Journal of Materials Processing Technology 209 (2009) 60396047

Contents lists available at ScienceDirect

Journal of Materials Processing Technology

journal homepage: www.elsevier.com/locate/jmatprotec

Computer-based monitoring of the polishing processes using LabView

F. Klocke, O. Dambon, U. Schneider, R. Zunke , D. Waechter
Department of Process Technology, Fraunhofer Institute for Production Technology IPT, Aachen, Germany

a r t i c l e

i n f o

Keywords:
Polishing
Process monitoring
Silicon nitride
Material removal mechanisms

a b s t r a c t
Despite of an extensive scientic research regarding the polishing technologies of a variety of materials,
the mechanisms and interactions of the single components in the process are still not fully understood. In
order to facilitate the research activities in the eld of polishing the usage of computer-based data acquisition and analysis is recommended. A high data rate provides the researcher an appropriate density of
information which can be used to enhance the evaluation of stated hypothesis. This is normally based on
various inspection methods which only take place after the processing of the sample, e.g. white light interferometry and scanning electron microscopy. A solution for further elaborating the scientic insight on
the removal mechanisms and relevant interactions during the process and enhancing the process stability and reproducibility is the monitoring of the signicant chemical, mechanical and thermal indicators.
Therefore, a variety of sensors and measurement devices are installed and used to gather data during
the process duration (e.g. pH value, conductivity, polishing work, coefcient of friction). The amount of
different devices and high data rates requires a computer-based tool to realise an adequate online process monitoring. Thus, a monitoring tool based on the LabView environment was implemented which
enables the researcher to use the whole opportunities of computer-based data processing. Examples of
the functionality are given for the validation of the polishing behaviour of silicon nitride.
2009 Elsevier B.V. All rights reserved.

1. Introduction
Polishing is the most frequently used technology if high surface
qualities in terms of low roughness, minimised subsurface damage
and high form accuracies are demanded. It is an essential step in
optics manufacturing and of increased usage in die and mould making, e.g. for applications in massive forming and injection moulding.
And is one of the enabling technologies in semiconductor manufacturing where polishing is also referred to as chemicalmechanical
planarisation (CMP), and numerous research activities were conducted in order to enhance the understanding of the interactions
based on the online monitoring of the process. The progress of
researchers in the eld of CMP technology has brought the scientic insight and the on it optimisation of machinery and process
to an advanced level (Oliver, 2004). Particularly in comparison to
the polishing technologies used for glass, advanced ceramics and
steel. In these elds the process can still be described as a black box
and stability, reproducibility as well as efcient processing is not
necessarily given.
Following this argumentation it is as well of interest to monitor
the processing of glass, advanced ceramics or polishing operations for any other material. The fundamental understanding forms

a necessary basis for any convenient process optimisation. Reliable data about the polishing process is needed either for the
enhancement of the scientic insight or for improvements of the
process stability in industrial applications. It is not sufcient to
gather data manually because of high efforts and inaccuracy as
well as non-uniform timing if periodical data is needed. Additionally, the manual monitoring of a high amount of measured process
parameters is not feasible. Particularly the monitoring in industrial
applications requires an online data acquisition and processing to
realise a standardised monitoring and to ensure the reliability of
the data. Online gathered data enables the operator or an automatic control system to react on recent process disturbances for
minimising rejects. The monitoring system also realises an exoneration of the operator. A computer-based data acquisition allows
the online calculation and visualisation of various process indicators and the identication of failures with implemented warning
signals. Thereby, an assistance of the operator is given so that he
can take immediate action in case of any aberration of the process. Furthermore, the data is gathered and stored in a standardised
approach which allows the analysis by the means of statistical methodology, i.e. process analysis by design of experiments
(DoE).
2. Published approaches to monitor polishing operations

Corresponding author at: Fraunhofer IPT, Steinbachstr. 17, 52074 Aachen, Germany. Tel.: +49 241 8904 137; fax: +49 241 8904 6137.
E-mail address: [email protected] (R. Zunke).
0924-0136/$ see front matter 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.jmatprotec.2009.08.014

Several research projects were already undertaken to enable the

scientist to monitor the signicant indicators of the polishing pro-

6040

F. Klocke et al. / Journal of Materials Processing Technology 209 (2009) 60396047

cess. Especially in CMP many researchers worked on a monitoring

solution which was investigated either to collect important online
data for monitoring in real time and analysis of the process condition or to enable the detection of the endpoint or even both
objectives. The indicators where derived from different point of
views which could be mechanical, thermal, chemical or even acoustical ones. The knowledge in the eld of CMP monitoring can be
used as a basis to elaborate a solution for other polishing processes.
In order to provide a short overview of the state of the art in monitoring of CMP processes some publications are highlighted in the
following. One example for already available solutions is the CMP
Slurry Monitor by Colloidal Dynamics 2000 (Colloidal Dynamics,
2000). This system as well as the system which was published by
Hunter (2001) present the capability of monitoring pH, temperature, conductivity, particle sizes, zeta potential, slurry pressure and
concentration of particles in the polishing slurry. Matsuzaki et al.
(2002) published a CMP machine which measures additionally the
torque of the spindles and derives the uid lm thickness. Oliver
(2004) covers the eld of the so called endpointing. The purpose of
endpointing is to detect the nish of the process by various methods. One way is the measurement of motor currents which was
already published in a patent by Sandhu et al. (1991). More recent
research results by Gitis et al. (2001) are based on the direct monitoring of the coefcient of friction. Another way of endpointing
was developed by Kojima et al. (2000) and uses the monitoring of
acoustic emissions to characteristic signals for the different process
conditions. Due to the advances in CMP research the technology
cannot be any more characterised by black arts but is a stable and
predictive process (Oliver, 2004). The progress in the eld of CMP
from research to industrial usage shows clearly that the process
monitoring of polishing operations was successful and useful and
that the transfer to other polishing applications should be considered.
As a matter of fact the various polishing technologies differ
greatly in detail due to the nano-scale characteristics of process
interactions and mechanisms. Therefore, many of the approaches
in CMP are not applicable for other applications. However, some
aspects can be transferred to support the investigations of other
polishing processes where stability is still an issue or where the
knowledge is still insufcient to fully understand the process.
First results for the process monitoring of polishing of optic
components were published by Hambcker (2001). Hambcker
highlighted the importance of several indicators of the slurry condition in order to examine the interactions and specify the according
material removal mechanisms in chemo-mechanical polishing of
glass. Klocke et al. (2005a,b) used a similar approach to investigate the polishing process of silicon and calcium uoride and
emphasised the importance of a well chosen polishing system
due to the complex interactions between polishing pad, polishing
agent, polishing slurry and workpiece surface. These interactions
can result in various mechanical and chemical reactions depending on the material properties and the chemical reactivity of the
involved materials. On every important objective of polishing is
the integrity of the surface layer which can be characterised by
minimised defects and cracks or even micro dislocations. These
can be induced by the mechanical and chemical conditions in
the working gap of the process. Dambon investigated the abrasive mechanisms of steel polishing using diamond slurries and
used the specic energy derived from the determined polishing
work in relation to the amount of removed material to amplify
the differences in mechanisms of polishing steel with various
hardness. In order to remove a similar amount of material the
steel with low hardness tended to consume much more energy
as the hardened steel. Dambon (2005) was able to prove that
these observations were due to micro ploughing and a surface
near plastication of the soft steel instead of micro cutting which

dominated the processing of the hardened steels (Klocke et al.,

2005a,b).
The above mentioned investigations on process interactions
and removal mechanisms where based on manually gathered data
which featured the disadvantages mentioned in Section 1. Nevertheless, the results showed the importance of several indicators in
order to investigate the process at the highest stage. In summary it
can be stated that the computer-based monitoring of the process
enables the scientist to gain insights on the running process and
its condition. Additionally, it allows a convenient process analysis
based on a high data density and comparability of the standardised
acquired and stored data.
3. Flexible and mobile monitoring of the polishing process
In this paper, an approach to a computer-based monitoring of
the polishing process is described. The objective is the monitoring
and online calculation of mechanical, chemical, thermal indicators of the process which allows the examination of the polishing
behaviour of various materials. The approach can be used to gain an
insight on the process interactions and different inuencing parameters and to monitor the stability of any conventional polishing
process. It is intended to be used in research activities as well as
in industrial environments where specic questions on removal
mechanisms are to be claried. The approach was implemented in
the well known programming environment called LabView. This
software provides powerful capabilities of data acquisition and
online calculations as well as online data analysis. The environment can be congured to read data from any sensor or bus and
does allow the online calculation and visualisation of the acquired
and computed data. The concept of the monitoring system is based
on high mobility and exibility. Therefore, all measurement devices
were chosen to allow the adaption on different polishing machinery and can easily be used not only in laboratory environments but
also in industrial ones.
The overall objective is the elaboration of standardised data
acquisition and storage to provide a broad data basis for technological investigations. Fig. 1 explains the approach which includes three
important steps: online data acquisition and visualisation, insertion of ofine gathered data and data analysis based on principals
following statistical methodology. Thus, the continuously acquired
data is supplemented by discontinuously acquired data to provide a
holistic data basis of the polishing process and to document the process consistently. The discontinuously gathered data does include
important indicators such as weight loss of the workpiece after the
machining operation to derive the amount of removed material and
results of optical surface measurements to evaluate the inuences
of parameters on the resulting surface quality. The total amount of
collected data can easily be imported in a software solution which
allows the analysis by statistical means. The consistent usage of
computer-based data acquisition, calculation and analysis is used as
an approach to support the investigations of technological aspects
of the polishing process.
4. Design and implementation of the computer-based
monitoring system
The polishing process is inuenced by a great variety of parameters (Fig. 2). In order to establish a monitoring with reasonable
efforts the main inuencing parameters and most signicant
indicators have to be ascertained. In this context, the material
properties of the workpiece and the polishing machine are considered as xed. The main parameters are the polishing system
consisting of polishing tool, polishing agent and polishing slurry,
as well as the machining parameters, e.g. pressure and relative
velocity.

F. Klocke et al. / Journal of Materials Processing Technology 209 (2009) 60396047

6041

Fig. 1. Approach to a computer-based data acquisition and analysis of the polishing processes.

4.1. Measured and derived indicators for monitoring the slurry

condition
Fig. 3 provides an overview about the installed measurement
devices. On the one hand, the proposed monitoring system focuses
on the description of the slurry condition, due to the signicant inuence on the performance of the polishing process and,
additionally, because the reow of used slurry can carry information about the mechanisms and interactions in the working gap.

The monitoring system measures important physical and chemical properties of the slurry, such as pH, electrical conductivity,
and zeta-potential. The slurry temperature is measured before and
after the process as the slurry is constantly pumped into the working gap between polishing tool and workpiece surface. On the
other hand, in order to gather an insight in the energetic characteristics of the process the power of the spindles is measured
by the monitoring system. It is used to derive the process work
and the energy in relation to the amount of material removal in a

Fig. 2. Overview of the great variety of inuencing parameters in polishing (Hambcker, 2001).

6042

F. Klocke et al. / Journal of Materials Processing Technology 209 (2009) 60396047

in LabView, using Eq. (1) (Delgado et al., 2005), where ESA represents the measured electro-kinetic sonic amplitude,  the volume
fracture of particle,  the difference of density between particle
material and polishing liquid, c the velocity of sound and the
frequency:
d (w) =

ESA
c

(1)

4.2. Measured and derived indicators for the monitoring of the

polishing process as a tribological system

Fig. 3. Overview of the monitored indicators and the placement of the according
measurement devices.

qualitative manner. The data also allows the online calculation of

the coefcient of friction and herewith a qualitative consideration
of the tribological condition in the working gap of the polishing
process.
4.1.1. Slurry temperature
The slurry temperature is measured both in the inow and in the
reow with conventional thermo couples. This enables not only the
measuring of the absolute temperature but the calculation of the
temperature difference, which indicates the heat development in
the working gap.
4.1.2. pH value of the slurry
The pH value is dened as the negative logarithm of the concentration of hydrogen ions and describes the acid or alkaline effect
of a liquid. A change of the pH indicates chemical reactions with
the formation of hydronium or hydroxide ions. The pH inuences
among others chemical reactions, solution processes and the surface charge of particles and forms a widely discussed property of
polishing slurries for chemo-mechanical polishing. In this application a glass membrane based pH electrode with liquid electrolyte
is used.
4.1.3. Electrical conductivity of the slurry
The electrical conductivity describes the ability of a material
to conduct electricity. The conductivity of solutions is based on
the movement of electrical charged particles, such as ions and
depends on the type of the solved ions and its relevant amount.
Given no change in the existing type of ions in a solution, a change
in conductivity can be taken as indicator for changes of the ions
concentration.
4.1.4. Calculation of the zeta potential of the particles in the slurry
The zeta potential characterises the physical stability of slurries. Stability means that the particle size distribution does not
change during a specied time interval. Therefore, agglomeration
and sedimentation form two driving mechanisms of instability of
a slurry.
Particularly in CMP where slurries with preferably small particle
sizes are used, large particles in the slurry are frequently accountable for microscratches (Aytes et al., 2003). Park stated aggregation
of slurry particles as a source of large particles (Park et al., 1998).
The zeta potential cannot be measured directly, but must be calculated. The integrated device measures the electro-kinetic sonic
amplitude. The calculation of the zeta potential was programmed

The main objective of the integrated power measurement is to

gain a qualitative view inside the energetic processes taking place
in the working gap. The employed polishing work can be calculated by an integration of the polishing power over the time. The
overall amount of energy can be employed in different actions in
the working gap. On the one hand, some of the energy is used
for the mechanical interactions between the polishing agent and
the workpiece surface, for example for plastic deformation or chip
formation. On the other hand, the energy which is brought into
the process could be transferred to heat and as well be needed
for activating chemical reactions. These considerations show that
the monitoring of the power of the spindles and the derivation of
the polishing power contributes to the process understanding. In
particular, the relationship between material removal and polishing power is of interest for the process analysis to characterise the
tribological condition in the working gap of the polishing process.
In the proposed monitoring system the performance of electric
drives of tool and workpiece spindle are measured separately by
capturing the currents and voltages at the frequency converters.
4.2.1. Determining the polishing power and deriving the specic
energy
The polishing power is calculated by adding the two measured
powers of the spindles minus the both idle powers measured without contact between workpiece and tool (Eq. (2)).
p = ptool + pworkpiece pidle

(2)

The integration of the polishing power is resulting in the polishing work using Eq. (3). The integration is approximated by the
following summation, because the analogue measured power is
rstly converted to a digital signal.

W =
W =

p dt (ideal relation),

p t (approximated formula)

(3)

A normalisation of the polishing work by dividing through the

volume of removed material allows to compare polishing processes
regarding the different energetic efforts and efciency. As already
mentioned in Section 2, the normalised value is called specic
energy.
4.2.2. Calculating the coefcient of friction
The elements of the polishing system, e.g. polishing pad, workpiece and polishing slurry, and the resulting interactions can be
regarded analogically as a tribological system (Dambon, 2005). This
assumption suggests the calculation of the coefcient of friction
which describes the tribological condition in the working gap of
the polishing process. The coefcient of friction is dened as the
quotient of friction force (Fr ) and normal force (Fn ). It is assumed
that frictional interactions cause the complete measured polishing power. In Eq. (4) ppolishing represents the polishing power, vrel
the relative velocity which is the velocity between workpiece and

F. Klocke et al. / Journal of Materials Processing Technology 209 (2009) 60396047

6043

Fig. 4. Graphical user interface of the monitoring system based on LabView (excerpt).

tool, p the load (pressure) and Af the area carrying the load in the
working gap.
Fr =

ppolishing

vrel

and Fn = p Af

(4)

This results in Eq. (5) which is used for the calculation of the
coefcient of friction in the proposed monitoring system.
=

ppolishing
p Af vrel

(5)

Additionally to the measured and calculated parameters, the

monitoring system calculates online several statistical values for
each run. This includes the average values and standard deviations
as well as minima and maxima of the indicators.
4.3. Implementation of the system and graphical user interface
The complete implementation of the monitoring system is based
on a nite state machine with the following main states: user
input, initialising of the system and the measurement devices, control of devices, measurement (including visualisation, calculation
and data recording), end of measurement (input of discontinuous
data), stop (after last measurement, further calculation and result
sheet writing). A complex graphical user interface (GUI) facilitates
the operation by users. At the start of the programme a pop-up
menu appears for the input of all required data by the user. On
the rst tab of this menu the user has to enter general data (user,
name of experiment), specications of the polishing system and
process parameter. Based on these data the monitoring systems
loads all required data, such as idle power of spindles, density of
the workpiece material, viscosity of the slurry etc, from a database.
If necessary, the user can edit the data. Fig. 4 shows the graphical
user interface. The graphs reveal the online development of temperature difference, pH value, electrical conductivity, zeta potential
and power and the graphs of the last two runs in order to examine
the reproducibility of the process.

5. Experimental setup for the evaluation of the proposed

monitoring system
The practicability of the approach is validated by investigations
of the polishing behaviour of silicon nitride which is primarily used
in adverse environments and applications, due to its extraordinary
hardness, corrosion resistance, low thermal expansion, high thermal shock resistance and low weight. Example applications are
bearing components and moulds and dies for sheet metal forming
or replication of optics. To ensure a high quality of the products in
the terms of tribological and optical functionality a high surface
quality and integrity has to be maintained by polishing operations. Beside of its numerous applications the material removal
mechanisms in polishing of silicon nitride is still not fully understand. Therefore, the computer-based data acquisition and analysis
is used to gain scientic insight on the relevant mechanisms and
interactions. The important ambition of the ongoing investigations is an efcient design of polishing strategies for a computer
controlled polishing process which allows the processing of free
formed geometries of complex ceramics geometries as shown by
Klocke et al. (2008).
5.1. Machine tool used for the implementation of the monitoring
system
The investigations were carried out on a Satisloh Synchrospeed
SL120 polishing machine. This machine is suitable for polishing planar and spherical lenses and is widely used in the manufacturing of
precision optics. The use of this equipment ensures that any undesirable inuences arising from kinematic instabilities are avoided
because it is state-of-the-art and featuring a high stability. The polishing slurry circulates in a closed loop. The slurry tank has a volume
of 5 l and the temperature is controlled by a heating/cooling device.
The investigations include the inspection of the polished
surfaces by optical measurement systems, i.e. white light
interferometer and laser interferometer, and scanning electron
microscopy. The material removal rate is quantied by weight

6044

F. Klocke et al. / Journal of Materials Processing Technology 209 (2009) 60396047

measurements. The special synchrospeed kinematic allows the

assumption of a homogeneous material removal on the complete
surface. Therefore, the amount of removed material can also be
expressed in a height difference, calculated on the basis of the
weight loss after each process step.

as a catalyst under the inuence of high pressure and tribological interactions in the working gap and the combination with the
so called chemical tooth ability of ceria (Hu et al., 1998) Whereas,
the removal mechanisms in polishing advanced ceramics using diamond slurry is assumed to be an abrasive process.

5.2. Materials and polishing systems

6. Results and discussion

In order to illustrate the capabilities of the computer-based

monitoring it is used to investigate the polishing behaviour of hot
isostatically pressed silicon nitride.
The objective is to elaborate scientic understanding of the process interactions. Based on this knowledge, the optimisation of the
material removal rate, the surface quality and the form accuracy
can be realised. Therefore, fundamental investigations of polishing
the silicon-based ceramics are undertaken and evaluated. The specimen were ground by using diamond grinding wheels with a cup
grinding kinematic. The polishing system consists of polyurethanebased polishing foils as tools and ceria slurry based on deionised
water (concentration 60 g/l).

This section provides examples, in which the monitoring system

was applied to enhance the scientic insight on material removal
mechanisms in polishing of advanced ceramics.

5.3. Experimental procedure

For the purpose of investigating the inuences of input parameters on the response of the output parameters, i.e. material removal
rate and surface quality, short-term experiments were conducted.
Each sample was polished for 20 min divided into three steps
of 5, 5, and 10 min. After each single step, the amount of material removal was measured by determining the weight loss. After
20 min the surface quality and form accuracy was measured as well.
The polyurethane foam was dressed to provide a fresh surface to the
following process step. Every specimen was polished for 60 min in
total in order to proof stability and reproducibility of the process. In
order to extend the knowledge about the polishing behaviour longterm experiments are realised with several steps (20 min each) up
to a total polishing time of 8 h.
The short-term experiments are conducted to identify the inuences of input parameters, e.g. pressure, velocity and concentration
of polishing agents in the slurry. The long-term experiments deliver
the basis to elaborate an understanding of, e.g. the life time of the
slurry and the stability of the process.
5.4. Hypothesis for material removal mechanisms
In general, the material removal mechanism in lapping and polishing range from pure mechanical character, e.g. abrasion, erosion
and fatigue, and pure chemical character, e.g. chemical reactions
such as oxidation or dissolution of the material (Evans et al., 2003).
Particularly due to the hardness of advanced ceramics, abrasive
polishing with diamonds represents a widely applied polishing
technique (Chinn, 2002).
Former investigations (Klocke et al., 2007) revealed satisfying
results using ceria and diamond based slurries for the nishing of
silicon nitride samples. Material removal rates of up 0.3 m/min
were achieved. Both polishing slurries realised highly reproducible
surface qualities in the range of several nanometers. However, ceria
slurry tended to result in a more efcient polishing operation than
the one with diamond slurry due to less machining efforts. The best
results with ceria slurry were achieved with the same pressure as
with diamond slurry but with less revolutions, and certainly less
costs for the polishing agent.
Following the argumentation by Jiang and Komanduri (1998)
the high efciency of polishing silicon nitride with ceria slurry
is based on an interaction of chemical and mechanical removal
mechanisms. The mechanism can be explained by hydrolysis of
the silicon nitride surface in aqueous solutions where ceria acts

6.1. Short-term experiments of polishing silicon nitride with ceria

To verify the stated hypothesis, experiments of polishing silicon nitride with ceria slurry were conducted. The computer-based
monitoring system was used to gain information about the inuence of signicant input parameters on the response of the
indicators dened in Section 4.
Taking the Preston Hypothesis (Preston, 1927) for granted an
increase of the material removal rate should easily be realised by
increasing the pressure p and the relative velocity vr as the main
machining parameters. In Eq. (6) dz represents the removed height
and dt the time interval. The Preston coefcient KP stands for various other inuences on the process which include the material
properties, the surface quality of the prior machining operation,
material properties of the polishing agent and the polishing pad as
well as other inuencing parameters as indicated in Fig. 2.
dz
= KP p vr
dt

(6)

The short-term experiments revealed that the pressure features

the most signicant inuence on the material removal rate. Following the DoE methodology, one of the rst diagrams of interest
for the statistical analysis of experimental results is the diagram
of the so called main effects. The left graph in Fig. 5 shows the
data which is the basis for the highly aggregated main effect diagrams. The right graph of Fig. 5 shows the main effect diagram of
the machining parameters pressure and relative velocity on the
material removal. The signicance of the parameter pressure is
clearly revealed by the high slope of the curve which nearly features linearity. However, the increase of the relative velocity shows
the tendency to inuence the surface quality of the silicon nitride
sample. The pattern of the main effect diagram can be logically
found in the detailed data given in the left graph. Additionally, Fig. 5
(left graph) points out the usual slide variations of polishing results
and the according standard deviations of the single experimental
data sets.
Fig. 6 shows two graphs to emphasise the average resulting surface quality in terms of the roughness parameters Ra (left graph)
and Rt (right graph) in dependence of the main machining parameters. The surface quality of the polished samples is very high,
featuring Ra values lower than 5 nm for all resulting surfaces and Rt
values mostly around 50 nm. This is due to the chemo-mechanical
character of the material removal mechanism (Klocke et al., 2007).
The analysis of the Ra graph shows that the lowest, i.e. best, values
were achieved by high pressure and low relative speed. The Rt graph
which shows the maximum distance between a measured depth
and a measured height of the surface roughness leads to the same
result. Beside of two values which have to be labelled as outliers
all resulting Rt values follow an anticipated pattern and stay within
the usual variation which is common for the roughness parameter Rt due to its high sensitivity. The combined interpretation of
Figs. 5 and 6 arises that high removal rates and low roughness
for polishing of silicon nitride with ceria can be achieved by high
pressure and low relative speeds. Therefore, this combination of

F. Klocke et al. / Journal of Materials Processing Technology 209 (2009) 60396047

6045

Fig. 5. Detailed showing all measured material removal rates and the according standard deviations (left graph), main effect diagrams for the material removal rate to
evaluate the dependency on the machining parameters (right graph).

Fig. 6. Inuence of the machining parameters on the resulting surface roughness in terms of arithmetic average of the roughness prole Ra (left graph) and maximum
roughness Rt (right graph).

parameters was chosen for the long-term experiments discussed

in the following section.
6.2. Long-term experiments of polishing silicon nitride with ceria
Due to circulation of the slurry the monitoring of timedependent changes of chemical and physical properties enables the
researcher to draw conclusions on the removal mechanisms. For a
better understanding a long-term experiment was conducted with
an overall polishing time of 8 h. The process was stopped every
20 min to determine the removal rate by measuring the weight
loss. The proposed monitoring system facilitates the survey of, e.g.
pH and the electrical conductivity as important indicators of the
slurry condition and its inuence on the material removal rate.
Any changes of the slurry condition are observed by the before
mentioned indicators and enable the researcher to draw online
conclusions on mechanisms and interactions as well as the process stability. The monitoring difference between the temperature
of slurry inow and outow should show constant values. If not the
researcher will be alerted that, e.g. the volume ow rate of slurry
inow has decreased which causes instabilities of the running process and renders the results that are not useful.
6.2.1. Monitoring the time-dependent evolution of the pH value
An indicator of the hydrolysis of silicon nitride stated as a mechanism in Section 5.4 is given by the observed increase of the pH
value (Fig. 7). Jiang and Komanduri (2001) discussed several stoichiometric calculations of possible chemical reactions. Dependent
on the environmental conditions it is assumed that nitrogen respectively ammonia is a reaction product which leads to the increase
of the pH value. In order to determine a second indicator for the
validity of the hypothesis, samples of the slurry were extracted

periodically during the long-term experiments. With these samples measurements of the concentration of ammonium hydroxide
were undertaken using photometric detection. The results revealed
an increase of the concentration of ammonium hydroxide from
0.1 mg/l at the beginning of the long-term experiments up to
13 mg/l after 8 h. Thus the validity of the hypothesis stated in Section 5.4 was afrmed. The pH graph in Fig. 7 shows three times
a slide decrease, i.e. at the duration of 120, 200, and 380 min. At
these time steps the long-term experiments had to be stopped for
a while. For the determination of the reasons for the described
decrease ongoing research is needed. Additionally, the hydroxide concentration was calculated based on the monitored pH with
the equilibrium constant of water (Eq. (7)). The second graph in
Fig. 7 shows the anticipated linear trend with three clearly visible
interrupts. It can be recognised that the hydroxide concentration
increases quite rapidly in the rst 300 min of the long-term experiments. After that point, a slide transition of the increase can be seen

Fig. 7. Time-dependent evolution of the pH value and correlation with the calculated hydroxide concentration in the long-term experiments.

6046

F. Klocke et al. / Journal of Materials Processing Technology 209 (2009) 60396047

Fig. 8. pH sensitivity of the material removal rate in polishing silicon nitride with
ceria slurry.

and it decelerates because the hydroxide concentration approaches

the saturation of hydroxide in ceria slurry. These ndings could not
have been detected in such a distinct way by solely analysing the
pH graph.
c(OH ) =

Kw
10pH

, Kw equilibrium constant of water

(7)

Furthermore, the investigations conrmed the pH sensitivity

of the material removal rate. Experiments with identical process
parameters but a variation of the pH value from acidic (pH around 3)
to alkaline (pH around 12) demonstrate that the material removal
rate decreases in acidic as well as in alkaline environment (Fig. 8).
These results correlate with the results of the above described
investigations where high removal rates were realised with a pH
value range from 5.5 to 8.5.
6.2.2. Monitoring the time-dependent evolution of the electrical
conductivity
A further clue on the hydrolysis is given by the increase of the
electrical conductivity, shown in Fig. 9. The increase indicates the
formation of ions, e.g. ammonium hydroxide. The second graph
shows the weight loss of the sample. The correlation between the
increase of conductivity and the amount of removed material support the hypothesis that the material removal is partly driven by
chemical reactions. The results argue for the validity of the hypothesis on the chemo-mechanical mechanisms which was stated in
Section 5.4.
6.2.3. Monitoring the process based on coefcient of friction and
polishing power
In Fig. 10 the time-related evolution of the material removal
rate and the coefcient of friction is given. The graphs clearly show
a correlation which allows to draw the conclusion that the operator
can achieve a rst impression on the material removal rate during
the running process by observing the results of the online calculated
coefcient of friction. During a single polishing step a stable process
accounts for a stable value of the coefcient of friction. Any aberrations in the graphs indicate aberrations of the running process and

Fig. 9. Time-dependent evolution of the electrical conductivity and amount of

removed material in the long-term experiments.

Fig. 10. Correlation of the material removal rate and the online calculated coefcient
of friction.

thus missing stability. By analysing the data of the long-term experiments in Fig. 10 the researcher learns that the height of the graph
of the coefcient of friction already gives an impression of the trend
of the material removal rate in the currently running step. Thus it
is possible to preview the process results before any weight measurements were taken to quantify the amount of removed material.
Alternatively, observing the monitored polishing power allows the
same conclusions. The sole difference is that the coefcient of friction which is calculated as shown in Eq. (5) also takes into account
the area of contact between tool and workpiece. Thus it is possible
to compare several investigations with different workpiece geometries because the value of the coefcient of friction is normalised
compared to the value of the polishing power.
In addition to the monitoring of the slurry condition the system is also applied to investigate the energetic coherences in the
working gap. Hambcker already mentioned a linear correlation
between the employed polishing work and the amount of removed
material for the chemicalmechanical polishing of different glass
types and thereby concluded that the process is generally reproducible (Hambcker, 2001). The long-term experiment conrms
the linear correlation, as shown in Fig. 11. This indicates, that no
signicant change in the energetic coherences took place.
The comparison of different data sets regarding the polishing
work enables further conclusions. As an example the energetic efciency of different concentrations of ceria slurries is given in Fig. 12.
No other experimental conditions were changed. In the left graph
of Fig. 12 it is shown that increasing the concentration of polishing agent induces an increase of the amount of removed material.
Nevertheless, the increase does not feature a linear behaviour. The
gain of doubling the concentration of ceria from 30 to 60 g/l is
much higher than the gain of 60120 and to 180 g/l, respectively.
The graph on the right in Fig. 12 analyses the specic energy of
the different concentrations. A low value is favourable but keeping
in mind that doubling the concentration also means doubling the
price of the slurry, the concentration of 120 and 180 g/l is regarded
as not efcient. The costs of the slurry are doubled but as the amount
of removed of material is increased by half the time needed for polishing decreases by one third. The drawn conclusion is that 60 g/l

Fig. 11. Evolution of the polishing work and specic energy.

F. Klocke et al. / Journal of Materials Processing Technology 209 (2009) 60396047

6047

Fig. 12. Using the monitored polishing work to describe the inuence of concentration of polishing agent in the slurry on the amount of removed material (left graph) and
deriving the process efciency by analysis of the specic polishing energy (right graph).

does lead to an overall acceptable performance regarding material

removal rate and costs for the slurry.
7. Conclusion
In order to elaborate a scientic insight regarding the process
interactions in polishing a computer-based data acquisition and
analysis was realised. The application is programmed in the well
known LabView programming environment. The monitoring system integrates numerous measurement devices and calculations
in order to facilitate the investigations and to indicate the process
stability and reproducibility. The monitored indicators are polishing power resp. work, pH value, conductivity, zeta potential and
temperature of the slurry. The online calculated indicators are coefcient of friction, the specic energy, and average values of the
monitored indicators for statistical analyses on a higher aggregation level. This monitoring system can detect disturbances of
process stability, which allows early identication or even the prevention of rejects in an industrial production.
The functionality of the presented system was veried by investigations of the polishing behaviour of silicon nitride. Therefore, the
majority of indicators were exemplary discussed in the paper.
The monitoring system can be used for a great variety of materials. Future work will implement a process control, e.g. for online
adjustment of the pH value in order to ensure the best performance
regarding material removal rate and resulting surface quality in pH
sensitive processes. Due to the mobility of the system it can be used
in industrial environments and easily adapted to various polishing
machinery. Beside the usage in ongoing research activities, the presented approach has already been used in industrial applications to
support the understanding of specic questions on the interactions
in serial production.
Acknowledgements
We sincerely thank the German Research Foundation (DFG)
for nancial support of the Collaborative Research Center SFB/TR4
(RWTH Aachen and University Bremen (Germany), Oklahoma State
University at Stillwater (USA)).
References
Aytes, S.D., Armstrong, J.S., Mortensen, K.A., Russell, C.W., et al., 2003. Experimental investigation of the mechanism for CMP micro-scratch formation. In:

University/Government/Industry Microelectronics Symposium. Proceedings of

the 15th Biennial, pp. 107109.
Chinn, R.E., 2002. Ceramography. ASM International, Materials Park.
Colloidal Dynamics Ltd., 2000. The CMP Slurry Monitor. Australia.
Dambon, O., 2005. Das Polieren von Stahl fr den Werkzeug- und Formenbau. Ph.D.
thesis. RWTH Aachen, Germany.
Delgado, A.V., Gonzalez-Caballero, F., Hunter, R.J., Koopal, L.K., Lyklema, J., 2005.
Measurement and interpretation of electrokinetic phenomena. Pure Appl. Chem.
77, 17531805.
Evans, C.J., Paul, E., Dornfeld, D., Lucca, D.A., Byrne, G., Tricard, M., Klocke, F., Dambon,
O., Mullany, B.A., 2003. Material removal mechanisms in lapping and polishing.
CIRP Ann. 52, 611633.
Gitis, N., Vinogrado, M., Gao, C., 2001. Quantitative evaluation of CMP processes and
materials using a CMP test with multiple sensors. In: Proceedings of the Second
International Conference on Microelectronics and Interfaces. American Vacuum
Society, Santa Clara.
Hambcker, S., 2001. Technologie der Politur sphrischer Optiken mit Hilfe der
Synchrospeed-Kinematik. Ph.D. thesis. RWTH Aachen, Germany.
Hu, Y.Z., Gutmann, R.J., Chow, T.P., 1998. Silicon nitride chemicalmechanical polishing mechanisms. J. Electrochem. Soc. 145, 39193925.
Hunter, R.J., 2001. Measuring zeta potential in concentrated industrial slurries. Colloids Surf. A: Physicochem. Eng. Aspects 195, 205214.
Jiang, M., Komanduri, R., 1998. On the nishing of Si3 N4 balls for bearing applications.
Wear 215, 267278.
Jiang, M., Komanduri, R., 2001. Chemical mechanical polishing in magnetic
oat polishing of advanced ceramic and glass. Key Eng. Mater. 202203,
114.
Klocke, F., Dambon, O., Schneider, U., 2005a. Removal mechanisms in ne polishing
calcium uoride and silicon. Prod. Eng. 12, 510.
Klocke, F., Dambon, O., Capudi Filho, G.G., 2005b. Inuence of the polishing process on the near-surface zone of hardened and unhardened steel. Wear 258,
17941803.
Klocke, F., Zunke, R., Dambon, O., 2007. Chemo-mechanical polishing of silicon
nitride. In: Proceedings of 7th euspen International Conference. Bremen, Germany, pp. 356359.
Klocke, F., Brecher, C., Zunke, R., Tuecks, R., 2008. Finishing complex surfaces with
zonal polishing tools. In: Mitsuishi, M., Ueda, K., Kimura, F. (Eds.), Manufacturing
Systems and Technologies for the New Frontier. The 41st CIRP Conference on
Manufacturing Systems. Tokyo, pp. 445448.
Kojima, T., Miyajima, M., Akaboshi, F., Yogo, T., Ishimoto, S., Okuda, A., 2000. Application of CMP process monitor to Cu polishing. IEEE Trans. Semicond. Manuf.
13, 293299.
Matsuzaki, S., Tanzawa, A., Igarashi, T., Suzuki, T., Tokushige, K., 2002. Development
of data logging system for chemical mechanical polishing and its application for
process control. IEEE Trans. Semicond. Manuf. 15, 438441.
Oliver, M.R. (Ed.), 2004. ChemicalMechanical Planarization of Semiconductor Materials. Springer, Berlin.
Park, H., Kim, K., Hang, C., Chung, U., Lee, M., 1998. Control of microscratches in
chemicalmechanical polishing process for shallow trench isolation. Jpn. J. Appl.
Phys. 37, 58495853.
Preston, F.W., 1927. The theory and design of plate glass polishing machines. J. Soc.
Glass Technol. 11, 214219.
Sandhu, G., Schultz, L., Doan, T., July 1991. US Patent 5,036,015.