J. Phys. D: Appl. Phys. 33 (2000) R173–R186. Printed in the UK
`
`PII: S0022-3727(00)64098-6
`
`REVIEW ARTICLE
`Recent developments in plasma
`assisted physical vapour deposition
`
`Jochen M Schneider†, Suzanne Rohde‡, William D Sproul§ and
`Allan Matthews(cid:2)
`
`† Thin Film Physics Division, Department of Physics, Link¨oping University, S-58183,
`Sweden
`‡ Department of Mechanical Engineering, University of Nebraska-Lincoln, 255 WSEC,
`Lincoln, NE 68588-0656, USA
`§ Reactive Sputtering, Inc., Santa Barbara, CA 93111, USA
`(cid:2) The Research Centre in Surface Engineering, Hull University, Hull, HU6 7RX, UK
`
`Received 26 April 2000
`
`Abstract. Recent developments in plasma assisted physical vapour deposition (PAPVD)
`processes are reviewed. A short section on milestones in advances in PAPVD covering the
`time period from 1938 when the first PAPVD system was patented to the end of the 1980s is
`followed by a more detailed discussion of some more recent advances, most of which have
`been related to increases in plasma density. It has been demonstrated that the state of the art
`−3. In this range a
`PAPVD processes operate in a plasma density range of 1011 to 1013 cm
`substantial fraction of the plasma consists of ionized film forming species. Hence, the energy
`of the condensing film forming species can be directly controlled, as opposed to utilizing
`indirect energy control with, for example, ionized inert gas bombardment. For a large variety
`of applications ranging from ceramic film synthesis at conditions far from thermodynamic
`equilibrium to state of the art metallization technology, such direct energy control of the
`condensing film forming species is of critical importance, and offers the possibility to
`engineer the coating microstructure and hence the coating properties.
`
`1. Introduction
`
`Plasma assisted physical vapour deposition (PAPVD)
`involves the condensation of vapour created from a solid
`source, in the presence of a glow discharge or plasma. Typical
`PAPVD processes are evaporative ion plating,
`reactive
`sputtering and some plasma/ion beam based and/or assisted
`deposition techniques. Historically,
`the first sputtering
`experiments were reported by Grove in 1852 [1], early reports
`on evaporation were made by Faraday in 1857 [2], and the
`first arc deposition was patented by Edison in 1892 [3].
`State of the art PAPVD processes allow the deposition
`of metals, alloys, ceramic and polymer thin films onto a wide
`range of substrate materials [4], and as such the use of PAPVD
`in science and technology has increased dramatically in the
`last two decades. New fields of applications have emerged,
`and the requirements on the synthesis techniques has been
`greatly increased. PAPVD techniques have been described
`in numerous books [5–8] as well as established international
`conferences [9].
`PAPVD processes can be classified according to their
`degree of ionization. Magnetron sputtering and ‘non-
`enhanced’ evaporation typically employ a weakly ionized
`plasma with the ionized fraction of (cid:3)10%, and more
`typically 0.01%. On the other hand, processes such as
`‘enhanced’ evaporation, cathodic arc deposition,
`ionized
`
`0022-3727/00/180173+14$30.00 © 2000 IOP Publishing Ltd
`
`magnetron sputtering and some electron cyclotron resonance
`(ECR) plasmas can have ionized fractions of >10% and
`are reported to approach 100% in some cases [10]. At the
`other extreme, high-pressure (atmospheric) plasma processes
`usually operate near thermal equilibrium, where the electron
`temperature, ion temperature and the temperature of the
`processing gas are approximately equal (in the range of 0.1 eV
`to 2 eV). These types of discharges are mainly used as heat
`sources and are not discussed in this paper.
`PAPVD processes on the other hand are low-pressure
`discharges and are not in thermal equilibrium. These plasmas
`are characterized by hot electrons (electron temperature of
`several eV) and ‘cold’ ion and gas species, with typically
`two orders of magnitude lower temperatures. ‘Hot’ electrons
`collide with ‘cold’ neutrals and create excited and ionic
`species which are highly reactive. Chemical reactions are
`thus enabled at significantly lower temperatures than those in
`equilibrium processes. Such film growth cannot be described
`by equilibrium thermodynamics, and as a consequence, the
`formation of metastable phases is often observed. The ability
`to synthesize materials far from thermodynamic equilibrium
`is attractive for materials research and materials processing
`of new materials. High plasma density PAPVD processes are
`characterized by a large fraction of ionized species (metal and
`non-metal). The ion energy at the substrate is given by the
`difference of the plasma potential to the substrate potential
`
`R173
`
`GILLETTE 1010
`
`

`
`J M Schneider et al
`
`(floating or biased) plus the initial kinetic energy. Hence,
`by controlling the substrate bias potential, the energy of the
`condensing ionic species, which actually form the films, can
`be directly controlled.
`Let us begin by briefly reviewing the major milestones
`in the development of PAPVD. The original PAPVD system
`was patented by Berghaus in 1938 [11], but it was not until
`the 1960s that the potential of the process was recognized
`when Mattox coined the term ‘ion plating’ and led research
`into ion plating of metallic films [12]. Likewise, PAPVD of
`ceramic thin films was pioneered by Bunshah and Raghuram
`[13] and Raghuram and Bunshah [14].
`It was soon found
`that it was necessary to directly bias the substrate [15–17] as
`well as to enhance the level of ionization in order to obtain
`dense ceramic coatings with highly desirable properties
`[16, 18, 19]. By the early 1980s evaporation systems were
`in widespread use for the production of ceramic coatings,
`especially for cutting tools, most notably by the Balzers
`company. At that time it was clear that the sputtering systems
`available to coat tools, could not deliver the same quality of
`coatings, due to the much lower levels of ionization being
`achieved [20].
`The situation has changed remarkably since the mid to
`end of the 1980s with the invention of both the unbalanced
`and closed field unbalanced magnetron configurations. These
`systems achieve ionization levels equal to and even greater
`than, those achieved using ‘enhanced’ evaporative systems.
`The subsequent developments in PAPVD will be
`reviewed in more detail in the present article. We have
`attempted to compile the recent, and for us, exciting
`developments in PAPVD. The present article is not intended
`as a ‘stand alone’ comprehensive review of recent advances
`in PAPVD around the world. It is simply not possible to cover
`all recent developments in PAPVD in a single journal article.
`We do hope however, that the reader finds that our selection
`of the recent advances stimulates their interest in the progress
`of this field. Section 2 of this review deals with advances in
`magnetron sputtering technology. Thin film growth utilizing
`massive Ar ion fluxes of low energy is discussed, as well as
`process control issues for the high rate reactive deposition of
`dielectric thin films. In section 3 the recent developments in
`high plasma density PVD are reviewed. Particular attention
`is given to ionized magnetron sputtering, self sputtering and
`high-power pulsed sputtering, as well as filtered cathodic arc
`techniques and plasma immersion ion implantation.
`
`2. Magnetron sputtering technology
`
`2.1. Introduction
`
`The advent of magnetron sputtering technology has attracted
`the interest of both research and industry, mainly due to the
`significant increase in deposition rate associated with the
`magnetic confinement of electrons. The magnetic ‘trapping’
`of electrons in the vicinity of the target surface results in a
`higher probability for electron impact ionization, and hence
`in an increased plasma density. As the ‘magnetic trapping’ is
`increased, the path length and residence time of the electrons
`in the near-cathode plasma region is multiplied and so is
`the probability to undergo ionizing collisions. Hence, the
`
`R174
`
`sputtered flux increases simply due to the fact that more ions
`are generated in the near-cathode region.
`A major drawback of magnetron deposition technology,
`prior to the invention of the unbalanced magnetron, was the
`strong decrease of the ion flux arriving at the substrate, with
`increasing source-to-substrate distances. In 1986 Windows
`and Savvides [21, 22] introduced the unbalanced magnetron,
`where magnetic stray fields are employed to increase the
`ion current density remote from the target surface. Early
`development of the unbalanced magnetron technology was
`carried out by Sproul et al [23], Rohde et al [24], Teer
`[25, 26], Spencer et al [27], Howson et al [28, 29] as well as
`Musil et al [30], Musil and Kadlec [31], Kadlec et al [32, 33],
`Kadlec and Musil [34]. Substantial increases in substrate ion
`current density can be achieved using unbalanced magnetron
`configurations. For example, Kadlec et al [35] have reported
`an increase in substrate ion current density by a factor 80 at a
`source-to-substrate distance of 190 mm as the operation mode
`of the magnetron was changed from balanced to unbalanced.
`A major breakthrough in terms of industrial application
`of unbalanced magnetron sputtering sources came with the
`work of Sproul et al [23] and Rohde et al [24]. Deposition rate
`is a key industrial requirement for the application of coating
`technology in general. By linking the magnetic stray fields of
`two unbalanced magnetron sources, electrons are trapped in
`the volume between the two plasma sources, hence precisely
`in the vicinity of the substrate [23]. Many overviews on these
`topics can be found in the literature, see for example [36, 37].
`It has been shown that magnetic trapping of electrons
`is instrumental in increasing the degree of ionization. This
`is important, since it was well known that the flux and
`energy of condensing ions at the substrate influence the
`microstructure and hence the properties of grown thin films
`[38–40]. Thus, the invention of the unbalanced magnetron
`and later the closed field configuration opened the door to
`exploration of the effect of controlled changes in ion current
`density on the evolution of the film microstructure, and hence
`film properties.
`In the following section, recent advances
`in the field of magnetron sputtering are discussed. Many
`of the advances deal with design of equipment to enhance
`ionization. Subsequent design solutions are geared towards
`controlling the magnitude of the ratio of the condensing ion
`flux to the condensing neutral flux by utilizing magnetic
`confinement in the vicinity of the growing film.
`
`2.2. Magnetic confinement in sputter deposition
`technology
`
`Magnetic plasma confinement with magnetic fields has been
`used in fusion devices, as well as in plasma based etch
`techniques to increase the plasma density and homogeneity,
`see for example [41] and references therein. In a typical low-
`pressure discharge the electron temperatures range from 2 to
`5 eV [42]. Electrons may be trapped at field strengths in the
`range of several mT. This is because the electron gyration
`radius for such field strengths is smaller than the discharge
`dimension. In the presence of magnetic fields, the residence
`time and path length of electrons in the plasma is multiplied,
`as is the probability of causing ionizing collisions.
`Kadlec
`[43] utilized multipolar magnetic
`et al
`confinement of the plasma between the magnetron source
`
`

`
`and the substrate to enhance ionization of an ion plating
`process.
`In this work, plasma confinement is achieved by
`using permanent magnets arranged in a cage between the
`magnetron source and the substrate. Figure 1 shows a
`schematic illustration of the unbalanced magnetron and the
`magnet cage for the multipolar magnetic confinement, which
`in this case is coupled to the field of the magnetron source.
`Kadlec et al [33] demonstrated a plasma homogeneity of
`10% at a source-to-substrate distance of 200 mm with an ion
`−2, that is a threefold increase
`current density of 3–4 mA cm
`compared with the ion current density achieved without
`magnetic field. The technology was applied to deposit high-
`quality TiN films [33].
`Similarly, Petrov et al [44] utilized an external pair of
`Helmholtz coils to create a uniform variable, axial magnetic
`field of ±600 G between the source and the substrate. The
`experimental setup is shown in figure 2. In this case, the ion
`flux to the substrate could be controlled over two orders of
`−2. The ion-to-atom arrival
`magnitude up to several mA cm
`rate ratio could be varied from 0.1 to 6. This configuration
`was used to explore the effect of the ion-to-atom arrival
`rate ratio (Ji /Ja) on the microstructure, phase composition
`and texture and chemical composition of titanium aluminium
`nitride films [45, 46]. It was found that the average energy
`per deposited atom Ed (Ed = Ei (Ji /Ja) where Ei is the ion
`energy) is not a universal parameter describing the effect of
`low-energy ion irradiation on the film microstructure during
`plasma assisted growth [46]. Furthermore, it was shown that
`the mechanistic pathways for the microstructure evolution of
`Ti0.5Al0.5N are different depending on whether Ed is varied
`through changes of Ei at constant Ji /Ja as compared with
`changes of Ji /Ja at constant Ei [45, 46].
`In a detailed plasma probe study with an internal
`magnetron–magnetic coil arrangement, Ivanov et al [47]
`reported that the ion-to-atom flux ratio could be varied from
`0.1 to 5. Furthermore, it was shown that the electron and ion
`density, floating potential and electron temperature could be
`controllably varied by changing the magnetic field strength
`in Ne, Ar as well as Kr ambients. For example, as the
`coil current was varied from 4 A to −4 A in a 3 mTorr Ar
`discharge, the plasma density changed by a factor 30 in the
`vicinity of the substrate, while the target current and voltage
`were unaffected.
`Johansson et al [48] reported the low-temperature
`growth of cBN utilizing an Ji /Ja of up to 27 using an
`internal coil based on the system described by Ivanov et al
`[47]. A system utilizing an internal magnetic coil for dual
`magnetron sputtering was designed and characterized with
`Langmuir probes by Engstr¨om et al [49]. Depending on
`the current supplied to the internal coil, the ion current
`density on the substrate could be varied between 0.2 and
`−2. As the ion current density was increased, the
`5.3 mA cm
`Ti film microstructure changed from an open/porous structure
`to a well defined dense structure. The system consists of
`two 3 inch magnetrons with opposing magnetic poles, where
`the internal magnetic coil can be magnetically coupled to
`the magnetic field of each of the two magnetrons. This
`is illustrated in figure 3, which shows the finite element
`simulations of the magnetic field configuration at a coil
`current of (a) 0 A and a coil current of (b) 5 A. Magnetic
`
`Plasma assisted physical vapour deposition
`
`coupling is achieved by reversing the direction of current
`flow through the internal coil. This results in a powerful
`experimental setup for the deposition of multilayers, since the
`ion-to-atom flux ratio of both targets can be independently
`controlled during the deposition of multilayer film stacks.
`The target shutters and current supply for the internal coil are
`microprocessor controlled. This was shown experimentally
`by Johansson et al [50] for the deposition of CNx/BN:C
`multilayer thin films. Both layer materials were deposited
`under previously optimized conditions [50].
`
`2.3. Pulsed direct current (dc) reactive sputtering for
`dielectric thin films
`
`In reactive magnetron sputtering, the reactive gas can form a
`compound layer on the target surface. The reactive gas partial
`pressure required for the formation of a particular compound
`at the substrate may result in a partially, or even fully,
`reacted target surface. Stable operation of the sputter source
`can be maintained if the compound formed is an electrical
`conductor.
`In the case of a non-conductive compound
`layer, such as TiO2, AlN or Al2O3,
`the positive ions
`impinging on the target cannot be neutralized by electrons
`from the target, which may lead to accumulation of charge
`at the target surface. Electrically this situation represents a
`capacitor. The accumulation of positive charge can result
`in a significant potential difference between the two sides
`of the dielectric compound layer and may increase, during
`Ar ion bombardment, to values as large as the dielectric
`breakdown strength of the compound layer. This in turn
`may result in electrical breakdown or arcing, where a large
`part of the discharge current is concentrated in a small
`surface volume segment (so-called cathode spots) and causes
`local evaporation [51]. Hence, the low-current–high-voltage
`magnetron discharge changes to a high-current–low-voltage
`glow discharge. The physics of this type of glow discharge
`is known as arc evaporation [52] of the target material, and is
`accompanied by two undesirable effects. First, the ejection
`of so-called ‘macroparticles’, leading to inhomogeneity and
`defects in the film; and second, a drastic change in processing
`parameters, such as the cathode potential and the metal to
`reactive gas concentration ratio. Arcing represents a problem
`for the control of the sputtering process and generally leads
`to instabilities.
`Until the advent of alternating current (ac) sputtering
`or medium frequency pulsed dc sputtering, insulating film
`materials (such as alumina or zirconia) were principally
`deposited by radio frequency (rf) sputtering. Since the sputter
`yield of the insulating compound is usually lower than the
`yield for the pure metal it is well known that the sputtered
`flux using rf sputtering (compound target) is at least a factor
`ten smaller than by dc sputtering (metallic target) [53].
`The historical developments that
`led to medium
`frequency pulsed dc sputtering as we know it today are
`outlined in [54], and are not described here. The basic
`principle of pulsed power supplies is that positive charge
`accumulation and hence arcing (as already described) is
`avoided by discharging the target surface with impinging
`electrons. This can be achieved by applying a target potential
`which is more positive than the floating potential in between
`
`R175
`
`

`
`J M Schneider et al
`
`Figure 1. A schematic illustration of an unbalanced magnetron and magnet cage for the multipolar magnetic confinement, after
`Kadlec et al [43].
`
`Figure 2. Schematic of a UHV sputtering system with an external pair of Helmholtz coils, after Adibi et al [45].
`
`sputtering cycles. Patents were issued in both the former East
`[55] and West Germany [56] for pulsed dc power supplies in
`the kHz frequency range in 1986 and 1988. In the same time
`
`period, Este and Westwood [57] published their work on rate
`enhanced sputtering of dielectric materials by a quasi-direct-
`current sputtering technique. The influence of the frequency
`
`R176
`
`

`
`Plasma assisted physical vapour deposition
`
`Figure 3. Finite element simulations of the magnetic field configuration at a coil current of (a) 0 A and (b) a coil current of 5 A.
`
`Figure 4. Normalized deposition rate against frequency after Este and Westwood [57].
`
`on the deposition rate was investigated for AlN films. The
`individual deposition rates were normalized to the deposition
`rate of the dc case and are given as a function of frequency in
`figure 4. Este and Westwood found that the rf deposition rate
`was half of the dc rate, and in the kHz range the deposition
`rate was reduced by 10% to 25% of the dc rate [57]. The film
`properties were reported not to be affected by the frequency.
`The frequency range of today’s commercial power supplies
`that will suppress the electrical breakdown of insulators
`−1) at normalized deposition rate
`(approximately 108 V m
`losses of less than 20% is shown in figure 4. These losses
`can be reduced, if the duty cycle can be increased.
`In 1991 Scherer et al [58] deposited Al2O3, SiO2 and
`Si3N4 by reactive ac magnetron sputtering. The deposition
`technique is based on two magnetron sources with the voltage
`output of the ac power supply driving the sources supplied
`◦
`with a phase shift of 180
`. In this way the two sputter sources
`are run periodically as a cathode and an anode.
`Today, many commercial pulsed dc power supplies with
`frequencies of up to 250 kHz are available. These pulsed
`
`dc power supplies produce either a bipolar symmetric or
`asymmetric pulse. With the symmetric pulsed dc power,
`the pulse height is of equal magnitude for both the positive
`and negative pulses.
`In between the pulses, there is an
`off time, and the width of the positive pulse is usually
`smaller than the width of the negative pulse. The concept
`behind this approach is to apply the positive pulse for just
`enough time to discharge the target surface, but since no
`sputtering takes place during the positive pulse, it should
`be as small as possible while still achieving this effect. For
`the asymmetric pulsed dc power supplies, the magnitude of
`the positive pulse is only a fraction of the magnitude of the
`negative pulse, and the sense of the pulse changes directly
`from positive to negative with no off time. Similar to the
`symmetric bipolar pulsed power, the width of the positive
`pulse is usually smaller than the negative pulse. With both the
`bipolar symmetric and asymmetric pulsed dc power supplies,
`sputtering takes place from the target only during the negative
`pulse, whereas discharging of the target surface takes place
`during the positive pulse.
`
`R177
`
`

`
`J M Schneider et al
`
`Figure 5. Voltage against time trace for asymmetric pulsed dc
`power from a switched source.
`
`Commercially, there is another way of achieving bipolar
`asymmetric pulsed dc power. Instead of producing a pulsed
`output directly from the power supply, a normal dc power
`supply is used, but a switching network is placed between
`the dc power supply and the sputtering target. This switching
`network produces a pulsed dc output from a conventional dc
`input.
`Both ways of producing bipolar asymmetric pulsed dc
`power work well, but there can be differences in the actual
`pulse shape at the target. A typical pulse pattern from the
`switched asymmetric pulsed dc supply is shown in figure 5
`[59]. This looks very close to the ideal schematic drawing.
`There is little overshoot with either the positive or negative
`pulses. However, for the asymmetric pulsed dc power
`delivered directly from the pulsed dc power supply, there can
`be significant overshoot particularly with the negative pulse,
`as is shown in figure 6. This overshoot is a function of the way
`that the power supply operates. The average target voltage
`for this example was about −450 V, but the actual peak-to-
`peak value was almost 1500 V. The voltage overshoot that
`occurs from this pulsed dc power supply is due to the fact
`that the power supply is actually two power supplies in one
`housing. A constant current power supply is used for the
`negative pulse whereas a constant voltage supply is used for
`the positive pulse. The large voltage overshoot occurs in
`the negative pulse when the constant current power supply
`initially turns on and is trying to reach the current operating
`set-point. The effects of this large overshoot in the negative
`pulse are still being determined, but it is safe to say that the
`average energy of the high-energy particles is larger in the
`direct pulsed dc discharge, than in the switched dc discharge.
`Due to the rather large peak currents, the plasma density is
`probably increased in the direct pulsed dc discharge.
`If pulsed dc power is used to bias a magnetron source
`and/or the substrate, it is important to recognize that both the
`ion and neutral energies, as well as the ion and neutral fluxes
`impinging on the growing film, are varying as a function of
`time. If more than one pulsed dc power supply is employed
`to drive a discharge, it is recommended to synchronize them
`in a master–slave setup. For example, if pulsed dc power
`is applied to both the target and the substrate, both power
`supplies should pulse positively at the same time with the
`same frequency, and both should pulse negatively together.
`This can be accomplished by having one of the pulsing units
`act as the master, and the other (slave) mimicking the pulse
`
`R178
`
`Figure 6. Voltage against time trace for asymmetric pulsed dc
`power showing a peak-to-peak voltage of ∼1500 V.
`
`pattern of the master. Such master–slave setups are available
`commercially.
`
`2.4. Process control for dielectric thin films
`
`The reactive dc magnetron sputter deposition of non-
`conducting oxides has almost been an impossible task until
`just recently. As discussed in section 2.3, reactive sputter
`deposition of the non-conducting oxides traditionally has
`been done at very slow rates using rf power. However, it
`was recently shown that by combining medium frequency
`pulsed dc power and partial pressure control of the reactive
`gas, it is possible to reactively sputter dielectric materials,
`such as TiO2 and Al2O3 at high deposition rates [60, 61].
`When flow control of the reactive gas is used for the reactive
`sputtering of a material such as TiO2, initially all of the
`reactive gas will be consumed in the reaction with the metal
`being sputtered from the target as the reactive gas flow is
`increased from a low level (at a constant target power), as is
`shown in figure 7. When the gas flow reaches a certain level,
`there will be sufficient reactive gas to form a compound on
`the substrate, but this same compound will also form on the
`target surface. This reaction of the reactive gas with the target
`surface occurs very quickly for many oxides, and then when
`the targets becomes covered with the compound (poisoned),
`the sputtering rate drops rapidly. Due to the lower sputtering
`rate, less reactive gas is consumed, and its partial pressure
`increases rapidly. With this flow control of the reactive gas,
`there is a whole range of forbidden compositions between
`points A and B in figure 7 that cannot be deposited with this
`technique.
`When the target is fully poisoned, any further increases
`in reactive gas partial pressure lead to a linear increase in the
`reactive gas partial pressure. When the gas flow is reduced,
`it takes time for the compound on the surface of the target
`to be removed by the sputtering process and for metal to be
`sputtered again. When the compound is broken through, the
`partial pressure of the reactive gas drops due to the reaction
`between the large flux of sputtered metal and the reactive gas.
`The drop in the partial pressure when the flow is reduced
`completes the hysteresis curve for this reactive deposition.
`When partial pressure control of the reactive gas is
`used in conjunction with pulsed dc power, the shape of the
`hysteresis curve is quite different to that when flow control
`
`

`
`Plasma assisted physical vapour deposition
`
`Figure 7. Hysteresis curve for reactive pulsed dc magnetron
`sputtering of Ti in an Ar/O2 atmosphere using flow control [60].
`
`Figure 9. AlOx hysteresis curve for reactive pulsed dc magnetron
`sputtering of Al in an Ar/O2 atmosphere using partial pressure
`control [61].
`
`Figure 8. Hysteresis curve for reactive pulsed dc magnetron
`sputtering of Ti in an Ar/O2 atmosphere using partial pressure
`control [60].
`
`alone is used. As is shown in figure 8 when the curve is plotted
`in the same manner as it is in figure 7, there is a negative slope
`region for the partial pressure between points A and B. In this
`region, the target is slowly, but controllably poisoned. At
`point A, the deposition rate is at the full metal rate, whereas
`at point B the target is fully poisoned, and the rate is a
`small fraction of the metal rate. With the partial pressure
`control, all points are accessible between points A and B,
`and there are no forbidden compositions. Such control is
`possible with the combination of partial pressure control of
`the reactive gas and pulsed dc power (or medium frequency
`ac), which together prevent arcing on the target surface and
`thus maintenance of uninterrupted sputtering. Many other
`materials, such as aluminum, zirconium, hafnium, or yttrium,
`when reactively sputtered in an oxygen/argon atmosphere
`have a similar hysteresis curve as is shown above for the
`TiOx system.
`Recently, Schneider et al [61] reported a high-rate
`deposition process for alumina thin films. Figure 9 shows
`the part of the hysteresis curve at the nose of the hysteresis
`curve which was studied. Oxide films produced at 0, 0.36,
`0.41 and 0.43 mTorr O2 partial pressure resulted in 100%,
`92%, 76% and 38% deposition rates relative to the metal
`deposition rate. The deposition rate in a pure Ar discharge
`−1. The rate data for the AlOx is given figure 9.
`is 9.9 Å s
`As expected, increasing the O2 partial pressure reduces the
`
`Figure 10. Relative alumina deposition rates for rf sputtering
`[53], dc sputtering with baffles [53], pulsed dc sputtering [62], and
`pulsed dc sputtering with partial pressure control are compared
`[61].
`
`deposition rate. These experiments showed clearly that this
`high rate can only be achieved by precisely maintaining a
`certain O2 partial pressure value (position on the hysteresis
`curve). In figure 10, the relative deposition rates with respect
`to the metal deposition rate (which is the highest that can
`be achieved for a given input power) of rf sputtering [53],
`dc sputtering with baffles [53], pulsed dc sputtering [62],
`and pulsed dc sputtering with partial pressure control are
`compared [61]. It can be seen that optimum high-rate results
`can be achieved by utilizing both medium frequency pulsed
`power and partial pressure control of the reactive gas. This
`technology was also utilized for the high-rate deposition of
`near stoichiometric zirconia films at a deposition rate relative
`to the metal rate of 82% [61].
`Recent modelling results by Macak [63] suggest that ac-
`tive process control is needed for the high-rate deposition of
`alumina by magnetron sputtering if large target power set-
`tings and short source-to-substrate distances are employed.
`These modelling results are consistent with what has been
`observed experimentally when pulsed dc power and closed
`loop control of the reactive gas partial pressure are used.
`
`R179
`
`

`
`J M Schneider et al
`
`These techniques represent an important advance that
`will continue to influence how reactive processes and process
`controls are applied in the future.
`
`2.5. Modelling of reactive sputtering
`
`Modelling of reactive sputtering processes was recently
`reviewed by Berg et al [64]. For the convenience of the reader
`only the major conclusions are summarized here. Berg’s
`model [65] of the reactive sputtering process is useful to
`predict variations in deposition rate and film composition
`as the reactive gas partial pressure is varied since it has been
`found to be consistent with experimental findings. According
`to Berg et al [64], the basic model was extended to simulate
`reactive co-sputtering systems [66], as well as systems
`utilizing two reactive gases [67]. Most recently, Macak et al
`[68] extended Berg’s model to take the spatial distribution of
`the magnitude of the sputtered flux into account.
`
`3. High plasma density physical vapour deposition
`
`3.1. Introduction
`
`The plasma density at the substrate during conventional
`−3 [69]. Petrov et al
`magnetron sputtering is typically 109 cm
`[70] have shown that the ionized fraction of the sputtered
`flux is of the order of a few percent, and does not contribute
`measurably to the deposition rate [70]. In section 2 of this
`review, several ways to increase the plasma density have been
`discussed:
`• utilizing an unbalanced magnetron with magnetic stray
`fields: ‘the unbalanced magnetron’ (see section 2.1);
`• coupling magnetic fields of two unbalanced magnetrons:
`‘the closed field configuration’ (see section 2.1);
`• as well as with additional magnetic trapping provided by
`internal or external Helmholtz coils (see section 2.2).
`
`The following section is devoted to alternative ways to
`increase the plasma density. First ionized PVD (I-PVD),
`where the ionized fraction of the sputtered flux can approach
`100% is discussed. Then the requirements and limitations of
`so-called ‘self sputtering’ will be reviewed. Then the initial
`reports on high-power pulsed magnetron sputtering will be
`discussed. Finally, recent progress in magnetic filtering of
`cathodic arc plasma and plasma immersion ion implantation
`will be reviewed.
`
`3.2. Ionized physical vapour deposition
`
`The state of the art in ionized physical vapour deposition for
`microelectronic applications has recently been reviewed by
`Rossnagel [71]. For the convenience of the reader a short
`synopsis of Rossnagel’s article is presented, which focuses
`on microelectronics applications. This is followed by a
`discussion on the use of ionized PVD for reactive compound
`synthesis.
`Evaporation [10] and sputtering [72] based electron
`cyclotron resonance (ECR) pro