31st ICPIG, July 14-19, 2013, Granada, Spain
6
Measurement of molecular chlorine density in a metal chloride reduction
chemical vapour deposition system by optical absorption spectroscopy
K. Sasaki1, K. Furuta2, Y. Tomita3, and H. Sakamoto3
1
Division of Quantum Science and Engineering, Hokkaido University
Kita 13, Nishi 8, Kita-ku, Sapporo 060-8628, Japan
2
Department of Electrical Engineering and Computer Science, Nagoya University
Furo-cho, Chikusa-ku, Nagoya 464-8603, Japan
3
Hang-Ichi Corporation, 315 Azuma build., 1-7 Hon-cho, Naka-ku, Yokohama 231-0005, Japan
We measured the density of molecular chlorine in Cl2/He mixture plasmas in the presence of copper
targets as the source for film deposition. The density of molecular chlorine was measured by
broadband optical absorption spectroscopy using an ultraviolet light emitting diode as the light
source. It was found that the density of the molecular chlorine decreased drastically and it became
almost negligible at 30 s after the ignition of the plasma. The recovery of the density of molecular
chlorine after the extinction of the plasma was very slow. The density of molecular chlorine without
the discharge was recovered at approximately 10 minutes after the extinction of the plasma. The
experimental result suggests the strong gettering of molecular chlorine by deposits on the substrate
and the chamber wall.
1. Introduction
Metal chloride reduction chemical vapour
deposition (MCRCVD) is a method for depositing
thin films of various metals [1,2]. It is a kind of
plasma-enhanced chemical vapour deposition, but it
does not need gaseous ingredients such as organic
metals. The ingredients in MCRCVD are solid-state
targets of metals. The solid-state metal targets are
etched by chlorine-based plasmas in MCRCVD.
When the volatility of the etch products is not high,
they are not evacuated from the chamber and deposit
on the substrate to form a thin film. Although the thin
film contains chlorine atoms, they are reduced from
the film due to reactions between the chloride film
surface and atomic chlorine transported from the
plasma.
Although MCRCVD is developed under the
assumption of the aforementioned ideal mechanism,
it should be examined experimentally by using
advanced plasma diagnostics. In this work, we
investigated the density of molecular chlorine in a
MCRCVD system. Since molecular chlorine has a
broadband absorption spectrum in an ultraviolet
wavelength range, we employed optical absorption
spectroscopy for measuring its density. In this work,
we used a light emitting diode (LED) as the light
source of broadband absorption spectroscopy [3],
instead of conventional Xe and D2 lamps.
2. Experiment
The system of MCRCVD is schematically shown
in Fig. 1. Inductively-coupled plasmas were
produced by applying various rf powers at 13.56
ICP
Antenna
Cl2/He
Quartz window
Gas Nozzle
Cu Target
CuCl
Substrate
Quartz tube
Figure 1 Schematic of MCRCVD system.
MHz to a spiral antenna via a matching network. The
rf antenna was placed on a quartz window at the top
of the vacuum chamber. The mixture of Cl2 and He
was introduced into the chamber from a nozzle
placed below the quartz window, and was evacuated
from the bottom of the chamber using a
turbomolecular pump. The mixing ratio of He and Cl2
was adjusted by changing their flow rates using mass
flow controllers. The total gas flow rate was 30 sccm.
The total gas pressure was controlled by changing the
pumping speed using a variable conductance valve. A
SiO2 substrate (100 mm diameter) was installed on a
substrate stage, and Cu targets were placed between
the quartz window and the substrate. The electric
potential of the targets was floating. The chamber
31st ICPIG, July 14-19, 2013, Granada, Spain
3.32
Plasma ON
3
-3
cm )
3.36
3.5
2
1.5
1
500 W
300 mTorr
Cl : 30%
2
3.28
2.5
15
500 W
300 mTorr
Cl2: 30%
Cl density (10
Transmitted intensity (arb. unit)
Valve Discharge Discharge
open
ON
OFF
2
3.24
3.2
0
0.5
0
200
200 400 600 800 1000 1200
Time (s)
Figure 2 Temporal variation of the transmitted
LED light intensity observed in a cycle of the
experiment.
was made of stainless-steel, but the surface of the
chamber was covered with a quartz inner tube.
An LED with ultraviolet emission at a wavelength
around 357 nm was used as the light source of
broadband absorption spectroscopy. The emission
side of the LED housing was flattened by mechanical
polishing to avoid its lens effect. Three lenses were
used for obtaining a collimated path of the LED light.
The LED light transmitted through the plasma was
detected using a photomultiplier tube via a
monochromator. The emission of LED was pulse
modulated by a rectangular current, and the output of
the photomultiplier tube was amplified using a
lock-in amplifier with the reference frequency of the
current modulation.
3. Results and discussion
Figure 2 shows a time history of the transmitted
LED light intensity, which was observed in a cycle of
the experiment. The substrate was not heated in this
experiment. At the beginning (t=0 s), the chamber
was evacuated below a background pressure of
5x10-6 Torr. The left edge of Fig. 2 shows the level of
the transmitted LED light intensity with no
absorption. When a valve was opened to introduce
the mixture of Cl2 and He into the chamber, the
transmitted LED light intensity decreased rapidly.
The decrease in the transmitted LED light intensity
was due to absorption of molecular chlorine. The
400
600
800
Time (s)
1000
Figure 3 Temporal variation of the absolute
density of molecular chlorine. The plasma was
sustained during the hatched period.
total pressure and the partial percentage of Cl2 were
300 mTorr and 30 %, respectively, in Fig. 2. An rf
power of 500 W was applied to the rf antenna and a
plasma was ignited at t=320 s. As shown in the figure,
the transmitted LED light intensity increased
drastically after the ignition of the plasma. The
original level (at t=0 s) of the transmitted LED light
intensity was recovered in the plasma. Another
interesting phenomenon was observed after the
extinction of the discharge. As shown in the figure,
the transmitted LED light intensity decreased very
slowly after the extinction of the discharge. The
decrease in the transmitted LED light intensity means
the increase in the Cl2 density. It is noted here that the
rate of the decrease in the transmitted LED light
intensity after the extinction of the plasma was much
slower than that observed after the opening of the gas
valve.
The temporal variation of the absolute Cl2 density
was calculated from the transmission curve shown in
Fig. 2. As shown in Fig. 3, the Cl2 density decreased
in the plasma significantly, and was almost negligible
at the steady state. The Cl2 density increased very
slowly after the extinction of the discharge.
A mechanism of the decrease in the density of
molecular chlorine in the plasma is dissociation into
atomic chlorine due to electron impact (Cl2 + e Cl
+ Cl + e). However, the slow recovery of the density
of molecular chlorine after the extinction of the
plasma indicates a loss process of Cl2 without the
plasma. Therefore, we should consider a loss process
31st ICPIG, July 14-19, 2013, Granada, Spain
result shown in Fig. 4 suggests that the deposits
formed by MCRCVD without heating the substrate
has a strong gettering effect of molecular chlorine.
-2 -1
1
Gettering flux (10
1.4
15
cm s )
1.6
1.2
0.8
0.6
0.4
Plasma ON
0.2
0
200
400
600
800
Time (s)
1000
Figure 4 Estimation of the flux of gettered
molecular chlorine by the deposits on the
chamber wall.
of Cl2 in addition to electron impact dissociation. A
possible mechanism for the loss of Cl2 is the gettering
by the deposits on the substrate and the chamber wall.
The pumping speed of the evacuation system for
Cl2 at the experimental condition is evaluated by
S0 =
Q
,
p0
(1)
where Q and p0 are the flow rate and the partial
pressure of Cl2 before the discharge, respectively. If
we ignore the loss of Cl2 due to electron impact
dissociation, the decrease in the Cl2 density after
starting the discharge is owing to the increase in the
effective pumping speed. Under the assumption that
the increase in the effective pumping speed is due to
the gettering by the deposits on the substrate and the
chamber wall, we evaluated the flux of gettered
molecular chlorine by
Γ=
Q − pS 0
,
A
(2)
where A is the surface area of the deposits and p is the
partial pressure of Cl2 given by p = nCl2 kTg with nCl2,
k, and Tg being the density of Cl2, the Boltzmann
constant, and the gas temperature, respectively.
Figure 4 shows the temporal variation of the flux
of gettered molecular chlorine. In this estimation, we
assumed that A is given by the surface area of the
chamber wall and Tg is kept at room temperature. The
4. Conclusions
In this work, we measured the temporal variation
of the density of molecular chlorine by broadband
absorption spectroscopy using a light emitting diode
as the light source. As a result, it was found that the
density of molecular chlorine decreased drastically
after the ignition of the discharge. The recovery of
the density of molecular chlorine after the extinction
of the plasma was very slow. The experimental result
suggests the strong gettering of molecular chlorine by
deposits on the substrate and the chamber wall.
5. Acknowledgement
The authors are grateful to Nader Sadeghi for
teaching them the details of broadband absorption
spectroscopy for measuring the density of molecular
chlorine.
6. References
[1] Y. Ogura, C. Kobayashi, Y. Ooba, H. Sakamoto,
N. Yahata, T. Nishimori, K. Hatayama, Jpn. J.
Appl. Phys. 43, L56 (2004).
[2] H. Sakamoto, Y. Ogura, Y. Ooba, T. Nishimori, N.
Yahata, J. Electrochem. Soc. 151(3), C200
(2004).
[3] G. Cunge, M. Mori, M. Kogelschatz, and N.
Sadeghi, Appl. Phys. Lett. 88, 051501 (2006).