recycled fumed silica slurries in ILD CMP applications
Philipossian and Farhang Shadman, University of Arizona;
Patrick Levy, Saied Tousi, and Barry Gotlinsky, Pall; and W.
Scott Rader, Paul Lefevre, and Isamu Koshiyama, Fujimi
polishing (CMP) technology has played an enabling role in attaining
near-perfect planarity of interconnection and metal layers, an essential
step for realizing and miniaturizing high-performance devices. To ensure
stable and high-performance CMP characteristics, optimization of the
slurry, the pad, and other consumables is critical. Additionally, the
relatively high cost of ownership associated with CMP consumables warrants
novel approaches to reduce these expenses.
studies have demonstrated the feasibility of regenerating slurry.1,2
One study in particular showed that high-purity colloidal silica could
be reclaimed because of its higher dispersion and lower tendency to
clog than fumed silica.3 For these reasons, fumed silica
slurry used in interlayer dielectric (ILD) CMP applications was chosen
for the study discussed in this article. The goals of the study were
to characterize the slurry, which was regenerated using filtration techniques,
and to determine whether it could be rendered analytically and functionally
equiv- alent to fresh slurries.
polishing experiments were performed on 100-mm thermally grown SiO2
wafers using a scaled version of a 472 polisher from SpeedFam-IPEC (now
Novellus Systems, San Jose). Details of the experimental apparatus are
described elsewhere.4 To measure the shear force between
the pad and the wafer during polishing, a sliding table was placed beneath
the polisher. The sliding table consisted of a bottom and a top plate
on which the polisher rested. As the wafer and pad were engaged, the
top plate slid with respect to the bottom plate in only one direction
because of friction between the pad and wafer. The degree of sliding
was quantified by coupling the two plates to a load cell. The load cell
was attached to a strain gauge amplifier that sent a voltage signal
to a data-acquisition board.
apparatus was calibrated to report the force associated with a particular
voltage. IC-1000 perforated pads from Rodel (now Rohm and Haas Electronic
Materials, Phoenix) were used. Before frictional and removal-rate data
were acquired, each pad was conditioned for 30 minutes using PL-4217
slurry from Fujimi (Tualatin, OR). The silica abrasive concentration
was 12.5% by weight. Pad conditioning was performed using a 100-grit
diamond disk at a pressure of 0.5 psi (3.5 kPa), a rotational velocity
of 30 rpm, and a disk sweep frequency of 20 per minute. Conditioning
was followed by a 2-minute pad break-in step using a silicon dummy wafer.
Other experimental conditions included:
In situ conditioning at 30 rpm and 20 oscillations per minute.
Applied wafer pressure of 6 psi (~41.5 kPa).
A relative pad-wafer average linear velocity of 0.62 m/sec.
A slurry flow rate of 60 cm3/min.
Counterclockwise wafer and platen rotations.
aggressive polishing conditions and the use of a perforated pad were
selected to ensure relatively high slurry utilization efficiencies of
approximately 15% and high slurry shearing. Slurry utilization in CMP
applications is known to be quite inefficient, ranging from 2 to 20%.5
Since the efficiency of the process depends on such variables as the
slurry flow rate, wafer pressure, linear velocities, and pad grooving,
it was important to choose polishing conditions that resulted in a more
efficient process. The more efficient process provided a worst-case
scenario in which more of the slurry being delivered to the pad reached
the wafer than would be the case under typical conditions, creating
a larger quantity of slurry for regeneration.
samples were collected after multiple polishings were performed and
the slurry was recycled 1, 3, and 5 times. The samples were characterized
analytically to determine pH, trace-metal levels, viscosity, specific
gravity, mean aggregate particle size, and large-particle counts (LPCs,
or particles >1.0 Ám) before and after depth filtration.
depth filtration, the entire depth (or thickness) of the melt-blown
polymeric filtration medium is used to retain particles. As the tortuous
path of the particle is increased, the probability that a filter will
retain the particle also increases. When CMP slurries are filtered,
the goal is to pass the native slurry particles through the filter while
trapping the oversized agglomerates and foreign materials. This procedure
is used more to classify particles than to filter them. Studies have
shown that LPC data have been correlated to wafer defects.6
Furthermore, it has been demonstrated that filtration can reduce defects
during CMP processing.5 Thus, depth filtration is used as
a means of removing agglomerated particles or foreign materials that
may have been introduced into the slurry as a result of ILD polishing.
slurries can undergo changes in viscosity when the shear stress to the
fluid is sufficient, affecting the stability of the slurry. Because
of the morphology of the particles they contain, fumed silica slurries
are especially prone to shearing effects. Hence, the flow rate and corresponding
differential pressure chosen for this study were lower than the values
in typical slurry applications.
addition to undergoing analytical characterization, the fresh and recycled
slurry samples were characterized functionally to determine their removal
rate and coefficient of friction (COF) both before and after filtration.
Before the polishing experiments were conducted, baseline trials were
performed using fresh and "bypassed" slurry, where the slurry
was allowed to flow over the rotating pad with the wafer disengaged.
All filtration studies were performed utilizing a peristaltic pump to
deliver the slurry to Profile Y010 (1.0-Ám) filter media from Pall (East
Hills, NY). The material was cut into 1-in. segments. All tests were
conducted in a single pass at a flow rate of 50 cm3/min to
minimize the chances of slurry shearing during filtration.
of Friction. COF is defined as the ratio of shear to normal
the study described here, shear force was determined experimentally
using the combination of the sliding table, the load cell, and the strain
gauge amplifier. Normal force was obtained by multiplying polishing
pressure by the total surface area of the wafer.
Analysis. For a given polishing run, the measured total unidirectional
shear force as a function of time can be divided into two components,
a mean force and a fluctuating force component, as shown in the following
= F + f(t)
1 presents an example of the total force measurement obtained during
a typical polishing run. Sampling time was 1 second and the sampling
frequency was 1000 Hz. For a 75-second polishing experiment, a total
of 75 such plots would be generated and analyzed for tribological attributes.
The mean force F, which represents the average of all 75,000
data points, is used in calculating COF, as defined in the first equation
and utilized in previous studies.1–4 For spectral analysis,
the measured total unidirectional shear force function (which includes
the fluctuating component) is converted into frequency domain via fast
1: Shear force in time domain measured during a 1-second polishing
2 shows an example of this transformation, where the x-axis represents
signal frequency (in hertz) and the y-axis indicates the amplitude of
the transformed function. On a qualitative basis, this force spectrum
identifies the extent and frequency at which stick-slip phenomena occur.
In the tribology of the CMP process, stick-slip (i.e., hydrodynamic
chattering) refers to cyclic fluctuations in the magnitudes of frictional
force and relative velocity between the wafer and the pad. It is usually
associated with a relaxation oscillation that depends on a decrease
of COF with increasing sliding velocity. True stick-slip, in which each
cycle consists of a stage of actual stick followed by a stage of overshoot
(i.e., slip), requires that the kinetic COF (i.e., the parameter being
measured in this study) is lower than the static COF (i.e., the maximum
friction force that must be overcome to initiate macroscopic motion
between the wafer and pad). In contrast, random variations in friction-force
measurements do not constitute stick-slip. In this study, the stick-slip
criterion was met.
2: Shear force in frequency domain measured during the same 1-second
polishing run shown in Figure 1.
form of stick-slip can be caused by the spatial periodicity of the friction
coefficient along the path of contact (i.e., pad grooves or microtrenches
that are created on the surface of the pad by the conditioner, or complex
films that form and are abraded on the surface of the wafer).
Figure 2, the area under the curve is the basis of another parameter
called the interfacial interaction index (γ), which is
determined empirically based on transformed data in the frequency domain
and is essentially a measure of the range of forces encountered during
polishing. In other words, g is similar to the variance of the distribution.
On a quantitative basis, the area under the curve can be viewed as representing
the total amount of mechanical energy caused by stick-slip phenomena.
Further studies must be conducted to distinguish among the various types
of stick-slip phenomena that occur during CMP. In the meantime, this
study treated all forms of stick-slip as one entity, represented by
Removal. The most widely adopted removal-rate equation is the
one proposed by Preston, which states that the removal rate (RR)
is proportional to the product of the applied pressure and the relative
velocity of the substrate in contact with the pad:8
= kPr X p X U
constant (kPr) itself depends on the various chemical
and mechanical attributes of the process.
3a and 3b show that slurry recycling had little effect on viscosity
and specific gravity. Fresh filtered slurry had a viscosity of 2.41
cP while filtered slurry that had been recycled five times had a viscosity
of 2.39 cP. Specific gravity was 1.081 for fresh slurry and ~1.077 for
slurry that had been recycled five times. In addition, the change in
specific gravity corresponded to only a 0.15% drop in solids content
(from 12.50 to 12.35% by weight). Figures 4a and 4b, on the other hand,
show that while slurry recycling had little effect on pH (11.00 compared
with ~10.85 for fresh versus recycled slurry), it had a notable effect
on trace-metal levels, as highlighted by the threefold increase shown
in Figure 4b (3 ppm as compared with 1 ppm for fresh versus recycled
3: Effect of slurry recycling on (a) viscosity and (b) specific
4: Effect of slurry recycling on (a) pH and (b) metals content.
results presented in Figure 5 show that slurry recycling had a significant
impact on both mean aggregate particle size and LPCs. Mean aggregate
particle size increased by 11%, from 138 to 150 nm, after filtered slurry
was reused five times. However, while the use of unfiltered recycled
slurry resulted in a 50-fold increase in LPCs over unfiltered fresh
slurry, the use of filtered recycled slurry resulted in only a threefold
increase in LPCs over filtered fresh slurry. Moreover, all of the filtered
slurry samples resulted in at least three-times-lower LPC values than
fresh unfiltered slurry.
5: Effect of slurry recycling on (a) mean aggregate particle size
and (b) LPCs.
filtration does not have any notable impact on the above metrics except
possibly on mean aggregate size (although it appears that multiple reuses
of the slurry have a greater impact), and large-particle counts. In
that case, filtration is quite effective in reducing the number of LPCs
(especially in the case of the recycled slurries where the reduction
was greater than two orders of magnitude).
6: Effect of slurry recycling on (a) ILD removal rate and (b) coefficient
7: Correlation between ILD removal rate and COF.
6 shows the effect of slurry recycling on ILD removal rate and COF.
The results indicated a nearly 40% drop in removal rate as a result
of recycling the fumed silica five times. The relative reduction in
the COF value was also close to 40%, as shown in Figure 7, indicating
a near-perfect correlation between removal rate and COF. Both ILD and
COF tests compared fresh unfiltered slurry to filtered slurry that had
been recycled five times. After being recycled five times, slurry is
unusable without filtering. These results are consistent with a previously
reported near-linear relationship between removal rate and COF.4
8: ILD removal rate as a function of slurry pH. The circles are
from this study, and the squares are from the study documented in
drop in removal rate presented in Figure 6 is too large to account for
the decrease in pH represented by the square plots shown in Figure 8,
which are data derived from another study.9 However, the
decrease in both removal rate and COF is quite possibly a result of
the observed increase in aggregate particle size illustrated in Figure
9 (the smaller contact area between the slurry's abrasive particles
and the wafer surface is associated with the larger-size abrasives).10
The reason for the increase in mean aggregate particle size with repeated
slurry recycling is likely a result of the increase in trace-metal levels
shown in Figure 4. It is well known that the presence of metal contamination
in silica-based alkaline slurries can cause adjacent silica particles
to aggregate through dehydration reactions between the hydroxyl groups
on the surfaces of the silica particles and metal hydroxides, as depicted
in Figure 10.
9: ILD removal rate as a function of mean aggregate particle size.
only about 15% of the slurry introduced onto the surface of the pad
actually enters the pad-wafer interface, an analysis of the raw shear
force data shown in Figure
11 indicates that the stick-slip properties of the slurry change
dramatically as the slurry is recycled. This phenomenon is evident in
the shear force data over time for fresh as well as recycled slurries
(the top spectra in Figure 11), where the shear force associated with
the fresh slurry has significantly higher variation than its recycled
counterparts. This observation supports the postulation that an increase
in aggregate particle size and the subsequent decrease in the contact
area between the slurry's abrasive particles and the wafer surface reduce
the stick-slip between the two bodies. The results of integrating the
shear force spectra in the frequency domain (the bottom spectra in Figure
11) are presented in Figure 12. These results clearly indicate that
the extent of the stick-slip associated with recycled slurries is 2–4
times less than that associated with fresh slurry, which reduces their
effectiveness in ILD removal applications.
10: Possible mechanism accounting for aggregate particle-size growth
in metal-contaminated silica-based alkaline slurries.
CMP polishing experiments were performed to analytically and functionally
characterize recycled fumed silica slurries after multiple CMP processes.
The effects of multiple polishes on viscosity, specific gravity, and
pH were negligible, but mean aggregate size, trace metals, and large-particle
counts were notably affected. Utilizing depth filtration to regenerate
the used slurry affected LPCs and possibly particle mean aggregate size.
Slurry filtration reduced the large particle counts of the used slurries
by more than a factor of 100. The removal rate decreased nearly 40%
after the slurry was recycled five times, while filtration had only
a minor impact on that decrease. Interestingly, COF data showed a nearly
perfect correlation to removal rate, indicating that repeated slurry
recycling and filtration causes COF to decrease significantly.
12: ILD removal rate as a function of gamma for fresh slurry, slurry
recycled once, slurry recycled three times, and slurry recycled
is likely that the increase in mean aggregate particle size, which lowers
the contact area between the abrasive particles and the wafer, had some
impact on removal rate results. This postulation was further supported
by the fact that the extent of stick-slip phenomena associated with
the process decreased by a factor of 2 to 4 when recycled slurry was
authors wish to express their gratitude to John Cheney, Stu Sawai, and
Keishi Seki of Fujimi for their support. The work presented in this
article was supported financially by the NSF/SRC Engineering Research
Center for Environmentally Benign Semiconductor Manufacturing.
TFA Bibby et al., "CMP COO Reduction: Slurry Reprocessing,"
Thin Solid Films 308–309 (1997): 538–542.
H-J Kim, D-H Eom, and J-G Park, "Physical and Chemical Characterization
of Reused Oxide Chemical Mechanical Planarization Slurry," Japanese
Journal of Applied Physics 40, part 1, no. 3a (2001): 1236–1239.
H Kodama, "A Reclaim Use of CMP Slurry," in Proceedings
of the 29th Symposium on ULSI Ultra Clean Technology (Tokyo: Ultra
Clean Society, 1996), 67–73.
S Olsen, "Tribological and Removal Rate Characterization
of ILD CMP," master's thesis, University of Arizona, 2002.
A Philipossian and E Mitchell, "Performing Mean Residence
Time Analysis of CMP Processes," MICRO 20, no. 7 (2002):
D Capitanio et al., "Defect Reduction during Chemical Mechanical
Planarization by Incorporation of Slurry Filtration" (paper presented
at the Workshop on Contamination in Liquid Chemical Distribution Systems,
San Francisco, CA, July 13–15, 1998).
E Brigham and H Oren, The Fast Fourier Transform and Its Applications
(Inglewood Cliffs, NJ: Prentice-Hall, 1988).
F Preston, "The Theory and Design of Plate Glass Polishing
Machines," Journal of the Society of Glass Technology
11 (1927), 214–256.
M-S Kim, "A Study on CMP Process by Regenerated Oxide Slurry
Using Filter Modules," master's thesis, Hangyang University, Korea,
N Brown, C Baker, and R Maney, "Optical Polishing of Metals,"
in Proceedings of SPIE 306 (Bellingham, WA: SPIE, 1982), 42–51.
Philipossian, PhD, has been the Koshiyama associate professor
of planarization in the department of chemical and environmental engineering
at the University of Arizona (Tucson) since January 2000. His current
areas of research include lubrication and wear, fluid dynamics, consumables,
equipment characterization and design, and thermal modeling as related
to various aspects of planarization and postplanarization cleaning processes.
From 1992 to 2000, Philipossian was a materials technology manager at
Intel, where he was responsible for the development and characterization
of planarization and postplanarization cleaning consumables, low-k dielectrics,
and electroplating chemicals. Philipossian is the author or coauthor
of more than 70 journal publications and more than 100 conference papers.
He holds 12 patents in the area of semiconductor processing and device
fabrication. He received BS, MS, and PhD degrees in chemical engineering
from Tufts University in Medford, MA. (Philipossian can be reached at
520/621-6101 or firstname.lastname@example.org.)
Shadman, PhD, is professor of chemical engineering with joint
appointment in optical sciences at the University of Arizona and is
also the director of the NSF/SRC Engineering Research Center for Environmentally
Benign Semiconductor Manufacturing. Before joining the University of
Arizona in 1979, he was a research engineer at the General Motors Research
Laboratory. Shadman specializes in chemical reaction engineering, particularly
as it applies to advanced material processing and semiconductor fabrication
technology. A fellow of the American Institute of Chemical Engineering,
he received a PhD in chemical engineering from the University of California,
Berkeley, in 1972. (Shadman can be reached at 520/621-6052 or email@example.com.)
Levy has worked at Pall for the past 15 years, most recently
as a project engineer. In this role, he has provided technical support
to the CMP industry. He is studying the mechanisms of particle capture
and transmission in filtration media. Previously, Levy worked as a visiting
research scientist at the University of Arizona, where his research
centered on the characterization of fresh, spent, and reprocessed fumed
silica slurries and filtration and separation technology. He received
a BS in physics from the State University of New York at Albany in 1988.
(Levy can be reached at 516/801-9279 or firstname.lastname@example.org.)
Tousi, PhD, is a senior vp of Pall and global director of the
company's scientific and laboratory services (SLS) department. He began
his career with the company in 1986 as a staff scientist and has held
many positions in the SLS department. Tousi helped pioneer Pall's global
technical support structure, which emphasizes regional laboratories
near the company's customer base. In this role, he directs scientific
and technical support efforts and leads the expansion of laboratories
into new areas. Additionally, he has responsibility for the quality
assurance and regulatory affairs department. Tousi performed early work
in aerosol monitoring for gas applications in the semiconductor industry
and has been widely published, presenting papers at conferences and
committees. He received BS, MS, and PhD degrees in chemical engineering
from the University of Tulsa in Tulsa, OK. (Tousi can be reached at
516/801-9407 or email@example.com.)
Gotlinsky, PhD, is vp of Pall's process technologies scientific
and laboratory services group, where he is responsible for global applications
support of filtration and purification technologies. He has been involved
in the semiconductor industry for more than 20 years, has published
numerous papers, and has given many presentations. He received a PhD
in chemistry from the City University of New York. (Gotlinsky can be
reached at 516/801-9260 or barry_gotlinsky
Scott Rader, PhD, is the R&D manager at Fujimi. For the
past seven years, he has been involved in the development and manufacturing
of polishing slurries used in high-technology applications.
Before joining the company, he worked in the R&D group at O.I. Analytical,
where he focused on flow injection and continuous-flow analysis. Rader
holds five patents in the area of polishing formulations. He received
a BS from George Fox University in Newberg, OR, and a PhD in inorganic
chemistry from the University of Nevada in Reno. (Rader can be reached
at 503/972-9440 or firstname.lastname@example.org.)
Lefevre has been CMP business development manager at Fujimi
since 2001. Before that, he worked at IBM for 10 years and at International
Sematech as an IBM assignee for 2 years. A member of the Materials Research
Society and the Electrochemical Society, he has contributed regularly
to technical conferences on copper CMP. In 1990, he received
an MS in industrial engineering from the Ecole Nationale Supérieure
d'Arts et Métiers in Paris. (Lefevre can be reached at 503/579-9479
Koshiyama is the former chairman and CEO of Fujimi. He joined
the company in 1964 as an R&D manager and was general manager of
the R&D department between 1970 and 1980. Between 1979 and 1991,
he was the president of Fujimi Abrasive Sales, and between 1988 and
2003, he was chairman and CEO of Fujimi. Koshiyama received numerous
prestigious awards, including an award from the minister of the national
science and technology board and an R&D award from Japan Industrial
Paper. He received an MS from Meiji University in 1962 and a BS from
Hosei University in 1960. In 2005, he submitted a doctoral dissertation
to Chubu University in Kasugai City, Japan, and is awaiting receipt
of his doctoral degree. (Koshiyama can be reached at +81 52 5038181