The fundamentals of sputtering were introduced in sputtering basics. Here we elaborate on Magnetron Sputtering which is one of the sub-categories of sputtering and some of the subtleties of the techniques.
It was stated earlier (see Diode Sputtering ) that DC diode sputtering has essentially fallen from use and has been superseded by DC magnetron sputtering. For RF sputtering it is not quite as clearcut, there are some applications where RF diode sputtering is used in preference over RF magnetron.
In the absence of a confining magnetic field the electrons within an RF system can pick up substantial energy in their oscillations between the electrodes. As a result the environment at the substrate in an RF diode system will have a high abundance of high energy (hot) electrons. These will give rise to substantial heating of the substrate. Some applications benefit from this environment.
DC Magnetron Sputtering
This technique has been widely adopted for the high-rate deposition of metal films. It is possible to achieve a deposition rate as high as 2μm.min-1(for some materials). One disadvantage of magnetron sputtering is that the enhanced erosion of the racetrack leads to relatively poor target utilisation. For inexpensive materials, this might not matter too much. For precious metals it can be a significant issue. Although the spent target will have a significant reclaim value it can lead to higher than desired inventory of precious metal targets. Over the past couple of decades, with the introduction of rare earth magnets, more sophisticated magnet arrays have been introduced that also mitigate the poor utilisation problems.
Using a classical planar magnetron electrode the target erodes preferentially below the region of the electron racetrack. At the periphery and the centre of the target there is net deposition of material. This can lead to particle shedding. Within the semiconductor industry this became an impetus for the development of moving magnet arrays that for many materials can achieve full face erosion.
Magnetron sputtering is significantly more efficient than the DC diode counterpart. This manifests as a relatively low voltage (below 1000 V) but high current operation. The power supply design for magnetron operation is therefore different from those for diode operation. The constraint of the operation voltage in the magnetron case is an advantage over the diode for high power operation. Many tens of kW may be delivered to a magnetron target while maintaining a cathode voltage comfortably below 1000 V. Similar power levels applied to a diode would result in a much higher operation voltage. The reduction in voltage level makes arcs less violent and easier to handle. Modern DC sputtering power supplies have sophisticated arc-handling capabilities.
When a fresh metal target is installed within a system and run for the first time, depending upon the nature of the target material, arcing can be prevalent. Those materials that form robust oxides (such as aluminium, titanium, tantalum etc) can take a while to stabilise, while the oxide layer is cleared. Small islands of oxide on the target surface tend to cause continued arcing. The traditional way of coping with the early operation arcing is to burn-in the target. This process involves operation at a relatively low power for a period of time and progressively ramping the power level. Once a power level above the intended operation level is achieved then arcs during processing should be avoided. For a system equipped with a vacuum load lock this is a workable solution. For a system without a load lock this burn-in procedure needs to be repeated following each exposure to air for sample loading and exchange.
For an equipment builder DC magnetrons are attractive. The reason is that the power delivery is simpler than for RF. The output from a single DC magnetron drive can readily be switched between multiple cathodes for sequential deposition of materials. The absence of an impedance matching unit tends to make DC systems more reliable than those equipped with RF cathodes.
RF Magnetron Sputtering
RF magnetron operation for sputtering electrodes has some attractive attributes especially for research focussed work. Both metal and dielectric materials can be sputtered. Oxide islands and asperities on target sources do not lead to arcing. This can be major benefit, compared to the DC case, where a system does not have vacuum load lock.
The power delivery train is more complex than for the DC case. Between the power supply which has a characteristic output impedance of 50 Ω and the load that has an unknown impedance there needs to be a transformer. At low power (up to 500 W) it is safe to switch the output after the impedance transformer. In such an instance the output from the matching transformer can be a coaxial cable and the unit can be positioned remotely (up to 1 m) from the load. Above this level, however, safe operation requires that each load (cathode) has a dedicated and directly mounted matching transformer. This adds cost and also makes the area adjacent to the cathodes congested. In any system equipped with an RF supply attention needs to be given to mitigation of RF interference, both radiated and conducted.
RF power supplies are less efficient than their DC counterparts. A 10 or even 20 kW DC supply typically is forced-air cooled and more than 90% efficient. Apart from on very large industrial systems it is rare to find RF units rated at more than 5 kW used on a system. Units above 1 kW are typically water-cooled and are less than 70% efficient.
Reactive RF Magnetron Sputtering
So far we have considered the deposition of materials sputtered by Argon. The argon flow may also be augmented with another gas, eg Oxygen or Nitrogen. Gases other than Oxygen and Nitrogen have been used in reactive sputtering including: ammonia, hydrogen sulphide, hydrocarbons and Hydrogen. Sputtering a metal target in the presence of a reactive gas can allow the reactive deposition of an oxide or nitride. The motivation is normally to achieve a higher deposition rate than might be achieved by RF magnetron deposition. As DC power supplies are less expensive for industrial applications, substitution of many RF cathodes with higher power DC ones is an attractive proposition. So for example Al2O3may be deposited by reactively sputtering aluminium metal in the presence of Oxygen. Reactive sputtering can also be used to replenish constituents (permanent gases) of compound targets lost by dissociation. This can be very helpful where the stoichiometry of the deposition film is important to the application.
In practice there are some complications, but through ingenuity they can generally be overcome. Reactions taking part in the gas phase can for the most part be ruled out, due to the requirements for conservation of energy and momentum. Simultaneous conservation of energy and momentum require that the reactions take place at a surface. Generalising, there are three surfaces available: the target surface, the substrate surface and the chamber (or chamber liner) surface. At a high target sputtering rate and low reactive gas partial pressure, it is well established that essentially all reactions occur at the substrate. In this instance the stoichiometry of the film depends on the relative rates of arrival at the substrate of metal vapour and reactive gas. For these conditions the rate of removal of compounds at the target surface is greater than the rate of formation. As the reactive gas partial pressure is increased and/or the target sputtering rate is decreased, however, a threshold is reached at which the rate of compound formation at the target surface exceeds their removal rate. For metal targets, this event is normally accompanied by an abrupt drop in the sputtering rate. At this point the target is often referred to as poisoned. There are several factors that contribute to this drop in rate: generally compounds have a lower deposition rate than metals; they also have higher secondary electron emission yields than metals; also with a higher partial pressure of the reactive gas the resulting mixture has a lower sputtering efficiency than pure Argon. Deposition will continue but at a substantially reduced rate.
Reactive RF magnetron sputtering was actively developed industrially in the 1980s. The characterisation of the events and a useful technique to mitigate the deposition rate loss are captured in a Patent filed by Borg Warner in 1983.
Diagram E represents a hysteresis curve mapping the excess reactive gas against the flow of reactive gas into the chamber. At the starting point Ⓐ the magnetron racetrack is fully metallic and at Ⓓ it is fully poisoned. In between these states the conversion of the racetrack surface from metallic to poisoned proceeds through transition regions Ⓑ and Ⓒ with increasing reactive gas flow. Intense ion bombardment within the racetrack delays the onset of fully reacted (poisoned) state. When reactive gas is first introduced its partial pressure remains low, the gas is rapidly consumed through reaction with the metal deposited in the system. In the region from Ⓐ to Ⓑ the deposition rate of the reacted product is that of the metal. At Ⓑ the partial pressure of the reactive gas starts to rise quickly. The racetrack surface at this point is becoming partially converted to the reacted product compound, which sputters at a much lower rate than the metal. At point Ⓓ the sputtering rate (of the compound) is very low compared to the metal.
Now when reducing the reactive gas flow hysteresis is observed and partial pressure does not start to rapidly decrease until the compound layer covering the racetrack is broken through, at Ⓔ.
The composition of the compound deposited also varies as we pass around this curve. The fully reactive stoichiometric compound is observed at point Ⓑ. The difficulty is that trying to operate in this region is like trying to balance an upside down pendulum. The solution is rapid feedback control of the partial pressure using a mass spectrometer enabling the system to sit at point Ⓑ, so exploiting the maximum achievable rate while maintained the desired film stoichiometry.
Reactive DC Magnetron Sputtering
Since DC power delivery is simpler and generally less expensive than an equivalent RF unit there is an incentive for adoption in industrial applications. This is an instance where the thin film industry has been served well by the development of electrical components and fast control of power supplies. When DC power supplies were thyristor controlled the options were limited. Today with very fast switched mode technologies and very fast processor control some very sophisticated power supplies are available.
DC cathodes have a tendency to arc. The arcs are detrimental to the quality of the films deposited. Early solid state power supplies introduced arc handling capabilities. These typically work by turning the output off, then on again after a few tens of milliseconds. DC magnetron electrodes tend to be sensitive to disturbances within the racetrack region. The tendency to arc may be exacerbated by the low impedance of the discharges. A unipolar arc can be established resulting from dielectric breakdown of insulating inclusions. These can be initiated at, for example, a small oxidised spot on the cathode target surface. They can result in virtually the entire discharge current to be concentrated into the arc spot on the target. These are in fact the type of events seen during the burn-in of a fresh target. Starting with a well conditioned and clean target (for example of Aluminium) when oxygen is introduced with an aim of reactive deposition of alumina Al2O3 unipolar arcing is encouraged. As the partial pressure is increased arcing can become dramatic. In the RF instance any charge build up is neutralised during the periodic cycle of supply.
Power supplies are available today that exploit switched mode topology with additional features to allow temporary periodic current reversal. These pulsed DC magnetron drives can mitigate the arcing in DC magnetron operation. Since the real advantage is for high power and industrial operation these units tend to have relatively high output power, ranging from 5 kW to 60 kW.
The Disappearing Anode
There is another practical consequence of depositing insulating films in a DC system. With increasing deposition time the whole of the internal structure of the chamber, including the wall in contact with the discharge will become coated with the isolating material. This deprives the discharge of a return current path as the anode becomes hidden beneath the non-conductive film. This phenomena is referred to as the disappearing anode. As the coating develops, the impedance of the load begins to rise, causing the discharge voltage to increase, and the power supply will eventually loose the ability to regulate the power. When the film is fully formed the plasma will extinguish. In a laboratory or small-scale application this may be a nuisance. In an industrial manufacturing environment it has a major impact.
In the 1970s work on arc free sputtering of TiO2 at Airco was published. This involved the use of a low frequency AC supply substituted for the normal DC unit. This demonstrated the adoption of an AC supply with a frequency in the range from 400 to 60,000 Hz. The required frequency is dependent upon the dielectric constant of the insulating film to be deposited.
While this clever approach addresses the problem with arcing the issue of the disappearing anode remains. The first publication of a solution to this was published a decade later by a group at Bell Northern Research. By connecting two cathodes (Dual Cathode) to a common AC power supply, continuous arc free deposition of AlN (aluminium nitride) was achieved. The technique also works for oxides.
More recently with advances in the power supply capabilities the goal of achieving arc-free continuous deposition of dielectric material by reactive DC magnetron sputtering may have also become a reality. Work published by Advanced Energy promotes the possibility of Dual-cathode pulsed DC magnetron sputtering with a floating anode.
In discussing pulsed DC magnetron sputtering there is one further class of technique to introduce: this is High Power Impulse Magnetron Sputtering (HiPIMS). The concept here is to deliver a similar average power to that employed in non-pulsed DC magnetron sputtering, but through periodic low duty-cycle pulses of intense power. The use of a very high power pulses lead to the generation of very high discharge density and a subsequent high fraction of ionisation of the sputtered target material. Now with a high fraction of the ions in front of the sputtering target comprising ionised target atoms we enter a regime of self-sputtering. This combined with the local recycling of sputtering gas are characteristic of HiPIMS.
A high fraction of ionised deposition material can be exploited to affect the morphology and properties of the growing film. Studies using synchronised bias have demonstrated it can be used to control the microstructure of a deposited film.
An attractive feature of the HiPIMS technique is that it requires little or no modification to a standard magnetron sputtering system, other than provision of the specialised power supply.