New Particle Metrology for CMP Slurries
S. Kim R. Williams1, Ilyong Park1, Edward E. Remsen2, and Mansour Moinpour3
1
Chemistry and Geochemistry, Colorado School of Mines, 1500 Illinois St., Golden, CO, 80401
2
Cabot Microelectronics Corporation, 870 North Commons Drive, Aurora, IL, 60504
3
Intel Corporation, 2200 Mission College Blvd., Santa Clara, CA, 95054
ABSTRACT
A new particle sizing and counting method based on the coupling of flow field-flow
fractionation (FFF) with dual-sensor, single particle optical sensing (SPOS) detection is reported.
The integration of FFF and SPOS systems was accomplished by means of a dilution interface
that preserved the resolution of FFF-separated particles. Analysis of a model mixture of
polystyrene latex standards of different diameters established that the FFF-SPOS system can
resolve particles into discrete peaks for subsequent particle counting. Application of this method
for the analysis of a colloidal silica standard demonstrated its use for materials commonly
employed as CMP abrasives. Further development and refinement of the technique will enable
compositional and structural analyses of heterogeneous large particle populations constituting
commercial CMP slurries.
INTRODUCTION
A central tenet of integrated circuit manufacture states that the larger particle fraction in
chemical mechanical planarization (CMP) slurries used to process surface films on silicon wafers
produces micro-scratches on commercial microelectronic devices.1 The most compelling
evidence for this assertion comes from experimental studies2, 3, 4 demonstrating the correlation
between the large particle count (LPC) and wafer scratch defect counts. However, the chemical
and structural nature of these particles can not be determined from SPOS measurements4, 5
commonly used to determine the LPC. This lack of detail prevents the assignment of specific
large particle types in the LPC as micro-scratch generators. In addition, a relative ranking of the
micro-scratching potential for the different large particle types is also beyond the reach of
conventional SPOS techniques.
Findings reported in the present study for model particle mixtures indicate that a new
method combining size separation of particles via flow field-flow fractionation (FFF)6 and
particle detection using dual-sensor SPOS can be applied in the detailed analysis of the larger
particle populations typically found in CMP slurries.
The FFF separation utilizes laminar flow conditions to induce a parabolic flow profile
with its characteristic maximum flow rate at the center of the channel and decreasing flow
velocities towards the channel walls. The FFF process commences with the transport of
particulate sample to a membrane accumulation wall by a crossflow of fluid (see Figure 1A).
Since the closest distance of approach to the wall is limited to one particle radius, small particles
will have a center of mass that is closer to the wall than large particles and will be transported to
the detector at a lower flow velocity. Consequently, small particles elute later than large
particles in steric mode FFF.7
EXPERIMENTAL
Test samples were dilutions of stabilized suspensions of size-certified, spherical
polystyrene latex beads ranging in diameter from 0.3 to 3 µm; and a 0.73 µm colloidal silica
standard (Duke Scientific Inc., Palo Alto, USA). Polystyrene standards with diameters < 1 µm
were diluted 1000x and standards ≥ 1 µm in diameter were diluted 100x with carrier liquid.
Flow FFF experiments were performed using a Model F-1000-FO universal fractionator
(PostNova Analytics Inc., Salt Lake City, USA). Two pumps were employed to drive the
channel and crossflow through the channel. The channel was outfitted with a Mylar spacer
which provided a channel thickness of 127 µm and a 0.1 µm pore size
poly(ethyleneterephthalate) (PETP) membrane. The FFF channel’s carrier and crossflow fluids
were 0.22 µm filtered, deionized water with 0.1% (v/v) FL-70 (Fisher Scientific, Fair Lawn,
USA). Typical flow rates for the carrier and crossflow fluids were 4.9 and 1.7 mL/min,
respectively. The sample volumes injected were 12 µL. The separated particles were monitored
as they eluted from the flow FFF channel using an on-line UV detector and two, size bincalibrated SPOS sensors Model Liquilaz S03 and S05 (Particle Measurement Systems Inc.,
Boulder, USA).
Figure 1. Schematic illustration of (A) steric mode flow FFF separation mechanism and (B) the
FFF-SPOS system.
The FFF-SPOS schematic diagram is shown in Figure 1B. The two instruments were
connected using a three way switching valve (Upchurch Scientific, Oak Harbor, USA) or a PMS
home-built interface. The latter consisted of a 1/4 inch tube with a 40 mil constriction in the
middle. The inlet and outlet were 1/4 inch fittings and the sample inlet on the upstream side of
the constriction was a 2 mm tube. Both interfaces allowed the continuous addition of filtered
deionized water to the stream of size-separated particles eluting from the flow FFF channel. This
dilution minimized the presence of multiple particles in the SPOS detector’s viewing area, and
consequently, the frequency of coincident particle counting which leads to erroneously large
particle sizes. The sum of the FFF channel and the diluent flowrates was set to match the 80
mL/min used in the factory-calibration of the S03 and S05 sensors.
Data acquisition for the SPOS sensors was provided by program FACILITY NET (Particle
measurement Systems Inc., Boulder, USA). Exported SPOS sensor data were reduced and
analyzed using program ORIGIN (version 7.5, OriginLab Corp., Northampton, USA).
RESULTS AND DISCUSSION
8000
3.0
0.966 µm
7000
6000
S03
S05
UV
5000
4000
3000
2000
2.0
1.5
1.0
2.062
0.5
1000
0.0
0
-1000
2.5
UV Response (mV)
Normalized Counts (#/mL)
The steric mode separation of polystyrene size standards is demonstrated in Figure 2 with
the larger particle eluting first. Overlapping fractograms for the UV and the two SPOS detector
responses were also found. In the case of the SPOS sensors, the cumulative particle count was
used to generate the fractograms. Cumulative particle counts included all particles between 0.3
and 3 µm; and 0.5 to 20 µm for the S03 and S05 sensors, respectively.
Near perfect overlap of UV and SPOS detector responses confirmed that the dilution of
FFF-separated particles in the FFF-SPOS interface did not broaden or time-shift the fractograms.
As a result, correction of UV and SPOS fractograms for these effects was not required in order to
directly compare the fractograms.
It is interesting to note that the relative heights of the two size standards are different for
the UV and the SPOS responses. This occurs as a result of large particles scattering more light
and yielding a higher intensity UV signal. In other words, the measured UV signal does not
provide accurate quantitative results for PS particles.
-0.5
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Time (min)
Figure 2. Fractograms produced by the FFF-SPOS system for 2.062 µm and 0.966 µm
polystyrene latex bead standards. Results are shown for both the UV (black line) and
SPOS detectors (red and blue lines).
8000
4000
2000
0
0.4
( m
)
0.3 t
o
Normalized
6000
Counts (#/m
L)
The resolving power of the FFF-SPOS system was further investigated by analyzing a
10,000 particle mixture composed of eight polystyrene latex standards ranging in diameter from
0.3 to 3.09 µm. Experiments were performed using SPOS only and FFF-SPOS. The identical
system was used in both sets of experiments with the exception that in the former case, the flow
FFF channel was replaced by a tube of similar volume. Unlike the data sets shown in Figure 2,
normalized particle counts in individual particle size bins were also examined. This analysis
produced 3D plots of normalized particle count versus bin size (particle diameter) versus FFF
elution time, as shown in Figures 3 and 4. In the absence of the flow FFF channel, there was no
retention or spatial separation of the particles (Figure 3). The single peak, obtained across the
time axis, is deconvoluted into different particle sizes by the SPOS. By contrast, flow FFF
produced spatially resolved peaks for the individual components of the mixture. Each size
component is then analyzed by SPOS. It is important to note that the FFF process distributed
the 10,000 particle sample among different size bins over the span of several minutes.
Consequently, samples of higher particle concentration can be analyzed by FFF-SPOS without
counting coincidence error and low levels of large particles in complex particle mixtures can be
readily detected.
e
Bi
n
Siz
e
Tim
12
:0
2:
00
Elu
tion
12
:0
1:
36
12
:0
1:
12
12
:0
0:
48
12
:0
0:
21
µ
2.5 t
o
3
Figure 3. 3D plot (elution time vs. bin size vs. normalized particle counts) for a 10,000 particle
mixture of eight polystyrene latex standards with an open tube instead of the flow
FFF channel (equivalent to zero cross flowrate or no separation). S03 sensor (0.3 to 3
µm) data are shown.
/mL)
Counts (#
200 0
1000
500
0
to 0
.4
( m
)
0.3
Normalize
d
1500
Elut
ion
T
Siz
e
Bin
11
:1
4:
24
11
:1
5:
12
11
:1
6:
0
1
11
:1
6:
49
11
:1
7:
38
11
:1
8:
28
11
:1
9:
18
11
:2
0:
07
11
:2
0:
53
µ
2.5
to
3
ime
Figure 4. 3D plot (elution time vs. bin size vs. normalized particle counts) for a 10,000 particle
mixture of eight polystyrene latex standards under applied field conditions (crossflow rate = 1.7 mL/min). S03 sensor data (0.3 to 3 µm) are shown.
The direct applicability of the FFF-SPOS method to the types of particles characteristic
of CMP slurries was demonstrated by the analysis of a colloidal silica standard. As shown in
Figure 5A, a 0.73 µm colloidal silica exhibited a UV detector elution time that was intermediate
between the values found for 0.806 and 0.596 µm polystyrene latex standards. The calculated
diameter was 0.73 µm based on the FFF retention time. However, the SPOS particle size
distribution for the silica standard (Figure 5B) yielded a diameter of 0.44 µm.
(A)
PS 0.806 µm
Normalized Counts (#/mL)
UV Detector Response (mV)
2
Silica 0.73 µm
1
PS 0.596 µm
0
0
1
2
3
4
5
Time (min)
6
7
8
0.44 µm
7000
(B)
6000
5000
4000
3000
2000
1000
0
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
1.1
Diameter ( µm)
Figure 5. (A) Overlay of fractograms for polystyrene latex and a colloidal silica standards.
(B) S03-based particle size distribution for the colloidal silica standard. Data (blue
bars) and Gaussian fit to the data (red line) are shown.
This apparently anomalous result is not an artifact due to calibration error for the SPOS
sensor, but is the product of silica’s less intense light scattering relative to the light scattering for
a polystyrene latex bead of the same diameter. Previous SPOS characterizations of fumed silica
CMP slurries have established that this effect is due to the higher refractive index (R.I.) of
polystyrene relative to the R.I. of silica.4 As a result, polystyrene latex-based calibration of the
size bins of a SPOS system, as was performed in the present study, underestimates the true
diameter of silica particles. Conversion of the polystyrene calibration to a silica sphere basis can
be used to correct the apparent SPOS size to a true particle size.4
CONCLUSIONS
Analyses of mixtures of polystyrene latex standards demonstrated conclusively that a
flow field-flow fractionation (FFF) separation can be successfully interfaced with dual-sensor,
single particle optical sensing (SPOS) detection. Operation of the FFF-SPOS system in a steric
FFF separation mode allowed the detection and quantification of polystyrene latex standards
ranging in diameter from 0.3 to 3 µm. Using the multi-channel capability of the SPOS sensors,
FFF-separated mixtures of latex standards of different diameters can be resolved into discrete
peaks corresponding to the individual standards in the mixture. Successful extension of the
technique to the analysis of colloidal silica confirmed that materials characteristic of the
abrasives used in CMP slurries can be analyzed via FFF-SPOS.
ACKNOWLEDGMENTS
The authors gratefully acknowledge the project support provided by the following corporate
sponsors: Intel, Cabot Microelectronics, Fujimi, DA Nanomaterials, Planar Solutions and
Particle Measurement Systems.
REFERENCES
1. J.M. Steigerwald, S.P. Muraka and R.J. Gutmann, Chemical Mechanical Planarization of
Microelectronic Materials (John Wiley & Sons, New York, 1997).
2. L.H. Hanus, S.A. Battafarano and A.R. Wank, MICRO 21, 71 (2003).
3. K. Nicholes, M.R. Litchy, E. Hood, W.G. Easter, V.R. Bhethanabotla, L.Cheema and
D.C. Grant, Proc. of the 8th CMP-MIC Conference, (2003), pp. 221-223.
4. E.E. Remsen, S. Anjur, D. Boldridge, M. Kamiti, S. Li, T. Johns, C. Dowell and
P. Feeney, J. Electrochem. Soc. 153, G453-G461 (2006).
5. J.P. Bare and T.A. Lemke, MICRO 15, 53 (1997).
6. J.C. Giddings, Science 260, 1456 (1993).
7. K.D. Jensen, S.K.R. Williams and J.C. Giddings, J. Chromatogr. 746, 136 (1996).