Pulsed-DC Reactive Sputtering of Dielectrics: Pulsing Parameter 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
`Al2O3
`
`ABSTRACT
`
`Pulsed DC power is used for reactive sputtering of dielectrics.
`At certain pulsing frequencies and duty cycles, deposition
`can be done without arcing and high 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. Arcing appears mainly on the portion of
`the target surface that is not strongly bombarded by positive
`ions and may therefore be covered with a dielectric layer. Such
`a dielectric layer accumulates positive charges, which if dense
`enough, may cause arcing. To avoid this arcing, reactive
`sputter-deposition of dielectric films is done using either
`pulsed-DC [1-14] or mid-frequency ac power [15-17]. Pulsed-
`DC power is implemented mainly with a single magnetron
`system, while ac power is generally used with a dual magne-
`tron arrangement. A waveform showing the effect of pulsed-
`DC power is shown in Figure 1. The pulsing parameters must
`satisfy certain conditions to avoid arcing. These conditions
`are usually satisfied when pulsing frequencies and duty cycle
`
`
`
`(defined as the ratio ␶on/(␶on +␶off) are chosen properly from the
`range of about 20-350 kHz and 0.5-0.9, respectively. Pulsing
`parameters such as duty 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 pulsing frequency and some details of pulsing
`parameter influences are described. Results are given for
`pulsed dc reactive sputtering of Al2O3 thin film.
`
`2. EXPERIMENTAL TECHNIQUE
`
`The experiments were done in a box coater with a planar
`rectangular unbalanced magnetron HRC-817 (BOC Coating
`Technology) previously described in [11,13,14]. The pulsed
`DC power was applied to the cathode using a DC power
`supply, model MDX-10 by Advanced Energy Industries (AEI),
`pulsed generator Sparcle®-V (AEI), and also by a PinaclePlus®
`unit (AEI). The DC power supply was run in the constant
`power and current modes. Another model of pulsed power
`supply, RPG (ENI), was used for comparison.
`
`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 voltage probe
`model P5100, and a current probe, model A6303, (all from
`Tektronix). The pulsed power consumed by the plasma was
`obtained by dynamic multiplication (point-by-point) of the
`current and voltage waveforms. Plasma emission was re-
`corded by two photomultipliers H5783 (Hamamatsu) with
`narrow filters attached to each of them to record Ar+ (425 nm)
`and atomic oxygen (777 nm) emission. The time constant of
`optical responses was less than 10 ns.
`
`3. PULSING PARAMETERS AND CATHODE
`MICROARCING
`
`It is commonly accepted that positive charges accumulated on
`the surface of a dielectric layer deposited (or grown) on the
`target surface, are the main source of arcs. Accumulation of
`these charges takes place during each on-time pulse, ␶
`on
`(Figure 1) as ions bombard the target surface. During the
`following off-time, ␶
`off, electrons from the residual plasma
`discharge the layer. If the discharge is not complete, addi-
`tional step-by-step charge accumulation during many sequen-
`tial periods takes place, and layer breakdown may still occur.
`Let us consider first only charge accumulation in a single
`on-time.
`
`If the electrical field in the dielectric layer created by the
`charge accumulated during an on-period exceeds the dielec-
`tric strength of the film, a breakthrough takes place in the film,
`and, as a result of a high level of electron emission from the
`
`86
`
`© 2000 Society of Vacuum Coaters 505/856-7188
`43rd Annual Technical Conference Proceedings—Denver, April 15–20, 2000 ISSN 0737-5921
`
`Page 1 of 5
`
`APPLIED MATERIALS EXHIBIT 1008
`
`

`

`Figure 2. First-approximation view of arcing-free conditions
`in the 2-d space of pulsing frequency and duty cycle and two
`different currents, I1 (solid lines) and I2 (dotted lines): I1 < I2.
`
`4. PULSING PARAMETERS AND DISCHARGE
`BEHAVIOUR
`
`Current and voltage pulses during reactive sputtering of a
`dielectric are shown schematically in Figure 3. An example of
`actual shapes of current and voltage pulses as well as optical
`emission pulses, is shown Figure 4 [11,13,14]. Let us consider
`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 plasma begins to
`decay through bipolar charge carrier diffusion to the wall of
`the chamber, and the electrodes. As the plasma density dimin-
`ishes, the electron current to the positive target decreases. As
`may be seen in Figure 3, the shape of the electron current
`waveform between times A and B may be described by
`est,1 = 30÷40 µs and
`two exponentials, with time constants of ␶
`
`st,2 = 3÷5 ms for the balanced magnetron and of ␶est,1 ⬇ 15 µs
`for the unbalanced magnetron [13,14]. The longest time
`constant is consistent with the overall plasma decay. These
`time constants are close to those observed in various high-
`density oxygen plasmas [19]. The larger time constant is
`consistent with Bohm bipolar plasma diffusion to the walls
`[20]. Note that the plasma emission in pulsed power reactive
`sputtering of Al2O3 recorded for the Al-line and O-line, also
`decay, but not at the same rate; to a first approximation, the
`emission lines decay exponentially with two time constants:
`0.2-0.3 µs and 1 µs. This may be explained by noting that the
`excitation of optical emission in the plasma requires involve-
`ment of electrons with energies above about 10 eV. It is to be
`expected that when the plasma starts to decay, the fast elec-
`trons disappear much faster the slow ones [20]. Although 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 plasma density at the end of a few
`microsecond-long off-time.
`
`␶e
`
`87
`
`breakthrough area, an arc appears. Estimations have been
`made [2,3,13,14,18] which show that, at the deposition
`conditions used in typical cases, a breakdown field
`of Ebr ⬇ 106 V/cm appears in the Al2O3 films in about
`
`r = 0.1-1 ms. To avoid such arcs, the on-time, ␶on, (Figure 1)
`should not larger than 0.1-1 ms and the pulsing frequency
`should be larger than the related critical frequency
`fcr.1 = 1-10 kHz.
`
`␶b
`
`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-
`track area. These might be called surface creepage break-
`down. The traces of these microarcs are seen on the target
`surface as scratches. Very little is known about their electrical
`parameters. To avoid this kind of arcing, 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, fcr.2. So, to avoid microarcing
`due to charge accumulation in a single on-time pulse, the on-
`time should less than a certain value, which is to say that the
`frequency should be larger than fcr.1 or fcr.2, whichever is
`higher. It is important to note that that fcr.1 or fcr.2 do not depend
`on the duty 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 during many sequential
`periods. To avoid such accumulation, the dielectric surfaces
`must be discharged completely during each off-time, ␶
`off (also
`called the “reverse time”, ␶
`rev). For this to occur, for each
`
`
`off,min exists to avoid arcing. It is obtainedperiod ␶on, a minimal ␶
`
`
`as a first approximation that ␶off,mi is linearly related to ␶on, and
`the relationship coefficient is directly proportional to the
`current [13,14]. The linear relationship between these dura-
`tions leads to a simple upper limitation on the duty cycle, dcr,
`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 on-time, off-time, or duty
`cycle, constant. The number of microarcs counted in a certain
`time is usually very low until the frequency is decreased to the
`critical frequency fcr. Continuous arcing is observed at fcr and
`lower [11,13,14]. In accordance with the discussed condi-
`tions, to avoid arcing and experimental data, the critical
`frequency depends on the duty as shown in Figure 2.
`
`Page 2 of 5
`
`

`

`Figure 4. Voltage, current, and optical emission of Ar and
`O oscillograms.
`
`5. PULSING PARAMETERS AND TAREGT STATE
`OF OXIDATION
`
`The target surface in the racetrack area of a magnetron in
`reactive sputtering can be in either of two stable states or
`modes: metallic and reactive. If the target is largely or com-
`pletely covered with oxide, it is said that the system is
`operating in the “oxide mode”. In the “metallic mode”, the
`racetrack area is practically clean from oxide, and sputtering
`of the metal is the dominant process. In addition to these stable
`
`Figure 3. Typical current and voltage oscillograms
`
`So, if the ␶
`off is of an order of a few microseconds or less, then
`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, ne,res depends on the off-time,
`
`ff; increasing ␶off decreases the density ne,res.
`
`␶o
`
`The negative voltage is turned on again at the end of the
`off-, or reverse-, time (Point B on the time scale, Figure 3). At
`this moment, the density ne,res determines the initial current, Io,
`and also the re-establishment time, tre-est (Figure 3). The higher
`the residual concentration of charge carriers ne,res (i.e., the
`smaller the off-time), the larger the current Io and lower the
`duration ␶
`re-est.
`
`Plasma re-establishment time depends on the residual plasma
`density ne,res, which increases with increasing cathode current
`re-est = 1-2 µs,
`and decreasing off-time. It takes, usually, a ␶
`sometimes even less, to recover to the original density. Such
`fast plasma re-establishment is promoted by a voltage over-
`shoot created by the dc power supply, which will occur if the
`power supply has any substantial output inductance.
`
`The unbalanced magnetron plasma decay time constants are
`higher than those of the balanced magnetron. This could be
`related to the fact that the dense part of the unbalanced
`magnetron plasma occupies much larger volume, and its
`decay thus requires a longer time.
`
`88
`
`Page 3 of 5
`
`

`

`modes, the target racetrack area can be maintained also in an
`intermediate mode, which is not ordinarily stable, by employ-
`ing a closed-loop control system. In all cases, the balance of
`sputtering and oxidation rates determines the racetrack area’s
`surface conditions. Sputtering takes place only during the on-
`time, while oxidation takes place in both parts of the cycle,
`during the off-time as well as the on-time. During the on-time,
`oxidation takes place mainly by atomic oxygen generated in
`the plasma. During 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).
`
`Figure 6. Deposition rate versus on-pulse average power at
`different constant duty cycles.
`
`Figure 5. Voltage versus off-time at constant on time (upper)
`and versus on-time at constant on-time (lower).
`
`Changing the duty cycle will change the equilibrium between
`oxidation and sputtering processes and result in a different
`average thickness of the oxide layer on the target surface. It is
`worthwhile to notice that the thickness of the oxide layer, even
`in the oxide mode, is very small; if this were not true, a
`substantial voltage drop would build up due to charging of the
`oxide layer, which would cut off the discharge. In spite of
`being very thin, however, slightly different thickness in the
`oxide layer leads to variation of the cathode voltage and
`deposition rate. An effect related to the target oxidation state
`is actually seen in the variation of the cathode voltage as
`shown in Figure 5. It is also seen in a variation of the linear
`relationship between the deposition rate and power consumed
`in the plasma at different constant duty cycles (Figure 6).
`
`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 pulsed power delivered to the plasma (i.e., the power
`delivered during the on-time) is slightly lower than the power
`generated by supply (Figure 7). The difference between the
`two increases with frequency; this is probably due to power
`loss in the pulsing device (the electronic switches dissipate 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 [13,14]
`that the deposition rate in pulsed systems is directly propor-
`tional to the power delivered to the plasma during the on-time.
`
`The power delivered to magnetron to generate the plasma is
`released mainly in sputtering of the target material, but also
`can be lost through heating all surfaces: the target, chamber
`walls, and substrates. The power distribution between the
`sputtering and other processes depends on the pulsing fre-
`
`89
`
`Page 4 of 5
`
`

`

`quency. For example, increasing the frequency increases
`substrate and wall heating, due to increased ion bombardment
`due to increased floating potential, in turn caused by increased
`electron temperature and plasma density at surfaces remote to
`the plasma.
`
`7. CONCLUSION
`
`Pulsing power is used to perform reactive sputtering of dielec-
`trics. Its implementation has discovered various processes
`that influence deposition characteristics. These processes
`include plasma dynamics, sputtering and oxidation of the
`target, deposition of an oxide layer on the anode surface and
`its resputtering, and others. The analysis shows that:
`
`• Reactive sputtering of dielectrics can be performed
`without arcing using pulsing frequencies that exceed the
`critical frequency, and a duty cycle that is less than the
`critical duty cycle.
`• The critical frequency increases with the power (or cur-
`rent) delivered to the target.
`• The critical duty cycle decreases as the target power or
`current increases.
`• Plasma dynamics determines the kinetics of the voltage
`and current pulses
`• Pulsing parameters (frequency and off-time) affect the
`principal reactive sputtering parameters: voltage, deposi-
`tion rate, substrate heating, etc.
`Increasing the duty cycle at constant power provided by the
`power supply decreases the oxidation state of the target
`surface and, therefore, increases the deposition rate
`Increasing the pulsing frequency increases power loss,
`both in the pulser switching devices and in substrate and
`wall heating. As the power measured by the power supply
`does not include these losses, there is an apparent decrease
`in specific deposition rate (deposition rate per watt) due to
`a decreasing fraction of the measured power being deliv-
`ered to the target. 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-
`ings, 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).
`
`3. R. Scholl, in 37th Annual Technical Conference Proceed-
`ings, Boston (Society of Vacuum Coaters) p. 312 (1994).
`
`90
`
`
`
`View publication statsView publication stats
`
`4. W.D. Sproul, M.E. Graham, M.S. Wong, S. Lopez, D. Li,
`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. Arnell, 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-
`ceedings, Philadelphia, PA (Society of Vacuum Coaters),
`p. 123(1996).
`
`9. S. Schiller, V. Kirchoff, K. Goedicke, and P. Frach, in 40th
`Annual Technical Conference Proceedings, New Orleans,
`LA (Society of Vacuum Coaters), 129(1997).
`
`10. O. Zywitzki and G. Hoetzsch, Surf.Coat.Technol., 94-95,
`303 (1997).
`
`11. A. Belkind, A. Freilich, and R. Scholl, in 41st Annual
`Technical Conference Proceedings, Boston, 1998 (Soci-
`ety of Vacuum Coaters), p. 321 (1998).
`
`12. J.M. Schneider and W.D. Sproul, in 98/1 Reactive Sput-
`tering. W.D. Westwood, ed. (Institute of Physics Publish-
`ing, Bristol and Philadelphia), p. A5.1:1 (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. Cormia 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).
`
`18. J. Szczyrbowski, G. Brauer, W. Dicken, M. Scherer,
`W. Maas, G. Teschner, A. Zmelty, Surf.Sci.Tcehnol., 93,
`14 (1997).
`
`19. J.P. Booth and N. Sadeghi, J.Appl.Phys., 70, 611 (1991).
`
`20. M.A. Lieberman, Plasma Sources Sci,Technol., 5, 145
`(1996).
`
`Page 5 of 5
`
`