Characterisation studies of the pulsed dual cathode magnetron sputtering process for oxide films

Surface and Coatings Technology 142 144 2001 621 627 Characterisation studies of the pulsed dual cathode magnetron sputtering process for oxide films J. O Brien, P.J. Kelly Centre for Materials Research, Uni ersity of Salford, Salford M5 4WT, UK Abstract Pulsed magnetron sputtering has become the leading industrial production process for large area thin film deposition due to its versatility, low environmental impact and ability to provide uniform coatings across large substrate areas. Such applications commonly employ oxides, including silica Ž SiO., alumina Ž Al O. and titania Ž TiO. 2 2 3 2. Although all of these materials can be produced by reactive DC-magnetron sputtering, until recently the deposition process was highly problematic. The industrial exploitation of the pulsing process is impeded by the fact that, during long-term deposition runs, eventually all surfaces will be covered with the insulator. Once the chamber and anode are covered with insulating material, there can be no average Ž DC. current flowing to the power supply leads. Moreover, if the anode chamber becomes covered with the dielectric film it will cease to operate as an effective ground Ž anode.. A more viable approach is to use two magnetrons in a dual bipolar arrangement Ždual cathode.. In this arrangement each magnetron acts alternatively as an anode and a cathode. The broad aim of this study is to investigate the inter-relationship between global parameters, pulse parameters, plasma parameters and in particular, film properties in relation to the dual cathode system. The spatial distribution of measured parameters will be considered. This paper describes the production of Al2O3 films by dual cathode unbalanced reactive magnetron sputtering, in particular, the effects of spatial distribution and pulse frequency on coating properties. In general, it was observed that, once hard arcs have been removed, all coating structures, coating surfaces and hardness show little variation. However, more variation was observed in critical loads during scratch adhesion testing for coatings deposited at different pulse frequencies. 2001 Elsevier Science B.V. All rights reserved. Keywords: Please supply keywords 1. Introduction High quality functional films on large area substrates are becoming steadily more important and widespread. Applications include low emissivity and solar control coatings on architectural glass, anti-reflective coatings on automobile windscreens and flat-panel displays and transparent films on food packaging 1. In all cases, high rate, stable deposition conditions are required. Pulsed magnetron sputtering has become the leading industrial production process for large area thin film Corresponding author. Tel.: 44-161-745-ext4009; fax: 44-161- 745-5108. deposition, due to its versatility, low environmental impact and ability to provide uniform coatings across large substrate areas Ž up to 4 m in width.. The coating materials used in the applications listed above are most commonly oxides, including silica Ž SiO. 2, alumina Ž Al O. and titania Ž TiO. 2 3 2. Although all of these materials can be produced by reactive DC magnetron sputtering, until recently the deposition process was highly problematic. The use of this technique for the deposition of highly insulating materials is limited by the intrinsic problems of target poisoning, and the consequent arcing and process instabilities 2. Arc events during reactive sputtering are a serious problem, because they can cause defects in the coating structure; affect the composition and properties in the growing 0257-8972 01 $ - see front matter 2001 Elsevier Science B.V. All rights reserved. PII: S 0 2 5 7-8 9 7 2 0 1 01058-1

622 ( ) J. O Brien, P.J. Kelly Surface and Coatings Technology 142 144 2001 621 627 film; and lead to damage of the magnetron power supply 3. The use of pulsed DC power has transformed the deposition of dielectric materials such as alumina. This process has been well described in the literature and will only be reviewed here 1,4. Arc events are suppressed and as a result, very significant improvements in structure and hence, other properties have been observed compared to DC sputtered films 5. In essence, during the pulsed sputtering process the pulseon time is limited, so that the charging of the insulating layers on the target does not reach the point were breakdown and arcing occur. The charge is then dissipated through the plasma during the pulse-off time. The industrial exploitation of the pulsing process is, however, impeded by the fact that, during long-term deposition runs, eventually all surfaces will be covered with the insulator unless new conductive, i.e. metallic, material is brought into the system constantly. Once the chamber and anode are covered with insulating material, there can be no average Ž DC. current flowing to the power supply leads. Moreover, if the anode chamber becomes covered with the dielectric film it will cease to operate as an effective ground Ž anode.. This gradual shifting of the electric field shape can generate serious changes in the deposition pattern of the magnetron and possibly wreck the deposition uniformity 6. A more viable approach is to use two magnetrons in a dual bipolar arrangement. In this case, both magnetrons are connected to the same alternating power supply and the target voltage is reversed during each cycle. Thus, each magnetron acts alternatively as an anode and a cathode. The periodic pole changing prevents arcing, but also effectively maintains clean target surfaces, allowing high rate, long term Ž 300 h., stable deposition conditions 7. This system has already overcome some of the difficulties involved in sputtering dielectric materials and is clearly a process with great potential. However, much development is still required before this system is optimised and the process can be considered routine. The broad aim of this study is to investigate the inter-relationship between global parameters, Žsuch as magnetron configuration and orientation, degree of unbalance and magnetic field strength.; pulse parameters Ž frequency, phase relationship.; plasma parameters Želectron temperature Ž T., ion density Ž n. e i, electron density Ž n.. and in particular, film properties Ž e optical, structural physical. in relation to the dual cathode system. In addition, the use of a separate ion assisted source will also be investigated. In all cases the spatial distribution of measured parameters will be considered. The present study looks at the production of Al 2O3 films by dual cathode unbalanced reactive magnetron sputtering, in particular the effects of spatial distribution and pulse frequency. Operating dual unbalanced magnetrons in a closed-field configuration can result in a high flux of relatively low energy Ž 50 ev. ions being drawn at the substrate 8. It was observed that varying pulse frequency can lead to notable variations in both the time-averaged ion current at the substrate, and in the time-resolved waveform. 2. Experimental For this work a Teer Coatings Ltd. UDP450 vacuum coating rig 9 was used fitted with 2 Gencoa Vtech variable magnetrons in the vertically opposed configuration. These magnetrons allow magnetic field strength and degree of balance to be varied in situ. For this study the Vtech inner and outer magnetic arrays were fixed in the fully forward position giving maximum field strength Ž 0.06 T at the target. and a mid range degree of unbalance for these magnetrons. Power to the targets was supplied via a dual channel Advanced Energy Pinnacle Plus magnetron driver. In the dual cathode mode this unit allows the magnetron discharge to be pulsed at frequencies from 100 to 350 khz, at 0.5 duty factor, at a nominal reverse voltage of 10%. Pulsed voltages are often schematically represented by a square wave format but in the case of the Pinnacle Plus significant voltage overshoots are observed in both directions. This is illustrated by the oscilloscope measurements shown in Fig. 1 and Fig. 2. These figures are voltage waveforms at one of the two targets during sputtering of aluminium. They also show typical substrate current waveforms when the substrate is floating, but attached to an MDX supply. These examples, at 100 and 350 khz, illustrate how the voltage waveforms differ greatly, along with the resulting substrate current waveform, both in magnitude and shape. In concurrent studies by Bradley et al. 10 it has been observed that the electron temperature increases with pulse frequency. This is attributed to the larger overshoot spike at higher pulse frequencies, marked and on Fig. 1 and Fig. 2. The aim of this work is to establish if any marked difference in coating properties accompany these changes due to pulse frequency in the dual cathode mode. In the dual cathode mode target voltages run 180 out of phase with each other, one acting as a cathode whilst the other is an anode and vice versa. The majority of work reported here has been completed using this arrangement. Additional depositions at 200 khz frequency have also been completed in the synchronous mode, were both targets run in phase. Clearly this mode of operation does not have the same advantage of the dual cathode system but is of interest due to the different resulting substrate current forms. The DC films were deposited using the

( ) J. O Brien, P.J. Kelly Surface and Coatings Technology 142 144 2001 621 627 623 Fig. 1. The voltage at one target and substrate current waveforms for metallic deposition 100 khz, 2 A, dual cathode. same Pinnacle Plus power supply with the pulse frequency set to zero. Alumina films were deposited by reactive sputtering in an Ar O2 atmosphere, using an optical emissions monitoring control system. Conditions for stoichiometric Al O and high rate deposition were determined 2 3 from previously compiled hysteresis curves for the system utilised 11. Target current was fixed at 4 A for all runs and coating pressure was fixed at 1.25 mtorr partial pressure of Ar. Substrates were rotated between targets at a target-to-substrate separation of 110 mm and samples were positioned at several positions along Fig. 2. The voltage at one target and substrate current waveforms for metallic deposition 350 khz, 2 A, dual cathode.

624 ( ) J. O Brien, P.J. Kelly Surface and Coatings Technology 142 144 2001 621 627 Fig. 3. The voltage waveforms at both targets in dual cathode mode, with resulting substrate current 150 khz, 4 A. the length of the substrate holder. The main substrate used was glass; run time was selected to give a coating thickness of approximately 2 m. A full experimental array was completed, including repeats, looking at the effect of pulse frequency on coating properties. Pulse frequencies of 100 350 khz were investigated in the dual cathode mode. Coating structure was investigated using SEM and surface roughness was measured by profilometry Ž Talysurf 10.. XRD Philips PW1729 X-ray generator Ž Cu K. was used to assess if the structure was amorphous or crystalline and EPMA ŽJEOL JXA 50A equipped with WDAX. determined the composition of the films. Hardness was measured by a Fischerscope microhardness tester fitted with a Vickers indentor. A load of 50 mn was used and hardness values were determined at 10% of the coating thickness. Single pass scratch testing was completed using a Teer Coatings Ltd. ST3001 Tribotester fitted with an acoustic emission module. 3. Results All pulsed depositions ran stabily with no hard arcs detected, the reactive sputtering system also gave very stable operation. Conversely, during DC deposition arcing occurred almost continuously, which caused some degree of instability with the reactive system. Deposition rates during reactive pulsing was found to be in the region of 65% of that for the DC reactive process and 43% of that for the metal under the same pulsed conditions. Fig. 3 and Fig. 4 illustrate the voltage waveforms at 150 khz pulse frequency at each target, for dual cathode and synchronous mode respectively, also illustrated are the substrate currents drawn. 1 It can be seen that the substrate current waveforms differ greatly, most predominantly in that the current waveform in the synchronous mode features a large current peak during every cycle. The ion current drawn at the substrate was measured using the bias supply, an advanced energy MDX unit, was found to be between 70 and 90% of that achieved running in DC mode during dual cathode pulsed operation. This is despite the fact that the duty factor is only 0.5. The actual value was dependent on pulse frequency. It was also observed that ion-to-atom ratio varied as pulse frequency increased. We estimate there are approximately 1.6 times more ions per atom at 200 khz, where the ion-to-atom ratio is at maximum, than during DC deposition. This is due both to a slight drop in deposition rate and also to an increase in the average ion current density from 100 to 350 khz pulse frequency. In the synchronous pulse mode the average ion current drawn at 0.5 duty factor is approximately 40 60% of the DC value, again dependent on pulse 1 At 100 V bias

( ) J. O Brien, P.J. Kelly Surface and Coatings Technology 142 144 2001 621 627 625 Fig. 4. The voltage waveforms at both targets in synchronous mode, with resulting substrate current 150 khz, 4A. frequency and has an additional ion current spike Ž refer to Fig. 4.. Structure did not vary over the substrate area. All coatings produced by the dual cathode and synchronous mode pulsed processes were found to have fully dense structures with no defects. Furthermore, the coatings were found to be amorphous and highly transparent, although no optical properties have been determined as yet. No inclusions were observed upon investigation with SEM. The glass-like structure of coatings produced during pulsed deposition is well illustrated by Fig. 5; all pulsed coatings show a similar structure. The DC coatings were again dense but contained a very high number of inclusions and splats Ž one such feature is marked A on Fig. 6b., DC coatings tended to be smoked in appearance and therefore, would have a reduced transparency. Figs. 6a and b show a typical DC coating structure. Fig. 7 shows profilometry traces at a 10 000 magnification. It can be seen that the surfaces of pulsed samples are very smooth. No real variation with pulsing frequency or substrate position was noted. Coatings produced by DC deposition exhibit a much increased roughness. Pulsed coatings were found to have a surface texture parameter, Ra 2, value of 0.004 0.008 m, whereas DC coatings gave an average Ra value of 0.25 0.3 m. Composition also varied a little with pulse frequency or spatial position. All coatings were stoichiometric Al 2O3 and all contain levels of included Argon. The rougher surfaces of DC coatings made an accurate measurement of composition difficult. The Ar could be included in the lattice in the place of Al or be held in small pockets between the structure 12. The DC coatings were found to give lower Al readings Ž 34%.. This could be due to one or a combination of four 2 The arithmetic mean of the departures of the roughness profile from the mean line. Fig. 5. The typical glass-like structure of Al O Ž 150 khz, 4 A.. 2 3 pulsed DC coatings

626 ( ) J. O Brien, P.J. Kelly Surface and Coatings Technology 142 144 2001 621 627 150 khz pulse frequency, to 3.5 N for 350 khz pulse frequency coatings. It should be noted, however, that to date scratch tests have not been completed on 100 and 200 khz specimens and further work is in progress to reaffirm results obtained. DC specimens failed at a critical load between 50 and 60% of that for the best pulsed specimens tested so far. 4. Discussion Fig. 6. Ž. a Fracture section of DC Al O coating showing defects. Ž. 2 3 b Top surface of DC Al2O3 coating, featuring splats marked A. reasons: Ž. a an actual lower percentage of Al; Ž. b increased inaccuracy due to scattering from the much rougher top surface; Ž. c an increased degree of Ar inclusion; and Ž d. an increased number of voids in the structure. Further studies will be carried out to try and ascertain the main factor effecting the DC EPMA reading; however, it is most likely that the low DC values are due to both scattering from the surface, giving greater inaccuracy and greater Ar content. Extensive Vickers microhardness tests revealed no variation in hardness over the substrate area, any variations were well within the error of the technique. In addition, no real trend with pulse frequency was established, although a slight peak in hardness at 100 khz was noted. Using Vickers hardness the hardness of pulsed coatings varied from 12.5 to 13.75 GPa Ž 1 Gpa. with a value of approximately 13.4 GPa for DC coatings. Initial scratch tests have been completed on a number of samples. The critical load for coating failure was found to decrease from 25 N, for coatings produced at Results would seem to imply so far that, under the conditions investigated, pulsing in the dual cathode or synchronous mode produces highly transparent, defect free, stoichiometric coatings that are produced by a very stable process. Results to date show no variation along the substrate length in structure, composition, thickness, Ar inclusion or hardness. In addition, no real major trends in the same properties with pulse frequency or mode of pulse can be discerned despite the differences in pulsing signal leading to variation in plasma properties and the energy flux of particles incident at the substrate. Trends have been established between pulse frequency and scratch test critical loads. A general decrease is observed with increasing pulse frequency although all values have not been collected to date. The decrease in the critical load with frequency may be due to either decreased adhesion between the coating and the substrate or increased stresses in the coatings. Extensive optical property measurements and further scratch and wear tests may reveal more variations between the coatings and these will be investigated in the next portion of the work. 5. Conclusions Highly transparent, defect-free alumina films have been grown in the dual cathode and synchronous pulsed sputtering modes. In both modes the deposition process is stable, at least in the short term Ž 10 h.. The DC process, under the same conditions, leads to hard arcs and a less stable process. Coatings produced using this process have a high defect density. Little variation in coating structure, surface roughness or coating hardness was observed with pulse frequency or mode of pulse. A decrease in critical load, as determined with scratch tests, was found with increasing pulse frequency. Optimal pulse coatings showed up to a two-fold improvement on critical load over DC coatings.

( ) J. O Brien, P.J. Kelly Surface and Coatings Technology 142 144 2001 621 627 627 Fig. 7. The surface roughness profile of typical DC and typical pulsed coating surface. References 1 G. Brauer, J. Szczyrbowski, G. Teschner, Surf. Coat. Tech. 94-95 Ž 1997. 658 662. 2 S. Schiller, K. Goedicke, J. Rescke, V. Kirchoff, S. Schneider, F. Milde, Surf. Coat. Tech. 61 Ž 1993. 331 337. 3 W.D. Sproul, M.E. Graham, M.S. Wong, S. Lopez, D. Li, R.A. Scholl, J. Vac. Sci. Technol. A13 Ž.Ž 3 1995. 1188 1191. 4 A. Belkind, A. Freilich, R. Scholl, J. Vac. Sci. Technol. A17 Ž. 4, Jul Aug 1999. 5 P.J. Kelly, O.A. Abu-Zeid, R.D. Arnell, J. Tang, Surf. Coat Tech. 86 87 Ž 1996. 28 32. 6 R.A. Scholl, Surf. Coat Tech. 98 Ž 1998. 823 827. 7 J.C. Sellers, Surf. Coat. Tech. 98 Ž 1998. 1245 1250. 8 P.J. Kelly, R.D. Arnell, JVST A16 Ž.Ž 5 1998. 2858 2869. 9 P.J. Kelly, R.D. Arnell, Surf. Coat. Technol. 108-109 Ž 1998. 317 322. 10 J.W. Bradley, H. Baeker, P.J Kelly, R.D. Arnell, accepted for publication in Surf. Coat. Tech. 11 P.J. Kelly, R.D. Arnell, JVST A17 Ž.Ž 3 1999. 945 953. 12 P.J. Kelly, PhD Thesis, University of Salford, UK, 1997.