Thin Solid Films 377᎐378 Ž2000.

394᎐400

Progress in determination of the area function of indenters used

for nanoindentation

K. Herrmann a,U , N.M. Jennett b , W. Wegener c , J. Meneve c , K. Hasche a , R. Seemanna

a
Physikalisch-Technische Bundesanstalt, Bundesallee 100, 38116 Braunschweig, Germany
b
National Physical Laboratory, Teddington, Middlesex TW11 0LW, UK
c
VITO-Vlaamse Instelling ¨ oor Technologisch Onderzoek, Boeretang 200, B-2400 Mol, Belgium

Abstract

It is generally accepted that the most significant source of uncertainty in nanoindentation measurement is the geometry of the
indenter tip. In this paper results are presented on the determination of indenter area functions by two methods: Ž1. direct
determination from co-ordinate measurements obtained using a scanning force microscope ŽSFM.; Ž2. indentation into reference
materials with known Young’s modulus and Poisson ratio. The direct measurement of geometry of a Berkovich indenter was
determined using a traceably calibrated Thermomicroscopes ‘Autoprobe CP’ SFM that of other mentioned indenters ŽBerkovich
and ball shaped indenters. was determined with a modified SFM Veritekt-3. The modification of the latter is that the movement
of the SFM scanner is directly calibrated by three laser interferometers fitted in the co-ordinate axes x, y and z, thus realizing
directly traceable measurements. Recently, the original arrangement of the laser-interferometers was improved such that an
Abbe error is minimized. The paper reports the results achieved with this arrangement and analyses the uncertainty of the area
functions obtained. Good knowledge of the machine compliance is a second prerequisite for the accurate determination of the
mechanical properties of thin films, in particular elastic modulus. Using the indenter area calculated from SFM measurements,
an iterative procedure of high load indentation into a stiff reference material is assessed. In the alternative method,
improvements in the determination of the area function by means of indentation into reference materials, based on the Oliver
and Pharr method ŽJ. Mater. Res. 7 Ž1992. 1564᎐1583., is reported. Due to the uncertainty of the thus determined area functions
essentially depending on the uncertainty of the machine compliance, a combined iterative approach using two reference materials
is described. The machine compliance again was determined by high load indentation into a stiff reference material Žas in the first
method.. The indenter area function, however, is derived from another reference material allowing accurate stiffness measure-
ments at shallow depths. This allows measurement of the area function near the indenter tip. Both methods are compared to
assess the accuracy of the indentation method for determining the indenter area function and the machine compliance. 䊚 2000
Elsevier Science B.V. All rights reserved.

Keywords: Indentation method; Indenter; Scanning force microscopy; Area function

1. Introduction the plastic and elastic properties of thin films, e.g.

At present, nanoindentation is the sole measuring
method capable of delivering information about both
U
Corresponding author. Tel.: q49-30-6441291; fax: q49-30-644-
1488.
E-mail address: konrad.herrmann@ptb.de ŽK. Herrmann..
hardness and elastic modulus. In order to safeguard the
reproducibility and intercomparability of instrumented
indentation results, a draft standard ISOrCD 14577
has been prepared. This standard primarily addresses
indentation into bulk materials. At the European level
a draft standard, addressing the case of indentation
into coatings, has been prepared by the Committee
CEN TC184rWG5 and is now available w2,3x. These

0040-6090r00r$ - see front matter 䊚 2000 Elsevier Science B.V. All rights reserved.
PII: S 0 0 4 0 - 6 0 9 0 Ž 0 0 . 0 1 3 6 7 - 5
 K. Herrmann et al. r Thin Solid Films 377᎐378 (2000) 394᎐400
standards contain within their scope indentations in the
nano-range which is defined as indentation depths h -
0.2 ␮m. Due to the additional peculiarities in de-
termining the mechanical properties of thin films by
indentation, in particular the need to separate film and
substrate properties from a composite indentation re-
sponse, pre-normative research into the nanoindenta-
tion of thin films is under way, for instance in the
project ‘INDICOAT’ w4x. Metrological analysis of the
nanoindentation method readily shows that the geome-
try of the indenter tip is almost always the largest
contribution to the uncertainty budget. Precise and
traceable determination of indenter geometry is, there-
fore, the key contributor to improving and validating
the accuracy of the nanoindentation test. Moreover,
the accurate representation of the indenter’s geometry
in an area function has proven worth w5x. Therefore, in
this paper, progress in the determination of the inden-
ter geometry and of indenter area functions by differ-
ent methods is presented.
2. Experimental details
2.1. Direct measurement of indenter area function by
metrological scanning force microscopy
The measurements of the indenter geometry were
made with metrological scanning force microscopes
ŽSFM.. In the case of the Berkovich measured, a ther-
momicroscope ‘Autoprobe CP’ SFM was used with a
10-␮m-range tube scanner operating in contact mode
with the Scanmaster TM closed loop scan linearizing
system enabled. Z data were derived from the indepen-
dent Z position sensor rather than the feedback volt-
ages. The Scanmaster TM Ž X and Y axes. position de-
tectors were calibrated using a traceably calibrated
artefact consisting of a 2-␮m period, one-dimensional
periodic array. The Z position detector calibration was
derived from a HeNe laser, Jamin interferometer. Abbe
Fig. 1. Abbe error free arrangement of the three laser interferometers in the SFM.
395
error was corrected by subtraction of a background
function obtained by fitting an image of a reference flat
surface. The total calibration uncertainty for the SFM
was - 1% in X᎐Y and - 1.6% in Z at the 95%
confidence level. For a further Berkovich and a spheri-
cal indenter, a modified Veritekt-3 Žmanufactured by
Carl Zeiss Jena. was used. This has three piezo-transla-
tors, one for each of the three co-ordinate axes x, y, z.
The piezo-translators each contain a capacitive mea-
surement system with feedback to control the move-
ment of the scanners. This SFM was additionally
equipped with three laser interferometers ŽMichelson
type. aligned with the x, y, z axes thus enabling the
correction of the non-linearities of the capacitive mea-
surement systems. Moreover, the measurement results
are directly traceable to the length standard w6x.
In a first set-up of the Veritekt-3, the z-interferome-
ter was arranged such that its measuring laser-beam
was aligned with the measuring point on the sample,
but the x- and y-interferometers had an Abbe offset
between the measuring laser-beam and the measuring
point on the sample. The Abbe error, resulting from
inclinations of the sample stage around the x- and
y-axis, was determined with a two-axis autocollimator
which has an uncertainty of measurement U␣ s 0.01⬙.
The uncertainty of inclination measurement meant
that a complete correction of the inclinations was not
possible and a noticeable contribution to the uncer-
tainty budget remained. Therefore, to fulfil the uncer-
tainty requirements for the calibration of indenters, a
new arrangement of the x- and y-laser interferometers
was developed, which reduced the Abbe errors to a
minimum w7,8xŽcf. Fig. 1..
The characteristic of this second arrangement is that
the measuring beams of the three laser interferometers
aligned with the x-, y- and z-axes, intersect close to the
measuring point on the sample. Thus, the contribution
of uncertainty from the calibration of the inclinations
of the measuring table is rendered negligible and may
396 K. Herrmann et al. r Thin Solid Films 377᎐378 (2000) 394᎐400

Table 1
Metrological properties of the modified SFM Veritekt-3

Parameter of the metrological properties Quantitative data

Measuring range Ž x, y, z . 70 ␮m, 15 ␮m, 15 ␮m

Resolution Ž x, y, z . 1.2 nm, 0.25 nm, 0.25 nm
Expanded uncertainty U in z-direction Us Ž1.1 . . . 1.2. nm q 10y4 h, h-height
Expanded uncertainty in x᎐y-plane Us 0.6 nm q 10y4 L, L-length
Žpreliminarily estimated.
Expanded uncertainty U for a distance within Us 1.5 nm q 10y4 L, L-length
the measuring range ŽAbbe error free
arrangement of laser interferometers .
Žpreliminarily estimated.

be safely excluded from the total uncertainty budget. mum load upon unloading was used to establish ther-
Table 1 summarizes the metrological properties of this mal drift.
modified SFM. Analysis of the indentation curves was carried out as
The Veritekt-3 expanded uncertainties in the z- follows, after the method proposed by Oliver and Pharr
direction and in the x᎐y-plane were each confirmed by w1x:
comparison measurements with step height gauges and After having corrected the measured data for dis-
one-dimensional sinusoidal gratings respectively. The placements due to the instrument compliance and ther-
gratings were compared with the reference method of mal drift, a power law is fitted to the upper 80% of the
optical diffraction measurement, whose uncertainty is unloading data:
Us Ž0.05 . . . 0.1. nm. The comparison with three grat-
ings whose period length is between 270 and 450 nm Ps aŽ h y b .
m
Ž1.
yielded differences, ⌬, between diffraction and SFM
measurements < ⌬ < F 0.11 nm w9x. Ps load, h s depth, a, b and m are fitting parameters.
For the measurement of indenters a scan is made The contact stiffness, S, is the derivative at peak
with 256 = 256 Žs 65 536. or 512 = 512 Žs 262 144. load:
equi-distant distributed scan points. At first, an Ž x, y .
scan 10 ␮m= 10 ␮m is made in order to determine the
overall indenter geometry. Then follows an Ž x, y . scan
1.5 ␮m= 1.5 ␮m in order to investigate the geometry
Ss ž dPdh / max
s am Ž h max y b .
my 1
Ž2.

near the indenter tip. Each scan is repeated at least

three times. The intercept depth, h i , is given by:

Pmax
2.2. Indentation into reference materials h i s h max y Ž3.
S

For the experimental indentation work, a NanoTest and the contact depth, h c , is calculated as:
600 instrument, manufactured by Micro Materials Ltd.,
was used. The load and depth resolutions are typically h c s h max y 0.75 Ž h max y h i . Ž4.
better than 1 ␮N and 0.1 nm, respectively.
Single crystal tungstenŽ100. and synthetic fused silica
ŽSuprasil 311. were used as reference materials to The contact area, A c , is calculated using the ap-
determine the frame compliance and indenter area propriate area function.
function, respectively. Literature elastic properties were
assumed, i.e. for amorphous SiO 2 : Es 72 GPa and Ac s F Ž hc . Ž5.
␯ s 0.17, for W: Es 410 GPa and ␯ s 0.28, and those
of the diamond indenter: Es 1141 GPa and ␯ s 0.07 The reduced modulus, Er , is given by:
w10x.
The loadingrunloading cycle used for each indent '␲ S
consisted of: load to peak load in 30 s, hold at peak Er s Ž6.
2 'Ac
load during 10 s, unload at the same rate as loading,
hold at 10% of the maximum load during 60 s. The
hold at peak load was used to allow plastic deformation which in turn is a function of the elastic properties of
to be completed, while the hold at 10% of the maxi- the specimen Žsubscript s. and the diamond indenter
 K. Herrmann et al. r Thin Solid Films 377᎐378 (2000) 394᎐400
Žsubscript i.:
1 1 y ␯s2 1 y ␯ i2
s q
Er Es Ei
A Matlab 䊛 script was developed to enable the fast,
automated, documented and reproducible implementa-
tion of this analysis procedure.
The SFM calibration of the indenter shape is de-
scribed in detail elsewhere w5,11x. In principle, the
indenter projected area was determined by counting
the pixels present at each height in a reconstructed,
three-dimensional representation of the indenter tip.
The indenter area function was obtained by fitting a
multiple knot spline to the projected area vs. depth
data set.
3. Determination of area functions
3.1. Determination of area function with SFM
measurements
From the two-dimensional data-sets using all scan
points the real area A of the indenter is calculated by
the summation of triangle areas calculated from the
scan points. The data table A i ; h i then can be used to
calculate suitable regression curves A s f Ž h..
Fig. 2 shows the area functions of a Berkovich inden-
ter:
'A s ah c q b
c
with a, b as regression coefficients.
The regression curves for the two ranges h c - 100
nm and 100 nm F h c F 1000 nm are depicted together
with the theoretical curve
Fig. 2. Area function of a Berkovich indenter used in a nanoindenta-
tion device.
397
Ž7.
Fig. 3. Relative standard deviation of repeated measurements of the
indenter area Žcf. Fig. 2. as function of the contact depth.
'A c s 24.5h 2c
' Ž9.
In this case, for h c - 100 nm the area function takes
a non-linear course mainly due to the tip rounding and
for large h, the gradient diverges from the theoretical
curve, indicating a slightly incorrect facet angle.
Fig. 3 represents the relative standard deviation of
the mean area
sA
sA re l s Ž 10 .
'n A c
Ž8. where sA is the standard deviation in each height of
repeated measurements Ž n s 8 measurements in each
of the two ranges. of the indenter area. It is assumed
here, that the pixellation uncertainty is randomly dis-
tributed and contained in the values of sA re l .
The relative standard deviation sA re l , which contains
the overall effect of sensing the indenter’s pyramid
planes with large inclination angles, already gives a
rough indication of the relative measurement uncer-
tainty. One must consider that the following relation-
ship is valid:
⌬ HrH s ⌬ ArAf sA re l Ž 11.
The uncertainty of the indenter area UA can be
estimated as follows:
UA s 2 sA2 re l q f pd
' 2
q f tr2
where f pd is the area deviation due to the scan point
distance and f tr is the area deviation due to the tip
radius of the SFM tip
As the three influences are dependent on the inden-
tation depth the uncertainty UA also is depth-depen-
dent. Take, for an example, an analysis with the limit
values: SFM tip radius rst s 20 nm, indenter tip radius
398 K. Herrmann et al. r Thin Solid Films 377᎐378 (2000) 394᎐400

r it s 100 nm and a scan point distance ⌬ xs 7 nm. This

yields the following relationships:

sA re l s 467 156hy0
c
.2677
for 10 nm F h c F 1000 nm
f pd - 1% negligible.
Ž
f tr s 14.10y 0.43 h y 0.0189 h2 for h - 6 nm

As a result the area function of the investigated

Berkovich indenter would have a relative uncertainty of
Urel s 8% at h c s 200 nm; this value decreases to Urel s
6% at h c s 1000 nm.
In nanoindentation ball-shaped indenters are gaining
increasing importance. Indenters having a tip radius of
the order of several ␮m allow purely elastic deforma-
tions, with stresses and displacements described by the
Hertzian equations. Moreover, since the average strain
in a spherical contact zone increases with depth of
penetration, it is possible to derive stress᎐strain rela-
tionships w12x. Fig. 5. Relative standard deviation of repeated radius determination
For a ball-shaped diamond indenter with 20 ␮m of a ball-shaped diamond indenter as a function of the indentation
nominal ball diameter Fig. 4 shows the ball radius, r, as depth.
function of the distance from the indenter tip, h. This
function was determined on the basis of the geometry two straight lines shows the degree of consistency of
measurement of the ball indenter with the SFM. One the SFM measuring values in both scanning ranges. As
can clearly see that from h s 100 nm where r f 8250 discussed for the investigated Berkovich indenter, the
nm the measured radius rises rapidly with increasing quantity srrel yields a rough estimation of the radius
indentation depth. At h s 300 nm a ball radius of uncertainty. In Fig. 5 it can be observed that the
r f 9000 nm is reached and the nominal value of 10 relative uncertainty at h s 20 nm is Urel s 2 srrel f 9%
␮m is only just being approached at h f 1000 nm. and decreases steadily until 0.3% at h s 1000 nm.
In Fig. 5 the relative standard deviation srrel of
repeated radius determinations by SFM is depicted. 3.2. Indentation techniques
The two straight lines present the evaluation results
due to the geometry measurements in two ranges as
Two approaches are presented. In the case that an
indicated in Section 3.2.1. The small jump between the
area function determined by SFM is available, a simple
method to determine the frame compliance is used. If
there is no area function available, a combined itera-
tive procedure to determine the frame compliance as
well as the area function is adopted. The indentation-
determined area function for the case of a Berkovich
indenter is compared with that obtained by SFM in Fig.
6. Comparison with the theoretical area function shows
clearly, that in common with most real indenters, it has
an aberrant tip geometry.

3.2.1. Determination of the frame compliance with the

Fig. 4. Radius of a ball-shaped diamond indenter with 20 ␮m nomi-
nal ball diameter as a function of the indentation depth Ždetermined
on the basis of geometry measurements with an SFM..
SFM area function
In the method proposed by Doerner and Nix w13x and
Oliver and Pharr w1x, the frame compliance is de-
termined by observation of the intercept of a plot of
'
total compliance, Ct , vs. 1r A c , where A c is the
contact area, for a series of indents at different loads.
The method follows by modelling the instrument and
specimen as two springs in series, in which case
C t s C f q Cs Ž 13 .
 K. Herrmann et al. r Thin Solid Films 377᎐378 (2000) 394᎐400 399

Tungsten was used as the reference material, be-

Fig. 6. Area functions of a Berkovich indenter determined by SFM
and indentation technique. The error bars are 2 times the standard
deviation in length Žat each side., and take into account the statistical
scatter of the indentation results as well as the uncertainty in the
frame compliance value. The theoretical Berkovich area function is
also depicted, showing that the indenter under investigation has an
aberrant tip geometry.
where Cf and Cs are the compliance of the instrument
frame and the specimen, respectively.
Since the specimen compliance is given by the in-
verse of the contact stiffness, S, Eq. Ž13. can be rewrit-
ten as:
'␲ 1
Ct s Cf q Ž 14 .
2 E A '
r c
Aluminium was used as a reference material because
of its low hardness, allowing stiff contacts which yield
data close to the intercept.
In the method presented here, only indents at a high
load Žas high as practically achievable with the instru-
ment. are used. A simple, iterative procedure is
adopted, assuming the elastic properties of the refer-
ence material. The procedure is listed below:
raw indentation data of material with known elastic
properties;
< r
cause it is an elastically isotropic, homogeneous mate-
rial, and it is less prone to handling damage compared
to aluminium. Moreover, it has a high modulus and
allows sufficient plastic deformation Žhigh A c ., yielding
better estimates of the frame compliance wright-hand
side of Eq. Ž15. is small, i.e. for stiff contactsx. Twelve
indents at a high maximum load of 100 mN were used.
It is important for the accuracy of this method that
the area function Žat distances relatively far from the
indenter tip. is well known, because the result is very
sensitive to that input. Indeed, when using the indenter
area function obtained by SFM, a frame compliance
value of 0.46" 0.032 Žone ␴ . nmrmN was found,
whereas when the theoretical Berkovich area function
was used Ž A c s 24.5h 2c ., the frame compliance became
0.37" 0.03 nmrmN, i.e. a difference of 20%. In the
method proposed by Doerner and Nix w13x and Oliver
and Pharr w1x, however, uncertainties in the area func-
tion play an even greater role since it must be known
over a range of distances from the indenter tip, since
different loads are used.
3.2.2. Determination of frame compliance and area
function with two different reference materials
This method is essentially the same as the combined
iterative method proposed by Oliver and Pharr, but the
alternative method to determine the frame compliance,
as explained in Section 3.2.1, is used. The frame com-
pliance is thus determined using WŽ100. as reference
material, while the indenter area function is derived
from indents in fused silica, allowing accurate stiffness
measurements at shallow depths, which in turn allows
measurement of the area function near the indenter
tip. Indentation loads ranged from 10 to 180 mN. The
area was calculated from the indentation data, knowing
the frame compliance, by rearranging Eq. Ž13.
␲ 1
As Ž 16 .
4 E 2 Ž C y C .2
t f

determine Ct and h c , and calculate A c ;

< The data were fitted to the following function, which
calculate the frame compliance using proved to perform adequately
<
c
'␲ 1 hyb
C f s C t y Cs s C t y Ž 15. Asa ž / Ž 17 .
2 E A ' h max
r c

<
where a, b and c are fitting parameters.
correct indentation data for the frame
After only two combined iterations, the determined
compliance;
area function was very close to that determined by
<
SFM Žmeasured in NPL., as shown in Fig. 6. The error
repeat until convergence. bars are two times the standard deviation in length Žat
400 K. Herrmann et al. r Thin Solid Films 377᎐378 (2000) 394᎐400
each side., and take into account the statistical scatter
of the indentation results as well as the uncertainty in
the frame compliance value. The theoretical Berkovich
area function is also shown, and it is clear that the area
function of the indenter under investigation deviated
significantly from the perfect geometry. For example,
the area function far from the indenter tip Ž h c ) 566
nm., as determined by SFM, was 'A s 5.3h c q 287 nm.
The frame compliance given by this combined itera-
tive procedure was 0.47" 0.047 nmrmN, which corre-
sponds to that when using the area function de-
termined by SFM, within the error margins.
These results clearly demonstrate that the combined
iterative method is able to yield reliable values of
frame compliance as well as valid estimates of the
indenter area function. The method does, however,
require considerable experimental effort. It would be
desirable, therefore, that indenters for nanoindentation
are certified for area function by SFM.
4. Conclusion
The key calibration problem at nanoindentation
measurements ᎏ the determination of the indenter tip
geometry ᎏ successfully can be solved with SFM mea-
surements. It is essential that the SFM is traceably
calibrated with laser interferometers and that Abbe
error is avoided.
When the area function of the indenter is known, an
iterative procedure by high load indentation in tung-
sten is able to yield reliable and reproducible values of
frame compliance. Using a combined iterative ap-
proach based on the Oliver and Pharr method, using
tungsten and fused silica as reference materials, the
area function obtained agreed well with that obtained
by SFM, in both cases detecting the aberrant tip
geometry of a Berkovich indenter. As a consequence,
the value of frame compliance obtained in the latter
method also corresponds to that obtained by using the
SFM area function directly.
The combined iterative approach requires consider-
able experimental effort, however. It is, therefore, re-
commended that indenters for nanoindentation are
calibrated for area function by SFM.
Acknowledgements
This work was funded by the EC Standards, Mea-
surements and Testing Programme ŽSMT. Contract
Number SMT4-CT98-2249 ‘Determination of hardness
and modulus of thin films and coatings by nanoindenta-
tion ᎏ INDICOAT’. D. Havermans at VITO is greatly
acknowledged for carrying out the indentation work.
References
w1x W.C. Oliver, G.M. Pharr, An improved technique for determin-
ing hardness and elastic modulus using load and displacement
sensing indentation experiment, J. Mater. Res. 7 Ž6. Ž1992.
1564᎐1583.
w2x ISOrCD 14577-1 to -3; Metallic material ᎏ Instrumented
indentation test for hardness and materials parameters.
w3x Draft of CEN TC184: Advanced Technical Ceramics, WG5:
Methods of Test for Ceramic Coatings, WI 121-132: Determi-
nation of Hardness and Young’s Modulus by Depth Sensing
Indentation.
w4x European Commission ᎏ Standards, Measurements and Test-
ing Programme ŽSMT.. Determination of hardness and modu-
lus of thin films and coatings by nanoindentation ᎏ INDI-
COAT Contract SMT4-CT98r2249.
w5x N.M. Jennett, J. Meneve, Depth sensing indentation of thin
hard films: a study of modulus measurement sensitivity to
indentation parameters, Proc. MRS Symp. 522 Ž1998. 239᎐244.
w6x M. Bienias, S. Gao, K. Hasche, R. Seemann, K. Thiele, A
metrological scanning force microscope used for thickness and
other topographical measurements, Appl. Phys. A 66 Ž1998.
S837᎐S842.
w7x K. Hasche, K. Herrmann, R. Seemann, H. Buchner,
¨ Traceable
measurement results from scanning probe microscopes by laser
interferometry, Proceedings of Laser Metrology 1999 5.30᎐5.37,
Florianopolis, Brazil, October, Ž1999..
w8x G. Jager,
¨ Laser-based measurement to nanometer scale accu-
racy, Proceedings of Laser Metrology 1999, 5.38᎐5.47, Flori-
anopolis, Brazil, October, Ž1999..
w9x K. Hasche, K. Herrmann, W. Mirande, ´ F. Pohlenz, R. See-
mann, Development of scanning force microscopy with sub-
nanometric capability, Proceedings of Quantitative Microscopy
2000, Semmering, Austria, 12᎐14 January 2000, PTB-F-39,
Braunschweig, Ž2000. 219᎐225.
w10x J.E. Field, R.H. Telling, The Young modulus and Poisson ratio
of diamond, Research Note, Cavendish Laboratory, Cambridge,
1999.
w11x N.M. Jennett, G. Shafirstein, S.R.J. Saunders, Comparison of
indenter tip shape measurement using a calibrated AFM and
indentation into fused silica, Hardness Testing in Theory and
Practice, VDI Berichte 1194, VDI-Verlag GmbH, Dusseldorf,
1995, pp. 201᎐210.
w12x J. Mencik, Mechanics of Components with Treated or Coated
Surfaces, Kluwer, Dordrecht, 1996, pp. 189᎐201.
w13x M.F. Doerner, W.D. Nix, A method for interpreting the data
from depth sensing indentation instruments, J. Mater. Res. 1
Ž1986. 601᎐609.