A. How to create a highly ionized discharge

In glow discharge processes such as magnetron sputtering, it is often difficult to achieve a large fraction of ionized sputtered material reaching the substrate.Reference Christou and Barber^19–Reference Konstantinidis, Ricard, Ganciu, Dauchot, Ranea and Hecq²¹ When the deposition flux consists of more ions than neutrals, the process is referred to as ionized PVD or IPVD.Reference Helmersson, Lattemann, Bohlmark, Ehiasarian and Gudmundsson⁵ There are a few different IPVD techniques available today, such as postvaporization ionization using a secondary plasma generated by, for example, an RF coil placed in the deposition chamber to create ions that can be accelerated to the substrate surface when applying a negative bias.Reference Nouvellon, Konstantinidis, Dauchot, Wautelet, Jouan, Ricard and Hecq²² Another technique is the previously mentioned cathodic arc evaporation (see references Reference Johnson, Vossen and Kern23 and 24), which uses the fact that very localized, extremely high current discharges can create a dense plasma resulting in a high degree of ionization around a particular spot. A third possibility is the use of hollow cathode magnetron sputtering.Reference Wang and Cohen^25, Reference Klawuhn, D’Couto, Ashtiani, Rymer, Biberger and Levy²⁶ The hollow cathode traps electrons in a hollow cylinder or between two parallel plates. It works like two electrostatic mirrors reflecting electrons between the sheaths until they are thermalized through collisions, thus increasing the plasma density and the probability of ionizing any material passing through.Reference Söderström²⁷

Another approach to IPVD is to use HiPIMS, which is the focus of the present review. HiPIMS can be set up on a conventional magnetron sputtering system by changing power supplies making it possible to deliver high power pulses to the magnetron in the range of a few kilowatts per square centimeter while keeping the time-averaged power on a DCMS level of typically watts per square centimeter to avoid damaging the magnetron.Reference Sarakinos, Alami and Konstantinidis¹⁸ Such a HiPIMS discharge pulse is shown in Fig. 1. The length of the pulse is often kept in the range 10–500 μsReference Konstantinidis, Dauchot, Ganciu, Ricard and Hecq^28, Reference Lundin, Brenning, Jädernäs, Larsson, Wallin, Lattemann, Raadu and Helmersson²⁹ with a pulse frequency from tens of hertz to kilohertz.Reference Helmersson, Lattemann, Bohlmark, Ehiasarian and Gudmundsson^5, Reference Wallin and Helmersson³⁰ The applied voltage during the pulse in most HiPIMS processes is usually around 500–1000 V, and the peak current density (peak discharge current/target area) reaches at most a few amperes per square centimeter (see Ref. Reference Alami, Sarakinos, Uslu and Wuttig31).

FIG. 1.

Voltage and current characteristic for a typical high-power impulse magnetron sputtering (HiPIMS) discharge pulse. This discharge was generated using a 6’’ Cu target at 1.33 Pa Ar pressure. The peak current density is about 1.6 A cm⁻².

B. Plasma conditions

The mechanism of transferring a process gas to a HiPIMS plasma can be seen as the analog of dielectric breakdown in an insulating solid, where the dielectrics will start conducting current at a critical voltage. In the case of the process gas it starts with free electrons, caused by background radiation or thermal energy, being accelerated toward the anode (chamber walls or the ground shield of the magnetron) by an electric field, created by the voltage difference applied between the cathode (target) and anode. The accelerated electrons will gain energy and eventually collide with neutral gas atoms, which in some cases will lead to ionization and the release of two free electrons per ionized atom (electron impact ionization). These two electrons can now collide with two other neutrals, whereas any gas ion present will be accelerated and collide with the cathode releasing, among other particles, electrons (referred to as secondary electrons). Eventually, this process leads to the breakdown of the process gas resulting in a plasma. Through the use of very high applied instantaneous power densities to the magnetron, there will be a tremendous increase of charge carriers in front of the target during the HiPIMS pulse. In numbers this means that for the HiPIMS discharge, the electron density in the ionization region close to the target surface is on the order of 10¹⁸–10¹⁹ m^-3.Reference Gudmundsson, Sigurjonsson, Larsson, Lundin and Helmersson^32, Reference Bohlmark, Gudmundsson, Alami, Lattemann and Helmersson³³ For an electron density around 10¹⁹ m^-3, the ionization mean free path of a sputtered metal atom is about 1 cm, whereas for an electron density of 10¹⁷ m^-3, commonly observed in a DCMS discharge, the ionization mean free path is ∼50 cm for typical discharge conditions.Reference Gudmundsson³⁴ Thus, given the high electron density in the HiPIMS discharge a significant fraction of the sputtered material is thereby ionized, which also has been verified in a great number of publications.Reference Kouznetsov, Macák, Schneider, Helmersson and Petrov^4, Reference Vlcek, Kudlacek, Burcalova and Musil^35–Reference Macák, Kouznetsov, Schneider, Helmersson and Petrov³⁷ Worth pointing out is that there are substantial differences in the degree of ionization of the sputtered material depending on what target material is used (from a couple of percent to almost fully ionizedReference Bohlmark, Alami, Christou, Ehiasarian and Helmersson^36, Reference DeKoven, Ward, Weiss, Christie, Scholl, Sproul, Tomasel and Anders³⁸). The reason for this behavior is mainly related to the fact that the ionization potential is material dependent and for metals used in magnetron sputtering commonly found in the range E _IP = 5.99 eV (Al) to E _IP = 11.26 eV (C), meaning that Al is more easily ionized than C. In plasma discharge modeling by both HopwoodReference Hopwood and Hopwood³⁹ and Samuelsson et al.,Reference Samuelsson, Lundin, Jensen, Raadu, Gudmundsson and Helmersson⁸ these trends are clearly shown and in Fig. 2, one such example is illustrated for HiPIMS using a 100 μs discharge pulse of −880 V and 30 A peak current (corresponding to a peak current density of about 1.5 A cm⁻²). As a reference, for DCMS discharges the degree of ionization is found to be 5% or less.Reference Hopwood and Hopwood^39, Reference Rossnagel and Hopwood⁴⁰ Further discussions on the impact of a high degree of ionization of the sputtered material on the thin film growth will be given in Section III. Before continuing, a word of caution should be given regarding ionization, since there is occasionally some misunderstanding concerning reported values of the ionization of sputtered material, which often has to do with the fact that the degree of ionization of sputtered material (in the plasma), n_i/(n_i + n_n), is not of the same value as the ionized flux fraction (flux of material reaching the substrate), Γ_i/(Γ_i + Γ_n). This is because the ion flux is governed by the electron temperature, T _e, and the neutral flux by the gas temperature, T _g. For partially ionized plasma discharges, such as the HiPIMS discharge, T _e is often significantly larger than T _g (in HiPIMS the dense process gas, n _Ar ∼ 10²⁰ m^-3, is far from being completely ionized thus affecting T _g). This means that the ion flux fraction is larger than the degree of ionized metal and therefore it is not uncommon to find that although the degree of ionization only reaches about 50%, the ionized flux fraction at the substrate can be more than 90%.Reference Samuelsson, Lundin, Jensen, Raadu, Gudmundsson and Helmersson⁸

FIG. 2.

A comparison of the evolution of the degree of metal ionization for different target materials from a global plasma model. The discharge current peaks at about 30 A at ∼50 μs into the pulse (the peak current density is about 1.5 A cm⁻²). A 100 μs square voltage pulsed of −880 V was used in the modeling (from Samuelsson et al.Reference Samuelsson, Lundin, Jensen, Raadu, Gudmundsson and Helmersson⁸ with permission from Elsevier).

The HiPIMS plasma has also shown the presence of multiply charged ions for various target materials,Reference Bohlmark, Lattemann, Gudmundsson, Ehiasarian, Aranda Gonzalvo, Brenning and Helmersson^7, Reference Vlcek, Kudlacek, Burcalova and Musil^41–Reference Andersson, Ehiasarian and Anders⁴⁴ which is rarely found in other types of magnetron discharges but often found in cathodic arc discharges.Reference Rosén, Anders, Mráz and Schneider⁴⁵ One implication of having this type of ions is that they enable sustained self-sputtering,Reference Anders⁶ meaning that they will sputter enough material for continued ionization and release enough secondary electrons to keep the plasma burning assuming that enough sputtered metal ions can be attracted back to the target. For high yield materials such as Cu this has led to the onset of self-sputtering regimes characterized by a second increase of the discharge current beyond the value of the initial peak current, if the pulse is sustained for a long enough time in combination with high negative discharge voltages.Reference Anders, Andersson and Ehiasarian^43, Reference Anders⁴⁶ The singly ionized sputtered atoms cannot contribute to this process, since they often do not exceed the ionization energy threshold required for the release of secondary electrons into the plasma. This has led to the possibility that HiPIMS processes can be run without the use of a process gas, which opens up the field of ultraclean sputtering.Reference Andersson and Anders⁴⁷ Having multiply charged ions in the plasma might also lead to a high energy bombardment of the growing film, since ions will be accelerated in the plasma sheath between the bulk plasma and the substrate. An applied substrate bias of −50 V will, for example, accelerate a doubly charged ion to 100 eV, which might lead to implantation of the primary (incoming) species.Reference Wei, Zhong-Quan, Ye and De-Ming⁴⁸ This in turn may induce compressive stress as the bombarding species are distorting the lattice of the sputtered films,Reference Ohring¹ which might cause the film to peel off. Therefore, one has to take care when running a process containing these types of ions.

Moreover, the metal ions generated in the HiPIMS plasma are found to be highly energetic.Reference Bohlmark, Lattemann, Gudmundsson, Ehiasarian, Aranda Gonzalvo, Brenning and Helmersson⁷ From ion-energy measurements in HiPIMS discharges, it has been found that the average ion energy in the bulk plasma during the discharge pulse is around 20 eV without the use of any substrate bias.Reference Bohlmark, Lattemann, Gudmundsson, Ehiasarian, Aranda Gonzalvo, Brenning and Helmersson^7, Reference Lundin, Larsson, Wallin, Lattemann, Brenning and Helmersson⁴⁹ As it turns out, ion energies in the range of 20–30 eV have been shown to have a densifying effect on thin films,Reference Eriksson, Ghafoor, Schäfers, Gullikson and Birch⁵⁰ which is a very fortunate result and will be explored further in Section III C dealing with thin film growth. In Fig. 3, the ion-energy distributions for HiPIMS and DCMS for Ti⁺ ions are compared. In the case of HiPIMS, a high energy part around 10–15 eV is clearly detected followed by a high energy tail. Last, from measurements by Hecimovic and EhisarianReference Hecimovic and Ehiasarian^51, Reference Hecimovic and Ehiasarian⁵² on the ion-energy distribution for various metals as well as for Ar, it is also seen that ions are long-lived in the HiPIMS discharge and in some cases present during the entire pulse-off time (up to 10 ms), although there is an exponential ion density decrease. The ion energy and life span of ions are found to depend strongly on collisions of the sputtered flux with the surrounding gas and as such depends on the target material, where light elements, such as C and Al, lose more energy in collisions with Ar atoms and thereby have a shorter lifetime than heavy elements, for example, Nb.Reference Hecimovic and Ehiasarian^51, Reference Hecimovic and Ehiasarian⁵² The implications of having ion bombardment of the thin film during the off-time of the HiPIMS pulse have so far not been investigated. One should also bear in mind that deposition takes place during the pulse-off time because of the time needed for sputtered material to reach the substrate.

FIG. 3.

Comparison between Ti⁺ ion-energy distributions from HiPIMS and direct current magnetron sputtering (DCMS) measured at 0.80 Pa Ar under equivalent process conditions at the same average power. The distributions have been normalized to fit into one single plot. Reprinted from Lundin et al.Reference Lundin, Larsson, Wallin, Lattemann, Brenning and Helmersson⁴⁹ with permission from the Institute of Physics.

Table I sums up a few important parameters for the HiPIMS process. They are not to be seen as exact values but rather a help to determine what type of discharge one is dealing with.

TABLE I.

A summary of important parameters for the high-power impulse magnetron sputtering discharge. The plasma parameters are obtained from the dense plasma region a few centimeters from the target, where the ionization is expected to be the strongest (ionization region).

Another important aspect of having an energetic flux of ions bombarding the substrate is the total energy flux, which affects the growing film. Furthermore, the heating due to energetic particle bombardment might melt thermally sensitive substrates, such as plastics. Comparative studies of the energy flux in HiPIMS and DCMS have shown that the total energy flux is of the same orderReference Leroy, Konstantinidis, Mahieu, Snyders and Depla⁵³ or lowerReference Lundin, Stahl, Kersten and Helmersson^54, Reference West, Kelly, Barker, Mishra and Bradley⁵⁵ in the case of HiPIMS sputtering. This difference can be explained by the lower deposition rate for HiPIMS resulting in fewer bombarding particles per time unit (see Section II D). As an example, taking the lower deposition rate into account in the case of Ti sputtering using HiPIMS and DCMS, it has been concluded that ∼90% more energy per incoming particle is deposited on the substrate in the HiPIMS case.Reference Lundin, Stahl, Kersten and Helmersson⁵⁴ Furthermore, from absolute temperature measurements at the substrate position during a HiPIMS process, it was found that the maximum equilibrium temperature reached about 70 °C.Reference Lundin, Stahl, Kersten and Helmersson⁵⁴ Although having an intense metal ion bombardment, this is well below the limit for several thermally sensitive substrates, which are thermally stable up to around 150 °C (Kapton).Reference Lundin, Stahl, Kersten and Helmersson⁵⁴ In this example, the average power was 500 W on a 6’’ Ti target and a peak current density of about 1.5 A cm⁻².

C. Gas dynamics

During the transport of both metal neutrals and ions, there is a certain probability that these particles will collide with the neutral gas background. These collisions lead to heating of the gas followed by expansion (decrease in gas density in front of the target) in a process that is called gas rarefaction. It has extensively been investigated in magnetron discharges both experimentally and theoreticallyReference Hoffman^56–Reference Kolev and Bogaerts⁶⁰ during the last three decades. The loss of process gas results in a reduction of ions available for sputtering (often Ar ions) leading to a reduced deposition rate as well as a reduction of plasma density, which means that the desired IPVD properties will be lost.Reference Lundin, Brenning, Jädernäs, Larsson, Wallin, Lattemann, Raadu and Helmersson²⁹

The reduction of gas density would not be much of a problem if the refill process would be fast enough. Unfortunately, this does not seem to be the case in all types of magnetron discharges,Reference Lundin, Brenning, Jädernäs, Larsson, Wallin, Lattemann, Raadu and Helmersson²⁹ and it is believed to be particularly serious in HiPIMS due to the high peak power densities,Reference Kadlec⁶¹ which is further investigated and discussed in more detail in Ref. Reference Lundin, Brenning, Jädernäs, Larsson, Wallin, Lattemann, Raadu and Helmersson29. Also, a fraction of reflected gas neutrals moving away from the target surface will collide with the neutral background and thereby enhance the gas rarefaction. Last, note that neutral gas atoms are also rapidly lost through direct ionization during the HiPIMS pulse.Reference Lundin, Brenning, Jädernäs, Larsson, Wallin, Lattemann, Raadu and Helmersson²⁹ This is a process, where gas ions arriving at the target become neutralized and reflected in the sputtering event, but are energetic enough to pass quickly through the near-cathode region and thus effectively drain the ionization region of process gas. To what extent it affects the overall gas depletion remains to be investigated. Conditions leading to loss of process gas will be discussed in Section II E regarding different pulse configurations.

D. Deposition rate

The lower deposition rate for HiPIMS compared to DCMS for the same average power is a drawback and is one of the most discussed topics in this field of research. In a recent report by Samuelsson et al.,Reference Samuelsson, Lundin, Jensen, Raadu, Gudmundsson and Helmersson⁸ it was found that the rates are typically in the range of 30–80% compared to DCMS depending on target material. In Fig. 4, reported deposition rates are given for various target materials based on the data from Helmersson et al.Reference Helmersson, Lattemann, Alami, Bohlmark, Ehiasarian and Gudmundsson⁶² and Samuelsson et al.Reference Samuelsson, Lundin, Jensen, Raadu, Gudmundsson and Helmersson⁸ The most common explanation to the reduction in deposition rate stems from a work by Christie,Reference Christie⁶³ where back-attraction of metal ions to the target followed by self-sputtering causes a reduction in the amount of sputtered particles reaching the substrate. Controlling and optimizing the potential profile in the cathode region will, therefore, greatly affect the number of metal ions incident on the target surface,Reference Poolcharuansin and Bradley⁶⁴ which so far has not been fully explored. Gas rarefaction, which was discussed in the previous section, is also likely to affect the deposition rate in HiPIMS processes. In addition, Konstantinidis et al.Reference Konstantinidis, Dauchot, Ganciu and Hecq⁶⁵ found that the amount of ions reaching the substrate is increased by increasing the plasma conductivity, which was seen by creating a secondary plasma between the magnetron and the substrate using an external inductive coil. Furthermore, Bugaev et al.Reference Bugaev, Koval, Sochugov and Zakharov⁶⁶ and Bohlmark et al.Reference Bohlmark, Östbye, Lattemann, Ljungcrantz, Rosell and Helmersson⁶⁷ showed that the magnetic confinement of the sputtered material affected the deposition rate. In a recent publication by Mishra et al.,Reference Mishra, Kelly and Bradley⁶⁸ it is reported that the deposition rate at the substrate position could be increased six times by weakening the magnetic field strength at the target surface by ∼33%. Emmerlich et al.Reference Emmerlich, Mráz, Snyders, Jiang and Schneider⁶⁹ highlighted the nonlinear energy dependence of the sputtering yield, meaning that it does not make sense to compare HiPIMS and DCMS deposition rates for the same average power, if not taking this dependence into account. When comparing with experimental results, they saw trends confirming their experiments, but it could not fully explain the differences. Another possibility is a recently discovered fast charged particle transport operating in pulsed plasma discharges such as HiPIMS.Reference Brenning, Merlino, Lundin, Raadu and Helmersson^70, Reference Lundin, Helmersson, Kirkpatrick, Rohde and Brenning⁷¹ It results in an increase of ions being transported parallel to the target surface and lost to the walls instead of arriving at the substrate position, and thus reducing the deposition rate.Reference Lundin, Larsson, Wallin, Lattemann, Brenning and Helmersson⁴⁹ Ongoing work on plasma modeling of HiPIMS discharges, where one can arbitrarily turn on and off mechanisms believed to affect the deposition rate, will most likely shed more light on the subject.Reference Brenning, Huo, Lundin, Raadu, Vitelaru, Stancu, Minea and Helmersson⁷² Last, it should be pointed out that there are many situations where the quality of the coating is far more important than the deposition rate and one should, as always, be careful when making these comparisons based on only one or a few properties.

FIG. 4.

Relative deposition rates for HiPIMS as compared to DCMS for various metal targets deposited at approximately the same average power. In the study by Samuelsson et al.,Reference Samuelsson, Lundin, Jensen, Raadu, Gudmundsson and Helmersson⁸ the depositions were carried out at 0.67 Pa Ar pressure at a HiPIMS peak current density of 1.5 A cm⁻². In the study by Helmersson et al.,Reference Helmersson, Lattemann, Alami, Bohlmark, Ehiasarian and Gudmundsson⁶² the process conditions could not be found (Data taken from Helmersson et al.Reference Helmersson, Lattemann, Alami, Bohlmark, Ehiasarian and Gudmundsson⁶² and Samuelsson et al.Reference Samuelsson, Lundin, Jensen, Raadu, Gudmundsson and Helmersson⁸ with permission from Elsevier).

E. Pulse configuration

Often the question regarding what type of HiPIMS pulse to use arises. One answer is that it depends on what one wants to achieve in the deposition process. Below is a short summary of what process conditions that should be expected for a given pulse-type. In all cases, we assume that the above definition of HiPIMS still holds, meaning that we still are working with a considerable peak power density compared to the average power.

• (1) t_pulse < 50 μs: In short pulses, Ar ions are the dominant sputtering particles.Reference Lundin, Brenning, Jädernäs, Larsson, Wallin, Lattemann, Raadu and Helmersson²⁹ Konstantinidis et al.Reference Konstantinidis, Dauchot, Ganciu, Ricard and Hecq²⁸ have shown that by decreasing the pulse length it is possible to increase the deposition rate, where ∼70% of the DCMS deposition rate is achieved for 5 μs pulses in the case of Ti compared to ∼20% of the DCMS deposition rate for 20 μs pulses, mainly due to less self-sputtering for short pulses. At the same time there is a tradeoff with the degree of ionization, where the highest reported values are found for the 20 μs pulses.Reference Konstantinidis, Dauchot, Ganciu, Ricard and Hecq²⁸ It is also reported that shorter pulses have beneficial consequences for reactive HiPIMS processes,Reference Aiempanakit, Kubart, Larssron, Sarakinos, Jensen and Helmersson⁷³ which are discussed in more detail in Section II F.

• (2) t_pulse ∼ 50–200 μs: The most commonly used pulse width is ∼80–100 μs, which is enough time to develop high peak currents leading to many of the desired properties discussed in this work. As the pulse is prolonged, it is more likely that a fraction of the ionized sputtered material is attracted back to the target surface and will participate in the sputtering process (self-sputtering).Reference Christie⁶³ This means that less sputtered material will reach the substrate, and the deposition rate is reduced.Reference Lazar, Vlcek and Rezek⁷⁴

• (3) t_pulse ∼ 200-500 μs: Here, gas heating leading to dissipation of the gas in front of the target may seriously affect the discharge conditions (see discussion in Section II C), which often leads to a dramatic drop of the discharge currentReference Lundin, Brenning, Jädernäs, Larsson, Wallin, Lattemann, Raadu and Helmersson^29, Reference Horwat and Anders⁷⁵ unless self-sputtering can compensate for the reduction of ions available for sputtering.

• (4) t_pulse > 500 μs: Longer pulsed regimes, where the discharge is sustained for several milliseconds, have been achieved when using the modulated pulsed power (MPP) technique, where the duty cycle can reach almost 30% compared to 1–10% for conventional HiPIMS discharges.Reference Lin, Moore, Sproul, Mishra, Rees, Wu, Chistyakov and Abraham⁷⁶ In general, this limits the peak current that can be achieved to avoid damaging the magnetron, which means that the peak plasma density is limited to around 10¹⁷–10¹⁸ m^-3,Reference Meng, Cloud, Jung and Ruzic⁷⁷ and the degree of metal ionization is found to be lower compared to shorter HiPIMS pulses.Reference Hála, Capek, Zabeida, Klemberg-Sapieha and Martinu⁷⁸ Still, the perspective of using multiple pulse packets within the same MPP pulse may result in facilitating the ignition of the plasma.Reference Hála, Capek, Zabeida, Klemberg-Sapieha and Martinu⁷⁸

In addition, the amplitude of the HiPIMS discharge current density is a good indicator for metal ionization, where an increasing discharge current density results in increased ionizationReference Alami, Sarakinos, Uslu and Wuttig³¹ and also an increased number of multiply charged ions.Reference Ross, Sangines, Treverrow, Bilek and McKenzie⁷⁹ On the other hand, the applied voltage determines the energy of the sputtering ion, meaning that an increased applied voltage will increase the ion energy and thereby increase the number of sputtered particles. It is important to keep in mind that the sputter yield roughly scales as ,Reference Emmerlich, Mráz, Snyders, Jiang and Schneider⁶⁹ where V _D is the applied discharge voltage, suggesting that the best choice to increase the deposition rate is not always to increase the applied voltage but, for example, increase the pulse frequency.

F. Reactive HiPIMS

In reactive sputtering processes, a reactive gas (e.g., O₂, N₂, etc) is used along with the buffer gas to synthesize compound films. Typical for these processes is that the formation of the compound material takes place on the surface of the target (referred to as target coverage or poisoning).Reference Berg and Nyberg^80, Reference Depla and De Gryse⁸¹ This layer is then sputtered and transported to the substrate to form the compound film. Deposition from a fully covered target (referred to as the compound sputtering mode) allows for growth of stoichiometric compound films, i.e., compound films with sufficient incorporation of the reactive gas atoms.Reference Berg and Nyberg^80, Reference Depla and De Gryse⁸¹ At these conditions, deposition rates lower than those obtained from an elemental (e.g., metallic) target are commonly achieved.Reference Berg and Nyberg^80, Reference Depla and De Gryse⁸¹ Growth of stoichiometric compound films with relatively high rates can be facilitated in the intermediate target coverage regime (referred to as transition zone) between the metallic and the compound mode.Reference Berg and Nyberg^80, Reference Depla and De Gryse⁸¹ In reactive DCMS, the transition sputtering zone is frequently unstable, and a hysteresis in the process parameters (i.e., target voltage, deposition rate, and reactive gas partial pressure) is often observed.Reference Berg and Nyberg^80, Reference Depla and De Gryse⁸¹ This is particularly pronounced during reactive DCMS of metal oxides.Reference Severin, Kappertz, Kubart, Nyberg, Berg, Pflug, Siemers and Wuttig⁸² As a consequence, stoichiometric films can only be obtained in the compound sputtering mode,Reference Berg and Nyberg^80, Reference Depla and De Gryse⁸¹ unless a feedback system for controlling the target coverage is used.Reference Sproul, Graham, Wong, Lopez, Li and Scholl⁸³ Investigations of the process characteristics during reactive HiPIMS of Al₂O₃,Reference Wallin and Helmersson^30, Reference Aiempanakit, Kubart, Larssron, Sarakinos, Jensen and Helmersson⁷³ ZrO₂,Reference Sarakinos, Alami, Klever and Wuttig⁸⁴ and CeO₂Reference Aiempanakit, Kubart, Larssron, Sarakinos, Jensen and Helmersson⁷³ have shown that these processes can exhibit a hysteresis-free and stable transition zone at deposition conditions which in the case of DCMS result in hysteresis and an unstable transition zone, see Fig. 5 for Al₂O₃. The stabilization of the transition zone allows for deposition of stoichiometric films at a lower target coverage when compared to the compound mode in DCMS.Reference Wallin and Helmersson^30, Reference Aiempanakit, Kubart, Larssron, Sarakinos, Jensen and Helmersson^73, Reference Sarakinos, Alami, Klever and Wuttig⁸⁴ This has been shown to result in deposition rates similarReference Wallin and Helmersson³⁰ or up to two times higherReference Sarakinos, Alami, Klever and Wuttig⁸⁴ than those obtained by DCMS. It is worth mentioning that all the studies on reactive HiPIMS have been performed in small-size laboratorial scale deposition systems. The ability of HiPIMS to stabilize the transition zone and suppress/eliminate the hysteresis effect for large industrial size systems has still to be demonstrated.

FIG. 5.

Deposition rates obtained from an Al target sputtered using DCMS and HiPIMS in an Ar–O₂ atmosphere as measured by a quartz crystal microbalance. The filled and hollow squares correspond to the DCMS rate for increasing and decreasing O₂ flow, respectively. It is seen that the DCMS process exhibits an unstable transition zone with pronounced hysteresis. On the contrary, the HiPIMS process (filled circles for both increasing and decreasing O₂ flow) is stable and hysteresis-free. This behavior allows for a deposition rate similar to that obtained by DCMS (Data taken from Wallin and HelmerssonReference Wallin and Helmersson³⁰).

In general, the stability of the transition zone in reactive sputtering processes is determined by the competition between the formation and the removal (sputtering) of the compound from the target surface.Reference Berg and Nyberg^80, Reference Depla and De Gryse⁸¹ According to the formalism developed by Berg and Nyberg,Reference Berg and Nyberg⁸⁰ the steady-state target coverage, ϑ_t,is equal to

(1)

where J _i is the ion target current density, Y _c the compound sputtering yield, q the elementary charge, a the sticking coefficient of the reactive gas, and F the flux of the reactive gas molecules toward the target. Stabilization of the transition mode can be achieved, if for a certain nominal partial pressure/flow of the reactive gas the term “” (removal of the compound) increases and/or the term “a2F(1 – ϑ_t)” (formation of the compound) decreases. It has been suggested that the pulsed character of the HiPIMS discharge affects both terms in Eq. (1). In addition, the absence of plasma during the pulse-off time results in a limited activation of the reactive species.Reference Depla and De Gryse⁸⁵ The expected life time of atomic species created during the HiPIMS discharge is ∼2 ms based on the experimental results by Clarenbach et al.Reference Clarenbach, Lorenz, Krämer and Sadeghi⁸⁶ Under these conditions, relatively high levels of reactive gas exposure are necessary for the formation of the compound on the target surface, i.e., this mechanism leads to lower effective values of the sticking coefficient a in Eq. (1). Furthermore, the high peak target current in HiPIMS results in rarefaction of the neutral species in the target’s vicinity, as discussed in Section II C, which affects not only Ar gas but also the reactive gas.Reference Sarakinos, Alami and Wuttig^87, Reference Kubart, Aiempanakit, Andersson, Nyberg, Berg and Helmersson⁸⁸ The rarefaction implies that the reactive species flux F is lower than the value corresponding to the nominal partial pressure of the reactive gas.Reference Sarakinos, Alami and Wuttig^87, Reference Kubart, Aiempanakit, Andersson, Nyberg, Berg and Helmersson⁸⁸

A. Deposition on complex-shaped substrates

The deposition of homogeneous films on substrates of complex geometry is a requirement for many technological applications, such as metallization of submicrometer patterns in optical and semiconductor devicesReference Weis, Müggenburg, Grosse, Herlitze, Friedrich and Wuttig^96, Reference Siemroth and Schülke⁹⁷ and deposition of thick protective layers on forming tools and turbine blades.Reference Alami, Bolz and Sarakinos^98, Reference Schuelke, Witke, Scheibe, Siemroth, Schultrich, Zimmer and Vetter⁹⁹ In conventional sputtering techniques such as DCMS the deposition flux is highly anisotropicReference Siemroth and Schülke^97, Reference Berisch¹⁰⁰ leading to inhomogeneous deposition, porosity, and poor coverage on substrate sites located along low flux directions.Reference Siemroth and Schülke⁹⁷ To alleviate these problems, highly ionized deposition fluxes can be used, since the trajectories of charged species can be manipulated by electric and magnetic fields. Different state-of-the-art attempts to achieve such conditions were discussed in Section II A.

HiPIMS provides an alternative approach to successfully deposit films on complex-shaped substrates. In Fig. 6, scanning electron microscopy (SEM) images of the crosssection of Ta films deposited both by DCMS and HiPIMS on a negatively biased (−50 V) Si substrate clamped on the side of a trench with an area of 1 cm² and a depth of 2 cm are presented.Reference Alami, Persson, Music, Gudmundsson, Bohlmark and Helmersson¹⁵ The SEM images were taken from a position at the middle of the trench as shown in the sketch in Fig. 6. The DCMS grown film exhibits a porous columnar structure with columns tilted from the normal of the Ta/Si interface [Fig. 6(b)], whereas the HiPIMS films are dense with columns growing perpendicularly to the Ta/Si interface [Fig. 6(a)]. This concept has been, for instance, used to deposit homogeneous hard nitride coatings on cutting inserts.Reference Alami, Bolz and Sarakinos⁹⁸ Smaller features, such as holes of several tens or hundreds of nanometers can also be successfully filledReference Chistyakov, Abraham, Sproul, Moore and Lin¹⁰¹ or homogeneously coated.Reference Kouznetsov, Macák, Schneider, Helmersson and Petrov⁴ It has to be pointed out here that all studies performed so far concerned deposition of metallic films on biased conductive substrates.

FIG. 6.

Cross-sectional scanning electron microscope (SEM) images of Ta films grown by (a) HiPIMS and (b) DCMS on a Si substrate clamped on the side of a trench with an area of 1 cm² and a depth of 2 cm. The HiPIMS deposited film is dense with columns growing perpendicular to the Ta/Si interface. The DCMS deposited films have a porous microstructure with columns inclined toward the flux direction. Reprinted from Alami et al.Reference Alami, Persson, Music, Gudmundsson, Bohlmark and Helmersson¹⁵ with permission from the American Institute of Physics.

B. Phase composition tailoring by HiPIMS

The phase composition of films is crucial for their mechanical, electrical, and optical performance. In sputtering processes, relatively low-growth temperatures and high deposition rates result in limited assembly kinetics.Reference Ohring^1, Reference Michely and Krug¹⁰² This in combination with extremely high cooling rates (10¹³ Ks^-1) during the condensation of the vapor on the substrate leads to nonequilibrium growth.Reference Ohring^1, Reference Michely and Krug¹⁰² Thus, the variation of both thermodynamic and kinetic conditions during deposition enables tailoring of the phase composition. Numerous studies have demonstrated that energetic ions can be used for this purpose since they can trigger surface and bulk diffusion processes,Reference Ohring^1, Reference Michely and Krug¹⁰² induce changes in the film structure and chemical composition,Reference Petrov, Adibi, Greene, Hultman and Sundgren¹⁰³ and cause generation of internal stresses.Reference Janssen and Kamminga^104–Reference Davis¹⁰⁷

The high fluxes of ionized material available in HiPIMS have been found to allow for control of the phase formation in both elemental and compound films. One example is the control of the phase composition in Ta films.Reference Alami, Eklund, Andersson, Lattemann, Wallin, Bohlmark, Persson and Helmersson¹² In this case, tailoring of the phase formation is achieved by controlling the magnitude of the internal stresses.Reference Alami, Eklund, Andersson, Lattemann, Wallin, Bohlmark, Persson and Helmersson¹² This process is particularly efficient in HiPIMS discharges owing to the degree of ionization of the Ta vapor of up to 70%.Reference Helmersson, Lattemann, Bohlmark, Ehiasarian and Gudmundsson⁵ Ta forms both a low resistivity body-centered cubic crystal structure (also known as the α phase) at elevated temperatures and a metastable high resistivity tetragonal phase (β-Ta) at room temperatures.Reference Clevenger, Mutscheller, Harper, Cabral and Barmak^108, Reference Face and Prober¹⁰⁹ The abundance of Ta⁺ ions in the deposition flux during the HiPIMS deposition of Ta films implies a more efficient momentum transfer to the growing surface.Reference Alami, Eklund, Andersson, Lattemann, Wallin, Bohlmark, Persson and Helmersson^12, Reference Dalla Torre, Gilmer, Windt, Kalyanaraman, Baumann, O’Sullivan, Sapjeta, Díaz de la Rubia and Djafari Rouhani⁹² This is in contrast to the growth by DCMS, where the majority of the ions in the deposition flux consists of the much lighter Ar⁺ ions.Reference Christou and Barber¹⁹ HiPIMS thus provides tools for better influencing the internal stresses of the growing Ta films and accordingly the possibility to deposit α-Ta films at room temperature (Fig. 7).Reference Alami, Eklund, Andersson, Lattemann, Wallin, Bohlmark, Persson and Helmersson¹²

FIG. 7.

X-ray diffraction patterns of Ta films deposited by HiPIMS on Si substrates biased at negative potentials of (a) −50, (b) −70 and (c) −90 V. The bcc α-Ta phase is obtained at −70 V due to the ion bombardment-induced compressive stress. The vertical lines indicate the position of the β-Ta (200) and α-Ta (110) peaks. Data taken from Alami et al.Reference Alami, Eklund, Andersson, Lattemann, Wallin, Bohlmark, Persson and Helmersson¹²

In HiPIMS discharges, part of the ions moves along off-normal, with respect to the target, directionsReference Lundin, Larsson, Wallin, Lattemann, Brenning and Helmersson⁴⁹ (see Section II D). This anomalous ion transport results in differences in the flux, the energy, and the composition of the deposited and bombarding species as functions of the deposition angle. Moreover, in the case of compound targets, the composition of the material flux along off-normal directions is largely determined by the ionization fraction of the target’s constituent elements.Reference Alami, Eklund, Emmerlich, Wilhelmsson, Jansson, Högberg, Hultman and Helmersson¹¹⁰ This composition is different from that along directions close to the target normal, which is also influenced by the angular distribution of the neutral species. The effect of the deposition angle on the phase composition has been studied for the ternary system Ti–Si–C,Reference Alami, Eklund, Emmerlich, Wilhelmsson, Jansson, Högberg, Hultman and Helmersson¹¹⁰ which can allow for the formation of the so-called MAX phases [M is a transition metal, A is an A-group element (mostly IIIA and IVA), and X is C or N].Reference Barsoum¹¹¹ The formation of MAX phases is strongly dependent on the relative fractions of the constituent elements, i.e., the chemical composition.Reference Barsoum¹¹¹ When a ternary Ti–Si–C target is sputtered, light elements like C are favored at the expense of heavier elements, such as Ti and Si, along the target normal.Reference Eklund, Beckers, Jansson, Högberg and Hultman¹¹² On the other hand, substrates placed at an angle of 90° with respect to the target experience a lower flux of C because of the lower ionization degree of C compared to Ti and Si, i.e., predominantly ballistic transport of C in the forward direction.Reference Alami, Eklund, Emmerlich, Wilhelmsson, Jansson, Högberg, Hultman and Helmersson¹¹⁰ Worth pointing out is that with increasing process gas pressure (>2 Pa), more pronounced scattering of light elements to off-normal directions will occur.Reference Alami, Eklund, Emmerlich, Wilhelmsson, Jansson, Högberg, Hultman and Helmersson¹¹⁰ In Fig. 8, the x-ray diffraction patterns of Ti–Si–C films grown from a Ti₂SiC₃ target employing HiPIMS are presented. The TiC phase is the main constituent for films grown on substrates parallel to the target surface [Fig. 8(a)]. At deposition angle of 90°, the so-called Nowotny’s Ti₅Si₃C_x phase is also formed, as shown in Fig. 8(b).

FIG. 8.

X-ray diffraction patterns of Ti–Si–C films deposited on MgO (111) substrates by HiPIMS on substrates oriented (a) parallel and (b) perpendicular with respect to the target surface. Peaks of the TiC_x and Ti₅Si₃C_x phases are indicated by arrows, respectively. S stands for substrate peaks. Data taken from Alami et al.Reference Alami, Eklund, Emmerlich, Wilhelmsson, Jansson, Högberg, Hultman and Helmersson¹¹⁰

Compound films can also be deposited from elemental targets when a reactive gas is added in the sputtering atmosphere. A typical example of a reactively synthesized material system in which HiPIMS can be used to control the phase composition is TiO₂.Reference Konstantinidis, Dauchot and Hecq^11, Reference Alami, Sarakinos, Uslu, Klever, Dukwen and Wuttig^113–Reference Bandorf, Vergöhl, Giesel, Wallendorf and Mark¹¹⁶ In general, TiO₂ films can grow in an amorphous and in two tetragonal crystalline structures; the rutile and the anatase phases, respectively.Reference Wiggins, Nelson and Aita¹¹⁷ The anatase phase exhibits interesting photocatalytic properties,Reference Wiggins, Nelson and Aita¹¹⁷ whereas the rutile phase exhibits one of the highest refractive indices in nature.Reference Wiggins, Nelson and Aita¹¹⁷ Deposition and/or annealing at high temperatures (700–900 °C) have been reported to lead to the formation of the rutile phase.Reference Wiggins, Nelson and Aita¹¹⁷ The anatase phase is generally obtained at lower temperatures,Reference Wiggins, Nelson and Aita¹¹⁷ whereas deposition at room temperature often leads to the formation of amorphous films.Reference Wiggins, Nelson and Aita¹¹⁷ In addition, the energetic bombardment by positively charged ions (facilitated by applying a negative bias voltage on the substrate) can promote the crystallization of TiO₂ films at room temperature.Reference Zhou, Zhong, Wu, Yuan, Shu, Li and Xia¹¹⁸ When HiPIMS is used, the rutile phase can be achieved even at room temperature, and the increase of the flux of energetic species toward the growing film favors the formation of the rutile at the expense of the anatase phase.Reference Konstantinidis, Dauchot and Hecq^11, Reference Alami, Sarakinos, Uslu, Klever, Dukwen and Wuttig^113–Reference Aiempanakit, Helmersson, Aijaz, Larsson, Magnusson, Jensen and Kubart¹¹⁵ This can be, for instance, achieved by increasing the peak target power (or current)Reference Aiempanakit, Helmersson, Aijaz, Larsson, Magnusson, Jensen and Kubart^115, Reference Bandorf, Vergöhl, Giesel, Wallendorf and Mark¹¹⁶ (see Fig. 9) or by decreasing the working pressure during deposition.Reference Alami, Sarakinos, Uslu, Klever, Dukwen and Wuttig^113, Reference Stranak, Quaas, Wulff, Hubicka, Wrehde, Tichy and Hippler¹¹⁴

FIG. 9.

X-ray diffraction patterns of TiO₂ films grown on Si substrates by HiPIMS at various values of peak target power. The increase of the peak target power and the subsequent resulting increase in the number of ions available during deposition promote the formation of the rutile at the expense of the anatase phase. Adapted from Aiempanakit et al.Reference Aiempanakit, Helmersson, Aijaz, Larsson, Magnusson, Jensen and Kubart¹¹⁵

C. Control of film microstructure and interface engineering

Polycrystalline films grown by PVD techniques exhibit a variety of microstructures with respect to the size, the morphology, and the relative orientation of the crystallites.Reference Petrov, Barna, Hultman and Greene^91, Reference Barna and Adamik¹¹⁹ These features have, for instance, implications for the mechanical strengthReference Koehler¹²⁰ and the electrical conductivityReference Tellier and Tosser^94, Reference Munoz, Finger, Arenas, Kremer and Moraga^95, Reference van Attekum, Woerlee, Verkade and Hoeben¹²¹ of the film. The microstructure is determined primarily by surface and bulk diffusion processes,Reference Petrov, Barna, Hultman and Greene^91, Reference Barna and Adamik¹¹⁹ which are controlled by the deposition temperature, the bombardment by energetic species, and the incorporation of impurities that act as inhibitors for the crystal and grain growth.Reference Petrov, Barna, Hultman and Greene^91, Reference Barna and Adamik^119, Reference Anders¹²² Deposition at relatively low temperatures (typically lower than 0.4T_m, where T_m is the melting temperature of the deposited material) using state-of-the-art sputtering techniques allows only for surface diffusion to be activated leading to the formation of films with a columnar microstructure and intercolumnar porosity [see Fig. 10(a) for CrN films].Reference Alami, Sarakinos, Uslu and Wuttig^31, Reference Petrov, Barna, Hultman and Greene^91, Reference Barna and Adamik¹¹⁹ Growth of films using HiPIMS is characterized by high ionic fluxes to the substrate (up to several hundreds of milliamperes per square centimeter) of relatively low energies (several tens of electronvolts), as was discussed in Section II B. These growth conditions trigger and/or enhance surface diffusion leading to film densificationReference Alami, Sarakinos, Uslu and Wuttig^31, Reference Greczynski, Jensen, Böhlmark and Hultman¹²³ as shown in Fig. 10(b). Upon increasing the flux of ions available at the substrate (achieved, e.g., by increasing the peak target current), repeated nucleation occursReference Alami, Sarakinos, Uslu and Wuttig^31, Reference Ehiasarian, Hovsepian, Hultman and Helmersson¹²⁴ resulting in suppression of the columnar structure and transition from a dense polycrystalline to a globular nanocrystalline microstructure,Reference Alami, Sarakinos, Uslu and Wuttig³¹ as shown in Figs. 10(c)–10(d). When high energetic fluxes are not available during growth, globular microstructures are consequence of segregation of impurity phases, which hinder the crystal and grain growth. It is, therefore, evident that the low-energy high-flux ion irradiation during HiPIMS can be used to overcome the characteristically underdense and rough microstructures and obtain morphologies unique for low-temperature sputter deposition.Reference Alami, Sarakinos, Uslu and Wuttig³¹ This, in turn, allows growth of films with higher hardness,Reference Paulitsch, Mayrhofer, Mitterer, Münz and Schenkel^13, Reference Chistyakov, Abraham, Sproul, Moore and Lin^101, Reference Ehiasarian, Hovsepian, Hultman and Helmersson¹²⁴ lower friction coefficient,Reference Paulitsch, Mayrhofer, Mitterer, Münz and Schenkel^13, Reference Ehiasarian, Münz, Hultman, Helmersson and Petrov¹²⁵ and improved scratch and wear as well as corrosion resistance,Reference Paulitsch, Mayrhofer, Mitterer, Münz and Schenkel^13, Reference Ehiasarian, Hovsepian, Hultman and Helmersson¹²⁴ as compared to films deposited by DCMS.

FIG. 10.

Cross-sectional SEM images of CrN films deposited on Si by (a) DCMS, as well as HiPIMS at a peak target current of (b) 44 A (1.0 A cm⁻²), (c) 74 A (1.7 A cm⁻²), and (d) 180 A (4.0 A cm⁻²). The increase of the peak target current results in a transition of a dense polycrystalline morphology to a nanocrystalline featureless one. The sketches next to each SEM image serve merely as a schematic representation for the reader’s convenience. Results taken from Alami et al.Reference Alami, Sarakinos, Uslu and Wuttig³¹

The highly ionized fluxes available in HiPIMS in combination with the use of a substrate bias result in ion energies in the order of several hundreds to thousands of electronvolts, which can be used to engineer the film–substrate interface and enhance the film adhesion.Reference Ehiasarian, Wen and Petrov¹⁴ Typical example is a CrN/NbN film deposited on steel substrates etched using HiPIMS (pretreatment by Nb⁺ ions at a substrate bias voltage of −1000 V) exhibited a scratch test critical load (L_c) of 56 N, which was higher than the values of 25 N obtained for films deposited on substrates etched by a DCMS plasma (substrate bias voltage of −1000 V at an Ar pressure of 0.8 Pa),Reference Ehiasarian, Wen and Petrov¹⁴ which exhibit a significantly lower ionization degree for both Ar⁺ and metal ions.Reference Christou and Barber¹⁹ It should also be pointed out here that the L_c values achieved on HiPIMS etched substrates are comparable to those for films grown on substrates cleaned using high fluxes of Ar⁺ ions generated by an external ionization source.Reference Vetter, Michler and Steuernagel¹²⁶ This fact indicates that HiPIMS can be used as an alternative process to improve the quality of the film–substrate interface and enhance the film adhesion, when no external source for increasing the Ar ionization is available.