Dual Emission Laser-Induced Fluorescence (DELIF) in Chemical Mechanical Planarization 1 Apone, D., 1 Gray, C., 1 Rogers, C., 1 Manno, V., 2 Philipossian, A., 3 Anjur, S., 4 Moinpour, M., 5 Barns, C. 1 Tufts University Medford, MA 2 University of Arizona, Tucson, AZ 3 Cabot Microelectronics, Aurora, IL 4 Intel Corporation, Santa Clara, CA 5 Intel Corporation, Hillsboro, OR 1 Abstract In this paper, we present a review of the measurements that Dual Emission Laser Induced Fluorescence (DELIF) has allowed us to make over the last 8 years. We have been able to measure many features of the slurry film between the wafer and the pad during polishing while simultaneously measuring the frictional drag. In particular, we measured slurry mean residence time, slurry temperature, and the slurry film thickness. All experiments are performed on a 1:2 scaled tabletop rotary polisher with variable pad speed, wafer downforce control, and in-situ conditioning. Mean residence times are strongly affected by conditioning. Slurry film thickness (about 20 μm on average, and accurate to within 1 μm) was a strong function of macro wafer curvature as well as downforce. Slurry temperature (accurate to 1 ◦ C) increase with downforce and correlate closely to frictional drag. The resulting data are discussed for wafers polished with a 3.1 wt percent abrasive concentration slurry solution on flat Freudenberg FX-9 polishing pads. Correlations for friction and temperature are discussed, as well as thickness versus pressure and rotation rates. 1
32
Embed
Dual Emission Laser-Induced Fluorescence (DELIF) in Chemical Mechanical · PDF file · 2007-05-22Dual Emission Laser-Induced Fluorescence (DELIF) in Chemical Mechanical Planarization
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Dual Emission Laser-Induced Fluorescence (DELIF) in ChemicalMechanical Planarization
1Apone, D.,1Gray, C.,1Rogers, C.,1Manno, V.,2Philipossian, A.,3Anjur, S.,4Moinpour, M.,5Barns, C.
1Tufts University Medford, MA2University of Arizona, Tucson, AZ3Cabot Microelectronics, Aurora, IL4Intel Corporation, Santa Clara, CA5Intel Corporation, Hillsboro, OR
1 Abstract
In this paper, we present a review of the measurements that Dual Emission Laser Induced Fluorescence (DELIF) has
allowed us to make over the last 8 years. We have been able to measure many features of the slurry film between the
wafer and the pad during polishing while simultaneously measuring the frictional drag. In particular, we measured
slurry mean residence time, slurry temperature, and the slurry film thickness. All experiments are performed on
a 1:2 scaled tabletop rotary polisher with variable pad speed, wafer downforce control, and in-situ conditioning.
Mean residence times are strongly affected by conditioning. Slurry film thickness (about 20µm on average, and
accurate to within 1µm) was a strong function of macro wafer curvature as well as downforce. Slurry temperature
(accurate to 1C) increase with downforce and correlate closely to frictional drag. The resulting data are discussed for
wafers polished with a 3.1 wt percent abrasive concentration slurry solution on flat Freudenberg FX-9 polishing pads.
Correlations for friction and temperature are discussed, as well as thickness versus pressure and rotation rates.
1
2 Introduction
Chemical Mechanical Planarization is a polishing process widely used in the production of integrated circuits. Tech-
nological advances have allowed the feature sizes on these circuits to decrease greatly in accordance with Moore’s
Law, which states that the number of transistors on a chip will double every 2 years. As these circuits continue to
shrink, the challenge is to obtain a smoother, flatter surface on which to deposit the metal traces that comprise the
chip. Chips are manufactured in a layered process, so each layer must be polished to a high degree before the next set
of metal traces can be added onto the silicon substrate. The work done at Tufts University has been directed at gaining
a better understanding of the nature of this polishing process by measuring slurry properties during an actual polishing
process. In particular, in this article we will present an overview of measurements of slurry thickness, temperature,
and mean slurry residence time.
Much work has been done to try to determine the exact nature of the wafer/pad contact. There are several lubri-
cation regimes that the CMP process is thought to encompass. Modelling work indicates that it is a delicate balance
between hydrodynamic lubrication, mixed solid-liquid contact and even direct solid-solid contact. [1] The hydrody-
namic lubrication occurs when the wafer is fully supported by a layer of a slurry, with no contact between the wafer
and polishing pad. Mixed solid-liquid contact is the case when the wafer is partly supported by the pressure in the
liquid layer, and partly resting on the polyurethane asperities that rise up out of the pad. Solid-solid contact would
occur most often in either static or super high downforce experimental cases, where the wafer is entirely supported by
the pad.
More complex models have begun to take into account the pad porosity and compressibility. Modelling efforts
have been employed to predict the film thickness and removal rate based on the contact regime. Higher removal rates
can be obtained with more solid-solid contact, but with a higher incidence of wafer scratching. Changing experimental
parameters to increase the film thickness will result in less scratching of the wafer, but also a lower removal rate. [2]
Other studies have shown that the pad properties change as it soaks up water. Compression tests on the pads have
shown that the elastic modulus decreases as the pad soaks longer. The soaking also increases the pad suction pressure,
which will contribute to the contact stress on the wafer. [3]
The friction in the polishing process generates heat to dissipate energy. This increase in slurry and pad temperature
2
is important for several reasons. First, the pad properties change with temperature. The pad is a porous polyurethane
material, and becomes softer with an increase in temperature. This leads to a higher percentage of the pad contacting
the wafer, which will affect removal rates and scratching. Second, the increase in temperature will speed the chemical
reactions involving the wafer and the slurry. Generally, for an increase of 10C a chemical reaction’s rate will double.
Third, the increase in temperature (above 50C) will cause the silica particles in the slurry to clump together and act as
much larger particles. The polish removal rate, however, is dominated by the increase in chemical reactions, rather than
the mechanical factors. [4] To measure the increase in temperature experimentally, an IR camera has been employed
by many research groups. This non-invasive approach allows in situ measurements of the temperature downstream of
the polishing wafer. [4] [5]
Measuring of film thicknesses is a difficult task, since their scale (50 microns) makes ordinary measurement
techniques useless. Light Induced Fluorescence (LIF) has been used to measure thicknesses, with a camera capturing
the fluorescent image. The brightness in the image is directly related to the thickness of the film. The more fluid
there is, the more dye is present, therefore a brighter image is captured. The problem is that the initial excitation
intensity may not be even across the image. This would result in inaccurate measurements. The solution is to use a
two-dye system. Initially one dye is excited, and the light emitted from it excites the second dye. Using two cameras
to capture the image (each camera fitted with a filter for one of the dyes) and then taking a ratio of the two camera
images allows for true fluid film measurements. [6] Switching to a temperature sensitive set of dyes allows for the
optical measurement of fluid temperature. [7]
3 Experimental Setup
A Struers RotoPol-31 table top polisher unit serves as a scaled version of an industrial rotary polisher such as the
SpeedFam-IPEC 472. The polisher was fitted with a Mitsubishi Freqrol frequency modulator that allowed a working
range from 20 rpm to 300 rpm±1 rpm. The polisher has a 12 inch rotating platen that carries the polishing pad (half
of the diameter of the 24 inch commercially employed pads). To ensure the scaled setup closely replicates industrial
planarization efforts, the wafer size is scaled to match the table top polisher. The ratio of the polishing pad area,Ap,
3
to the wafer area,Aw, is held constant, resulting in a 3 inch wafer simulating a 300 mm industrial wafer.
The applied wafer pressures match the IPEC 472 and are higher than current industrial levels (from 4 - 8 psi, 26
- 52 kPa) and the relative velocities of the wafer to pad is scaled as well. The rotational velocities of the wafer,Ωw,
and pad,Ωp, are matched so that the velocity under the wafer is comparable to commercial polishing. Most runs are at
polishing pad speeds of 60 rpm, the corresponding relative velocity across the wafer is approximately 0.50 m/s. This is
typical of older polishers, with more recent polishes operating closer to 1 m/s. Measured removal rates and operating
temperatures are typical of commercial CMP as well.
For experimental procedures, an array of control and measurement systems are integrated into table top polisher.
Independently controlled parameters include wafer applied pressure, wafer rotational velocity, slurry flow rate, and pad
conditioner oscillation and rotational speeds. Throughout the experimental runs, measurement systems characterize
friction, fluid film thickness, and temperature at the wafer-slurry-polishing pad interface. Figure 1 is a schematic of
the integrated system.
[Figure 1 about here.]
3.1 Polisher Head
A Sears 20 inch drill press replaces the standard wafer carrier head mechanism of the RotoPol-31. A Dayton12 HP DC
motor controller allows drill press rotational velocities to range from 0 rpm to 160 rpm. Manual velocity adjustments
are made by adjusting the motor’s amplifier. Absolute rotational velocity is calibrated with a tachometer.
A weighted traverse is integrated into the polisher head system to achieve variable applied wafer pressures. Located
atop the drill press, this system consists of a lead screw driven carriage riding on two rails. A standard 25 lb weight
mounts on top of the carriage, positioned by a Servo Systems precision stepper motor. On one end, the traverse rests on
the rotating quill that directly attaches to the wafer while the other end is free to pivot about the main drill press support
column. Thus, the load is applied directly to the wafer with minimal external moments. Carriage position, which is
controlled by a LabVIEW interface, is calibrated before a run to wafer applied pressures using a force transducer.
4
3.2 Wafer
The DELIF optical measurement technique requires transparent BK-7 glass wafers as a replacement for traditional
opaque silicon wafers. BK-7 glass was chosen because it is structurally similar to silicon, widely available, relatively
inexpensive, and disposable. These wafers have a diameter of 75 mm (3 inch) and thickness of 12.5 mm (1/2 inch),
scaled down from 300 mm wafers. A gimbal joint mounts the wafer directly to the drill press drive shaft. In general we
found that the glass wafers are not perfectly flat, but were either slightly convex, possessing a shape that conforms to
the pad surface, (2(a)) or slightly concave, possessing a non-conforming shape as shown in figure 2(b). A profilometer
is used to perform two perpendicular line scans across the wafer diameter to characterize each wafer. If they were
more than 5µm out of flat they were discarded.
[Figure 2 about here.]
3.3 Polishing Pads and Conditioning
This research exclusively uses Freudenberg’s flat FX-9 polishing pads. The FX-9 is similar to the industry standard
Rodel IC-1000 in that it is composed of a closed cell polyurethane pore structure. In addition, the FX-9 polishing pads
are dyed black for these experiments, which minimizes the pad florescence. To date, no measurable difference due to
the dying has been found; dyed and undyed pads produce the same friction data, wear at the same rate, and polish at
the same rate. [8]
During CMP, not only do polishing pad asperities compress, but also slurry debris fills and clogs pores. The result is
a ”glazed” surface, which adversely affects planarization performance. To prevent glazing, the polishing pad surface is
conditioned by a 100 grit diamond disc to remove slurry particles while simultaneously reopening pores. The diamond
disk is mounted on a rotating gimbal type PPS (polyphenylene sulfide) carrier, and rotates as it is swept radially across
the polishing pad surface (see figure 3). Sweep and rotational rates are also controlled through a software interface.
Conditioning may be either conducted ex situ, between wafer polishes, or in situ, during the polishing process. In this
work, all conditioning is done in situ. The oscillation due to the sweeping motion of the conditioning arm periodically
starves or overloads the wafer with slurry. These periodic fluctuations in the fluid lubrication can affect planarization
5
and are clearly seen in the friction measurements.
[Figure 3 about here.]
3.4 Slurry Delivery
The slurry used is Cabot Microelectronic’s Cab-O-Sperse SC-1 in a 3.1 wt% solution with small concentrations of
dyes added to obtain fluorescence measurements. Concentrations of 1.0 g/l and 0.25 g/l of Calcein and Coumarin
4, respectively, are used for thickness measurements. 0.25 g/l of 2,3-Dicyanohydroquinone and 0.5 g/l of Brilliant
Sulfaflavin are used for temperature measurements. For the Mean Residence Time measurements, slurry is tagged
with DHPN and considered ”new” slurry while slurry tagged with Coumarin is considered ”old” slurry. The slurry
consists of 100nm fumed silica particles suspended in potassium hydroxide (KOH). In these experiments, a .3 wt%
solution was used to minimize the polish. These runs behaved similarly to the 3.1 wt% cases and did not have the
wafer changing shape over the course of the run. The 3.1 wt% solution has a viscosity of 1.13 cp1 and a shear rate
of 8200 1sec . The slurry is delivered to the CMP system by a Masterflex peristaltic pump. Its non-intrusive nature is
critical for preventing contamination, corrosion, and clogging. With 14 gauge Norprene Masterflex tubing, the pump
is capable of flow rates of up to 250 mL/min. The delivery tubing and all other slurry solution containers are opaque
to prevent photobleaching of the added dyes, a condition where a dye loses its sensitivity to light after being exposed
for a period of time. Although the actual flow rate is manually dialed in, an optical encoder on the pump enables real
time flow rate monitoring in the LabVIEW interface. The flow rates are precise to within±0.5 ml/min.
3.5 System Integration
A custom National Instruments LabVIEW program functions as the interface and controller for the scale laboratory
CMP system. Each variable may be individually controlled via the software instrument panel. All operations are
performed in the following sequence in order to avoid hysteresis and ensure repeatability.