Metal-ion subplantation: A game changer for controlling nanostructure and phase formation during film growth by physical vapor deposition

During low-temperature film growth by magnetron sputtering, ion irradiation of the growing film surface is commonly used to eliminate porosity and provide a means for changing layer nanostructure, crystal phase, and texture.^1–5 Since conventionally used direct-current magnetron sputtering (DCMS) has a very low ionization of the sputter-ejected material,⁶ these advantages are almost exclusively due to the gas ions from the sputtering gas mixture. As it is often the case, benefits come at a price, which, in this case, is particularly steep—entrapment of energetic noble gas ions at interstitial sites resulting in an increase in the residual compressive stress state,^7,8 and, eventually, in the adhesive and/or cohesive film failure.^9,10

It has been recognized in the early 1990s that irradiation by ions of the sputtered flux (referred to as metal-ions in the remaining part of this Perspective) is free from this drawback,¹¹ as ionized film-forming species are primarily incorporated at lattice sites, which decreases the ion-induced compressive stress in as-deposited layers, while allowing to maintain all benefits of gas-ion bombardment.¹¹ In addition, a better mass match between incident ion and the film-forming atoms implies more efficient energy and momentum transfer to the growing film surface, which enables lower growth temperatures by substituting thermally induced mobility for that provided by ion irradiation.^12,13

These attractive features of metal-ion irradiation were to some extent utilized in ionized physical vapor deposition,¹¹ cathodic arc deposition,¹⁴ plasma immersion ion implantation and deposition,¹⁵ and, eventually, in high-power impulse magnetron sputtering (HiPIMS).^16,17 The degree of ionization of the metal flux obtained in cathodic arc deposition is generally larger than in both non-reactive and reactive¹⁸ ionized PVD as well as HiPIMS. The advantage of the latter technique as compared to cathodic arc is that macro particle emission can be avoided.¹⁹ This, together with the simplicity of use which does not require any modifications of existing sputter deposition setups to operate in HiPIMS mode, created a broad interest in the thin film community, as manifested by more than 1200 papers published²⁰ following the original paper by Kouznetsov et al.¹⁶ 

As there are several excellent review articles on HiPIMS,^21–27 we highlight below the two most essential aspects for the topic of this paper, namely, (i) high ionization of the material flux sputtered from the target²⁸ and (ii) the inherent time separation between gas and metal-ions at the substrate.²⁹ 

High ionization levels in HiPIMS are due to high plasma density in front of (and close to) the sputtering target,³⁰ which enhances the probability for electron-impact ionization of target atoms as proven by numerous groups with evidence coming from various analytical techniques.^28,31–36 The ionized fraction of the sputtered flux varies in a wide range from 4.5% for C (Ref. 37) to more than 90% for Ti.²⁸ The reasons for this large spread are threefold: (i) material-specific properties such as the first ionization potential $I P M e 1$ of the sputtered species with respect to that of the sputtering gas $I P g 1$⁠,³⁸ or the cross section for electron-impact ionization, (ii) the operating conditions that include primarily the peak target current density (which often varies widely from 0.1 A/cm² to several A/cm² with a direct effect on the plasma density in front of the target), and (iii) the technique used to estimate the ionization degree as well as the details of the experimental setup (e.g., probe position and orientation with respect to the target).

From the novel film growth perspective described in this review, equally important as the significant ionization of the sputtered flux is the ability to separate metal-ion and gas-ion fluxes at the substrate. This is possible to accomplish by extensive gas rarefaction caused by the high temporal fluxes of sputtered species,^39–41 which results in the metal/gas-ion ratio incident at the growing film surface being able to temporarily reach high values.⁴² This effect was first observed by time-resolved optical emission spectroscopy (OES) for Ti₅₀Al₅₀ target sputtered in Ar at the peak target current density of 1.1 A/cm².²⁹ 

The above phenomenon opened new research directions for thin film growth by magnetron sputtering by enabling selective control of metal-ion energy and momentum.⁴³ This is achieved by the application of substrate bias pulses synchronized to the metal-ion-rich portion of the HiPIMS pulses using as an input the results of time-resolved ion mass spectrometry analyses of ion fluxes incident at the growing film surface, the technique we refer to as metal-ion-synchronized HiPIMS.^44–47 The metal-ion energy and momentum can then be controlled by simply varying the amplitude of the bias pulse. The impact of gas-ion irradiation on the nanostructure evolution is minimized by maintaining the substrate at floating voltage during the time gas ions dominate the flux to the substrate such that their incident energy is lower than the lattice displacement threshold. One example of the new possibilities offered by this approach is the growth of nanostructured N-doped bcc-CrN_0.05 films, with atomically smooth surfaces, which combine metallic (bcc-Cr crystal structure, electrical resistivity, and toughness) and ceramic (high hardness) properties.⁴⁸ 

Much of the potential of metal-ion fluxes available in HiPIMS has been realized by using a hybrid approach in which HiPIMS is combined with a second magnetron operating in the DCMS mode to grow multinary compounds. This configuration provides important degrees of freedom as the type of metal-ion irradiation can be selected by proper choice of the sputtering target operating in the HiPIMS mode, while the DCMS source (or sources) provides a continuous flux of neutrals. Hence, fundamental differences among the effects of different metal-ion fluxes were revealed.^44,46 Irradiation with lower-mass metal-ions (e.g., Al⁺ or Si⁺) has been shown to yield fully dense, hard, and stress-free Ti_0.38Al_0.62N layers,⁴⁹ single-phase cubic Ti_1−xSi_xN with record-high SiN concentrations,⁵⁰ and unprecedented AlN supersaturation in single-phase NaCl-structure V_1−xAl_xN.^51–54 At the other extreme, bombardment of the growing film with heavy metal-ion fluxes (e.g., Ta⁺) was shown to provide fully dense, low stress, Ti_1−xTa_xN, and Ti_1−x−yAl_xTa_yN compound films without using external substrate heating.^12,13 This is because of effective low-energy recoil generation, overlapping collision cascades, and enhanced near-surface atomic mixing, which overtake segregation driven by thermal diffusion.

With the above perspective, we want to highlight the newly reported and exciting phenomenon discovered during metal-ion-synchronized HiPIMS/DCMS film growth experiments involving lighter metal-ions, namely, metal-ion subplantation.⁵¹ It enables unprecedented control over nanostructure and phase formation in the case of transition metal (TM)-based nitrides and we see no obvious physical limitation that would prevent its application to other materials systems.^53,54 Hence, we start in Sec. II by revisiting essential results from time-dependent ion mass spectrometry studies conducted at the substrate position during reactive sputtering of Groups IVB and VIB TM targets in Ar/N₂ mixtures. This knowledge is needed to exploit the effect of metal-ion irradiation on the film properties by selecting the optimum bias pulse parameters. The basic film growth concept, which relies on the subplantation effect, is described in Sec. III, followed by key examples (Sec. IV). We end with a summary and outlook for possible future developments in this field.

In order to fully exploit the advantages of metal-ion-synchronized HiPIMS/DCMS film growth, a detailed knowledge of the time evolution of metal- and gas-ion fluxes incident at the substrate is crucial for two reasons.^42,55 First, in addition to the tuning of the metal-ion energy and momentum by varying the amplitude of the substrate bias, precise timing of the bias duration and phase shift with respect to the HiPIMS pulse at the cathode allows for intentional application of the substrate bias only when metal-ion flux dominates over that of gas ions. Second, it is essential for revealing the relationship between metal-ion mass, energy, momentum and flux, and the film properties.⁴³ 

The crucial feature of film growth with the metal-ion-synchronized HiPIMS/DCMS approach is that the substrate is maintained at a floating potential between the metal-ion-rich pulses (95%–99% of the deposition time), i.e., during the time when low-energy gas-ion irradiation dominates, resulting in the later species being incident at the growing film surface with energies below the lattice displacement threshold (∼20–50 eV, depending upon the ion and film species involved).^56,57 Hence, structure evolution is determined to a large extent by energy and momentum transfer from metal-ion irradiation during the metal-ion-rich phases of the HiPIMS pulses, both controlled by the choice of metal-ion mass and the amplitude of the synchronized bias pulse. Moreover, the low incident energy of gas ions efficiently prevents Ar trapping in the interstitial positions, which has been demonstrated to result in low residual stress levels.⁴⁴ 

Recent extensive time-dependent in situ mass- and energy-spectrometry analyses of HiPIMS sputtering of Groups IVB and VIB TM targets in Ar and Ar/N₂ gas mixtures allowed for the identification of key mechanisms that govern the time evolution of ion fluxes incident at the substrate.^42,55 Below we discuss the most representative examples of these essential findings. In all experiments, carried out using a Hiden Analytical EQP1000 instrument in a CemeCon CC800/9 magnetron sputtering system,⁵⁸ the spectrometer orifice was placed at the substrate position and 18 cm away from the target. Rectangular 8.8 × 50 cm² targets were used, and the gas pressure was set at 3 mTorr (0.42 Pa). Time-dependent ion energy distribution functions $I( E i,t)$ were obtained for Meⁿ⁺ (n = 1, 2, …) metal, as well as gas, ions during 100 consecutive pulses such that the total acquisition time per data point was 1 ms. The ion energy was scanned in 0.5 eV steps from $E i=0$ to 50 eV. Data in Figs. 1 and 3 are corrected for the ion time-of-flight from the orifice to the detector using the formula supplied by the manufacturer (Hiden Analytical, UK), which is discussed in Ref. 59. Additional details are given in Ref. 60. The time-dependent ion fluxes $F M e n +(t)$ were obtained by integrating $I( E i,t)$ over the entire energy range recorded $F M e n +(t)=∫ I M e n + ( E i , t ) d E i$⁠.

Figure 1 shows time-dependent energy-integrated ion fluxes $F T i +(t)$⁠, $F T i 2 +(t)$⁠, $F A r +(t)$⁠, $F N +(t)$⁠, and $F N 2 +(t)$ recorded at the substrate plane during HiPIMS of a Ti target in Ar/N₂ atmosphere with a N₂ gas component optimized to provide stoichiometric TiN films.⁵⁵ Zero on the time axis corresponds to the onset of the cathode voltage pulse, while each data point at time t represents the number of ions collected during the interval from t − 5 to t + 5 μs. Results are shown as a function of peak target current density $J T , p e a k$⁠, which ranges from 0.2 to 2.0 A/cm², varied by changing the average power to the magnetron at the constant frequency of 300 Hz. The preset HiPIMS pulse length was 150 μs, while the actual duration of the current pulse varied as a function of $J T , p e a k$⁠, from ∼140 μs with 0.2 A/cm² to ∼80 μs with 2.0 A/cm², due to the limited size of the capacitor bank with respect to the target area.

The most critical aspect for TM nitride film growth is the interplay between Ti⁺/Ti²⁺ and Ar⁺ ion fluxes at the substrate plane. No overlap between $F T i +(t)$ and $F A r +(t)$ would allow for truly independent (from the gas ions) control of metal-ion energy and momentum by employing substrate bias pulses synchronized to the metal-ion-rich portion of HiPIMS pulses. Overlap between $F T i +(t)$ and $F N +(t)$ or $F T i +(t)$ and $F N 2 +(t)$ is not critical as all these ions are film-forming species. As can be seen in Fig. 1, the $F T i +(t)$– $F A r +(t)$ overlap clearly depends on the peak current density. Although Ar⁺ flux dominates the ion flux to the substrate for all peak current density values, several distinct changes can be observed with increasing $J T , p e a k$ at constant frequency.

Second, the current peak, denoted as I_T(t) in Fig. 1, shifts toward t = 0 μs (corresponding to the ignition of the HiPIMS pulse) with increasing $J T , p e a k$⁠. As a result of this, the $F T i +(t)$ peak moves toward earlier times and becomes narrower, leading to a smaller overlap with the Ar ion flux (which is less dependent on the peak target current density) at higher $J T , p e a k$ values. Thus, $J T , p e a k=0.5 A / c m 2$ presents more suitable conditions for metal-ion-synchronized film growth. For example, in the case of $J T , p e a k=0.5 A / c m 2$ setting the bias pulse V_s from 0 to 100 μs results in the metal/Ar ion ratio $N M e +/ N A r +=3.9$⁠, which is ∼20 times higher than the average for the entire 300 μs time period. Under such conditions, the fraction of accelerated Ar⁺ incident at the growing film is kept to a minimum.

An overview of how the $N M e +/ N A r +$ ratio changes as a function of substrate bias offset τ_offset and length τ_length is shown in Fig. 2 for all four peak target current density values. 2D color maps are obtained by integrating ion fluxes shown in Fig. 1 over corresponding time intervals. The green–blue transition indicates the border line between gas-ion and metal-ion dominated regions in the τ_offset–τ_length parameter space. Clearly, under the selected experimental conditions, sputtering with moderate $J T , p e a k$ in the range 0.5–1.0 A/cm² provides both largest metal-ion dominated areas $( N M e +/ N A r + ≳1)$ and highest $N M e +/ N A r +$ values. This is a direct consequence of the fact that with increasing $J T , p e a k$ to 2.0 A/cm², an intense $F A r +(t)$ peak appears during the initial phase of the HiPIMS pulse.

Plots of the type shown in Fig. 2 are required for efficient experimental design as they enable the identification of interesting conditions in the enormous process parameter space. They do not, however, give a definite answer as to what parameters should be selected to process coatings for a particular application. For example, in the case of $J T , p e a k=0.5 A / c m 2$⁠, the choice of τ_offset in the range 60–70 μs makes most sense as this is the point where $N M e +/ N A r +$ increases sharply. Shorter offset times result in biasing during the time when ion fluxes incident at the substrate have low intensities, which does not provide any benefits. The choice of τ_length is not at all that obvious. Metal-ion irradiation dominates up to τ_length = 70 μs (with τ_offset = 60 μs); however, such a biasing scenario implies that a part of the Ar⁺ ion flux that increases strongly for t > 100 μs is accelerated by the electric field of the substrate. Whether or not this is acceptable depends on the particular application and, in the case of metastable systems like V_1 − xAl_xN discussed below, may often come to a compromise between increase in the solubility limit and increase in the residual stress. In any case, a delicate interplay between $F T i +(t)$ drop and the rapid increase in the $F A r +(t)$ intensity, both taking place for t > 100 μs, is crucial. For the most effective metal-ion subplantation (see Sec. III), the substrate is biased during the entire metal-ion-rich period supplying a significant $F M e +(t)$ to the growing film surface, to ensure that as large metal fraction as possible is subplanted below the high-mobility surface zone, resulting in the formation of supersaturated alloys. Contrarily, too high Me fraction deposited at the surface during the time of low substrate bias facilitates the formation of thermodynamically favored phases, as discussed in detail in Secs. III and IV. At the same time, high fluxes of energetic Ar⁺ ions should be avoided as they lead to higher stresses from residual defect creation, which may, in turn, promote second-phase precipitates. Thus, one has to find a good compromise for a particular application by investigating the most interesting range in the parameter space outlined by τ_offset−τ_length plots.

A third observation that can be made based on the ion fluxes presented in Fig. 1 is that N⁺ ion flux increases with increasing peak target current density. A detailed analysis of corresponding ion energy distribution functions (IEDFs)⁵⁵ (not shown) reveals large similarity to metal-ion fluxes, which strongly suggests that N⁺ ions originate from the target, in agreement with other reports,⁶⁰ via a combination of both sputter-ejected N atoms, and reflected N atoms arising from dissociative N₂⁺ collisions at the target surface. $F N +(t)$ peak precedes that of $F T i +(t)$ due to the lower mass of nitrogen ions (m_N = 14.01 amu vs m_Ti = 47.87 amu) resulting in, on average, shorter transit time from the target to the substrate.

The fourth point is that the separation between $I T(t)$ maxima and subsequent $F T i +(t)$ peaks, which reflects the average ion transit times t_TOF from the target to the substrate, decreases with increasing peak target current density, from 55 μs with $J T , p e a k=0.2 A / c m 2$ to 30 μs with $J T , p e a k=2.0 A / c m 2$⁠. The effect is due to the reduced number of collisions of the sputtered atoms with gas particles as gas rarefaction increases, leading to an increase in the average metal sputter-ejection energy as sputtering is primarily occurring by metal-ions which, in turn, further increases gas rarefaction.^61,62

To illustrate the dependence of ion fluxes incident at the substrate during HiPIMS in the Ar (39.95 amu) and N₂ (28.01 amu) gas mixture on the choice of target material, the results for Ti (47.87 amu) and W (183.84 amu) are plotted in Fig. 3. Both TM targets were operated with a peak target current density of 1.0 A/cm² and sputtering was performed in Ar/N₂ atmosphere with a N₂/Ar flow ratio of 0.11, chosen to provide stoichiometric TM nitride films.⁵⁵ The difference is dramatic: while in the case of Ti sputtering, Ar⁺ dominates the ion flux, gas irradiation is essentially absent during W-HiPIMS. The primary reason for the observed difference is the extent of gas rarefaction effects taking place during HiPIMS pulse. W is characterized by a higher sputtering yield (⁠ $S A r → M e=1.45$ for W and 0.62 for Ti) and a lower reactivity toward N₂ (nitride formation enthalpy, $Δ H f 0=−0.2 eV / atom$ for W vs −3.4 eV/atom for Ti)^63, both leading to a much higher flux of sputter-ejected species than that encountered during Ti-HiPIMS (under similar conditions of peak current, pressure, etc.) and, hence, to a higher overall momentum transfer to the gas atoms upon collisions, resulting in more severe rarefaction.

Further illustration of the drastic decrease in gas density during W-HiPIMS is shown in Fig. 4, where time-resolved IEDFs $I M e +( E i), I A r +( E i), I N +( E i)$⁠, and $I N 2 +( E i)$ were acquired during 50-μs consecutive time intervals for all primary ions detected during the sputtering of Ti and W targets in HiPIMS mode. In the case of Ti-HiPIMS, $I T i +( E i)$ evolves from broad Sigmund–Thompson sputtered-species energy distributions^64,65 observed for t ≤ 100 μs to narrow, low-energy, peaks at 2–3 eV (the potential energy difference between the bulk plasma potential and the grounded mass-spectrometer orifice), as a result of inelastic collisions between sputtered atoms and gas species, the process often referred to as thermalization. In contrast, $I W +( E i)$ original energy distributions are preserved even 400 μs after the ignition of the HiPIMS pulse since the collision rate with gas species is very low, indicative of severe rarefaction. Interestingly, this fundamental difference between Ti-HiPIMS and W-HiPIMS is reflected in the corresponding $I N +( E i)$ distributions, which show thermalization only in the former case. The reason is that N originates primarily from the target, as discussed above.

The distinction made in Figs. 3 and 4 is typical for all Group IVB and VIB targets.⁵⁵ The general observation is that the overlap between $F M e +(t)$ and $F g +(t)$ distributions is much larger in the former case, thus making the selective manipulation of metal-ion energy and momentum via synchronized biasing more challenging. Since N⁺ is a primary film-forming ion, the overlap with metal-ion fluxes is not an issue. In contrast, $F A r +(t)$ fluxes are far more problematic as they can result in trapped Ar interstitials leading to higher residual film stress^66,67 and, as a consequence, to adhesive and/or cohesive film failure.^9,10 This can be overcome by using shorter HiPIMS pulses, while maintaining the peak target current density constant.⁶⁸ For example, $N T i +/ N A r +$ has been shown to increase from 1 with $τ O N=120μ s$ to 60 with $τ O N=30μ s$ for a Ti target operated at 1 A/cm². In this case, terminating the target current pulse once $I T(t)$ reaches a maximum resulted in much stronger gas rarefaction, while preventing significant Ar ionization during Ar refill, which takes place during longer pulses [characterized by a gradual decay in $I T(t)$ resulting from the limited size of the capacitor bank with respect to the target area].

In the case of Group VIB targets, one has a complete freedom in setting the bias length and offset based upon measured $F M e +(t)$ distributions, as Ar⁺ ion fluxes are nearly absent. In the latter case, the bias pulse width can be increased with increasing ion mass for a given target/substrate separation to account for longer transit times. The “first order approximation” guidelines for this are provided in Fig. 5, where the average ion transit times t_TOF, calculated based on the separation between $I T(t)$ and corresponding $F M e +(t)$ peaks, are plotted for different values of peak target current density. Although they are subject to change with varying processing parameters, the provided values give an important indication of what sort of variation can be expected based on the choice of target material at the source-to-substrate distance of 18 cm.

The precise tuning of the metal-ion energy through variation of the negative bias pulse amplitude V_s during metal-ion-synchronized HiPIMS requires, apart from synchronization to the metal-ion-rich portion of the ion flux incident at the substrate, also control over the metal-ion charge state. This is especially the case if the second ionization potential $I P M e 2$ of the sputtered metal is lower than the first ionization potential $I P g 1$ of the noble gas, which results in high fluxes of multiply charged Meⁿ⁺ (n > 1) metal-ions.⁴⁴ Under such circumstances, the metal-ion energy, and, hence, the subplantation depth, is expected to show a much larger spread due to the fact that doubly charged ions impinge on the film with, approximately, twice the energy of the singly charged ions. The effect of varying the fraction of doubly charged metal-ions in the incident flux on the metal implantation profiles was simulated by Monte Carlo TRIDYN methods for the case of Al⁺ and Al²⁺ incident onto VN film biased negatively at V_s = −200 V. The results are shown in Fig. 6. As intuitively expected, the film thickness over which Al atoms are distributed (width of the implantation profile) increases with increasing the Al²⁺ fraction $ξ A l 2 +$ from 9.8 Å with $ξ A l 2 +=0$ to 14.2 Å with $ξ A l 2 +=1$⁠. In the case of TM-based nitrides, this implies that the extent to which one can manipulate the phase formation (see Secs. III and IV) depends on the fraction of doubly ionized metal flux.

An effective way of controlling the metal-ion charge state relies on the use of noble gas mixtures involving gases with different first ionization potentials $I P g 1$⁠.³⁸ To illustrate this approach, Ti⁺ and Ti²⁺ ion fluxes from a Ti target sputtered in Ne $(I P N e 1=21.63 eV )$⁠, Ar $(I P A r 1=15.76 eV )$⁠, Kr $(I P K r 1=14.04 eV )$⁠, and Xe $(I P X e 1=12.16 eV )$⁠, at the average HiPIMS power P_a = 1.0 kW and pulse frequency f = 100 Hz (2% duty cycle), were studied by ion mass spectrometry performed at substrate position.³⁸  Figure 7 is a plot of the doubly to singly charged Ti ion flux ratio $F T i 2 +/ F T i +$ as a function of sputtering gas composition $χ g= f g/( f g+ f A r)$⁠, in which g stands for the noble gas Ne, Kr, or Xe. Very good correlation between $F T i 2 +/ F T i +$ and $I P g 1$ is evident: $F T i 2 +/ F T i +=0.62$ for pure Ne, which has the highest first ionization potential, and decreases to 0.17 in pure Ar, 0.03 in pure Kr, and 0.01 in pure Xe (which has the lowest ionization potential). All intermediate $F T i 2 +/ F T i +$ values can be obtained by using gas mixtures (see Fig. 7), which proves that the Ti charge state can be controlled over a wide range. The key point here is the relation between the second ionization potential of a sputtered species $I P M e 2$ and the average first ionization potential of the sputtering gas $I P g 1$⁠. The production of doubly charged metal-ions requires electrons with energy $E e>I P M e 2$⁠; thus, the Me²⁺ creation rate depends on whether $I P g 1$ is larger or smaller than $I P M e 2$⁠.

For Ti sputtered in Ar (⁠ $I P T i 2=13.62 eV$ and $I P A r 1=15.76 eV$⁠), $I P M e 2<I P g 1$⁠, there is a significant electron population in the discharge with energies in the range $I P T i 2< E e<I P A r 1$⁠, i.e., too low to ionize Ar, yet high enough to produce Ti²⁺. The opposite is true if $I P M e 2>I P g 1$⁠, which is the case for Ti sputtered in Xe $(I P X e 1=12.16 eV )$⁠. Here, Xe ionization depletes the population of electrons with $E e>I P X e 1$ and the Ti²⁺ creation rate decreases. As both $I P X e 1$ and $I P A r 1$ are considerably higher than $I P T i 1$ (6.85 eV), the impact on Ti⁺ production is small. Consequently, varying the Xe fraction in mixed Ar/Xe discharges has a large effect on the energy-integrated $F T i 2 +$ flux incident at the film growth surface, but affects $F T i +$ to a far less extent. Increasing the difference between $I P M e 2$ and $I P g 1$⁠, as for Ti sputtered in Ne $(I P N e 1=21.63 eV )$⁠, increases the population of electrons with energy in the range $I P T i 2< E e<I P N e 1$⁠, which leads to an increased production of Ti²⁺ as observed experimentally $( F T i 2 +/ F T i +=0.62)$⁠.

The idea presented above is not limited to Ti as the criteria $I P M e 1<I P X e 1$ and $I P X e 1<I P M e 2<I P N e 1$ are satisfied by the majority of metals.

A direct illustration of the phenomena discussed above is the tremendous difference in the fraction of detected doubly charged metal-ions during HiPIMS sputtering of Ti and Al targets in Ar/N₂ mixtures.⁴⁵ As shown in Fig. 8, the Ti²⁺ fraction recorded during Ti-HiPIMS increases with increasing peak target current density from 2% with $J T , p e a k=0.06 A / c m 2$ to 26% for $J T , p e a k=1.60 A / c m 2$⁠. This is two to three orders of magnitude higher than the Al²⁺ fraction during Al-HiPIMS, which reaches a maximum value of 0.04 at $J T , p e a k$ as high as 1.5 A/cm². The primary reason for the large difference between $F T i 2 +$ and $F A l 2 +$ is the correspondingly large difference in second ionization potentials: $I P A l 2=18.83 eV$ is significantly higher than $I P A r 1$ (15.76 eV), $I P N 2 1$ (15.55 eV), and $I P N 1$ (14.50 eV),⁶⁹ while the opposite is true for $I P T i 2$ (13.62 eV). Since ion energy distribution functions during Al-HiPIMS and Ti-HiPIMS are similar, the probability for multiple ionization events is further decreased for Al atoms, which are lighter than Ti atoms (m_Al = 26.98 amu vs m_Ti = 47.87 amu) and, hence, their transit time through the dense plasma region is shorter. Thus, upon application of a moderate substrate bias voltage, the asymmetry in Ti²⁺ vs Al²⁺ ion fluxes plays an important role during the growth of Ti_1−xAl_xN films as discussed in Sec. IV A.

The vast majority of film growth experiments involving metal-ion fluxes from HiPIMS sources utilized a hybrid HiPIMS/DCMS co-sputtering arrangement with elemental targets, denoted as Me₁ and Me₂ in Fig. 9, sputtered in Ar/N₂ gas mixtures to obtain ternary nitrides of the form (Me₁)_1−x(Me₂)_xN. Here, HiPIMS, which serves as a pulsed source of $Me 2 +$ metal-ions, is combined with a second magnetron operating in DCMS mode supplying a continuous flux of $M e 1$ metal atoms (Me₂-HiPIMS/Me₁-DCMS). The positions of Me₁ and Me₂ targets can then be switched to Me₁-HiPIMS and Me₂-DCMS to examine the differences observed when using $Me 1 +$ metal-ion pulses. The success of this approach for revealing different effects of selected individual metal-ion fluxes on layer nanostructure, phase content, and physical properties,⁴³ relies on well-documented differences in the intensity and composition of ion fluxes incident at the growing film surface during DCMS vs HiPIMS.^28,30–36 In a typical experiment series (see Sec. IV), films were grown in both target arrangements as a function of x by varying power to the DC magnetron, while maintaining the HiPIMS settings, and, hence, the incident metal-ion flux, constant. In this way, the properties of (Me₁)_1−x(Me₂)_xN layers with the same stoichiometry, but grown with either $Me 1 +$ or $Me 2 +$ metal-ion flux, could be directly compared. The above approach, although tested exclusively for nitrides, should be possible to apply to most other materials systems.

The essential ingredient of any subplantation-based thin film growth experiment utilizing a hybrid setup as shown in Fig. 9 is the pulsed substrate bias potential with negative amplitude V_s which is synchronized with the metal-ion-rich portion of the HiPIMS flux, as determined by energy- and time-dependent mass spectrometry analyses performed at the substrate position (see Sec. II). It is applied in order to exploit the effect of metal-ions and minimize the role of gas-ion irradiation, which dominates inbetween metal-ion-rich pulses, i.e., during 95%–99% of the total deposition time. During the latter film growth phase the substrate is intentionally set at the floating condition, as confirmed by the oscilloscope waveforms recorded in situ, to minimize the incident gas-ion energy so that it is maintained below the lattice displacement threshold. Consequently, the detrimental effects of noble gas-ion irradiation, in the form of higher residual stress levels^9,10 resulting from trapped Ar interstitials, are minimized. Hence, film structural evolution is predominantly determined by energy and momentum transfer from metal-ion irradiation during the HiPIMS pulses, easily controlled via the choice of metal target and the V_s amplitude.

The metal-ion-synchronized film growth in a hybrid HiPIMS/DCMS configuration employing light metal-ions offers the possibility of utilizing ion subplantation for obtaining supersaturated structures.^51,53,54 The basic idea behind the subplantation mechanism is illustrated in Fig. 10. This novel film growth concept relies on an intentional separation of film-forming material fluxes originating from DCMS and HiPIMS sources, Me₁-DCMS and Me₂-HiPIMS, in time and energy domains by utilizing the inherent asymmetry between corresponding ionized fractions. Unlike during DCMS growth, Me₁- and Me₂-type adatoms are not allowed to coexist at the very surface where the mobility is high due to elevated substrate temperature as well as continuous irradiation by working gas ions (used to overcome underdense microstructures with rough surfaces characteristic of low-temperature film growth with T_s/T_m < 0.3, where T_s and T_m: growth and melting temperature in K), which favors the formation of thermodynamically stable phases. Instead, the flux of Me₁ neutrals (ionization levels during DCMS are low)⁶ is deposited at the topmost surface atomic layer for 95%–99% of the deposition time when the substrate is electrically floating, while the ionized fraction of the Me₂-HiPIMS flux, $α= Me 2 +/( M e 2+ Me 2 +)$⁠, is accelerated by the high-amplitude (200–400 V) synchronized bias pulse (1%–5% duty cycle) and consequently subplanted under the advancing growth surface. Thus, on average, the surface is always Me₁-rich even for the cases when the time-averaged Me₂ flux exceeds that of Me₁. This intentional separation of film-forming species opens up new opportunities, particularly for the growth of supersaturated phases that can include metallic alloys (Me₁)_1−x(Me₂)_x (if experiments are conducted in pure Ar) or compound films (Me₁)_1−x(Me₂)_xA, in which A = C, O, B, or N (if reactive gases are used). Here, for simplicity, we assume that the film composition on the metal lattice is entirely determined by the metal fluxes (hence, e.g., resputtering effects are neglected). Thermodynamically favored Me₁-rich phases forming continuously at the surface serve as a template for the subsequent $Me 2 +$ implantation in the sub-surface region during the metal-ion-rich part of the HiPIMS pulse. Effective subplantation takes place in the surface-near volume determined by the $Me 2 +$ penetration depth which extends to 15–30 Å, depending on the species and incident energies involved. In this volume, the incident metal-ion flux triggers local diffusion on the cation sublattice, since the ion energy is larger than the lattice displacement threshold (∼20–50 eV depending upon the ion and film species involved). This enables the formation of supersaturated phases (see Sec. IV for examples) due to larger activation energy for bulk than for surface diffusion and, hence, insufficient mobility to cause nucleation and growth of thermodynamically favored compounds.

An additional advantage is, that in the case of highly supersaturated phases, the actual phase composition can be easily controlled by tuning the $Me 2 +$ penetration depth which is done by varying V_s amplitude (see Sec. IV C).⁵⁴ At low V_s values, $Me 2 +$ is essentially deposited at the surface and phase formation proceeds along the same paths as during DCMS, while higher V_s values increase the separation of film-forming species.

TRIDYN simulations of the light metal-ion subplantation in TM-based nitrides indicate that the incident ion energy necessary to implant ions below the high-mobility surface requires V_s amplitudes of the order of 200–400 V. As an example, the most probable and maximum ion implantation depths are plotted in Fig. 11 as a function of ion energy $E A l +$ for Al⁺ incident on the VN surface. The most probable implantation depth (corresponding to the maximum in the Al distribution profile) varies from 3.2 Å with $E A l +=60 eV$ to 8.9 Å with $E A l +=400 eV$⁠, while the maximum Al implantation depth (corresponding to the distance from the surface at which the implantation profile decreases to 10% of its maximum value) increases from 8.2 Å with $E A l +=60 eV$ to 21.6 Å with $E A l +=400 eV$⁠. Taking into account that the relaxed lattice parameter for cubic VN is 4.14 Å (Ref. 70), the Al⁺ incident energy necessary to implant the majority of incident Al⁺ below the first two to three surface layers where mobility is high (hence where Al content should be kept to minimum to avoid w-AlN formation) exceeds 150 eV. The corresponding V_s amplitude appears high from the conventional point of view (DCMS), in which case severe resputtering and gas-ion implantation are expected to take place.⁷¹ This notion is, however, not applicable for metal-ion subplantation growth as the bias is applied only during a small fraction, 1%–5%, of the deposition time.

Figure 12 is a 2D contour map showing the relative Me₂ concentration at the surface δ as a function of relative Me₂ fraction in the flux to the substrate x and Me₂ ionization degree α. δ isolines are also plotted in steps of 0.1. Critical solubilities x_max,sub that are obtained by subplantation for a given materials system characterized under DCMS growth conditions by x_max,kin, and for a given ionization of the Me₂ flux from the HiPIMS source, can readily be identified as “x” values corresponding to the cross points of δ = x_max,kin curves with the corresponding α lines. In the limit of no Me₂ ionization (DCMS like growth), α = 0 and x_max,sub = x_max,kin. With increasing α, all isolines move toward higher x values due to the subplantation effect. For example, for α = 0.5, the maximum kinetic solubility limit x_max,kin can be increased from 0.20 (α = 0) to 0.34, from 0.50 (α = 0) to 0.66, or from 0.70 (α = 0) to 0.82. Hence, independent of the solubility limit which can be achieved with conventional DCMS, the gain due to the subplantation effect is significant, although the relative increase x_max,sub/x_max,kin decreases with increasing x_max,kin. This is quite expected, as systems characterized by low x_max,kin are the ones with the highest driving forces for decomposition (active at the high-mobility surface), and, hence, separation of film-forming species by implantation is expected to give the largest effect on solubility. The benefits of subplantation obviously increase with α. In the limit of very high ionization, α = 0.9, x_max,kin can be increased from 0.20 (α = 0) to 0.71 (3.5 times), from 0.50 (α = 0) to 0.91, or from 0.70 (α = 0) to 0.96. The latter numbers indicate that the situation depicted in Fig. 12 is the ideal scenario. In reality, x_max,sub is likely significantly lower, especially at higher α values. Me₂ ionization fraction as high as 90% together with x > 0.9 implies that more than 80% of the film-forming metal flux (Me₁ + Me₂) is ionized. Hence, one can no longer consider the high-mobility zone to be restricted to the first few monolayers (MLs), as depicted in Fig. 10. Instead, highly overlapping collision cascades resulting from an intense ion irradiation extending down to tens of Å (defined by the incident metal-ion energy) are expected to provide high mobility within the region where Me₂ is subplanted, which must have a negative impact on x_max,sub.

Another critical factor during subplantation is the precise synchronization of the bias pulse to the $Me 2 +$-rich flux (cf. Sec. II). If the bias pulses are shorter than the metal-ion-rich portion of the HiPIMS pulse, a significant fraction of the $Me 2 +$ flux is not subplanted, and, hence, is deposited onto the very surface where high mobility removes kinetic constraints, resulting in the formation of thermodynamically favored phases. This results in a similar scenario as described above for the non-complete ionization of the Me₂ flux. For bias pulses extending outside the time period dominated by the $Me 2 +$ flux at the substrate, a significant contribution from high-energy gas-ion bombardment leads to the formation of point defects⁷² that may serve as nucleation sites for second-phase precipitation. In addition, energetic gas-ion irradiation results in high compressive stresses due to gas ions trapped in the interstitial sites.

In this section, we discuss cases of subplantation growth using TM-based nitride model systems. Chronologically, subplantation effects were first manifested during the growth of Ti_1−xAl_xN and Ti_1-xSi_xN films in a hybrid HiPIMS/DCMS configuration with Al⁺ or Si⁺ as the incident ion and a substrate bias applied synchronously with the metal-ion-rich portion of HiPIMS pulses.^45,50 However, the full potential of this technique has not been tested until a series of experiments involving V_1−xAl_xN and high-amplitude bias pulses (300 V or more) was conducted.⁵² That study revealed the extraordinary role of Al subplantation depth and provided an understanding of the critical parameters that govern nanostructure evolution and crystalline phase formation in metastable compounds.⁵⁴ 

Good control over the phase formation of TM nitride-based coatings is highly desired as it directly affects film performance in a variety of applications ranging from wear-protective coatings on cutting tools to wear- and corrosion-protection of components in automotive engines.^73–75 The most prominent and widely studied example is the precipitation of the thermodynamically favored w-AlN phase in metastable NaCl-structure TiAlN-based layers, which is known to negatively affect mechanical properties,^76,77 compared to a retained cubic solid solution, while cubic-structure spinodal decomposition leads to increased hardening.⁷⁶ This presents a serious challenge in the design of next-generation functional coatings with improved thermal and chemical stability, for which alloying with AlN to obtain single-phase NaCl-structure films with improved high-temperature oxidation resistance is a proven strategy.^78,79

Ti_1−xAl_xN is by far the most studied materials system in the family of TM-based nitrides. Such large research interest is driven by numerous wear-protective applications that range from cutting tools to mechanical components in the aerospace industry.^80–83 This is caused by their unique combination of properties which include relatively high hardness (typically ∼30 GPa) and good high-temperature oxidation resistance.⁸⁴ Enhanced performance in single-phase NaCl-structure compound films is connected with high AlN fractions. The growth of such layers, however, is challenging due to Ti_1−xAl_xN enthalpies of mixing being positive over the entire range of compositions, with a maximum (corresponding to the largest driving force for decomposition) at x = 0.68.⁸⁵ Nevertheless, metastable NaCl-structure compounds can be obtained by physical vapor deposition due to low-temperature kinetically limited growth and continuous dynamic low-energy ion irradiation-induced mixing in the near-surface region.⁸⁶ Typically, x_max,kin ∼ 0.50 at growth temperatures near 500 °C.^87,88

To investigate the role of Al⁺ vs Ti⁺/Ti²⁺ metal-ion irradiation on coating nanostructure, phase content, and mechanical properties two series of metastable Ti_1−xAl_xN compound films were grown in a hybrid HiPIMS/DCMS configuration (cf. Fig. 9).⁴⁵ In the first case, pulsed Al⁺ flux from a HiPIMS source was superimposed onto the continuous flux of Ti neutrals (Al-HiPIMS/Ti-DCMS). After that, the positions of elemental targets were switched to grow a second series employing Ti⁺/Ti²⁺ metal-ion irradiation and Al neutral flux (Ti-HiPIMS/Al-DCMS). Film compositions covered a wide range, 0.4 ≤ x ≤ 0.74, as determined by time-of-flight elastic recoil detection analysis (ToF-ERDA).^44,46 Negative substrate bias synchronized to HiPIMS pulses was used with the pulse amplitude V_s = 60 V and a pulse length of 200 μs. Between HiPIMS pulses (during the DCMS phase), the substrates were electrically floating at V_s = V_f = –10 V. The total Ar/N₂ sputtering pressure was maintained at 3 mTorr (0.4 Pa) with a N₂/Ar flow ratio of 0.2. All films were deposited on Si(001) substrates at T_s = 500 °C and film thickness ranged from 0.8 to 2.9 μm. Additional experimental details are given in Ref. 45.

A comprehensive summary of obtained results, emphasizing the link between the phase content evolution and mechanical properties, is shown in Fig. 13. Nanoindentation hardness H is plotted as a function of Al content on the metal lattice x for both series of Ti_1−xAl_xN layers along with XTEM images and selected area electron diffraction (SAED) patterns for two most instructive samples: Ti_0.41Al_0.59N film grown by Al-HiPIMS/Ti-DCMS and Ti_0.47Al_0.53N layer grown in the Ti-HiPIMS/Al-DCMS configuration. While XTEM images indicate that both samples exhibit dense columnar microstructures with no intercolumnar porosity, the SAED patterns reveal an essential difference in the layer phase content: the Ti_0.47Al_0.53N film deposited with Ti⁺/Ti²⁺ metal-ion irradiation contains both cubic TiN and wurtzite AlN crystallites. In contrast, the SAED pattern obtained from the Ti_0.41Al_0.59N layer grown with Al⁺ irradiation indicates single-phase NaCl-structure film. This difference is even more striking if one takes into account that the latter film has a significantly higher Al content, and, hence, a larger driving force for decomposition. Different phase content has a direct and pronounced effect on mechanical properties as illustrated by H(x) plots over the intermediate Al concentration range, 0.40 < x < 0.67. For Ti_1−xAl_xN layers grown with Al⁺ irradiation, H increases with x, reaching a maximum value of ∼30 GPa with 0.55 ≤ x ≤ 0.60 and then decreases rapidly at higher x values to ∼20 GPa as a softer w-AlN phase appears. In contrast, H(x) for films deposited with Ti⁺/Ti²⁺ ions decreases to ∼19 GPa with x > 0.4 and is approximately constant throughout the remaining concentration range.

The striking differences in the phase content and resulting mechanical properties of Ti_1−xAl_xN layers grown by Al-HiPIMS/Ti-DCMS and Ti-HiPIMS/Al-DCMS is a direct consequence of the inherent asymmetry in the ionization of target atoms during film growth, Al⁺/Ti vs Tiⁿ⁺/Al (n = 1, 2), as well as the intentional time and energy separation of film-forming species imposed by the hybrid configuration. In the former case, the low-energy Ti neutrals from DC magnetron are deposited at the film growth surface where high adatom mobility secured by elevated growth temperature and continuous low-energy (∼10 eV) gas-ion bombardment results in the formation of a dense NaCl-structure Ti_1−δAl_δN layer with $δ≪x$⁠. The latter is TiN-rich and contains a minor Al fraction δ due to the non-ionized portion of the HiPIMS flux. This layer serves as a template for energetic Al⁺ and N⁺ ions which are implanted in the sub-surface region during HiPIMS pulses, when the synchronized bias is applied, thus, enabling the formation of single-phase films with high Al content, even for the case where time-averaged Al flux exceeds that of Ti. The supersaturated solid solution does not decompose into w-AlN and cubic TiN due to the much larger activation energies for bulk vs surface diffusion as well as 20% larger molar volume of w-AlN vs that of NaCl-structure TiN. Consequently, H(x) with approximately 30 GPa is high in this composition range, 0.55 ≤ x ≤ 0.60, as a result of solid-solution hardening. Corresponding elastic moduli (not shown—see Ref. 45) range from 350 to 410 GPa. However, with increasing x, which is achieved by lowering the Ti flux from DCMS magnetron, Al concentration in the surface template layer increases due to the incomplete ionization of the Al-HiPIMS flux and the relatively low Al⁺ incident energy (∼60 eV) and eventually exceeds the kinetic w-AlN precipitation threshold (δ > x_max,kin≃ 0.5), resulting in the appearance of a second-phase wurtzite-structure AlN for x ≥ x_max,sub = 0.64.

In contrast, during Ti-HiPIMS/Al-DCMS Ti_1−xAl_xN film growth, Al arrives as neutrals, while the Ti ion flux contains a large Ti²⁺ fraction (see Fig. 8, $F T i 2 +/ F T i +=0.42$ during the most energetic phase of the high-power pulse) with the mean energy exceeding 140 eV.⁸⁹ As a consequence, the Al concentration in the near-surface region δ, subject to continuous intermixing, is relatively high and eventually exceeds x_max,kin ≃ 0.5, even though the time-averaged Ti flux is still higher than that of Al. Thus, the thermodynamically favored w-AlN phase is formed for x < 0.5, which is even lower than that for DCMS (x_max,kin ∼ 0.5).^87,88

The possibility of growing single-phase NaCl-structure Ti_1−xAl_xN films with high Al content by using a hybrid Al-HiPIMS/Ti-DCMS approach together with Al⁺-synchronized substrate bias enables an improvement of high-temperature oxidation resistance, which scales with x (up to x ≃ 0.7)⁹⁰ without sacrificing mechanical quality.

Another widely studied TM-based nitride system is Ti_1−xSi_xN.^91–94 In this case, a nanocomposite structure forming as a result of self-organization during film growth gives rise to superhard coatings with H > 40 GPa.^95–98 The metastable SiN solubility in NaCl-structure TiN is significantly smaller than for AlN in TiN.⁹⁹ At higher film growth temperatures with Si fractions on the cation lattice above ∼15%, there is strong SiN_z (here z denotes the N fraction in the SiN phase) surface segregation giving rise to a nanostructure consisting of TiN-rich crystallites encapsulated by a few monolayers (MLs) of a disordered SiN_z tissue phase.⁹³ This occurs dynamically as Si segregation to the surface of TiN-rich grains to precipitate SiN_z in turn drives Ti segregation away from the surface of the just-formed SiN_z, to renucleate TiN, given a buffered N for each reaction.^93,100 The stepwise process is repeated throughout the film deposition, while effectively promoting a smooth film surface and eliminating columnar formation. The encapsulation layers also constrain further growth of the TiN-rich nanograins. The maximum SiN content in metastable supersaturated cubic solution Ti_1−xSi_xN is 9%–14%.^99,101

Two sets of Ti_1−xSi_xN_y films were grown with metal-ion irradiation being Si⁺/Si²⁺ (Si-HiPIMS/Ti-DCMS)⁵⁰ and Ti⁺/Ti²⁺ (Ti-HiPIMS/Si-DCMS). Ti_1−xSi_xN_y compositions in the range 0 ≤ x ≤ 0.57 were obtained by varying power to the DCMS magnetron, while keeping the HiPIMS settings constant to ensure the same ionization of the sputtered flux. The system base pressure was <2.3 × 10⁶ Torr (0.3 mPa), and the total pressure p during deposition was 3 mTorr with a N₂/Ar flow ratio of 0.2. The substrate temperature T_s was 500 °C. A 200-μs-long negative pulsed substrate bias, with amplitude V_s = 60 V synchronized to the entire HiPIMS pulse, was used in all experiments. Between HiPIMS pulses, the substrate was at floating potential, V_f = –10 V. More experimental details are given in Ref. 50.

Typical XTEM and plan-view STEM micrographs with corresponding SAED patterns and EDX elemental maps of Si-HiPIMS/Ti-DCMS and Ti-HiPIMS/Si-DCMS Ti_1−xSi_xN_y films with essentially the same SiN content, x = 0.25 ± 0.01, are shown in Fig. 14 along with the H(x) plots for both film series.⁵⁰ In the case of x = 0.26, Ti-HiPIMS/Si-DCMS film plan-view elemental EDX/STEM map of the spatial Ti, Si, and N distribution acquired from the outlined area reveals SiN_z-rich layers encapsulating TiN-rich columns in agreement with reports for Ti_1−xSi_xN_y films with lower Si fractions, 0.05 ≤ x ≤ 0.14, grown by cathodic arc evaporation.¹⁰² 

In contrast, cross-sectional EDX elemental distribution maps obtained from the Si-HiPIMS/Ti-DCMS x = 0.24 sample together with a lattice-resolution STEM image obtained along the 110 zone axis from a single NaCl-structure column reveal no indication of a SiN_z tissue phase, not even at the sub-nm scale. The collage of Ti and Si elemental EDX maps superimposed onto the lattice-resolution STEM image provides evidence that Ti_0.76Si_0.24N films grown under Si⁺/Si²⁺ ion irradiation from the Si-HiPIMS source are a NaCl-structure with local sub-nm scale cubic Si-rich and Ti-rich regions. Differences between layers grown with Ti⁺/Ti²⁺ or Si⁺/Si²⁺ ion irradiation are confirmed by the evolution of film elemental compositions and relaxed lattice parameters with increasing x.⁵⁰ 

The difference in nanostructure and phase content has a direct effect on film mechanical properties; Ti-HiPIMS/Si-DCMS Ti_1−xSi_xN_y nanocomposite films exhibit superhardness over a composition range, 0.04 ≤ x ≤ 0.26, that is significantly wider than previously reported. In contrast, single-phase NaCl-structure Si-HiPIMS/Ti-DCMS layers are softer, with H around 14 GPa in the corresponding compositional range 0.03 ≤ x ≤ 0.24, which is a consequence of a decreased elastic modulus.

Also for this materials system, the crystalline phase content and nanostructure of the resulting films is clearly determined by the type of the incident metal-ion irradiation and separation of the film-forming species, Tiⁿ⁺ and Si vs Siⁿ⁺ and Ti (n = 1, 2), during film growth. With the former target configuration (Ti-HiPIMS/Si-DCMS), Si arrives as neutrals, while the Ti flux is primarily ions with an intense Ti²⁺ component, leading to a high Si concentration at the surface which results in the growth of thermodynamically favored phase-separated nanostructures consisting of TiN crystallites encapsulated in a SiN_z matrix. Hence, the film growth scenario is analogical to that observed for Ti_1−xAl_xN layers obtained in the Ti-HiPIMS/Al-DCMS configuration for which a high tendency toward forming thermodynamically stable phases (c-TiN and w-AlN) is observed at relatively low Al concentrations (i.e., lower than during conventional DCMS growth).⁴⁵ 

In contrast, no nanocomposite formation takes place during Si-HiPIMS/Ti-DCMS Ti_1−xSi_xN_y film growth for any x value tested. Instead, the NaCl-structure forms with a dynamic concentration gradient in the Ti_1−δSi_δN surface layer (⁠ $δ≪x$⁠) due to the Ti neutrals from the DCMS source and the non-ionized fraction of the Si flux from the HiPIMS target. With increasing the Si/Ti flux ratio (realized here by decreasing the Ti-DCMS power), effective Siⁿ⁺ implantation below the high-mobility mixing layer into the cubic Ti_1−δSi_δN template results in single-phase films being formed with the Si content far above the kinetic solubility limit x_max,kin. Here, the significant Si²⁺ ion fraction, which is characterized by a longer implantation range due to high incident energy exceeding 2 × V_s, plays an important role. This scenario is active as long as δ < x_max,kin. With further increasing the Si/Ti flux ratio, eventually the latter condition is no longer satisfied and SiN_x begins to form at the surface, which limits the maximum solubility achievable by Si subplantation x_max,sub. Under the experimental conditions used in Ref. 50  x_max,sub = 0.24 which is 1.7–2.7 times higher than solubilities reported for conventional experiments involving film growth by DCMS or by cathodic arc evaporation with x_max,kin in the range 0.09–0.14.^97,101

The most thorough and systematic study to date of subplantation effects concerns V_1−xAl_xN.^51,52,54 It represents a cubic-phase system, which is thermodynamically immiscible and much less investigated than Ti_1−xAl_xN and Ti_1−xSi_xN, with only a few publications,^103–105 yet interesting for its combination of high hardness and low friction coefficient.¹⁰³ 

In all experiments (Refs. 51, 52, and 54), V_1−xAl_xN films were grown on Si(001) substrates mounted symmetrically with respect to the 8.8 × 50 cm² V and Al targets arranged in a co-sputtering geometry such that the angle between the substrate normal and the normal to each target was ∼28°, with a target-to-substrate distance of 18 cm. Two Advanced Energy Pinnacle Plus power supplies were used for DCMS co-sputtering. In addition, two external Melec SIPP2000USB-10-500-S pulsers with 10 kW ADL GX 100/1000 DC power supplies were employed for HiPIMS co-sputtering and pulsed substrate bias. The system base pressure was less than 5.6 × 10⁻⁶ Torr (0.75 mPa), the substrate temperature was 500 °C, and the total pressure during deposition was p = 3 mTorr (0.42 Pa) with the nitrogen flow fraction N₂/(N₂ + Ar) varied from 0.29 to 0.32. Following film deposition, the vacuum chamber was only vented after the substrate temperature was less than 180 °C, in order to control the film surface chemistry upon air exposure.¹⁰⁶ Additional experimental details are given in Ref. 52.

In the initial experiments aiming at the comparison of Al⁺ vs V⁺/V²⁺ effects on the nanostructure,⁵² phase content, and related physical properties of V_1−xAl_xN films (Al-HiPIMS/V-DCMS vs V-HiPIMS/Al-DCMS), the negative synchronous substrate bias with the amplitude of 100 V and 200 μs length was used, while the substrate was electrically floating at –10 V between HiPIMS pulses, as revealed by the oscilloscope waveforms recorded during film growth (see Fig. 1 in Ref. 52). In the following papers devoted to studies of Al⁺ subplantation effects,^51,54 V_s was precisely synchronized to the Al⁺-rich fraction of each HiPIMS pulse using as an input the results of in situ ion mass spectrometry studies conducted at the substrate position. Hence, in those papers, 100-μs-long bias pulses starting 30 μs after the ignition of each HiPIMS pulse were used corresponding to the time interval during which the metal/gas-ion flux ratio at the substrate was at its maximum.⁵⁴ 

Figure 15 shows XTEM images along with the corresponding SAED patterns obtained from Al-HiPIMS/V-DCMS and V-HiPIMS/Al-DCMS V_1−xAl_xN films with essentially the same Al content, x = 0.57 ± 0.01. In addition, the H(x) plots for both sample series grown with either Al⁺ or V⁺/V²⁺ ion irradiation are shown.⁵² Both films have a dense columnar nanostructure with no open voids between columns. The average column width is 40 ± 10 nm for the Al-HiPIMS V_0.42Al_0.58N layer and 35 ± 15 nm for the V-HiPIMS V_0.43Al_0.57N film. However, SAED obtained from the former sample exhibits only 111, 002, and 220 NaCl-phase reflections, while in the case of the V-HiPIMS film, $11 2 ¯0$ and $10 1 ¯0$ w-AlN peaks are also detected.

A difference in the phase content between both V_1−xAl_xN film series is reflected in their mechanical properties, as revealed by the plots of nanoindentation hardness H(x) over the Al concentration range, 0.15 < x < 0.75. Films grown with V⁺/V²⁺ ion irradiation exhibit high H values up to x = 0.51 (29.6 ± 1.8 GPa), after which a sharp decay is observed to 19.9 ± 1.3 GPa with x = 0.57 due to a softer w-AlN phase being present (cf. SAED in Fig. 15), and stays low at 14.7 ± 2.2 GPa for x ≥ 0.61. In the case of V_1−xAl_xN layers grown with Al⁺ irradiation, the hardness decay takes place at much higher Al concentrations. H is very high at 30.2 ± 1.2 GPa with x = 0.58, i.e., for essentially the same Al content for which a sudden softening is observed if the heavier metal-ions Vⁿ⁺ (n = 1, 2) are used instead (V-HiPIMS/Al-DCMS series). Even with x = 0.64, hardness is only marginally lower at 27.8 ± 0.2 GPa. At the Al content, x = 0.72 and H = 20.3 ± 2.0 GPa, which is similar to the film grown with V⁺/V²⁺ irradiation with x = 0.57. The elastic modulus E(x) (not shown) exhibits similar trends to those observed for H(x), which can be exemplified by the fact that E = 398 ± 18 GPa for the Al-HiPIMS/V-DCMS V_0.42Al_0.58N layer and only 302 ± 26 GPa for the V-HiPIMS/Al-DCMS V_0.43Al_0.57N film.⁵² As a consequence, the H³/E² ratio is an important parameter that reflects resistance against plastic deformation,^107,108 differs significantly between the two layers, and amounts to 0.18 ± 0.003 and 0.08 ± 0.01, respectively.

The H(x) plots shown in Fig. 15, together with the E(x) trends (not shown), mirror the changes in phase content revealed by detailed XRD analyses performed as a function of the sample tilt angle (the angle between the surface normal and the diffraction plane containing the incoming and diffracted x-ray beams).⁵² The latter experiments allowed to estimate the kinetic solubility limits for Al in the NaCl-structure VN at 0.55 and 0.63 for V-HiPIMS and Al-HiPIMS, respectively. The inherent asymmetry in the ionization of target atoms during film growth, Al⁺/V vs Vⁿ⁺/Al, as well as the intentional time and energy separation of film-forming species utilizing both the hybrid configuration and the synchronized substrate bias potential are key factors to explain the differences obtained regarding phase formation and mechanical properties.

A point to emphasize here is that the V_1−xAl_xN experiments were conducted in a different sputtering system to that used to grow Ti_1−xAl_xN and Ti_1−xSi_xN films described above. Both cathode configuration and power supplies were different. In particular, the larger size of the capacitor bank used during the V_1−xAl_xN experiments resulted in HiPIMS pulses with a constant voltage throughout the entire pulse length, which was not the case during Ti_1−xAl_xN and Ti_1−xSi_xN depositions. The fact that the type of metal-ion irradiation has a decisive influence on the deposited layers independently of the HiPIMS pulse shape (as long as the peak current density is sufficiently high to ensure significant ionization of the sputter-ejected flux) or cathode configuration, reveals the fundamental nature of described phenomena and demonstrates the versatility of the novel film growth method.

Convincing evidence for the Al⁺ subplantation effects in V_1−xAl_xN was obtained from a set of experiments employing in situ time-resolved ion mass spectrometry analyses of ion fluxes to the substrate position during Al-HiPIMS/V-DCMS sputtering.⁵⁴ The study revealed that the metal/gas-ion flux ratio at the substrate is the highest during the time interval from 30 to 130 μs after the ignition of the HiPIMS pulse,⁵⁴ which allowed for precise synchronization of the substrate bias pulse. Hence, the growing film was set at V_s only during the time Al⁺ dominated the incident ion flux, while at all other times, i.e., when gas-ion irradiation was dominant, the substrate was electrically floating at approximately –10 V.

Figure 16 is a two-dimensional map of the XRD-determined phase content (cubic, cubic + hexagonal, and hexagonal) for V_1−xAl_xN films, plotted with x varied from 0.53 to 0.80 and 15 ≤ |V_s| ≤ 400 V.⁵⁴ The dashed curve indicates the maximum Al concentration that can be accommodated in the single-phase NaCl-structure x_max,sub and separates the regions of single-phase cubic films (solid squares) from two-phase layers (open triangles) containing both cubic-VAlN and w-AlN grains. Interestingly, x_max,sub exhibits a dependence on the amplitude of the synchronous negative bias pulse and increases from 0.59 with |V_s| = 15 V to 0.66 at |V_s| = 100 V, 0.72 at |V_s| = 200 V, 0.75 at |V_s| = 300 V, and 0.77 with |V_s| ≥ 400 V. The Al implantation effect saturates at higher V_s values as revealed by the slope of the x_max,sub(V_s) curve.

The observed changes in kinetic solubility limits are a direct consequence of varying Al implantation depth which is controlled with the amplitude of applied bias pulse, as depicted in Fig. 11 where the results of TRIDYN-simulated Al implantation in VN are summarized. At low incident energy corresponding to lower V_s values, the majority of the incident Al⁺ flux is essentially deposited at the high-mobility surface zone (together with the non-ionized portion of the HiPIMS flux), which results in a relatively low x_max,sub, only slightly higher than that obtained with DCMS growth (x_max,kin = 0.52).⁵² With increasing Al⁺ energy (increasing V_s amplitude), the implantation depth increases, resulting in a better separation between film-forming species. The flux of V neutrals deposited at the surface provides the cubic VN-rich template for Al subplantation which occurs at greater depths, below the high-mobility zone, thus mitigating constrains imposed by the laws of thermodynamics. Eventually, in the limit of the highest V_s values tested, all Al⁺ is implanted and x_max,sub saturates at the value determined to a large extent by the ionization degree of the Al-HiPIMS flux and the Al/V flux ratio, as discussed in Sec. 3.2 (cf. Fig. 12). Importantly, the growth of the NaCl-structure V_1−xAl_xN template with x < 0.5 at the surface requires that the incident V neutral flux be significantly larger than the non-ionized portion of the Al-HiPIMS flux. If this condition is not satisfied such that the Al concentration in the surface layer exceeds that of V, second-phase w-AlN is formed (cf. Eq. (3)).

The results presented in Fig. 11 imply that metal-ion-synchronized HiPIMS/DCMS film growth allows for a wide-range control over the phase formation in V_1−xAl_xN compounds simply by varying the V_s amplitude. This is especially valid for Al concentrations in the range x_max,kin < x < x_max,sub, i.e., for supersaturated films where Al implantation depth plays a key role. A good illustration of this is shown in Figs. 17(a) and 17(b) where two sets of tilt angle-dependent XRD θ‐2θ scans are displayed for V_0.26Al_0.74N films grown with V_s of either –60 V or –300 V.⁵³ The only diffraction peaks in the former case are those from w-AlN, indicative of a XRD-single-phase film. With increasing |V_s| to 300 V, while keeping all other growth parameters constant, single-phase NaCl-structure layers are obtained. Such complete control over the phase content is not achieved if x > x_max,sub. Still, the effects of Al subplantation are very clear as, for example, in the case of Ti_0.31Al_0.69N films for which XRD data shown in Figs. 17(c) and 17(d) indicate that the relative cubic-phase content varies from 9% with V_s = –60 V to 98% for films grown with V_s = –300 V.⁵³ 

Al subplantation is expected to give no benefits for x < x_max,kin since in this composition range single-phase NaCl-structure films are grown irrespective of whether Al is deposited at the surface or subplanted at greater depths.

The precise bias pulse synchronization to the metal-ion-rich portion of the HiPIMS pulse, determined from the time-resolved ion mass spectrometry analyses conducted at the substrate plane, is very critical for efficient metal-ion subplantation. As shown in Fig. 18 in the case of V_0.26Al_0.74N films, too narrow (50–80 μs) or too broad (30–160 μs) bias pulses result in the wurtzite AlN precipitates. As the length of the V_s pulses become shorter than the duration of the Al⁺-ion-rich portion of the HiPIMS pulse, the fraction of the Al⁺ flux subplanted below the near-surface ion-mixing layer is reduced, while non-accelerated Al⁺ ions that arrive outside the 50–80 μs bias window (thus at V_f,) are deposited in the near-surface region, resulting in the precipitation of w‐AlN. In the case of longer V_s pulses extending into the gas-ion-rich region of the HiPIMS pulse, the majority of Al⁺ ions are subplanted, but so are also gas ions (primarily Ar⁺ and N₂⁺). The high-energy gas-ion irradiation results in the creation of point defects⁷² and defect complexes, which serve as nucleation sites for the formation of w-AlN precipitates, even at relatively low AlN concentrations. Thus, detuning the synchronous bias by as little as 30 μs (in either direction) has an immediate effect on the phase content of metastable TMN layers.

The benefits of precise synchronization of the substrate bias pulse to the Al⁺-rich fraction of the ion flux over the application of 200-μs-long synchronous bias pulses are clear: the solubility limit is increased from 0.63 with V_s = –100 V (0–200 μs) to ∼0.77 with V_s = –400 V (30–130 μs). This difference has large implications for the mechanical properties of V_1−xAl_xN layers as shown in Fig. 19 in which (a) H(x) and (b) H³/E²(x) are plotted for both biasing scenarios together with the results for layers grown with DCMS. The superior effects of Al⁺ subplantation are evident: H = 29.7 ± 1.1 GPa with x = 0.74 and decays slowly with a further increase in Al concentration to reach 28.3 ± 0.3 GPa with x as high as 0.84. For comparison, in the case of conventional DCMS growth with –100 V DC bias, H does not exceed 28 GPa for x > 0.45. Similar dramatic improvement is also observed for H³/E², which for V_1−xAl_xN films grown by the subplantation technique is at 0.29 with x = 0.80 and 0.31 with x = 0.84. In sharp contrast, for DCMS layers, H³/E²= 0.02 ± 0.01 with x > 0.51 and does not exceed 0.15 for any Al concentration.

Until recently thin film growth by magnetron sputtering relied on the use of gas-ion irradiation to increase film density at low deposition temperatures. The inherently low ionization degree of the sputtered flux during DCMS, owing to relatively low plasma densities involved, prevented the utilization of metal-ions for affecting nanostructure evolution, phase formation, and, hence, film property optimization. Following the invention of HiPIMS, significant fluxes of metal-ions can be utilized during growth and, in addition, the collateral damage of noble gas-ion irradiation can be excluded. The latter is possible due to extensive gas rarefaction effects, stemming from very high temporal material fluxes from the target, which lead to large variations in the metal/gas-ion ratio in the flux incident at the growing film during HiPIMS pulses. To make full use of this phenomenon, the substrate bias is best applied only during the time when the metal/gas-ion ratio at the substrate is at maximum, using input from time-resolved ion mass spectrometry analyses, while at all other times the substrate is left electrically floating (so called “metal-ion-synchronized HiPIMS”). As a consequence, the film growth process is to a large extent determined by the flux of metal-ions with the energy and momentum easily controlled by the amplitude of the synchronous bias pulses. This has a number of benefits as metal-ions, in contrast to noble gas ions, are film-forming species, and, hence, they are primarily incorporated at lattice sites resulting in much lower compressive stress levels, as noble gas trapping at the interstitial sites is avoided. In addition, metal-ions provide more efficient energy and momentum transfer to the growing film surface due to better mass match with the film-forming species, which allows for yet lower growth temperatures.¹⁰⁹ 

In recent years, we performed in-depth systematic studies of the metal-ion irradiation effects on thin films using Groups IVB and VIB transition metal nitrides as model systems. It was established that the combination of target material properties like mass, ionization potentials (with respect to that of sputtering gas), sputter yields, and reactivity toward the reactive component of the sputtering gas, together with the selected sputtering parameters (primarily the peak target current density which affects ionization degree), determines the time evolution of metal- and gas-ion fluxes through their impact on the gas rarefaction.

While metal-ion-synchronized HiPIMS enables new film growth pathways during deposition from a single target, it offers even more possibilities if combined with a second source operating in the DCMS mode (a hybrid HiPIMS/DCMS co-sputtering). In such a configuration, the film growth proceeds from the continuous near-thermal fluxes of neutrals supplied by the DCMS source, which are deposited at the high-mobility surface, and from the pulsed fluxes of metal-ions from HiPIMS magnetron that are intentionally made energetic by the application of synchronous bias pulses.

The metal-ion-synchronized film growth experiments in a hybrid HiPIMS/DCMS configuration involving light metal-ions (Al⁺ or Si⁺) revealed the key role of the ion subplantation mechanism for achieving unprecedented solubility limits in metastable TM-based nitrides. An intentional separation of film-forming species originating from HiPIMS and DCMS sources in the time and energy domain opens up new opportunities. This is particularly so for the growth of supersaturated phases, which can include either metallic alloys (Me₁)_1−x(Me₂)_x for experiments conducted in pure Ar or compound films (Me₁)_1−x(Me₂)_xA, in which A = C, O, B, or N, if reactive gases are used. The impact of this processing concept for materials systems other than nitrides and other metal-ion-based growth techniques (in place of HiPIMS), however, remains to be demonstrated.

The implementation of the metal-ion-synchronized HiPIMS/DCMS growth on the industrial scale must meet the challenge of precise control over the metal-ion time-of-flight from the target to the substrate, especially for applications in which the target–substrate separation varies continuously (e.g., cutting tool industry). In the latter case, the HiPIMS pulse-to-bias-pulse timing adds to the coating nonuniformity challenge arising from the complex substrate rotation schemes.^110–112 Computer simulations of the same type as presented in Ref. 110 may be used to optimize the effective deposition time when substrates are in-phase with the metal-ion irradiation for a given coating system setup (target arrangement, turntable gear ratio, substrate position, etc.). While it is clear that the full benefits offered by synchronized metal-ion irradiation require a thorough redesign of the coating units, further experiments are necessary to determine how large a fraction of the total deposition time the substrates must, in fact, spend in-phase with the metal-ion flux in order to gain positive effects. The recent results indicate the significant advantage of W⁺ ion irradiation for the growth of Ti_0.40Al_0.27W_0.33N nanocomposite films with no external heating despite the continuously varying target–substrate distance.¹¹³ In many other application areas, e.g., large area glass coating for flat-panel displays, with the constant target–substrate separation, the metal-ion subplantation effects can be a game changer that does not require hardware modifications.

The authors most gratefully acknowledge the financial support of the Knut and Alice Wallenberg Foundation Scholar Grant (No. KAW2016.0358), the VINN Excellence Center Functional Nanoscale Materials (FunMat-2) Grant (No. 2016-05156), the Swedish Research Council VR Grant (No. 2018-03957), the VINNOVA Grant (No. 2019-04882), the Åforsk Foundation Grant (No. 16-359), and the Carl Tryggers Stiftelse Contract (No. CTS 17:166). J.M.S. most gratefully acknowledges funding from the German Research Foundation (DFG) within No. SFB-TR 87.