`
`Surface and CoatingsTechnology 86-87 (1996) 28-32
`
`SURFACE
`&GOATIN6S
`IFGHKOLOGY
`
`The deposition of aluminium oxide coatings by reactive unbalanced
`magnetron sputtering
`PJ. Kelly a,*, a.A. Abu-Zeid b, R.D. Arnell a, 1. Tong C
`a Research Institute for Design, Manufacture and Marketing, Uniuersity ofSalford, Salford M5 4WT, UK
`b Department ofMechanical Engineering, University ofthe United Arab Emirates, Al-Ain, PO Box 17555, United Arab Emirates
`C School ofAgricultural Machinery, Jilin University of Technology, Changchun 130025, PR China
`
`Abstract
`
`The problems associated with the reactive d.c. sputtering of highly insulating materials, such as alumina, are welI documented.
`Deposition rates are low and an insulating layer can build up on the surface of the target, causing arcing. Arc events prevent
`stable operation and can result in droplets of material being ejected from the target. Such droplets can cause defects in the growing
`film. However, studies have shown that the formation of arcs can be significantly reduced if the magnetron discharge is pulsed at
`a frequency in the 10-200 kHz range. In this investigation, AIOx (where 0.7;$; x:::;1.5) coatings were deposited by reactive
`unbalanced magnetron sputtering using either a d.c. power supply in series with a fixed 20 kHz pulse unit, or a variable frequency
`supply with a maximum frequency of 33 kHz (for comparison purposes, coatings were also deposited by reactive d.c. sputtering,
`without pulsing the discharge). Deposition parameters were varied systematically to produce a range of coating compositions and
`properties. The resulting coatings ranged from extremely dense, stoichiometric Al203 films, with Knoop microhardness values
`>2500 kg mm r
`' , to very soft «100 kg mm V) columnar, sub-stoichiometric films. Deposition rates varied from 4 to 20 um h"".
`Some initial results of wear tests carried out on these coatings are also reported. The pulsed power supplies were found to be very
`stable in operation, with very few arc events being observed.
`
`Keywords: Reactive unbalanced magnetron sputtering; Aluiminium oxide coatings; Pulsed magnetron sputtering process
`
`1. Introduction
`
`The technique of closed-field unbalanced magnetron
`sputtering (CFUBMS) has become established as a
`versatile, commercially viable method of depositing high
`quality metal, alloy and multi-layer coatings onto com(cid:173)
`plex components [1]. It has also successfully been used
`to deposit a wide range of ceramic coatings, including
`titanium nitride, alloy nitrides and diamond-like carbon,
`by reactive sputtering from metallic targets [2-4].
`CFUBMS is generally considered to be a high rate
`deposition process. Metallic coatings can be deposited
`at rates in the microns per minute range [5]. However,
`when operating in the reactive sputtering mode, depos(cid:173)
`ition rates are relatively low, and can be in the microns
`per hour range [2].
`Arc discharges at the target are another problem that
`can occur during reactive sputtering, particularly during
`
`* Corresponding author. Tel.: 44 161 745 5000, ext. 4009; fax: 44 161
`7455108; e-mail: p.kelly@aeromech.salford.ac.uk
`
`0257-8972/96/$15.00 © 1996Elsevier Science SA All rights reserved
`PII S0257 -8972 (96) 02997·0
`
`the deposition of highly insulating materials, such as
`alumina. As the coating process proceeds, areas of the
`targets away from the main racetrack become covered
`with reaction products, as do the target earth shields.
`This can lead to arc discharges on the target. Droplets
`of material can be ejected from the target and cause
`defects in the coating. Also, the damaged area on the
`target can become a source of further arc discharges.
`This results in an increasing frequency of arcing, which
`prevents stable operation. The reactive sputtering pro(cid:173)
`cess is controlled by a feedback loop. Instabilities caused
`by arcing can cause fluctuations in the coating parame(cid:173)
`ters which, in turn, can effect the stoichiometry of the
`resulting film [6].
`These problems limit some of the applications of the
`CFUBMS system. Low deposition rates are commer(cid:173)
`cially unattractive. The presence of defects in films can
`be unacceptable in many optical and microelectronic
`applications, and can effect the performance of a film as
`a corrosion, or
`thermal barrier. Also, variations in
`stoichiometry can giverise to anisotropic film properties.
`
`Page 1 of 5
`
`APPLIED MATERIALS EXHIBIT 1048
`
`
`
`P.J. Kelly et al.ISurfaceand Coatings Technology 86 -87 ( 1996) 28-32
`
`29
`
`RJ. sputtering is generally considered too slow and
`complex a process for large scale commercial applica(cid:173)
`tions [7,8].
`the pulsed magnetron sputtering process
`However,
`(PMS) offers the potential to overcome the problems
`encountered when operating in the reactive sputtering
`mode with the CFUBMS system. Initial studies have
`indicated that pulsing the magnetron discharge at
`medium frequencies
`(10-200 kHz), when depositing
`highly insulating materials, can significantly reduce the
`formation of arcs and, consequently reduce the number
`of defects in the resulting film [6-11 J. For example,
`Schiller [9,10J found that during the reactive sputtering
`of A12 0 3 , raising the pulse rate from 10 to 50 kHz
`reduced the defect density of the coating by several
`orders of magnitude. Furthermore, deposition rates of
`4-5 nm S-1 were achieved, which compares with less
`that 1 nm S-1 for RF sputtering of A120 3. This rate
`amounted to about 60% of the rate achieved by Schiller
`the non-reactive sputtering of pure aluminium.
`for
`Pulsing was also found to stabilise the discharge. This
`allowed Frach [7]
`to deposit virtually defect-free
`Al20 3 coatings up to 50 urn thick.
`If a single magnetron discharge is pulsed,
`then the
`system is described as unipolar pulsed sputtering. In this
`situation, the pulse-on time is limited so that the charging
`of the insulating layers does reach the point where
`breakdown and, therefore, arcing occurs. The discharge
`is dissipated during the pulse-off time through the
`plasma. If two magnetrons are connected to the same
`pulse supply then the configuration is described as
`bipolar pulsed sputtering. Each magnetron source then
`alternately acts as an anode and a cathode of a discharge.
`The periodic pole changing promotes discharge of the
`insulating layers, hence preventing arcing.
`The high rate deposition of defect-free ceramic coat(cid:173)
`ings onto complex components would be a commercially
`attractive process. In view of this, the PMS process is
`being increasingly studied. This paper, therefore, reports
`on work carried out at Salford University to investigate
`the deposition of alumina coatings in a closed-field
`unbalanced magnetron system, utilising the PMS
`process.
`
`2. Experimental
`
`Alumina coatings were deposited by reactive magnet(cid:173)
`ron sputtering in a Teer Coatings UDP 450 rig. The rig
`is equipped with two 300 x 100 mm vertically opposed
`unbalanced magnetrons installed in a closed-field con(cid:173)
`figuration. The aluminium sputter targets were 99.5%
`pure and were also obtained from Teer Coatings.
`Coatings were deposited onto silicon wafers, polished
`aluminium SEM pin stubs and ground stainless steel
`coupons.
`The reactive sputtering process was controlled using
`
`spectral line monitoring [12-14]. The optical emission
`monitor (OEM) was tuned to the 396nm line in the
`aluminium emission spectrum. The target current was
`ramped up and pure Al films were deposited for 2 min.
`The OEM signal at this point was taken as the "100%"
`metal signal. The reactive gas was then allowed into the
`chamber until the OEM signal fell to a pre-determined
`proportion of the initial 100% metal signal. The value
`of the "turn-down" signal was maintained by the feed(cid:173)
`back loop throughout the remainder of the deposition
`run. After each reactive deposition,
`the targets were
`the OEM signal returned to its
`sputter cleaned until
`initial value.
`The coatings were deposited using, either a d.c. power
`supply in series with a fixed 20 kHz pulse unit, or a
`supply with a variable frequency in the range 0.05 Hz
`to 33 kHz (for comparison purposes, coatings were also
`deposited by reactive d.c. sputtering, without pulsing the
`discharge). The 20 kHz unit was an Advanced Energy
`SPARC-LE unit, which was connected in series with the
`existing Advanced Energy MDX magnetron driver [8J.
`The magnitude of the positive pulse is fixed at about
`10% of the magnitude of the negative pulse. The variable
`frequency supply was a Magtron unit [15]. This unit is
`more sophisticated than the SPARC-LE. It can be used
`in unipolar or bipolar mode and the pulse-on and pulse(cid:173)
`offtimes can be varied independently. Only the unipolar
`configuration was used in this investigation.
`All
`the coatings were deposited at a substrate(cid:173)
`to-target separation of 110 mm and a pressure of
`1.25 mTorr. Target current, substrate bias and OEM
`turn-down signal were varied. The run conditions are
`summarised in Table 1.The coatings deposited on silicon
`wafers were fractured to allow the coating structure to
`be examined in the SEM. Thickness measurements were
`taken from SEM micrographs of the fracture sections.
`Knoop microhardness measurements were also made of
`these coatings. Measurements were taken at three appro(cid:173)
`priate loads. The results were then extrapolated to zero
`load to remove any influence of the substrate material
`on the apparent hardness of the coatings. The composi(cid:173)
`tion of the coatings, deposited on polished pin stubs,
`was determined using a JEOL JXA-50A microanalyser,
`equipped with WDAX. The accuracy of this machine,
`as quoted by the manufacturer, is between 1 and 5 at.%.
`Pinyon-disc wear tests were carried out on selected
`specimens, deposited onto ground stainless steel cou(cid:173)
`pons. The "pin" was a 6.35 mm diameter hardened steel
`ball. The normal load was 3 N; the sliding speed 4.4
`m min -1; and the sliding distance 61.6 m. Profilometry
`tests and SEM examination were carried out on the
`specimens after testing. Some initial results are presented.
`
`3. Results
`
`The deposition rates, microhardnesses and composi(cid:173)
`tions of the coatings are listed in Table 1. As expected,
`
`Page 2 of 5
`
`
`
`30
`
`P.J. Kelly et al.fSurface and Coatings Technology 86-87 ( 1996) 28-32
`
`Table 1
`Run conditions and properties of aluminium oxide coatings depositedby d.o. and pulsed magnetron sputtering
`
`Run
`no.
`
`Target
`current (A)
`
`Substrate
`bias (V)
`
`Tum down
`signal (%)
`
`Thickness
`(urn]
`
`Dep. rate
`(urn min -1)
`
`Hk
`(kg mm t
`
`' )
`
`1
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`
`6
`6
`6
`6
`6
`6
`8
`6
`6
`6
`3
`3
`3
`
`-50 rf
`-50 rf
`-50 rf
`-30 rf
`-30 rf
`-100 rf
`-50 rf
`self-bias (-19)
`-50 de
`-50 de
`-100 de
`-50 de
`-30 de
`
`60
`50
`50
`30
`30
`15
`15
`25
`20
`20
`15
`20
`15
`
`13.9
`7.7
`5.3
`3.1
`2.0
`3.3
`3.8
`40.0
`13.0
`10.0
`4.9
`8.1
`5.0
`
`0.52
`0.28
`0.19
`0.11
`0.07
`0.07
`0.07
`0.31
`0.13
`0.18
`0.06
`0.07
`0.06
`
`90
`210
`320
`2650
`1240
`1180
`2480
`270
`1940
`1710
`1020
`1510
`2010
`
`AI/O
`(at.%)
`
`51/49
`51/49
`58/42
`45/55
`44/56
`37/63
`41/59
`54/46
`42/58
`43/57
`41/59
`40/60
`37/63
`
`Power
`supply
`
`d.c, only
`d.c. only
`d.c.+ SPARC-LE
`d.c.+SPARC-LE
`d.c.+SPARC-LE
`d.c, + SPARC-LE
`d.c.+SPARC-LE
`Magtron 15.4 kHz
`d.c.+SPARC-LE
`Magtron (15.4 kHz)
`Magtron (15.4 kHz)
`Magtron (25 kHz)
`Magtron (25 kHz)
`
`the d.c. reactive sputtering of aluminium oxide coatings
`proved extremely difficult. Arcing took place from the
`target throughout the runs, and the process was highly
`unstable, even at tum-down signals of 60% of the pure
`metal signal, i.e., relatively low levels of target poisoning.
`The structure of one of these coatings (run no. 1) is
`shown in Fig. 1. As can be seen,
`the coating has a
`granular, porous structure. Reference to Table 1 indicates
`a
`sub-stoichiometric
`composition
`and
`very
`low
`microhardness.
`By contrast, when operating with the SPARC-LE
`units the process was very stable, with few arc events at
`the target. This was found to be the case, even at turn(cid:173)
`down signals of 15%, i.e., sputtering from a heavily
`poisoned target. Figs. 2 and 3 show SEM micrographs
`of the fracture sections of coatings 7 and 9, respectively.
`Both coatings are fully dense with no discernible struc(cid:173)
`tural aspects on the fracture surface. Also, both coatings
`remained well adhered to the substrate after fracture.
`The composition of these coatings is very close to
`
`Fig. 2. SEM micrograph of fracture section of aluminium oxide coating
`number 7, deposited by d.c, magnetron sputtering with SPARC-LE
`pulse unit attachment. The substrate is a silicon wafer.
`
`Fig. .I. SEM micrograph of fracture section of aluminium oxide coating
`number 1, deposited by d.c, magnetron sputtering onto silicon wafer.
`
`Fig. 3. SEM micrograph of fracture section of aluminium oxide coating
`number 9, deposited by d.c. magnetron sputtering with SPARC-LE
`pulse unit attachment. The substrate is a silicon wafer.
`
`Page 3 of 5
`
`
`
`r.: Kelly et {ll./Surface and Coatings Technology 86-87 ( 1996) 28-32
`
`31
`
`Table 2
`Results of pin-on-disc tests on selected aluminium oxide coatings
`deposited on stainless steel coupons
`
`Coating number
`
`7
`
`0.25
`0.83
`
`1.02E-3
`
`8
`
`0.1
`0.33
`
`0.11
`
`9
`
`0.26
`0.87
`
`12
`
`0.2
`0.63
`
`13
`
`0.26
`0.87
`
`1.23E·3
`
`Frictional force (N)
`Steady state coef.
`of friction (us)
`Wear volume (mnr')
`
`Fig. 5. SEM micrograph of surface of aluminium oxide coating number
`8, showing part of pin-on-disc wear track.
`
`stoichiometric Al20 3 • Both have high microhardness
`values (2480 and 1940kg mm -2, respectively). However,
`the deposition rate for run 9 was nearly twice that of
`run 7, despite the fact that the target current was lower.
`This, presumably, reflects the greater degree of target
`poisoning (i.e., the lower OEM signal) during run 7.
`It proved difficult to optimize the performance of the
`Magtron unit. When operating at a target current of 6
`A and a pulse frequency of 15.4kHz, arcing occurred at
`the target throughout the run, with the frequency of arcs
`increasing with run time. The supply operated most
`successfully when delivering a target current of 3 A, at
`a frequency of 20 kHz with identical pulse-on and pulse(cid:173)
`off times. Fig.4 shows an SEM micrograph of the
`fracture section of coating 13, deposited using the
`Magtron supply. Again, the coating has a fully dense
`structure and good coating-to-substrate adhesion. The
`microhardness, deposition rate and composition of this
`coating are very similar to coating 7, deposited using
`the SPARC-LE unit. However, coating 7 was deposited
`at a target current of 8 A, whereas, coating 13 was
`deposited at a target current of 3 A. The similarity in
`deposition rates, despite the significant difference in
`target powers between these two coatings cannot be
`explained at
`this stage, particularly as both coatings
`were deposited at the same turn-down signal.
`Pin-on-disc tests were carried out on a number of
`selected specimens, as described earlier. The results of
`these tests are listed in Table 2. Profilometry measure(cid:173)
`ments were made of the wear tracks for coatings 7, 8
`and 13 and, from this, wear volumes were calculated.
`Fig. 5 shows a SEM micrograph of the surface of coating
`8, showing part of the wear track. The wear track for
`coating 9 was within the original surface roughness and,
`therefore, a wear volume could not be calculated for this
`specimen. A section of the wear track on the surface of
`coating 9 is shown in Fig. 6. For coating 12, material
`
`Fig. 6. SEM micrograph of surface of aluminium oxide coating number
`9, showing part of pin-on-disc wear track.
`
`to the coating
`transfer occurred from the steel ball
`surface. SEM examination of the wear track showed
`that only transferred material was present, and no wear
`of the coating was observed. Figs. 7 and 8 are SEM
`micrographs of the wear track region for coating 12, in
`which transferred material can clearly be seen. Fig. 7
`also demonstrates how closely the topography of the
`coating surface matches the topography of the substrate.
`
`Fig. 4. SEM micrograph of fracture section of aluminium oxide coating
`number 13, deposited by pulsed magnetron sputtering onto silicon
`wafer.
`
`Page 4 of 5
`
`
`
`32
`
`P.J. Kelly et al.fSurface and Coatings Technology 86-87 ( 1996) 28-32
`
`pure aluminium films using the SPARC-LE attachment
`and for d.c. sputtering alone under otherwise identical
`conditions.
`Both of the pulse units investigated were stable in
`operation (once optimized), with very few arcs being
`observed at the target. Both allowed control over the
`reactive sputtering process to be established. Thus, the
`composition of the coating and, therefore, its properties,
`could be controlled.
`
`5. Conclusions
`
`This investigation has demonstrated that fully dense
`coatings of alumina can be deposited in a CFUBMS
`system at rates of about 40% of that obtained for the
`d.c. sputtering of aluminium. The pulsed magnetron
`sputtering process is a major development in the reactive
`sputtering field. The prevention of arcs at
`the target
`provides stability to the reactive sputtering process. This,
`in turn, permits the coating composition and properties
`to be controlled. The ability to deposit defect-free oxide
`coatings at high rates offers the potential to improve the
`performance and extend the range of applications of
`these coatings.
`
`References
`
`[IJ R.1. Bates and R.D. Arnell, Surf Coat. Technol., 68/69 (1994)
`686-690.
`[2J S.L. Rohde, W.D. Sproul and J.R. Rohde, J. Vac. Sci. Technol.;
`A 9(3) (1991) 1178-1183.
`[3J D.P. Monaghan, D.G. Teer, K.c. Laing, 1. Efeoglu and R.D.
`Arnell, Surf Coat. Techno/., 59 (1993) 21-25.
`[4J D.P. Monaghan, D.G. Teer, P.A. Logan, 1. Efeoglu and R.D.
`Arnell, Surf Coat. Teclmo/., 60 (1993) 525-530.
`[5J D.P. Monaghan and RD. Arnell, Surf. Coat. Technol.; 49
`(1991) 298-303.
`[6 J D.A. Glocker, J. Vac. Sci. Technoi., A 11(6) (1993) 2989-2993.
`[7J P. Frach, U. Heisig, C. Gottfreid and H. Walde, Surf Coat. Tech(cid:173)
`no/., 59 (1993) 177-182.
`[8J W.D. Sproul, M.E. Graham, M.S. Wong, S. Lopez, D. Li and
`R.A. Scholl, J. Vac. Sci. Technol., A 13 (6) (1995) 1188-1191.
`[9J S. Schiller, K. Goedicke, J. Reschke, V. Kirchkoff, S. Schnieder
`and F. Milde, Surf Coat. Technol., 61 (1993) 331-337.
`[IOJ S. Schiller, K. Goedicke and C. Metzner. Paper presented at the
`Int. Conf. on Metallurgical Coatings and Thin Films (ICMCTFJ,
`April 1994, San Diego.
`[l1J B. Stauder, F. Perry and C. Frantz, Sur]. Coat. Technol., 74/75
`(1995) 320-325.
`[12J R.P. Howson, A.G. Spencer, K. Oka and R.W. Lewin, J. Vac. Sci.
`Technol., A 7(3) (1989) 1230-1234.
`[13J Z. Pang, M. Boumerzoug, R.V. Kruzelecky, P. Mascher, J.G.
`Simmons and D.A. Thompson, J. Vac. Sci. Technol., A 12(1)
`(1994) 83-89.
`[14J S. Inoue, K. Tomianga, R.P. Howson, K. Kusaka, J. Vac. Sci.
`Technol., A 13(6) (1995) 2808-2813.
`[15J Technical Data Sheet, DC-Bipo/ar Pulse Power Supply compact
`version. Magtron, GmbH, Guterslrasse 21, D-77833 Otters(cid:173)
`weier, Germany.
`
`Fig. 7. SEM micrograph of surface of aluminium oxide coating number
`12, showing part of pin-on-disc wear track and topographical detail.
`
`Fig. 8. SEM micrograph of wear track region of aluminium oxide coat(cid:173)
`ing number 12, showing transferred material from steel ball.
`
`Grinding marks on the substrate are perfectly repro(cid:173)
`duced through a coating thickness of (in this case) 8 urn.
`Based on these results described above, the coatings
`were ranked in the following order of increasing wear
`resistance; run 8, run 13, run 7, run 9 and run 12. As
`mentioned earlier,
`these are preliminary results. The
`tribological properties of these coatings will be investi(cid:173)
`gated in more detail in the future.
`
`4. Discussion
`
`This investigation has demonstrated that fully dense,
`stoichiometric alumina coatings can be deposited at
`relatively high rates by closed-field unbalanced magnet(cid:173)
`ron sputtering, provided the magnetron discharge is
`pulsed. Extremely dense coatings with high microhard(cid:173)
`rates of up to 0.13
`ness values were deposited at
`urn min -t. This rate is equivalent to 47.5 and 39.4%,
`respectively, of the rates obtained for the deposition of
`
`Page 5 of 5
`
`