1. FT33- 06_5
CRITICAL
COMPONENTS
& SUB SYSTEMS
Effect of pump type on the health of
various CMP slurries
Mark R. Litchy, Donald C. Grant, CT Associates, Inc., USA, & Reto Schoeb, Levitronix GmbH, Switzerland
Source: Mykrolis, MIT LL
ABSTRACT
Delivery systems used to supply
slurry to CMP planarization tools can
damage slurry. In this experiment, four
slurries were circulated in a simulated
slurry delivery loop at a fixed flow
rate and pressure using a variety
of pumps (bellows, diaphragm, and
magnetically levitated centrifugal) to
determine the effect of circulation on
the slurry health. During each test, a
number of slurry health parameters
were monitored including the size
distribution of the particles in the
slurry. Most slurry health parameters
were unaffected during the tests.
However, significant changes in the
large particle tail (particles ≥ 0.5µm)
of the slurry particle size distributions Figure 1. Test system schematic.
(PSD) were observed. Both the pump
Experimental procedure
accepted as low shear devices due to
and slurry type played important
their relatively low speeds of operation.
roles in the magnitude of the change. A schematic of the test system used
Centrifugal pumps have not been widely
In some slurries, large increases in in this evaluation is shown in Figure
used for delivery of agglomeration-
the large particle concentrations 1. Each pump was used to circulate
sensitive slurries since they are perceived
were observed during circulation 12 liters of slurry at a flow rate of
as high shear devices, due to their
with diaphragm and bellows pumps, approximately 26-30 lpm (7-8 gpm) and
relatively high speeds of operation.
while in other slurries increases were outlet pressure of approximately 30 psig
Previous studies have shown significant
not observed. With the magnetically (2.1 bars). Settling of the slurry in the
increases in large particle concentrations
levitated centrifugal pumps, minimal tank was minimized by drawing from
when certain types of pumps (diaphragm
changes were observed, regardless of the bottom of a conical bottom tank
and bellows) were used to circulate
the type of slurry tested. and by turning the volume of slurry in
slurry [2-5]. However, these tests were the tank over in less than 30 seconds.
performed using a slurry known to The return line to the slurry tank was
be sensitive to particle agglomeration. submerged below the liquid level of
Introduction In practice, there are many CMP the slurry to avoid entraining gas into
planarization applications incorporating
Delivery systems are often used to the slurry. The return line was also
many different types of slurries.
supply slurry during wafer planarization. positioned to minimize the formation
While it is not feasible to evaluate all
These systems pressurize and circulate of a large vortex in the tank, which can
available slurries with all types of pumps,
the slurry to deliver it to the tools entrain gas into the slurry. No valves
this set of experiments was undertaken
and keep the particles in suspension. were used to generate backpressure at
to determine the effect of pump type
Pressur ization and circulation are the outlet of the pump. Instead, a long
on the health of a small number, but
accomplished by var ious means length of 1⁄2quot; PFA tubing was used to
wide variety, of slurry types. The slurries
including pumping and pressure- gradually reduce the pressure at the
selected for this experiment include
vacuum technology. Typically, the slurry pump outlet to ambient pressure at the
a fumed silica slurry, a colloidal silica
passes through the equipment providing end of the return line to the tank.
slurry, an alumina slurry, and a ceria
the motive force approximately The air pressure supplied to the
slurry. The pumps in this experiment
100 times before it is used to polish diaphragm and bellows pumps was
included a diaphragm pump, a bellows
wafers, i.e. the slurry is ‘turned-over’ adjusted to achieve the desired flow
pump, and two magnetically levitated
approximately 100 times [1]. Some rate and outlet pressure. Meanwhile, the
centrifugal pumps. The sizes of the
CMP slurries are susceptible to damage speed of the small and large centrifugal
pumps were chosen such that each was
caused by mechanical handling. For pumps was varied to achieve the desired
capable of delivering slurry at a flow
example, particle agglomerates may flow rate and pressure. In each test, the
rate of 30 lpm at an outlet pressure of 30
form that can limit the life of filters or slurry was circulated until approximately
psig. Manufacturer-recommended pulse
reduce yield by causing wafer defects. 1,000 tank turnovers were achieved.
dampeners were installed downstream
Diaphragm and bellows pumps The test system was constructed of PFA,
have been commonly used for bulk of both the bellows and diaphragm except for the conical bottom tank that
CMP slurry delivery. These pumps are pumps to minimize pulsation. was constructed of polyethylene. The
2 SEMICONDUC T O R F A B T E C H – 3 3 RD E D I T I O N W W W .FABTECH .ORG

2. FT33- 06_5 CRITICAL
COMPONENTS
& SUB SYSTEMS
Barbara, CA) that determines particle
size by dynamic light scattering. The size
distribution of the large particle tail was
measured using a Particle Sizing Systems
AccuSizer 780 sensor.
The AccuSizer 780 sensor uses a
combination of light scattering and
light extinction to measure the size
distribution of particles ≥ 0.56µm. The
size measurements were performed by
injecting the slurry sample into a flowing
stream of filtered deionized water using
the test system shown in Figure 2. The
dilution ratio was varied by adjusting
the slurry injection rate. The slurry types
required different dilution ratios, which
varied from approximately 800:1 to
Figure 2. Slurry dilution system schematic. 32,000:1. Prior to starting the pump tests,
each slurry was thoroughly analyzed to
slurry used in each test was taken from dehydration can both result in particle determine the proper dilution ratio for
the same lot of slurry. agglomeration in the slurry. A chiller and accurate measurement of the large particle
The tank holding the slurry was stainless steel coil were used to maintain tail of the slurry PSD. Each slurry contains
blanketed with nitrogen to prevent the slurry at 22 ± 2°C during the test. a very high concentration of the ‘working’
absorption of carbon dioxide from the Samples were drawn from the system particles, which are responsible for the
air, which can change the pH of the at selected times for analysis. The PSD mechanical portion of the polishing. Care
slurry. The nitrogen was humidified to was measured using two techniques. The must be taken to ensure that each slurry
prevent dehydration of the slurry. The size of the ‘working’ particles (typically ~ is diluted sufficiently such that scattering
relative humidity in the tank was > 90% 0.05-0.5µm) was measured using Particle from particles < 0.56µm does not interfere
throughout the test. Shifts in the pH and Sizing Systems’ NICOMP 380ZLS (Santa with the particle size analysis.
Figure 3. Cumulative PSDs measured in the fumed silica slurry.
3
S E M I C O N D U C T O R F A B T E C H – 3 3 RD E D I T I ON

3. FT33- 06_5
CRITICAL
COMPONENTS
& SUB SYSTEMS
Figure 4. Cumulative PSDs measured in the alumina slurry.
Figure 5. Cumulative PSDs measured in the colloidal silica slurry.
4 SEMICONDUC T O R F A B T E C H – 3 3 RD E D I T I O N W W W .FABTECH .ORG

4. FT33- 06_5 CRITICAL
COMPONENTS
& SUB SYSTEMS
Figure 6. Cumulative PSDs measured in the ceria slurry.
The working PSD and zeta potential measured with each slurry. The four concentration of particles ≥ 0.56µm
graphs in each figure present the results increased during the diaphragm and
measurements were both made using
from the four pumps tested. The initial bellows pump tests, but to a lesser
the NICOMP 380ZLS. The samples
PSD, measured prior to the start of each extent than during tests with the fumed
were diluted approximately 40:1 into
test, is presented in each graph as well silica and alumina slurries.
deionized water and analyzed at 23°C.
as the PSDs after selected numbers of As with the previous slurries, no
Each PSD measurement was made over
turnovers. changes were observed during the
a period of 10 minutes, while each zeta
The fumed silica slurry, which has centrifugal pump tests in the ceria slurry
potential measurement was made over
been known to be sensitive to particle (Figure 6), while significant increases
2 minutes. The PSD and zeta potential
agglomeration [2-5], exhibited large in the concentration of particles larger
measurements of each sample were
increases in the concentration of large than about 2µm in size were observed
performed in triplicate and quintuplicate,
particles when subjected to multiple during the diaphragm and bellows
respectively. The size measurement data
passes through the diaphragm or bellows pump tests.
were analyzed using the instrument’s
pumps (Figure 3). The concentration To better visualize the changes in
gaussian distribution assumption.
increases occurred over a wide particle the PSDs during the tests, the ratios of
Other slur r y health parameters
size range from ≥ 0.56µm to more than the particle concentrations at each test
measured included zeta potential, total
10µm. However, when this slurry was point to the corresponding particle
percent solids, specific gravity, and pH.
subjected to many passes through the concentration at the start of each
Measurements of each sample were
magnetically levitated centrifugal pumps, test were plotted. An example of the
performed in triplicate.
little change in the concentration of results is presented in Figure 7 for the
Results and discussion large particles was observed. alumina slurry tests. The four graphs in
Similar to the fumed silica slurry, Figure 7 show concentration ratios for
The results presented will focus on
little change was observed dur ing selected size channels as a function of
the measurements of the slurry large
the centr ifugal pump tests in the tank turnovers during each pump test.
particle tail since substantial changes
alumina slurry as shown in Figure 4. The results for the other three slurry
were observed in the tail during these
However, large increases in the particle tests are included in reference [6]. In this
experiments. Results for the other
concentration, particularly for particles slurry, significant changes in the large
parameters were not included due to
> 1µm, were observed dur ing the particle tail, particularly for particles ≥
the fact that none of those parameters
diaphragm and bellows pump tests. 1.5µm in size, were observed during the
changed significantly during the 1,000-
For the colloidal silica slurry (Figure diaphragm and bellows pump tests in less
turnover test.
5), little change was observed during than 10 turnovers. The concentration
Figures 3-6 present the cumulative
the centrifugal pump tests, while the of particles ≥ 1.5µm tended to increase
PSDs of the slurry large particle tail
5
S E M I C O N D U C T O R F A B T E C H – 3 3 RD E D I T I ON

5. FT33- 06_5
CRITICAL
COMPONENTS
& SUB SYSTEMS
Figure 7. Concentrations relative to initial concentration during the alumina slurry tests.
Figure 8. Concentration increases measured during all tests after 100 turnovers.
6 SEMICONDUC T O R F A B T E C H – 3 3 RD E D I T I O N W W W .FABTECH .ORG

6. FT33- 06_5 CRITICAL
COMPONENTS
& SUB SYSTEMS
Figure 9. Concentration increases measured during all tests after 1,000 turnovers.
linearly with increasing turnovers. After
TABLE I: SUMMARY OF CONCENTRATION CHANGES IN SUBMICRON
1,000 turnovers, the concentration of
AND SUPERMICRON PARTICLE SIZES DURING ALL TESTS
particles ≥ 2µm increased approximately
35- and 70- fold during the diaphragm
and bellows pump tests, respectively. This
equates to about a 3.5% and 7% increase
in particle concentrations each time
the slurry passes through these pumps,
respectively. Meanwhile, concentrations
measured during both centrifugal pump
tests tended to decrease with increasing
turnovers.
Figures 8 and 9 present a subset of
the data collected in this study. The
data are organized by slurry type rather
a decrease in the particle concentrations in three of the four slurries tested. In
than by pump type to more easily
was observed in some of the slurries the fumed and colloidal silica slurries,
compare the effects of the pumps on
during the centrifugal pump tests. For the concentration increases occurred
each of the slurries. For each slurry,
example, in the alumina slurry centrifugal over a wide range of particle sizes, from
the concentration ratios are presented
pump tests, the particle concentrations 0.56µm to larger than 10µm. However,
as a function of particle size for each
decreased on the order of 30-50% for concentration increases were very large
pump after approximately 100 and
particle sizes ranging from 0.56µm to during the diaphragm and bellows
1,000 turnovers. These points in time
more than 5µm. Ratios were not plotted pump tests in fumed silica slurry, while
were chosen since slurry is typically
for some of the large particle sizes in only small increases were observed in
turned over on the order of 100 times
the alumina and ceria centrifugal pump the colloidal silica slurry. Meanwhile, in
prior to use, with 1,000 probably being
tests since the particle concentrations the alumina and ceria slurries, little or
a conservative upper limit in most
were so low, but no increases in particle no particle concentration increase was
delivery systems.
concentrations were apparent. observed for particles < 1.0µm in size,
There were no observed significant
Meanwhile, the concentrations of but very large concentration increases
increases in the particle concentrations
particles dramatically increased during were observed with both the diaphragm
with either centrifugal pump at any
the diaphragm and bellows pump tests and bellows pumps for par ticles
particle size in any of the slurries. In fact,
7
S E M I C O N D U C T O R F A B T E C H – 3 3 RD E D I T I ON

7. FT33- 06_5
CRITICAL
COMPONENTS
& SUB SYSTEMS
≥ 2.0µm in size. (For the ceria slurry, the four slurries tested (fumed silica, REFERENCES
alumina, and ceria) during both the
the actual concentration ratios are likely
[1] Personnel communication with J.
diaphragm and bellows pump tests. For
to be even higher than indicated for
Kvalheim, BOC Edwards Chemical
the fumed silica slurry, the concentration
particles ≥ 3µm, since the initial particle
Management Division, Chanhassen,
increases were much greater than an
concentration for these large sizes was
MN, March 2003.
order of magnitude for both submicron
near the background of the test system.)
[2] Litchy MR and R Schoeb, ‘Effect
and supermicron particle sizes.
The particle sizes at which
of shear stress and pump method
concentration increases occur red
Summary on CMP slurry,’ Semiconductor
during the diaphragm and bellows
International, 27 (12), 87-90 (2004).
pump tests were remarkably similar in Previous work has shown that silica-
[3] Nicholes K, R Singh, DC Grant, and
all slurries. In general, the concentration based CMP slurries are sensitive to
MR Litchy, (2001). ‘Measuring particles
agglomeration induced by extensive
increases observed during the bellows
in CMP Slurries,’ Semiconductor
slurry handling. This study has shown
pump tests were larger than those seen
International, 24(8): 201-206.
that non-silica-based CMP slurries
during the diaphragm pump tests. The
[4] Litchy MR and R Schoeb (2005),
such as alumina and cer ia are also
concentration increases observed with
‘Effect of particle size distribution on
sensitive. Both slurry type and pump
the bellows pump were typically about
filter lifetime in three slurry pump
type are key factors influencing the
twice the level of increase with the
systems,’ Materials Research Society
magnitude of agglomeration during
diaphragm pump, except during the
Symposium Proceedings Vol. 867,
slur ry handling. Minimal changes
ceria slurry tests in which the increases
W2.8.1, (2005).
in the large particle tail of the slurry
appeared to be even higher.
[5] Litchy MR, DC Grant and G Van
PSD were observed during tests with
Table I shows a summary of the
Schooneveld (2006). ‘Effects of fluid
magnetically levitated centr ifugal
concentration changes relative to the
handling components on slurry
pumps, while large increases in the
initial concentration during each test
health,’ Transaction on Electrical and
large par ticle tail were obser ved
for both submicron (0.56 to 1.0µm)
Electronic Materials of the Korean
in silica, alumina, and cer ia-based
and super micron (≥ 3µm) particle
Institute of Electrical and Electronic
slurries during circulation tests with
sizes. Minimal changes in particle
Material Engineers.
either diaphragm or bellows pumps.
concentrations were observed during
[6] Litchy MR, DC Grant and R Schoeb
The bellows pump caused more
the 1,000 tur nover test for both
(2007). ‘Effect of pump type on
agglomeration than the diaphragm
centr ifugal pumps. For submicron
various CMP slurries,’ Proceedings of
pump during most of the slurry tests.
particles sizes, minimal changes were
the 26th Annual Semiconductor Pure
No significant changes were observed
also observed with the diaphragm
Water and Chemicals Conference,
in the other slurry health parameters
and centrifugal pumps in both the
Sunnyvale, CA, pp 119-134.
(working PSD, zeta potential, total
alumina and ceria slurries. However,
percent solids, pH, and specific gravity)
concentration increases greater than Reprinted with permission from the
for any of the pumps during any of the
an order of magnitude were observed SPWCC (Semiconductor Pure Water
slurry tests.
for supermicron particles in three of and Chemicals Conference).
ABOUT THE AUTHORS ENQUIRIES
Mark Litchy is a Research Engineer for CT contamination control, particle measurement CT Associates, Inc.
7121 Shady Oak Road, Eden Prairie, MN 55344
Associates, Inc. He has more than 10 years of and control, filtration, permeation and chemical
USA
experience in particle measurement and control engineering. He has more than 30 years of
Tel: +1 (952) 470-0166
in high purity liquid chemicals and gases and experience in analysis and purification of fluids
Website: www.ctassociatesinc.com
CMP slurry characterization. He is the author of and is the author or co-author of more than 130
more than 20 publications and presentations. He technical papers and presentations. He has a
has a M.S. in mechanical engineering from the M.S. in mechanical engineering from the Particle
Levitronix GmbH
Particle Technology Laboratory at the University of Technology Laboratory at the University of
Technoparkstrasse
Minnesota and a B.A. in physics from St. John’s Minnesota and a B.S. in chemical engineering from
CH-8005 Zurich
University. Case Western Reserve University.
Switzerland
Don Grant is Founder and President of CT Reto Schoeb is the CEO and Founding Member of
Website: www.levitronix.com
Associates, Inc., which performs contract research Levitronix GmbH, a pioneer in bearingless pump
and development and testing services related to technology.
8 SEMICONDUC T O R F A B T E C H – 3 3 RD E D I T I O N W W W .FABTECH .ORG