`process for alumina films
`
`Cite as: Journal of Vacuum Science & Technology A 18, 2890 (2000); https://doi.org/10.1116/1.1319679
`Submitted: 17 March 2000 . Accepted: 28 August 2000 . Published Online: 10 November 2000
`
`P. J. Kelly, P. S. Henderson, R. D. Arnell, G. A. Roche, and D. Carter
`
`ARTICLES YOU MAY BE INTERESTED IN
`
`Control of the structure and properties of aluminum oxide coatings deposited by pulsed
`magnetron sputtering
`Journal of Vacuum Science & Technology A 17, 945 (1999); https://doi.org/10.1116/1.581669
`
`Reactive direct current magnetron sputtering of aluminum oxide coatings
`Journal of Vacuum Science & Technology A 13, 1188 (1995); https://doi.org/10.1116/1.579859
`
` Tutorial: Reactive high power impulse magnetron sputtering (R-HiPIMS)
`Journal of Applied Physics 121, 171101 (2017); https://doi.org/10.1063/1.4978350
`
`Journal of Vacuum Science & Technology A 18, 2890 (2000); https://doi.org/10.1116/1.1319679
`
`18, 2890
`
`© 2000 American Vacuum Society.
`
`Page 1 of 8
`
`INTEL EXHIBIT 1059
`
`
`
`Reactive pulsed magnetron sputtering process for alumina films
`P. J. Kelly,a) P. S. Henderson, and R. D. Arnell
`Centre for Advanced Materials and Surface Engineering, University of Salford, Salford, M5 4WT,
`United Kingdom
`G. A. Roche and D. Carter
`Advanced Energy Industries Inc., Fort Collins, Colorado 80525
`~Received 17 March 2000; accepted 28 August 2000!
`The pulsed magnetron sputtering ~PMS! process is now among the leading techniques for the
`deposition of oxide films. In particular, the use of pulsed dc power has transformed the deposition
`of dielectric materials, such as alumina. The periodic target voltage reversals during the PMS
`process effectively discharge poisoned regions on the target. This significantly reduces the
`occurrence of arc events at the target and stabilizes the deposition process. Many researchers have
`now shown that pulsed dc reactive magnetron sputtering can be routinely used to produce fully
`dense, defect-free oxide films. Despite the success of the PMS process, few detailed studies have
`been carried out on the role played by parameters such as pulse frequency, duty cycle, and reverse
`voltage in the deposition process. In this study, therefore, alumina films were deposited by reactive
`pulsed dc magnetron sputtering. Operating conditions were systematically varied and the deposition
`process monitored throughout. The aim was to investigate the influence of the pulse parameters on
`the deposition process, and the interrelationships between the occurrence of arc events and the
`parameters chosen. As a result of this investigation, optimum conditions for the production of
`high-quality alumina films under hard arc-free conditions were also identified. © 2000 American
`Vacuum Society. @S0734-2101~00!04806-5#
`
`I. INTRODUCTION
`
`Since the initial development work in the early 1990s,1–4
`the pulsed magnetron sputtering ~PMS! process has become
`established as one of the leading techniques for the deposi-
`tion of oxide films. In particular, the use of pulsed dc power
`has transformed the deposition of dielectric materials, such
`as alumina.1–3,5–9 The process itself has been well described
`in various review articles,3,6,8–14 and no repetition is required
`here. It is sufficient to state that pulsed dc reactive magnetron
`sputtering offers significant advantages over conventional,
`continuous dc processing.14 If the magnetron discharge is
`pulsed in the bipolar mode ~see Fig. 1! at frequencies, usu-
`ally, in the range 10–200 kHz, the periodic target voltage
`reversals effectively discharge poisoned regions on the tar-
`get. This significantly reduces the occurrence of arc events at
`the target and stabilizes the deposition process. Many re-
`searchers have now shown that pulsed dc reactive magnetron
`sputtering can be routinely used to produce fully dense,
`defect-free oxide films. All stoichiometries are available,5,6,8
`arc events are suppressed,1–3,6–9,15–17 deposition rates can ap-
`proach those obtained for metallic films,2,3,7,15,16 and in dual-
`cathode systems, very long-term ~.300 h! process stability
`is attainable.18,19 As a consequence, very significant
`im-
`structure,5,7,8
`provements have been observed in the
`hardness,7,8 and optical properties6,13 of PMS alumina films,
`compared to dc sputtered films.
`The target voltage wave form during asymmetric bipolar
`pulsed dc sputtering is shown schematically in Fig. 1. Refer-
`ring to Fig. 1, the critical parameters which make up the
`
`a!Electronic mail: p.kelly@salford.ac.uk
`
`wave form are the pulse frequency, duty factor, and reverse
`voltage. Duty factor is the relative proportion of the pulse
`cycle made up of the ‘‘pulse-on’’ period, when the target
`voltage is negative and sputtering is occurring. The reverse
`voltage is the nominal positive target voltage achieved dur-
`ing the ‘‘pulse-off’’ period, often expressed as a percentage
`of the mean-negative voltage during the pulse-on period. The
`schematic wave form in Fig. 1 shows a pulse frequency of
`100 kHz, with a duty factor of 80%, and the reverse voltage
`set at 20% of the pulse-on voltage. In practice, this ‘‘square’’
`wave form is not achieved due to the inherent characteristics
`of the plasma and the power delivery system, with both posi-
`tive and negative voltage overshoots being observed.20 These
`artifacts can be clearly seen in Fig. 2, an oscilloscope trace of
`the target voltage wave form obtained when actually operat-
`ing under the conditions defined previously.
`Reference has already been made to the many examples
`in the literature of the success of the PMS process. However,
`as yet, few detailed studies have been published on the role
`played by the pulse parameters in the deposition process.
`Belkind, Freilich, and Scholl,9,10 derived an expression
`showing that the critical pulse frequency for arc-free opera-
`tion depends on the discharge current and the pulse-off time.
`Although not explicitly stated, their study indicates that, for a
`given discharge current, the duty factor is actually the most
`critical parameter in establishing arc-free conditions. Also,
`these studies did not consider time-dependent effects, since
`arc counting was only carried out for 3 min per run. In situ-
`ations where, during each pulse-off cycle, the parameters se-
`lected only partially discharge the poisoned regions on the
`target, a residual charge will accumulate until, eventually,
`
`2890
`
`J. Vac. Sci. Technol. A 18(cid:132)6(cid:133), Nov(cid:213)Dec 2000
`
`0734-2101(cid:213)2000(cid:213)18(cid:132)6(cid:133)(cid:213)2890(cid:213)7(cid:213)$17.00 ©2000 American Vacuum Society
`
`2890
`
`Page 2 of 8
`
`
`
`2891
`
`Kelly etal.: Reactive pulsed magnetron sputtering process
`
`2891
`
`FIG. 1. Schematic representation of the target voltage wave form during
`asymmetric bipolar pulsed sputtering ~pulse frequency5100 kHz, reverse
`time52 ms, duty580%, and reverse voltage520%!.
`
`FIG. 2. Oscilloscope trace of the target voltage wave form when operating in
`asymmetric bipolar pulsed mode at 100 kHz ~80% duty and 20% reverse
`voltage!.
`
`arcing occurs. Thus, conditions which appear to prevent arc-
`ing at the beginning of a deposition run can prove ineffective
`as the run progresses.
`In this study, alumina films were deposited by reactive
`pulsed dc magnetron sputtering. Operating conditions were
`systematically varied and the deposition process monitored
`throughout. The aim was to investigate the influence of the
`pulse parameters, such as pulse frequency, duty factor, and
`reverse voltage, and the interrelationships between the occur-
`rence of arc events and the parameters chosen. As a result of
`this investigation, optimum conditions for the production of
`high-quality alumina films under hard arc-free conditions
`were also identified.
`
`II. EXPERIMENT
`The commercial interest generated by the PMS process
`has led to the development of new power delivery systems.
`These include ac supplies, single- and dual-channel pulsed
`dc supplies, and pulse units which can be connected in series
`with the output from standard dc magnetron drivers. This
`article concentrates on the use of this latter type of system, in
`which the magnetron discharge could be pulsed over the fre-
`quency range 1–100 kHz. Parallel studies are also being
`made of the latest generation of pulsed dc supply which ex-
`tends the maximum pulse frequency up to 350 kHz.21,22
`The dc power supplies used in this study were the Ad-
`vanced Energy MDX and Pinnacle magnetron drivers. These
`power supplies were used in conjunction with the Advanced
`Energy Sparc-le V pulse unit. The Sparc-le V unit allows the
`pulse parameters to be varied over the following ranges; fre-
`quency: 1–100 kHz, reverse time: 1–10 ms, and reverse volt-
`age: 10%–20%. The dc supplies were operated in current
`regulation mode.
`The Sparc-le V unit allows both hard arc and microarc
`events to be monitored. Hard arcs are generally considered to
`be a discharge which takes place between a region on the
`cathode and an earthed surface, whereas microarcs are dis-
`charges between different sites on the cathode. While micro-
`
`JVST A - Vacuum, Surfaces, and Films
`
`arcs can normally be tolerated, hard arc events are extremely
`detrimental to the deposition process.3,8 Thus, in this study
`only the incidence of hard arcs was monitored.
`The work performed here was carried out in a Teer Coat-
`ings Ltd. UDP 450 closed-field unbalanced magnetron sput-
`tering rig, which has been described in detail elsewhere.7,8
`Alumina films were deposited by reactive unbalanced mag-
`netron sputtering from a 99.5% pure Al target. In all cases
`the base pressure was ,231025 mbar, the argon flow rate
`was adjusted to give a chamber pressure of 231023 mbar
`prior to deposition, and the target current was set to 6 A. The
`target was precleaned with the substrates shuttered, but no
`sputter cleaning of the substrates themselves was carried out.
`In fact, the substrate holder was allowed to float electrically
`throughout. The flow of reactive gas was controlled by an
`optical emissions monitoring ~OEM! system tuned to the 396
`nm line in the Al emission spectrum. An OEM turn-down
`signal of 25% was used for all depositions, i.e., reactive gas
`was allowed into the chamber until the OEM signal had
`fallen to 25% of the initial 100% metal signal. A feedback
`loop then maintained the OEM signal at this value for the
`duration of the deposition run, which was typically 90 min.
`Previous experience had shown that such conditions would
`produce stoichiometric Al2O3 films.8
`Figure 3 shows the characteristic hysteresis behavior of
`this system as the oxygen flow rate is varied. As the oxygen
`flow is increased initially, the target voltage rises slightly.
`Operating in this ‘‘metallic’’ regime could result in the for-
`mation of a substoichiometric aluminum oxide film. At a
`flow rate of approximately 13 sccm of oxygen, the target
`poisons rapidly and the negative target voltage falls from 395
`to 250 V. The target then remains poisoned until the O2 flow
`rate is reduced to ,4 sccm. Operating in the ‘‘poisoned’’
`regime would produce stoichiometric films, but at very much
`reduced deposition rates. The OEM system allows control to
`be maintained at any point on the hysteresis curve. Figure 4
`shows the relationship between target voltage and the OEM
`setting, expressed as a percentage of the 100% metal signal.
`As can be seen, operating at a turn-down signal of 25%
`
`Page 3 of 8
`
`
`
`2892
`
`Kelly etal.: Reactive pulsed magnetron sputtering process
`
`2892
`
`TABLE I. Experimental Taguchi L9 array for the investigation of alumina
`films.
`
`Run
`No.
`
`Pulse frequency
`~kHz!
`
`Reverse time
`~ms!
`
`Reverse voltage
`~%!
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`
`20
`20
`20
`35
`35
`35
`50
`50
`50
`
`1
`5
`10
`1
`5
`10
`1
`5
`10
`
`10
`15
`20
`15
`20
`10
`20
`10
`15
`
`pulse frequency ~at levels 20, 35, and 50 kHz!, reverse time
`~1, 5, and 10 ms!, and reverse voltage ~10%, 15%, and 20%
`of the nominal sputtering voltage!. This range of frequencies
`was chosen because the Sparc-le V limits the maximum re-
`verse time which can be selected at frequencies greater than
`50 kHz. Higher frequencies were explored in a second array.
`The initial experimental array is summarized in Table I.
`The alumina films were deposited onto precleaned glass
`substrates which were subsequently sectioned for analytical
`purposes. The coating structures were examined by scanning
`electron microscopy ~SEM!, with the thickness of each coat-
`ing being measured from fracture section micrographs.
`Deposition rates were then calculated from these measure-
`ments. The composition of the coatings was determined us-
`ing a JEOL JXA-50A microanalyzer equipped with WDAX.
`A high-purity aluminum standard was used in the analysis,
`with oxygen content being determined by difference. X-ray
`analyses were carried out using a Philips system, operating
`in u–2umode ~Cu Ka radiation!, and the resistivity of the
`coatings was measured using a four-point probe.
`Following this, a second array of experiments was carried
`out. In this case, coatings were deposited over an extended
`range of pulse frequencies, up to 100 kHz. Also, the MDX
`magnetron driver was used as the dc supply to allow com-
`parison with the Pinnacle unit. Deposition runs were re-
`peated under, otherwise, identical conditions, but at different
`levels of duty factor. Care was taken between runs to sputter
`clean the target, such that all runs started with the target in a
`similar condition. Run times were varied to ensure that the
`total pulse-on time was consistent, i.e., the total sputtering
`time was constant. The reverse voltage was fixed at 20% of
`the nominal sputtering voltage. The number of hard arcs dis-
`played by the Sparc-le V was recorded at regular intervals,
`both to monitor the onset of arcing, and to give the total
`cumulative number of arc events for each set of conditions.
`The coating structures and properties were investigated as for
`the preceding array.
`
`III. RESULTS
`The deposition rates and total number of hard arcs re-
`corded during each of the Taguchi array runs are listed in
`Table II. The deposition rates have been normalized to target
`current to give the rate per minute, per A. The maximum
`
`FIG. 3. Hysteresis behavior displayed during reactive sputtering of alumina.
`
`maintains the target between the metallic and poisoned re-
`gimes in a ‘‘partially poisoned’’ mode. This allows stoichio-
`metric Al2O3 films to be deposited at acceptable rates.
`The first stage of this investigation was to deposit a series
`of alumina films under systematically varied conditions us-
`ing the Pinnacle/Sparc-le V combination referred to above.
`For each run, the total number of hard arcs detected by the
`Sparc-le V was recorded. The film properties were then in-
`vestigated, and the effectiveness of the deposition conditions
`at arc suppression was considered. The Taguchi method23
`was used to design this experiment. This method utilizes
`fractional factorial arrays which are designed to optimize the
`amount of information obtained from a limited number of
`experiments, and, as such, it is a very efficient experimental
`technique. The Taguchi L9 array was selected, which allows
`up to four factors to be varied at three levels, although only
`three factors were actually used. The factors chosen were
`
`FIG. 4. Relationship between optical emission ~OEM! signal and target volt-
`age during reactive sputtering of alumina.
`
`J. Vac. Sci. Technol. A, Vol. 18, No. 6, Nov(cid:213)Dec 2000
`
`Page 4 of 8
`
`
`
`2893
`
`Kelly etal.: Reactive pulsed magnetron sputtering process
`
`2893
`
`TABLE II. Taguchi L9 array data table.
`
`Run
`No.
`
`Duty factor,
`%
`
`No. of hard arcs
`recorded
`
`Coating thickness
`~mm!
`
`Normalized dep’n
`rate ~nm/min/A!
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`
`98
`90
`80
`96.5
`82.5
`65
`95
`75
`50
`
`.10 000
`1823
`5
`.10 000
`492
`0
`.10 000
`2754
`0
`
`4.5
`3.0
`1.4
`2.0
`4.0
`0.75
`3.75
`1.4
`1.8
`
`10.0
`9.3
`4.6
`6.5
`8.8
`2.3
`11.6
`4.6
`4.0
`
`number of arcs which can be displayed by the counter on the
`Sparc-le V is 10 000. Where this value was reached before
`the end of a run, a value of .10 000 has been inserted in
`Table II. Also listed in Table II are the duty factors for each
`run, arising from the array settings of pulse frequency and
`reverse time. Statistical analyses were carried out on these
`data using a software package from the American Supplier
`Institute, entitled ANOVA-TM. This package was used to com-
`pute the level averages using deposition rate and number of
`hard arcs as response variables, i.e., to compute the average
`response of each variable at each level of each factor. The
`results of these analyses are shown graphically in Figs. 5 and
`7, respectively. It appears from Fig. 5 that reverse time and
`reverse voltage both have significant, but opposite influ-
`ences, on deposition rate. In the case of reverse time, this is
`simply because, as this factor is increased, so the pulse-off
`time becomes a greater proportion of the total pulse cycle,
`i.e., the duty factor is reduced and sputtering takes place for
`a lesser proportion of each cycle. This is illustrated in Fig. 6,
`which shows the positive correlation between the duty factor
`and normalized deposition rate ~correlation coefficient, r
`
`FIG. 6. Relationship between the duty factor and normalized deposition rate
`for reactive pulsed sputtered alumina films.
`
`50.77!. Thus, it could be argued that, of the variables inves-
`tigated, reverse voltage actually has the most significant in-
`fluence on deposition rate. As reverse voltage is increased
`from 10% to 20%, the level average for the normalized depo-
`sition rate increases from 5.6 to 8.3 nm/min/A, a factor of
`approximately 1.5 times.
`The Taguchi analysis using the total number of hard arcs
`detected as the response variable is shown in Fig. 7. Rather
`surprisingly, pulse frequency and reverse voltage do not ap-
`pear to influence the response variable, whereas the level
`average for reverse time varies from 10 000 to virtually zero
`as this parameter is increased from 1 to 10 ms. Clearly, vary-
`ing the reverse, or pulse-off time can have a very significant
`
`FIG. 5. Taguchi analysis of alumina films, using the normalized deposition
`rate as the response variable.
`
`FIG. 7. Taguchi analysis of alumina films, using the total number of hard
`arcs detected as the response variable.
`
`JVST A - Vacuum, Surfaces, and Films
`
`Page 5 of 8
`
`
`
`2894
`
`Kelly etal.: Reactive pulsed magnetron sputtering process
`
`2894
`
`FIG. 8. Relationship between the duty factor and the number of hard arcs
`detected during the deposition of the Taguchi array alumina films.
`
`effect on the occurrence of arc events. Again, though, vary-
`ing the reverse time has the effect of varying the duty factor.
`Figure 8, therefore, shows the relationship between the duty
`factor and the number of arc events recorded for the Taguchi
`array runs. At duty factors of 95% and higher, greater than
`10 000 hard arc events were recorded, independent of the
`pulse frequency and reverse voltage selected. At lower duty
`factors the number of arc events decreases exponentially un-
`til at 65% and below zero arcs were recorded. At intermedi-
`ate duty factors, arc events were reduced substantially, but
`not eliminated. There is some scatter in these data; at a duty
`factor of 75% the arc count was unexpectedly high compared
`to the counts at 80% and 82.5% duty. Reference to Table I
`reveals that in the former case the reverse voltage was set at
`10% of the nominal sputtering voltage, whereas in the latter
`case it was set to 20%. It may, perhaps, be the case that
`reverse voltage exerts a second-order influence on the occur-
`rence of arcs. This suggestion is merely speculative at this
`stage. Finally, in these analyses the anticipated interaction
`between pulse frequency and reverse time was not observed.
`This may well have been due to the limited range of pulse
`frequencies investigated.
`When the films themselves were examined, very little
`run-to-run variation was observed. By way of example, Fig.
`9 is a SEM micrograph of the fracture section of array coat-
`ing run 1. In this case, as in all other cases, the coatings were
`fully dense and defect free, with glass-like featureless struc-
`tures. Compositional analysis, x-ray diffraction, and four-
`point probe measurements also showed a consistent pattern.
`In all cases, within the accuracy of the equipment, the com-
`positions were found to be stoichiometric Al2O3 . X-ray
`analysis indicated that these coatings were amorphous. This
`would be expected, as their deposition temperatures did not
`exceed 250 °C. Finally, four-point probe measurements con-
`firmed that the coatings were highly insulating. All resistivity
`
`J. Vac. Sci. Technol. A, Vol. 18, No. 6, Nov(cid:213)Dec 2000
`
`FIG. 9. SEM micrograph of the fracture section of an alumina film deposited
`on a glass substrate: Taguchi array run 1.
`
`readings exceeded 20 MV cm, which is the maximum value
`that could be measured by the probe.
`To investigate further the parameters influencing arcing,
`the second array, described earlier, was carried out. Table III
`lists the pulse frequencies, reverse times, and duty factors
`investigated ~the reverse voltage was set to 20% throughout!.
`Also included in Table III is the total number of hard arcs
`displayed by the Sparc-le V at the conclusion of the deposi-
`tion run. The arc count was also monitored at regular inter-
`vals during each run. Figure 10 shows the incidence of arc
`events during a series of runs carried out at 60 kHz pulse
`frequency. In these runs, the reverse times were varied from
`2 to 6 ms, giving duty factors ranging from 88% to 64%. It is
`again clear from Fig. 10 that there is a strong relationship
`between the duty factor and the occurrence of hard arcs. As
`the duty factor is lowered, the incidence of arcing is signifi-
`cantly reduced. Indeed, at 64% duty, hard arc events were
`completely suppressed for the duration of the deposition run.
`At other duty factors there was still an initial arc-free period
`lasting for several minutes. However, in these cases, charge
`accumulation eventually reached the point where breakdown
`occurred. Beyond this point the incidence of arcing increased
`at an exponential rate.
`
`TABLE III. Run conditions and hard arc counts for second alumina array.
`
`Run
`No.
`
`Pulse frequency
`~kHz!
`
`Reverse time
`~ms!
`
`Duty factor
`~%!
`
`Total hard arcs
`detected
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`
`60
`60
`60
`60
`20
`35
`70
`80
`100
`
`6
`4
`3
`2
`5
`5
`4
`2
`2
`
`64
`76
`82
`87.5
`90
`82.5
`72
`84
`80
`
`0
`37
`385
`5784
`1545
`492
`497
`3998
`1041
`
`Page 6 of 8
`
`
`
`2895
`
`Kelly etal.: Reactive pulsed magnetron sputtering process
`
`2895
`
`FIG. 11. Oscilloscope trace of the target voltage wave form capturing an arc
`event during pulsed reactive sputtering of alumina films ~pulse frequency
`5100 kHz and duty580%!.
`
`the range tested pulse frequency alone does not significantly
`influence deposition rate, or the incidence of hard arcs during
`the deposition of alumina films. The point has already been
`made that a greater interaction with other parameters might,
`perhaps, be expected if the range of frequencies was ex-
`tended. In the case of deposition rate, reverse voltage is the
`critical
`factor at any given duty factor.
`It has been
`suggested12 that this may be a result of preferential target
`cleaning arising from the bipolar nature of the target voltage.
`At the end of each pulse-off period the target voltage is re-
`versed. At that instant, ions in the vicinity of the target will
`be accelerated by the normal negative sputtering voltage,
`plus the positive pulse-off voltage. Thus, at the beginning of
`each pulse-on period there will be a flux of ions incident at
`the target with a higher than average energy. Such a flux
`would preferentially sputter clean poisoned regions of the
`
`FIG. 12. SEM micrograph of an alumina film deposited onto a glass sub-
`strate at 80 kHz pulse frequency and 84% duty.
`
`FIG. 10. Influence of the duty factor on the incidence of hard arc events
`during reactive pulsed sputtering of alumina films ~pulse frequency560
`kHz!.
`
`Overall, the arc counts for the second array were gener-
`ally lower than those for the Taguchi array. While there may
`be a number of reasons for this, the second array runs were
`all carried out at a reverse voltage of 20%. This may, again,
`be weak evidence that reverse voltage can influence arc sup-
`pression.
`To confirm that the events recorded by the Sparc-le V unit
`were indeed arcs, and not merely artifacts of the arc-counting
`circuitry, the target voltage wave forms were investigated
`using an oscilloscope. By triggering the oscilloscope on tar-
`get current, it was possible to capture actual arc events. Fig-
`ure 11 shows a typical example. At the onset of the arc event,
`the discharge voltage collapses and the current rises signifi-
`cantly. In this example, it is at least two pulse cycles before
`the discharge is reestablished.
`Coating structures and properties were investigated for
`the second array coatings, as for the initial array. Again, all
`coatings were x-ray amorphous with stoichiometric alumina
`compositions. An example of the structures of these coatings
`is given in Fig. 12, which shows a SEM micrograph of the
`fracture section of the coating deposited at 80 kHz ~duty
`584%!. Interestingly, the high number of arcs recorded dur-
`ing the deposition of each of these coatings does not seem to
`have had a detrimental effect on the structures, which still
`appear fully dense and defect free. Once again though, fur-
`ther analysis of these films is planned.
`
`IV. DISCUSSION
`A number of interesting points have emerged from this
`investigation. The first Taguchi array demonstrated that over
`
`JVST A - Vacuum, Surfaces, and Films
`
`Page 7 of 8
`
`
`
`2896
`
`Kelly etal.: Reactive pulsed magnetron sputtering process
`
`2896
`
`target. Since the sputtering rate from a metallic target is
`higher than the rate from a poisoned target, this would have
`the effect of raising the deposition rate. Clearly, the effec-
`tiveness of the target cleaning would be increased as the
`magnitude of the reverse voltage is increased, giving rise to
`the trend observed here. Further studies, including the use of
`a time-resolved Langmuir probe, are planned to investigate
`this in more detail.
`Both arrays have demonstrated the very strong depen-
`dence of hard arc events on the duty factor selected. It ap-
`pears, therefore, that it would be more appropriate to con-
`sider a critical duty factor for arc-free operation, rather than
`a critical frequency ~accepting, again, the limited range of
`frequencies tested!. From these experiments, a duty factor of
`70% or lower is necessary, independent of pulse frequency,
`if arc suppression throughout the duration of a deposition run
`is the prime concern. The second array did show that limited
`periods of arc-free operation can be achieved at higher duty
`factors. This finding is in agreement with Belkind, Freilich,
`and Scholl,9,10 who also obtained arc-free reactive sputtering
`of alumina for short time periods at duties greater than 90%.
`However, this study indicated that such conditions do not
`remain arc free and that breakdown soon occurs. Once this
`has happened, the incidence of arcing then increases at an
`exponential rate. The scatter observed in the data presented
`here probably reflects the difficulty in replicating target con-
`ditions at the beginning of each run. The target condition is
`certainly an important, but currently unquantified, factor. Fi-
`nally, on this subject, and underlining the comments made
`about the target condition, there may also be some evidence
`to suggest that increasing the reverse voltage can be benefi-
`cial in reducing arcing. In a manner analogous to the influ-
`ence of reverse voltage on deposition rate, the mechanism for
`this may again be preferential cleaning of the poisoned re-
`gions of the target at the beginning of each pulse-on period.
`This is somewhat speculative at this stage, and any actual
`effect is very much second order, compared to the duty fac-
`tor.
`The other surprising point to come out of this work is the
`apparent insensitivity of the coating structures and properties
`to the incidence of arcing. The alumina films showed a great
`deal of similarity at the relatively superficial level of exami-
`nation used here. All coatings were x-ray amorphous with
`stoichiometric Al2O3 compositions. All structures were fully
`dense and defect free. More sophisticated analysis of these
`films is planned for the future, including nanohardness mea-
`surements and surface roughness measurements.
`To summarize the findings of this work, high-quality alu-
`mina films can be deposited by pulsed reactive magnetron
`sputtering over a broad range of conditions. No significant
`differences in performance were observed between the two
`dc magnetron drivers used. The optimum conditions to
`achieve hard arc-free operation throughout the course of a
`deposition run, using the power delivery systems and depo-
`sition conditions employed here, and for pulse frequencies in
`the range 20–100 kHz, are to select a duty factor of 70%,
`with the reverse voltage set to 20%.
`
`J. Vac. Sci. Technol. A, Vol. 18, No. 6, Nov(cid:213)Dec 2000
`
`V. CONCLUSIONS
`
`High-quality defect-free alumina films have been depos-
`ited by pulsed reactive magnetron sputtering over a broad
`range of conditions. A systematic study of the deposition
`conditions demonstrated that the incidence of hard arcs is
`largely controlled by the duty factor selected, and is indepen-
`dent of pulse frequency ~over the range tested!. It is more
`appropriate, therefore, to consider the concept of a critical
`duty factor for arc-free operation, rather than a critical fre-
`quency. This study indicates that for the deposition of alu-
`mina films a duty factor of 70% or lower is necessary for
`medium-term ~i.e., several hours! arc-free operation. The
`deposition rate also appeared to be independent of pulse fre-
`quency, but to increase with reverse voltage at any given
`duty factor.
`
`1M. Scherer, J. Schmitt, R. Latz, and M. Schanz, J. Vac. Sci. Technol. A
`10, 1772 ~1992!.
`2P. Frach, U. Heisig, C. Gottfried, and H. Walde, Surf. Coat. Technol. 59,
`177 ~1993!.
`3S. Schiller, K. Goedicke, J. Reschke, V. Kirkhoff, S. Schneider, and F.
`Milde, Surf. Coat. Technol. 61, 331 ~1993!.
`4D. A. Glocker, J. Vac. Sci. Technol. A 11, 2989 ~1993!.
`5B. Stauder, F. Perry, and C. Frantz, Surf. Coat. Technol. 74-75, 320
`~1995!.
`6W. D. Sproul, M. E. Graham, M. S. Wong, S. Lopez, D. Li, and R. A.
`Scholl, J. Vac. Sci. Technol. A 13, 1188 ~1995!.
`7P. J. Kelly, O. A. Abu-Zeid, R. D. Arnell, and J. Tong, Surf. Coat.
`Technol. 86-87, 28~1996!.
`8P. J. Kelly and R. D. Arnell, J. Vac. Sci. Technol. A 17, 945~1999!.
`9A. Belkind, A. Freilich, and R. A. Scholl, Surf. Coat. Technol. 108-109,
`558 ~1998!.
`10A. Belkind, A. Freilich, and R. A. Scholl, J. Vac. Sci. Technol. A 17,
`1934 ~1999!.
`11W. D. Sproul, Vacuum 51, 641 ~1998!.
`12J. C. Sellers, Surf. Coat. Technol. 98, 1245~1998!.
`13P. J. Kelly and R. D. Arnell, Vacuum 56, 159 ~2000!.
`14G. Roche and L. Mahoney, Vacuum Solutions 12, 11~1999!.
`15M. S. Wong, W. J. Chia, P. Yashar, J. M. Schneider, W. D. Sproul, and S.
`A. Barnett, Surf. Coat. Technol. 86-87, 381 ~1996!.
`16K. Koski, J. Holsa, and P. Juliet, Surf. Coat. Technol. 116-119, 716
`~1999!.
`17K. Koski, J. Holsa, and P. Juliet, Surf. Coat. Technol. 120-121, 303
`~1999!.
`18G. Brauer, J. Szczyrbowski, G. Teschner, Surf. Coat. Technol. 94-95, 658
`~1997!.
`19G. Brauer, M. Ruske, J. Szczyrbowski, G. Teschner, and A. Zmelty,
`Vacuum 51, 655 ~1998!.
`20J. M. Schneider and W. D. Sproul, in Handbook of Thin Film Process
`Technology: 98/1 Reactive Sputtering, edited by W. D. Westwood ~IOP,
`Bristol, 1998!.
`21D. Carter, G. McDonough, L. Mahoney, G. A. Roche, and H. Walde,
`AVS 46th International Symposium, Seattle, Washington, 25–29 October
`~1999! ~unpublished!.
`22P. J. Kelly, P. S. Henderson, R. D. Arnell, G. A. Roche, and D. Carter,
`AVS 46th International Symposium, Seattle, Washington, 25–29 October
`~1999! ~unpublished!.
`23R. Roy, A Primer on the Taguchi Method ~Van Nostrand Reinhold, New
`York, 1990!.
`
`Page 8 of 8
`
`