The Impact of Power Supply Arc Response on
Production Yield and Field Reliability
D. J. Christie, Advanced Energy Industries, Inc., Fort Collins, CO
ABSTRACT
MAGNETRON ARCING
Magnetron sputtering is widely used to deposit layer systems in
large area processes coating architectural and automotive glass
and plastic films for various applications. The system power
supply must operate the process at the desired set point, and
must also respond to arcs which are a natural outgrowth of the
sputtering process. The size of the macroparticles produced by
arcing depends in part on the energy the power supply delivers
to the arc. Some of these macroparticles which become inclusions in the film can negatively impact the field reliability of the
layer system. It is important for the power supply to manage
arcs correctly in order to minimize the number and size of the
macroparticles. Arc response duration is important for reactive
sputtering, particularly for processes stabilized in the transition
region. The disturbance in the process can last longer than the
power supply shut off time, resulting in visible banding. It is
important for the arc response duration to be short enough to
eliminate any visible effects in the final product. Modeling can
be used to study the maximum tolerable arc response time for
reactive sputtering processes. The physical basis of the effect
of macroparticles on layer system integrity is reviewed.
In the magnetron sputtering process, the desired glow discharge
mode is sustained by secondary emission of electrons induced
by ion impact at the target surface, as an individual process. An
individual process means that one ion incident on the surface
results in emission of some number of secondary electrons (the
ion impact being primary), with some probability. These secondary electrons perform bulk ionization of process gas neutrals by
electron impact [4] and possibly sequential secondary processes
such as Penning ionization and multiple body collisions. The
undesired cathodic arc discharge mode is sustained by explosive
emission of ions and electrons from small craters on the target
surface in what could be considered a collective process [2].
It is a collective process because the current flowing in the arc
provides heat to the arc spot which in turn causes melting and
explosive emission of a “collection” of target material atoms.
In the cathodic arc mode, target material macro-particles are
also explosively emitted from arc craters, often landing on the
substrate, resulting in product yield issues.
INTRODUCTION
Magnetron sputtering is an important technique for production
of both consumer and commercial goods [1]. In large area
coating, it is widely used to deposit low-emissivity (“low-e”)
coatings on architectural glass in large in-line coaters. Other
applications include mirrors, flat panel displays, and antireflection (AR) coated glass. It is also employed in roll coaters for depositing optical coatings on plastic films as well as
oxygen barriers for plastic food packaging films.
One of the challenges in the practical realization of DC and
pulsed mid-frequency magnetron sputtering is that it will support two stable discharge modes. The desired discharge mode
is a glow or abnormal glow discharge, at low current density.
The undesired discharge mode is a cathodic arc discharge, at
very high current density [2-4]. The cathodic arc mode causes
damage to both target and work piece, so it must be detected
and quenched by the sputtering power generator. An additional
challenge is real-time assessment of the health of the process.
As process conditions degrade, it is necessary at some point
to stop the process in order to maintain the process equipment
and restore it to an acceptable state.
Electrical discharge devices with the ability to operate in
arc and glow regimes have been studied for some time now.
Early work on the transition from the glow to the cathodic arc
mode showed the importance of oxide on the target surface
for sustaining an arc [5-9]. When ultra pure noble gases were
used, it was essentially impossible to sustain an arc. In some
experiments, the argon gas was purified in situ with the arc
operating. When a high level of purity was attained, the arc
mode discharge ceased and only a glow discharge was possible. This result was attributed to formation of oxides on the
surface. When the gas was purified, the oxides were eventually
removed by the arc. The motivation for this work was likely to
understand how to make arc sources work better. Now we are
interested in keeping sputtering processes out of the arc regime.
This early work at least suggests the importance of process gas
and target material purity in metal sputtering processes, and
perhaps sets the expectation that target arcing could develop
when reactively sputtering oxides. Even so, formation of arcs
in the magnetron sputtering process is still not well understood
fundamentally. Issues with arcing in sputtering processes are
handled predominantly with the “know-how” of power supply
designers and sputtering practitioners.
© 2006 Society of Vacuum Coaters 505/856-7188
49th Annual Technical Conference Proceedings (2006) ISSN 0737-5921
183
EFFECT OF ARC SHUTOFF TIME FOR REACTIVE
SPUTTERING PROCESSES
When arcing occurs, the magnetron power supply must shut off
to quench the arc. In the case of reactive sputtering this presents
some subtle challenges. The effects of the interruption in power
can last longer than the shutdown due to the dynamics of the
reactive sputtering process, especially in the transition mode.
These effects were studied for a typical large area TiO2 process
[10, 11]. The dynamical version of the Berg model was used
to simulate the process response to power supply shutdown
for arc handling [12-14]. The model used three process states:
partial pressure, target coverage fraction, and chamber surface
coverage fraction. It is based on continuity of flow and competing processes at target and chamber surfaces. Rate was taken
as the metal atom removal rate (from the target), both native
target material and compound (TiO2) on the target surface.
Numerical simulations were performed using MATLAB® and
Simulink® [15]. The process was stabilized in the naturally
unstable transition region for simulations by a feedback controller designed with the MATLAB® Control System Toolbox.
The results showed that when the power supply shuts off for
arc handling, the reactive gas partial pressure increases and
the target coverage fraction (with compound) increases. As a
consequence, the sputtering rate is decreased when the power
supply turns on again. Some time is required for rate to recover
to its value prior to shut-down, so the process effect is not just
time off. There is lost rate due to arcing and lost time when
power is off. It would be expected that for practical reactive
sputtering partial pressure control systems, process stabilization
becomes increasingly difficult with longer off time. The reduced
rate when the power supply turns on again is due to increased
coverage fraction; it takes time to reduce the coverage fraction
to the equilibrium level and achieve steady state rate again.
The worst case is where there is a large difference in sputtering yield between target metal and compound, for example:
STi ≈ 0.7, STiO2 ≈ 0.03. Figure 1 shows how an increase in
coverage fraction from about 75% to 90% can cut rate by
almost 50% for TiOx reactive sputtering.
184
Figure 1: Normalized rate as a function of target coverage
fraction for TiO2 reactive sputtering.
MACRO-PARTICLES GENERATED BY PROCESS
ARCING
Large area magnetron sputtering processes operate at high powers, up to 100 kW to 200 kW, or even higher powers in some
rare cases. Power supplies capable of delivering these high
powers have considerable arc energies, and deliver considerable peak currents to process arcs. It could be expected that
a single arc has locally similar conditions, such as magnetic
field, target surface temperature, and process gas pressure, as
for low power (1 kW to 10 kW) processes. However, available energy that could be delivered to an arc may be ten or
more times higher. The expected result is larger target material
particles (macro-particles) generated by the arc as arc energies
increase, as shown in Figure 2. This suggests that for large area
coating, there is a greatly increased importance of minimizing
arc energy. The generation of macro-particles by sputtering
magnetron arcing has been studied by several researchers
[16-20]. Based on published data, macro-particles as large as
10 µm or larger could be expected. The details are difficult to
quantify by analysis but would seem to be clearly dependent
on process conditions, target material, and the details of the
magnetron design, as well as power supply arc handling. Large
area coating presents special challenges for arc energy and arc
rate. The expected general trend of macro-particle size distribution and quantity scaling is shown in Figure 3. Arc energy
should be reduced to minimize arc induced macro-particle
size. Arc frequency (rate) should be reduced to minimize
arc-induced macro-particle quantity. Appropriate choice and
setup of pulsed DC equipment can minimize arc energy and arc
frequency (rate) [21]. Macro-particles generated by arcing can
be included in the layer stack, as shown in Figure 4, resulting
in layer system defects and a possible reduction of field reliability, characterized by premature degradation of appearance
and/or reduction in performance. Therefore, it is desirable to
minimize macro-particle size and quantity.
ACKNOWLEDGEMENTS
The author would like to acknowledge the contributions to this
paper of the many (too numerous to mention by name) co-workers at Advanced Energy who have contributed to the development of arc handling technology over the past 25 years.
REFERENCES
Figure 2: Expected trend of macro-particle size with arc
energy.
Figure 3: Expected trend of macro-particle size distribution
and quantity with arc energy and frequency (rate).
Figure 4: Macro-particles generated by target arcing can be
included in the thin film layer stack, resulting in defects and
possible reduction of field reliability.
1.
W.D. Westwood, Sputter Deposition, AVS, New York,
2003.
2.
A. Anders, “Physics of arcing, and implications to sputter
deposition,” Proc. 5th Intl. Conf. on Coatings on Glass,
p. 59, 2004.
3.
J.D. Cobine, Gaseous Conductors, Dover, New York,
1958.
4.
M.A. Lieberman, A.J. Lichtenberg, Principles of Plasma
Discharges and Materials Processing, Wiley, New York,
1994.
5.
G.M. Schrum, H.G. Wiest, “Experiments with short arcs,”
Electrical Engineering 50, p. 827, 1931.
6.
G.E. Doan, J.L. Myer, “Arc discharge not obtained in pure
argon gas,” Phys. Rev. 40, p. 36, 1932.
7.
G.E. Doan, A.M. Thorne, “Arcs in inert gases. II,” Phys.
Rev. 46, p. 49, 1934.
8.
M.J. Druyvesteyn, “Electron emission of the cathode of
an arc,” Nature, p. 580, 1 April 1936.
9.
C.G. Suits, J.P. Hocker, “Role of oxidation in arc cathodes,”
Phys. Rev. 53, p. 670, 1938.
10. P. Greene, R. Dannenberg, “Modeling of production scale
reactive deposition,” 42nd Annual Technical Conference
Proceedings of the Society of Vacuum Coaters, p. 23,
1999.
11. D.J. Christie, E.A. Seymour, “Power system requirements
for enhanced mid-frequency process stability,” 46th Annual Technical Conference Proceedings of the Society of
Vacuum Coaters, p. 257, 2003.
185
12. S. Berg, et al., “Modeling of reactive sputtering of compound materials,” JVST A 5(2), Mar/Apr, p. 202, 1987.
13. H. Bartzsch, P. Frach, “Modeling the stability of reactive
sputtering processes,” Surface and Coatings Technology,
142-144, p. 192, 2001.
14. N. Malkomes, N. Vergöhl, “Dynamic simulation of process
control of the reactive sputter process and experimental
results,” Journal of Applied Physics, vol. 89, no. 1, p. 732,
1 Jan, 2001.
15. Using Simulink®, The Mathworks, Natick, 2002.
16. B. Jüttner, “On the variety of cathode craters of vacuum
arcs, and the influence of the cathode temperature,” Physica
114C, 255 (1982).
17. C.E. Wickersham, Jr., J.E. Poole, J.S. Fan, L. Zhu, “Video
analysis of inclusion induced macroparticle emission
from aluminum sputtering targets,” JVST A 19(6), 2741
(2001).
186
18. C.E. Wickersham, Jr., J.E. Poole, A. Leybovich, L. Zhu,
“Measurements of the critical inclusion size for arcing and
macroparticle ejection from aluminum sputtering targets,”
JVST A 19(6), 2767 (2001).
19. C.E. Wickersham, Jr., J.E. Poole, J.S. Fan, “Arc generation
from sputtering plasma-dielectric inclusion interactions,”
JVST A 20(3), 833 (2002).
20. K. Koski, J. Hölsä, P. Juliet, “Surface defects and arc
generation in reactive magnetron sputtering of aluminium
oxide thin films,” Surface and Coatings Technol. 115, 163
(1999).
21. D. Carter, H. Walde, G. McDonough, G. Roche, “Parameter optimization in pulsed DC reactive sputter deposition
of aluminum oxide,” 45th Annual Technical Conference
Proceedings of the Society of Vacuum Coaters, p. 570,
2002.