`Case 6:20-cv-00636—ADA Document 87-2 Filed 03/29/21 Page 1 of 6
`
`EXHIBIT 23
`
`EXHIBIT 23
`
`
`
`Case 6:20-cv-00636-ADA Document 87-2 Filed 03/29/21 Page 2 of 6
`
`Pulsed-DC Reactive Sputtering of Dielectrics: Pulsing Para.meter Effects
`
`A. Belkind and Z. Zhao, Stevens Institute of Technology, Hoboken, NJ;
`and D. Carter, L. Mahoney, G. McDonough, G. Roche, R. Scholl and H. Walde,
`Advanced Energy Industries, Inc., Fort Collins, CO
`
`Key Words:
`
`Pulsed deposition
`Pulsed plasma
`
`Reactive sputtering
`AIO
`
`ABSTRACT
`
`of dielectrics.
`Pulsed DC power is used for reactive
`At certain pulsing frequencies and duty cycles, deposition
`can be done without arcing and
`deposition rates can be
`achieved. In this work, the influence of frequency, on-time
`and off-time durations on the deposition process are
`investigated.
`
`I. INTRODUCTION
`
`DC reactive sputtering of dielectrics is often accompanied
`with strong
`Arcing appears mainly on the portion of
`the
`surface that is not strongly bombarded by positive
`ions and may therefore be covered with a dielectric
`Such
`a dielectric layer accumulates positive
`which if dense
`enough, may cause
`To avoid this arcing, reactive
`sputter-deposition of dielectric films is done
`either
`pulsed-DC
`or mid-frequency ac power [ 15-1 7]. Pulsed-
`DC power is implemented
`with a single magnetron
`system, while ac power is
`used with a dual magne-
`tron arrangement. A waveform showing the effect of pulsed(cid:173)
`DC power is shown in
`1. The pulsing parameters must
`satisfy certain conditions to avoid
`These conditions
`are
`satisfied when pulsing frequencies and
`cycle
`( defined as the ratio Tor/( T
`,, +T011) are chosen properly from the
`range of about 20-350 kHz and 0.5-0.9, respectively. Pulsing
`parameters such as
`cycle and off-time influence also the
`forms of voltage and current peaks, deposition rate and the
`properties of the deposited thin films. In this paper, the proper
`choice of
`frequency and some details of
`parameter influences are described. Results are given for
`pulsed de reactive sputtering of Alp3 thin film.
`
`0
`
`2. EXPERIMENTAL TECHNIQUE
`
`The experiments were done in a box coater with a planar
`unbalanced magnetron HRC-817 (BOC Coating
`Technology) previously described in [11
`The pulsed
`DC power was applied to the cathode
`a DC power
`supply, modelMDX-lOby Advanced Energy Industries (AEI),
`pulsed generator Sparcle(g)-V (AEI), and also by a PinaclePlus®
`was nm in the constant
`unit (AEI). The DC power
`power and current modes. Another model of pulsed power
`supply, RPG (ENI), was used for comparison.
`
`..---- Discharging the dielectric surfaces
`
`T<J'Cle
`Sputtet- deposit.ion of dielectric layers and
`charging up dielectric surfaces
`
`Figure 1. Pulsed voltage used to power a cathode
`
`The current and voltage pulse forms were recorded using an
`oscilloscope model TDS 340, connected to a
`probe
`model P5100, and a current probe, model A6303, (all from
`Tektronix). The
`power consumed by the
`was
`obtained by dynamic
`(point-by-point) of the
`current and voltage waveforms. Plasma emission was re(cid:173)
`corded by two photomultipliers H5783 (Hamamatsu) with
`narrow filters attached to each of them to record Ar+ ( 425 nm)
`and atomic oxygen
`nm) emission. The time constant of
`optical responses was less than 10 ns.
`
`3. PULSING PARAMETERS AND CATHODE
`MICROARCING
`
`charges accumulated on
`It is commonly accepted that
`the surface of a dielectric layer deposited (or
`on the
`are the main source of arcs. Accumulation of
`these charges takes place
`each on-time
`T
`(Figure 1) as ions bombard the target surface. During the
`following
`T "" electrons from the residual plasma
`OJI
`discharge the layer. If the discharge is not complete, addi-
`tional step-by-step
`accumulation
`many sequen-
`tial periods takes
`and layer breakdown may still occur.
`Let us consider first only charge accumulation in a single
`on-time.
`
`,,
`
`0
`
`created by the
`If the electrical field in the dielectric
`an on-period exceeds the dielec-
`takes place in the
`tric strength of the
`and, as a result of a high level of electron emission from the
`
`86
`
`© 2000 Society of Vacm1m Coaters 505/856-7188
`43rd Annual Technical Conference Proceedings-Denver, April 15-20, 2000
`ISSN 0737-5921
`
`DEFTS-PA_0001100
`
`
`
`Case 6:20-cv-00636-ADA Document 87-2 Filed 03/29/21 Page 3 of 6
`
`breakthrough area, an arc appears. Estimations have been
`made [2,3,13,14,
`which show
`at the deposition
`conditions used in
`cases, a breakdown field
`of Eb, = 106 V/cm appears in the Alp3 films in about
`T = 0.1-1 ms. To avoid such arcs, the ,__.,~rn Ton' (Figure 1)
`br
`should not larger than 0.1-1 ms and the
`frequency
`should be larger than the related critical frequency
`= 1-10 kHz.
`
`Another kind of microarc can appear when a high electrical
`field is created between positive charges accumulated on the
`dielectric layer and a nearby open metal surface of the race(cid:173)
`track area. These
`be called surface creepage break(cid:173)
`down. The traces of these microarcs are seen on the
`surface as scratches. Very little is known about their electrical
`To avoid this kind of
`the on-time should
`not be larger than some duration dependent upon surface
`conditions; that is to say, the pulsing frequency should be
`larger than a critical frequency,j~rr
`to avoid
`due to
`accumulation in a single on-time pulse, the on-
`time should less than a certain
`which is to say that the
`frequency should be larger than
`whichever is
`higher. It is
`to note that
`do not depend
`on the
`cycle, because they relate to an arc formed by
`charge buildup in a single on-time period.
`
`Another limitation for pulsing frequency comes from possible
`step-by-step charge accumulation
`many sequential
`periods. To avoid such
`the dielectric surfaces
`must be discharged completely
`each off-time,
`(also
`called the "reverse time",
`For this to occur,
`each
`It is Obtained
`period Ton' a minimal
`as a first
`the relationship coefficient
`current [13,
`The linear relationship between these dura-
`tions leads to a
`upper limitation on the
`cycle,
`which decreases as the current increases.
`
`Experimentally, the conditions for avoiding microarcs may be
`verified by counting microarcs at different pulsing frequen-
`cies. This may be done keeping
`or
`cycle, constant. The number of microarcs counted in a certain
`time is
`very low until the frequency is decreased to the
`critical
`Continuous
`is observed atfc, and
`lower [11,13,14]. In accordance with the discussed condi(cid:173)
`tions, to avoid
`and experimental
`the critical
`frequency depends on the duty as shown in Figure 2.
`
`CT)
`C
`
`:::;
`0...
`
`IXX;:<XJ<)()Z>CXX.x.xx_::,,::_;.:,;,~x.x.x,,
`E>ielectric film
`breakthrough
`limitation
`~~~~~...;:;.;:..~Ul~..:i.:-....
`
`Duty cycle
`
`view of arcing-free conditions
`Figure 2.
`frequency and
`and two
`in the 2-d space of
`different currents, I 1 (solid lines) and I2 (dotted lines): I 1 < I2
`•
`
`4. PULSING PARAMETERS AND DISCHARGE
`BEHAVIOUR
`
`reactive sputtering of a
`Current and
`in Figure 3. An example of
`dielectric are shown
`pulses as well as
`actual shapes of current and
`4 [ 11,
`Let us consider
`emission pulses, is shown
`a moment when the cathode voltage is switched to a positive
`value (Point A on the time scale, Figure 3). Since at this
`moment the plasma support is turned off, the
`to
`decay
`bipolar
`carrier diffusion to the wall of
`the
`and the electrodes. As the
`dimin-
`ishes, the electron current to the
`target decreases. As
`may be seen in
`3, the shape of the electron current
`waveform between times A and B may be described by
`= 30+40 µs and
`two exponentials, with time constants of
`T = 3+5 ms for the balanced magnetron
`= 15 ~Ls
`of
`est,2
`.
`for the unbalanced magnetron
`The longest t1me
`constant is consistent with the overall
`These
`time constants are close to those observed in various
`density oxygen
`The larger time constant is
`consistent with Bohm
`diffusion to the walls
`[20]. Note that the plasma emission in pulsed power reactive
`of Al?O., recorded for the Al-line and 0-line, also
`decay, but not at the same rate; to a first
`the
`emission lines decay exponentially with two time constants:
`0.2-0.3 µs and 1 ~Ls. This may be explained by
`that the
`excitation of optical emission in the plasma requires involve(cid:173)
`ment of electrons with energies above about 10 e V. It is to be
`expected that when the plasma starts to decay, the fast elec(cid:173)
`trons disappear much faster the slow ones
`the
`time constants obtained for optical emission are less that the
`ones obtained for electron current decay, they still confirm
`that there is a substantial
`density at the end of a few
`microsecond-long off-time.
`
`87
`
`DEFTS-PA_0001101
`
`
`
`Case 6:20-cv-00636-ADA Document 87-2 Filed 03/29/21 Page 4 of 6
`
`/Tdf
`
`0
`
`IC
`
`,I.>
`
`:;::
`il.l
`
`~ u
`
`Time
`
`Figure 3. Typical current and voltage oscillograms
`
`of an order of a few microseconds or less, then
`So, if the
`at this moment a plasma still exists in the space between the
`cathode and anode, although its density has fallen off. The
`density of this residual plasma,
`depends on the
`· increasing
`decreases the
`
`The negative voltage is turned on again at the end of the
`off-, or reverse-, time (Point Bon the time scale, Figure 3). At
`this moment, the density
`determines the initial current, 1
`and also the re-establishment
`(Figure The higher
`the residual concentration of charge carriers ne.res (i.e., the
`smaller the off-time), the larger the current / 0 and lower the
`duration Tre-est"
`
`0
`
`,
`
`Plasma re-establishment time depends on the residual
`density n
`, which increases with increasing cathode current
`= 1-2 µs,
`and decr~~~ing off-time. It takes, usually, a
`sometimes even less, to recover to the
`Such
`fast plasma re-establishment is promoted
`a voltage over-
`shoot created by the de power supply, which will occur if the
`power supply has any substantial
`inductance.
`
`decay time constants are
`The unbalanced magnetron
`This could be
`higher than those of the balanced
`related to the fact that the dense part of the unbalanced
`magnetron
`occupies much larger volume, and its
`decay thus requires a longer time.
`
`88
`
`Figure 4. Voltage, current, and optical emission of Ar and
`0 oscillograms.
`
`5. PULSING PARAMETERS AND TAREGT STATE
`OF OXIDATION
`
`The target surface in the racetrack area of a magnetron in
`reactive
`can be in either of two stable states or
`modes: metallic and reactive. If the
`is largely or com(cid:173)
`pletely covered with oxide, it is said that the
`is
`in the "oxide mode". In the "metallic mode", the
`racetrack area is practically clean from
`and
`of the metal is the dominant process. In addition to these stable
`
`DEFTS-PA_0001102
`
`
`
`Case 6:20-cv-00636-ADA Document 87-2 Filed 03/29/21 Page 5 of 6
`
`racetrack area can be maintained also in an
`the
`intermediate mode, which is not ordinarily
`by
`ing a closed-loop control
`In all cases, the balance of
`sputtering and oxidation rates determines the racetrack area's
`surface conditions. Sputtering takes place only during the on(cid:173)
`time, while oxidation takes place in both
`of the cycle,
`the off-time as well as the on-time. During the on-time,
`oxidation takes place
`by atomic oxygen generated in
`the
`the off-time, oxidation is provided by the
`decaying concentration of atomic oxygen that at the end of an
`off-time is still high (emission of atomic oxygen line is still
`strong).
`
`2 -
`
`~ 2.4 -
`
`l
`
`Q.)
`
`~
`'"'
`,::1
`0
`:-E
`"' 1.6 -
`0
`0...
`Q.)
`Ci
`
`1.2
`
`2
`
`i 90%, y = 0 .0088x
`
`70%, y = 0.0051x
`y
`
`11111
`
`11111 y-
`
`.A.
`
`.;;;;;
`1111
`
`50%, y = 0 .0036x Ar.
`'
`4
`3
`On-pulse average power (',1,T)
`
`I
`
`5
`
`300
`~
`o.O 250
`~
`;!:::l
`0 :>
`
`(D
`
`200
`
`~
`
`•D
`ti.I)
`,:,;:;
`;!:::l
`
`0 :>
`
`0
`
`2
`Off-time (µ.s)
`
`4
`
`280
`
`260 -
`
`240
`
`4
`
`•<>!'I·= 2 µ.s /
`
`I
`
`6
`On-time (µ.s)
`
`'
`8
`
`5. Voltage versus off-time at constant on time (upper)
`and versus on-time at constant on-time (lower).
`
`between
`Changing the duty cycle will change the
`oxidation and sputtering processes and result in a different
`average thickness of the oxide
`on the
`surface. It is
`worthwhile to notice that the thickness of the oxide layer, even
`in the oxide mode, is very
`if this were not true, a
`substantial voltage drop would build up due to charging of the
`oxide layer, which would cut off the
`In spite of
`being very
`however, slightly different thickness in the
`oxide layer leads to variation of the cathode
`and
`uv1Jvc,un.m rate. An effect related to the target oxidation state
`is actually seen in the variation of the cathode voltage as
`shown in
`5. It is also seen in a variation of the linear
`relationship between the
`rate and power consumed
`in the plasma at different constant duty cycles (Figure 6).
`
`""'" = 6 µs
`
`6. Deposition rate versus on-pulse average power at
`different constant duty cycles.
`
`4.0
`
`s--
`>
`6
`'- 3.5
`Q)
`:§:
`0
`o_
`
`•
`
`0
`
`2·1 00
`~
`~
`
`Q)
`
`18
`
`...,_
`
`•
`
`b --()
`
`-
`
`C
`0
`
`t'l
`0
`D..
`0
`
`3.0
`
`0
`
`300
`200
`100
`Frequency (Hz)
`
`15
`400
`
`Figure 7. On-pulse average power (closed diamonds) and
`deposition rate (open circles) versus frequency at constant
`power of 2.5 kW and 0.7 duty cycle.
`
`6. PULSING PARAMETERS AND POWER
`DISSIPATION
`
`The
`
`the power
`power delivered to the plasma
`the
`is slightly lower than the power
`generated by supply
`The difference between the
`two increases with frequency; this is probably due to power
`loss in the pulsing device (the electronic switches
`a
`little energy for each switch operation, and thereby power loss
`increases with frequency). Deposition rate, measured in the
`power supply's constant power mode, also decreases with
`increasing frequency (Figure 7). It has been shown
`that the deposition rate in pulsed systems is directly propor(cid:173)
`tional to the power delivered to the
`during the on-time.
`
`is
`uiu.j',rn~u,,u to generate the
`but also
`of the target
`chamber
`all surfaces: the
`walls, and substrates. The power distribution between the
`sputtering and other processes depends on the pulsing fre-
`
`89
`
`DEFTS-PA_0001103
`
`
`
`Case 6:20-cv-00636-ADA Document 87-2 Filed 03/29/21 Page 6 of 6
`
`quency. For example, increasing the frequency increases
`substrate and wall heating, due to increased ion bombardment
`due to increased
`in turn caused by increased
`electron
`density at surfaces remote to
`the plasma.
`
`7. CONCLUSION
`
`Pulsing power is used to perform reactive sputtering of dielec(cid:173)
`trics. Its implementation has discovered various processes
`that influence deposition characteristics. TI1ese processes
`include
`and oxidation of the
`deposition of an oxide layer on the anode surface and
`its resputtering, and others. TI1e
`shows that:
`
`of dielectrics can be performed
`• Reactive
`using pulsing frequencies that exceed the
`without
`critical frequency, and a
`cycle that is less than the
`critical
`cycle.
`• The critical frequency increases with the power (or cur(cid:173)
`rent) delivered to the
`• The critical
`cycle decreases as the
`current increases.
`• Plasma
`
`determines the kinetics of the voltage
`
`power or
`
`affect the
`
`• Pulsing
`principal reactive
`etc.
`tion rate, substrate
`• Increasing the duty cycle at constant power provided by the
`power supply decreases the oxidation state of the target
`surface
`therefore, increases the deposition rate
`• Increasing the
`frequency increases power loss,
`both in the pulser switching devices and in substrate and
`wall
`As the power measured
`the power supply
`does not include these losses, there is an apparent decrease
`in specific deposition rate (deposition rate per
`due to
`a decreasing fraction of the measured power being deliv(cid:173)
`ered to the
`This effect disappears when the actual
`power delivered to the plasma is measured and used as the
`power figure.
`
`REFERENCES
`
`1. R. Scholl, in 36th Annual Technical Conference Proceed(cid:173)
`in gs, Dallas (Society of Vacuum Coaters), p. 405 (1993).
`
`2. S. Schiller, K. Goedicke, J. Reschke, V. Kirchke,
`S. Schneider, and F. Milde, Surf.Coat.Technol., 61, 331
`(1993).
`
`M.S. Wong, S. Lopez, D.
`4. W.D. Sproul, M.E.
`and R.A. Scholl, J.Vac.Sci.Technol., A 13, 1188 (1995).
`
`5. F. Fietzke, K. Goedicke, and W. Hempel, Surf.
`Coat.Technol., 86-87, 657 (1996).
`
`6. P.J. Kelly, O.A. Abu-Zeid, R.D. Amell, and J. Tong, Surf.
`Coat.Technol., 86-87, 28 (1996).
`
`7. V. Kirchoff and T. Kopte, in 39th Annual Technical
`Conference Proceedings, Philadelphia, PA (Society of
`Vacuum Coaters), p. 117 (1996).
`
`8.
`
`J.C. Sellers. in 39th Annual Technical Conference Pro(cid:173)
`ceedings, Philadelphia, PA (Society of Vacuum Coaters),
`p. 123(1996).
`
`V. Kirchoff, K. Goedicke, and P.
`9. S.
`Annual Technical Conference Proceedings, New Orleans,
`LA (Society of Vacuum Coaters), 129(1997).
`
`10. 0. Zywitzki and G. Hoetzsch, Surf.Coat.Technol., 94-95,
`303 (1997).
`
`11. A. Belkind, A. Freilich, and R. Scholl, in 41 st Annual
`Technical Conference Proceedings, Boston, 1998 (Soci(cid:173)
`ety of Vacuum Coaters), p. 321 (1998).
`
`12. J.M. Schneider and W.D. Sproul, in"-""'~==-'-"-="'--'(cid:173)
`tering. W.D. Westwood, ed. (Institute of Physics Publish(cid:173)
`ing, Bristol and Philadelphia), p. A5.l:l (1998).
`
`13. A. Belkind, A. Freilich, and R. Scholl, Surf. Coat. Technol.,
`108-109, 558 (1998).
`
`14. A. Belkind, A. Freilich, and R. Scholl, J. Vac.Sci.Technol.,
`A 17, 1934 (1999).
`
`15. R.L. Com1ia and T. Tumbly, US Patent 4,046,659 (Sept.
`1977).
`
`16. G. Este and W.D. Westwood, J.Vac.Sci.Technol., A 6,
`1845 (1988).
`
`17. D.A. Glocker, J.Vac.Sci.Technol., A 11, 2989 (1993).
`
`G. Brauer, W. Dicken, M. Scherer,
`18. J.
`W. Maas, G. Teschner, A. Zmelty, Surf.Sci.Tcehnol., 93,
`14 (1997).
`
`3. R. Scholl, in 37th Annual Technical Conference Proceed(cid:173)
`ings, Boston (Society of Vacuum Coaters) p. 312 (1994).
`
`19. J.P. Booth and N. Sadeghi, J.Appl.Phys., 70, 611 (1991).
`
`20. M.A. Lieberman, Plasma Sources Sci.Technol., 5, 145
`(1996).
`
`90
`
`DEFTS-PA_0001104
`
`