Professional Documents
Culture Documents
Application of Nanoindentation Technique For Investigation of Elasto-Plasticproperties of The Selected Thin Film Materials PDF
Application of Nanoindentation Technique For Investigation of Elasto-Plasticproperties of The Selected Thin Film Materials PDF
Microelectronics Reliability
journal homepage: www.elsevier.com/locate/microrel
a r t i c l e i n f o a b s t r a c t
Article history: Nanoindentation is one of the most known method for investigating the properties of thin films. The
Received 23 June 2012 materials can be assessed by means of elastic mechanical properties (hardness and Young’s modulus).
Received in revised form 7 October 2012 However, the author’s research work shows that it is possible to obtain the elastic as well as the plastic
Accepted 23 October 2012
material behavior of the investigated thin layer. It can be done by using the nanoindentation experiment
Available online 24 November 2012
and the numerical simulations.
This paper focuses then on investigation of thin metal layers by nanoindentation with a support of
numerical methods, such as finite element method and numerical optimization processes. Additionally,
the 3-level, full factorial design of experiment (DOE) process was applied. In order to carry out such
experiment 27 samples were prepared and taken into account: three different materials with three dif-
ferent thickness’s values sputtered on three different substrates. The results were then processed by the
numerical methods in order to achieve more information about the materials – mainly the plastic
behavior.
Ó 2012 Elsevier Ltd. All rights reserved.
0026-2714/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.microrel.2012.10.009
444 A. Wymysłowski, Ł. Dowhań / Microelectronics Reliability 53 (2013) 443–451
2. Design of experiments
Fig. 1. Illustration of nanoindentation test with the Berkovivh tip (left) and the
remained print (right). The first step of this research was DoE analysis based on the 3-
level full factorial scheme. The selected factors were: thin film
material, thickness of this film and the substrate. For each factor
to a variety of tests and analytical definition of hardness value. The three levels were defined (Table 1).
indenter that is the most suitable for thin layer assessment is the The designed DOE process consisted therefore of 27 experi-
indenting test with the so-called Berkovich tip. It is a diamond ments and such amount of samples was prepared using the mag-
three-sided pyramidal indenter shape (Fig. 1). This kind of indenter netron sputtering deposition technique. Every sample was
tip assures high spatial resolutions [10]. measured 32 times with different load values. Additionally, the
The indentation area in the material depends on the indentation in situ scans of the remained print were performed. In order to ob-
depth. The simplified formula is given by: tain the stress–strains characteristics of the investigated materials
Aðhc Þ ¼ 24:56hc
2
ð1Þ only the elastic modulus was taken into account. It is a well know
fact in nanoindentation method that measured Young’s modulus of
After the test the loading/unloading curve in the function of the the thin layer is influenced by the interaction with the substrate
indentation depth is obtained (Fig. 2). material. In order to minimize the above influence a special tech-
Considering the unloading part of the curve it is possible to esti- nique was used. The above technique is based on making a number
mate the following elastic material properties: of indents with different depths and then by extrapolating the
measurement results of Young’s modulus values E0 [6]. Unfortu-
Stiffness S nately the mean value E0 , which is normally given by the nanoind-
enter after a series of measurements, due to the influence of a
dF
S ¼ ð2Þ substrate, is not a good solution. The idea of the extrapolation tech-
dh F¼F max nique is given on Fig. 3.
Before the main experiment it was decided to measure with the
where Fmax is a maximum force during loading and unloading
nanoindentation technique the elastic properties, which is Young’s
indentation process,
modulus, of the selected substrates: silicon, glass and polycarbon-
Hardness H
F max ate. The above data are included in Table 2 along with the litera-
H¼ ð3Þ ture equivalents.
Ac
The results of measurements and corresponding extrapolation
where Ac is the contact area and can be assessed on the hc depth of Young’s modulus values Em are presented in Figs. 4–6 (for better
value according to the relation (1), comparison the Y scale is the same). As it can be seen the polycar-
Effective
pffiffiffiffimodulus of elasticity Eeff bonate substrate influenced the results the most. The obtained val-
p S ues are about 5 times smaller than in case of glass and silicon
Eeff ¼ pffiffiffiffiffi ð4Þ
2 Ac substrate. This is mainly caused by the high elasticity of polycar-
bonate (about 4.5 GPa).
where Eeff takes into account displacement in both materials:
The measurements were the initial stage to the numerical anal-
specimen and indenter.
ysis. The goal of this analysis was to obtain the plastic material
Using the nanoindentation test it is only possible to obtain the
parameters of the investigated thin-film materials. Nevertheless
elastic mechanical properties such as hardness and elastic
(Young’s) modulus. However, the loading/unloading curve
Table 1
Designed 3-level full factorial experiment.
Table 2
The measured and literature values of Young’s modulus of the selected substrates
[16].
Fig. 3. Method of Young’s modulus extraction with extrapolation technique, where h is the indentation depth, tf is the thin film thickness, E0 is the measured Young’s modulus
and Em is the extrapolated Young’s modulus.
in all cases was lower than 1% of the mean value. The same the
both tip location and surface grains seemed to have small influence
on the final results. Another interesting behavior is in case of silver
deposited on silicon substrate, which shows the opposite trend
from all other tests (including silver on other substrates). Accord-
ing to the authors the above behavior could be explained by quality
of the deposited thin film by magnetron sputtering deposition
technology. Unfortunately the above was not confirmed by struc-
tural investigations or visual inspection due to e.g. grain size
distribution.
3.1. FE-model of nanoindentation test tion process in order to analyze and compare the numerical results
to the experiments. The material properties were described by
In order to obtain the results in the optimization process, the elastic and plastic models. The plastic model was chosen as the
numerical FE-model of the nanoindentation test was prepared. Ramberg–Osgood relation:
The model is fully parametric and was created in Abaqus software. nh
r r r
According to the research already done in the past, the model was e¼ þa y ð6Þ
designed as a 2D axisymmetric part (Fig. 7) [13,14]. The 2D model E E ry
though less accurate that 3D model was found out to be a good
where e – strain, r – stress, E – Young’s modulus, ry – yield stress, a,
trade off between simulation accuracy and computing time.
nh – hardening parameters. However, the so-called yield offset a(ry/
Especially in case of multi-criteria optimization. In our case, multi-
E) in Eq. (6) can be represented as constant 0.002, which stands for
criteria optimization in each instance required hundreds of simula-
the 0.2% of the nominal material strain (before loading). Therefore,
tions in order to extract elasto-plastic material properties.
the numerical loading/unloading curve was fitted to the measured
The applied FE-model consisted of 2585 elements. The elements
one by changing three material parameters: Young’s modulus, yield
were defined according to the axisymmetric model as 4-node bilin-
stress and hardening module (Table 3).
ear quadrilateral structures with reduced integration (CAX4R). The
The elastic modulus was changed from the minus 5% value of
model consisted of three areas: the diamond tip (366 elements),
the measured Young’s modulus Em to the plus value of Em (see
the thin film area (980 elements) and the substrate (1239 ele-
Fig. 4–6). The nominal value was set as the mean value. The elastic
ments). The bottom of the substrate was pinned (all degrees of
values were taken individual for each investigated sample. The
freedom set to 0) and the left edge was defined as the axis of rota-
plastic parameters were fixed as it is shown in Table 3.
tion. The key aspect in the prepared model was the contact defini-
As the output parameters, which are the optimization’s objec-
tion between the diamond tip and the thin layer. The contact was
tives three criteria were defined. Criterion 1 and Criterion 2 refer
defined as the surface-to-surface interaction with the ‘‘hard
to the loading/unloading curve in order to find the best fit between
contact’’ properties [9]. Additionally, the influence of friction coef-
the numerical and experimental data (Fig. 8). Criterion 1 is the dif-
ficient was tested and explored during the evaluation of the
ference between the contact depths after unloading. Criterion 2 is
FE-model and it did not have any significant effects on the final
the difference between the depth values at maximal load. Criterion
results.
3 refers to the pile-up effect and is defined as the difference be-
The indenter shape was designed according to the area function
tween the heights of the experimental value (in situ scanning)
in its complex form [10]:
and the numerical (Fig. 9). The goal is to find the minimal values
2 1=2 1=4 1=8 1=16
of the selected criteria.
Aðhc Þ ¼ C 0 hc þ C 1 hc þ C 2 hc þ C 3 hc þ C 4 hc þ C 5 hc The presented optimization process is carried out by the self-
ð5Þ developed software (Fig. 10). It combines the FE-models with the
implemented numerical multi-objective algorithms.
where the Ci coefficients were obtained in the calibration phase In order to extract numerically the plastic behavior of the inves-
from the nanoindentation equipment. tigated material, the FE-model has to be analyzed by the optimiza-
The material properties of the thin layer were defined as elastic tion algorithm. The goal of the optimization is to find the best fit
and plastic. The plastic behavior is described by the kinematic between the numerical and experimental data. The fitting is made
hardening model. This model can properly describe the material by changing the elastic and plastic material parameters in the
plastic behavior after applying the load and then releasing it
[12]. Such defined numerical model was applied to the optimiza- Table 3
Values of the selected input parameters.
processed FE-model. The numerical loading/unloading curve and where fi(X) are the objective function and the F(X) is the optimiza-
the pile-up are compared to their experimental equivalents which tion criterion based on the set of m objective functions; X – vector of
carry the information about the plasticity [8,11] (detailed descrip- optimization variables; ci – equality and inequality constraints
tion of the numerical extraction method in [10]). functions. The optimal solution can be formulated in a form of a
The multi-objective optimization scheme takes into consider- set, which is optimal in the Pareto sense. The set means that there
ation more than one optimization criterion and therefore the result is not only one optimal solution but a whole set of solutions, which
becomes optimal in reference to several criteria at the same time are equally acceptable from an engineering point of view and equiv-
(in this case: loading/unloading curve and pile-up effect). In case alent from mathematical point of view. The solution is optimal in
of a multi-objective optimization problems (usually complex) the Pareto sense, if there is not a better solution in reference to at
there is a need for finding a compromise between different objec- least one criterion without worsening the solution in reference to
tive functions in order to find the optimal solution in reference to all other ones. The idea of Pareto set is shown in Fig. 11. The values
all the selected objective functions. In contrast to the single objec- x and y on the right hand-side represent the input space and the
tive optimization, the multi-objective optimization can be defined f(x,y), g(x,y) ar0065 the objective functions dependent on these in-
as follows: put (x,y) values [13,14].
Fig. 15. Example stress–strains curves for the investigated silver thin films on
polycarbonate substrate.
Fig. 12. Fitted experimental and numerical response of the silver thin-film on glass
substrate.
Fig. 13. Example stress–strains curves for the investigated silver thin films on glass Fig. 16. Example stress–strains curves for the investigated copper thin films on
substrate. glass substrate.
A. Wymysłowski, Ł. Dowhań / Microelectronics Reliability 53 (2013) 443–451 449
Fig. 17. Example stress–strains curves for the investigated copper thin films on Fig. 20. Example stress–strains curves for the investigated titanium thin films on
silicon substrate. silicon substrate.
Fig. 18. Example stress–strains curves for the investigated copper thin films on
polycarbonate substrate.
Fig. 21. Example stress–strains curves for the investigated titanium thin films on
polycarbonate substrate.
Table 4
Numerical results for all investigated thin-films.
Fig. 24. Yield stress and hardening parameter versus the film thickness for Cu on
given substrates.
It can be noticed that the curves are similar and the total error
between simulation and tensile test results are around 10%. There-
fore it can be concluded that the elaborated and applied numerical
Fig. 22. Yield stress and hardening parameter versus the film thickness for Ag on
given substrates.
5. Verification experiment