On The Evaluation of Coating-Substrate Adhesion by Indentation
On The Evaluation of Coating-Substrate Adhesion by Indentation
On The Evaluation of Coating-Substrate Adhesion by Indentation
187-204
Ed. K.L. Mittal
0VSP 2001
B. ROTHER*
MAT GmbH Dresden, Reisstr. 3, 0 - 0 1 2 5 7 Dresden, Germany
1. INTRODUCTION
The technological starting point of interest in adhesion of coatings and thin
films is usually the absence of adhesion or, in other words, coating delamina-
tion. Such failure would correlate to the zero point of empirical scales of adhe-
sion and a full-scale value could be the fracture of either the coating or the sub-
strate material. The principles of adhesion measurement (see for example [ 13)
can be classified into (i) destructive principles with the definition of loading
conditions and finding of loading parameters leading to delamination, and (ii)
non-destructive or nearly non-destructive analysis of adhesion-related re-
sponses of coating-substrate systems under defined loading conditions.
The second of the above classifications is of particular importance and it is
the aim of the present paper to show the potential of depth-sensing indentation
measurements as a nearly non-destructive principle for the evaluation of adhe-
sion of physical vapor deposited (PVD) coatings.
* Phone: +49 (0) 351 20772-17, Fax: +49 (0) 351 20772-22, E-mail: rothermmat-dresden.de
188 B. Rother
Normal Tangential
Elementary
loading
conditions
Examples
Tension Friction
1 Pi
Mixed loading
conditions,
examples pP -ae Peeling Indentation
From the elementary loading conditions shown in Figure 1 forces parallel to the
interface with resulting shear stress in the interface region are of particular practi-
cal relevance. In this connection, important effects for coatings and thin films are:
(i) internal coating stresses parallel to the interface as a result of energy deposition
during film growth as well as thermally induced stresses as a result of different
thermal expansions of coating and substrate materials, (ii) friction, scratch or in-
dentation loading of the coatings with resulting force components parallel to the
surface, and (iii) deformation of the substrate. These effects can be quantified by
stress measurement by bending analysis [2] or x-ray diffraction [3], by scratch test
with spallation pattern evaluation [4] or with acoustical emission analysis [ 5 ] ,by
depth-sensing indentation measurements with indentation depths lower than the
coating thickness [6] or higher than the coating thickness with spallation pattern
characterization at the rim of the indent [7] as well as bending tests on coated
samples [ 5 ] .
Ideal adhesion quantification should employ non-destructive measurement
methods to real samples with loading conditions close to the application of the
samples. Depth-sensing indentation measurements with indentation depths
smaller than the coating thickness are close to these ultimate requirements. They
additionally provide an access to investigations of different classes of coating-
substrate systems such as metal or ceramic coatings on metals or on plastics.
penetration depth h,, and the penetration depth at peak load h is converted into
an increase of internal energy within a limited region of the probed material.
This limited sample region is designated as virtual indenter. The concept of a
virtual indenter has recently been confirmed by finite element calculations
[IO].
190 B. Rother
(ii) The virtual indenter is characterized by its volume Viand its outer face area A i
-
as well as by the mean energy densities wv for the volume related increase
of internal energy density and w, for the face area related increase of inter-
nal energy density. For homogeneous materials the virtual indenter is as-
sumed to be in a most general case a segment of a spheroid with the projec-
tion face on the sample surface being determined by the diagonal(s) of the
indent [l 11. A schematic of such model is shown in Figure 2. During penetra-
tion of the real indenter, the virtual indenter expands against the resistance of
the surrounding material and the volume of the virtual indenter is linearly re-
lated to the volume of the sample material replaced by the real indenter.
(iii) The penetration of the real indenter is considered to be quasistationary which
permits to neglect the dynamic effects.
Starting from these assumptions, the balance of the displacement energy of the
indenter with the increase of the internal energy of the sample region designated
as virtual indenter can be written as
W, =wvVi+w, Ai
with Viand A ibeing the volume and the surface area of the virtual indenter, re-
spectively. With respect to Figure 2 and by neglecting the real indenter, the vol-
Figure 2. Schematic illustration of the indentation model with a real and a virtual indenter approxi-
mated by a cut-off of a spheroid. a, b: semi-axes of the spheroid; s: shift of the spheroid center with
respect to the surface of the sample.
On the evaluation of coating-substrate adhesion by indentation experiments 191
s
U
ume and the surface area of the virtual indenter are Vi= n y 2 d x and
-s
The two semi-axes a and b of the spheroid as well as the shift s of the spheroid
have now to be correlated to the indentation process. To that aim, for pyramidal
indenters and homogeneous sample materials, a linear relation between the two
semi-axes according the equation a = rb with the proportionality factor r is ex-
pected. Simultaneously, for the short semi-axis, a linear relation to the indentation
depth h according to b = kih with the proportionality constant ki is assumed. The
shift s of the spheroid is additionally assumed to be constant during penetration of
a pyramidal indenter into homogeneous materials.
For the energy balance of Eq. 1 we thus obtain the expression
with a polynomial expression for the volume term involving one cubic, one quad-
ratic and one linear term of the indentation depth as well as with the transcenden-
tal expression C(h)of the indentation depth for the surface area term. The expres-
sion is further simplified by the introduction of an indenter specific energy
2- 3
density e, which is e , = - w, n r k i . Equation 2 can now be written as
3
Eq. 2 can now be applied to the experimentally measurable quantities, load and
indentation depth, by the second derivative of W, in relation to h which is identi-
cal to the first derivative of the load in relation to h:
d2Wd d P
--- s -d2
=6evh+6e,-+ w,-(C(h)) (3)
dh2 dh rki dh2
The contributions of the three additive terms in Eq. 3 to dPldh can be analyzed
by the plot of experimental dPldh values versus h. This approach proves, how-
ever, not to be trivial as the values of P and h are recorded, for the DSI measure-
ments performed, at the limits of digital resolution of the available instruments.
Consequently, a combination of averaging, numerical differentiation and data
smoothing was developed which was designated as differential load and feed
(DLF) analysis. To distinguish the results of this procedure from the pure differ-
entiation dP/dh, the term "DLF plot" was introduced. To indicate the modification
of dP/dh in the scheme of the DLF analysis, the term {dP/dhJDLF is used for the y-
axis designation in the graphs.
The experimental analysis of Eq. 3 with data gained by DSI measurements with
pyramidal indenters showed a linear behavior of the DLF plots for homogeneous
bulk materials [9,12,13]. This experimental result suggests that the surface term
-
of the mean internal energy density increase w, can be neglected in relation to
wv which permits to neglect the contributions of the nonlinear term of Eq. 3
which can now, following an earlier convention [9], be simplified to
The specific energy density e,, is closely related to the hardness under load and
was, therefore, designated as hardness equivalent [9].
3.1.2. Coatings. The above considerations for homogeneous bulk materials hold,
in principle, also for coating-substrate systems. This is illustrated in Figure 3. The
expansion of the virtual indenter within the coating before reaching the interface re-
gion can be described according to Eq. 4 with the values of e,, r, kiand s being char-
acteristics of the coating material. The situation changes after the virtual indenter has
passed through the interface. Its further expansion within the coating is now limited
by the coating thickness h, whereas the expansion in the substrate follows the same
rules as for bulk materials. The volume of the virtual indenter within the coating
equals now the volume V, of a zone of a spheroid defined by the semi-axes a and b
as shown in Figure 2 as well as by the zone thickness which is identical to the coat-
ing thickness h,. The volume of the virtual indenter within the coating Vi,is thus
2
+h,s2 -h,
3
On the evaluation of coating-substrate adhesion by indentation experiments 193
real indenter
,
bulk approximation
for the expansion
of the virtual indenter
in the coating
virtual indenter
approximation for
the expansion of
the virtual indenter
in layered structures
Figure 3. Schematic illustration of the penetration of a real indenter and the formation of the virtual
indenter in a coating-substrate system.
with the indices c and s referring to coating and substrate, respectively, and the
indexed terms are analogous to the explanations in 3.1.1. Equation 6 is again a
linear function of dPldh versus h where the slope is determined only by the hard-
ness equivalent of the substrate material. It should once more be noted that Eq. 6
stands for the simultaneous expansion of the virtual indenter in both the coating
and substrate which is identical to the situation after the virtual indenter has
194 B. Rother
passed the interface between coating and substrate. The expected DLF plots of
coating-substrate systems should thus be determined by two successive linear
ranges. Starting with h from zero, the slope of the first range should correspond to
the hardness equivalent of the coating and the slope of the second range to the
hardness equivalent of the substrate. An experimental confirmation for this ex-
pectation is shown in Figure 4.
I I I I I I
400 -
300 -
200 -
coating-substrate system
100 -
- e - - uncoated substrate (HSS)
I .......... regression lines
1 I I I I I
Figure 4. Experimental DLF plot for (Ti,Al)N on hardened high speed steel (HSS) together with the
plot of the uncoated substrate material. e,: hardness equivalent, see Eqs. 2 and 2a.
causes positive vertical shift of the DLF plots [14]. This vertical shift correlates to
the value of the absolute term of the linear equation which is, with respect to Eq.
4, ev ( d r k J . In particular, if the slope of the vertically shifted DLF plots remains
constant, changes of their absolute term will, with respect to Figure 2 and the dis-
cussion in 3.1.1., correlate only to the term s/rk,. This term describes the shape
and shift of the virtual indenter independent of the indentation depth h. The above
discussed changes of the term d r k , thus support the concept of the virtual indenter
which expands against the resistance of the surrounding material with compres-
sive stress in the surrounding material being one component of the resistance. A
quantitative relation could, however, not be determined.
The crack events at the penetration of an indenter into coating-substrate sys-
tems are particularly evident with hard and brittle materials. These crack events
are related to a stress relaxation which corresponds, in the picture of the virtual
indenter, to a decrease of the internal energy density and, consequently, to a
negative slope of the DLF plots. Such negative slopes are part of negative peaks
of DLF plots. The existence of such negative peaks can be seen from the results
published for example in [ 151 where the so-called pop-ins of load-indentation
plots were correlated to crack events under the penetrating indenter. It should be
mentioned, however, that such crack events do not provide information on
whether the crack propagates along the interface or not. The correlation between
the negative peaks in the DLF plots and the crack events in multilayer coatings as
described in [ 161 provides, therefore, no access to the assessment of adhesion
between the layers.
Besides the vertical shift of DLF plots and negative peaks, another important
feature of the curves is a plateau region instead of a bending point between the
coating and the substrate dominated plot ranges as shown in Figure 4. This pla-
teau range was qualitatively discussed in terms of the resistance of the interface
against the penetration of the boundary of the virtual indenter characterized by an
indenter specific energy consumption [6]. The problems with this approach are,
however, the definition of the plateau region as well as a comparison of the results
to other adhesion assessment techniques such as scratch test [6] or Rockwell-C
indent method [ 171.
Hence summarizing, adhesion related features of the DLF curves are vertical
shifts, negative peaks and plateau ranges between the coating and substrate domi-
nated plot ranges. A schematic illustration of these features is given in Figure 5
and the evaluation principle of the energy consumption of plateau ranges as pre-
sented in [6] is illustrated in Figure 6.
196 B. Rother
I negative peak
h, h*
Indentation depth [arb. units] -
Figure 5. Schematic illustration of adhesion-related features of the DLF curves. CR: plot range
where the slope is determined by the coating material; SR: plot range where the slope is dominated
by the substrate material; h l : indentation depth of the real indenter marking the end of the coating
determined plot range CR; h2: indentation depth of the real indenter marking the beginning of the
substrate dominated plot range SR.
h, h2
Indentation depth [arb. units] *
Figure 6. Schematic illustration of the evaluation principle for the energy consumption in relation to
plateau ranges of the DLF plots: Calculation of indenter specific energy consumption by integration
over AP.CR, SR, h , , hl: see Figure 5 ; AP:difference area (= difference force) between the linearly ex-
trapolated coating range of an experimental DLF plot and the plateau range of the experimental plot.
On the evaluation of coating-substrate adhesion by indentation experiments 197
140
120
-5
CI
100
z€ 80
Y
60
0
0 20 40 60 80 100 120 140 160
Indentation depth [nm]
200 l ' i ' l ' l ' l ' l ' l ' i
/
1-1
180 ,/
n 0
E 160 - 0
0
/
L
140
Y
120 -
d
A
= 100 -
T1 after ion cleaning
n
80 - - - - without ion cleaning - ?
60' I I I * I I
0 20 40 60 80 100 120 140
Indentation depth [nm]
Figure 7.Typical DLF plots for the investigated layer arrangements of inkjet systems. Indenter used
for DSI measurements: Knoop pyramid.
On the evaluatioiz of coating-substrate adhesion by indentation experiments 199
3.2.2. Hard Coatings for Wear Protection. A large field for the application of
PVD and CVD hard coatings is the wear protection of tools and machinery com-
ponents. Typical coating-substrate systems for this purpose consist of coating
materials such as TiN. (Ti,Al)N, or, generally, transition metal nitrides or carbides
and substrate materials such as high speed steel or cemented carbide. The adhe-
sion of the coatings is of substantial importance to the performance of the sys-
tems. Process steps of the deposition with direct relation to coating adhesion are
wet cleaning and ion cleaning prior to deposition as well as ion mixing between
the coating and the substrate by ion bombardment during interface formation.
The efficiency of wet cleaning procedures (combination of hydrocarbon
cleaning with electrochemically enhanced degreasing) was analyzed in [ 171 for
magnetron sputtered (Ti,Zr)N coatings deposited on hardened high speed steel.
Prior to deposition, the samples were exposed to three different wet cleaning pro-
cedures, and the deposition of the (Ti,Zr)N coatings was performed in one run
with identical parameter sets. The different wet cleaning procedures could thus be
expected to affect mainly coating adhesion with only minor influence on other
properties of the coating-substrate systems. The expected effects were analyzed
by DSI measurements with subsequent DLF analysis as well as by Rockwell-C
indentation tests with a semiquantitative evaluation of the delamination patterns
of the coating at the rim of the indent. Typical DLF plots are shown in Figure 8
where the dotted and dashed auxiliary lines shown in the bottom diagram are in
relation to the scheme given in Figure 6 and indicate the area over which integra-
tion is performed to calculate the specific energy consumption for the penetration
of the virtual indenter boundary through the interface. A summary of these results
together with the evaluation of the Rockwell-C indents is given in Table 1.
Table 1.
Adhesion evaluation for (Ti,Zr)N coatings on high speed steel of different wet cleaning stages [16].
Calculation results of a specific penetration energy of the virtual indenter through the interface and
semiqualitative evaluation of Rockwell-C indents
degreasing. B: As for A, without neutralization between cathodic and anodic degreasing. C: 3-stage
procedure of ultrasonically enhanced hydrocarbon cleaning.
*) Classification numbers according to [ 7 ] .
200 B. Rother
Original plots
-E 600 1
I
$
Y
400
2 200
--- - - - .- C
I 1 I , I I
- I I I
- ------- .
I
Regression lines (CR)
I I I
Besides wet cleaning, ion cleaning is a further process step in PVD which is di-
rectly linked to adhesion. The effects of ion cleaning time have been analyzed in
connection with sputter deposited TiN on hardened high speed steel. For this pur-
pose, the samples were prepared with identical surface finish and wet cleaning
procedures. Two samples were placed in one deposition run and each of them was
On the evaluation of muting-substrate adhesion by indentation experiments 201
exposed for different durations to an Ar ion etch at -300 V substrate bias. The
samples were then identically coated by magnetron sputtering [ 181. The effects of
the different ion cleanings on adhesion were evaluated by DSI measurements and
DLF analysis. The results of these investigations are shown in Figure 9. The char-
acteristic result for all investigated samples was the slight but reproducible verti-
cal shift of those plots with lower cleaning time.
I I I I I
-E
-
-
I
+
z -
U
E
-
-
11 min ion cleaning -
1 min ion cleaning
-0 '
- _ e - - ~
I I I
A powerful tool for the increase of adhesion between PVD coatings and metal-
lic as well as ceramic substrates is the high energy ion bombardment during the
interface formation, Ion energies in the range of some keV are thereby suitable to
initiate implantation and recoil implantation of the film forming particles into the
substrate material which increases generally the potential for increased adhesion
[19]. This process step was again investigated with respect to its effect on DSI
measurements. For this purpose, different high speed steel samples were arranged
in pairs in the deposition chamber and only one of them was exposed to an Ar ion
bombardment during the first minute of TiN deposition by magnetron sputtering.
Before and after the high energy ion bombardment the samples were treated iden-
tically. A typical result for a pair of samples coated under identical parameters ex-
cluding the high energy ion bombardment during interface formation is shown in
Figure 10. A typical result of these investigations was again a vertical shift in the
DLF plot for the sample which was deposited without ion bombardment during
the formation of the interface.
202 B. Rother
250
n
E
1
200
t
-E 150
LL
.----
Regression lines
for the substrate material
3.3. Discussion
The results presented here are a summary of the work on the investigation of em-
pirical correlations between adhesion-relevant PVD process steps and features of
the response of coating-substrate systems to indentation experiments. This re-
sponse was analyzed by an energetic interpretation of the indentation process with
the first derivative of the load in relation to the indentation depth plotted versus
the indentation depth. The plot features particularly considered in relation to ad-
hesion were (i) vertical shifts which could be expected to correlate with internal
stresses, (ii) negative peaks which are known from the literature to correlate to
crack events, and (iii) a plateau region between the DLF plot ranges dominated by
the coating and by the substrate responses. An illustration of these features was
given in Figure 5 .
The dominating effect of the above features was the vertical shift of the DLF
plots. The positive shifts thereby correlate with an expected reduced adhesion
between the coating and underlayer. This effect is very important for hardness
measurements on coatings as it is in contradiction to the generally acknowledged
Buckle-rule which claims that the effect of the substrate is relevant to hardness
values for indentation depths larger than 1/10 of the coating thickness [20]. The
On the evaluation of coating-substrate adhesion by indentation experiments 203
results illustrated in Figures 9 and 10 and particularly in Figure 7 show that the ef-
fects of the interface are already evident at the beginning of the penetration proc-
ess. Compared to conventional hardness calculations as load per area, these posi-
tive shifts are identical to higher hardness values.
At the present state, the vertical shift of the DLF plots as an effect of adhesion
can be discussed in relation to compressive stress in the coatings. This correlates
to earlier results with TIN coatings deposited with different bias voltages gener-
ating different internal stresses which were determined independently from the
DSI measurements [ 141. The correlation between internal coating stress and adhe-
sion follows from a qualitative interpretation of ion enhanced film growth where
the ions are generally assumed to generate internal stress in the growing film [21].
This stress could be balanced by the substrate, and the transformation of the
coating stress to the substrate is the more efficient the better the interface perfec-
tion is. A reduced interface perfection thus leads to accumulation of internal stress
in the coating which again correlates with the origin of the vertical shift of the
DLF plots as discussed above.
The consideration of the negative peaks in the DLF plots as crack events was
introduced above as a further feature for an evaluation of adhesion effects. These
negative peaks do not, however, distinguish between the crack propagation along
or perpendicular to the interface. The negative peaks in the DLF plots are thus a
more comprehensive criterion of coating-substrate systems. Therefore, in the pre-
sent paper it was excluded from an extended discussion.
The plateau ranges in the DLF plots of coating-substrate systems was the feature
of the plots which was first investigated for quantitative correlations with interface
effects [6]. The reliability of the suggested quantification procedure depends, how-
ever, on the reliable determination of the integration boundaries and this is not fully
solved yet [22]. The application of plateau criterion to the quantitative evaluation of
coating adhesion is, therefore, the subject of further investigations.
4. CONCLUSIONS
The results presented illustrate the state of adhesion assessment by depth sensing
indentation measurements with an energetic evaluation of the load-indentation
data. It was shown that the results were particularly suitable for comparative as-
sessment of interface effects. The sensitivity of the approach to interface effects is
dependent on the nature and intensity of the effects as well as of the material
combinations considered.
Effective assessments of adhesion were demonstrated for multilayer arrange-
ments of inkjet systems. Application of the approach to hard coating evaluations
was also demonstrated. At this stage, the approach can already be considered as a
useful tool for adhesion assessments. Further developments of the approach re-
quire further understanding of the model of the virtual indenter combined with
suitable experimental investigations.
204 B. Rother
REFERENCES
1 . K.L. Mittal (Ed.), Adhesion Measurement of Films and Coatings, VSP, Utrechr, The Nether-
lands (1995).
2. C.N. Kouyurndjiev, I.V. Ivanov and R.R. Tanov, Surface Coatings Technol., 113, 113-119
(1999).
3. R. Wiedemann, H. Oettel and M. Jerenz, Surface Coatings Technol. 97,313-321 (1997).
4. P.J. Bumett and D.S. Rickerby, Thin Solid Films 154,403 (1987).
5. H. Ollendorfand D. Schneider, Surface Coatings Technol. 113,86-102 (1999).
6. B. Rother and D.A. Dietrich, Thin Solid Films, 250, 181-186 (1994).
7. Association of German Engineers, Coating (CVD, PVD) of Cold Forging Tools, VDI Guideline
VDI 3198, Beuth Verlag, Berlin 1992 (in German).
8. W.C. Oliver and G.M. Pharr, J. Mater. Res., 7, 1564-1583 (1992).
9. B. Rother and D.A. Dietrich, Phys. Stat. Sol. 142,389-407 (1994).
10. N. Schwarzer, F. Richter and G. Hecht, Surface Coatings Technol. 114,292-304 (1999).
1 1. J.M. Olaf, PhD Thesis, Albert-Ludwigs-Universitat, Freiburg 1992 (in German).
12. B. Rother, A. Steiner, D.A. Dietrich, J. Haupt and W. Gissler, J. Mater. Res. 13,2071-2076 (1998).
13. B. Rother, J. Mater. Sci., 30, 5394-5398 (1995).
14. B. Rother, Materialwiss. Werkstofftech., 26,477-482 (1995).
15. T.F. Page and S.V. Hainsworth, Surface Coatings Technol., 61,201 -208 (1 993).
16. M.K. Kazmanli, B. Rother, M. Urgen and C. Mitterer, Surface Coatings Technol., 107, 65-75
(1998).
17. B. Rother, L.A. Donohue and H. Kappl, Surface Coatings Technol. 82,214-217 (1996).
18. B. Rother in B. Michel and T. Winkler (Eds.): Proceedings of the International Conference on
Micrornaterials, pp. 214-217, ddp Goldenbogen, Dresden 1997 (ISBN 3-932434-05-6).
19. G.K. Wolf, P. Engel, D. Heyden, E. Vera and F. Stippich, Materialwiss. Werkstofftech. 29,518-524
(1998).
20. H. Buckle, Microhardness Measurements and Their Application, Berliner Union Verlag, Stuttgart
(1965) (in German).
21. H. Oettel and R. Wiedemann, Surface Coatings Technol., 76/77,265-273 (1995).
22. B. Rother and M.K. Kazrnanli, Surface Coatings Technol. 99, 31 1-318 (1998).