forces can also be measured simultaneously (MTS,
2001). The additional hardware for the lateral force
option includes a set of proximity (capacitance) probes
for measurement of lateral displacement or force in
the two lateral directions along x and y, and a special
scratch collar which mounts around the indenter
shaft with hardness indenter (Fig. 2b). A scratch block
is mounted on the end of the indenter shaft, in line
with the proximity probes and the positioning screws.
The scratch tip is attached to the scratch block with
two Allen head screws. The scratch tip can be a
Berkovich indenter or a conventional conical diamond tip with a tip radius of about 15 mm and an
included angle of 6090 (typically 1 mm of tip radius
with 60 of included angle; Bhushan et al., 1995). A
larger included angle of 90 may be desirable for a
more durable tip. The tip radius should not be very
small as it will readily become blunt.
During scratching, a load is applied up to a speci ed indentation load or up to a speci ed indentation
depth, and the lateral motion of the sample is measured. In addition, of course, load and indentation
depth are monitored. Scratches can be made either
at the constant load or at ramp-up load. Measurement
of lateral force allows the calculation of the coeYcient
of friction during scratching. The resolution of the
proximity probe provides resolution of lateral force
of about 2 mN; therefore, a minimum load of about
20 mN can be measured. Consequently, a minimum
normal load of about 02 mN should be used for a
sample with a coeYcient of friction of about 01.
Microscopy of the scratch produced at ramp-up load
allows the measurement of critical load required
to break up the lm (if any) and scratch width,
and general observations of scratch morphology.
Typically, draw acceleration (mm s 2), and draw
velocity (mm s 1) are 10 mm s 2 and 5 mm s 1,
respectively.
Analysis of indentation data
An indentation curve is the relationship between load
W and displacement (or indentation depth or penetration depth) h, which is continuously monitored and
recorded during indentation. Stressstrain curves,
typical indentation curves, deformed surfaces after tip
removal, and residual impressions of indentation for
ideal elastic, rigidperfectly plastic, and elasticperfectly plastic and real elasticplastic solids are shown
in Fig. 5. For an elastic solid, the sample deforms
elastically according to Youngs modulus, and the
deformation is recovered during unloading. As a
result, there is no impression of the indentation after
unloading. For a rigidperfectly plastic solid, no
deformation occurs until yield stress is reached, when
plastic ow takes place. There is no recovery during
unloading and the impression remains unchanged. In
the case of an elasticplastic solid, it deforms elastically according to Youngs modulus and then it
deforms plastically. The elastic deformation is re-
Bhushan and Li
Schematics of a stressstrain curves, b typical

indentation curves, c deformed surfaces after
tip removal, and d residual impressions of
indentation, for ideal elastic, rigidperfectly
plastic, elasticperfectly plastic (ideal) and real
elasticplastic solids
covered during unloading. In the case of an elastic

perfectly plastic solid, there is no work hardening.
All engineering surfaces follow real elasticplastic
deformation behaviour with work hardening
(Johnson, 1985; Bhushan, 1999a, 2002). The deformation pattern of a real elasticplastic sample during
and after indentation is shown schematically in Fig. 6a
(Oliver and Pharr, 1992; Hainsworth et al., 1996;
Oliver, 2001). In this gure, the contact depth h
c
represents the depth of indenter in contact with the
sample under load. The depth measured during the
indentation h includes the depression of the sample
around the indentation in addition to the contact
depth. The depression of the sample around the
indentation h =h h is caused by elastic displaces
c
ments and must be subtracted from the data to obtain
the actual depth of indentation or actual hardness.
At peak load, the load and displacement are W
max
and h
respectively, and the radius of the contact
max
circle is a. Upon unloading, the elastic displacements
in the contact region are recovered and when the
indenter is fully withdrawn, the nal depth of the
residual hardness impression is h . A schematic of a
f
loaddisplacement curve is shown in Fig. 6b.
Based on the work of Sneddon (1965) to predict
the de ection of the surface at the contact perimeter
for a conical indenter and a paraboloid of revolution,
Oliver and Pharr (1992) developed an expression for
h at maximum load (required for hardness calcuc
lation) from h
max
131
a schematic representation of indentation process

illustrating depression of sample around indentation and
decrease in indentation depth upon unloading (Oliver and
Pharr, 1992); b schematic of loaddisplacement curve
Indentation data: example for real elastic

plastic sample
h =h eW /S
. . . . . . . . (3a)
c
max
max max
where the geometrical factor e =072 for the conical
indenter, e =075 for the paraboloid of revolution,
and e =1 for the at punch; S
is the stiVness
maxof the unloading
(=1/compliance) equal to the slope
curve (dW /dh) at maximum load. (Hay et al. (1999)
have proposed corrections to Sneddons equations.)
Oliver and Pharr (1992) assumed that the behaviour
of the Berkovich indenter is similar to that of the
conical indenter, since the cross-sectional area of both
types of indenter varies as the square of the contact
depth and their geometries are singular at the tip.
Therefore, for the Berkovich indenter, e # 072. Thus,
h is slightly larger than the plastic indentation depth
hc , which is given by
p
h =h W /S
. . . . . . . . (3b)
p
max
max max
It should be mentioned here that Doerner and Nix
(1986) underestimated h by assuming that h =h .
c
p
Based on nite element canalysis of the indentation
process, Laursen and Simo (1992) showed that h
c
cannot be assumed equal to h for indenters which
p
do not have at punch geometry.
For a Vickers indenter with ideal pyramidal
geometry (ideally sharp tip), the projected contact
area/depth relationship is given as (Doerner and Nix,
1986; Bhushan, 1996; Swadener et al., 2002)
A=245h2 . . . . . . . . . . . . (4a)
c
Since the area/depth relationship is equivalent for
both typical Berkovich and Vickers pyramids, equaInternational Materials Reviews 2003
Vol. 48
No. 3
132
Bhushan and Li
relationship (assuming no elastic recovery) between

the projected contact area of the indenter and the
actual contact depth. Figure 7b shows the measured
contact area versus indentation depth data of Pethica
et al. (1983) for nickel and of Doerner and Nix (1986)
for annealed a-brass. From this gure, it seems that
a tip radius of 1 mm ts the data best. If there is
elastic recovery, the experimental data are smaller
than they should be, and then the tip radius would
be even larger than 1 mm. Shih et al. (1991) used the
nite element method to simulate an indentation test.
They showed that loadindentation depth data
obtained for nickel by Pethica et al. (1983) using a
nanoindenter can be tted with a simulated pro le
for a tip radius of about 1 mm (Fig. 7c).
As shown in Fig. 7, the actual indentation depth h
c
produces a larger contact area than would be expected
for an indenter with an ideal shape. For the real
indenter used in the measurements, the nominal shape
is characterised by an area function F(h ), which
relates projected contact area of indenter toc contact
depth (equation (4a))
A1/2=F(h ) . . . . . . . . . . . . (4b)
c
The functional form must be established experimentally prior to the analysis (to be described below).
Hardness
Berkovich hardness HB (or H ) is de ned as the load
B
divided by the projected contact
area of the indentation. It is the mean pressure that a material will
support under load. From the indentation curve,
hardness at maximum load can be obtained as
/A
. . . . . . . . . . . (5)
max
where W
is the maximum indentation load and A
max
is the projected
contact area at peak load. The contact
area at peak load is determined by the geometry of
the indenter and the corresponding contact depth h
using equations (3a) and (4b). A plot of hardness asc
a function of indentation depth for polished single
crystal silicon (111), with and without tip shape
calibration, is shown in Fig. 8. It should be noted
that, for this example, tip shape calibration is necessary and the hardness is independent of corrected
depth.
It should be pointed out that hardness measured
using this de nition may be diVerent from that
obtained from the more conventional de nition in
which the area is determined by direct measurement
of the size of the residual hardness impression. The
reason for the diVerence is that, in some materials, a
small portion of the contact area under load is not
plastically deformed, and, as a result, the contact area
measured by observation of the residual hardness
impression may be less than that at peak load.
However, for most materials, measurements using the
two techniques give similar results.
Stresses during the indentation process can induce
phase transformation in some ceramic materials, such
as yttria stabilised ZrO . Stress induced phase trans2
formation can cause a change in volume of indented
material that will in turn aVect the loaddisplacement
data. A relationship between indentation pressure and
phase transformation and resulting change in volume
HB=W
a schematic of indenter tip with non-ideal shape: contact

depth and effective depth are also shown; b predicted
projected contact area as function of indentation depth
curves for various tip radii and measured data; c predicted
load as function of indentation depth curves for various tip
radii and measured data (Shih et al., 1991)
Effect of tip radius
tion (4a) holds for the Berkovich indenter as well.

Although a slightly diVerent expression for A(h) has
been derived (see equation (2)) for the assumed
Berkovich indenter geometry, in this section equation
(4a) is used for A(h), as this relationship is most
commonly employed in the analysis of indentation
hardness data.
The indenter tip is generally rounded so that ideal
geometry is not maintained near the tip (Fig. 7a). To
study the eVect of tip radius on the elasticplastic
deformation (load versus displacement curve), Shih
et al. (1991) modelled the blunt tip geometry by a
spherical tip of various radii. They derived a geometric
Vol. 48
No. 3
Bhushan and Li
9
8
Hardness as function of indentation depth for

polished single crystal silicon (111) calculated
from area function with and without tip shape
calibration (Doerner and Nix, 1986)
of indented material during the indentation would

need to be studied.
Modulus of elasticity
Even though, during loading, a sample undergoes
elasticplastic deformation, the initial unloading is an
elastic event. Therefore, the Youngs modulus of elasticity or, simply, the elastic modulus of the specimen
can be inferred from the initial slope of the unloading
curve dW /dh called stiVness (1/compliance) (at maximum load ) (Fig. 6b). It should be noted that the
contact stiVness is measured only at the maximum
load, and no restrictions are placed on the unloading
data being linear during any portion of the unloading.
If the area in contact remains constant during
initial unloading, an approximate elastic solution is
obtained by analysing a at punch for which the area
in contact with the specimen is equal to the projected
area of the actual punch. Based on the analysis by
Sneddon (1965) of indentation of an elastic half space
by a at cylindrical punch, Loubet et al. (1984)
calculated the elastic deformation of an isotropic
elastic material with a at-ended cylindrical punch.
They obtained an approximate relationship for compliance dh/dW for the Vickers (square) indenter. King
(1987) solved the problem of at-ended cylindrical,
quadrilateral (Vickers and Knoop), and triangular
(Berkovich) punches indenting an elastic half space.
He found that the compliance for the indenter is
approximately independent of the shape (with a variation of at most 3%) if the projected area is xed.
Pharr et al. (1992) also veri ed that the compliance
of a paraboloid of revolution of a smooth function is
the same as that of a spherical or a at-ended
cylindrical punch. The relationship for the compliance
C (inverse of stiVness S) for an indenter (Vickers,
Knoop, and Berkovich) is given as
AB
1
dh
1 p 1/2
C= =
#
S dW
2E A
r
where
1
1 n2 1 n2
s+
i
=
E
E
E
r
s
i
. (6)
133
Compliance as function of inverse of indentation

depth for tungsten with and without tip shape
calibration (constant modulus with 1/depth
would be as indicated by straight line): slope of
corrected curve is 480 GPa, which compares
reasonably well with known modulus of
tungsten (420 GPa), and small y intercept of
about 03 nm/mN is attributed to load frame
compliance, not removed (Doerner and Nix,
1986)
and dW /dh is the slope of the unloading curve at

maximum load (Fig. 6b), E is the reduced modulus,
r
E and E are the elastic moduli of the specimen and
s
i
the indenter, n and n are the Poisson ratios of the
s
i
specimen and indenter, C (or S) is the experimentally
measured compliance (or stiVness) at maximum load
during unloading, and A is the projected contact area
at maximum load.
The contact depth h is related to the projected
c
area of the indentation A for a real indenter by
equation (4b). A plot of the measured compliance
dh/dW versus the reciprocal of the corrected indentation depth obtained from various indentation curves
(one data point at maximum load for each curve)
should yield a straight line with slope proportional
to 1/E (Fig. 9) (Doerner and Nix, 1986). The elastic
r
modulus of the specimen E can then be calculated,
s
provided Poissons ratio is known with great precision, to obtain a good value of the modulus. For a
diamond indenter, E =1140 GPa and n =007 are
i
i
taken. In addition, the y intercept of the compliance
versus the reciprocal indentation depth plot should
give any additional compliance that is independent
of the contact area. The compliance of the loading
column is generally removed from the loaddisplacement curve; measurement techniques for the former
will be described below.
The most frequently used method to measure initial
unloading stiVness S is now discussed. Doerner and
Nix (1986) measured S by tting a straight line to
the upper about one-third portion of the unloading
curve. The problem with this is that for non-linear
loading data, the measured stiVness depends on which
of the data are used in the t. Oliver and Pharr
(1992) proposed a new procedure. They found that
all the unloading data are well described by a simple
power law relation
Vol. 48
No. 3
134
Bhushan and Li
W =B(h
h )m
. . . . . . . . . . (7)
f
where the constants B and m are determined by a
least squares t. The initial unloading slope is then
found analytically, diVerentiating this expression and
evaluating the derivative at the maximum load and
maximum depth. As pointed out above, unloading
data used for the calculations should be obtained
after several loading/unloading cycles and with peak
hold periods.
This analysis is based on an elastic solution which
only accounts for sink-in (the indented material
around the indenter below its original surface).
However, in the more realistic case of elasticplastic
contact, sink-in or pile-up (the indented material
around the indenter above its original surface) can
occur depending on the speci c mechanical properties
of the material. For pile-up situations, the just
described OliverPharr method would underestimate
the true contact area by as much as 50%. This in
turn leads to overestimation of the hardness and
elastic modulus. Based on some modelling, pile-up is
signi cant only when h /h >07 and the material
f max
does not appreciably work harden. Note that
h /h =0 corresponds to fully elastic deformation
f max
and
h /h =1 corresponds to rigidplastic behavf max
iour. Compressive residual stresses result in pile-up
whereas tensile stresses result in sink-in. Although
some correction procedures have been proposed
(Pharr, 1998; Tsui and Pharr, 1999), the real contact
area measurement requires imaging of indentation
impressions.
Determination of load frame compliance and
indenter area function
As stated above, measured displacements are the sum
of the indentation depths in the specimen and the
displacements of suspending springs and the displacements associated with the measuring instruments,
referred to as load frame compliance. Therefore, to
determine accurately the specimen depth, load frame
compliance must be known. This is especially important for large indentations made with high modulus
for which the load frame displacement can be a
signi cant fraction of the total displacement. The
exact shape of the diamond indenter tip needs to be
measured because the hardness and Youngs modulus
of elasticity depend on the contact areas derived from
measured depths. The tip becomes blunt (Fig. 7a) and
its shape signi cantly aVects the prediction of mechanical properties (Figs. 89).
Oliver and Pharr (1992) proposed a method for
determining area functions. Their method is based
only on one assumption that Youngs modulus is
independent of indentation depth. They also proposed
a method to determine load frame compliance. Here,
the methods for determining load frame compliance
are described rst, followed by the method for area
function. Oliver and Pharr modelled the load frame
and the specimen as two springs in series; thus
C =C +C . . . . . . . . . . . . (8)
s
f
where C, C , and C are the total measured coms
f
pliance, specimen compliance, and load frame compliance respectively. From equations (6) and (8)
Vol. 48
No. 3
10
Plot of C Cf as function of A 1/2 for

aluminium: error bars are two standard
deviations in length (Oliver and Pharr, 1992)
AB
1 p 1/2
C=C +
. . . . . . . . . (9)
f 2E A
r
From equation (9), it is noted that if the modulus of
elasticity is constant, a plot of C as a function of
A 1/2 is linear and the vertical intercept gives C . It
is obvious that most accurate values of C f are
f
obtained when the specimen compliance is small, i.e.
for large indentations.
To determine the area function and the load frame
compliance, Oliver and Pharr made relatively large
indentations in aluminium because of its low hardness. In addition, for the larger aluminium indentations (typically 700 to 4000 nm deep), the area
function for a perfect Berkovich indenter (equation
(4a)) can be used to provide a rst estimate of the
contact. Values of C and E are thus obtained by
f of A r
plotting C as a function
1/2 for the large indentations (Fig. 10).
Using the measured C value, they calculated conf made at shallow depths
tact areas for indentations
on the aluminium with measured E , and/or on a
r
harder fused silica surface with published values of
E , by rewriting equation (9) as
r
p 1
1
A=
. . . . . . . . . (10)
4 E2 (C C )2
r
f
from which an initial guess at the area function was
made by tting A as a function h data to an eighth
c
order polynomial
A=245h2 +C h +C h1/2
c
1 c
2 c
+C h1/4 + +C h1/128 . . . . . . (11)
3 c
8 c
where C C are constants. The rst term describes
1 shape
8
the perfect
of the indenter, the others describe
deviations from the Berkovich geometry due to blunting of the tip. A convenient tting routine is that
contained in the Kaleidagraph software. A weighted
procedure can be used to ensure that data points
with small and large magnitudes are of equal importance. An iterative approach can be used to re ne the
values of C and E further.
f
r procedure, recommended by
The step-by-step
Oliver and Pharr (1992), for determination of load
frame compliance and indenter area function, is now
Bhushan and Li
described in detail. It involves making a series of

indentations in two standard materials aluminium
and fused quartz and relies on the facts that both
these materials are elastically isotropic, their moduli
are well known, and their moduli are independent of
indentation depth. The rst step is to determine
precisely the load frame compliance. This is best
accomplished by indenting a well annealed, high
purity aluminium which is chosen because it is readily
available, has a low hardness, and is nearly elastically
isotropic. Some care must be exercised in preparing
the aluminium to ensure that its surface is smooth
and unaVected by work hardening. Typically, indentation depths range from 700 to 4000 nm. The load
time history recommended by Oliver and Pharr is as
follows: (1) approach and contact surface, (2) load to
peak load, (3) unload to 90% of peak load and hold
for 100 s, (4) reload to peak load and hold for 10 s,
and (5) unload completely. The lower hold is used to
establish thermal drift and the upper hold to minimise
time dependent plastic eVects. The nal unloading
data are used to determine the unloading compliances
using the power law tting procedure described above.
The load frame compliance is determined from the
aluminium data by plotting the measured compliance
as a function of area calculated, assuming the ideal
Berkovich indenter. Calculated E is checked with
r
known elastic constants for aluminium,
E=704 GPa
and n=0347. The values of the elastic constants used
in this review for the diamond indenter are E =
i
1141 GPa and n =007.
i
The problem with using aluminium to extend the
area function to small depths is that, because of its
low hardness, small indentations in aluminium require
very small loads, and a limit is set by the force
resolution of the indentation system. This problem
can be avoided by making the small indentations in
fused quartz, a much harder, isotropic material available in optically nished plate form. The standard
procedure that Oliver and Pharr recommend for
determining the area function involves making a series
of indentations in fused quartz using a set of six peak
loads. (Typically, measurements are made at depths
ranging from about 15 to 700 nm. For depths
exceeding 700 nm, the indenter can be assumed to
have a perfect shape.) The contact areas and contact
depths are then determined using equation (11) and
h in conjunction with the reduced modulus computed
c
from
the elastic constants for fused quartz, E =
72 GPa and n=0170. The machine compliance is
known from the aluminium analysis. The area function is only good for the depth range used in the
calculations. Typical data of contact areas as a function of contact depths for six materials are shown
in Fig. 11.
Measurement of hardness and elastic modulus at
an indentation depth of less than 20 nm using a
conventional indenter remains a technical challenge.
To solve this problem, an ideally sharp indenter tip
is needed to perform indentations at ultralow loads.
A sharp indenter with a tip radius of less than 20 nm
can be fabricated by using the focus ion beam technique, but it may be very costly. Attempts using
atomic force microscopy have been successful to make
measurements with indentation depth as shallow as
11
135
Computed contact areas as function of contact

depths for six materials: error bars are
two standard deviations in length (Oliver and
Pharr, 1992)
5 nm or less (Bhushan et al., 1996b). In situ transmission electron microscope indentation tests are
expected to give an in-depth understanding of elastic/
plastic deformation and nucleation and motion of
dislocations. Atomistic dynamic simulation may help
us in understanding the very early stage of contact
between the tip and the sample.
Hardness/modulus 2 parameter
Calculations of hardness and modulus described so
far require calculations of the projected contact area
of the indentation from the indentation depth, which
are based on the assumption that the test surface be
smooth to dimensions much smaller than the projected area. Therefore, data obtained from rough
samples show considerable scatter. Joslin and Oliver
(1990) developed an alternative method for data
analysis without requiring calculations of the projected contact area of the indentation. This method
introduces measurement of a parameter hardness/
modulus2, which provides a measure of the resistance
of the material to plastic penetration.
Joslin and Oliver showed that for several types of
rigid punch (cone, at punch, parabola of revolution,
and sphere) as long as there is a single contact
between the indenter and the specimen
H/E2 =(4/p)(W /S2) . . . . . . . . . (12)
r
where S is the stiVness obtained from the unloading
curve. Modulus E is related to E by a factor of
r with moduli ssigni cantly less
(1 n2 ) for materials
s
than diamond (equation (6)). The H/E2 parameter
s
represents the resistance of a material to plastic
penetration. It can clearly be seen that calculation of
projected contact area and knowledge of area function
are not required. However, this method does not give
the hardness and modulus values separately.
Continuous stiffness measurements (CSM)
Oliver and Pethica (1989) and Pethica and Oliver
(1989) developed a dynamic technique for continuous
measurement of sample stiVness during indentation
Vol. 48
No. 3
136
Bhushan and Li
without the need for discrete unloading cycles, and

with a time constant that is at least three orders of
magnitude smaller than the time constant of the more
conventional method of determining stiVness from
the slope of an unloading curve. Furthermore, the
measurements can be made at exceedingly small
penetration depths. Thus, their method is ideal for
determining the stiVness and, hence, the elastic modulus and hardness of lms a few tens of nanometres
thick. Furthermore, its small time constant makes it
especially useful in measuring the properties of some
polymeric materials.
Measurement
of
continuous
stiVness
is
accomplished by the superposition of a very small ac
current of a known relatively high frequency (typically
693 Hz) on the loading coil of the indenter. This
current, which is much smaller than the dc current
that determines the nominal load on the indenter,
causes the indenter to vibrate with a frequency related
to the stiVness of the sample and to the indenter
contact area. A comparison of the phase and amplitude of the indenter vibrations (determined with a
lock-in ampli er) with the phase and amplitude of
the imposed ac signal allows the stiVness to be
calculated either in terms of amplitude or phase.
Displacements as small as 0001 nm can be measured
using frequency-speci c ampli cation. The time constant of about 033 s provides a good combination of
low noise and dynamic response.
To calculate the stiVness of the contact zone, the
dynamic response of the indentation system has to
be determined. The relevant components are the mass
m of the indenter, the spring constant K of the leaf
springs that support the indenter ( load fframe stiVness), the stiVness of the indenter frame K , and the
s
damping constant D due to the air in the gaps
of the
i
capacitor plate displacement-sensing system. These
combine with the sample stiVness S, and damping
constant of the contact zone D , as shown sches overall response.
matically in Fig. 12, to produce the
If the imposed driving force is F(t)=F exp(ivt) and
0
the displacement response of the indenter
is z(t)=
z exp[i(vt w)], the ratio of amplitudes of the
0
imposed force and the displacement response is given
by (Pethica and Oliver, 1989)
K K C
F
1/2
0 = (K mv2)2+v2D2
. . . . (13)
z
0
and the phase angle w between the driving force and
the response is
tan w=vD/(K
mv2)
. (14)
Equations (13) and (14) may be solved simultaneously

for K and D. The stiVness and damping of the contact
are given by
S=
1
F z 1 cos w
0 0
(K
s
mv2)
1
K
(15a)
and
F
D v= 0 sin w D v . . . . . . .
(15b)
s
i
z
0
With the exception of S and D terms, all the terms
s
in equation (15) can be measured independently. The
Vol. 48
No. 3
12
a schematic of nanoindenter with each

component represented in simple harmonic
oscillator models and b dynamic model of
indentation system
parameters m, K , and D are determined by analysing

s
i
the dynamic response of the system when the indenter
is hanging free, which is done in the factory. Details
on K have been presented above. In a CSM experiment,f the excitation frequency v is set. The ac input
to the force coil is generated with a standard ac signal
generator, and any frequency between about 10 and
150 Hz may be selected. The displacement amplitude
z and phase angle w (using a lock-in ampli er) are
0
measured.
Using equation (15), S and D v are
s
calculated.
Modulus of elasticity by cantilever deflection
measurement
The submicrometre indentation of thin lms on substrates can lead to quick and accurate measurement
of hardness, but the large pressures under the indenter
may alter the structure of thin lms being tested. For
submicrometre (<200 nm) lms, the in uence of the
substrate must be considered. With wafer curvature
techniques, the average stress and strain in a lm can
be measured, but the range of stresses is limited by
the thermal expansion and/or growth mismatch of
the substrate and the lm. Thus, for a given lm and
substrate, one cannot dictate the stress to be applied
to the lm. In an eVort to avoid some of these
diYculties, Weihs et al. (1988) developed a new technique based on the de ection of free-standing cantilever microbeams. Brie y, the technique involves the
fabrication of the beams using silicon micromachining
techniques and the de ection of the beams using a
nanoindenter. The nanoindenter mechanically bends
the beams while continuously monitoring the loading
and the de ection of the beams. Using this technique,
both elastic modulus and yield strength can be
studied.
Bhushan and Li
137
deformation can be recognised in the plot of load

versus de ection by a deviation from linearity. The
load that marks this deviation is de ned as the yield
load W , and the yield strength is given by
y
s =6W l/(bt2)
. . . . . . . . . . (17)
y
y
Following yielding, the beam continues to bend as
more load is applied, and some of the strain in the
beam is plastic and thus unrecoverable during
unloading. The shape of the load versus de ection
curves after yielding and prior to unloading depends
on the elastic and plastic properties of the material.
13
Schematic of cantilever microbeam deflected

by nanoindenter loading mechanism for
measurement of modulus of elasticity and
yield strength of microbeam material (Weihs
et al., 1988)
Weihs et al. measured properties of free-standing

beams of SiO , LTO ( lead titanium oxide), and gold
coatings used2 in VLSI (very large scale integration)
fabrication cut into small (1515 mm) sections
(Fig. 13). The microbeams were de ected at velocities
ranging from 3 to 60 nm s 1. Typical dimensions of
the beam free to de ect were 40 mm length, 20 mm
width, and 1 mm thickness. The modulus of elasticity
is calculated from the following equation valid for
small de ection of a thin lm
h =4W l3(1
n2)/(bt3E ) .
. (16)
where h is the vertical de ection, W is the force

applied, l is the eVective length, n is Poissons ratio,
b is the width, t is the thickness, and E is the Youngs
modulus of the beam. As mentioned above, the nanoindenter records the load required to displace its own
spring and the microbeam. Consequently, the load
required by the springs of the nanoindenter must be
subtracted from the total applied load to determine
the load on the beam. For good accuracy, the spring
constant of the microbeam must be approximately
equal to or greater than 10% of the spring constant
of the nanoindenter.
The application of simple beam theory to the load
de ection data of beams also enables one to determine
the yield strength of the microbeam material. For a
homogeneous cantilever beam under load, the strain
in the beam varies linearly through the thickness such
that the maximum strain at a given length occurs at
the top and at the bottom of the beam. For the
loading shown in Fig. 13, the top of the beam is in
tension and the bottom of the beam is in compression.
In addition, the maximum stress in the beam is
located at the xed end of the beam (B) where the
applied moment is greatest. When this maximum
stress reaches the yield strength of the material, the
beam begins to deform plastically. The onset of such
Determination of hardness and modulus of

elasticity of thin films from composite response of
film and substrate
As mentioned above, for a thin lm on a substrate, if
the indentation exceeds about 30% of the lm thickness, measured hardness is aVected by the substrate
properties. A number of researchers have attempted
to derive expressions that relate thin lm hardness to
substrate hardness, composite hardness (measured on
the coated substrate) and lm thickness, and so allow
the calculation of these quantities given the remaining
three (Buckle, 1973; Jonsson and Hogmark, 1984;
Sargent, 1986; Burnett and Rickerby, 1987b;
Bhattacharya and Nix, 1988a,b; Bull and Rickerby,
1990; Vinci and Vlassak, 1996; Korsunsky et al., 1998;
McGurk and Page, 1999; Tsui and Pharr, 1999).
Here, two models are discussed based on the volume
law of mixtures (volume fraction model ) and nite
element simulation.
Sargent (1986) suggested that the hardness of a
lmsubstrate composite is determined by a weighted
average of the volume of plastically deformed material
in the lm V and that in the substrate V
f
s
V
V
H =H f +H s
. . . . . . . . . (18)
fV
sV
where V =V +V . The deformed volumes of lm
and substratef can sbe calculated using the expanding
spherical cavity model (Johnson, 1985). Burnett and
Rickerby (1987b) found it necessary to incorporate a
further weighting factor to deforming volume to
obtain a reasonable t to experimental data. This
factor accounts for the diVerences in relative sizes of
the plastic zones in the lm and substrate. Equation
(18) is modi ed for a soft lm on a hard substrate
V
V
H =H X3 f +H s
f
s
V
V
. (19a)
. (19b)
and for a hard lm on a soft substrate

H =H
V
V
f +H X3 s
fV
s
V
where X is the ratio of plastic zone volumes given as
A B
EH n
f s
. . . . . . . . . . (19c)
HE
f s
They found that n, determined empirically, ranged
from 1/2 to 1/3.
Bhattacharya and Nix (1988a) modelled the indentation process using the nite element method to
study the elasticplastic response of materials.
Bhattacharya and Nix (1988b) calculated elastic and
X=
Vol. 48
No. 3
138
Bhushan and Li
plastic deformation associated with submicrometre

indentation, by a conical indenter, of thin lms on
substrates, using the nite element method. The eVects
of the elastic and plastic properties of both the lm
and substrate on the hardness of the lmsubstrate
composite were studied by determining the average
pressure under the indenter as a function of the
indentation depth. They developed empirical equations for lmsubstrate combinations for which the
substrate is either harder or softer than the lm. For
the case of a soft lm on a harder substrate, the eVect
of substrate on lm hardness can be described as
H
H
f
=1+
H
H
s
s
B C
1 exp
(s /s )
f s (h /t )2
(E /E ) c f
f s
. . . . . . .
(20a)
where E and E are the Youngs moduli, s and s

f
s
f
s
are the yield strengths, and H and H are the hardf
s
nesses of the lm and substrate respectively, H is the
hardness of the composite, h is the contact indenc
tation depth, and t is the lm thickness. Similarly,
f
for the case of a hard lm on a softer substrate, the
hardness can be expressed as
H
H
f
=1+
H
H
s
s
B C
1 exp
(H /H )
f s
(h /t )
(s /s )(E /E )1/2 c f
f s f s
. . . . . . . (20b)
Composite hardness results were found to depend

only very weakly on Poissons ratio n. For cases in
which the lm and substrate have diVerent Youngs
moduli, hardness is observed to be independent of the
substrate for indentation depths less than about 03
of the lm thickness, after which the hardness slowly
increases/decreases because of the presence of the
substrate (Fig. 14). For cases in which the lm and
substrate have diVerent yield strengths, it is observed
that the variation of hardness with depth of indentation is qualitatively similar to cases in which the
lm and substrate have diVerent Youngs moduli.
Doerner and Nix (1986) empirically modelled the
in uence of the substrate on the elastic measurement
of very thin lm in an indentation test using the
following expression for the compliance
A B G C A BD
A B H
dh
1 p 1/2 1 n2
at
f 1 exp
f
=
dW
2 A
E
A1/2
f
1 n2
at
1 n2
s exp
f +
i +b
+
E
A1/2
E
s
i
. . . . . . . . (21)
C=
where the subscripts f, s, and i refer to the lm,

substrate, and indenter respectively. The term A1/2 is
equal to (245)1/2h for the Vickers or Berkovich
c thickness is t , and b is
indenter. The lm
f
the y intercept for the compliance versus 1/depth
plot, obtained for the bulk substrate, which can be
neglected in most cases. The weighting factors
[1 exp( at /A1/2)] and exp( at /A1/2) have been
f
added to account
for the changingf contributions of
the substrate and lm to the compliance. The factor
a can be determined empirically.
Analysis carried out by King (1987) veri ed that
equation (21) is an excellent functional form for
describing the in uence of the substrate, and theoretInternational Materials Reviews 2003
Vol. 48
No. 3
14
Effect of relative Youngs moduli of film and

substrate Ef /Es on composite hardness for soft
film on hard substrate and hard film on soft
substrate: yield strengths s and Poissons ratio
m are same for both substrates and films
(Bhattacharya and Nix, 1988b)
ically determined the values of a for various indenter

shapes. The value of a was found to depend on the
indenter shape and size and lm thickness and was
found to be independent of E /E . The values of a as
s
a function of A1/2/t for iBerkovich
(triangular)
f
indenters are shown in Fig. 15. The values of a were
found to be similar for square and triangular indenters.
Bhattacharya and Nix (1988b) analysed the deformations of a layered medium in contact with a conical
indenter using the nite element method. Their analysis also veri ed the relationship given in equation (21).
Examples of measured mechanical

properties of engineering materials
In this section, to illustrate the usage of nanoindentation techniques, typical data obtained on various
Bhushan and Li
15
139
Parameter a as function of normalised indenter

size for Berkovich indenter indenting layered
solid surface (King, 1987)
materials, coatings, and materials after surface treatments are presented.

Loaddisplacement curves
A variety of mechanical phenomena, such as transition from elastic to plastic deformation, creep
deformation, formation of subsurface cracks, and
crystallographic phase transition, can be studied by
analysing the loaddisplacement curves obtained
at diVerent loading conditions (Pethica et al., 1983;
Doerner and Nix, 1986; Stone et al., 1988; LaFontaine
et al., 1990c, 1991; Pharr et al., 1990; Page et al., 1992;
Oliver and Pharr, 1992; Pharr, 1992; Whitehead and
Page, 1992; Gupta et al., 1993, 1994; Gupta and
Bhushan, 1994, 1995a,b; Bhushan et al., 1995, 1996a;
Bhushan and Gupta, 1995). The loadtime sequence
used by Oliver and Pharr (1992) for study of various
materials included three loadingunloading cycles,
hold for 100 s at 10% of the peak load, reload, hold
for 100 s, and unload. Loaddisplacement curves for
electropolished, single crystal tungsten, fused silica
and (110) single crystal silicon are shown in
Figs. 1618. Only two loadingunloading cycles were
included for silicon. The softest material is tungsten,
with peak depth of almost 1000 nm at 120 mN load,
while the hardest is silicon which was penetrated to
a depth of only about 800 nm. The tungsten data are
typical of materials in which the hardness is relatively
small compared to the modulus, as is observed in
most metals; most of the indenter displacement in
these metals is accommodated plastically and only a
small portion is recovered on unloading. Fused silica
and silicon are harder and show larger elastic recovery
during unloading, the largest being that for fused
silica.
The unloadingreloading behaviours of the various
materials are diVerent. For tungsten, Fig. 16 shows
that the peak load displacements shift to higher values
in successive loading/unloading cycles. In addition,
the relatively large displacement just before nal
unloading is due to creep during the 100 s hold period
at peak load. Indentation at a very low load of
05 mN caused only elastic displacements (Fig. 16b).
At higher peak loads, the indentation is not just
elastic (Fig. 16c). When a threshold load of about
1 mN is reached, a sudden jump in displacement
corresponding to the onset of plasticity is observed,
and a permanent hardness impression is formed.
Tungsten at low loads (Fig. 16c) exhibits distinct
16
Load versus displacement plot for electropolished single crystal tungsten at peak load
of a 120 mN, b 05 mN (elastic contact), and
c 15 mN showing yield point (Oliver and
Pharr, 1992)
hysteresis loops, as might be expected if there were a

small amount of reverse plasticity upon loading.
However, the looping degenerates with cycling after
three or four cycles, and the loaddisplacement behaviour is largely elastic. The unloading/reloading curves
for fused silica are almost the same (Fig. 17). The
near perfect reversibility suggests that at peak loads
of 120 and 45 mN, deformation after initial unloading
is almost entirely elastic.
The behaviour of silicon, shown in Fig. 18, is in
sharp contrast to other materials (e.g. the fused silica
data shown in Fig. 17) (Pharr et al., 1989, 1990; Page
et al., 1992; Pharr, 1992). The data presented in
Fig. 18 were taken over two cycles of loading and
unloading. At high peak loads, the initial unloading
curve for silicon is not at all smooth, but exhibits a
Vol. 48
No. 3
140
17
Bhushan and Li
Load versus displacement plots for fused silica

at peak load of a 120 mN and b 45 mN (Pharr,
1992)
discontinuity (Fig. 18a). At lower peak loads, the

behaviour changes, and below some critical value the
discontinuity is no longer observed (Fig. 18b).
However, at this load the loaddisplacement behaviour shows another anomalous feature: a large hysteresis which shows no sign of degeneration through
several cycles of deformation. The fact that the curves
are highly hysteretic implies that deformation is not
entirely elastic. The discontinuity at high loads and
the non-degenerative hysteresis at low loads are quite
unique to silicon and are observed in each of the
(100), (110), and (111) orientations. The load below
which the discontinuity disappears and the hysteresis
becomes apparent is generally in the range 520 mN.
Pharr (1992) concluded that larger hysteresis
observed in the unloading curve at low loads is due
to a pressure induced phase transformation from
its normal diamond cubic form to a b-tin metal
phase. At some point in the transformation, an
amorphous phase is formed evidence of which is
reported by Callahan and Morris (1992). The discontinuity in displacement observed during unloading at peak loads of greater than about 15 mN is
believed to be due to formation of lateral cracks
at the base of the median crack which results in
the surface of the specimen being thrust upward
(Fig. 19). Lateral cracking is aided by the phase
transformations.
Vol. 48
No. 3
18
Load versus displacement plots for silicon (110)

at peak load of a 120 mN and b 45 mN (Pharr,
1992)
Weihs et al. (1992) indented silicon (100) with

120 mN of load. The loaddisplacement curve is
shown in Fig. 20a and the rst AE signal that was
recorded for this test is shown in Fig. 20b. The AE
signals were sharp and they correlated with small
jumps in tip displacement. The rise time for the signal
in Fig. 20b is 15 ms. After testing, radial cracks were
visible at indentation corners. AE events such as that
shown in Fig. 20b were recorded at applied load as
low as 48 mN on loading. In the nal stages of
19
Schematic illustration of coupling of phase

transformation and cracking during indentation of single crystal silicon (Pharr, 1992)
Bhushan and Li
141
20
a applied force as function of tip displacement for silicon (100) and b AE amplitude as function of
time during loading (Weihs et al., 1992)
21
a applied force as function of tip displacement for nickel film on glass substrate and b AE amplitude
as function of time during loading (Weihs et al., 1992)
unloading, small AE signals that had an inverted

shape compared to Fig. 20b were detected occasionally. Goken and Kempf (2001) reported that the
loaddisplacement curves obtained by nanoindentations on most metallic and intermetallic materials
show discontinuities or pop-ins during the initial
stages of loading. These pop-ins mark a sharp transition from pure elastic loading to a plastic deformation of the specimen surface, and thus correspond to
an initial yield point. On smooth surfaces, pop-ins
are observed frequently, but not on surfaces with a
high roughness. Step edges on the surface are believed
to act as dislocation sources for the initial yield
events. It should be noted that the pop-in load
determines a minimum necessary load to generate
plastic indentation.
Gupta et al. (1993, 1994) and Gupta and Bhushan
(1994) reported that hysteresis in cyclic indentation
and discontinuity kinks in the unloading curve are
considerably reduced by ion implantation of compound forming species O+ and N+ into single crystal
silicon. They further reported that amorphous silicon
lms did not exhibit either hysteresis in the cyclic
indentation or a discontinuity kink for the indentation
loads ranging from 1 to 90 mN. This suggests that
for a perfectly amorphous structure, the hysteresis
does not occur because of absence of crystallographic
pressure induced phase transition during cyclic
loading and unloading. In addition, the disordered
structure does not allow the nucleation and propagation of the lateral cracks beneath the indentation.
They have also reported that ion implantation has
an insigni cant eVect on hardness and elastic modulus. Improvement in the hardness is about 10%
as a result of implantation with carbon and boron
ions. Gupta et al. (1993, 1994) reported that ion
bombarded silicon exhibits a very low coeYcient
of friction (~005) and a low wear factor
(~10 6 mm3 N 1 m 1) on sliding against alumina
and 52100 steel balls.
Weihs et al. (1992) indented nickel lms on a glass
substrate (Fig. 21). The nickel lms debonded from
their substrates at forces ranging between 130 and
250 mN. The debonding events were marked by the
indenter tip jumping downward as chunks of the
nickel lm buckled away from underneath the indenter. The indenter tip jumped a distance equal to
the lm thickness as it initially debonded. Figure 21
also shows the corresponding AE trace with a rise
time of 18 ms. In this particular test, debonding
continued at higher forces and a second AE event
was recorded. After each test, optical microscopy
con rmed the delamination of the lm from underneath the indenter.
Li and Bhushan (1999c) indented 100 nm thick
DLC (diamond-like carbon) coatings deposited, on a
silicon substrate, by four diVerent deposition techniques: ltered cathodic arc (FCA), ion beam (IB),
electron cyclotron resonancechemical vapour deposition (ECRCVD), and sputtering (SP) (Fig. 22). The
indentation depths at peak load vary from 14 to
24 nm, i.e. less than the coating thickness. All of the
coatings exhibited indentation depths at peak load
slightly higher than that of the silicon substrate.
Vol. 48
No. 3
142
Bhushan and Li
FCA: filtered cathodic arc; IB: ion beam; ECRCVD: electron

cyclotron resonancechemical vapour deposition; SP
sputtering
22
Representative loaddisplacement plots of

indentations made at 02 mN peak indentation
load on various 100 nm thick DLC coatings and
on uncoated silicon substrate (Bhushan, 1999c)
Continuous stiffness measurements

Continuous stiVness measurements (CSM) are used
for continuous measurement of sample stiVness or
compliance (related to elastic modulus) during indentation. Oliver and Pharr (1992) measured continuous
stiVness of electropolished single crystal tungsten at
low peak loads of 05 mN or less, for purely elastic
contact. Continuous stiVness data for the indentation data in Fig. 16b are presented in Fig. 23a.
Comparisons of these data show that the measured
contact stiVness and thus the contact area does
increase and decrease in the way that would be
expected based on the loading history. The continuous
stiVness data for the indentation at 15 mN load
(plastic contact, Fig. 16c) are shown in Fig. 23b. It is
seen that for each of the four unloadings, the contact
stiVness changes immediately and continuously as the
specimen is unloaded. Thus, the contact area, which
varies in the same way as the contact stiVness, is not
constant during the unloading of the plastic hardness
impression, even during the initial stages of unloading.
The CSM technique oVers a direct measure of
dynamic contact stiVness during the loading portion
of an indentation test, and, being somewhat insensitive
to thermal drift, allows accurate observation of small
volume deformation. Li and Bhushan (2000a,b, 2001a)
have developed a CSM analytical method for studying
graded materials and multilayered structures. For a
uniform material, from equations (6) and (4a), contact
stiVness S is linearly proportional to contact depth
h . For a non-uniform material, the linear relationship
c
Vol. 48
No. 3
23
Contact stiffness versus time for a fully elastic

contact and b fully plastic contact on
electropolished
single
crystal
tungsten
measured
using
continuous
stiffness
technique (Oliver and Pharr, 1992)
between S and h does not exist. Therefore, the CSM

technique can bec used to study the mechanical properties of graded materials (such as magnetic tapes)
and multilayered structures (such as magnetic rigid
disks) by monitoring the change in contact stiVness,
elastic modulus and hardness as a function of indentation contact depth (Li and Bhushan, 2000a,b,
2001a,b, 2002a; Li et al., 2001). The contact stiVness,
elastic modulus and hardness as a function of contact
depth for a magnetic rigid disk and a magnetic tape
with multilayered structures are shown in Fig. 24a, b.
For the magnetic disk, from the variations in contact
stiVness, elastic modulus and hardness, it is easy to
distinguish one layer from another. In the case of the
magnetic tape, contact stiVness as a function of contact depth and its elastic modulus and hardness values
can be measured. An observed continuous decrease
in contact stiVness suggests that the tape coating has
graded properties. The elastic modulus values
obtained at a shallow indentation depth are the same
as those measured from tensile tests for the tape.
For metals and ceramics at elevated temperatures
and for polymers under most conditions, time dependent deformation occurs under the application of
load. This phenomenon is termed creep. The CSM
technique has been used to perform nanoscale indentation creep testing (Syed Asif and Pethica, 1997; Li
and Bhushan, 2000b). In an indentation creep test, a
constant load is applied to the indenter and the
change in indentation depth (size) is monitored as a
Bhushan and Li
143
a contact stiffness, elastic modulus and hardness as function of contact depth for magnetic rigid disk; b contact stiffness as
function of contact depth, and elastic modulus (E) and hardness (H) data at contact depth of 15 nm for magnetic tape A;
c indentation displacement, mean stress, and contact stiffness as function of time for magnetic tape A
24
Results of continuous stiffness measurements for multilayered structures (Li and Bhushan, 2002a)
function of time. Compared to conventional tensile

creep tests, CSM indentation creep experiments are
particularly useful as they simulate creep resulting
from asperity contact. The CSM technique has been
used to study the creep behaviour of bulk materials
(Syed Asif and Pethica, 1999), graded materials, and
multilayered solids (Li and Bhushan, 2000b, 2001a,b,
2002a; Li et al., 2001). Figure 24c shows the CSM
indentation creep results for a magnetic tape. The
tape exhibits an increase in indentation displacement
and a decrease in mean stress with time, indicating
that stress relaxation occurred during the hold segment. The contact stiVness remains almost constant
during the 600 s hold segment, indicating that the
contact between the tip and the tape does not change.
Nanoscale damage caused by fatigue is of critical
importance to the reliability of ultrathin protective
overcoats and micro/nanostructures. The cyclic loading used in CSM makes the technique useful for
evaluation of nanofatigue, which will be described
below.
Hardness and elastic modulus measurements
A nanoindenter is commonly used to measure surface
mechanical properties of bulk materials (Oliver and
Pharr, 1992; Bhushan, 1999b). Hardness and elastic
moduli for six bulk materials, three materials used as
a substrate for the construction of magnetic rigid
disks, and single crystal silicon are presented in

Fig. 25. The data show that, for several of the materials, indentation size has very little eVect on the
hardness values. In the case of aluminium and tungsten in Fig. 25a, there is a modest increase in hardness
at low loads. According to Oliver and Pharr (1992),
this could be due to surface-localised cold work
resulting from polishing. The modulus data also show
that there is very little evidence for an indentation
size eVect; i.e. the moduli remain more or less constant
over the entire range of load (Fig. 25b). Oliver and
Pharr (1992) reported that the hardness and modulus
values at the two highest loads are comparable with
literature values. Published values and measured
values of the magnetic disk substrates in Fig. 25cf
can be compared in Table 3. (Fracture toughness data
will be discussed below.)
A nanoindentation hardness value of single crystal
silicon (110) is found to be about 115 GPa at a peak
indentation depth of 750 nm. This value is slightly
higher than the macrohardness value of silicon of
910 GPa (INSPEC, 1988). Using an atomic force
microscope, Bhushan and Koinkar (1994) and
Bhushan et al. (1996b) measured the hardness of
single crystal silicon at an ultralow normal load and
a corresponding residual indentation depth of 25 nm
and reported a value of 135 GPa (Fig. 26). It is clear
that the hardness of silicon increases with a decrease
Vol. 48
No. 3
144
Bhushan and Li
a hardness and b elastic modulus as function of load for six bulk materials: mechanically polished single crystal aluminium,
electropolished single crystal tungsten, soda lime glass, fused silica, (001) single crystal sapphire, and (001) single crystal
quartz (Oliver and Pharr, 1992); cf hardness and elastic modulus as function of indentation depth (load ranging from 01 to
25 mN) for mechanically polished 10 mm thick electroless NiP film on AlMg alloy 5086, chemically strengthened alkali
aluminosilicate glass, chain silicate glassceramic (polycrystalline) (Canasite by Corning), and single crystal silicon (111)
(Bhushan and Gupta, 1995)
25
Measurement of mechanical properties using nanoindenter
Table 3
RMS roughness values measured by

atomic force microscope, hardness and
elastic modulus values measured by using
nanoindenter (at indentation depth of
approximately
20 nm), and fracture
toughness measured by using Vickers
microindenter
RMS
roughness, Hardness*,
nm
GPa
Elastic
Crack
modulus*, length,
GPa
c, mm
NiP/AlMg
36
60 (55)
130 (200)
Chemically strengthened glass
11
60 (58)
85 (73)
Glassceramic
61
85 (55)
100 (83)
Single crystal silicon (111)
095
110 (910) 200 (180)
Fracture
toughness*,
MPa m1/2
No cracks
Significant
cracking
(09)
594
065 (40)
* For hardness, elastic modulus, and fracture toughness, values

in parentheses are reported values measured by conventional
Vickers indentation method, conventional tensile pull test
method, and chevron notched short bar method respectively.
Vol. 48
No. 3
in the load and corresponding indentation depth. The

hardness is aVected by indentation size. An increase
in hardness at lower indentation depths may result
from contributions of the surface lms. At smaller
volumes, there is a lower probability of encountering
material defects. Furthermore, at small volumes, there
is an increase in the stress necessary to operate
dislocation sources (Gane and Cox, 1970; Sargent,
1986). According to the strain gradient plasticity
theory advanced by Fleck et al. (1994), the large
strain gradients inherent in small indentations lead
to the accumulation of the geometrically necessary
dislocations that cause enhanced hardening. These
are some of the plausible explanations for the increase
in hardness at smaller volumes.
A number of investigators have used the nanoindenter to study the eVect of ion implantation on the
hardness and elastic modulus of metals (Nastasi et al.,
1988; Lee and Mansur, 1989; Bourcier et al., 1990,
1991; Was, 1990), silicon (Pharr et al., 1989, 1990;
Pharr, 1992; Gupta et al., 1993, 1994; Gupta and
Bhushan and Li
26
Indentation hardness as function of residual

indentation depth for silicon (100) (Bhushan
et al., 1996b)
Bhushan, 1994; Bhushan and Koinkar, 1994; Li

and Bhushan, 1999c), ceramics (McHargue, 1989;
McHargue et al., 1990; OHern et al., 1990; Was and
DeKoven, 1991) and polymers (Lee et al., 1992, 1993;
Rao et al, 1993; VanLandingham et al., 2001).
Generally the ion implanted zone is shallow and the
nanoindenter is capable of measuring the changes in
mechanical properties.
The nanoindenter is ideal for measurement of
mechanical properties of thin lms and composite
structures. A number of investigators have reported
the hardness of composite structures with thin lms
on a substrate (Pethica et al., 1985; Burnett and
Rickerby, 1987c; Stone et al., 1988; Wu et al., 1988;
OHern et al., 1989; Bhushan and Doerner, 1989; Cho
et al., 1990; Rubin et al., 1990; Gissler et al., 1991;
Stone et al., 1991; Bhushan et al., 1992, 1995; Fabes
et al., 1992; Nastasi et al., 1993; Gupta and Bhushan,
1995a,b; Patton and Bhushan, 1996; Korsunsky et al.,
1998; Li and Bhushan, 1998b, 1999a,b,c; Bhushan,
1999b,c; McGurk and Page, 1999).
It is widely accepted that to measure the true
hardness of the lms, the indentation depth should
not exceed 10% of the lm thickness (Tabor, 1951).
Based on a nite element analysis of the indentation
of thin lms of various thicknesses by Bhattacharya
and Nix (1988b), Bhushan (1996, 1999a, 2002) concluded that the true hardness of the lms can be
obtained if the indentation depth does not exceed
about 30% of the lm thickness. At greater indentation depths, the composite hardness changes with
the indentation depth. Measured hardness values of
soft titanium lms on a hard sapphire substrate are
presented in Fig. 27. It can be seen that hardness
increases with a decrease in the lm thickness or an
increase in the indentation depth, as expected. The
lm hardness is the steady state hardness independent
of the indentation depth.
Microscratch resistance measurement

of bulk materials using micro/
nanoscratch technique
The microscratch technique is commonly used for
screening of bulk materials for wear resistance. As
27
145
Indentation hardness as function of indenter

displacement for titanium coatings on
sapphire substrates: numbers next to each set
of data correspond to coating thickness (Fabes
et al., 1992)
stated above, normal load applied to the scratch tip

is gradually increased during scratching until the
material is damaged. Friction force is sometimes
measured during the scratch test (Bhushan and
Gupta, 1995; Bhushan et al., 1996a). After the scratch
test, the morphology of the scratch region including
debris is observed in a SEM. Based on the combination of the changes in the friction force as a function
of normal load and SEM observations, the critical
load is determined and the deformation mode is
identi ed. Any damage to the material surface as a
result of scratching at a critical ramp-up load results
in an abrupt or gradual increase in friction. The
material may deform either by plastic deformation or
fracture. Ductile materials (all metals) deform primarily by plastic deformation, resulting in signi cant
ploughing during scratching. Tracks are produced the
width and depth of which increase with an increase
in the normal load. Ploughing results in a continuous
increase in the coeYcient of friction with an increase
in the normal load during scratching. Debris is generally ribbon-like or curly. By contrast, brittle materials
deform primarily by brittle fracture with some plastic
deformation. In the brittle fracture mode, the
coeYcient of friction increases very little until a critical
load is reached, at which the material fails catastrophically and produces ne rounded debris, and above
which the coeYcient of friction increases rapidly.
The coeYcient of friction pro les as a function of
normal load for scratches made on various ceramic
substrates and corresponding magnetic disks (substrates coated with 75 nm thick, sputtered CoPtNi
magnetic lm and 20 nm thick, sputtered hydrogenated, DLC lm) can be compared in Fig. 28. This
gure also includes the friction force pro le for a
single crystal silicon substrate for comparison. The
SEM images of two regions of 500 mm long scratches
made at 112 mN normal load on various samples
can also be compared in Fig. 28 (presented under
their corresponding friction pro le in each case). The
upper images in the sets of two SEM images for
each sample correspond to a region where friction
Vol. 48
No. 3
Bhushan and Li
Vol. 48
No. 3
28 Coefficient of friction profiles as function of normal load and corresponding SEM images (Bhushan and Gupta, 1995)
top rows: friction profiles for 500 mm long scratches made using diamond tip (1 mm tip radius) at normal load ranging from 112 mN on various ceramic substrates, corresponding magnetic
disks, and single crystal silicon; bottom rows: upper SEM images in sets of two images for each sample corresponding to location or normal load where friction increased abruptly and/or damage
began to occur (points indicated A in profiles) and lower SEM images corresponding to location close to end of scratch (~11 mN) (points indicated B in profiles): scratching direction was from
left to right (scale bars apply to all images)
146
Bhushan and Li
increased abruptly. These are the points indicated by

A in the friction pro les. The lower images in each
SEM set correspond to the region that is very close
to the end of a scratch. These are the points indicated
by B in the friction pro les. The extent of damage in
a scratch is estimated by the width and depth of the
scratch and by the amount of debris generated toward
the end of the scratch.
Single crystal silicon exhibited the lowest friction
with little ploughing at a low load and cracking at
higher loads. This observation suggests that scratching of silicon took place primarily by brittle deformation. In the case of the NiP coated AlMg substrate,
friction increase was continuous from the beginning
of scratching (Fig. 28, friction pro le). The SEM
images of the NiP coated AlMg substrate presented
in Fig. 28, show that material removal occurred by
ploughing with formation of curly ductile chips. It is
evident that scratching took place primarily by plastic
deformation typical of ductile materials. Ploughing is
responsible for the continuous increase in friction for
this substrate. Glass and glassceramic substrates and
corresponding disks and the NiP coated AlMg disk
exhibited relatively low friction with a sudden increase
at higher load. The glass substrate exhibited the
lowest friction followed by the glassceramic substrate. In the case of the NiP coated AlMg disk,
the load at which friction increased was lower than
that for the glass and glassceramic substrates. SEM
images of these samples exhibit ploughing in addition
to the formation of ne debris. There is no evidence
of cracking of ceramic substrates or the ceramic
overcoats used in all disks at magni cations as high
as50 000. Glass is chemically strengthened in order
to produce signi cant compressive stresses in the
glass surface. Glassceramic consists of ne grained
polycrystalline material in a glass matrix. Chemical
strengthening and the crystals add to the fracture
toughness of the material. Thus, both ceramic substrate materials are expected to deform with ductile
and brittle deformation modes. Ductile deformation
results in ploughing, whereas brittle deformation aids
in debris generation. Lower values of the coeYcient
of friction before a sudden increase, as compared to
the NiP coated AlMg substrate, suggest that brittle
fracture contributes to overall deformation. Hard
overcoats generally contain signi cant compressive
residual stresses. It is these compressive stresses that
allow ductile deformation with little cracking. It
should further be noted that a sudden increase in the
coeYcient of friction for ceramic substrates and for
all disks at some load results from signi cant damage
to the bulk material or to the coating surface (Fig. 28,
SEM images).
Based on the friction data, the width and depth of
scratches, the amount of debris generated, and scratch
morphology, glass substrates and corresponding
disks exhibit a lower coeYcient of friction against a
diamond tip and a superior resistance to scratch,
followed by glassceramic substrates and corresponding disks.
This example clearly suggests that deformation
modes and critical load to failure can be identi ed
using the scratch technique.
147
Nanoindentation and microscratch

techniques for adhesion
measurements, residual stress
measurements, and materials
characterisation of thin films
Adhesion describes the sticking together of two materials. Adhesion strength, in a practical sense, is the
stress required to remove a coating from a substrate.
Indentation and scratch on the micro- and nanoscales
are the two commonly used techniques to measure
adhesion of thin hard lms with good adhesion
(>70 MPa) to the substrate (Campbell, 1970; Mittal,
1978; Blau and Lawn, 1986; Bhushan, 1987; Bhushan
and Gupta, 1997). Nearly all coatings, by whatever
means they are produced, and surface layers of treated
parts are found to be in a state of residual (intrinsic
or internal ) stress. These are elastic stresses that exist
in the absence of external forces and are produced
through the diVerential action of plastic ow, thermal
contraction, and/or changes in volume created by
phase transformation. Microindentation and nanoindentation techniques are also used to measure
residual stresses (Bhushan and Gupta, 1997).
Microscratch and nanoscratch techniques (using a
nanoindenter) are also used to measure the scratch
resistance of surfaces of bulk materials (Bhushan and
Gupta, 1995; Bhushan et al., 1996a). The nanoindenter
has also been modi ed to conduct microwear (Wu
and Lee, 1994) and microimpact studies (Beake
et al., 2001).
In this section, adhesion measurements and residual
stress measurements using nanoindentation and microscratch apparatuses are described. Typical examples
are also presented.
Adhesion strength and durability
measurements using nanoindentation
In the indentation test method, the coating sample is
indented at various loads. At low loads, the coating
deforms with the substrate. However, if the load is
suYciently high, a lateral crack is initiated and propagates along the coating/substrate interface. The lateral
crack length increases with the indention load. The
minimum load at which coating fracture is observed
is called the critical load and is employed as the
measure of coating adhesion (Fig. 29). For relatively
thick lms, the indentation is generally made using a
Brinell hardness tester with a diamond sphere of
20 mm radius (Tangena and Hurkx, 1986), a Rockwell
hardness tester with a Rockwell C 120 cone with a
tip radius of 200 mm (Mehrotra and Quinto, 1985) or
a Vickers pyramidal indenter (Chiang et al., 1981; Lin
et al., 1990; Alba et al., 1993). However, for extremely
29
Schematic illustration of indentation method

for adhesion measurement
Vol. 48
No. 3
148
Bhushan and Li
thin lms, a Berkovich indenter (Stone et al., 1988)

or a conical diamond indenter with a tip radius of
5 mm and of 30 included angle (Bhushan et al., 1995)
is used in a nanoindenter.
It should be noted that the measured critical load
W is a function of hardness and fracture toughness
cr
in addition to the adhesion of the coating. Chiang
et al. (1981) have related the measured crack length
during indentation, the applied load, and the critical
load (at which coating fracture is observed ) to the
fracture toughness of the substrate/coating interface.
A semianalytical relationship derived between the
measured crack length c and the applied load W is
c=a 1
W 1/2
cr
W 1/4 .
W
. (22)
where
a t3/2 H1/2
a2= 1 c
(K )
Ic interface
and a is a numerical constant, t is the coating
1 H is the mean hardness, and
c (K )
thickness,
is
Ic interface
the fracture toughness of the substrate/coating
interface. Mehrotra and Quinto (1985) used this
analysis to calculate the fracture toughness of the
interface.
Marshall and Oliver (1987) estimated adhesion of
composites by measuring the magnitude of shear
(friction) stresses at bre/matrix interfaces in composites. They used a Berkovich indenter to push on the
end of an individual bre, and measured the resulting
displacement of the surface of the bre below the
matrix surface (due to sliding). The shear stress was
calculated from the forcedisplacement relation
obtained by analysis of the frictional sliding. The
force and displacement measurements were obtained
only at the peak of the load cycle, and the sliding
analysis was based on sliding at constant shear resistance at the interface. These experiments provided
measurements of average shear stresses at individual
bres.
Adhesion strength and durability
measurements using microscratch technique
Scratching the surface with a ngernail or a knife is
probably one of the oldest methods for determining
the adhesion of paints and other coatings. In 1822,
Friedrich Mohs used resistance to scratch as a measure of hardness. Scratch tests to measure adhesion of
lms was rst introduced by Heavens in 1950
(Heavens, 1950). A smoothly round chromium steel
stylus with a tungsten carbide or Rockwell C diamond
tip (in the form of 120 cone with a hemispherical tip
of 200 mm radius) (Perry, 1983; Mehrotra and Quinto,
1985; Valli, 1986) or a Vickers pyramidal indenter
(Burnett and Rickerby, 1987a; Bull and Rickerby,
1990; Venkataraman et al., 1992) for macro- and
microscratching, or a conical diamond indenter (with
1 or 5 mm tip radius and 60 included angle) for
nanoscratching (Wu, 1991; Bhushan et al., 1995), is
drawn across the coating surface. A normal load is
applied to the scratch tip and is gradually increased
during scratching until the coating is completely
removed. The minimum or critical load at which the
Vol. 48
No. 3
30
Geometry
of
scratch
measurement technique
for
adhesion
coating is detached or completely removed is used as

a measure of adhesion (Benjamin and Weaver, 1960;
Campbell, 1970; Greene et al., 1974; Ahn et al., 1978;
Mittal, 1978; Laugier, 1981; Mehrotra and Quinto,
1985; Valli, 1986; Burnett and Rickerby, 1987a; Sekler
et al., 1988; Wu, 1991; Bull and Rickerby, 1990; Cheng
et al., 1990; White et al., 1993; Bhushan et al., 1995;
Gupta and Bhushan, 1995a,b; Patton and Bhushan,
1996; Bhushan and Patton, 1996; Bhushan and Li,
1997; Li and Bhushan, 1999a,b,c; Huang et al., 2002).
It is a most commonly used technique to measure
adhesion of hard coatings with strong interfacial
adhesion (>70 MPa).
For the scratch geometry shown in Fig. 30, surface
hardness H is given by
H=
W
cr
pa2
. (23)
and adhesion strength t is given by (Benjamin and

Weaver, 1960)
t=H tan h
=
W
a
cr
pa2 (R2 a2)1/2
(24a)
W
cr
paR
R a .
(24b)
or
t=
if
where W is the critical normal load, a is the contact

cr R is the stylus radius.
radius, and
Burnett and Rickerby (1987a) and Bull and
Rickerby (1990) analysed the scratch test of a coated
sample in terms of three contributions: (1 ) a ploughing
contribution which will depend on the indentation
stress eld and the eVective ow stress in the surface
region, (2) an adhesive friction contribution due to
interactions at the indenter/sample interface, and (3)
an internal stress contribution since any internal stress
will oppose the passage of the indenter through the
surface, thereby eVectively modifying the surface ow
stress. They derived a relationship between the critical
normal load W and the work of adhesion W
cr
ad
pa2 2EW 1/2
ad
W =
. . . . . . . . (25)
cr
2
t
where E is Youngs modulus of elasticity and t is the
Bhushan and Li
coating thickness. Plotting W as a function of a2/t1/2

cr slope p(2EW /t)1/2/2
should give a straight line of
ad
from which W can be calculated. Bull and Rickerby
ad
suggested that either the line slope (interface toughness) or W could be used as a measure of adhesion.
ad
An accurate
determination of critical load W is
cr
sometimes diYcult. Several techniques, such as (1)
microscopic observation (optical or SEM) during the
test, (2) chemical analysis of the bottom of the scratch
channel (with electron microprobes), and (3) acoustic
emission, have been used to obtain the critical load
(Perry, 1983; Valli, 1986; Sekler et al., 1988; Wu,
1991). In some instruments, tangential (or friction)
force is measured during scratching to obtain the
critical load (Jacobson et al., 1983; Valli, 1986; Wu,
1991; MTS, 2001; Bhushan et al., 1995, 1997; Bhushan
and Li, 1997; Gupta and Bhushan, 1995a,b; Patton
and Bhushan, 1996; Bhushan and Patton, 1996; Li
and Bhushan 1998b, 1999a,b,c). The acoustic emission and friction force techniques have been reported
to be very sensitive in determining critical load.
Acoustic emission and friction force start to increase
as soon as cracks begin to form perpendicular to the
direction of the moving stylus.
Venkataraman et al. (1992) studied the adhesion of
as sputtered platinum thin lms on NiO single crystals. They calculated the work of adhesion of the
PtNiO system based on the width of the scratch
track at the point at which the lm was delaminated
from the substrate, the critical load required for
delamination, and the area of the delaminated region.
Their model used an elastic contact mechanics
approach to relate the stresses acting in a scratch
experiment to the strain energy released during lm
delamination. Was et al. (1993) identi ed all the major
scratch failure modes of palladium and tungsten
coatings.
Bhushan et al. (1995, 1997), Bhushan and Li (1997),
Gupta and Bhushan (1995a,b), Patton and Bhushan
(1996), Bhushan and Patton (1996), and Li and
Bhushan (1998b, 1999a,b,c) have used the scratch
technique to study adhesion and scratch resistance
(mechanical durability) of various ceramic lms.
Scratch tests conducted with a sharp diamond tip
simulate a sharp asperity contact. Bhushan and
co-workers have also conducted accelerated friction
and wear (ball-on-coated disk) and functional tests
and have found good correlation between the scratch
resistance and wear resistance measured using accelerated tests (Bhushan et al., 1995a, 1997; Gupta and
Bhushan, 1995b; Patton and Bhushan, 1996; Bhushan
and Patton, 1996; Li and Bhushan, 1998b, 1999a,b)
and functional tests (Patton and Bhushan, 1996;
Bhushan and Patton, 1996; Bhushan et al., 1997).
Based on this work, scratch tests can be successfully
used to screen materials and coatings for wear
applications.
Li and Bhushan (1999a) conducted scratch tests on
ultrathin DLC coatings of thicknesses ranging from
20 to 35 nm deposited by several techniques on
silicon (100) substrates. Results of these tests can be
compared in Fig. 31, which shows coeYcient of friction pro les as a function of increasing normal load,
and three-dimensional line AFM images of regions
over scratches at the respective critical loads (indi-
149
cated by the arrows in the friction pro les and AFM

images), for the various coatings of diVerent thicknesses and the uncoated silicon substrate. From the
AFM images in Fig. 31, the debris can be easily seen
and the appearance of the rst debris is indicated by
the arrows, which correspond to critical loads. It is
observed from Fig. 31 that there exists a well de ned
critical load for each coating. Before critical load, the
ECRCVD coatings exhibit a lower coeYcient of
friction of about 01, whereas the other coatings
exhibit a higher coeYcient of friction of about 0203.
The SP coatings exhibit a large variation in coeYcient
of friction. For the SP coatings, the coeYcient of
friction drops to zero just before critical load. This
may be attributed to the diVerence in mechanical
properties between the interface region and the coating itself. The AFM images clearly show that below
critical load the coatings were ploughed by the scratch
tip, associated with the plastic ow of materials. At
and after critical load, debris (chips) or buckling was
observed on the sides of scratches. Delamination or
buckling can be observed around or after critical
load. This suggests that damage of these coatings
starts from delamination and buckling. For the 35
and 5 nm thick FCA coatings, before critical load,
small debris is observed on the sides of scratches.
This suggests that the thinner FCA coatings may be
more brittle and easily broken. Also, the 35 and 5 nm
thick FCA coatings show more delamination and
buckling events than the other coatings. This indicates
that the thinner FCA coatings have higher interfacial
and residual stresses than the other coatings. It is
obvious that for a given deposition method, critical
load increases with increasing coating thickness. This
indicates that critical load is determined not only by
the adhesive strength to the substrate, but also by the
coating thickness. It can be seen from Fig. 31 that the
uncoated silicon substrate exhibits a lower coeYcient
of friction of about 01 before critical load. The AFM
image shows that the uncoated silicon substrate was
damaged by ploughing, associated with the plastic
ow of materials. At and after critical load, small and
uniform debris is observed and the amount of debris
increases with increasing normal load.
The critical loads estimated from the friction pro les and AFM images for the various coatings of
diVerent thicknesses and the uncoated silicon substrate can be compared in Fig. 32. For the 20 and
10 nm thick coatings, the ECRCVD coatings exhibit
the highest critical load, followed by the FCA, IB,
and SP coatings. The critical load of the 10 nm thick
FCA coating is comparable to that of the 10 nm thick
IB coating. For a given deposition method, the thicker
the coating, the higher is the critical load. For the
5 nm thick coatings, the IB coating exhibits the
highest critical load, followed by ECRCVD, FCA,
and SP coatings. At 35 nm, the critical load for all
coatings drops to about 04 mN, which is about the
same as the critical load for the uncoated silicon
substrate.
The optical images of wear tracks and debris
formed on all samples when slid against a 3 mm
diameter sapphire ball after a sliding distance of 5 m
can be compared in Fig. 33. The normal load used
for the 20 and 10 nm thick coatings was 200 mN and
Vol. 48
No. 3
150
31
Bhushan and Li
Coefficient of friction profiles as function of increasing normal load, and three-dimensional line AFM
images of regions over scratches at respective critical loads (indicated by arrows in friction profiles
and AFM images), for scratches made on various coatings of different thicknesses and on uncoated
silicon substrate (Li and Bhushan, 1999a)
Vol. 48
No. 3
Bhushan and Li
32
Critical loads estimated from friction profiles

and AFM images for various coatings of
different thicknesses and uncoated silicon
substrate (Li and Bhushan, 1999a)
the normal load used for the 5 and 35 nm thick

coatings and silicon substrate was 150 mN. Among
the 20 nm thick coatings, the optical images show
that the SP coating has a larger wear track and more
debris than the IB coating. No wear track and debris
were found on the 20 nm thick FCA and ECRCVD
coatings. Among the 10 nm thick coatings, the optical
images show that the SP coating was severely damaged, showing a large wear track with scratches and
much debris. The FCA and ECRCVD coatings show
smaller wear tracks and less debris than the IB
coating. The poor wear resistance of the 5 nm thick
FCA coating is in good agreement with the low
scratch critical load, which may be due to the higher
interfacial and residual stresses as well as brittleness
of the coating. At 35 nm, the FCA coating provides
no wear resistance, failing instantly like the uncoated
silicon substrate. Large block-like debris is observed
on the sides of the wear track of the FCA coating.
This indicates that large region delamination and
buckling occurred during sliding, resulting in large
33
151
block-like debris. The IB and ECRCVD coatings

are able to provide some protection against wear
at 35 nm.
Magnetic thin lm head sliders made with
Al O TiC substrate are used in magnetic storage
2 3
applications
(Bhushan, 1996). The multilayered, thin
lm, pole tip structure present on the head slider
surface wears more rapidly than the Al O TiC sub2 3
strate, which is much harder. Pole tip recession (PTR)
is a serious concern in magnetic storage. Two of the
DLC coatings superior in mechanical properties (ion
beam and cathodic arc carbon) were deposited on the
air bearing surfaces of Al O TiC head sliders. The
2 3 by running a metal
functional tests were conducted
particle (MP) tape in a computer tape drive. Average
PTR values as a function of sliding distance data are
presented in Fig. 34. It can be seen that PTR increases
for the uncoated head, whereas for the coated heads
there is a slight increase in PTR in early sliding
followed by little change. Thus, coatings provide
protection.
This example clearly suggests that material characterisations (hardness, elastic modulus, and scratch
resistance) are powerful ways of screening materials
and the data correlates well with the functional friction and wear performance.
Residual stress measurements using
nanoindentation
Indentation measurements similar to those used to
determine the hardness and elastic modulus of a lm
can also be used to measure the residual stresses in
the lm. When a compressive force on a biaxially
stressed lm during indentation is applied in a direction perpendicular to the lm, yielding will occur at
Optical images of wear tracks and debris formed on all samples when slid against sapphire ball after
sliding distance of 5 m (Li and Bhushan, 1999a)
Vol. 48
No. 3
152
Bhushan and Li
35
34
Pole tip recession as function of sliding

distance as measured with AFM for a uncoated
and ion beam carbon coated and b uncoated
and cathodic arc carbon coated Al2 O3 TiC
heads run against MP (metal particle) tapes
a smaller applied compressive force while the lm is

stressed in biaxial tension as compared with the
unstressed lm. Thus, the biaxial tension decreases
hardness and the biaxial compression increases hardness (Swain et al., 1977; Vitovec, 1986). LaFontaine
et al. (1990c, 1991) used the nanoindentation technique to measure the eVect of residual stresses on the
hardness of thin lms. For samples that do not
undergo large structural changes, changes in hardness
with time re ect a change in residual stress in the
lm (LaFontaine et al., 1990c, 1991). LaFontaine
et al. (1991) measured the stress relaxation in thin
aluminium lms and the residual stresses on identical
lms using an X-ray stress measurement technique.
Results of the indentation and X-ray stress measurements compared closely implying that decrease in
hardness with time resulted from the relaxation of
residual stresses. Thus, the indentation measurements
can be used to investigate stresses in thin lms.
Bolshakov et al. (1996) and Tsui et al. (1996) studied
the eVect of residual stresses on hardness and elastic
modulus. Based on indentation measurements and
nite element analysis of the indentation process, they
reported that, once pile-up is accounted for, the
residual stresses have little eVect on hardness and
elastic modulus. Suresh and Giannakopoulos (1998)
developed a method, based on nite element simulations, to estimate pre-existing residual stress/strain
elds. Their technique can be directly used for residual
stress measurements for engineered surfaces having
properties that are altered by such methods as
mechanical working, etching, ion implantation, case
hardening, laser treatments, shot peening or laser
shock peening.
Tsukamoto et al. (1987) measured the de ection at
the centre of a bent beam (bent as a result of residual
stresses in the lm) by pressing the beam at with a
nanoindenter. The bent beam is placed on a at glass
surface supported by two fulcrums, and a load
de ection curve is generated (Fig. 35). The distance
h can be estimated from the in ection point in the
a
curve.
Because of the limited atness of most substrates, the lm is removed from the substrate, and
then the initial de ection is measured. The true
Vol. 48
No. 3
a
schematic
diagram
of
deflection
measurement
of
bent
beam
using
nanoindenter and b loaddeflection curve for
warped composite beam and its substrate
de ection resulting from residual stresses in the lm

is equal to h h (see Fig. 35b). The curvature 1/R
a can
b then be calculated by the geoof the substrate
metric relationship
L2
. . . . . . . . . . . (26)
8(h h )
a
b
where L is the span.
Hong et al. (1990) used another de ection measurement technique. In this technique, a circular section
of the substrate is removed from beneath the lm to
produce a drumhead-like membrane and the load is
applied at its centre. The stiVness of the membrane
( lm) is a sensitive function of the biaxial tension in
it. The de ection h is related to load W as
R=
h=
W a2
g(k)
16pD
. (27)
where
D=
Et3
12(1 n2)
and a is the radius of the membrane, t is its thickness,

and the function g(k) depends on the membrane and
its geometry. If the geometry and elastic constants of
the membrane are known, the tension can be accurately evaluated. This technique can only be used to
study tensile residual stresses since compressive
stresses buckle the membrane when the substrate is
removed.
Other applications of nanoindentation

techniques
Nanoindentation techniques have been used for
measurement of time dependent viscoelastic/plastic
properties (creep/relaxation), nanofracture toughness,
and nanofatigue.
Time dependent viscoelastic/plastic
properties
Most materials including ceramics and even diamond
are found to creep at temperatures well below half
their melting point, even at room temperature.
Indentation creep and indentation load relaxation
(ILR) tests are used for measurement of the time
dependent ow of materials. These oVer the advantage
of being able to probe the deformation properties of
Bhushan and Li
a thin lm as a function of indentation depth and

location.
In the indentation creep test, the hardness indenter
maintains its load over a period of time under well
controlled conditions, and changes in indentation size
are monitored (Westbrook, 1957; Mulhearn and
Tabor, 19601961; Atkins et al., 1966; Walker, 1973;
Hooper and Brookes, 1984; Li et al., 1991). The
analysis of creep is more complex than the analysis
of creep data obtained using a conventional technique:
because of the shape of the tip, indentation stress
acting on the sample decreases with time as the
contact area increases. Chu and Li (1977, 1979, 1980)
developed an impression creep test on a macroscale
which used a circular tip with a at end.
Log (hardness) decreases (linearly for most metals)
with log(time at load). Nanoindentors are also used
for indentation creep studies (Li et al., 1991; Lucas
and Oliver, 1992; Raman and Berriche, 1992).
Figure 36a shows the plot of log(indentation hardness) as a function of log(time) for silicon at diVerent
temperatures. Indentation creep is in uenced by a
large number of variables such as the plastic deformation properties of the material, diVusion constants,
normal load of indenter, duration of the indentation,
and the test temperature. Li et al. (1991) reported
that for temperatures between 27C and melting, the
mechanism of dislocation glide plasticity dominates
the indentation creep process.
In a typical ILR test, the indenter is rst pushed
into the sample at a xed displacement rate until a
predetermined load or displacement is achieved and
the position of the indenter is then xed. The material
below the indenter is elastically supported and will
continue to deform in a non-elastic manner, thereby
tending to push the indenter farther into the sample.
Load relaxation is achieved by conversion of elastic
strain in the sample into inelastic strain in the sample.
During the test, the load and position of the indenter
and the specimen are continuously monitored.
Typically, the indenter motion is held constant and
the changes in the load are monitored as a function
of time. It is possible to obtain the plastic indentation
rate from the indentation load and total depth information during the relaxation run (Hart and Solomon,
1973; Chu and Li, 1980; Nastasi et al., 1993). The
resulting load relaxation data are reported in the
form of log(indentation pressure) as a function of
log(plastic indentation strain rate) (Hannula et al.,
1985; LaFontaine et al., 1990a; Wu, 1991).
The indentation pressure is calculated by dividing
the load by the projected area of the indenter. Once
the plastic indentation depth is known as a function
of time, the projected area is determined experimentally, as described above. The plastic indentation
strain rate (h 1(dh/dt), where h is the current indentation depth) is calculated in a manner similar to that
for bulk relaxation data. First, the plastic indentation
depth is calculated which is equal to the total depth
minus its elastic depth based on load versus time
data. The plastic indentation depthtime data are
divided into segments. Over each segment, the plastic
indentation depth is assumed to vary linearly with
time. At the midpoint of each segment the slope of
the plastic indentation depth curve is determined to
153
Tm is melting temperature; l is load; xo is plastic indentation

depth at beginning of relaxation; xis plastic indentation
rate
36
a indentation creep data for silicon at different

temperatures (Li et al., 1991) and b indentation
load relaxation data of (111) single crystal
aluminium sample at 25C (LaFontaine et al.,
1990b)
obtain the plastic indentation rate. The average plastic

indentation depth is used along with the corresponding average load over the same time interval to
determine the indentation pressure (LaFontaine et al.,
1990b). Figure 36b shows typical log (indentation
pressure) versus log(x/x ) (i.e. plastic indentation rate
o
normalised with the depth at the beginning of the
relaxation experiment) ow curves for a (111) single
crystal aluminium sample.
Another technique to measure strain rate sensitivity
of submicrometre lms was developed by Mayo and
Nix. (1988). They developed two procedures. In the
rst procedure, known as the constant rate of loading
test, individual indentations are performed at a prescribed loading rate that is varied from one indentation to another by about a factor of 2. The values
of the indentation pressure and strain rate from tests
performed at diVerent loading rates were compared
at a common indentation depth. In a second related
technique known as the loading rate change test, the
loading rate is held constant at a speci ed depth, but
is suddenly changed to a new value and the subInternational Materials Reviews 2003
Vol. 48
No. 3
154
Bhushan and Li
sequent changes in pressure and strain rate are

monitored.
Mayo et al. (1990) developed yet another procedure
for determining the strain rate sensitivity of nanophase materials in order to conduct tests at the high
loading rates required for very hard materials. In this
procedure, as soon as the indenter contacts the sample
surface, the indenter loading rate is instantaneously
increased to a high value and this value is maintained
until the indenter reaches a prescribed displacement.
The load is held constant at this point, and the
displacement is monitored as a function of time. The
initial fast descent rate of the indenter produces a
substantial amount of creep during a constant hold
period. A range of descent rates is realised ranging
from high values at the beginning of the hold period
to smaller values as the material stops deforming.
These correspond to a range of strain rates, and
several stressstrain pairs can be obtained (Raman
and Berriche, 1992).
The strain rate sensitivity of materials is measured
in terms of the stress exponent n, which is de ned by
the equation
plastic indentation rate =A(indentation pressure)n
. . . . . . . . (28)
where A and n are constants. The stress exponent is
found as the slope of a loglog plot of plastic indentation rate (or strain rate) and indentation pressure.
In the ILR test, the continuous change in the contact
area results in continuous changes in both plastic
indentation rate and pressure. Thus, data from a
single indentation test, which may span several orders
of magnitude in both strain rate and pressure, are
suYcient to determine the stress exponent. Stress
exponent can be used to de ne the superplasticity of
a material. The variations in stress exponent re ect
the changes that may take place when the substructure generated at high strain rate approaches the
equilibrium condition (Mayo and Nix, 1988).
Nanofracture toughness
Fracture toughness K of a material is a measure of
Ic
its resistance to the propagation of cracks and the
ratio H/K is an index of brittleness, where H is the
hardness. Ic
Resistance to fracture is a strong function
of crack pattern. It is typically measured in a test in
which a specimen containing a sharp crack of known
length c is subjected to an applied stress s (Fig. 37),
which is increased during the test until the sample
fractures (Lawn, 1993). The magnitude of the stresses
near the crack tip is determined by the stress intensity
factor K which, in turn, depends on s, c, and specimen
geometryI (A)
K =As(pc)1/2 . . . . . . . . . . . (29)
I
The term A provides correction for the
thickness/width ratio of the material. Units of stress
intensity factor are MPa m1/2. With more intense
stress or with deeper cracks, the stress intensity
becomes suYcient for the fracture to progress spontaneously. This threshold stress intensity is a property
of the material and is called the critical stress intensity
factor K or the fracture toughness of the material.
Ic
Ceramics generally have relatively low fracture toughInternational Materials Reviews 2003
Vol. 48
No. 3
37
Schematic of standard specimen used for

measurement of fracture toughness of
materials in tension
ness; consequently, it is an important property to be

considered for the selection of ceramics for industrial
applications.
The indentation fracture toughness technique is a
simple method for the determination of fracture
toughness (Palmquist, 1957; Lawn and Wilshaw,
1975; Evans and Charles, 1976; Lawn and Evans,
1977; Lawn and Marshall, 1979; Lawn et al., 1980;
Antis et al., 1981; Chantikul et al., 1981; Chiang et al.,
1981, 1982; Henshall and Brookes, 1985; Cheng et al.,
1990; Cook and Pharr, 1990; Choi and Salem, 1993;
de Boer et al., 1993; Lawn, 1993; Pharr et al.,
1993; Bhushan et al., 1996a; Hainsworth et al., 1998).
The indentation cracking method is especially useful
for measurement of fracture toughness of thin lms
or small volumes. This method is quite diVerent from
conventional techniques in that no special specimen
geometry is required. Rather, the method relies on
the fact that when indented with a sharp indenter,
most brittle materials form radial cracks and the
lengths of the surface traces of the radial cracks
(Fig. 38b; for de nition of crack length, see Fig. 38a)
have been found to correlate reasonably well with
fracture toughness. By using simple empirical equations, fracture toughness can then be determined from
simple measurement of crack length. Although this
indentation fracture toughness measurement technique has been widely used in practical applications,
the stressstrain state is still, to a large extent,
unknown. Stressstrain simulation on the indentation
fracture with Vickers, Berkovich, and cube corner
indenters is needed for developing indentation fracture models and modifying the existing measurement
methodology.
In microindentation, cracks at relatively high indentation loads of several hundred grams are on the
order of 100 mm in length and can be measured
optically. However, to measure the toughness of very
thin lms or small volumes, much smaller indentations are required. However, a problem exists in
extending the method to the nanoindentation regime
in that there are well de ned loads, called cracking
thresholds, below which indentation cracking does
not occur in most brittle materials (Lankford, 1981).
For a Vickers indenter, cracking thresholds in most
ceramics are about 25 g. Pharr et al. (1993), Li et al.
(1997) and Li and Bhushan (1998a, 1999d) have
found that the Berkovich indenter (a three-sided
pyramid) with the same depth/area ratio as a Vickers
indenter (a four-sided pyramid), has cracking thresholds very similar to those of the Vickers indenter.
They showed that cracking thresholds can be substan-
Bhushan and Li
155
39
38
a schematic of Vickers indentation with radial

cracks and b optical image of Vickers
indentation made on glassceramic substrate
at 500 g load
tially reduced by using sharp indenters; i.e. indenters

with smaller included tip angles, such as a three-sided
indenter with the geometry of the corner of a cube.
Studies using a three-sided indenter with the geometry
of a corner of a cube have revealed that cracking
thresholds can be reduced to loads as small as 05 g,
for which indentations and crack lengths in most
materials are submicrometre in dimension.
Based on fracture mechanics analysis, Lawn et al.
(1980) developed a mathematical relationship
between fracture toughness and indentation crack
length, given as
AB A B
E 1/2 W
K =B
Ic
H
c3/2
. (30)
where W is the applied load and B is an empirical

constant depending upon the geometry of the indenter
(also see Lawn, 1993; Pharr et al., 1993). Antis et al.
(1981) conducted a study on a number of brittle
materials chosen to span a wide range of toughnesses.
They indented with a Vickers indenter at several
loads and measured crack length optically. They
found a value of B=0016 to give good correlation
Table 4
Typical mechanical properties of materials

tested by Pharr et al. (1993)
Material
E, GPa
H, GPa
KIc , MPa m1/2
Soda lime glass

Fused quartz
(111) silicon
(111) sapphire
Si3 N4
70
72
168
403
300
55
89
93
216
163
070
058
070
22
40
Indentations in fused quartz made with cube

corner indenter showing radial cracking at
indentation loads of a 12 g and b 045 g (Pharr
et al., 1993)
between the toughness values measured from the

crack length and those obtained using more conventional methods. Mehrotra and Quinto (1985) used a
Vickers indenter to measure the fracture toughness of
coatings. Pharr et al. (1993) tested several bulk ceramics (listed in Table 4) using Vickers, Berkovich,
and cube corner indenters. They found that the fracture toughness equation can be applied for the data
obtained with all three indenters provided a diVerent
empirical constant was used for a cube corner
indenter. The constant B for the Vickers and
Berkovich indenters was found to be about 0016 and
for the cube corner indenter it was about 0032. It
should be noted that indentation cracking thresholds
can be signi cantly reduced by using a cube corner
indenter. This indicates that there are possible
indenter geometries that lead to even lower cracking
thresholds. Pharr et al. (1993) further reported that
predominant cracks formed with Vickers or
Berkovich indenters are cone cracks and with a cube
corner indenter, predominant cracks are radial cracks
(Fig. 39). Bhushan et al. (1996a) reported that cracks
formed in microcrystalline ceramic material (glass
ceramic) with a Vickers indenter are radial cracks
(Fig. 38b). Note that cracks propagate in a zigzag
manner. The interlocked crystal morphology is
responsible for propagation in a zigzag manner. Using
equation (31) below, the fracture toughness for this
material has been calculated (see Table 3).
Chantikul et al. (1981) developed a relationship
between fracture toughness and the indentation fracture strength and applied load, given as
K =c(E/H )1/8(s W 1/3)3/4 . . . . . . . (31)
Ic
f
where s is the fracture strength after indentation at
f
a given load and c is an empirical constant (059).
The advantage of this analysis is that the measurement of crack length is not required. Mecholsky et al.
(1992) used this analysis to calculate the fracture
toughness of diamond lms on silicon. They indented
the lms at various indentation loads of 39 kg and
then fractured the lms in four-point exure to measure fracture strength. The data were then used to
obtain fracture toughness. Equation (31) was found
to hold for the measurements. They reported the
Vol. 48
No. 3
156
Bhushan and Li
fracture toughness of 6 and 12 mm thick diamond

lms on silicon to be on the order of 2 MPa m1/2.
For fracture toughness measurement of ultrathin
lms ranging from 100 nm to a few micrometres,
indentation or four-point exure techniques cannot
be used. Because of the shallow indentation depths
required in the indentation technique, it is diYcult to
measure a radial crack length even under SEM. Li
et al. (1997) and Li and Bhushan (1998a, 1999d)
developed a novel technique based on nanoindentation in which through thickness cracking in the
coating is detected from a discontinuity observed in
the loaddisplacement curve and the energy released
during the cracking is obtained from the curve. Based
on the energy released, fracture mechanics analysis is
then used to calculate fracture toughness. A cube
corner is preferred because the through thickness
cracking of hard lms can be accomplished at lower
loads (Li and Bhushan, 1998a).
Loaddisplacement curves of indentations made at
30, 100, and 200 mN peak indentation loads together
with SEM micrographs of indentations on 0.4 mm
thick cathodic arc carbon coating on silicon are
shown in Fig. 40. Steps are found in all loading curves
as indicated by arrows in Fig. 40a. In the 30 mN
SEM micrograph, in addition to several radial cracks,
ring-like through thickness cracking is observed with
small lips of material overhanging the edge of the
indentation. The step at about 23 mN in the loading
curves of indentations made at 30 and 100 mN peak
indentation loads result from the ring-like through
thickness cracking. The step at 175 mN in the loading
curve of indentation made at 200 mN peak indentation load is caused by spalling.
No steps were observed in the loading curve of
indentation made at 20 mN peak indentation load
(not shown here)which suggests that the coating under
the indenter was not separated instantaneously from
the bulk coating via the ring-like through thickness
cracking but occurred over a period of time. At
30 mN peak indentation load, partial ring-like spalling is observed around the indenter and the other
parts of the lm bulge upward. This partial ring-like
spalling is believed to result in the step in the loading
curve. The absence of long steps in the loading curve
for uncoated silicon reported by Li et al. (1997) and
Li and Bhushan (1998a) suggests that the steps in the
loading curve for the coating result from the lm
cracking. Based on their work, the fracture process
progresses in three stages: (1) primary ring-like
through thickness cracks form around the indenter
by high stresses in the contact area, (2 ) delamination
and buckling occur around the contact area at the
lm/substrate interface by high lateral pressure, (3)
secondary ring-like through thickness cracks and
spalling are generated by high bending stresses at the
edges of the buckled lm (see Fig. 41a). In the rst
stage, if the lm under the indenter is separated from
the bulk lm via the primary ring-like through thickness cracking, a corresponding step will be present in
the loading curve. If discontinuous cracks form and
the lm under the indenter is not separated from the
remaining lm, no step appears in the loading curve
because the lm still supports the indenter and the
indenter cannot suddenly advance into the material.
Vol. 48
No. 3
40
a loaddisplacement curves of indentations

made at 30, 100, and 200 mN peak indentation
loads using cube corner indenter (arrows
indicate steps during loading portion of load
displacement curve) and b SEM micrographs
of indentations on 0.4 mm thick cathodic arc
carbon film on silicon (Li et al., 1997)
In the second stage, for the cathodic arc carbon lms

under discussion here, the advances of the indenter
during the radial cracking, delamination, and buckling are not large enough to form steps in the loading
curve because the lm around the indenter still supports the indenter, but generate discontinuities which
change the slope of the loading curve with increasing
indentation loads. In the third stage, the stress concentration at the end of the interfacial crack cannot be
relaxed by the propagation of the interfacial crack.
With an increase in indentation depth, the height of
the bulged lm increases. When the height reaches a
critical value, the bending stresses caused by the
bulged lm around the indenter will result in the
secondary ring-like through thickness crack formation and spalling at the edge of the buckled lm, as
shown in Fig. 41a, which leads to a step in the loading
curve. This is a single event and results in the separation of the part of the lm around the indenter from
the bulk lm via cracking through lms. The step in
the loading curve is totally from the lm cracking
and not from the interfacial cracking or the substrate
cracking.
The area under the loaddisplacement curve is the
work performed by the indenter during elasticplastic
deformation of the lmsubstrate system. The strain
energy release in the primary/secondary ring-like
Bhushan and Li
157
OBF. For point C, the elasticplastic energy stored

in the lmsubstrate system should be OACF.
Therefore, the energy diVerence before and after the
crack generation is the area of ABC; i.e. this energy
stored in ABC will be released as strain energy to
create the ring-like through thickness crack.
According to the theoretical analysis by Li et al.
(1997), the fracture toughness of thin lms can be
written as
41
a schematic of various stages in nanoindentation fracture for filmsubstrate system

and b schematic of loaddisplacement curve
showing step during loading cycle and
associated energy release
cracking and spalling can be calculated from the

corresponding steps in the loading curve. Figure 41b
shows a modelled loaddisplacement curve: OACD
is the loading curve, DE is the unloading. Since the
primary ring-like through thickness cracking does not
always lead to a step in the loading curve in some
lms, the secondary ring-like through thickness crack
should be considered. It should be emphasised that
the edge of the buckled lm is far from the indenter;
therefore, it does not matter if the indentation depth
exceeds the lm thickness or if deformation of the
substrate occurs around the indenter when fracture
toughness of the lm is measured from the released
energy during the secondary ring-like through thickness cracking (spalling). Suppose that the secondary
ring-like through thickness cracking occurs at AC.
Now, let us consider the loading curve OAC. If the
secondary ring-like through thickness crack does not
occur, it can be understood that OA will be extended
to OB to reach the same displacement as OC. This
means that the crack formation changes the loading
curve OAB into OAC. For point B, the elasticplastic
energy stored in the lmsubstrate system should be
E
U 1/2
. . . . . . (32)
n2)2pC t
R
where E is the elastic modulus, n is the Poissons
ratio, 2pC is the crack length in the lm plane, U is
the strain Renergy diVerence before and after cracking,
and t is the lm thickness.
Using equation (32), the fracture toughness of the
04 mm thick ltered cathodic arc carbon coating is
calculated. The strain energy diVerence U of 71 nN m
is assessed from the steps in Fig. 40a at the peak
indentation load of 200 mN. The loading curve is
extrapolated from the starting point of the step, up
to the same displacement as the step. The area
between the extrapolated line and the step is the
estimated strain energy diVerence before and after
cracking. A value for C of 70 mm is measured from
the SEM micrograph in RFig. 40b. The secondary ringlike crack is where the spalling occurs. For an E value
of about 300 GPa measured using nanoindentation
(see Table 5, below) and an assumed value of 025 for
n, fracture toughness is calculated as about
118 MPa m1/2.
This technique can be used to evaluate the brittleness of ultrathin lms with a thickness on the order
of 100 nm. However, theoretical analysis about the
driving force and stressstrain state during indentation cracking is needed. Dao et al. (2001) and
Vaidyanathan et al. (2001) studied the dependence of
uniaxial and multiaxial deformation on both shear
stresses and normal stresses. They found that constrained deformation of the material around the
indenter results in incomplete circular patterns of
shear bands. These results help in understanding the
driving force that causes cracking in indentation.
K =
Ic
(1
Nanofatigue
Fatigue fracturing progresses through a material via
changes within the material at the tip of a crack,
where there is a high stress intensity. There are several
situations: cyclic fatigue, stress corrosion, and static
fatigue. Cyclic fatigue results from cyclic loading of
machine components; e.g. the stresses cycle from
tension and compression that occurs in a loaded
rotating shaft. Fatigue also can occur with uctuating
stresses of the same sign, as occurs in a leaf spring,
in a dividing board. In a low ying slider in a
head/disk interface, isolated asperity contacts occur
during use and the fatigue failure occurs in the
multilayered thin lm structure of the magnetic disk
(Bhushan, 1996). Asperity contacts can be simulated
using a sharp diamond tip in an oscillating contact
with the thin lm disk.
Li and Chu (1979) developed an indentation fatigue
test, called impression fatigue. In this test, a cylindrical
Vol. 48
No. 3
158
Bhushan and Li
P: cyclic load; Pmean : mean load; Po : oscillation load

amplitude; v: oscillation frequency
42
Schematic of fatigue test on coatingsubstrate

system using continuous stiffness measurement (CSM) technique
indenter with a at end was pressed onto the surface

of the test material with a cyclic load and the rate of
plastic zone propagation was measured.
Wu et al. (1991) developed a nanoindentation
fatigue test by modifying their nanoindenter. The
cyclic indentation was implemented by servo controlling the PZT stack to drive the indenter. Li and
Bhushan (2002b,c) developed a novel technique to
perform nanofatigue experiments. This technique utilises a depth-sensing nanoindenter using a harmonic
force. A conical diamond indenter with a nominal
1 mm radius tip is used. The CSM technique provides
load cycles of a sinusoidal shape at high frequencies
that can be used to perform nanoscale fatigue tests.
The fatigue behaviour of coatings can be studied by
monitoring the change in contact stiVness, since the
contact stiVness is sensitive to the damage formation.
To obtain deformation and damage during fatigue
loading, large amplitude oscillations were used. The
numbers of cycles were determined from the elapsed
time. Figure 42 shows the schematic of a fatigue test
on a coatingsubstrate system using the CSM technique. Load cycles are applied to the coating, resulting
in a cyclic stress.
The following results can be obtained: (1) endurance limit, i.e. the maximum load below which there
is no coating failure for a preset number of cycles; (2)
number of cycles at which the coating failure occurs;
and (3) changes in contact stiVness measured by using
the unloading slope of each cycle which can be used
to monitor the propagation of the interfacial cracks
during the cyclic fatigue process.
Figure 43a shows the contact stiVness as a function
of the number of cycles for 20 nm thick FCA coatings
cyclically deformed by various oscillation load amplitudes with a mean load of 10 mN at a frequency of
45 Hz. At 4 mN load amplitude, no change in contact
stiVness was found. This indicates that 4 mN load
amplitude is not high enough to damage the coatings.
At 6 mN load amplitude, an abrupt decrease in contact
stiVness was found at a certain number of cycles,
indicating that fatigue damage has occurred. With
increasing load amplitude, the number of cycles to
failure N decreases for all coatings. Load amplitude
versus N f, a so called SN curve, is plotted in Fig. 43b.
f load amplitude, below which no fatigue
The critical
damage occurs (an endurance limit), was identi ed
for each coating. This critical load amplitude together
with mean load is of critical importance to the design
Vol. 48
No. 3
of head/disk interfaces or MEMS/NEMS device

interfaces.
To compare the fatigue lifetimes of diVerent coatings studied, the contact stiVness as a function of the
number of cycles for 20 nm thick FCA, IB,
ECRCVD, and SP coatings cyclically deformed by
an oscillation load amplitude of 8 mN with a mean
load of 10 mN at a frequency of 45 Hz is shown in
Fig. 43c. The FCA coating has the longest N , folf
lowed by ECRCVD, IB, and SP coatings. In
addition, after the N , the contact stiVness of the FCA
f decrease than the other coatcoating shows a slower
ings. This indicates that after the N , the FCA coating
f
had less damage than the others. The fatigue behaviour of FCA and ECRCVD coatings of diVerent
thicknesses can be compared in Fig. 43d. For both
coatings, N decreases with decreasing coating thickness. At 10 fnm, FCA and ECRCVD coatings have
almost the same fatigue life. At 5 nm, the ECRCVD
coating shows a slightly longer fatigue life than the
FCA coating. This indicates that even for nanometre
thick DLC coatings the microstructure and residual
stresses are not uniform across the thickness direction.
Thinner coatings are more in uenced by interfacial
stresses than thicker coatings.
Figure 44 shows the high magni cation SEM
images of 20 nm thick FCA coatings before, at, and
after N . In the SEM images, the net-like structure is
f
the gold lm coated on the DLC coating, which
should be ignored in analysing the indentation fatigue
damage. Before the N , no delamination or buckling
f
was found except the residual indentation mark at
magni cations up to 1 200 000 using SEM. This
suggests that only plastic deformation occurred before
the N . At the N , the coating around the indenter
f
bulgedf upwards, indicating
delamination and buckling. Therefore, it is believed that the decrease in
contact stiVness at the N results from delamination
f from the substrate. After
and buckling of the coating
the N , the buckled coating was broken down around
f
the edge of the buckled area, forming a ring-like
crack. The remaining coating overhung at the edge
of the buckled area. It is noted that the indentation
size increases with increasing number of cycles. This
indicates that deformation, delamination and buckling, and ring-like crack formation occurred over a
period of time.
The schematic diagrams in Fig. 44 show the various
stages of indentation fatigue damage for a coating
substrate system. Based on this study, three stages of
indentation fatigue damage appear to exist: (1) indentation induced compression; (2) delamination and
buckling; (3) ring-like crack formation at the edge of
the buckled coating. Residual stresses are often
induced in coatings by the deposition process. The
model shown in the schematics of Fig. 44 considers a
coating with a uniform biaxial residual compression
s . In the rst stage, indentation induces elastic
r
plastic deformation, exerting an outward acting pressure on the coating around the indenter. Interfacial
defects such as voids and impurities act as original
cracks. These cracks propagate and link up as the
indentation compressive stress increases. At this stage,
the coating, which is under the indenter and above
the interfacial crack (with a crack length of 2a), still
Bhushan and Li
159
a contact stiffness as function of number of cycles for 20 nm thick FCA coatings cyclically deformed by various oscillation load
amplitudes with mean load of 10 mN at frequency of 45 Hz; b plot of load amplitude versus number of cycles to failure Nf ; c
contact stiffness as function of number of cycles for four types of (20 nm thick) coating; d contact stiffness as function of
number of cycles for two types of coating of different thicknesses (Li and Bhushan, 2002b)
43
Use of contact stiffness measurement to assess fatigue life
maintains a solid contact with the substrate; the

substrate still fully supports the coating. Therefore,
this interfacial crack does not lead to an abrupt
decrease in contact stiVness, but gives a rise to a
slight decrease in contact stiVness, as shown in Fig. 43.
The coating above the interfacial crack is treated as
a rigidly clamped disc. It is assumed that the crack
radius a is large compared to the coating thickness t.
Since the coating thickness ranges from 20 to 5 nm,
this assumption is easily satis ed in this study (radius
of the delaminated and buckled area, shown in the
SEMs of Fig. 44, is on the order of 100 nm). The
compressive stress caused by indentation is given as
(Marshall and Evans, 1984)
s=
i (1
EV
i
e=
n) i 2pta2(1
n)
. (33)
where n and E are the Poissons ratio and elastic

modulus of the coating, V is the indentation volume,
t is the coating thickness,i and a is the crack radius.
With increasing number of cycles, the indentation
volume V increases. Therefore, the indentation comi
pressive stress s increases accordingly. In the second

i
stage, buckling occurs during the unloading segment
of the fatigue testing cycle when the sum of indentation compressive stress s and residual stress s
i
r
exceeds the critical buckling stress s for the delamib
nated circular section as given by (Evans and
Hutchinson, 1984)
AB
m2E
t 2
s =
b 12(1 n2) a
. (34)
where the constant m equals 4267 for a circular

clamped plate with a constrained centre point and
1468 when the centre is unconstrained. The buckled
coating acts as a cantilever. In this case, the indenter
indents a cantilever rather than a coatingsubstrate
system. This ultrathin coating cantilever has much
less contact stiVness than the coatingsubstrate
system. Therefore, the contact stiVness shows an
abrupt decrease at the N . In the third stage, with
f the size of the delamiincreasing number of cycles,
nated and buckled area increases, resulting in a further
decrease in contact stiVness since the cantilever beam
Vol. 48
No. 3
160
44
Bhushan and Li
High magnification SEM images of 20 nm FCA coatings and schematic diagrams of various stages in
indentation fatigue damage for coatingsubstrate system (Li and Bhushan, 2002b)
length increases. On the other hand, a high bending

stress acts at the edge of the buckled coating. The
larger the buckled area, the higher is the bending
stress. The cyclical bending stress causes fatigue
damage at the edge of the buckled coating, forming
a ring-like crack. The coating under the indenter is
separated from the bulk coating (caused by the ringlike crack at the edge of the buckled coating) and the
substrate (caused by the delamination and buckling
in the second stage). Therefore, the coating under the
Table 5
indenter is not constrained but is free to move with

the indenter during fatigue testing. At this point, the
sharp nature of the indenter is lost because the coating
under the indenter sticks to the indenter. The indentation fatigue experiment results in the contact of a
relatively huge blunt tip with the substrate. This
results in a low contact stiVness value.
Compressive residual stresses assist in delamination
and buckling. A coating with higher adhesion strength
and a less compressive residual stress is required for
Mechanical properties for FCA, IB, ECRCVD, and SP coatings* (hardness, elastic modulus and
fracture toughness were measured on 100 nm thick coatings; fatigue life and critical load during
scratch testing were measured on 20 nm thick coatings) and for single crystal silicon substrate
Coating
Hardness
(Li and
Bhushan,
1999a),
GPa
Elastic
modulus
(Li and
Bhushan,
1999a), GPa
Fracture
toughness
(Li and Bhushan,
1999d), MPa m1/2
Fatigue life
(Li and
Bhushan,
2002b), 104
Critical load
during scratch test
(Li and
Bhushan,
1999a), mN
FCA
IB
ECRCVD
SP
Silicon
24
19
22
15
11
280
140
180
140
220
118
43
64
28
075
20
08
12
02
38
23
53
11
06
* FCA: filtered cathodic arc; IB: ion beam; ECRCVD: electron cyclotron resonancechemical vapour deposited; SP: sputtered.
Fatigue life Nf was obtained at mean load of 10 mN and load amplitude of 8 mN.
Vol. 48
No. 3
Bhushan and Li
a higher fatigue life. Interfacial defects should be

avoided in the coating deposition process. It is known
that the ring-like crack formation occurs in the coating. The formation of fatigue cracks in the coating
depends upon the hardness and fracture toughness.
It is more diYcult to form and propagate cracks
in coatings having higher strength and fracture
toughness.
It is now accepted that long fatigue life in a coating
substrate system almost always involves living with
cracks, that the threshold or limit condition is associated with the non-propagation of existing cracks or
defects, even though these cracks may be undetectable
(Suresh, 1991). For all coatings studied here, at 4 mN,
contact stiVness does not change much. This indicates
that delamination and buckling did not occur within
the number of cycles tested in this study. This is
probably because the indentation induced compressive stress was not high enough to allow the cracks to
propagate and link up under the indenter or the sum
of indentation compressive stress s and residual stress
i
s did not exceed the critical buckling
stress s .
r
b
Figure 45 and Table 5 summarise the hardness,
elastic modulus, fracture toughness, fatigue life, and
critical load during scratch testing of all coatings
studied and of single crystal silicon substrate. A good
correlation exists between fatigue life and other
mechanical properties. Higher mechanical properties
result in a longer fatigue life. The mechanical properties of DLC coatings are controlled by the sp3/sp2
ratio. The sp3 bonded carbon exhibits the outstanding
properties of diamond (Bhushan, 1999c). A higher
deposition kinetic energy will result in a larger fraction of sp3 bonded carbon in an amorphous network.
Thus, the higher kinetic energy for the FCA could be
responsible for its better carbon structure and higher
mechanical properties (Bhushan, 1999c). Higher
adhesion strength between the FCA coating and
substrate makes the FCA coating more diYcult to
delaminate from the substrate.
The load cycles used in CSM have been successfully
used to perform nanoscale fatigue tests on ultrathin
coatings and nanostructures. Such tests hold promise
for applications in magnetic storage and MEMS
devices. Future development of the CSM technique
requires the measurement of contact stiVness over a
wide frequency range as well as a precise calibration
of thermal drift during the hold segment.
Summary
In this paper, the nanoindentation technique and
its methodologies are reviewed and discussed.
Nanoindentation can be used to measure hardness,
elastic modulus, lmsubstrate adhesion, residual
stresses, time dependent creep and relaxation properties, fracture toughness, and fatigue of surface layers
of bulk materials and thin coatings. The lateral force
sensor attached to the nanoindenter and motorised
sample stage allow nanoscratch and friction and wear
tests. The recently developed continuous stiVness
measurement (CSM) technique oVers a signi cant
improvement in nanoindentation testing. The CSM
technique in situ probes the mechanical property
changes during indentation, and provides useful
45
161
Bar charts summarising data of various

coatings and single crystal silicon substrate:
hardness, elastic modulus and fracture
toughness were measured on 100 nm thick
coatings, and fatigue life and critical load
during scratch testing were measured on
20 nm thick coatings (Li and Bhushan, 2002b)
information for layered materials and non-homogeneous (such as graded) composites. The CSM
indentation creep tests can detect creep displacement
and stress relaxation at small volumes. Load cycles
used in the CSM can be used to perform nanoscale
fatigue tests.
Many theoretical and technical issues in the indentation process are still, to a large extent, unresolved.
Indentation at an indentation depth less than 20 nm
requires sharp indenters with a tip radius less than
20 nm and the use of ultralow loads as well as
atomistic modelling. The contribution of friction force
needs to be considered as indentation load reduces
down to the order of nanonewton. The fracture
toughness measurement technique is expected to be
improved by the advances of nanoscale fracture mechanics and innovations regarding indenter geometry.
Vol. 48
No. 3
162
Bhushan and Li
Future development of the CSM technique requires

the measurement of contact stiVness over a wide
frequency range and the development of procedures
for viscoelastic materials that take into account viscoelastic deformation. With the rapid development of
instruments and analytical procedures, additional
material properties will be measured or estimated
using nanoindentation in the near future.
References
ahn, j., mittal, k. l. and macqueen, r. h. (1978): in Adhesion
measurement of thin lms, thick lms, and bulk coatings, (ed.
K. L. Mittal), STP 640, 134157; Philadelphia, PA, ASTM.
alba, s., loubet, j. l. and vovelle, l. (1993): J. Adhesion Sci.
T echnol., 7, 131140.
alekhin, v. p., berlin, g. s., isaev, a. v., kalei, g. n., merkulov,
v. a., skvortsov, v. n., ternovskii, a. p., krushchov, m. m.,
shnyrev, g. d. and shorshorov, m. kh. (1972): Zavod. L ab.,
38, 619621.
antis, g. r., chantikul, p., lawn, b. r. and marshall, d. b.
(1981): J. Am. Ceram. Soc., 64, 533538.
asme (1979): Standard test method for microhardness of materials,
ASME designation: E38473, 359379.
atkins, a. g., silverio, a. and tabor, d. (1966): J. Inst. Met., 94, 369.
baker, s. p., ross c. a., townsend, p. h., volkert, c. a. and
brgesen, p. (eds.) (1994): Thin lms: stresses and mechanical
properties V, MRS Symp. Proc. 356; Pittsburgh, PA, Materials
Research Society.
baker, s. p., cook, r. f., corcoran, s. g. and moody, n. r. (eds.)
(2000): Fundamentals of nanoindentation and nanotribology
II, MRS Symp. Proc. 649; Pittsburgh, PA, Materials Research
Society.
beake, b. d., goodes, s. r. and smith, j. f. (2001): Surf. Eng.,
17, 187192.
bell, t. j., field, j. s. and swain, m. v. (1992): T hin Solid Films,
220, 289294.
benjamin, p. and weaver, c. (1960): Proc. R. Soc. (L ondon) A,
254, 163176.
berkovich, e. s. (1951): Ind. Diamond Rev., 11, 129132.
bhattacharya, a. k. and nix, w. d. (1988a): Int. J. Solids Struct.,
24, 881891.
bhattacharya, a. k. and nix, w. d. (1988b): Int. J. Solids Struct.,
24, 12871298.
bhushan, b. (1987): in Testing of metallic and inorganic coatings,
(ed. W. B. Harding and G. A. DiBari), STP 947, 310319;
Philadelphia, PA, ASTM.
bhushan, b. (1996): Tribology and mechanics of magnetic storage
devices, 2nd edn; New York, Springer-Verlag.
bhushan, b. (1999a): Principles and applications of tribology;
New York, Wiley.
bhushan, b. (1999b): Handbook of micro/nanotribology, 2nd edn;
Boca Raton, FA, CRC Press.
bhushan, b. (1999c): Diam. Relat. Mater., 8, 19852015.
bhushan, b. (2001): Modern tribology handbook, Vols. 1 and 2;
Boca Raton, FA, CRC Press.
bhushan, b. (2002): Introduction to Tribology; New York, Wiley.
bhushan, b. and doerner, m. f. (1989): J. T ribology (T rans.
ASME ), 111, 452458.
bhushan, b. and gupta, b. k. (1995): Adv. Inf. Storage Syst.,
6, 193208.
bhushan, b. and gupta, b. k. (1997): Handbook of tribology:
materials, coatings and surface treatments; 1991, New York,
McGraw-Hill/reprint edn: 1997, Malabar, FA, Krieger
Publishing Co.
bhushan, b. and koinkar, v. n. (1994): Appl. Phys. L ett., 64,
16531655.
bhushan, b. and li, x. (1997): J. Mater. Res., 12, 5463.
bhushan, b. and patton, s. t. (1996): J. Appl. Phys., 79, 59165918.
bhushan, b., landesman, a. l., shack, r. v., vukobratovich, d.
and walters, v. s. (1985): IBM T ech. Disclos. Bull., 28,
29752976.
bhushan, b., williams, v. s. and shack, r. v. (1988): J. T ribology
(T rans. ASME), 110, 563571.
bhushan, b., kellock, a. j., cho, n. h. and ager, j. w. (1992):
J. Mater. Res., 7, 404410.
Vol. 48
No. 3
bhushan, b., gupta, b. k. and azarian, m. h. (1995): Wear,

181183, 743758.
bhushan, b., chyung, k. and miller, r. a. (1996a): Adv. Inf.
Storage Syst., 7, 316.
bhushan, b., kulkarni, a. v., bonin, w. and wyrobek, j. t.
(1996b): Philos. Mag., 74, 11171128.
bhushan, b., theunissen, g. s. a. m. and li, x. (1997): T hin Solid
Films, 311, 6780.
blau, p. j. and lawn, b. r. (eds.) (1986): Microindentation techniques in materials science and engineering, STP 889;
blau, p. j., oliver, w. c. and snead, l. (1997): T ribol. Int.,
30, 483490.
bolshakov, a, oliver, w. c. and pharr, g. m. (1996): J. Mater.
Res., 11, 760768.
bourcier, r. j., myers, s. m. and polonis, d. h. (1990): Nucl.
Instrum. Methods Phys. Res. B, 44, 278288.
bourcier, r. j., follstaedt, d. m., dugger, m. t. and myers, s. m.
(1991): Nucl. Instrum. Methods Phys. Res. B, 59/60, 905908.
bravman, j. c., nix, w. d., barnett, d. m. and smith, d. a. (eds.)
(1989): Thin lms: stresses and mechanical properties, MRS
Symp. Proc. 130; Pittsburgh, PA, Materials Research Society.
buckle, h. (1973): in The science of hardness testing and its
research applications, (ed. J. W. Westbrook and H. Conrad),
453491; Metals Park, OH, ASM.
bull, s. j. and rickerby, d. s. (1990): Surf. Coat. T echnol.,
42, 149164.
bulychev, s. i., alekhin, v. p., shorshorov, m. kh., ternovskii,
a. p. and shnyrev, g. d. (1975): Zavod L ab., 41, 9.
bulychev, s. i., alekhin, v. p. and shorshorov, m. kh. (1979):
Fizika Khim. Obrab. Mater., (5).
burnett, p. j. and rickerby, d. s. (1987a): T hin Solid Films,
154, 403416.
burnett, p. j. and rickerby, d. s. (1987b): T hin Solid Films,
148, 4150.
burnett, p. j. and rickerby, d. s. (1987c): T hin Solid Films,
148, 5165.
callahan, d. l. and morris, j. c. (1992): J. Mater. Res., 7,
16141617.
cammarata, r. c., nastasi, m. a., busso, e. p. and oliver, w. c.
(eds.) (1997): Thin lms: stresses and mechanical properties VII,
MRS Symp. Proc. 505; Pittsburgh, PA, Materials Research
Society.
campbell, d. s. (1970): in Handbook of thin lm technology, (ed.
L. I. Maissel and R. Glang), chap. 12; New York, McGraw-Hill.
chantikul, p., anstis, g. r., lawn, b. r. and marshall, d. b.
(1981): J. Am. Ceram. Soc., 64, 539543.
cheng, w., ling, e. and finnie, i. (1990): J. Am. Ceram. Soc.,
73, 580586.
chiang, s. s., marshall, d. b. and evans, a. g. (1981): in Surfaces
and interfaces in ceramics and ceramicmetal systems, (ed.
J. Pask and A. G. Evans), 603612; New York, Plenum.
chiang, s. s., marshall, d. b. and evans, a. g. (1982): J. Appl.
Phys., 53, 298311.
cho, n. h., krishnan, k. m., veirs, d. k., rubin, m. b., hopper,
c. b., bhushan, b. and bogy, d. b. (1990): J. Mater. Res., 5,
25432554.
choi, s. r. and salem, j. a. (1993): J. Mater. Res., 8, 32103217.
chu, s. n. g. and li, j. c. m. (1977): J. Mater. Sci., 12, 22002208.
chu, s. n. g. and li, j. c. m. (1979): Mater. Sci. Eng., 39, 110.
chu, s. n. g. and li, j. c. m. (1980): Mater. Sci. Eng., 45, 167171.
cook, r. f. and pharr, g. m. (1990): J. Am. Ceram. Soc., 73, 787817.
dao, m., chollacoop, n., van vliet, k. j., venkatesh, t. a. and
suresh, s. (2001): Acta Mater., 49, 38993918.
de boer, m. p., huang, h., nelson, j. c., jiang, z. p. and gerberich,
w. w. (1993): in Thin lms: stresses and mechanical properties
IV, (ed. P. H. Townsend et al.), MRS Symp. Proc. 308, 647652;
Pittsburgh, PA, Materials Research Society.
doerner, m. f. and nix, w. d. (1986): J. Mater. Res., 1, 601609.
doerner, m. f. gardner, d. s. and nix, w. d. (1986): J. Mater. Res.,
1, 845851.
doerner, m. f., oliver, w. c., pharr, g. m. and brotzen, f. r.
(eds.) (1990): Thin lms: stresses and mechanical properties II,
Society.
drory, m. d., bogy, d. b., donley, m. s. and field, j. e. (eds.) (1995):
Mechanical behavior of diamond and other forms of carbon,
Society.
Bhushan and Li
evans, a. g. and charles, e. a. (1976): J. Am. Ceram. Soc.,

59, 371372.
evans, a. g. and hutchinson, j. w. (1984): Int. J. Solids Struct.,
20, 455466.
fabes, b. d., oliver, w. c., mckee, r. a. and walker, f. j. (1992):
J. Mater. Res., 7, 30563064.
fischer-cripps, a. c. (2002): Nanoindentation; New York,
Springer-Verlag.
fleck, n. a., muller, g. m., ashby, m. f. and hutchinson, j. w.
(1994): Acta Metall. Mater., 42, 475487.
gane, n. and cox, j. m. (1970): Philos. Mag., 22, 881891.
gerberich, w. w., gao, h., sundgren, j.-e. and baker, s. p. (eds.)
(1996): Thin lms: stresses and mechanical properties VI, MRS
gissler, w., haupt, j., crabb, t. a., gibson, p. n. and rickerby,
d. g. (1991): Mater. Sci. Eng. A, 139, 284289.
goken, m. and kempf, m. (2001): Z. Metallkd., 92, 10611067.
greene, j. e., woodhouse, j. and pestes, m. (1974): Rev. Sci.
Instrum., 45, 747749.
gupta, b. k. and bhushan, b. (1994): Surf. Coat. T echnol., 68/69,
564570.
gupta, b. k. and bhushan, b. (1995a): T hin Solid Films, 270,
391398.
gupta, b. k. and bhushan, b. (1995b): Wear, 190, 110122.
gupta, b. k., chevallier, j. and bhushan, b. (1993): J. T ribology
(T rans. ASME), 115, 392399.
gupta, b. k., bhushan, b. and chevallier, j. (1994): T ribol. T rans.,
37, 601607.
hainsworth, s. v., chandler, h. w. and page, t. f. (1996): J. Mater.
Res., 14, 22832295.
hainsworth, s. v., mcgurk, m. r. and page, t. f. (1998): Surf.
Coat. T echnol., 102, 97107.
hannula, s. p., stone, d. and li, c. y. (1985): in Electronic
packaging materials science, (ed. E. A. Giess et al.), MRS Symp.
Proc. 40, 217224; Pittsburgh, PA, Materials Research Society.
hannula, s. p., wanagel, j. and li, c. y. (1986): in The use of
small-scale specimens for testing irradiated material, (ed. W. R.
Corwin and G. E. Lucas), STP 888, 233251; Philadelphia,
PA, ASTM.
hart, e. w. and solomon, h. d. (1973): Acta Metall. 21, 195200.
hay, j. c., bolshakov, a. and pharr, g. m. (1999): J. Mater. Res.,
14, 22962305.
heavens, o. s. (1950): J. Phys. Rad., 11, 355.
henshall, j. l. and brookes, c. a. (1985): J. Mater. Sci. L ett.,
4, 783786.
hong, s., weihs, t. p., bravman, j. c. and nix, w. d. (1990):
J. Electronic Mater., 19, 903.
hooper, r. m. and brookes, c. a. (1984): J. Mater. Sci., 19, 4057.
huang, l. y., zhao, j. w., xu, k. w. and lu, j. (2002): Diam. Relat.
Mater., 11, 14541459.
inspec (1988): Properties of silicon, EMIS Data Reviews Series
no. 4, INSPEC, The Institution of Electrical Engineers, London.
jacobson, s., jonsson, b. and sundquist, b. (1983): T hin Solid
Films, 107, 8998.
johnson, k. l. (1985): Contact mechanics; Cambridge, Cambridge
University Press.
jonsson, b. and hogmark, s. (1984): T hin Solid Films 114, 257269.
joslin, d. l. and oliver, w. c. (1990): J. Mater. Res., 5, 123126.
king, r. b. (1987): Int. J. Solids Struct., 23, 16571664.
korsunsky, a. m., mcgurk, m. r., bull, s. j. and page, t. f. (1998):
Surf. Coat. T echnol., 99, 171183.
lafontaine, w. r., yost, b., black, r. d. and li, c. (1990a): in
Thin lms: stresses and mechanical properties II, (ed. M. F.
Doerner et al.), MRS Symp. Proc. 188, 165170; Pittsburgh, PA,
Materials Research Society.
lafontaine, w. r., yost, b., black, r. d. and li, c. y. (1990b):
J. Mater. Res., 5, 21001206.
lafontaine, w. r., yost, b. and li, c. y. (1990c): J. Mater. Res.,
5, 776783.
lafontaine, w. r., paszkiet, c. a., korhonen, m. a. and li, c. y.
(1991): J. Mater. Res., 6, 20842090.
lankford, j. (1981): J. Mater. Sci., 16, 11771182.
laugier, m. (1981): T hin Solid Films, 76, 289294.
laursen, t. a. and simo, j. c. (1992): J. Mater. Res., 7, 618626.
lawn, b. r. (1993): Fracture of brittle solids, 2nd edn; Cambridge,
Cambridge University Press.
lawn, b. r. and evans, a. g. (1977): J. Mater. Sci., 12, 21952199.
lawn, b. r. and marshall, d. b. (1979): J. Am. Ceram. Soc.,
62, 347350.
163
lawn, b. r. and wilshaw, r. (1975): J. Mater. Sci., 10, 10491081.

lawn, b. r., evans, a. g. and marshall, d. b. (1980): J. Am. Ceram.
Soc., 63, 574581.
lee, e. h. and mansur, l. k. (1989): J. Mater. Res., 4, 13711378.
lee, e. h., rao, g. r. and mansur, l. k. (1992): J. Mater. Res., 7,
19001911.
lee, e. h., lee, y., oliver, w. c. and mansur, l. k. (1993): J. Mater.
Res., 8, 377387.
li, j. c. m. and chu, s. n. g. (1979): Scr. Metall., 13, 10211026.
li, w. b., henshall, j. l., hooper, r. m. and easterling, k. e.
(1991): Acta Metall. Mater., 39, 30993110.
li, x. and bhushan, b. (1998a): T hin Solid Films, 315, 214221.
li, x. and bhushan, b. (1998b): Wear, 220, 5158.
li, x. and bhushan, b. (1999a): J. Mater. Res., 14, 23282337.
li, x. and bhushan, b. (1999b): Z. Metallkd., 90, 820830.
li, x. and bhushan, b. (1999c): T hin Solid Films, 340, 210217.
li, x. and bhushan, b. (1999d): T hin Solid Films, 355356, 330336.
li, x. and bhushan, b. (2000a): Scr. Mater., 42, 929935.
li, x. and bhushan, b. (2000b): T hin Solid Films, 377378, 401406.
li, x. and bhushan, b. (2001a): J. Inf. Storage Proc. Syst., 3, 131142.
li, x. and bhushan, b. (2001b): IEEE T rans. Magn., 37, 16161619.
li, x. and bhushan, b. (2002a): Mater. Charact., 48, 1136.
li, x. and bhushan, b. (2002b): Scr. Mater., 47, 473479.
li, x. and bhushan, b. (2002c): J. Appl. Phys., 91, 83348336.
li, x., diao, d. and bhushan, b. (1997): Acta. Mater., 45, 44534461.
li, x., bhushan, b. and inoue, m. (2001): Wear, 251, 11501158.
lin, m. r., ritter, j. e., rosenfeld, l. and lardner, t. j. (1990):
J. Mater. Res., 5, 11101117.
loubet, j. l., georges, j. m., marchesini, o. and meille, g. (1984):
J. T ribology (T rans. ASME ), 106, 4348.
loubet, j. l., bauer, m., tonck, a., bec, s. and gauthier-manuel,
b. (1993): in Mechanical properties and deformation behavior
of materials having ultra- ne microstructures, (ed. M. Nastasi
et al.), 429447; Dordrecht, Kluwer Academic.
lucas, b. n. and oliver, w. c. (1992): in Thin lms: stresses and
mechanical properties III, (W. D. Nix et al.), MRS Symp. Proc.
239, 337341; Pittsburgh, PA, Materials Research Society.
lysaght, v. e. (1949): Indentation hardness testing; New York,
Reinhold.
marshall, d. b. and evans, a. g. (1984): J. Appl Phys., 15,
26322638.
marshall, d. b. and lawn, b. r. (1979): J. Mater. Sci., 14,
20012012.
marshall, d. b. and oliver, w. c. (1987): J. Am. Ceramic Soc.,
70, 542548.
mayo, m. j. and nix, w. d. (1988): Acta Metall., 36, 21832192.
mayo, m. j., siegel, r. w., narayanasamy, a. and nix, w. d. (1990):
J. Mater. Res., 5, 10731082.
mcgurk, m. r. and page, t. f. (1999): J. Mater. Res., 14, 22832295.
mchargue, c. j. (1989): Structureproperty relationships in surface-modi ed ceramics, (ed. C. J. McHargue et al.), 253273,
Dordrecht, Kluwer Academic.
mchargue, c. j., ohern, m. e. and joslin, d. l. (1990): in Thin
lms: stresses and mechanical properties II, (ed. M. F. Doerner
et al.), MRS Symp. Proc. 188, 111120; Pittsburgh, PA, Materials
Research Society.
mecholsky, j. j., tsai, y. l. and drawl, w. r. (1992): J. Appl. Phys.,
71, 48754881.
mehrotra, p. k. and quinto, d. t. (1985): J. Vac. Sci. T echnol. A,
3, 24012405.
mittal, k. l. (ed.) (1978): Adhesion measurement of thin lms,
thick lms, and bulk coatings, STP 640; Philadelphia, PA,
ASTM.
moody, n. r., gerberich, w. w., burnham, n. and baker, s. p.
(eds.) (1998): Fundamentals of nanoindentation and nanotribology, MRS Symp. Proc. 522; Pittsburgh, PA, Materials
Research Society.
mott, b. w. (1957): Microindentation hardness testing, London,
Butterworths.
mts (2001): Customer care kit: the Nano Indenter XP, MTS Nano
Instruments
Innovation
Center,
Oak
Ridge,
TN
(www .mts.com/nano/).
mulhearn, t. o. and tabor, d. (19601961): J. Inst. Met., 87, 7.
nastasi, m., hirvonen, j. p., jervis, t. r., pharr, g. m. and oliver,
w. c. (1988): J. Mater. Res., 3, 226232.
nastasi, m., parkin, d. m. and gleiter, h. (eds.) (1993): Mechanical
properties and deformation behavior of materials having ultra ne microstructures; Dordrecht, Kluwer Academic.
newey, d., wilkins, m. a. and pollock, h. m. (1982): J. Phys. E,
Sci. Instrum., 15, 119122.
Vol. 48
No. 3
164
Bhushan and Li
nix, w. d. (1989): Metall. T rans. A, 20A, 22172245.

nix, w. d., bravman, j. c., arzt, e. and freund, l. b. (eds.) (1992):
Thin lms: stresses and mechanical properties III, MRS Symp.
Proc. 239; Pittsburgh, PA, Materials Research Society.
ohern, m., parrish, r. h. and oliver, w. c. (1989): T hin Solid
Films, 181, 357363.
ohern, m. e., mchargue, c. j., white, c. w. and farlow, g. c.
(1990): Nucl. Instrum. Methods Phys. Res. B, 46, 171175.
oliver, w. c. (2001): J. Mater. Res., 16, 32023206.
oliver, w. c. and pethica, j. b. (1989): Methods for continuous
determination of the elastic stiVness of contact between two
bodies, US Patent 4,848,141, July 18.
oliver, w. c. and pharr, g. m. (1992): J. Mater. Res., 7, 15641583.
oliver, w. c., hutchings, r. and pethica, j. b. (1986): in
Microindentation techniques in materials science and engineering, (ed. by P. J. Blau and B. R. Lawn), STP 889, 90108;
oneill, h. (1967): Hardness measurement of metals and alloys;
London, Chapman and Hall.
ozkan, c. s., cammarata, r. c., freund, l. b. and gao, h. (eds.)
(2001): Thin lms: stresses and mechanical properties IX, MRS
page, t. f., oliver, w. c. and mchargue, c. j. (1992): J. Mater.
Res., 7, 450473.
palmquist, s. (1957): Jernkontorets Ann., 141, 300.
patton, s. t. and bhushan, b. (1996): Wear, 202, 99109.
perry, a. j. (1981): T hin Solid Films, 78, 7793.
perry, a. j. (1983): T hin Solid Films, 197, 167180.
pethica, j. b. and oliver, w. c. (1989): in Thin lms: stresses and
mechanical properties, (ed. J. C. Bravman et al.), MRS Symp.
Proc. 130, 1323; Pittsburgh, PA, Materials Research Society.
pethica, j. b. hutchings, r. and oliver, w. c. (1983): Philos. Mag
A, 48, 593606.
pethica, j. b., koidl, p., gobrecht, j. and schuller, c. (1985):
J. Vac. Sci. T echnol A, 3, 23912393.
pharr, g. m. (1992): in Thin lm: stresses and mechanical properties III, (ed. W. D. Nix et al.), MRS Symp. Proc. 239, 301312;
Pittsburgh, PA, Materials Research Society.
pharr, g. m. (1998): Mat. Sci. Eng. A, 253, 151159.
pharr, g. m., oliver, w. c. and clarke, d. r. (1989): Scr. Metall.,
23, 19491952.
pharr, g. m., oliver, w. c. and clarke, d. r. (1990): J. Electron.
Mater., 19, 881887.
pharr, g. m., oliver, w. c. and brotzen, f. r. (1992): J. Mater.
Res., 7, 613617.
pharr, g. m., harding, d. s. and oliver, w. c. (1993): in
Mechanical properties and deformation behavior of materials
having ultra- ne microsctructures, (ed. M. Nastasi et al.),
449461; Dordrecht, Kluwer Academic.
raman, v. and berriche, r. (1992): J. Mater. Res., 7, 627638.
randall, n. x., cristoph, r., droz, s. and juliaschmutz,
c. (1996): T hin Solid Films, 290291, 348354.
rao, g. r., lee, e. h. and mansur, l. k. (1993): Wear, 162164,
739747.
rubin, m. b., hopper, c. b., cho, n. h. and bhushan, b. (1990):
J. Mater. Res., 5, 25382542.
sargent, p. m. (1986): in Microindentation techniques in materials
science and engineering, (ed. P. J. Blau and B. R. Lawn), STP
889, 160174; Philadelphia, PA, ASTM.
scruby, c. b. (1987): J. Phys. E, Sci. Instrum., 20, 946953.
sekler, j., steinmann, p. a. and hintermann, h. e. (1988): Surf.
Coat. T echnol., 36, 519529.
shih, c. w., yang, m. and li, j. c. m. (1991): J. Mater. Res., 6,
26232628.
sneddon, i. n. (1965): Int. J. Eng. Sci., 3, 4757.
stilwell, n. a. and tabor, d. (1961): Proc. Phys. Soc., 78, 169179.
stone, d., lafontaine, w. r., alexopoulos, p. s., wu, t. w. and li,
c. y. (1988): J. Mater. Res., 3, 141147.
stone, d. s., yoder, k. b. and sproul, w. d. (1991): J. Vac. Sci.
T echnol. A, 9, 25432547.
suresh, s. (1991): Fatige of materials; Cambridge, Cambridge
University Press.
suresh, s. and giannakopoulos, a. e. (1998): Acta Mater., 46,
57555767.
swadener, j. g., george, e. p. and pharr, g. m. (2002): J. Mech.
Phys. Solids, 50, 681694.
Vol. 48
No. 3
swain, m. v., hagan, j. t. and field, j. e. (1977): J. Mater. Sci., 12,

19141917.
syed asif, s. a. and pethica, j. b. (1997): in Thin lms: stresses
and mechanical properties VI, (ed. W. W. Gerberich et al.),
MRS Symp. Proc. 436, 201206; Pittsburgh, PA, Materials
Research Society.
tabor, d. (1951): The hardness of metals; Oxford, Clarendon Press.
tabor, d. (1970): Rev. Phys. T echnol., 1, 145179.
tangena, a. g. and hurkx, g. a. m. (1986): J. Eng. Mater. T echnol.,
108, 230232.
ternovskii, a. p., alekhin, v. p., shorshorov, m. kh., khrushchov,
m. m. and skvortsov, v. n. (1973): Zavod. L ab., 39, 16201624.
townsend, p. h., weihs, t. p., sanchez, j. e. and borgesen, p. (eds.)
(1993): Thin lms: stresses and mechanical properties IV, MRS
tsui, t. y. and pharr g. m. (1999): J. Mater. Res., 14, 292301.
tsui, t. y., oliver, w. c. and pharr, g. m. (1996): J. Mater. Res.,
11, 752759.
tsukamoto, y., yamaguchi, h. and yanagisawa, m. (1987): T hin
Solid Films, 154, 171181.
vaidyanathan, r., dao, m., ravichandran, g. and suresh,
s. (2001): Acta Mater., 49, 37813789.
valli, j. (1986): J. Vac. Sci. T echnol., A4, 30073014.
vanlandingham, m. r., villarrubia, j. s., guthrie, w. f. and
meyers, g. f. (2001): Macromol. Symp., 167, 1543.
venkataraman, s., kohlstedt, d. l. and gerberich, w. w. (1992):
J. Mater. Res., 1, 11261132.
vinci, r. p. and vlassak, j. j. (1996): Ann. Rev. Mater. Sci.,
26, 431462.
vinci, r., kraft, o., moody, n., besser, p. and shaffer. e. (eds.)
(1999): Thin lms: stresses and mechanical properties VIII,
Society.
vitovec, f. h. (1986): in Microindentation techniques in materials
science and engineering, (ed. P. J. Blau and B. R. Lawn), STP
889, 175185; Philadelphia, PA, ASTM.
walker, w. w. (1973): in The science of hardness testing and its
research applications, (ed. J. H. Westbrook and H. Conrad),
258273; Metals Park, OH, ASM.
was, g. s. (1990): J. Mater. Res., 5, 16681683.
was, g. s. and dekoven, b. m. (1991): Nucl. Instrum. Methods Phys.
Res. B, 59/60, 802805.
was, g. s., mcintyre, m. e. and ocken, h. (1993): Proc. Int. Conf.
on Beam processing of advanced materials, (ed. J. Singh and
S. M. Copley), 489; Warrendale, PA, TMS.
weihs, t. p., hong, s., bravman, j. c. and nix, w. d. (1988):
J. Mater. Res., 3, 931942.
weihs, t. p., lawrence, c. w., derby, c. b. and pethica, j. b.
(1992): in Thin lms: stresses and mechanical properties III,
(ed. W. D. Nix et al.), MRS Symp. Proc. 239, 361370; Pittsburgh,
PA, Materials Research Society.
westbrook. j. h. (1957): Proc. AST M, 57, 873.
westbrook, j. h. and conrad, h. (eds.) (1973): The science of
hardness and its research applications; Metals Park, OH, ASM.
white, r. l., nelson, j. and gerberich, w. w. (1993): in Thin
lms: stresses and mechanical properties IV, (ed. P. H. Townsend
et al.), MRS Symp. Proc. 308, 141146; Pittsburgh, PA, Materials
Research Society.
whitehead, a. j. and page, t. f. (1992): T hin Solid Films, 220,
277283.
wierenga, p. e. and franken, a. j. j. (1984): J. Appl. Phys., 55,
42444247.
wierenga, p. e. and van der linden, j. h. m. (1986): in Tribology
and mechanics of magnetic storage systems, (ed. B. Bhushan
and N. S. Eiss), Vol. 3, 3137, SP21; Park Ridge, IL, American
Society of Lubrication Engineers.
williams, v. s., landesman, a. l., shack, r. v., vukobratovich,
d. and bhushan, b. (1988): Appl. Opt., 27, 541546.
wu, t. w. (1991): J. Mater. Res., 6, 407426.
wu, t. w. and lee, c. k. (1994): J. Mater. Res., 9, 805811.
wu, t. w., hwang, c., lo, j. and alexopoulos, p. (1988): T hin
Solid Films, 166, 299308.
wu, t. w., shull, a. l. and berriche, r. (1991): Surf. Coat.
T echnol., 47, 696709.
yanagisawa, m. and motomura, y. (1987): L ubr. Eng., 43, 5256.
yeack-scranton, c. e. (1986): IEEE T rans. Magn., 22, 10111016.

Nanomechanical Characterization of Solid Surfaces and Thin Films PDF

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Nanomechanical Characterization of Solid Surfaces and Thin Films PDF

Încărcat de

Drepturi de autor:

Formate disponibile

Nanomechanical characterisation of solid surfaces

and thin films

Depth-sensing nanoindentation measurement

are used in many applications including magnetic

International Materials Reviews 2003

Nanomechanical characterisation of solid surfaces and thin films

scratched by a sharp stylus made of hard material,

Extended load range of static indentation

ments have, however, a major weakness arising from

Nanomechanical characterisation of solid surfaces and thin films

testers available is about 10 mN. Loads on the order

force microscope, and an xyz motorised precision

General description and principle of operation

Selected details of commercial nanoindenters* according to model (and

Nano Indenter XP (MTS, Nano Instruments Innovation

International Materials Reviews 2003

Nanomechanical characterisation of solid surfaces and thin films

Two interchangeable indenter heads are available

International Materials Reviews 2003

microscopy. The position of an indentation on a

a major components: indenter head, xyz motorised

Schematics of Nano Indenter XP

Nanomechanical characterisation of solid surfaces and thin films

pyramids and have a slight oVset (051 mm) (Tabor,

The relationship h(l) is dependent on the shape of

The projected contact area A for the assumed

Another three-sided pyramidal indenter, the cube

a schematic; b indent impression; c photograph of

SEM is diYcult. Determination of tip area function

Nanomechanical characterisation of solid surfaces and thin films

Representative loaddisplacement curve during

unloading 90% of peak load; holding the indenter

Bhushan and Gupta, 1995; Gupta and Bhushan,

Nanomechanical characterisation of solid surfaces and thin films

Schematics of a stressstrain curves, b typical

covered during unloading. In the case of an elastic

a schematic representation of indentation process

Indentation data: example for real elastic

Nanomechanical characterisation of solid surfaces and thin films

relationship (assuming no elastic recovery) between

a schematic of indenter tip with non-ideal shape: contact

Effect of tip radius

tion (4a) holds for the Berkovich indenter as well.

Nanomechanical characterisation of solid surfaces and thin films

Hardness as function of indentation depth for

of indented material during the indentation would

Compliance as function of inverse of indentation

and dW /dh is the slope of the unloading curve at

Nanomechanical characterisation of solid surfaces and thin films

Plot of C Cf as function of A 1/2 for

Nanomechanical characterisation of solid surfaces and thin films

described in detail. It involves making a series of

Computed contact areas as function of contact

Nanomechanical characterisation of solid surfaces and thin films

without the need for discrete unloading cycles, and

Equations (13) and (14) may be solved simultaneously

a schematic of nanoindenter with each

parameters m, K , and D are determined by analysing

Nanomechanical characterisation of solid surfaces and thin films

deformation can be recognised in the plot of load

Schematic of cantilever microbeam deflected

Weihs et al. measured properties of free-standing

where h is the vertical de ection, W is the force