, ADVANCED
`
`ENERGY®
`
`By Thomas C. Grove, Advanced Energy Industries, Inc.
`
`ARCING PROBLEM5
`
`______________________________________________________________________________________________
`
`ENCOUNTERED DURING
`
`SPUTTER DEPOSITION
`
`OF ALUMINUM
`
`The physics of plasmas and particularly the physics of plasma arcs and their
`genesis has received new attention in the last five years. Researchers working on high—
`power lasers, fusion, charge particle accelerators and others have required high energy
`pulsed power. Many of these applications use switching technology based upon devices
`such as thyratron tubes and spark gaps. These devices cause a gas to undergo a transition
`from normally insulating to a conducting state, a purposeful use of plasma arcs. As demands
`for increased switching speed at higher and higher current densities grew, greater
`understanding of arc formation became necessary. While no single, complete theory covers all
`aspects of arc formation, much working knowledge has been gained. The physical problem of
`interest is primarily the growth of the ionization of a gas in an electric field and the subsequent
`breakdown of the insulating qualities of the gas (a very good, current reference on this is in’”).
`Although a plasma or glow discharge is already a conducting gas (i.e., it has broken down the
`insulating properties of a neutral gas), it has not reached current saturation and can be caused to
`conduct even more current. This research has significant implications for the lower energy uses of
`plasma commonly found in thin film processes such as etching and sputter deposition where arcs are
`sometimes troublesome.
`
`Arcs, as used here, are local events within the sputtering
`chamber that are detrimental to the process. Arcs are high
`power density short circuits which have the effect of miniature
`explosions. When they occur on or near the surfaces of the
`target material or chamber fixtures they can cause local melting.
`This material is ejected and can damage the material being
`processed and accumulates on other surfaces. This erosion can
`contaminate the source as well as degrade the structure.
`
`This article is meant to provide an introduction to the
`causes, mechanisms, and some cures for arcing in the sputtering
`environment. Since the application and distribution of power is
`central to the sputtering process, the new understanding of
`arcing phenomena will be related to advances in power supply
`design. In the following sections we will look at sputtering,
`simplified arc formation, some common causes of arcing and
`how to minimize them, and how power supply design is
`important to arc minimization. _
`
`WK Mfifli’iii’tiiigfi savraoanrsr
`
`.
`Figure 1 is a- diagram of a simple sputtering chamber. '
`A power supply is connected so that a low pressure gas is ionized
`by the voltage supplied. The ions of gas produced are
`accelerated toward the target surface by the voltage where they
`collide with the atoms of the target. The kinetic energy of the
`ions is transferred to these atoms, some of which are ejected and
`drift across the chamber where they are deposited as a thin film
`on the substrate material. Other impacted target atoms are
`simply heated. This is removed as waste energy. Still other
`impacts on the target surface produce secondary electrons.
`It is these electrons which maintain the electron supply and
`sustain the glow discharge‘“.
`
`A schematic drawing of a
`planar magnetron is shown in Figure
`2. The magnetron type source is
`commonly used to deposit aluminum
`and is noted for its sputtering efficiency.
`It also has a greater propensity to arc”.
`The source is designed such that as
`uniform an electric field gradient as
`possible is maintained across the active
`sputtering region of the target material. This
`determines the local power density which in
`turn has an impact on arc formationm“.
`
`Vacuum
`Chamber
`
`_
`
`Target Material
`
`Glow Discharge
`
`
`
`[:1
`
`Substrate Support
`
`Sputtering
`Gas Feed
`
`Vacuum
`Pump
`
`Figure 1. The sputtering environment
`
`GILLETTE 1021
`
`1
`
`GILLETTE 1021
`
`

`

`
`
`E Field Lines
`
`Magnet Poles
`
`Magnet Field Lines
`
` Hopping Electrons
`
`Glow Discharge
`
`Cathode
`
`Figure 2. Magnetron
`
`The magnets are placed so that the magnetic field lines
`are normal to the target material surface at the point of entry.
`The field is usually parallel to the surface in the sputtering
`regions. It is common for the shield surrounding the target to
`be at anode potential.
`
`Electrons are charged particles. They also have a magnetic
`field because of their internal spin momentum. When they
`are ejected from the target surface, the electron's charge and
`magnetic field interact with the electric field and source
`magnetic field which causes the electron to remain near the
`surface‘”. Furthermore, their path near the surface is somewhat
`lengthened due to electric field gradients in the dark space.
`This path has two beneficial effects, 1) it provides a greater
`probability of producing more ions near the surface thereby
`increasing the sputtering efficiency and 2) the electron current
`is trapped near the cathode“. The latter minimizes electron
`heating and other damage to the substrate.
`
`Magnetrons require careful attention to cooling because of
`this more concentrated electron current and higher ionization
`efficiency”. The sputtering area is more confined and controlled
`which in turn greatly increases the power density. Magnetrons
`also erode faster in the confinement area. This erosion causes
`
`impedance changes because the plasma to source magnet
`distance changes”. Erosion also changes surface texture and
`electric field gradients. All of these factors can contribute to
`arcing as we shall see later.
`
`Other chamber factors cannot be ignored. The flow of the
`sputtering gas through the chamber, the chamber pressure, the
`power supplied and the source temperature must be carefully
`controlled as they too can contribute to arc production.
`
`QRQW fiifimfifififiifi Ptfit‘ffiflifisfi
`
`We will now look more closely at the glow discharge
`space and the types of energies, potentials and particle species
`involved. The drawing of Figure 3 gives a relational view of
`charge versus potential in a typical negative glow discharge
`configuration.
`
`Target
`
`Anode
`
`4‘
`
`'i‘
`Anode Dark Space
`
`
`
`600 + VP
`Sheath Voltage
`
`VP
`Sheath Voltage
`
`Figure 3. Glow discharge potentials
`
`Through various mechanisms, the areas near both
`electrodes are striped of electrons and a surface charge is built
`up. As distance increases away from the electrode the charge
`balance returns to a more neutral value. This region is depicted
`by the + and 0 symbols in drawing 3. The potential across the
`glow region is nearly constant: almost all of the electric field
`potential occurs in the sheath. The cathode sheath region is of
`particular interest because it supplies the basic energy for
`sputtering. Most ions are produced in the negative glow region.
`These ions are accelerated by the sheath potential and bombard
`the target surface. This bombardment is what ejects the target
`material. it also produces a host of other particles and energies
`which contribute to sustaining the glow discharge.
`
`Ions
`Accelerating
`in Electric
`
`Field
`
`Source Surface
`
`Figure 4. No arc
`
`The surface/nearrsurface area is a busy place. We have a
`significant potential voltage supplied by the power supply, high
`voltage RF transients produced by magnetron oscillations in the
`plasma, a confining magnetic field, electrons colliding with the
`
`

`

`
`
`surface and gas atoms, photons exciting atoms to become ions,
`ions hitting atoms which produce more ions and thermal energy
`exciting atoms at local hot spots. All this with high average
`power density! At any given time these activities are in balance
`with degenerative phenomena and normal operation of the
`sputtering process is observed. This condition is depicted in
`Figure 4. The electric field lines are uniformly distributed across
`the surface and no major anomalies are present at the surface or
`in the dark space immediately above.
`
`fiiiififii
`
`Theory deals with two stages of arc formation. The first
`stage is the primary transition from normal gas state to that of
`a glow discharge. Let us suppose a dc electric field is applied
`between two electrodes (a spark gap) in a gas. It we then
`illuminate the cathode with light, a photoelectric current will
`be generated at the cathode. We may then move the electrodes
`together and at some distance the current will become self»
`sustaining without the light. This first level of current flow is
`known as a Townsend discharge“. Townsend proposed that the
`advancing primary electrons produced by photon excitation
`generated some multiple of electrons by electron/neutral atom
`collisions“. Later researchers‘ml have shown that secondary
`events such as ion impact on the cathode and the photoelectric
`effect of the photons generated in the discharge were large
`contributors. At some point the secondary events happen in
`numbers sufficient to insure the regeneration of the primary
`electrons thus continuing current flow.
`
`A second theory, labeled Streamer Theory, was proposed
`independently by Meek”, Raetherm' and Loebm'. They argue
`that as the primary photoelectrons avalanche toward the anode
`they reach a critical size such that their combined charge starts
`to generate secondary electrons just ahead of the avalanche by
`photoionization. The avalanche space charge produces electrons
`efficiently which in turn generates a space charge cloud in front
`of the avalanche. This process repeats until the anode boundary
`is reached. This progression is called a streamer. Once the anode
`is reached a similar process begins at the cathode end of the
`parent avalanche. There the electrons are accelerated towards
`the avalanche which extends the ion sheath of the avalanche to
`the cathode. Breakdown occurs immediately upon the space
`charge cloud reaching the cathode”).
`
`If the resistance is low enough, the current will increase
`indefinitely (limited only by the power supply and the positive
`column impedance), which brings us to the second stage of arc
`formation. The current flowing in the gas will produce photons,
`heat and other reactions as we have seen which enhance the
`current flow until a visible glow discharge is produced.
`Depending upon the power available and the energy losses due
`to the chamber design, the process may stabilize at this point.
`in our sputtering chambers the cathode resembles an assembly
`
`of an infinite number of spark gaps which operate in this stable
`condition; however, if the current continues to increase an arc
`is formed. The avalanche path of electrons and associated ion
`sheath operate at a certain sustaining voltage determined by the
`gas, gas pressure and voltage potential. The increasing current
`flow and subsequent secondary activity continue to increase the
`local space charge. This causes a concentration of the electric
`field in the vicinity of the current flow. When the local
`potential reaches a value approximately 20% greater than the
`sustaining voltage a second breakdown occurs which lowers the
`gas discharge impedance to its lowest possible level...essentially
`a short circuit in which all energy transfer takes place. In a
`typical aluminum sputtering application using argon as the
`sputtering gas and a planar magnetron, this may happen with an
`incremental voltage change of as little as 72 volts.
`
`Based upon the foregoing, it appears that a glow discharge
`process is never more than 20% away from an arc condition
`somewhere in the chamber. This is certainly true in the
`surface/near—surface area of the magnetron. This margin can
`become very much smaller as the design power density of the
`magnetron source is approached.
`
`Power Shut Down \
`
`
`
`
`/PowerTurn Off
`/ Overshoot
`
`Re-lgnition
`
`/
`
`500
`400
`oo
`3
`200
`100
`
`50
`40
`o
`3
`20
`10
`
`
`
`15
`Microseconds
`Nominal Operation 500 V @ 8A=4 kW
`
`20
`
`25
`
`Figure 5. Typical arc profile
`
`Figure 5 presents a profile of an arc in an operating
`sputtering system. The voltage drops in approximately
`5 nanoseconds to an initial value. The load impedance at this
`point is limited by the power supply output, cable, chamber, etc.
`This voltage is maintained while the surface charge energy and
`power supply output capacitor discharge. In some older constant
`current supplies this current will continue to be supplied until
`outside factors cause the impedance to increase and extinguish
`the arc. If this doesn't happen, severe arcing called racetrack
`arcing can result. As soon as the current goes negative, the arc
`is turned off. The discharge and charge times and the current
`values are dependent upon the energy supplied by the surface
`charge surrounding the arc location and the stored energy in
`the output stage of the power supply.
`
`

`

`
`
`M26 Sfifiiifiii‘a
`
`In general we may say that all arcs are the result of electric
`field anomalies, however, for convenience we will look at three
`classes of arc sources. These classes are mechanical, electric field
`
`anomalies and disruptions of surface equilibrium.
`
`Mechanical sources are primarily flakes that short circuit
`various parts of the structure, commonly target to chamber, by
`bridging the dark space or by contacting the shield. Problems of
`this nature are cured by cleanliness and reducing actions which
`produce mechanical bumps and vibrations
`
`Surface
`Blemish
`
`
`
`
`Dielectric inclusion
`
`Source
`Surface
`
`the bulk material. Arcs isolated to a specific area may be due
`to a poor bond between the target material and the source, a
`blocked cooling channel in the source or the source design itself
`causing elevated temperatures. These higher temperature areas
`are one of the principal causes of arcs, either because they are
`producing thermal electrons or are much closer to doing so.
`Thermal electrons then contribute to an increase in the
`
`secondary ion production cycle and associated power density
`increase and possible melting and/or sublimation.
`
`Systems that have been modified or are new designs
`sometimes exhibit significant arcing associated with their
`structure. Older systems which have significant encrustations
`of sputtered material and/or severely eroded surfaces may have
`problems too. These can be cured by cleaning and smoothing
`surfaces to approximately 600800 grit. System fixtures should
`have beveled or rounded edges and interfaces with target, dark
`space shield, insulators and supports to reduce electric field
`gradients. Some systems may require auxiliary magnets or
`secondary electrostatic shielding to eliminate consistent
`arcing in a specific place.
`
`Figure 6. Potential arc source
`
`
`
`I am using "electric field anomaly" to designate a small
`scale local disruption. In this category most of the sources are
`associated with the target material. The source must be of
`uniform composition and internal form. Such inclusions as gas
`pockets, voids, dielectric clumps, grain boundaries and surface
`blemishes can each cause an arc to form. Obviously the
`selection of the correct material is important. The material
`will vary from lot to lot from any vendor, but, a consistent
`problem may have to do with the material manufacturing
`process and other forms should be considered. Figure 6 shows
`the effect of a surface blemish and an oxide inclusion. Both of
`
`these defects cause a warp in the electric field in their vicinity
`which increases the power density. Other surface problems are
`caused by magnetic dirt attracted by the magnetic field of the
`source, local thermal hot spots on the target, local pressure
`fluctuations due to several molecules of water or other
`
`contaminants drifting by and the target's previous operating
`history. Target history is important. An operating target will
`settle into a certain set of equilibrium conditions including
`power distribution, heat distribution and re—deposition rate.
`A step change in power due to power line perturbations or loss
`of regulation can cause arcing until equilibrium is reaestablished.
`Targets that have been exposed to air must be preconditioned
`by slowly increasing the power level. This burns off accumulated
`oxides and drives off absorbed water and other contaminants.
`
`Temperature variations can be caused by defects in the
`target surface which produce higher electric field densities.
`The surface of the target is always at a higher temperature than
`
`It may be said that the power supply feeding the system is
`the ultimate cause of and cure for arcing. In the discussions
`above we have covered many of the details of arc formation.
`In this section we will discuss the management of the power
`supplied to the process to minimize the occurrence and
`severity of arcs.
`
`Looking again at Figure 5, we see a large current spike
`as the arc discharges the chamber and the power supply output
`filter. Since the chamber normally presents a very small
`capacitance the majority of the stored energy is due to the
`power supply output. Anything we can do to reduce this energy
`directly reduces potential arc damage. One technique which is
`very effective is to raise the frequency that must be filtered.
`As an example: for equal levels of voltage and current at the
`output the amount of energy per cycle which must be supplied
`at 360 Hz (60 Hz 3 phase) is 100 V x 5 A x 2.8 millisec. =
`1.4 Joules while at 100 kHz it is 100 V x 5 A x 5 microsec. =
`2.5 millijoules. The smaller energy at higher frequency allows
`smaller value components to be used. The ripple voltage at the
`output can be made much smaller which reduces the chance of
`the power density exceeding the breakdown potential
`somewhere on the target.
`
`Higher frequencies allow much faster response to problems
`on both the input and output of the power supply. Again, a
`60 Hz, 3 phase power supply can reduce a 10% input voltage
`increase to 1% in about 20 milliseconds while a 100 kHz supply
`can do the same in 30 microseconds. Arc inception times are in
`the range of 106 seconds'l” which puts the l 00 kHz supplies
`response in the same decade, certainly an advantage. This same
`
`

`

`
`
`rapid response holds true for loadainduced transients such
`as impedance changes.
`
`Knowledge of arc mechanisms has improved circuitry
`for suppressing them. It has been found experimentally'll’
`that an area where an arc has occurred takes approximately
`5 milliseconds to return to equilibrium conditions (cool down
`the surface or gas, disperse particles and fields) after the
`discharge current has fallen to zero. Building in a short delay
`of this order before power is reapplied Virtually eliminates a
`second are at that site””.
`
`Increasing the accuracy of instrumentation available
`enables the power supply to detect arc initiation ever closer to
`the fact. Tighter measurements allow greater repeatability and
`stability.
`
`We have taken a brief look at arc mechanisms, their causes
`
`in the sputtering environment and the improvements in power
`application made possible by the study of them. As always,
`research continues.
`
`QSEELREKCES arse EW‘WEKR REASWQ
`1
`j. M. Meek and I. D. Craggs, eds., ELECTRICAL BREAKDOWN OF GASES
`New York: Wiley, 197
`
`2 B. Chapman, GLOW DISCHARGE PROCESSES New York: Wiley, 1980
`
`3 l. S. Chopin, RES,./DEV. 25(1),37(1974)
`4 /. L. Vossen,/. VAC. SCI. TECHNOL. 8,512, (1971)
`
`5 L. r. Lamont, lr., VAR/AN VAC. VIEWS 9(3), 2(1975)
`6 L. T. Lamont, In, /. VAC. SCI. TECHNOL. 14,12 (1977)
`
`7 R. K. Waits in "THIN FILM PROCESSES " 0. L. Vossen, w. Kern, eds)
`Academic Press, 1976
`
`8 E Llewellyn lanes, ION/ZATION AND BREAKDOWN IN GASES London,
`England: Methuen, 1966
`
`9 l. S. Townsend, THE THEORY OF ION/ZATION 0F GASES BY COLLISION
`London, England: Constable, 191 o
`
`10 E Llewellyn lanes, "Ionization growth and breakdown, " HAND. PHYS., vol.
`22. pp 192,195
`
`11 E. D. Lozanskii, "Photoionizing radiation in the streamer breakdown of a
`gas. " 50V. PHYS. TECH. PHYS, vol. 13, pp 1269-1272, Mar. 1969
`12 L. B. Loeb, "Electrical breakdown ofgases with steady of direct current
`impulse potentials, " HAND. PHYS., vol. 22, pp 445-53o,1956
`13 I. M- Meek. ”A theory of spark discharge, " PHYS
`REM, vol. 57, pp. 722-730, Mar. 1940
`
`14 W. Rogowski, ”Impulse potential and breakdown
`in gases, ” ARCH. ELECTROTECH., vol. 20, pp 1928
`
`15 L. B. Loeb, FUNDAMENTAL PROCESSES OF ELECTRICAL DISCHARGES IN
`GASES New York: Wiley, 1939
`
`16 G. M. Petropoulos, ”Avalanche transformation during breakdown in
`uniform field, ” PHYS. REM vol. 78, pp 250-253, May 1950
`
`17 R. C. Fletcher, "Impulse breakdown in the 1 o -9 sec range of air
`atmospheric pressure, " PHYS. REV., vol. 76, pp 1501-151 1, Aug. 1945
`18 A. Daran and /. Meyer, "Photographic and oscillographic investigations of
`spark discharge in hydrogen, " BRIT. /. APPL. PHYS., vol. 18, pp 793—799,
`Mar. 1967
`
`19 A. H. Guenther and /. R. Bettis, "The laser triggering of high-voltage
`switches, "j. PHYS. APPL. PHYS, vol. 1, pp 1577-1613, lune 196
`
`20 E R. Dickey, lr., "Contribution to the theory of impulse breakdown, " /.
`APPL. PHYS., vol. 23, p 1336-1339, Dec. 1952
`
`21 K. McDonald, M. Newton, E. Kunhardt, M. Kristiansen and A. H. Guenther,
`"An electron beam triggered spark gap, " IEEE TRANS. PLASMA 50., vol.
`P5~8(3) 181, Sept. 1980
`
`22 D. Schatz, "The MDX as a strategic tool in reducing arcing, " APPLICATION
`NOTES, Advanced Energy Industries, Inc., 1983
`
`23 PHYSICAL VAPOR DEPOSIT/0N (R. 1. Hill, ed.)
`
`

`

`
`
`

`

`
`
`

`

`
`
` ADVANCED
`
`ENERGY®
`
`© Advanced Energy Industries, inc. 2000
`All righrs reserved. Printed in U.S.A.
`SL'WHITES-Z'lOvOl 1M 08/00
`
`Advanced Energy Indhstries, Inc.
`1625 Sharp Point Drive
`Fort Collins, Colorado 80525
`800.446.9167
`970.221.4670
`970.221.5583 (fax)
`support@aei.com
`www.3dvanced—energy.com
`
`California
`T: 408.263.8784
`F: 408.263.8992
`
`New Jersey
`T: 856.627.6100
`F: 856.627.6159
`
`United Kingdom
`T: 44.1869.320022
`F: 44.1869.325004
`
`Germany
`T: 49.711.77927-0
`F: 49.711.7778700
`
`T: 82.31.705.2100
`F: 82.31.705.2766
`
`Japan
`T: 81.3.32351511
`F: 81.3.32353580
`
`Taiwan
`T: 886282215599
`F: 886282215050
`
`8
`
`