Towards lowering energy consumption during magnetron sputtering: Benefits of high-mass metal ion irradiation Open Access

Thin film growth by magnetron sputtering is an essential technology in modern materials processing with numerous application areas and over 200 years long history reviewed by Greene in 2017.¹ The corresponding techniques are crucial for sustainable development as vast majority of produced coatings are used in applications motivated by minimizing energy consumption (e.g., low-friction coatings for fuel-injections and roller bearings),^2–6 material waste (e.g., wear protective coatings on cutting tools and corrosion- and oxidation-resistant coatings),^7–9 or in applications related to green energy production (solar cell panels, fuel cells, etc.).^10–12 In addition, magnetron sputtering is clean and dry. Unlike the case for other coating methods such as electroplating or chemical vapor deposition, no hazardous materials are involved, and no waste chemicals are generated. Thus, being more environmentally friendly at onset, magnetron sputtering is the technique of choice to meet most important technological requirements necessary for green transition.

In view of the above, it is essential to optimize surface engineering based on magnetron sputtering processes to make them energy- and resource-efficient. In this Perspective article, we focus on the energy-saving technical aspects, although some comments on resource-saving options are also made in the Outlook section.

Figure 1 shows a schematic illustration of all essential steps in the typical coating production cycle together with plots of power consumption and substrate temperature. The most energy-consuming steps are heating, etching, and coating.¹³ The former serves a double purpose: first to ensure that a clean deposition condition is achieved by getting rid of water adsorbents inside the vacuum vessel. At higher surface temperatures, the water desorption rate increases, which, together with effective pumping (turbo pumps often operating at full speed during this phase) results in lower background pressure prior to substrate etching and film deposition. Consequently, C, O, and H contaminations in deposited coatings are reduced to sub-at. % levels. The second reason for extensive heating is to ensure high substrate temperature, typically 400–600 °C, during the coating phase. This provides sufficient mobility for atoms landing on the surface to fill empty lattice sites such that dense layers can be obtained.¹⁴ Films grown with no external heating are typically columnar and underdense, which results in poor mechanical and optical properties as well as high resistivity.¹⁵ While such extensive heating is commonly used, it obviously limits the range of possible applications, as heat-sensitive substrates cannot be coated. Many modern engineering materials such as polymers, tempered steel, or lightweight Al- and Mg-based alloys are, thus, excluded. The process energy fraction consumed during the heating phase is relatively large as (i) the resistive heaters are typically used (rather slow and ineffective heat transfer), (ii) apart from the load also the entire vacuum vessel has to be heated up, and (iii) the phase is relatively long.

The etching step, essential for improving film/substrate adhesion,^16–19 traditionally employs gas ions; however, variants involving metal ions from the cathodic arc²⁰ or high-power impulse magnetron sputtering (HiPIMS)²¹ have been shown to be more advantageous. Energy consumption at this stage is due to the generation of etching plasma and accelerating ions.

Energy consumption is typically highest during the coating phase, as in addition to resistive heaters, also the sputtering cathodes need to be powered up. Power applied to heaters can be somewhat lowered with respect to that used during the heating phase as one takes advantage of plasma heating (i.e., energy carried by the flux of condensing atoms, ionized species, and electrons, see Sec. IV for more details). Nevertheless, it is crucial for the growth of high-quality layers that the surface temperature is sufficiently high. Typically, a homologous temperature $T h= T s/ T m>0.3$ is required to ensure sufficient adatom mobility (T_s and T_m: growth and melting temperatures in K), which corresponds to T_s> 800 °C for TiN.^22–24 The classic solution in direct current magnetron sputtering (DCMS), for which sputter-metal atom ionization is ≤1%,²⁵ is to employ gas-ion (e.g., Ar⁺ and N₂⁺) bombardment to densify the films.²⁶ This allows for lowering T_s to 400–500 °C (i.e., T_h > 0.21–0.24) at the expense of higher compressive stress due to significant concentrations of trapped gas atoms,²⁷ which is highly undesired as it results in cohesive film failure and delamination.

According to the case study of TiN grown on cemented carbide substrates in a four-cathode setup (twofold rotation) by pulsed DCMS, ∼90% of the process energy is consumed by resistive heaters and cathodes.¹³ While the latter is hard to be reduced without affecting process economics, the former one presents an opportunity for energy savings as outlined below.

The heating step that precedes the coating phase serves two purposes: it ensures cleaner growth conditions and provides high adatom mobility. It is, however, important to realize that the temperature requirements for these two differ substantially. While for the latter, T_s of 400–500 °C is necessary, the acceptable residual pressure can be achieved provided that the temperature of the vacuum chamber is kept above 100 °C for reasonably long time. This is illustrated in Fig. 2(a), which shows the changes in background pressure p_bckg for two heating scenarios. In the first case (conventional approach shown in red color), the full power of 17.6 kW is applied to resistive heaters, while in the second case (black curve) the reduced power of 4 kW is applied only during the first 60 min of the 2h long sequence. The resulting substrate temperature T_s, monitored with a calibrated²⁸ thermocouple attached to a sacrificial dummy substrate holder placed next to the sample, is plotted for both heating scenarios in Fig. 2(b). For conventional heating, p_bckg, which is determined by the water vapor desorption rate from all surfaces inside the vacuum chamber and the pumping capacity, shows initially a spike due to the sudden release of H₂O molecules adsorbed on heaters. After that, the pressure drops for a few minutes but then increases again as the outgassing from chamber walls is higher than the pumping capacity. This initial phase is completed within first 10 min, and eventually p_bckg exhibits monotonic decrease to reach 1 mPa after 2h of heating. Substrate temperature shows a steady increase during the entire heating phase exceeding 100 °C after 15 min and eventually reaching 450 °C. In the second scenario, the substrate temperature obviously increases with lower rate due to lower heating power used. Nevertheless, T_s reaches 100 °C after 50 min and stays above that for the remaining part of the heating sequence (reaching the maximum of 150 °C after ∼90 min.) even though the heating is turned off after 60 min. Lower heating power results also in lower H₂O desorption rate from all surfaces; therefore, the background pressure is 2–3 times lower during the entire heating step. At the end of the 2h long sequence, the substrate temperature is ∼120 °C and p_bckg is only 0.3 mPa, which is not only acceptable, but more than three times better than for the conventional process. The concern that arises at this point is—will the plasma heating during the deposition phase lead to additional temperature increase and a following (uncontrolled) outgassing that would result in higher contamination levels in films deposited with the second heating scenario? Experiments described in the following sections prove that this is not the case even if the full cathode power is used. It, however, eventually depends on the specific setup used for film growth. Hence, the primary challenge for energy-efficient film growth is to find substitutes for thermally driven adatom diffusion while controlling residual gas contamination.

A possible route forward was proposed in 2014 and is based on the use of a high mass $( m M e + ≈ 180 amu)$ metal ions Me⁺ generated by the magnetron operating in the HiPIMS mode in the absence of intentional substrate heating.²⁹ The incident ion energy is controlled by varying the amplitude of the pulsed negative substrate bias V_s pulses that are synchronized to the Me⁺-rich portions of ion fluxes at the substrate using the input from time-resolved ion mass spectrometry analyses performed at the substrate position (see Sec. IV).^30–32 The basic concept relies on the notion schematically illustrated in Fig. 3(a). The primary Me⁺ ions with the mass significantly larger than the average mass of film constituents, $m M e +≫ m f$ (here, we assume low Me concentrations such that m_f is determined by other film components), scatter at small angles and, thus, penetrate deeper into the near-surface region to dynamically fill occasional vacancies, which result from very-low-temperature growth. Another consequence of the large mass difference between the incident ion and film constituents is the generation of a large number of low-energy recoils as heavy Me⁺ ions cannot lose energy and momentum in just few collisions.³³ For example, assuming elastic head-on collisions, the energy transfer function, $4 m f m M e +/ ( m f + m M e + ) 2$⁠, for W⁺ ion (m_W = 183.84 amu) impinging onto the Ti_0.50Al_0.50N surface is 0.66, 0.45, and 0.26 for W⁺ → Ti (m_Ti = 47.87 amu), W⁺ → Al (m_Al = 26.98 amu), and W⁺ → N (m_N = 14.00 amu), respectively. The corresponding numbers for the momentum transfer function, $2 m f/( m f + m M e +)$⁠, are 0.42, 0.26, and 0.14. Both effects provide adatom mobility in the near-surface layer with the thickness that is determined by the incident Me⁺ ion energy $E M e +$⁠, which can be easily controlled by varying the V_s amplitude.

Another case illustrated in Fig. 3(b) is when the mass of the incident metal ions is similar to that of film constituents $m M e +≈ m f$⁠. Now both energy and momentum transfer to the film constituents are more efficient resulting in fewer collisions, thus producing, on average, lower number of higher energy recoils per incident ion. In the third scenario sketched in Fig. 3(c), $m M e +≪ m f$ leads to ion reflection with minimal energy and momentum transfer to the growing film.

A practical implementation of the concept illustrated in Fig. 3(a) was demonstrated in the hybrid HiPIMS/DCMS co-sputtering setup [see Fig. 4(a)] for the growth of transition metal nitride films in Ar/N₂ gas mixtures (for details, see Sec. V).²⁹ The primary metal targets Me₁ operate in the DCMS regime to provide predominantly neutral fluxes as the ionization of the sputter-ejected atoms is low during DCMS.²⁵ Consequently, the deposition rate is high as ion back attraction to the target, characteristic of HiPIMS processing, is avoided.³⁴ A high-mass target Me₂ powered by HiPIMS supplies predominantly energetic Me₂⁺ ion irradiation to densify the film deposited from the DCMS sources. The substrate temperature T_s is kept low as no intentional substrate heating is applied. The energy efficiency of the entire process is increased significantly since the majority of the supplied energy is consumed by the sputtering cathodes rather than being wasted for heating the entire vacuum vessel as is the case of conventional processing.

In magnetron sputtering, the sheaths around negatively bias surfaces are collisionless; the average energy E_i and momentum p_i of ions incident at the growing film can be expressed as^26,35 $E i= E i o+ne( V s − V p l)$ and $p i= p i o+ 2 n e m M e + ( V s − V p l )$⁠, in which $E i o$ and $p i o$ denote the average energy and momentum of ions entering the anode sheath, n accounts for the charge state of the ion, V_s is the amplitude of the applied substrate potential, and V_pl is the plasma potential (typically ±5 V).³⁶ Since $E i o$⁠, $p i o$⁠, and V_pl are constant for the given processing conditions, the metal-ion energy and momentum can be controlled in a wide range by varying V_s.

Importantly, during the DCMS phase of the hybrid HiPIMS/DCMS growth (i.e., between HiPIMS pulses), the substrate is intentionally kept at the floating potential to minimize the role of gas ions that arrive with energies lower than the lattice displacement threshold (∼20–50 eV, depending upon the ion and film species involved).³⁷ Thus, the impact of gas ion irradiation on film nanostructure evolution is minimized and so is the trapping of gas atoms at interstitial lattice sites. In this way, the buildup of high compressive stress levels is avoided.^38,39 The substrate heating due to ion bombardment is not significant for two reasons: (1) the fluxes of high mass metal ions are low in comparison to fluxes of neutrals from DCMS sources, and (2) the substrate bias duty cycle is very low and typically does not exceed 2%.

An additional benefit of the low-T_s growth concept presented here is that the cooling step that typically follows the coating phase (cf. Fig. 1) can be shortened. The main purpose of cooling is to ensure that the freshly deposited films are not exposed to air at too high temperature that could result in extensive oxidation.⁴⁰ Obviously, this is not of a concern if the growth temperatures do not exceed 150 °C; hence, the cooling cycle can be significantly shortened, which improves process economy.

The film microstructure, and in particular nanostructure evolution, is often described using the nomenclature of the structure zone models (SZM), which use homologous temperature T_h as a universal parameter for activation of surface and bulk diffusion processes.^14,41–46 These models apply for films deposited on non-matching, smooth substrates, that is, excluding epitaxial growth and rough or patterned substrates. Over the decades, most of the attention has been devoted to deposit dense films with bulk-like properties at low substrate temperatures T_h < 0.2.

The first SZM model was proposed by Movchan and Demchishin (MD) in 1969.⁴¹ MD found that the microstructures of electron-beam evaporated coatings of Ti, Ni, W, ZrO₂, and Al₂O₃ could be represented schematically in three zones, each with its own characteristic structure and physical properties. The low temperature (T_h < 0.3) Zone 1 structure is columnar with both inter- and intra-columnar voids and rough surface. Bulk and surface diffusion are suppressed; adatom stick where they land; self-shadowing applies. The Zone 2 structure (0.3 < T_h < 0.5) consists of columnar grains with dense boundaries; the films have bulk-like properties. The high temperature Zone 3 (T_h > 0.5 structure) featured of equiaxed grains with more or less flat surfaces due to surface tension minimization.

Barna and Adamik⁴² (BA) amended the model for evaporated metal films by inserting a transition Zone T between Zone 1 and Zone 2. In Zone T, 0.2 < T_h < 0.4, the structure is inhomogeneous along the film thickness. It is fine-crystalline and dense near the substrate, followed by competitive column growth with surface diffusion active to sustain local epitaxial growth on individual columns. Column tops become pointed due to kinetic roughening, which combined with inter-grain shading leads to open column boundaries. Zone 2 is similar to the one described by MD with columns extending from the bottom to the top of the film; however, in BA model, Zone 2 extends to T_m. The equiaxed structure of Zone 3 in the MD model is ascribed to impurities and excluded from the SZM for pure metals and compounds.⁴⁷ 

For pure metal films deposited by magnetron sputtering, Thornton¹⁴ proposed a modified MD model by adding a pressure axis that later he⁴⁴ interpreted as energy axis to account for energy provided by fast particles: that is the modification of the microstructure at low pressures is a consequence of primarily the bombarding effect of working gas ions that are neutralized and reflected at the cathode surface as well as the superthermal energy of the sputtered particles. Later other researchers^44–46 ascribed energy of gas ion bombardment to the pressure axis. Energetic particle bombardment densifies the films of Zone 1, thus creating a Transition Zone T extending down to T_h ∼ 0.1 at the low-pressure range of 1 mTorr. Thornton's Zone T microstructure is defined as “a dense array of poorly defined fibrous grains” and “the Zone T structure is therefore a Zone 1 structure with crystal sizes that are difficult to resolve and appear fibrous and with boundaries that are sufficiently dense to yield respectable material properties.” The Zone T coatings are characterized by high optical reflectance, low resistivities, and compressive internal stresses, and often contain considerable concentrations of entrapped working gas.

The film microstructure of Thornton’ s Zone T is uniform along the thickness of the film and is, thus, quite different from the Barna's Zone T microstructure that evolves from fine-grain, dense near the substrate toward large columnar grains with faceted tops and open grain boundaries as the film thickness increases.

Anders⁴⁸ extended Thornton's model to include explicitly the effects of gas ion-irradiation during bias sputtering. The linear pressure (thermalization) axis is replaced by a logarithmic energy axis while the unlabeled thickness axis is extended to include resputtering and etching effects. The temperature axis is modified to include condensation and plasma heating. Zone T in Anders' model is identical to Thornton's Zone T. The boundary Zone 1/Zone T again is shown to move to lower temperature as the energetic bombardment increases and reaches below 0.2 at energies that provide a considerable resputtering, represented by reduced films’ thickness. It is also noted in the diagram that this is accompanied by compressive stresses due to inert gas incorporation.

Anders does mention that the mass ratio of the ion to the films constitutes will play a role, but it is not elaborated how.

The approach described in this Perspective specifically exploits the use of heavy metal ions to create a large number of low-energy recoils from the lighter main constituents of the films in order to create dense layers. The heavy ions are incorporated into the film matrix, thus eliminating excessive compressive strains. Densification is achieved with compressive stress ≤1 GPa. The microstructure that we report is clearly different from Thornton/Anders Zone T, which lacks local epitaxial growth. It consists of densely packed columns with local epitaxial growth on individual column. The columns tops, however, are relatively flat, i.e., giving evidence for reduced or eliminated kinetic roughening and inter-grain shading and pronounced V-shape column growth. Thus, it is different from Barna's Zone T. The resulting microstructure would be best described as expanding Zone 2 structure down to T_h∼ 0.12.

The low-energy metal-ion-assisted deposition effectively expands the applicable Zone 2 structure toward lower temperatures by virtue of recoil implantation without much heating of the film (since fractions of high mass ions are relatively low compared to neutrals) and without causing excessive residual stress (due to bias synchronization). This approach has been successfully used to grow dense and hard Ti_1−x−yAl_xTa_yN,⁴⁹ Ti_1−xTa_xN,²⁹ or Ti_0.40Al_0.27W_0.33N,⁵⁰ films with Ta⁺ or W⁺ metal ions and at the substrate temperatures not exceeding 150 °C. Examples are discussed in Secs. V–VIII. However, before going into details of low-temperature film growth, analyses of ion fluxes incident at the growing film surface need to be introduced.

In order to establish the causal relationship between plasma parameters and properties of deposited layers, data for the type, flux, and energy of ions incident at the growing film surface are required.^32,51–53 In the case of HiPIMS/DCMS processing, the time evolution of metal- and gas-ion fluxes at the substrate is of primary interest,^54–56 as it allows for selective tuning of metal-ion energy and momentum by applying the substrate bias pulses only during the time when ion flux incident at the film surface is dominated by metal ions (metal-ion-synchronized HiPIMS).³⁰ 

Figure 5 shows examples of time-dependent ion energy distribution functions (IEDFs) recorded in 50 μs intervals (0 μs indicates the onset of the HiPIMS cathode pulse) at the substrate position during HiPIMS of (a) Ti and (b) W targets (HiPIMS pulse length: 120 μs, frequency: 300 Hz, and peak target current density: 1 A/cm²). Experiments were conducted in Ar/N₂ gas mixtures (Ar/N₂ flow ratio of 0.11 and the total pressure of 3 mTorr) with a Hiden Analytical EQP1000 instrument in a CemeCon CC800/9 magnetron sputtering system equipped with rectangular 8.8 × 50 cm² targets. Details are given in Ref. 32.

Ti⁺ IEDFs recorded during Ti-HiPIMS in the time interval 0–100 μs resemble Sigmund–Thompson energy distributions^57,58 with a peak of several electron volts and energy tail extending to dozens of eV. This indicates that a significant fraction of sputter-ejected Ti atoms, ionized while passing through the dense plasma region,⁵⁹ does not undergo inelastic collisions on the way to the detector. At later times, 150–300 μs, IEDFs collapse into the low-energy peak since vast majority of Ti⁺ ions that arrive so late after the HiPIMS pulse became thermalized, i.e., lose essentially all initial energy in numerous inelastic collisions with Ar and N₂ molecules.

Different evolution of IEDFs is observed for W⁺ ions during W-HiPIMS: the high-energy tail is preserved even after 400 μs from pulse ignition, and the only effect is the decrease in the signal intensity. Such complete lack of the low energy peak indicates that W⁺ thermalization does not occur. This is explained by extensive gas rarefaction^60–62 caused by high temporal fluxes of sputtered W atoms. Two factors are crucial. First, W sputtering yield is higher than that of Ti: S_Ar→Me = 1.45 for W and 0.62 for Ti for $E A r +=750 and480 eV$⁠, respectively (W target operates at significantly higher voltage to reach similar peak target current, hence correspondingly higher $E A r +$ value used in this comparison). The self-sputtering yield is also higher for W target: $S M e → M e=1.02 vs0.51$ for Ti target (calculated for the same ion energies as above). Second, reactivity toward N₂ is lower (nitride formation enthalpy, $Δ H f 0=−0.2 eV/ atom$ for W vs −3.4 eV/atom for Ti).⁶³ Both factors contribute to a much higher flux of sputter-ejected species during W-HiPIMS than that encountered during Ti-HiPIMS (under similar conditions of peak current, pressure, etc.). Thus, momentum transfer to the gas atoms upon collisions is higher in the W case resulting in more pronounced gas rarefaction.^64–67 

IEDFs of the type shown in Fig. 5 can be recorded with higher time resolution of, e.g., 10 μs, integrated over the entire ion energy range and plotted as a function of time. If performed for all ion types, such plots allow us to determine time intervals with the highest metal-ion/gas-ion flux ratio. Based on this essential input, the length and the offset of synchronized bias pulses is selected to ensure that the electric field acts predominantly on metal ions, which are then responsible for the film nanostructure modification. This principle is not always straightforward to implement as illustrated in Fig. 6 for the most abundant energy-integrated ion fluxes detected at the substrate plane during HiPIMS of group IVB (Ti, Zr, and Hf) and VIB (Cr, Mo, and W) targets in Ar/N₂ atmosphere.³² Clearly, for the IVB targets, gas ion fluxes are very intense, which makes the selective control of metal-ions essentially impossible, and the best option is to find a compromise where the gas ion contribution is minimized. For example, in the case of the Ti target, the bias pulses from 40 to 100 μs can be used to maximize the Ti⁺ effects on film properties, while avoiding high fluxes of Ar⁺ ions as they result in trapped Ar interstitials leading to higher residual film stress^68,69 with risk of adhesive and/or cohesive film failure.^70,71 The overlap between Ti⁺ and N⁺ ions [cf. Fig. 6(a)] is not an issue as the latter are film-forming species.

Group VIB targets render an entirely different situation, where, except for Cr, all gas ion fluxes are very low resulting in large flexibility in the choice of substrate bias length and offset. Higher metal-ion/gas-ion fractions for VIB targets (if comparing elements with similar mass) are caused by more severe gas rarefaction effects due to higher temporal fluxes of sputter-ejected species. The latter takes place because VIB metals have higher sputter yields and lower reactivity toward N₂ (resulting in lower rate of nitride formation). The gas rarefaction is also responsible for the increase in the metal-ion/gas-ion ratio with increasing metal ion mass $m M e +$⁠, which is observed for both IVB and VIB transition metal targets. The high metal-ion/gas-ion ratios found for HiPIMS of group VIB transition metals are the main motivations for selecting these elements for systematic studies discussed in Sec. VI.

The first report on the effects of high-mass metal-ion irradiation during growth with no external substrate heating²⁹ utilized Ta⁺ (m_Ta = 181.0 amu) ions from a HiPIMS source to densify TiN films deposited from a second magnetron operating in DCMS mode [Ta-HiPIMS/Ti-DCMS, with reference to Fig. 4(a)—only the upper DCMS magnetron is used with Me₁ = Ti and Me₂ = Ta]. The time-resolved ion mass spectrometry analyses performed at the substrate plane revealed that (1) the metal ion fluxes contain both Ta⁺ and Ta²⁺ ions with the doubly ionized fraction Ta²⁺/(Ta²⁺+Ta⁺) = 0.36, and (2) the Ta⁺/Ta²⁺-rich phase lasts from 80 to 280 μs from the ignition of the HiPIMS pulse. The metal-ion-synchronized substrate bias V_s was applied accordingly, with the amplitude varying from 20 to 280 V to probe effects of Ta⁺/Ta²⁺ energy and momentum on film properties. To achieve low background pressure prior to film growth, a low power of 0.5 kW was applied to resistive heaters for 1h, resulting in a chamber temperature of 110 °C at the substrate position, while in the next hour, the power was switched off, resulting in T_s of 65 °C and a background pressure as low as 3.8 × 10⁻⁷ Torr (0.05 mPa).

Figures 7(a) and 7(b) show cross-sectional SEM (XSEM) images from TiN and a Ti_0.92Ta_0.08N layer, respectively. The former film was grown with the substrate electrically floating to serve as a reference to elucidate effects of Ta⁺/Ta²⁺ bombardment, while the metal-ion-synchronized substrate bias V_s = 160 V was used during Ta-HiPIMS/Ti-DCMS alloy growth. Importantly, in both cases, the substrate temperature T_s did not exceed 120 °C (T_h < 0.12), as no heating was used during coating, and the bias duty cycle was very low (2% only). Low adatom mobility due to such low $T h$^15,72 results in porous TiN films as evidenced by a pronounced network of inter- and intra-columnar voids. Surfaces are rough due to kinetic roughening,^73–75 exacerbated by atomic shadowing.⁷⁶ The structure corresponds to Zone 1 in the Thornton's SZM diagram,⁴⁴ characterized by low average adatom mobility and essentially no bulk diffusivity. In distinct contrast, the inter- and intra-columnar porosity disappears once pulsed Ta⁺/Ta²⁺ ion irradiation is employed for Ti_0.92Ta_0.08N growth. The columns are wider (on average 200 vs 100 nm for TiN), the surface is smoother, and the growth is typical for the Zone 2 structure.

To illustrate sharp transition from an open Zone 1 to dense Zone 2 structure, Fig. 8 presents a bright-field XTEM image of a trilayer (DCMS TiN)/(Ta-HIPIMS/Ti-DCMS Ti_0.92Ta_0.08N)/(DCMS TiN) sample grown on a Si(001) wafer without external heating. The Ti_0.92Ta_0.08N alloy layer is deposited with a metal-ion-synchronized substrate bias V_s = 160 V, while DCMS TiN layers are grown with the substrate electrically floating. The DCMS TiN underlayer has a columnar nanostructure and pronounced intercolumnar voids as well as an intracolumnar network of voids [see Fig. 8(b)]. This is a signature of Zone 1 microstructure with low adatom surface and bulk mobilities.

The intracolumnar voids disappear almost instantaneously upon switching on pulsed Ta⁺/Ta²⁺ metal-ion irradiation-induced atomic mixing. However, the intercolumnar voids persist for 500–1000 Å while the pointed column tops are smoothened by ion-bombardment. Figure 8(c) shows that the tops of the Ti_0.92Ta_0.08N alloy layer columns are rounded with dense column boundaries. Thus, this layer microstructure can be characterized as Zone 2 structure. Figure 8(c) also illustrates that the intercolumnar porosity reemerges immediately upon discontinuing metal ion irradiation, i.e., the transition from Zone 2 to Zone 1 structure is abrupt.

All these positive effects of irradiation-induced atomic mixing are achieved with low Ta⁺/Ta²⁺ fluxes (relative to the Ti flux from the DCMS cathode) as evident from the low Ta content in the films of 4 at. %.

To prove that the densification effect is solely due to Ta⁺/Ta²⁺ irradiation and is not caused by alloying with Ta, complementary experiments were performed, using TiAlN as a base layer [referring to Fig. 4(a)—the Al target was mounted on the upper DCMS magnetron, Ti target on the bottom DCMS magnetron and Me₂ = Ta].⁴⁹ Ta target was operated either in the DCMS or in the HiPIMS mode to test the effects of Ta ionization degree on the density of resulting Ti_0.41Al_0.51Ta_0.08N layers. In order to maintain a composition, the Ta target power was increased in the latter case to compensate for a deposition rate loss caused by the back attraction of Ta⁺ and Ta²⁺ ions as observed during HiPIMS.^34,77 Metal-ion-synchronized bias V_s = 160 V was used during deposition involving Ta-HiPIMS, while substrate was floating for purely DCMS growth. T_s reached a maximum 150 °C at the end of the deposition sequence due to plasma heating. Figures 7(c) and 7(d) are fracture XSEM images from Ti_0.41Al_0.51Ta_0.08N films grown with the Ta target operated in the DCMS (Ta only) or in the HiPIMS mode (high Ta⁺/Ta²⁺ fraction). The densification effects of Ta⁺/Ta²⁺ irradiation are very apparent and of the same quality and extent as discussed for TiN and Ti_0.92Ta_0.08N above.

The effects of Ta⁺/Ta²⁺ ion energy and momentum transfer to the growing film surface are illustrated in Fig. 9, where (a) nanoindentation hardness H, (b) elastic modulus E, and (c) residual stress σ are plotted as a function of Ta⁺-synchronized V_s for Ti_0.92Ta_0.08N layers grown by Ta-HiPIMS/Ti-DCMS at T_s < 120 °C. In addition, results for several reference samples grown by DCMS (thus with no Ta⁺ irradiation) are shown: (1) TiN grown at T_s = 500 °C with V_s = 60 V, (2) TiN grown at T_s < 120 °C and at a floating bias of −10 V, and (3) Ti_0.92Ta_0.08N grown at T_s < 120 °C and at a floating bias of −10 V.

As expected from the XSEM image in Fig. 7(a), highly porous TiN layers deposited with no Ta⁺/Ta²⁺ bombardment at T_s = 120 °C are soft, H = 7.8 GPa, and exhibit relatively low E of 248 GPa. Essentially, the same poor mechanical properties are measured for the reference Ti_0.92Ta_0.08N film grown by DCMS (H = 9.8 GPa, E = 286 GPa), which proves that alloying TiN with such low Ta concentrations has very limited effect. What makes much more difference, however, is the energy and momentum transfer due to Ta⁺ irradiation—both H and E of Ti_0.92Ta_0.08N layers grown by Ta-HiPIMS/Ti-DCMS exhibit strong dependence on the V_s amplitude. Hardness increases from 14.4 GPa with V_s = 20 V to ∼26 GPa for V_s ≥ 160 V, while the corresponding increase in elastic modulus is from 346 to ∼500 GPa. This pronounced improvement in mechanical properties is directly connected to Ta⁺/Ta²⁺-induced densification, which scales with increasing Ta⁺/Ta²⁺ energy and momentum as both the projected ion range and the number of recoils per incident ion increase. Provided that the Ta⁺/Ta²⁺ energy is high enough, both inter- and intracolumn porosity is eliminated due to effective near-surface atomic mixing. In fact, the results in Figs. 9(a) and 9(b) are in excellent agreement with Monte Carlo TRIM^78,79 simulations of ion/surface collision cascades.²⁹ 

It is important to note that the excellent mechanical properties of Ti_0.92Ta_0.08N films grown with V_s ≥ 160 V that result from dense nanostructure [cf. Fig. 7(b)] are in fact better than those of the reference TiN film grown by DCMS at 500 °C (H = 19.4 GPa, E = 509 GPa). Actually, quantification by x-ray reflectivity using a stoichiometric, single-crystal TiN/MgO(001) reference sample⁸⁰ showed that DCMS TiN layers grown at T_s < 120 °C have a relative density ρ_r of 65% and for the corresponding TiN film deposited at 500 °C ρ_r = 92%. The density of Ti_0.92Ta_0.08N films grown with V_s = 20 and 160 V is 76% and 98%, respectively.²⁹ 

It is well known that the intense ion bombardment of a growing film surface often leads to the buildup of high compressive stress, especially if the volume density of residual defects and incorporated Ar is high.^15,27 This aspect is addressed in Fig. 9(c), in which residual stress values σ obtained from sin²ψ analyses (corrected for the tensile differential thermal contraction stress) are plotted as a function of V_s. Porous TiN reference films grown by DCMS at T_s < 120 °C are essentially stress-free, σ = −0.1 GPa, while TiN layers deposited at 500 °C with DC bias of 60 V exhibit a compressive stress of −1.3 GPa due to Ar incorporation as a continuous Ar⁺ ion bombardment is present during growth. For the Ti_0.92Ta_0.08N series deposited with Ta-HiPIMS/Ti-DCMS and 20 ≤ V_s ≤ 280 V, residual stresses are relatively low, ranging from −0.5 GPa with V_s = 20 V to −0.7 GPa with V_s = 160 V. With V_s ≥ 200 V, σ for Ti_0.92Ta_0.08N increases to −1.4 ± 0.15 GPa, likely due to the accumulation of residual defects. Thus, it can be concluded that the optimum V_s for the growth of dense, hard, and almost stress-free Ti_0.92Ta_0.08N films at T_s < 120 °C is in the range of ∼120 to ∼180 V, corresponding to the light green region in Fig. 9. These numbers should not uncritically be transferred to other material systems as the metal-ion flux intensity is also expected to play a role if, for example, overlapping collision cascades need to be considered, as demonstrated in Sec. VII for the case of TiAlWN.

A key factor allowing to obtain dense films, combining high hardness and low residual stress, in the absence of thermally activated adatom mobility (T_h < 0.12), is that Ta ions are primarily incorporated into the lattice (as evident by an increased lattice parameter),^29,49 hence forming solid solutions with low residual defect concentration. Such results cannot be obtained by irradiation with heavy noble-gas ions (e.g., Xe⁺, m_Xe = 131.29 amu) of similar energy, since they would become incorporated as interstitials leading to build up of high residual stress levels at relatively low V_s.

To achieve full densification, it is critical that atom mobility in the DCMS layer deposited between two consecutive Ta-HiPIMS pulses is active. This is the case if (1) the ion penetration range and the average depth of resulting collision cascades are longer than the thickness of the newly grown DCMS layer, and (2) the number of recoils created per metal atom deposited between the HiPIMS pulses is sufficiently high so that the lateral overlap of collision cascades is realized. In the Ti_0.92Ta_0.08N model experiments discussed in this section, the TiN thickness deposited between pulses is only 0.022 monolayer (ML); thus, the condition (1) is easily satisfied by virtue of adatom lower activation energy (∼1–10 eV). Moreover, the TiN and TaN thicknesses deposited during each HiPIMS pulse are 0.000 45 and 0.002 Ml, respectively. Thus, the real-time TaN deposition rate is 4.4× higher than that of TiN, resulting in strongly overlapping collision cascades fulfilling condition (2).

For better control of metal ion energy and momentum, it is desirable to minimize the fraction of doubly ionized metal-ions (36% in the case of Ti_0.92Ta_0.08N growth discussed here). This can be realized by using sputtering gases with the first ionization potential $I P g 1$ lower than the second ionization potential of sputtered metal $I P M e 2$⁠, which was shown to be a very effective way to reduce the fraction of doubly ionized Ti species during Ti-HiPIMS.⁵⁵ In the case of Ta-HiPIMS operated in Ar/N₂ mixtures, the Ta²⁺ fraction is relatively high since $I P T a 2=15.60 eV$⁠, which is slightly lower than $I P A r 1=15.76 eV$⁠.⁸¹ The population of electrons with energies high enough to create Ta²⁺ through electron impact ionization can be effectively reduced by exchanging Ar for Kr $( I P K r 1 = 14.00 eV)$ or even Xe $( I P X e 1 = 12.16 eV)$⁠.

The high quality of the TM nitride layer grown with metal-ion assistance and no external heating was confirmed by application studies in which Ti_1−xTa_xN films performed exceptionally well as Cu diffusion barriers on Si(001) (Ref. 82) and provided good corrosion protection for stainless-steel substrates, thus attesting to full film densification.⁸³ 

The basics presented in Sec. II suggest that the metal ion mass relative to that of film constituents would have a decisive role on the extent of densification effects in the absence of thermally activated surface mobility. This aspect was recently addressed in systematic studies involving VIB TM ions: Me = Cr (m_Cr = 52.0 amu), Mo (m_Mo = 96.0 amu), and W (m_W = 183.8 amu) from respective elemental targets operated in HiPIMS at the similar peak target current density.⁸⁴ As mentioned in Sec. IV, VIB metals are particularly well-suited for such experiments as gas-ion fluxes detected at the substrate plane are relatively low due to pronounced rarefaction, which means that the energy and momentum of metal-ions can be precisely controlled by V_s without affecting corresponding parameters of the gas ions (Ar⁺, N₂⁺, and N⁺). Metastable NaCl-structure Ti_0.50Al_0.50N thin films were selected as a model materials’ system due to well-known challenges for phase stability control^85–87 and high relevance for industrial applications.⁸⁸ The experimental arrangement is as sketched in Fig. 4(a), with two Ti_0.5Al_0.5 targets on DCMS positions and VIB elemental targets operating in the HiPIMS mode. Three film series were grown for each type of metal ion by altering the DCMS powers: (Ti_1−yAl_y)_1−xCr_xN, (Ti_1−yAl_y)_1−xMo_xN, and (Ti_1−yAl_y)_1−xW_xN, with the relative VIB metal content on the cation lattice x = Me/(Me + Ti + Al) being the main variable. A metal-ion-synchronized substrate bias with the amplitude of 120 V was used. T_s was lower than 130 °C (corresponding to T_h < 0.12) with no external heating applied.

The O concentrations c_o in all films measured by ERDA are plotted in Fig. 10(a). Since all layers are deposited at relatively low background pressure [cf. Fig. 2(a) and related discussion], the majority of detected oxygen is from inward diffusion along inter- and intracolumnar voids upon air exposure. Hence, as the air exposure time is the same for all samples, c_o is a good measure for film porosity.

Results presented in Fig. 10 reveal a strong dependence of film properties on the mass of the incident metal ions. First, Ti_0.50Al_0.50N reference films show properties typical of DCMS nitrides grown by DCMS with no substrate heating (cf. Sec. V): c_o is very high at 3.8 at. % revealing high porosity, and not surprisingly, the nanoindentation hardness is very low (8.4 GPa).

(Ti_1−yAl_y)_1−xCr_xN layers (0.07 ≤ x ≤ 0.23, 0.46 ≤ y ≤ 0.53) deposited with low-mass Cr⁺ irradiation exhibit high oxygen content, 2.0 ≤ c_o ≤ 3.2 at. %, irrespective of x—indicative of high porosity—and, consequently, poor mechanical properties, 14.4 ≤ H ≤ 15.8 GPa. y exhibits an increase with x from 0.46 with x = 0.07 to 0.53 with x = 0.23, revealing resputtering of Ti by similar mass Cr ions.

For (Ti_1−yAl_y)_1−xMo_xN films (0.10 ≤ x ≤ 0.32, y = 0.46 ± 0.01) grown with heavier Mo⁺ ion irradiation, c_o shows a clear dependence on x: O content decreases from 2.3 at. % with x = 0.10 to 0.7 at. % with x = 0.32. This indicates that film porosity decreases with increasing average number of Mo⁺ ions incident per deposited metal atom, which results in that hardness increases from 21.3 to 27.8 GPa. The relative Al content is lower than in the reference Ti_0.50Al_0.50N film, which is explained by preferential sputtering of lighter Al atoms.

The best results are, however, obtained for the highest mass W⁺ ions. All (Ti_1−yAl_y)_1−xW_xN films (0.09 ≤ x ≤ 0.27, y = 0.45 ± 0.01) are fully dense, even with the lowest W concentration x = 0.09, as revealed by c_o = 0.4 at. %, which corresponds to O incorporated during film growth. These films exhibit state-of the-art mechanical properties, 31.1 ≤ H ≤ 31.8 GPa, typical of Ti_0.50Al_0.50N grown at 500 °C with the assistance of energetic gas ion bombardment.^89–94 

The above findings are confirmed by plan-view TEM images shown in Fig. 11 for (a) the reference Ti_0.50Al_0.50N film as well as (Ti_1−yAl_y)_1−xMe_xN layers with x ∼ 0.1 and Me = (b) Cr, (c) Mo, and (d) W. A 5–20 nm wide pores are observed for the reference film, which then decrease with the increasing metal-ion mass. A crucial point is revealed by selected-area electron diffraction (SAED) patterns shown as insets—all layers have a single-phase NaCl structure. Thus, even the highest mass W⁺ irradiation does not cause precipitation of the softer equilibrium hexagonal AlN phase that is so detrimental for cutting applications.⁹⁵ This result was confirmed also for a (Ti_1−yAl_y)_1−xW_xN film with the highest W content of x = 0.27.⁸⁴ 

TRIM simulations of ∼120 eV Me⁺ (Me = Cr, Mo, and W) ion bombardment of Ti_0.50Al_0.50N layers provide more insights into the effects of varying metal ion mass $m M e +$ on ion ranges, depth profiles, and energy distributions of primary recoils. Results summarized in Fig. 12 show several important features: (1) primary ions penetrate deeper into the film with increasing $m M e +$⁠: the maximum of the Me⁺ distribution in Figs. 12(a)–12(c) shifts from 10 Å for Cr ions to 14 Å for Mo⁺ and to 18 Å for W⁺; (2) the recoil ranges are independent of $m M e +$⁠, which results in that Cr⁺ ions reside within the high mobility surface zone, while W⁺ ions get trapped at larger depths and cause additional densification; (3) the number and energy of N and Al recoils decrease with increasing $m M e +$ due to an increasing mass mismatch between lighter film constituents and projectiles; and (4) for W⁺, the incident ion energy is almost entirely transferred to the cation lattice [cf. Figs. 12(d)–12(f)] resulting in a larger number of low-energy Ti and Al recoils.

In the examples discussed in this section, the overall energy consumption for the typical 3h long process (including all three steps marked in red in Fig. 1) was lowered by 64% with respect to standard Ti_0.50Al_0.50N deposition conducted at 500 °C. This saving is due to the fact that in the novel approach, the extra power applied to HiPIMS cathodes is much lower than that required by resistive heaters to reach and maintain the conventional substrate temperature of ∼500 °C (2 kW vs 2 × 8.8 kW). Moreover, the HiPIMS power is used very efficiently due to the operation with low duty cycle and the application of synchronized substrate bias to master ion-surface interactions and densify the films. Thus, the supplied process energy is utilized where it is most needed, i.e., at the growing film surface, resulting in that films grown with W⁺ irradiation have excellent properties despite T_s < 130 °C (and T_h < 0.12).

It might be expected that metal-ion-induced densification effects should depend on incident ion energy $E i$⁠, momentum $p i$⁠, and the ion flux [or the number of incident ions per deposited metal atom, which for complete ionization and TiAlMeN growth equals to x = Me/(Me + Ti + Al)]. Results presented in Fig. 10 reveal that varying the relative concentration of metal ions on the metal lattice in a rather wide range (0.1 ≲ x ≲ 0.3) with V_s = 120 V leads to visible changes in film density only for Mo ions. For the Cr⁺ series, $m M e +$ is apparently too low and perhaps a combination of higher substrate bias and even higher ion dose would result in dense films. Using higher metal-ion concentrations, however, contradicts the basic application premise that the composition of the basic layer deposited by DCMS sources is not significantly altered. In this respect, W⁺ irradiation gives the most promising results, as high $m W +$ clearly results in that fully dense layers are obtained even with the lowest ion concentrations tested. Thus, the (Ti_1−yAl_y)_1−xW_xN system appears most interesting for exploring the relation between $E i$⁠, $p i$⁠, and x.

Such experiments were conducted in the same setup as the one described in Sec. VI. A total of 28 (Ti_1−yAl_y)_1−xW_xN films were grown as a function of two variables: the amplitude of the metal-ion-synchronized substrate bias 60 ≤ V_s ≤ 600 V and the W concentration on the metal lattice 0.02 ≤ x ≤ 0.12, controlled by varying HiPIMS pulse width, while keeping the peak target current and pulsing frequency constant (to maintain similar ionization of the sputtered W flux).⁹⁶ Results shown in Fig. 13 reveal that a strong coupling exists between the W⁺ incident energy/momentum (controlled by V_s) and the minimum W concentration required to grow dense layers. Again, good correlation is found between oxygen content in the films [Fig. 13(a)] and nanoindentation hardness [Fig. 13(c)]. For films deposited with the lowest x = 0.02, O content (indicative of porosity) is relatively high; however, even with such low W⁺ concentrations, certain densification takes place with increasing W⁺ energy and momentum, as evident by the fact that c_o drops from 3.5 at. % with V_s ≤ 120 V to 2.3 at. % with V_s = 540 V. Nanoindentation hardness is correspondingly low and ranges between 15.7 and 21.0 GPa with a slight tendency to increase with increasing V_s. A further increase in V_s beyond 600 V could possibly yield complete densification; however, such high bias values are not practical in applications involving coatings over sharp edges (e.g., cutting inserts), in which case local electrical fields could reach extreme values leading to film deterioration and/or adhesion failure.

The c_o(V_s) plot is much steeper for the x = 0.04 series, with O content varying from 2.8 at. % with V_s = 60 V to 0.8 at. % at the highest V_s = 600 V. As a consequence, also H shows a stronger dependence on the bias amplitude and increases from 16.0 to 28.3 GPa in the same V_s range. Irrespective of the applied V_s, residual stresses are low [cf. Fig. 13(d)], meaning that layers deposited with 540 ≤ V_s ≤ 600 V are hard and stress-free.

With even higher W⁺ dose (x = 0.07), the rapid drop in c_o(V_s) takes place at significantly lower V_s values: already with 180 V bias, the oxygen content is down to 0.7 at. % and levels out for higher V_s values at 0.5 at. %. H = 25.2 GPa already with the lowest substrate bias of 60 V and increases rapidly to above 30 GPa for all films deposited with V_s  ≥ 180 V. The maximum H = 31.5 GPa is obtained for films grown with V_s = 300 V. Residual stress σ exhibits a gradual increase toward compressive state with increasing V_s. However, σ is only −0.9 GPa for films grown at 180 V bias, so also in this sample series the combination of high hardness and low residual stress can be realized. Stresses increase to −1.5 GPa with V_s = 300 V and −1.8 ± 0.2 GPa for even higher substrate bias.

c_o values are even lower in (Ti_1−yAl_y)_1−xW_xN films with the highest W content x = 0.12. The oxygen concentration is at 1 at. % with the lowest V_s of 60 V and decreases to only 0.3 at. % for layers grown with V_s ≥ 120 V, while hardness is high at 29.0 ± 1.0 GPa for V_s ≥ 120 V. A drawback of the higher ion dose is however, the much steeper increase in the σ(V_s) plot [cf. Fig. 13(d)]. σ = −2.2 GPa already with V_s = 120 V and increases to −4.4 GPa with V_s = 480 V. At even higher V_s, stresses become so high that films delaminate from their substrates.

Importantly, no precipitation of the wurtzite AlN phase is observed for any combination of V_s and x. Even the (Ti_1−yAl_y)_1−xW_xN films with the highest W content of x = 0.12 deposited with the 480 V bias are of single cubic phase. For all layers, the relative Al content y is lower than in the reference Ti_0.52Al_0.48N film [cf. Fig. 13(b)] due to preferential sputtering of lighter Al atoms.

To summarize, in order to obtain (Ti_1−yAl_y)_1−xW_xN films with the same quality as those grown at high temperatures, x ≥ 0.04 is necessary. The higher the W concentration, the lower the amplitude of the synchronized bias that gives dense films with high hardness. Thus, the following set of parameters results in high quality layers: x = 0.04/V_s = 540 V and x = 0.07/V_s = 300 V. Both combinations give hard films with relatively low residual stress level. For x ≥ 0.07, the residual stresses show a steep increase with V_s, which is the main reason why layers with x = 0.12 are not the first choice although they are dense and hard. It is also worth mentioning that the σ(V_s) plot for (Ti_1−yAl_y)_1−xW_xN films with x = 0.07 is very similar to that described in Sec. V for Ti_0.92Ta_0.08N films grown with Ta⁺ ion irradiation [cf. Fig. 9(c)], which suggests that trends outlined in this section would apply to a broad range of TM-based nitrides.

This far, no distinctions were made between terms metal-ion energy and metal-ion momentum transfer while describing the densification effects. Previous experiments involving intense metal-ion fluxes revealed that the phase composition and the mechanical properties of Ti_1−xAl_xN alloys with x ∼ 0.6 grown by the hybrid HiPIMS/DCMS co-sputtering are determined primarily by the average metal-ion momentum transfer per deposited atom $⟨ p d ⟩$ rather than by the average metal-ion energy per deposited atom $⟨ E d ⟩$⁠.⁹⁷ The former parameter was also found more adequate to describe the phase evolution of Ti_1−xSi_xN films (0 ≤ x ≤ 0.26).⁹⁸ Those layers were, however, grown by a conventional approach, i.e., at T_s = 500 °C and with no high-mass ion irradiation.

To compare the role of the two parameters under specific growth conditions that apply if the surface mobility is provided by high mass metal-ion irradiation at low T_s values, we use results acquired from (Ti_1−yAl_y)_1−xW_xN film series described in Sec. VII, as they constitute a systematic set of samples grown under identical process conditions (gas composition, pressure, temperature, etc.). In Fig. 14, concentrations of oxygen adsorbed during air exposure, used here to assess film density, are plotted as a function of (a) $x V s$ and (b) $x V s$⁠.

Within this simplified treatment, which assumes that the effects of gas ions can be neglected as they arrive with relatively low energy, the former expression is proportional to the average energy deposited per metal atom, $⟨ E d ⟩= E i J i/ J M$⁠, in which $E i≈e V s$ is the average energy of metal ions. $J i/ J M$ is the ratio of the energetic metal-ion flux to the flux of condensing metal atoms that can be expressed as αx, where α corresponds to the ionized fraction of the W flux, $W +/( W + + W)$⁠, and x = W/(W + Al + Ti). Thus, $⟨ E d ⟩=αex V s$ is proportional to $x V s$⁠. In a similar way, the average momentum deposited per metal atom $⟨ p d ⟩= 2 E i m M e + J i/ J M=αx 2 e V s m M e +$ is proportional to $x V s$ for the same mass of the incident metal ion.

The unknown variable in the above equations is the ionized fraction of the W flux from the HiPIMS cathode α. However, the important point is that α is not expected to vary significantly for all films included in this comparison due to the fact that the peak target current density and, hence, plasma density in front of the W-HiPIMS target (which directly impacts the ionization probability)⁵⁹ are maintained constant.

Data plotted in Figs. 14(a) and 14(b) suggest that densification of (Ti_1−yAl_y)_1−xW_xN films grown by hybrid HiPIMS/DCMS co-sputtering in the absence of external substrate heating is predominantly determined by W⁺ momentum transfer rather than by the energy transfer per deposited metal atom. This interpretation is based on the fact that data points plotted as a function of $x V s$ show obvious trend, which contrasts with the significant scatter in the $c o−( x V s)$ plot. The residual oxygen content in Fig. 14(b) shows a gradual decrease with increasing $x V s$ from values as high as 4.2 at. % for the reference Ti_0.52Al_0.48N layer down to 0.3–0.5 at. % with $x V s ≳1.2$⁠. The latter marks the threshold value: all films deposited with $x V s ≳1.2$ are dense, and their c_o values are determined by O adsorption during film growth under high-vacuum conditions. However, as noted above, residual stresses tend to increase with increasing the amplitude of synchronous substrate bias, especially in films with higher x; thus, the optimum layers are obtained if $x V s$ only slightly exceeds the threshold value.

The above criterion obviously is not generally applicable. For a given material system and the incident metal-ion mass, it allows us, however, to predict the required substrate bias amplitude for a given metal-ion concentration (or the other way around). Such approach was employed to grow dense single-phase Ti_0.32Al_0.63W_0.05N films at substrate temperature not exceeding 130 °C.⁹⁹ Despite high Al content and energetic W⁺ irradiation, the precipitation of softer hexagonal AlN phase was not observed by SAED. The nanoindentation hardness was comparable to that of TiAlN films grown at 400–500 °C, while the residual stresses were very low. Importantly, the energy consumption by resistive heaters was reduced from 38 kWh in the conventional DCMS process (used to grow reference Ti_0.36Al_0.64N layers) to only 4 kWh, resulting in that the overall energy consumption (heating + etching + coating) was lowered by 83%.

All experiments described in Secs. IV–VIII were performed in the stationary configuration illustrated in Fig. 4(a). While such setup facilitates basic studies of metal-ion/surface interactions, it does not reflect all the complexity characteristics of film growth under industrial conditions. For example, to achieve conformal coverage of complex objects such as cutting inserts or drill bits, a substrate rotation is applied. The latter can be onefold or multifold with substrates making a planetary-like motion pattern with several cathodes placed on the outer circumference of the rotating table [cf. Fig. 4(b)]. The target–substrate distance varies continuously and so does the average ion transit time from the target to the substrate. Hence, the complete synchronization of substrate bias pulses to metal-ion fluxes from HiPIMS sources, as during film growth with stationary substrates, is not practically possible. V_s can, however, still be applied in phase with Me⁺ fluxes during certain periods, e.g., when the substrate is passing in front of the HiPIMS target and is exposed to the most intense metal-ion flux.

Another consequence of substrate rotation is that the separation of HiPIMS metal-ion fluxes and the atom fluxes from DCMS sources varies during deposition, likely resulting in compositional modulation.

Both effects likely decrease the extent of densification, which can be achieved with high mass metal irradiation. In addition, the situation in industrial setups with (say) threefold rotation is even more complicated due to the fact that each substrate has its individual history that is determined by its starting position.¹⁰⁰ 

The above observations imply that the analysis of the real-life deposition process becomes quite complicated. In order to develop basic understanding of the key factors that govern the film growth with substrate rotation, the problem complexity has to be reduced. The first (already quite dramatic) step is to consider onefold rotation with multiple HiPIMS/DCMS sources. The essential question is as follows: Can dense films be grown with no external substrate heating if the growing film surface is only periodically exposed to high mass metal-ion irradiation?

To answer that, (Ti_1−yAl_y)_1−xW_xN films with the volume-averaged x varying from 0.04 to 0.14 and 0.49 ≤ y ≤ 0.54 were deposited with onefold substrate rotation using two W-HiPIMS and four Ti_0.50Al_0.50-DCMS sources.¹⁰¹ In such setup [cf. Fig. 4(b)], rotating substrates are exposed to Al and Ti atomic fluxes when passing close to DCMS targets or to predominantly W⁺ ions in front of HiPIMS cathodes. Because of this uniformity in material fluxes arriving at the growing film surface, a multilayered structure is expected to form with the periodicity determined by the applied cathode powers and rotation speed.

Two variables that should have a key effect on densification (and, hence, on film properties) were considered, namely, the thickness of the DCMS layer deposited between two consecutive exposures to the W⁺ ion flux (controlled here by varying the power applied to DCMS cathodes) and the thickness of the W⁺-modified layer, i.e., the ion penetration range (controlled by adjusting the amplitude of the bias pulse that is synchronized to W⁺ fluxes at the shortest target-substrate distance of 6 cm when the substrate is in front of the HiPIMS cathode). We discuss below the typical behavior using as an example Ti_0.42Al_0.48W_0.10N films grown with 15 ≤ V_s ≤ 600 V and no external heating (T_s < 150 °C). Experiments with reference TiAlN films grown at self-bias on rotating substrates revealed that, under such conditions (DCMS powers and rotation speed), 2.4 nm is deposited during each substrate passage in front of two DCMS targets. To obtain high-quality films, this layer has to be effectively densified upon exposure to the W⁺ flux from the HiPIMS source.

XTEM images shown in Fig. 15(a) for films grown on floating substrates (V_s = 15 V) reveal both intra- and inter-columnar voids of similar type as those observed for samples deposited in the stationary configuration if no heating was used [cf. Figs. 7(a), 7(c), and 11(a)]. Markedly, both kinds of porosity disappear in films deposited with V_s = 180 V [Fig. 15(b)]. Lattice-resolution TEM (HRTEM) acquired form selected grains and corresponding inverse fast-Fourier Transform (FFT) images shown in the same figure visualize lattice fringes extending over layers in columnar grains, irrespective of V_s. Thus, the W⁺-induced intermixing that leads to densification does not interrupt the process of local epitaxial growth resulting in that the (subsequently) deposited nanolayers have coherent interfaces. Corresponding STEM images shown in Fig. 15(c) confirm the presence of pronounced porosity in the Ti_0.42Al_0.48W_0.10N film grown with V_s = 15 V. In this case, these low-density regions are visible as dark areas extending along the growth direction. These features are about absent in films grown with V_s = 180 V [Fig. 15(d)]. The layering is, however, observed for both samples, with W-rich layers appearing brighter due to higher average atomic mass. Importantly, SAED patterns (not shown) confirm that Ti_0.42Al_0.48W_0.10N films are single phase with NaCl structure; thus, similar to the stationary case, energetic W⁺ irradiation applied at low T_s does not induce nucleation of thermodynamically favored hexagonal AlN.¹⁰¹ 

The bilayer period Λ determined from the intensity line profiles recorded close to the film/substrate interface decreases from 3.9 nm with V_s = 15 V to 3.6 nm with V_s = 180 V. These numbers agree with those calculated from positions of satellite peaks observed around 111 reflections in XRD θ–2θ diffractograms acquired from Ti_0.42Al_0.48W_0.10N films and shown in Fig. 16. Here, a decrease of Λ from 3.5 to 3.1 nm is observed when the substrate bias increases from floating to 240 V (both numbers being in perfect agreement with those calculated based on the total film thickness and the number of revolutions). For V_s ≥ 300 V, no satellite peaks are detected as the square-wave composition modulation smears out indicative of intense intermixing induced by high-energy W⁺ ions.

The reduction in the bilayer period confirmed by XRD and STEM could be potentially attributed to resputtering. However, as the Al/metal ratio does not decrease with increasing V_s amplitude, the decrease in Λ is more likely due to W⁺ implantation into the TiAlN-rich layers deposited during exposure to DCMS cathodes and the following densification. TRIM simulations (see Ref. 101) reveal that the increase in V_s from 60 to 180 V results in that the W⁺ penetration range increases from 1.9 to 2.8 nm, thus exceeding the thickness of the TiAlN layer deposited during each passage in front of DCMS cathodes (calculated to be 2.4 nm based on the thickness of the reference TiAlN film grown at selfbias). The effective cascade depth (the sum of projected range and straggle of primary Ti, Al, and N recoils) is 2.0 nm with V_s = 180 V, which corresponds to ∼80% of the freshly deposited DCMS volume. Thus, results of TRIM simulations are in quantitative agreement with STEM/TEM observations, which show vanishing porosity in films grown with V_s = 180 V. It can also be concluded that using even higher bias amplitude does not provide any benefits. On the contrary, a higher W⁺ energy (V_s > 300 V) leads to higher defect density, subsequent renucleation, and grain refinement manifested by broadening of Bragg peaks and disappearance of satellite signature from multilayers.¹⁰¹ 

Densification effects are also confirmed by the decrease in the concentration of oxygen adsorbed during air exposure. As shown in Fig. 17(a), c_o decreases from 1.9 at. % with V_s = 60 V to 0.8 at. % with V_s = 180 V. Markedly, the Ar content increases rapidly for V_s > 240 V, which is ascribed to the fact that the pulsed bias is applied during the entire deposition cycle, i.e., including the time when substrates are facing DCMS targets. In agreement with results of stationary experiments described in Secs. VI and VII, also in the case of films grown with rotation, low O content (indicative of higher density) correlates with high hardness. The latter, plotted in Fig. 17(b), exhibits a steep increase from 18.6 GPa with V_s = 15 V to 24.5 GPa with V_s = 180 V. Further increase in the substrate bias amplitude does not give any increase in H and results only in the buildup of compressive stress [cf. Fig. 17(c)] due to increasing defect creation and incorporation of Ar at interstitial sites. In fact, hardness decreases slightly for V_s > 300 V, which can be attributed to the increase in the interface width (as evidenced by disappearing satellite peaks in Fig. 16). The latter together with the difference between layer shear moduli has a strong influence on hardness of compositionally modulated films.¹⁰² 

The key parameters that influence the (Ti_1−yAl_y)_1−xW_xN film quality during hybrid W-HiPIMS/TiAl-DCMS growth with no external heating and substrate rotation are summarized in Fig. 18. The layer porosity is determined by the relation between the thickness of the DCMS layer deposited between exposures to the W-HiPIMS flux and the W⁺ ion penetration range. The former is controlled by the power applied to DCMS cathodes and the rotation speed, while the latter by the amplitude of the substrate bias pulses. Dense films are obtained provided the W⁺ penetration range is sufficiently long. However, moving further away from the diagonal line that marks the border between underdense and dense layers leads to buildup of compressive stresses. The optimum region is indicated in green, while yellow insets show compositional profiles obtained at various combinations of V_s and x. Going to multifold substrate rotation is likely to narrow the applicability of the time-domain-controlled metal-ion impingement growth benefactor engineering, yet worthwhile to pursue.

Results discussed in this Perspective article demonstrate that it is possible to significantly reduce the energy consumption during magnetron sputtering of TM nitride films by exploring growth paths such as the one offered by high mass metal ion irradiation. More conventional approaches that rely on extensive heating of the entire vacuum vessel imply large energy losses that can be avoided if the supplied process energy is utilized where it is most needed, i.e., at the growing film surface. For example, in the case of Ti_0.32Al_0.63W_0.05N layer growth by W-HiPIMS/AlTi-DCMS at T_s < 130 °C, the energy consumed by resistive heaters is reduced from 38 kWh in the traditional DCMS process used to grow reference Ti_0.36Al_0.64N films to only 4 kWh. Hence, the energy cost (considering heating, etching, and coating phase) is lowered by 83%. Most importantly, while Ti_0.32Al_0.63W_0.05N films are dense, heavy metal ion bombardment at low T_s does not cause precipitation of softer hexagonal AlN phase despite relatively high Al content. The nanoindentation hardness is comparable to that of reference TiAlN films grown at 400–500 °C. At the same time, residual stresses are very low as the role of Ar ions is suppressed by metal-ion-selective biasing.

Apart from significant energy savings, the novel approach enables the growth of high-quality coatings on substrates such as polymers, various types of steel, or Al and Mg alloys that cannot be coated by conventional techniques due to high process temperatures involved. All these material classes have either low melting points or undergo phase transformation at relatively low temperatures, which prevents the use of DC sputtering with its demands for extensive substrate heating. Hence, the replacement of external heating by irradiation with heavy metal ions together with metal-ion-synchronized biasing considerably expands the range of possible applications.

Several aspects that require further studies are identified:

• Experiments that involve substrate rotation (Sec. IX) demonstrate challenges and limitations of the presented approach. To take the full advantage of the high mass metal-ion fluxes, novel sputtering system designs should be considered that would ensure the synchronization of substrate bias pulses to metal-ion fluxes during the entire deposition process. For that, the substrate-to-target distance has to be constant. Possible solutions could consider in-line geometries with rows of HiPIMS/DCMS-operated magnetrons (for good flux overlap) on both sides of rotating substrates [cf. Fig. 19(a)]. Another alternative could maintain the circular geometry with an additional magnetron in the middle [Fig. 19(b)].

• The role of background pressure and the residual gas composition on the film properties has to be addressed. Preliminary results presented here reveal that the gentle heating sequence tends to lower the background pressure prior to the film growth with obvious benefits for coating purity. No evidence for additional outgassing taking place in the subsequent coating phase has been found. This, however, does not need to be the general condition so potential effects of higher levels of contaminants such as H₂O, O₂, CO, etc., need to be evaluated for specific materials system, not the least ceramic oxide systems.

• A parameter that is critical for effective densification not directly measured in experiments described in this Perspective article is the ionization degree of the HiPIMS metal flux at the substrate plane. As illustrated by results in Sec. V [Figs. 9(a) and 9(b)], film porosity decreases with increasing ion energy; thus, the non-ionized fraction of the heavy metal-ion flux, arriving with the sputter energy of several eV, does not perform any function and should, therefore, be minimized. In order to establish optimum HiPIMS pulse parameters that provide the highest density of collision cascades at the lowest metal-ion concentration, systematic measurements of the ionization degree are required. Such studies can, for example, be conducted with a gridless quartz crystal microbalance sensor under relevant process conditions.^103,104

• The film adhesion to relevant substrates is a key requirement in most applications. Further studies are necessary to understand the effect of substrate temperature on the effectiveness of ion etching to that end. Here, the new perspective is that the high mass HiPIMS ions can be tested for substrate pretreatments.

• While metal-ion-induced densification seems to be similar for all TM-based nitrides studied (as evidenced by similar substrate bias voltage and the relative heavy ion flux required to obtain dense layers for Ta⁺ → TiN, Ta⁺ → TiAlN, and W⁺ → TiAlN), the extension to other materials classes like TM diborides is yet to be demonstrated.

• Structure Zone Model (SZM) diagrams for PVD films need to be modified to capture also determining effects of mass- and momentum-tuned metal ion-assisted deposition whereby the Zone 2 can be extended to homologous temperature T_h as low as 0.12.

The authors most gratefully acknowledge the financial support of the Swedish Research Council VR Grant No. 2018-03957, the Swedish Energy Agency under Project No. 51201-1, the Knut and Alice Wallenberg Foundation Scholar Grant No. KAW2019.0290, and the WISE-AP01-PD18 grant, the Competence Center Functional Nanoscale Materials (FunMat-II) VINNOVA Grant No. 2022-03071, the Swedish Government Strategic Research Area in Materials Science on Advanced Functional Materials at Linköping University (Faculty Grant SFO-Mat-LiU No. 2009-00971), the Carl Tryggers Stiftelse contract CTS 20:150, the Åforsk Foundation Grant 22-4, and the Olle Engkvist Foundation Grant No. 222-0053.

The authors have no conflicts to disclose.

G. Greczynski: Conceptualization (equal); Data curation (lead); Formal analysis (lead); Funding acquisition (equal); Investigation (lead); Project administration (lead); Writing – original draft (lead). L. Hultman: Conceptualization (equal); Funding acquisition (lead); Resources (lead); Supervision (equal); Writing – review & editing (equal). I. Petrov: Conceptualization (equal); Formal analysis (equal); Methodology (equal); Supervision (equal); Writing – review & editing (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.