Metrology Tool for Microstructure Control on 300 mm
Wafers During Damascene Copper Processing
K. J. Kozaczek, D. S. Kurtz, P. R. Moran, R. I. Martin, L-Y. Huang, A. Stratilatov
Hypernex Inc., 3006 Research Drive, State College, PA 16801
Abstract. The rapid adoption of damascene copper processing has brought about an increased need to understand and control
microstructure in the barrier, seed and electroplated copper layers during manufacture. We have developed a fully automated
metrology tool for rapidly characterizing thin film polycrystalline microstructures on 300 mm silicon oxidized substrates. This xray based metrology tool measures crystallographic texture, phase composition, film thickness, and other microstructural
characteristics of blanket and patterned films with a throughput suitable for in-line applications. The acquired data can be used as
a direct measure of the deposition process in terms of film quality, reproducibility, and stability over time. The spatial
distribution of crystallographic texture and phase can be measured on a single wafer in order to check wafer uniformity. More
importantly, the same measurements can be carried out at predetermined intervals on wafers from a single deposition tool, and
the results used to create a database that can be applied to trend charting and tool qualification. The tool satisfies all safety,
automation, and contamination standards applicable in the semiconductor industry. Examples of applications in damascene
copper processing are presented.
precision determination of texture and a subsequent
elimination of its effects on XRD data [1,2]. By using
a large area detector, we map a significant portion of
the reciprocal space. This map is used to determine the
orientation distribution function (ODF) of crystallites
for each phase present in the film stack, for example
tantalum barrier layer, and PVD or electroplated
copper. This crystallographic approach is then coupled
with a diffraction approach. The ODF is used to
eliminate the texture from the diffraction spectrum
reducing it therefore to an equivalent spectrum of a
mixture of powders, to which case all known methods
of powder diffraction apply.
Therefore, the
quantitative phase analysis, stochiometry analysis,
diffraction line profile analysis and film thickness
determination may be used with high level of
confidence. The following on-the-fly simultaneous
analysis options are available:
1. Crystallographic texture:
-Multiple pole figures on multiple materials (layers)
-High resolution orientation distribution function for
quantifying dominant texture components
2. Phase analysis:
-Quantitative phase on multiple layers
-Alloy composition and solid solution analysis
-Heteroepitaxy and misfit stresses
-Grain growth and microstrain relief
3. Film thickness:
-Multiple layer capability using calibration standards.
INTRODUCTION
The microstructure control plays an
increasingly important role in improving the
performance and reliability of ULSI devices that use
the damascene copper technology at 0.13-µm node and
below. The problems related to delamination, stress
voiding, and electromigration failures could be
mitigated by the selection of proper materials,
processing methods, and manufacturing tools. The
optimum process would result in a tailored
microstructure of barrier/seed/electroplated copper
aggregate. At the same time, the microstructure could
be used as an internal sensor, sensitive to process
excursions and providing guidance for the corrective
actions. X-ray diffraction (XRD) has been used
extensively in the failure analysis labs for studying the
relationships between the failure modes and
microstructural features of thin films and interconnects.
We took the diffraction methods one step further to the
fab, sacrificing some of the flexibilities of an analytical
instrument for the benefit of speed and automation.
We present a XRD metrology tool for microstructure
mapping on films deposited on 300 mm wafers.
CAPABILITIES AND PERFORMANCE
Crystallographic texture has always been an
issue while applying XRD to thin films, hindering the
classical powder diffraction techniques and forcing
researchers to develop methods of mitigating texture
effects on diffraction analyses. Our method uses a high
The tool, shown in Figure 1, satisfies all
applicable safety, automation, and contamination
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
490
standards for 300 mm wafers. It is also available as a
bridge 200/300 mm tool.
EXAMPLES OF APPLICATIONS IN
DAMASCENE COPPER PROCESSING
Typical applications of the XRD metrology
tool are listed below:
Process development and qualification:
Barrier and seed layers - selection of materials and
deposition conditions such as gas flow, temperature,
and pressure,
Electroplating – optimization of bath chemistry,
rotation speed, flow and current density,
Annealing - optimization of temperature, ramp, time.
Process control:
Stability of a process in terms of consistent barrier
layer, seed layer and electroplated copper
microstructure,
Process excursion and post maintenance stability.
Tool qualification:
Performance of a production tool.
On-line R&D:
Improvement of device performance and reliability
through microstructure control,
Fundamental studies of microstructure development
and evolution during processing.
The examples of applications of the XRD
metrology tool will refer (but are not limited to) to
materials and processes typical of damascene copper
technology for ULSI. A typical processing route
includes the deposition of a barrier layer and copper
seed layer (in some instances in one cluster tool),
followed by copper electroplate, anneal and chemicalmechanical planarization. All the processing steps
affect the microstructure of the annealed copper, and
therefore affect directly performance of interconnects.
Each processing step could be monitored by XRD and
adjusted in such a way as to produce the desired
microstructure. Each processing step affects the
subsequent one, therefore all of them need to be
considered and optimized as a one processing route.
The following section shows this interdependence of
all processing steps.
The type of material used as a barrier layer,
selected for its diffusion and adhesion properties,
controls the microstructure (such as grain size and
crystallographic texture) of the copper seed layer.
Figure 2 shows the dependence of copper texture in
0.18 micron lines on the type of barrier layer material
and the deposition method. As a quantitative measure
of texture we use the volume fraction of grains having
particular (hkl) planes parallel to wafer’s surface, with
a tolerance of 5 degrees. The most typical texture of
PVD and electroplated (EP) copper is a (111) fiber.
Other orientations of interest include (511) and (57 13)
Figure 1. 300 mm XRD metrology tool with dual cassette
holders operating in a mini-environment.
The basic performance data is summarized as
follows:
Detection Limits:
Blanket films: Cu-10nm, Ta-3nm, TaN-10nm, Co7nm, W-5nm,
Patterned films: depends on line width and coverage
density, e.g. 10 nm Ta in 0.18 µm trenches and 20%
coverage density is detectable.
Spatial resolution:
Software controllable beam diameter from < 50 µm to
1 mm.
Total System Measurement Error at 99% confidence
level:
1% for blanket films and lower spatial resolution,
10% for low coverage density and small beam size
(smaller than 100 µm).
Throughput:
2 sec/point for blanket films up to 120 sec/point for
low coverage density structures.
Data Storage:
Advanced database storage protocol with multiple
mining capabilities.
Machine Vision System for pattern recognition.
491
Ta
TaN
TaN/Ta-m ethod A
TaN/Ta-m ethod B
TaN/Ta +resputter
TaN/Ta flash
Other
30
25
20
15
the wafer, with the periphery usually more prone to
higher fluctuations.
XRD monitoring of phase
composition may be used for tool qualification or for
monitoring of process stability.
250000
10
150000
α-Ta
50A TiSiN/400A Ta
50A TiN/400A Ta
α-Ta
100000
0.55α+0.45β -Ta
50000
5
Liner type
Liner composition
200000
Intensity (a.u.)
Volume fraction of (111) fiber
in seed layer
fibers that are the first and second generation of copper
annealing twins, respectively.
50A TaN/400A Ta
β-Ta
400A Ta
β-Ta
Cu/400A Ta
0
30
0
35
40
45
50
55
60
65
70
75
80
2-theta (degrees)
Liener type and deposition method
Figure 4. Dependence of Ta phase on matrials used in a biliner.
Figure 2. Effect of liner type and deposition method on
copper texture inside 0.18-micron trenches.
60
50
100
α+β-Ta liner
Volume fraction of (111) Cu fiber (%)
90
30
20
10
0
Barrier Seed only
β-Ta liner
α-Ta
60
α-Ta liner
30
20
10
0
Bilayer
N2=45sccm
49
59
73
79
88
94
The quality of Cu seed layer (grain size,
texture, density of lattice defects) is controlled by the
deposition process and it’s parameters (temperature,
pressure). Figure 6 shows the effect of deposition
method on texture of Cu seed layer deposited on β-Ta
liner.
The parameters (chemistry, flow, current
density, rotation speed) of the subsequent processing
step, electroplating of copper, control its
microstructure. Figure 7 shows the effect of plating
method on copper texture.
Annealing method and conditions may be
used to control the final microstructure of copper.
Figure 8 shows the effect of annealing on copper
texture. By monitoring texture components that are
essential for the performance of the device the
annealing process can be optimized.
In the proceeding sections, we have outlined
the relationships between process parameters and the
microstructure of liners, and PVD and electroplated
copper. We have discussed only the texture and phase
composition issues.
Other factors, such as
heteroepitaxy, misfit stresses, and film thickness also
40
Bilayer
N2=25sccm
25
Figure 5. TaN/Ta/Cu liner composition change over time,
wafer’s mid-radius.
50
Bilayer
N2=15sccm
10
Time (days)
Post ECP Anneal
70
Ta only
40
0
β-Ta liner
80
% of β-Ta in liner
Properties of a liner are controlled by the
deposition conditions. The quality of a typical TaN/Ta
bi-liner may be controlled by the N2 flow. Low N2
flow produces a β-Ta liner, intermediate rates produce
a mixture of α-Ta and β-Ta, and high N2 flow produces
α-Ta liner, as shown in Figure 3. The liner quality has
an impact on grain size and texture of copper seed
layer and annealed electroplated copper.
Bilayer
N2=60sccm
Figure 3. Effect of nitrogen flow on texture of copper seed
(light shade) and electroplated copper (dark shade) after
anneal.
The Ta phase in the bi-liner depends on the material
used for the intermediate layer between Ta and the
dielectric. An example of such relationship is shown in
Figure 4.
Process excursions or tool maintenance may
change the liner deposition conditions and as a result,
the liner quality may vary over time. Figure 5 shows
the evolution of TaN/Ta liner over time changing from
pure α-Ta to almost 60% of β-Ta. The magnitude of
liner composition variation depends on the location on
492
70
contribute to the microstructure control. They are also
measured by the XRD tool, and in many instances may
provide a meaningful insight and serve as essential
monitoring quantities. Figure 9 shows an example of the
hetroepitaxial relationship between NiSi2 and Si.
50
(111) fiber
(511) fiber
40
30
20
10
0
A
B
A+EP
B+EP
Method of Cu seed deposition
Figure 6. Texture of copper seed and subsequently of
electroplated copper (EP) is controlled by the seed deposition
method (A vs. B). Room temperature recrystallization of EP
copper. Error bars show one standard deviation of a 49-point
map on a wafer.
Figure 9. (103) pole figure of 30 nm thick NiSi2 deposited on
Si.
20
Volume fraction of (111) fiber
18
An important issue is the wafer mapping capability.
The spatial distribution of measured quantities provides
information about the tool and process stability that, in
general, would be missed if the monitoring were
reduced to a single location on the wafer. Figure 10
shows an example of a large texture variation across
the wafer.
16
14
12
10
8
6
4
2
0
A
B
C
D
Plating method
Figure 7. Plating method controls copper texture. TaN/Ta
liner, Cu seed, and anneal were identical for all plating
methods. Error bars show one standard deviation of a 49point map on a wafer.
40
35
Volume fraction (%)
Volume fraction (%)
60
(111) fiber
(511) fiber
57 13) fiber
30
25
Figure 10. Distribution of (111) copper fiber volume
fraction (%), 0.2 micron interconnect lines, 50% coverage
density, TaN/Ta liner.
20
15
REFERENCES
10
5
1.
K.J. Kozaczek, et al., “Methodology of Quantitative
Texture Analysis in Thin Films and Interconnects,” in
Advanced Metallization Conference 2000, ed. D. Edelstein et
al. MRS, Warrendale, PA, 2000, pp. 193-197.
2. K.J. Kozaczek, et al., “Crystallographic Texture and
Phase Metrology During Damascene Copper Processing,” in
Materials Research Society Symp. Proc. Vol. 721 MRS,
Warrendale, PA, 2002, pp. 5-16.
0
A
B
Anneal type
C
Figure 8. Texture of electroplated copper is controlled by
annealing method. Error bars show one standard deviation of
a 49-point map on a wafer.
493