Definition: Reactive Sputtering is a process where a target of one chemical composition (e.g. elemental Si) is sputtered in the presence of a gas or a mixture of gasses (e.g. Ar + O2) that will react with the target material to form a coating of a different chemical composition (e.g. compound SiO2). Argon is in most cases the main gas and the amount of a reactive gas introduced into a process chamber is controlled to either achieve certain amount of doping or produce a fully reacted compound.
Today Reactive Sputtering deposition is a well established technique and is widely used for industrial coating deposition as well as for research and development. Reactive Sputtering is employed for production of thin layers for high added value products, such as flat panel displays, solar cells, optical components and decorative finishes.
The driving force behind the widespread use of this technology is the fact that operation in the so called ‘transition’ region (explained in the later chapters) allows high coating deposition rates to be obtained using AC or DC power, thus providing a much superior alternative to sputtering ceramic compound targets with RF power. Reactive sputtering processes can provide significant cost savings as well as production rate advantages.
Reactive Sputtering is most often practised using one or more magnetron sputtering cathodes. Most popular geometries include planar and cylindrical cathodes (figure 1), although other geometries exist (one example is inverted cylindrical cathode as shown in figure 1c).
Simple and inexpensive sputtering cathode designs that are intended for DC magnetron sputtering of metal films are in most cases well suited for Reactive Sputtering processes.
A significant cost saving that Reactive Sputtering technology offers is due to readily available and relatively inexpensive target materials. Metal targets can be used to produce compound materials. For example, a Ti target can be used to produce coatings such as TiO2, TiN and Ti-O-N. The same Ti target can be used to produce any of the aforementioned different compositions as well as boride and carbide films. A further advantage of using a metal target, as compared to a compound target, is that for the same target thickness metal target life time is longer. Some examples of sputter targets are shown in figure 2.
Figure 2. Sputtering target examples: (a) rectangular Ti target bonded on a Cu plate (courtesy of Angstrom Sciences, Inc.) (b) sub-oxide TiOx rotary target (courtesy of Process Materials, Inc.) (c) cylindrical CuGa and In sputter targets (courtesy of Indium Corporation), (d) AZO cylindrical target (courtesy of Process Materials, Inc.)
An important category of sputter targets used in Reactive Sputtering for oxide film deposition is the so called sub-oxide targets (figure 2b). An excellent example is a TiO1.8 case where the oxygen deficient target material is electrically conductive enough (~ 0.3 Ohm cm) to apply significant amount of DC or AC power to it. Reactive gas is still needed to complete formation of the compound being deposited and to adjust and improve coating thickness and property uniformity.
Reactive gas is a type of gas which undergoes chemical reactions with materials in contact. Gases that are inert in ambient conditions (e.g. molecular N2) become reactive when present in a plasma discharge due to collisions with energetic particles and subsequent dissociation into atomic neutral (e.g. N) and charged (e.g. N+, N-, N2+) components.
Most often used reactive gasses are:
- Oxygen (O2) - deposition of oxide films (e.g. Al2O3, SiO2, TiO2, HfO2, ZrO2, Nb2O5, AZO, ITO).
- Nitrogen (N2) – deposition of nitride films (e.g. TiN, ZrN, CrN, AlN, Si3N4, AlCrN, TiAlN)
- Carbon dioxide (CO2) – deposition of oxide coatings.
- Acetylene (C2H2) – deposition of metal-DLC, hydrogenated carbide, carbo-nitride films.
- Methane (CH4) – similar applications as for C2H2.
Several reactive gasses can be mixed (e.g. O2 + N2) in order to deposit multi-component functional thin films, such as oxy-nitrides (e.g. Cr-O-N) or oxy-carbides (e.g. Si-O-C). Additional reactive gas is sometimes used to enhance a certain deposition process (e.g. addition of N2 in SiO2 reactive sputtering process).
Reaction of gas with target material
Once reactive gas is introduced into a process chamber it reacts with the unpassivated surfaces, such as chamber walls and sputtering target. Intensive erosion (due to sputtering) in the magnetron racetrack delays complete poisoning (compound formation) of the racetrack surface. Racetrack poisoning therefore normally proceeds through different target states as a function of partial pressure of reactive gas and time, as, for example, is shown schematically in Figure 3. The starting stage is a clean metallic target racetrack (‘metal’ sputtering mode) and the final stage is a fully reacted racetrack surface (‘compound’ or ‘fully poisoned’ sputtering mode). The stages in between are called the ‘transition region’.
Figure 3. Target poisoning during reactive sputtering (after Sproul ).
Poisoning of the racetrack surface is associated with a number of effects, of which the following two have great importance:
- Change in secondary electron emission, which results in a change of the discharge impedance. Consequently, at the same discharge power, the current and voltage can change substantially as reactive gas is introduced (see figure 4).
- Drop in sputtering rate due to typically lower sputter yield of elements from a compound. For example Al2O3 sputtering rate is an order of magnitude lower than that of Al. Figure 5 Hollands  provides a nice example of this effect for reactive tantalum sputtering in oxygen/argon atmosphere.
Figure 4. Ti target voltage change as a result of oxygen flow and consequent racetrack oxidation.
It is known that three magnetron sputtering target poisoning mechanisms can be taking place [3, 4]:
- ion implantation and
The thickness of the compound formed depends on the location in the racetrack. It is generally accepted that AC and DC processes at typical cathode operation conditions can result in a few nanometre compound layers (mostly by chemisorption and ion implantation). “Hot” processes (characterised by an elevated sputter target bulk and/or surface temperature), such as Hot Target Sputtering (HTS)  or High Power Impulse Magnetron Sputtering (HIPIMS) , can have pronounced diffusion taking place, with diffusion profiles extending as deep as a few tens or even hundreds of nanometres .
Figure 5. (a) Deposition rate decrease due to target poisoning (after Hollands ), (b) an arc in a magnetron discharge (courtesy of Deposition Sciences, Inc.), (c) a TEM micrograph showing an example of a droplet incorporated in a ceramic coating and resultant structure evolution, (d) SEM micrograph showing arc damage on the surface of a ceramic coating.
DC or AC sputtering of a metal target in a ‘fully poisoned’ regime is generally associated with not only low deposition rates, but also with significant arcing (figure 5b), poor process stability and less than ideal coating quality. RF power can be used to sputter electrically non-conducting (i.e. fully poisoned) surfaces, but deposition rates are low and this power technology is more complicated and expensive, due to necessity to use matching boxes.
Models have been developed  that can reproduce to some degree the behaviour of Reactive Sputtering processes as observed in production or research systems. The degree of accuracy of predictions based on such models can vary greatly.
The fast transition and hysteresis loops
The fast transition from ‘metal’ to ‘fully poisoned’ target states is one major instability that is inherent to many reactive sputtering processes. The other is severe arcing, mostly observed in reactive dielectric film sputter deposition processes. While arcing is always related with defects in products and poor process performance, the transition region provides opportunities, such as to produce better coatings faster and to fine tune film chemical composition and properties.
Fast transition phenomena can be observed clearly by increasing steadily flow of a reactive gas and observing the target state through the response of a sensor signal. Figure 6a shows one such experiment. If the reactive gas flow is then reduced, at the same rate as it was increased, another fast transition in an opposite direction takes place (figure 6b). The region between the ‘metal’ and ‘fully poisoned’ states is called the ‘transition region’, regardless of the direction of the transition.
Hysteresis plots are sometimes used as a process characteristic and it is typically generated by flowing reactive gas and recording a sensor signal in the following manner:
- increasing reactive gas flow at a constant rate till the target fully poisons,
- decreasing the reactive gas flow at the same rate till the gas flow rate is zero,
- waiting (if needed) for the target to get to the original clean ‘metal’ state,
- plotting ‘sensor signal vs. reactive gas flow’ and/or ‘sensor signal vs. time’ plot.
Figure 6. Target state transitions during reactive sputtering a) transition from ‘metal’ to ‘fully poisoned’ state b) transition ‘fully poisoned’ to ‘metal’ state. The sensor signal is optical emission from a magnetron plasma discharge.
Figures 7a and 7b show two examples of hysteresis plots, which have been produced using the data recorded for the same process as shown in figures 6a and 6b. Hysteresis loops are useful for characterising reactive sputtering processes, however, a simple reactive flow ramp (figure 6) allows to see the transition and discern the process window quickly and easily too with no additional data processing. Most of the times, this provides all the information required by a process engineer setting up a reactive sputtering-based deposition process.
Figure 7. Hysteresis loops plotted in two different ways: (a) sensor signal vs. oxygen flow rate and (b) sensor signal vs. time. The latter hysteresis plot method has been introduced by Audronis et al.  and it can often be useful in revealing the full extent of the hysteresis.
It is common to visualise hysteresis loops by plotting sensor signal vs. reactive gas flow. However, since the physical meaning of the hysteresis is a time delay as a result of metal target surface poisoning and depoisoning, hysteresis can be also revealed by plotting ‘sensor signal’ vs. 'time’ (figure 7b), provided that there was a linear relationship (for clarity and convenience) between time and gas flow during the reactive gas flow increase/decrease experiment.
Various sensors can be used to characterise reactive sputtering processes. Different sensors will result in differently looking hysteresis loops. If the sensor used for an experiment is not appropriate for the given process or misplaced in a process chamber, the hysteresis can appear “reduced” or not be observed at all.
One important parameter, which can slow down the fast transition and (consequently) have a stabilising effect, plus change the shape of the hysteresis loop, is pumping speed. It has been known for decades that it is possible to “pump out” the hysteresis (i.e. make the transition less sudden and reduce the time delay between the different target states ). Substantial increase in pumping speed is required to achieve that, which is impractical for the majority of industrial coating systems. Hence, this stabilisation method through very high pumping speed of the process chamber is rarely implemented in medium-to-large size production systems.
The fast transition and hysteresis effect in reactive sputtering processes is more pronounced in the situations when the reactive gas used has high chemical affinity for a metal sputter target (e.g. the case of sputtering Aluminium in Argon/Oxygen atmosphere).
Control of Reactive sputtering processes
Reactive sputter deposition of high quality thin films at high deposition rates requires operation in the transition region. Closed loop process control systems are employed to provide stable operation in the inherently unstable transition region. Modern systems have a capability to control a few reactive sputtering processes at the same time (e.g. multi zone processing in large area coating systems). Proprietary PID and PDF+ control algorithms are commonly used for closed loop control. Fuzzy logic-based algorithms are practised, but are not as widespread.
Process control systems require input or inputs (e.g. signals from sensors) that indicate the state of the process environment. Usually it is one of the three basic (or a combination of any of these three) feedback signals that are used to control the reactive gas during deposition:
- target voltage (function of the discharge impedance),
- optical emission from plasma (reactive gas or target material; plasma can also be generated remotely); this technique is often called Optical Monitoring (OM) or Plasma Emission Monitoring (PEM), and
- partial pressure of reactive gas (e.g. as measured by a mass spectrometer or λ-sensor).
Of the above three options, the target voltage is the cheapest as no additional measurement or monitoring devices are required. Of the oxygen partial pressure sensor based systems, mass spectrometers can exhibit good response times, between 30 and 60 ms, but are the most expensive, while λ-sensors are much slower - response times typically are between 100 and 200 ms. The PM based techniques are low cost and can be very effective since changes in the reactive environment can be detected and updated on the order of every 1 ms.
Process control systems ensure high short and long term process stability. Long term process drifts can occur due to reasons, such as target material consumption and associated discharge parameter change, change in process environment (cleanliness, gaseous atmosphere) and change in sensor outputs. Long term process drifts can be tackled using modern process control systems, proprietary software solutions and good process engineering practice.
High rate reactive sputter deposition
One of the biggest advantages of operation in the transition region is substantially increased deposition rate. The gain in deposition rate is not 5%, 10% or 15%, it is normally a number of times (e.g. 2 or 3 times). Production rate of a coating production line can therefore be increased proportionally. Production rate improvements of approximately 3 times are common as compared to reactive sputtering processes with no control or when sputtering a compound target.
Actual deposition rate of the compound thin film, as well as its chemical composition, depend strongly on the chosen operation set-point in the transition region. This is illustrated in figure 8.
Figure 8. AlOx process: deposition rate dependence on the set-point.
Another great advantage of operation in the transition region is flexibility in fine tuning coating chemical composition and ability to optimise and improve substantially film properties. The following papers provide a few examples of such improvements:
- Malkomes, N., Vergöhl, M., Szyszka, B. “Properties of aluminum-doped zinc oxide films deposited by high rate mid-frequency reactive magnetron sputtering” Journal of Vacuum Science and Technology, Part A: Vacuum, Surfaces and Films, Vol. 19, March 2001, Pages 414-419.
- May, C., Strümpfel, J. “ITO coating by reactive magnetron sputtering-comparison of properties from DC and MF processing”, Thin Solid Films, Vol. 351, 30 August 1999, Pages 48-52.
- Scherer, M., Wirz, P. “Reactive high rate d.c. sputtering of oxides”, Thin Solid Films, Vol. 119, 14 September 1984, Pages 203-209.
Reactive sputtering-based coating systems
Coating systems that often integrate Reactive Sputtering processes can be divided into the following four groups:
- In-line large area glass or metal sheet coaters (available in horizontal, vertical or vertical inclined, configurations; applications include architectural glass, decorative metal sheets, flat panel displays, etc.) – figure 9a and 9c.
- Roll-to-roll (R2R) web coaters (used for coating polymeric web, textile, metal foil, etc.) – figure 9b,
- Batch coaters (optics, machining components) – figure 9d and 9e and
- Cluster coaters (OLED displays).
Figure 9. Examples of coating systems: (a) Large area glass in-line coater (courtesy of Leybold Optics GmbH) (b) web coater (courtesy of Mustang Vacuum, Inc.) (c) in-line coater (courtesy of Semicore, Inc.) (d) batch coater (courtesy of Kolzer S.r.l.) and (e) batch coater (courtesy of P&P Holding)
Application of reactively sputtered coatings
Coatings produced by Reactive Magnetron Sputtering are used in a large variety of products. Below are given a few application examples:
- In OLED displays ultra barrier coatings are used to encapsulate an OLED device (figure 10 a-b).
- In food packaging barrier coatings are used to stop water and oxygen vapour transmission thus preserving food for a longer time.
- Multilayer oxide antireflective coatings are used to reduce the amount of light reflected light (figure 10c).
- Scratch resistant coatings (figure 10c),
- Multilayer oxide filters for lasers (figure 10d).
- TCO layers are used in FPDs and solar cells as transparent conducting electrodes (figure 10 e-f).
- Oxide and nitride coatings are used for decorative purposes (figure 10g).
- Metal-DLC coatings are used as hard and wear resistant layers on various mechanical components such as machining components (e.g. drill bits, mills), automotive components (e.g. piston rings, shafts, gears), consumer items (e.g. knife blades, watch cases). Figure 10h shows an example.
- Nitride coatings for tools (Figure 10i).
- Oxy-nitride coatings have too many applications to list them all here. One interesting application is a solar absorber layer (e.g. using Cr-O-N, or Ti-O-N) in a thermal solar cell. Figure 10j shows an example.
Figure 10. Examples of reactively sputtered coatings: (a) and (b) devices with an OLED screen, (c) anti-reflective coatings, (d) laser optics elements (e) ITO coated glass, (f) ITO coated polymeric web, (g) decorative oxide coatings (h) TiN coated taps, (i) DLC coated end-mill, (j) Ti-O-N thermal solar absorber.
- Sproul, W. D. and Tomashek, J. 1983 CA1198084A1
- Hollands, E., Campbell, D.S. “The mechanism of reactive sputtering” Journal of Materials Science, Vol. 3, September 1968, Pages 544-552.
- Depla, D., Heirwegh, S., Mahieu, S., De Gryse, R. “Towards a more complete model for reactive magnetron sputtering” Journal of Physics D: Applied Physics, Vol. 40, 7 April 2007, Article number 019, Pages 1957-1965.
- M. Audronis, G. Abrasonis, F. Munnik, R. Heller, P. Chapon, V. Bellido-Gonzalez, “Diffusive racetrack oxidation in a Ti sputter target by reactive high power impulse magnetron sputtering” Journal of Physics D: Applied Physics, Vol. 45, 19 September 2012, Article number 375203.
- Mercs, D., Perry, F., Billard, A. “Hot target sputtering: A new way for high-rate deposition of stoichiometric ceramic films” Surface and Coatings Technology, Vol. 201, 4 December 2006, Pages 2276-2281.
- Kouznetsov, V., Macak, K., Schneider, J.M., Helmersson, U., Petrov, I. “A novel pulsed magnetron sputter technique utilizing very high target power densities” Surface and Coatings Technology Vol. 122, 15 December 1999, Pages 290-293
- Berg, S., Larsson, T., Nender, C., Blom, H.-O. “Predicting thin-film stoichiometry in reactive sputtering” Journal of Applied Physics, Vol. 63, 1988, Pages 887-891.
- Audronis, M., Bellido-Gonzalez, V “Investigation of reactive high power impulse magnetron sputtering processes using various target material-reactive gas combinations” Surface and Coatings Technology, Vol. 205, 15 March 2011, Pages 3613-3620.
- Kadlec, S., Musil, J., Vyskocil, H. “Hysteresis effect in reactive sputtering: A problem of system stability” Journal of Physics D: Applied Physics, Vol. 19, 1986, Article number 004, Pages L187-L190.