Novel Applications of Gas-Phase Analytical Methods to
Semiconductor Process Emissions
Brian Goolsby and Victor H. Vartanian
Motorola, Semiconductor Products Sector
3501 Ed Bluestein Blvd.
M/D K10
Physical Analysis Laboratory
Austin, TX 78721
Abstract. The semiconductor industry currently faces technical challenges in transistor design as traditional
materials used for decades are being driven to their physical limits. High-k materials (k>7 for Si3N4) are
being developed as gate oxides for sub 100 nm MOSFETs to prevent electron tunneling between source and
drain. Organometallic precursors under consideration could produce hazardous byproducts. Low-k
materials (k<3.9 for SiO2) are being developed as insulators or barriers in the dielectric stack to reduce RC
time delays and cross talk between adjacent conductors. Precursors containing carbon or fluorine may
increase the emission of CF4 during chamber cleans. Heavily doped polysilicon or metals currently in use as
gate electrodes may be replaced with metals or metal oxides having greater corrosion resistance or other
advantageous properties. All of these new materials must be characterized from the standoint of process
byproduct emissions and abatement performance. Gas-phase analysis is critical to the safe and timely
incorporation of these novel materials. Several new applications of Fourier transform infra-red spectroscopy
(FTIR) are presented, including techniques being applied to address some of the current challenges facing
the semiconductor industry. This report describes the characterization of various chemical vapor deposition
(CVD) processes. Applications of gas-phase analytical methods to process optimization are also described.
compatibility with substrates, which helps in decisionmaking
about
tool
throughput,
abatement,
environmental, health, and safety (EHS) and
infrastructure impacts, as well as appropriate process
control mechanisms.
INTRODUCTION
Traditional scaling techniques used for shrinking
semiconductor devices are reaching a critical state. A
consequence of making all parts of a chip smaller, as
described in Moore’s law, is that circuit components
are also placed closer together, posing serious
electrical problems for common semiconductor
materials [1]. In an effort to create materials that
display the desired physical properties at smaller
dimensions, a large number of chemicals, or
“precursors” are being studied in chemical vapor
deposition (CVD) processes [2]. Along with these
new precursor molecules arises the need to
characterize the processes that utilize them. Thorough
characterization will speed their acceptance and
integration by providing process windows with respect
to film growth rates, purity levels, by-products, and
While post-processing metrology provides a wealth
of information, real-time monitoring with gas-phase
analytical instruments provides the most thorough
picture of what happens in the process chamber. For
example, the identification and quantification of
gaseous CVD reaction by-products can not be
accomplished by measuring a film on a wafer.
Furthermore, characterization of components in a
CVD chamber’s exhaust makes it possible to fine-tune
recipes to minimize potentially harmful emissions,
thus reducing safety and environmental concerns.
This includes the emission of gases that have a high
CP683, Characterization and Metrology for ULSI Technology: 2003 International Conference,
edited by D. G. Seiler, A. C. Diebold, T. J. Shaffner, R. McDonald, S. Zollner, R. P. Khosla, and E. M. Secula
© 2003 American Institute of Physics 0-7354-0152-7/03/$20.00
611
deposition process in progress, and the chamber clean run
immediately after the deposition.
global warming potential (GWP), which the
semiconductor industry has made a commitment to
reduce. The same analytical techniques used to
characterize a CVD process can be use to determine
the effectiveness of a particular point-of-use (POU)
abatement device.
In addition to CVD, the chamber clean step after
deposition of a low-k film has been a major focus for
improvement. Recipe optimization to decrease NF3
use has been accomplished through the testing of
numerous recipes designed to better utilize the reactive
fluorine species. FTIR is an ideal tool for this
application because it provides not only an endpoint
marker, but also an emissions profile, which can be
integrated to determine the global warming impact of a
particular recipe. This property is a function of a
molecule’s ability to absorb infrared radiation as well
as its atmospheric lifetime. An example of this is
shown in Figure 2, where spectra from two different
clean recipes are compared. The optimized recipe has
much lower CF4 output, cutting its overall
fluorocarbon emissions significantly.
Extractive Fourier transform infrared (FTIR)
spectroscopy has tremendous utility as an analytical
technique in semiconductor process characterization.
Some compelling features include its ability to
quantify a wide range of analyte concentrations,
uniquely identify numerous gas-phase organic or
inorganic compounds, and provide rapid analysis
relative to the time-scale of most processes. In
addition, the technique is transparent to normal tool
operation, and can be implemented well downstream
of a process chamber, typically sampling just after the
roughing pump. While FTIR is by no means a novel
technique, it is under-utilized for on-tool applications
and real-time data collection for process development.
Several new applications of extractive FTIR to recent
advances in semiconductor processing are presented
herein.
Std. SiCOH Clean
Species
scc
C2F6
2.378795
CF4
78.19469
CO
16.59687
CO2
23.56738
COF2
73.28214
HF
1441.702
NF3
22.96662
SiF4
254.3715
1.2
1
.8
ADVANCED PROCESSES
.6
.4
Low-k CVD
.2
The chemical vapor deposition of several low-k
films containing Si, C, O, and H in various
stoichiometric ratios has been characterized with
FTIR. The spectrum in Figure 1 shows an example of
this application. Note that several bands indicative of
absorbance by hydrocarbon reaction by-products can
be identified using this spectrum, and the utilization
efficiency of the precursor material, used to make the
film, can be measured.
0
4500
4000
3500
3000
1
.8
.6
.4
Absorbance
1.5
.2
.5
C2H4
precursor
2000
1500
1000
CF4
Optimized Clean
Species
scc
C2F6
0.629538
CF4
18.24143
CO
68.90243
CO2
65.23438
COF2
74.69355
HF
1356.296
NF3
11.66531
SiF4
256.0147
Low-k precursor spectrum
Deposition spectrum
C2H4
2500
Optimized Recipe
1.2
2
1
CF4
Std. BKM
0
4500
4000
3500
3000
2500
2000
1500
1000
C2H4
0
-.5
Clean spectrum
HF
-1
4500
4000
3500
precursor
3000
CO2
2500
CF4
precursor
2000
1500
SiF4
NF3
FIGURE 2. The top spectrum shows the emissions profile
from a standard chamber clean recipe, while the lower
spectrum is from an optimized process. CF4, a molecule
with a high GWP, has been reduced significantly in the
optimized process.
1000
Wavenumber (cm-1)
FIGURE 1. FTIR spectra collected during CVD of a low-k
film using a silicon and hydrocarbon precursor. The spectra
(from top to bottom) represent the precursor only, the
612
fluorocarbon (UFC) compounds were evaluated for
global warming emissions and process performance on
a medium density etch chamber for silicon oxide. A
wide range of IR-absorbing perfluorocarbons can be
identified using an FTIR spectrum from a dielectric
etch process. A series of such spectra can then be used
to calculate the global warming potential of the entire
process.
High-k Deposition
Atomic layer deposition (ALD) represents a major
development in the ability to grow high-k dielectric
films with good control of uniformity and
composition. The growth of one monolayer at a time
using ALD provides the control necessary to interleave
different materials together even in very thin films.
The extremely low quantity of chemical precursor
consumed in an individual cycle, resulting in the
growth of one atomic layer, is a benefit to process
engineers but would be expected to pose a challenge
for downstream detection of reaction byproducts.
Figure 3 shows a series of measurements made with
extractive FTIR, demonstrating the ability of this
technique to resolve miniscule pulses from the tool’s
chemical delivery system. The plot shows by-products
(CH4 and HCl) from two different precursors being
tracked, showing how the layers are interleaved. For
example, one of the precursors, a metal chloride, is
converted to HCl in a reaction at the wafer’s surface.
The top line in Figure 3 shows how the quantity of
HCl in the chamber effluent changes with each of 15
pulses. These types of data may also be used as a
diagnostic for successful delivery of a precursor.
Unexpected disappearance of one of the reaction byproducts from the FTIR spectrum might indicate that
the chemical ampoule is empty or that a delivery line
is clogged.
5900
6000
6100
6200
6300
6400
6500
0
6600
c-C4F6
HF13B
5
HF2B
0.5
c-C5F8
10
C3F6
1.0
c-C4F8O
15
OF2B
CH4 Concentration (ppm)
20
1.5
HCl Concentration (ppm)
25
C4F10
30
2.0
0.0
5800
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
c-C4F8
2.5
The test processes and their emissions results are
compared to a traditional, baseline (PFC)
perfluorocompound-based process. For oxide etching,
global warming emissions reduction as high as 88%
were attained compared to a C3F8-based process, with
similar process performance (etch rate and via profile)
as determined by scanning electron microscopy (SEM)
of the via cross-section [3]. In the C3F8 process, a
large percentage of the total emissions (>50%) are due
to unreacted C3F8, a high GWP feed gas. By simply
switching to a more reactive, lower-GWP gas such as
hexafluoro-1,3-butadiene,
the
emissions
from
unreacted feed gas are eliminated almost entirely. In
addition the C4F6 process resulted in lower CHF3
emissions, which is likely due to lower photoresist
erosion, as photoresist is the major source of hydrogen
for the formation of CHF3. Results from several of the
tested compounds are presented in Figure 4.
FIGURE 4. Global warming potential (GWP) reductions
relative to C3F8 in identical processes using 9 alternative
gases. Lightly shaded bars represent the overall percent
reduction; dark bars are normalized reduction based on etch
depth.
Time (sec)
FIGURE 3. Extractive FTIR sampling downstream of an
ALD process’s exhaust pump is able to resolve byproducts
from short pulses of two different precursors. Top line is
HCl, lower is CH4.
Etch Recipe Development for Novel Gate
Electrodes
Dielectric Etch of Via Patterns Using
Unsaturated Fluorocarbons
Ruthenium and ruthenium oxide have been
proposed as metal oxide gate electrodes because they
are thermally stable and resistant to corrosion, but as
with most materials, there are process and integration
In an effort to develop via etch chemistries with a
reduced environmental impact, several unsaturated
613
other species. The collected data were used in
conjunction with quantitative FTIR data to generate a
mass balance for the process.
issues that must be addressed. Characterization of
early process attempts can provide valuable data to
this end. The chemical environment and plasma
conditions during which the toxic byproduct
ruthenium tetroxide (highly injurious to eyes, inflames
mucous membranes, and causes respiratory distress) is
formed, can be determined by FTIR [4]. A series of
attempts were made to etch RuO2 films with various
plasma chemistries. The FTIR spectrum shown in
Figure 5 was collected during an Ar/O2 recipe,
revealing the possible formation of RuO4 for this test
case.
Concentration (ppm) .
1.2
1
0.8
0.6
0.4
0.2
0.020
922 cm-1
likely identifies RuO4
0.015
0.010
0.005
0
0
-0.010
-0.015
90 sccm O2
110 sccm Ar
4 mTorr
75 W bias
500 W source
90 sec
-0.025
-0.030
-0.035
4000
3500
3000
2500
40
60
80
100
120
140
FIGURE 6.
Fluorine concentrations in tool exhaust
measured by a fluorine chemical sensor (FCS) during a
chamber clean.
-0.005
-0.020
20
Scan # (1 sec / scan)
residual HF from a
different etch recipe
0.000
Absorbance
1.4
CONCLUSIONS
2000
1500
1000
Analytical systems on semiconductor processing
tools provide real-time data that extends beyond
simple endpointing.
Characterization of mature
processes can lead to a better understanding of
potential environmental impacts, and in some cases,
possible improvements can be identified. Moreover,
the early learning provided by capable techniques such
as FTIR can ease the development and integration of
new materials, which are critical to the next generation
of semiconductor devices.
Wavenumber (cm-1)
FIGURE 5. FTIR spectrum collected during an attempt at
etching RuO2.
COMPLEMENTARY ANALYSIS
TECHNIQUES
A thorough process picture requires quantitative
techniques capable of identifying or measuring all byproducts. This can not necessarily be achieved with
just one technique. For example, ionized plasma
species are short-lived, and thus more easily
characterized in situ using optical emission
spectroscopy (OES), while homoatomic dimers (N2,
O2, etc.) are undetectable by typical FTIR
spectroscopy. Strongly absorbing species in complex
mixtures may overwhelm other features in an FTIR
spectrum. For a critical gas such as fluorine (F2),
which is present in abundance in chamber clean
processes, an alternative measurement technique is
needed.
ACKNOWLEDGMENTS
The authors gratefully acknowledge contributions to the
preceding work by: Laurie Beu, Laura Mendicino, Dina
Triyoso, Darrell Roan, and Terry Sparks of Motorola,
Semiconductor Products Sector, APRDL. Simon Karecki,
Ritwik Chatterjee*, and Rafael Reif of Massachusetts
Institute of Technology.
Curt Laush of URS Corp.
*Currently at Motorola.
REFERENCES
A commercially available fluorine chemical sensor
(FCS) was exposed to exhaust from a CVD
manufacturing tool during a cleaning process (data
shown in Figure 6). This portable measurement device
relies on fluorine’s photoemissive reaction with an
organic substrate for operation, and it displayed ppb
sensitivity, fast response, and no interferences from
1. Moore, G. E., Electronics 38, pp. 114-117 (1965).
2. International Technology Roadmap for Semiconductors
2001 Edition, WWW Document, (http://public.itrs.net).
614
3. Chatterjee R., Karecki S., Reif R., Sparks T., Vartanian
V. H., Goolsby B. J., “The Evaluation of
Hexafluorobenzene as an Environmentally Benign
Dielectric Etch Chemistry” in Environmental Issues with
Materials and Processes for the Electronics and
Semiconductor Industries V, edited by L. A. Mendicino,
199th Electrochemical Society Conference Proceedings,
Pennington, New Jersey, 2002, pp. 99-113.
4. Sax’s Dangerous Properties of Industrial Materials,
edited by R. J. Lewis, Wiley, New York, 10th ed., 2000,
p. 3172.
615