Ion kinetic energy-and ion flux-dependent mechanical properties and thermal stability of (Ti,Al)N thin films

Ion-irradiation-induced changes in structure, elastic properties, and thermal stability of metastable c-(Ti,Al)N thin films synthesized by high-power pulsed magnetron sputtering (HPPMS) and cathodic arc deposition (CAD) are systematically investigated by experiments and density functional theory (DFT) simulations. While films deposited by HPPMS show a random orientation at ion kinetic energies ( E k ) > 105 eV, an evolution towards (111) orientation is observed in CAD films for E k > 144 eV. The measured ion energy flux at the growing film surface is 3.3 times larger for CAD compared to HPPMS. Hence, it is inferred that formation of the strong (111) texture in CAD films is caused by the ion flux-and ion energy-induced strain energy minimization in defective c-(Ti,Al)N. The ion energy-dependent elastic modulus can be rationalized by considering the ion energy-and orientation-dependent formation of point defects from DFT predictions: The balancing effects of bombardment-induced Frenkel defects formation and the concurrent evolution of compressive intrinsic stress result in the apparent independence of the elastic modulus from E k for HPPMS films without preferential orientation. However, an ion energy-dependent elastic modulus reduction of ~18% for the CAD films can be understood by considering the 34% higher Frenkel pair concentration formed at E k = 182 eV upon irradiation of the experimentally observed (111)-oriented (Ti,Al)N in comparison to the (200)-configuration at similar E k . Moreover, the effect of Frenkel pair concentration on the thermal stability of metastable c-(Ti,Al)N is investigated by differential scanning calorimetry: Ion-irradiation-induced increase in Frenkel pairs concentration retards the wurtzite formation temperature by up to 206 ◦ C.


Introduction
Metastable cubic titanium aluminum nitride (c-(Ti,Al)N, space group Fm3m, NaCl prototype structure) [1,2] is the current benchmark coating material system in cutting and forming [3,4] as well as in oxidation protection applications [1,5,6].The importance of (Ti,Al)N for such applications originates from the combination of high hardness and elasticity [2], low wear [1], large thermal stability [7,8], and superior oxidation resistance [9].These unique properties are enabled by the formation of metastable cubic (Ti,Al)N solid solutions which require kinetically limited low-temperature growth conditions [10].The state-of-the-art physical vapor deposition techniques such as high power pulsed magnetron sputtering (HPPMS) [11,12] and cathodic arc deposition (CAD) [13,14] are frequently employed to synthesize c-(Ti,Al)N.
Due to the high power density in HPPMS [15] and CAD processes [14], high-energy ionized particle fluxes are directed towards the growing film surface.Additionally, a substrate bias potential ranging from tens to hundreds of volts is often applied to control the ion kinetic energy (E k ) during film growth.Therefore, structural defects are expected to form and annihilate, of which some are preserved during growth.In transmission electron microscopy (TEM) studies by Hultman et al. [16,17] of TiN thin films (isostructural to (Ti,Al)N), it has been proposed that dislocation loops observed on the (111) plane are the primary defect types and that they are formed by agglomeration of ion irradiation-induced point defects [17].However, the energetics of forming these defects preferentially on the (111) planes were not discussed.Since then, follow-up studies showed elastic anisotropy of cubic transition metal nitrides (c-TMN) [18] and cubic transition metal aluminum nitrides (c-(TM,Al)N) [19,20].For c-TiN, the largest elastic modulus has been predicted along (200) orientation and the lowest along (111) orientation.However, increasing the Al-content strengthens cubic (Ti,Al)N along (200) according to the previous theoretical work [19].Dependence of other physical properties of c-(Ti,Al)N on preffered orentation remains unfathomed even though it can be learned from other systems that for example thermoelectricity [21], magnetism [22], electrical, and optical properties [23], are affected.
In recent years, formation and stability of different types of point defects such as vacancies [8,[24][25][26][27][28], interstitials [24,26], Schottky defects [26], and Frenkel pairs [25,[27][28][29] in c-TMN and c-(TM,Al)N have been investigated.Point defects cause significant alterations in the structural and mechanical properties of nitrides.For example, vacancies as the lowest-energy defects [26] stabilize the mechanically unstable c-MoN [30,31] and WN [32], reduce the elastic modulus of Ti 0.5 Al 0.5 N [24], and increase the elastic modulus of NbN [33].In addition to vacancies, Frenkel pair defect types have also been shown to energetically stabilize the cubic structure of binary MoN [34].By correlating experiments and ab initio simulations for c-(Cr,Al)N [29], c-(V,Al)N [27], and c-(Ti,Al)N [28], it has been demonstrated that Frenkel pairs and vacancies form during high-energy ion irradiation.The kinetic energy E k determines the rate of point defect formation and annihilation.The concentration of Frenkel defects and vacancies also directly impacts the physical properties of ternary nitrides such as stress state and elastic modulous, as it has systematically been investigated for c-(V,Al)N [27].
At elevated temperatures, metastable c-(Ti,Al)N has been reported to exhibit age hardening [7] within the temperature range of 600-1000 • C, which can be explained by spinodal decomposition of the metastable solid solution into isostructural c-TiN-and c-AlN-rich domains [35].Further increase in annealing temperature results in the transformation of metastable c-AlN-rich domains into the thermodynamically stable wurtzite (w-)AlN phase [7], which is accompanied by an undesirable volume increase of 22% [36] and hardness reduction up to 35% [37].While many parameters affect the thermal stability of c-(Ti,Al)N such as the Al [37] and N [8,38] content, microstructure [39,40], stress state [41,42], and synthesis processing route [8,38,43], the effect of point defects on thermal stability has been just recently investigated [8,38,44].The focus in these studies was mainly on the presence of vacancies which accommodate the thin film off-stoichiometry.For magnetron sputtered (Ti,Al)N x [8], minimization of the vacancy concentration resulted in enhanced thermal stability.In contrast, for cathodic arc-deposited (Ti,Al)N x thin films [44], the highest thermal stability was obtained for thin films with N vacancies.Recently, Holzapfel et al. [28] have investigated the effect of ion irradiation-induced Frenkel pair defects, generated within the E k range from 64 to 91 eV, on the thermal stability of c-(Ti,Al)N deposited by CAD.It has been demonstrated that the presence of Frenkel pairs formed by ion irradiation enhances the thermal stability by retarding the formation of the wurtzite solid solution.This enhancement was attributed to the much larger activation energy for bulk diffusion of Al in the presence of Frenkel pairs, based on ab initio predictions [28].
Here, we systematically investigate the effect of ion irradiation on the structure evolution, mechanical properties, and thermal stability of c-(Ti,Al)N thin films deposited by different processing routes, HPPMS and CAD, within the ion kinetic energy range of 19-236 eV.The experimental findings are correlated with ion irradiation simulations of c-(Ti,Al)N using the thermal spike model [27][28][29].It is demonstrated that the ion irradiation-induced changes in elastic modulus and thermal stability of c-(Ti,Al)N thin films deposited by HPPMS and CAD can be understood based on the ion kinetic energy and, for the first time, based on the orientation-dependent formation of Frenkel defects.

Experimental details
(Ti,Al)N thin films were deposited by two physical vapor deposition (PVD) processing routeshigh power pulsed magnetron sputtering (HPPMS) and cathodic arc deposition (CAD) -at comparable ion kinetic energy (E k ) ranges to explore the effect thereof on the evolution of structure, mechanical properties, and thermal stability.
For the first series of (Ti,Al)N thin films, a HPPMS process was used in the industrial deposition system CemeCon CC 800/9 (Würselen, Germany) employing a Melec SIPP2000USB-10-500-S pulser in combination with a 10 kW ADL GX 100/1000 DC power supply.A rectangular magnetron with a Ti 0.5 Al 0.5 composite target (> 99.7% purity, Plansee Composite Materials GmbH, Germany) with the dimensions 8.8 × cm 2 was used for the depositions in a stationary geometry at a target-tosubstrate distance of 10 cm.α-Al 2 O 3 (0001) substrates (1 × 1 cm 2 ) and Fe foils (99.5% purity) were mounted on a Cu holder facing the target.Prior to depositions, the substrates were heated to the deposition temperature (T s ) of ~450 • C and etched in Ar atmosphere at 1.0 Pa for 30 min, using a mid-frequency Advanced Energy Pinnacle Plus generator operating at a voltage of -650 V with a frequency of 250 kHz and pulse off-time of 1600 ns.The chamber base pressure at the deposition temperature was below 0.75 mPa.A total pressure of 0.43 Pa was employed for sputtering with a constant N 2 /(Ar+N 2 ) flow ratio of 0.13.The HPPMS power supply was operated with a pulse on-time of τ HPPMS = 50 μs at a duty cycle of 2.5% corresponding to a frequency of f = 500 Hz.A timeaveraged power of 2500 W was used, which resulted in the peak power density of 630 W/cm 2 .A pulsed substrate bias potential V s was varied in the range from -20 V (floating) to -200 V with a pulse on-time of τ s = 100 μs, synchronized to the target HPPMS pulse with a +30 μs phase shift, see Fig. S1 of the supplementary materials.The substrate bias phase shift was calculated considering the formation time of metal ions within each HPPMS pulse and the time-of-flight of the ions from the target to the substrate [45].Therefore, majority of the ions accelerated towards substrate are expected to be film-forming species rather than Ar + .The substrates were kept electrically floating between the HPPMS pulses.The deposition time was 90 min resulting in films with a thickness of ~2.2 μm.
The second series of (Ti,Al)N thin films was deposited by CAD using the industrial deposition plant Ingenia P3e TM (Oerlikon Surface Solutions, Liechtenstein).α-Al 2 O 3 (0001) substrates (1 × 1 cm 2 ) and Fe foils (99.5% purity) were mounted onto the sample holder.When a base pressure < 3 × 10 − 4 Pa was reached, the substrates were heated to T s = 450 • C and etched at a pressure of 0.5 Pa in an Ar/H 2 atmosphere for min.The thin films were deposited in a stationary geometry with a cathode-to-substrate distance of 16 cm.A Ti 0.5 Al 0.5 composite cathode was used.The N 2 pressure in the deposition chamber was kept constant at 8.0 Pa.A DC substrate bias potential was varied between -11 V (floating) and -200 V.The deposition time was 20 min resulting in films with a thickness of ~3.5 μm.The venting temperature was always < 100 • C for both processing routes to avoid surface chemistry modification due to air exposure [46].
The Fe foil substrates were etched in nitric acid solution with HNO 3 / deionized H 2 O volume ratio of 1/5 and subsequently dried.Then, the thin film flakes were milled into a powder using a mortar, and powder samples corresponding to thin film samples (on the Al 2 O 3 substrates) were produced.
An energy-resolving mass spectrometer (PPM 422, Pfeiffer Vacuum, Asslar, Germany) was employed to characterize the plasma chemistry through a mass-to-charge scan at a fixed energy.Time-averaged ion energy distribution functions (IEDFs) of the most dominant positive ions, identified in the mass scans, were subsequently obtained through energy scans at fixed mass-to-charge ratios.The spectrometer orifice with a diameter of 100 μm was grounded in all measurements.For the analysis of the HPPMS discharge, the measurements were performed under the same conditions used for the film growth, but without heating.For the CAD discharge, the data were collected at the same pressuredistance product of 128 Pa cm as in the growth condition of the thin films.
The energy fluxes of the individual ions are calculated using: in which E is the ion energy.The kinetic energy of ionized species arriving at a biased growing film can be expressed as: where E ∘ i denotes the initial kinetic energy of the ions prior to entering the substrate sheath.The kinetic energy gained from acceleration across the substrate sheath is qe(V plasma − V s ), where q represents the ion charge state, e is the elementary charge, V plasma is the plasma potential, and V s is the negative substrate bias potential.
The chemical composition of the thin films was characterized by time-of-flight elastic recoil detection analysis (ToF-ERDA) at the Tandem Laboratory of Uppsala University [47].36 MeV 127 I 8+ primary ions were directed onto the sample at an angle of 22.5 • with respect to the surface.Scattered and recoiling target particles were detected at 22.5 • relative to the sample surface, in IBM geometry.The detection telescope was consequently located at 45 • with respect to the primary ion beam and the time-of-flight was measured using thin C foils [48].Depth profiles were obtained from time-energy coincidence spectra by using the CONTES code [49].Total maximum measurement uncertainties, obtained by combination of ToF-ERDA and Rutherford backscattering spectrometry (RBS) was 4% relative deviation of the deduced values for N and aliquot fractions for Ti and Al [28].
Structural analysis of the thin films were performed with a Siemens D5000 X-ray diffraction (XRD) system (Munich, Germany) using a Cu K α radiation source, operated at a voltage and current of 40 kV and 40 mA, respectively.The X-ray source and the detector were coupled in θ-2θ scans (Bragg-Brentano geometry), scanning a 2θ range from 30 • to 70 • at a step size of 0.05 • , and dwell time of 2 s per step.High energy XRD measurements on the powdered coatings were done in transmission geometry at the beamline P02.1 [50] of the Deutsches Elektronen-Synchrotron (DESY) in Hamburg, Germany, using X-rays with a wavelength of 0.20701 Å.The powders were sealed in fused silica capillaries with a diameter of 1 mm and a wall thickness of 20 μm under Ar (99.9999% purity) atmosphere.
X-ray stress and lattice parameter analyses were performed on the thin films using the sin 2 ψ method in a ψ geometry (ψ represents the tilt axis in the diffraction plane) over the (200) diffraction peak with a Bruker D8 Discover General Area Diffraction Detection System (Billerica, MA, USA) by assuming a biaxial stress state [51].ψ was varied from 0 to 60 where ν is the Poisson's ratio and E is the elastic (Young's) modulus of the thin films.
The thermal stress σ thermal contribution to the total stress state arising during cooling of the samples from T s to room temperature, was estimated by: where Δα is the difference between the linear coefficients of thermal expansion of the film (9.7 × 10 − 6 K − 1 for Ti 0.56 Al 0.44 N and 9.9 × 10 − K − 1 for Ti 0.50 Al 0.50 N, calculated in this study) and substrate (5.5 × 10 − K − 1 for α-Al 2 O 3 , provided by the substrate supplier).The temperature decrease ΔT after cooling from T s to room temperature was ~ 430 K for all depositions.Finally, the intrinsic stress (σ intrinsic ) was calculated by subtracting the σ thermal contribution from the σ residual .
Elastic modulus E was determined by nanoindentation using a Hysitron (Minneapolis, MN, USA) TI-900 TriboIndenter equipped with a Berkovich geometry-diamond tip with a 100 nm radius.At least quasistatic indents, with a maximum load of 8 mN, corresponding to contact depths between 98 and 108 nm, were performed.The tip area function was determined with a fused silica reference before each measurement series and verified to remain unchanged thereafter.The reduced modulus was acquired from the unloading segment of loaddisplacement curves using the method of Oliver and Pharr [52].The elastic moduli of the films were then attained from the reduced moduli data using the Poisson's ratio of v s = 0.214 [24].
Cross-sectional thin film lamellae were prepared by focused ion beam techniques in a FEI Helios Nanolab 660 dual-beam microscope (Hillsboro, OR, USA).The microstructure was characterized using scanning transmission electron microscopy (STEM) at 30 kV and 50 pA within the same microscope and bright field (BF) micrographs were acquired with a STEM III detector.
Transmission electron microscopy (TEM) analyses were performed using a probe C s -corrected FEI Titan 3 G2 60-300 (Hillsboro, OR, USA) operating at 300 kV accelerating voltage.BF-TEM images and selected area electron diffraction (SAED) patterns for the cross-sections of the samples were obtained with a Gatan CCD camera (Gatan Inc., Pleasanton, CA, USA).
The thermal response of (Ti,Al)N powders was investigated in a differential scanning calorimeter (DSC) Jupiter® STA 449 C produced by Netzsch (Selb, Germany) in a continuous heating mode and Ar atmosphere with a purity of 99.9999%.Initially, the powders were outgassed for 30 min at 150 • C. The measurements were then started by heating up the powders to 1500 • C in 30 sccm flow of Ar.The heating and cooling rates were kept at 20 and 40 K/min, respectively.An integrated oxygen trap system (OTS®) enabled a low oxygen concentration (<1 ppm) in the measurement atmosphere.For temperature and sensitivity calibration of the system, high purity standards (≥99.9%)(Zn, Sn, Al, Ag, Au, Pd) were melted.
The local chemical composition at the nanometer scale was determined by atom probe tomography (APT) using a local electrode atom probe 4000X HR (CAMECA, Madison, WI, USA).Field evaporation was assisted by 30 pJ laser pulsing at 200 kHz frequency and the base temperature was 60 K.The detection rate was set to 0.5%.Data reconstruction was performed with the IVAS 3.8.8,(CAMECA) using the evolution of the standing voltage and shank angle protocol.The frequency distribution functions (FDF) were obtained at a bin size of 50, excluding unranged ions.

Theoretical details
The plasma-surface interactions were investigated by DFT-based [53] molecular dynamics (MD) simulations together with the DFT thermal spike model [29].The OpenMX code was used for MD simulations due to its computational speed and simultaneously sufficient accuracy [54].The basis functions were in a form of a linear combination of localized pseudoatomic orbitals [55] in conjunction with the generalized gradient approximation (GGA) [56], and generated using a confinement scheme [55,57] specified as: Ti7.0-s3p2d1, Al7.0-s2p2d1, N6.0-s2p2d1, and Ar5.0-s2p1.Here, Ti, Al, N, and Ar designate the chemical name followed by the cutoff radius (Bohr radius unit) and the corresponding primitive orbitals.A total energy precision of 10 − 6 Ry atom − 1 was reached by choosing an energy cutoff of 150 Ry.Slab models of 512 atoms for (200) surface (equivalent to (100) for the present structures) and 384 atoms for (111) surface (N-terminated) of S. Karimi Aghda et al.
Ti 0.50 Al 0.50 N and Ti 0.56 Al 0.44 N were used, where the bottom layer of each configuration was frozen to mimic an infinite bulk.The clustered configuration C#3 for c-(Ti,Al)N was considered as proposed by Mayrhofer et al. [2], where the number of Ti-Al second order nearest neighbors was minimized.To describe the Ti 0.5 Al 0.5 N(200), Ti 0.56 Al 0.44 N(200), and Ti 0.56 Al 0.44 N(111) surfaces in the slab configuration, a 10 Å thick vacuum layer was introduced.Ar atoms with E k = 30 eV and E k = 60 eV were primarily placed at atop positions above Ti, Al, and N surface sites at a distance of 3 Å.The MD simulations were performed for a period of 400 fs with a timestep of 1 fs at 723 K (velocity scaling thermostat, canonical ensemble), corresponding to the growth temperature of (Ti,Al)N thin films in this work.The defect structures obtained from the low kinetic energy DFT-based MD simulations were used to define the instantaneous energy (E d ), based on the Kinchin-Pease equation [58], Here, the number of point defects (N), obtained from the structures after the MD runs, is directly correlated to the initial kinetic energy (E k ) of the irradiating ions.The calculated E d values for Ti 0.5 Al 0.5 N(200), Ti 0.56 Al 0.44 N(200), and Ti 0.56 Al 0.44 N(111) irradiated surfaces are 4.4 ± 0.4, 4.6 ± 0.5, and 3.8 ± 0.4 eV, respectively.E d has been shown to be close to the activation energy for diffusion [25,29].
In order to consider the impact of ion irradiation with E k covering hundreds of eV, the DFT-based thermal spike model was employed as fully described in our previous works [27,29].The Visualization for Electronic and Structural Analysis (VESTA) code was used for the evaluation of the final atomic configurations [59].Full structural relaxation of the slabs at 0 K was performed for each defect configuration.Subsequently, in order to obtain the equilibrium volume of the defect configurations, the irradiation-induced resputtered atoms, which moved above the pristine surface were removed (creating single vacancy sites within the supercells) and structural relaxation at 0 K was performed.The Birch-Murnaghan equation of state [60] was then used to obtain the equilibrium volume.The cohesive energy (E coh ) of each irradiated supercell was calculated based on: where E Ti , E Al , and E N , are total DFT energies of single (isolated) Ti, Al, and N atoms, respectively, m, n, and k, are the number of atoms in the supercells, and E TiAlN , is the total energy of (Ti,Al)N supercells.To determine the temperature-dependent equilibrium volume (lattice parameter), bulk modulus, and linear coefficient of thermal expansion, the Debye-Grüneisen model [61] was used, as described elsewhere [29].
The ion irradiation-induced intrinsic stress state evolution was calculated for each bombarded supercell by considering the volume change with respect to the equilibrium volume of the ideal defect-free structure at E k = 0 eV.Here, the expansion (contraction) of the supercell due to the presence of point defects results in the generation of compressive (tensile) stress.The elastic modulus corresponding to each defect structure was calculated from the bulk modulus within the isotropic approximation, using a Poisson's ratio of ν = 0.214 [24].In order to obtain the stress dependence of the elastic moduli, a volume offset from the equilibrium volume was considered for each thermal spike configuration.
The surface energies were obtained using DFT calculations performed with the Vienna ab initio simulation package (VASP, University of Vienna, version 5.4.4) [62][63][64], within the framework of the GGA, utilizing projector-augmented waves for basis set representation [65], with the cutoff energy set to 500 eV.Exchange and correlation effects were accounted for using the well-established functional by Perdew, Burke, and Ernzerhof (PBE) [56] and, hence, the same framework as for the OpenMX code.Brillouin zone integration was handled via Blöchl's tetrahedron method [66] in a k-mesh of sufficient density to ensure energetic convergence, depending on supercell size, generated with the Monkhorst-Pack approach [67].Spin polarization was not considered as c-(Ti,Al)N is not magnetically active.
For construction of the surface slab models, the three surfaces (200), (220), and (111) were considered and distinct supercells of bulk c-(Ti, Al)N were generated with 128 atoms, 32 atoms and 96 atoms in the (200), ( 220) and (111) case, respectively, with the surface to be created always parallel to the ab plane of the simulation cell.We have chosen orthorhombic unit cells for (220) and (111) cases for consistency with structural models used in preceding calculation series.These structures were then fully optimized at 0 K with the lattice parameters of the resulting supercells (a, b, c) of (8.35 Å, 8.35 Å, 16.70 Å), (6.01 Å, 4.14 Å, 11.80 Å), and (5.90 Å, 10.22 Å, 14.46 Å), for (200), (220), and (111) structural models, respectively.From the optimized structures, slab models of the (200), (220), and (111) surfaces were constructed by adding a vacuum region of 10 Å in the c direction of the supercell.The convergence tests were done for different slab thicknesses by successfully adding or removing further atomic layers, with energetic convergence within numerical limits of the utilized method obtained for all systems using the structural models described above.Thus, the chosen size yields optimum tradeoff between runtime and accuracy.In the case of the (111) surface, the N-terminated variant, corresponding to the orthorhombic structural models used for the thermal spike calculations, was considered.These slab models were in turn optimized with respect to the atomic positions, while keeping cell shape and size constant.The resulting surface energies were then calculated according to the following formula: with E surf as the surface energy, E tot,bulk and E tot,slab the total DFT energies after optimization of the bulk and corresponding slab models, respectively, and a and b as the lattice parameters of the respective slab model's simulation cell.The factor 2 accounts for the cleavage of the bulk crystal and the corresponding generation of two equivalent surfaces.

Plasma analysis
The charge states, kinetic energies, and ion fluxes of generated positive ions were characterized and compared for HPPMS and CAD plasmas in this study.The time-averaged IEDFs of gaseous as well as metallic species of Ti and Al in HPPMS and CAD discharges are shown in Fig. 1(a-d).Fig. 1 shows that, independent of the plasma source under the here chosen process conditions, the most probable ion energy is in the range of ~ 5 ± 1 eV corresponding to ~80 and ~85% of the total ion population for the HPPMS and CAD processes, respectively.This most probable ion energy is followed by an energy tail extending up to ~70 eV for the metallic species.The extended energy tails originate from the mechanisms of the sputtering or cathodic arc processes and can be associated with the remaining kinetic energy of the species after transport through the gas phase with no collisions, and hence without change in kinetic ion energy.In addition, in HPPMS, the high energy ions might also originate from the time-variation of the plasma potential during HPPMS pulses and the subsequent acceleration of ions over the modulated sheath potential [68].In HPPMS (Fig. 1(a) and (b)), the majority of the ions are singly charged Ar + , N 2 + , N + , Ti + , and Al + ions with the population of detected doubly charged Ar 2+ , Ti 2+ and Al 2+ ions being less than 2% of the overall ionic plasma composition.Hence, the HPPMS plasma is mainly dominated by singly charged ions, with Ar + being the most abundant species with ~52% of the total ionic plasma composition.The shape and energy range of the Ti + and Al + EDFs are comparable for both HPPMS and CAE plasmas.However, in CAD (Fig. 1(c) and (d)), there is a significant contribution of doubly charged Ti 2+ ions accounting for ~15% of the energetic high-intensity fluxes.In addition, a significant amount of N 3 + ions was detected as the gaseous species.The large energy density provided in CAD process is the primary reason for the larger population of the detected doubly charged Ti 2+ and molecular N 3 + ions [69].
The average charge states, atomic masses, and ion energies (Eq.( 2)) are calculated from the IEDFs and summarized in Table 1.There is a slight increase in average charge state from 1.01 to 1.15 and in average atomic mass from 39.3 to 42.7 amu comparing the HPPMS and CAD plasmas, respectively.These values are directly connected to the presence of ~15% doubly charged Ti 2+ ions in the CAD plasma, which accounts for a higher charge state as well as higher average atomic mass.In addition, as the most probable energy range for both processing routes account for more than 80% of the total ion population, for the discussion of ion kinetic energy we have adopted the concept of average energy for simplicity.The average ion energy increases only slightly from 4.9 to 5.4 for HPPMS and CAD plasmas, respectively.Based on the average charge states and ion energies obtained from the IEDFs, by changing the magnitude of substrate bias potential, the most probable ion kinetic energy (E k ) ranging from 25 to 205 eV for HPPMS and 19 to 236 eV for CAD, are investigated in this work.
Furthermore, the differences in the ion flux between the processing routes were analyzed by measuring the ion current density at the substrate and the results are shown in Table 1.The substrate ion current density in the HPPMS process is 0.9 mA/cm 2 .This is calculated by integrating the measured ion current at the substrate within 100 µs from the target pulse on-time, as the majority of the ions are expected to arrive at the substrate within a short time-span [70].In contrast, the higher ionization degree in the CAD mode resulted in a continuous ion flux with an ion-current density of 3.0 mA/cm 2 which is 3.3 times higher than in the case of HPPMS.Therefore, the ion flux arriving at the growing film surface is significantly larger for CAD than HPPMS, while  the difference in the average charge state, atomic mass, and ion energy are comparatively small between the processing routes.

Thin film composition
Average chemical compositions of the thin films were obtained from homogeneous ERDA depth profiles and are presented in Fig. 2. The oxygen content in all thin films was below 0.5 at%.Independent of the processing route and the magnitude of E k , (Ti,Al)N thin films exhibit overstoichiometric compositions with respect to nitrogen.The measured N content is between 51 and 52 at.% for all the here investigated thin films and this variation is within the maximum total measurement uncertainty of 2 at.%.However, the composition of metallic species depends on the magnitude of E k .In the following, Al concentrations are presented with the sum of systematic and statistic uncertainties.The Al content in the thin films deposited by HPPMS changes from 25 ± 1 at% at E k = 25 eV to 23 ± 1 at.% at E k = 205 eV.A less significant reduction in Al content is observed for the CAD thin films from 22 ± 1 at% to 21 ± 1 at% as the E k is increased from 19 to 236 eV, respectively.The here observed reduction in the Al content by increasing the magnitude of E k in case of HPPMS can be rationalized by the ion irradiation-induced preferential resputtering of lighter Al atoms from the surface of the growing film [71,72].The energy of formation for Al vacancies was calculated to be the lowest amongst all the species in c-(Ti,Al)N [73].While the target composition for both processing routes was identical (Ti 0.5 Al 0.5 ), the CAD thin films exhibit, on average, ~13% lower Al content as compared to HPPMS thin films.This can be understood by comparing the pressure-distance product (pd) [74] difference between the HPPMS (pd = 5 Pa cm) and CAD (pd = 128 Pa cm) conditions.The significantly higher pd of CAD may result in stronger scattering of the lighter Al compared to Ti, hence, less Al is incorporated during growth.Based on the data presented in Fig. 2, the dashed lines indicate the average compositions of (Ti 0.50 Al 0.50 ) 0.48 N 0.52 and (Ti 0.56 Al 0.44 ) 0.48 N 0.52 as calculated for HPPMS and CAD thin films, respectively, which are used for the remaining discussion.

Thin film structural analysis
The ion energy-dependent phase formation of (Ti,Al)N was characterized by XRD and is presented in Fig. 3 for the thin films synthesized on Al 2 O 3 substrates by (a) HPPMS and (b) CAD.The diffractograms show the formation of a single-phase metastable cubic (Ti,Al)N solid solution, without presence of the wurtzite AlN, independent of the processing route and magnitude of E k .However, there are distinct discernible differences in ion-energy-dependent evolution of preferred orientation for HPPMS and CAD thin films.In HPPMS, at E k = 25 eV, the formation of film with a (111) preferred orientation is observed.As the E k is increased, the intensity of the (111) diffraction peak reduces and a random orientation is present at E k ≥ 155 eV, which is accompanied by peak broadening.The TEM analysis and SAED pattern of the thin film deposited by HPPMS at E k = 205 eV are shown in Fig. S2 of the supplementary materials.This data reveal formation of a single-phase cubic structure of thin film deposited under the maximum ion kinetic energy of 205 eV in HPPMS process.Therefore, the shoulders observed for the peaks at (111) and (220) diffractions might be due to the lattice parameter distortions.In contrast, in CAD, at E k = 19 eV, a close-torandom orientation is observed, while the intensity of the (200) diffraction peak dominants at E k = 52 eV.Further increase of E k to ≥ 144 eV results in a very strong (111) preferred orientation.The observed structural (preferred orientation) evolution of the thin films deposited on Al 2 O 3 substrates are comparable to the thin films which are deposited on Fe foils for the purpose of powered-sample preparation, see Figure S3 of the supplementary materials.Generally, the evolution of preferred orientation during thin film growth is understood to be both thermodynamically and kinetically controlled [75][76][77].in thermodynamic equilibrium, the preferred orientation is defined by the minimization of the sum of surface energy E surf = 3.20 J/m 2 .This is consistent with the surface energy data calculated by Forslund et al. [78].Additionally, the directional elastic anisotropy of c-(Ti,Al)N [20] accounts for the strain energy anisotropy as E (111) Therefore, for a defect-free system the (200) plane shows the lowest surface energy as well as the lowest strain energy.However, our calculations for the high kinetic energy ion irradiation-induced defect structures show that the cohesive energy is defect structure-dependent.We have shown in Fig. 6 (see below) that the ion irradiation-induced generation of point defects would result in a lower cohesive energy along (111) direction in comparison to (200) direction, contrary to the calculated data of defect-free systems by Tasnádi et al. [19].Hence, the defect structure-dependent changes in elastic anisotropy, proportional to cohesive energies, results in a lower strain energy along (111) direction for (Ti,Al)N.Therefore, as thermodynamic equilibrium is approached, both E the kinetically limited, upward growth of (111) planes [76,79,80].Increasing E k in HPPMS provides enough surface and subsurface mobility to incorporate atoms into the planes with more open crystal channeling, namely (200) and (220) planes in comparison to (111) [75].These results are consistent with other magnetron sputter-deposited TiN [76], (Ti,Al)N [75,81] and (V,Al)N [27] thin films.In contrast, the 3.3 times higher substrate ion current density accounts for a significantly larger ion flux in CAD.The larger ion flux enhances surface adatom mobility at low ion kinetic energies E k ≤ 52 eV.Here, the surface adatom mobility is sufficient to grow low-energy (200)-orientated crystallites during nucleation and coalescence phase [81].However, at higher irradiation energies E k ≥ 144 eV in CAD, where the kinetic limitations should be even smaller than the depositions with low kinetic energies, the (111) evolution is dominant.Here the preferred orientation of (Ti, Al)N film is mainly controlled by the strain energy, hence, the preferred orientation corresponds to the (111) plane with the lowest strain energy.

Defect structure
To investigate the ion-surface-interactions at the atomistic scale, the thermal spike model [27][28][29] was used by considering the experimentally obtained composition (Fig. 2) and preferred orientation evolution (Fig. 3) of the synthesized thin films.Here, the nominal theoretical compositions of Ti 0.50 Al 0.50 N and Ti 0.56 Al 0.44 N were chosen to represent the average chemical compositions of (Ti 0.50 Al 0.50 ) 0.48 N 0.52 and (Ti 0.56 Al 0.44 ) 0.48 N 0.52 corresponding to HPPMS and CAD thin films, respectively.The ion irradiation effects on the final defect structures were systematically investigated in the E k range from 0 to 203, as well as 0 to 219, and 0 to 205 eV for Ti 0.50 Al 0.50 N with a (200) orientation and for Ti 0.56 Al 0.44 N with (200) and (111) orientations, respectively.Such comparison is only possible due to minute differences in the average charge state, atomic mass, and ion energy for the HPPMS and CAE processing routes, as argued above.For each of these supercell configurations, the initial and exemplary final atomic configurations after thermal spikes are shown in Fig. 4(a-c).It is evident that ion irradiation with similar kinetic energy results in surface as well as sub-surface diffusion of the atoms, independently of the here investigated compositions and surface orientations of the supercells.Formation of self-interstitial/vacancy point-defect pairs (Frenkel pairs) are the dominant type of point defects.We observe not only N but also metallic species as stable Frenkel defects within the c-(Ti,Al)N supercells.Additionally, formation of individual lattice vacancies observed as the high-energy ion irradiation also provides enough mobility for the surface atoms to move above the pristine surface.For the irradiation of the (200) surface (Fig. 4(a) and (b)), resputtering of surface Al atoms is observed.Conversely, the N-terminated (111) polar surface of (Ti,Al)N (Fig. 4(c)) appears to hinder resputtering of Al.This is consistent with the chemical composition analysis in Fig. 2, as the HPPMS thin films showed a significant reduction in Al content with respect to E k , while the CAD thin films did not show a significant Al reduction and exhibit a dominant (111) surface orientation.
The ion energy-dependent point defect concentrations were evalu-ated based on the DFT thermal spike model and the results are summarized in Fig. 5(a-f).For (200)-terminated Ti 0.50 Al 0.50 N, there is a significant increase in the concentration of Frenkel pairs, Fig. 5(a), up to a maximum of 3.7% as E k is increased to 154 eV, which is followed by a minor reduction to 3.5% at E k = 203 eV.Here, the dominant Frenkel pairs present in the supercells at E k ≥ 154 eV consist of N and Al species.Concurrently, formation of vacancies is evident and their concentration is directly connected to the magnitude of E k , see Fig.  structures along ( 111) and ( 200) directions are calculated and the results are depicted in Fig. 6.At E k = 0 eV, corresponding to the defect-free supercells, comparable E coh values of 8.23 and 8.22 eV/atom for the (111) and ( 200)-oriented structures, respectively, are obtained.However, by increasing E k , which consequently results in the formation of point defects, the cohesive energy of (111)-oriented irradiated structure becomes significantly lower than that of the (200)-oriented irradiated one.These results indicate that the elastic anisotropy of Ti 0.56 Al 0.44 N is clearly defect structure dependent.
The ion energy-dependent defect structures discussed in Figs.200) (red squares) shows an increase of 0.7% in the E k range from 0 to 154 eV, followed by a reduction at higher E k , which is in excellent agreement with the experimentally measured values for lattice parameter evolution.We have recently shown that Frenkel pairs increase the lattice parameter and vacancies contract the lattice [27].It is evident from the chemical composition data in Fig. 2, that no significant ion irradiation-induced chemical composition change is observed for the here investigated thin films.Therefore, the maximum increase in the lattice parameter by 0.7% in Ti 0.50 Al 0.50 N(200) is caused by the Frenkel pair concentration of 3.7% that overcompensates the concurrent vacancy-induced reduction in lattice parameter, see Fig. 5 (a) and (b).Slight deviations between the absolute values of experimentally measured and theoretically obtained lattice parameters with respect to E k might arise from the deviations between the absolute defect concentrations between the thin films and supercells as the thermal spike model describes one single irradiation event.A similar evolution in lattice parameters of the (Ti 0.56 Al 0.44 ) 0.48 N 0.52 thin films deposited by CAD (black circles) is observed, Fig. 7(a).However, the maximum increase in the lattice parameter of 1.5% from 4.170 to 4.235 Å upon increasing E k from 19 to 144 eV is ~2 times larger than the lattice parameter increase measured for the HPPMS deposited thin films.The calculated lattice parameters for Ti 0.56 Al 0.44 N based on the irradiation of (200) surface (pink circles) show a 0.8% increase within a similar ion energy range.However, by considering the irradiated (111) surface of Ti 0.56 Al 0.44 N (dark purple circles), as the preferred orientation observed for the CAD thin films at high E k , see Fig. 3 (b), the lattice parameter increase reaches a maximum of 1.1% from 4.196 to 4.241 Å as the E k is increased from 0 to 182 eV.This larger increase in case of the irradiated (111) surface in comparison to the case of the (200) surface is governed by the 34% higher Frenkel pair concentration at a corresponding E k .Fig. 7(b) and (c) display the measured lattice parameter of the HPPMS and CAD powders, respectively, from synchrotron diffraction data together with the theoretically obtained values as a function of E k .The powders diffraction data (not shown) exhibit polycrystalline cubic structure of (Ti,Al)N without any secondary phase for all the powders investigated here.The lattice parameter increase in the (Ti 0.50 Al 0.50 ) 0.48 N 0.52 /HPPMS powder is 0.5% as E k is increased from 25 to 155 eV, while for (Ti 0.56 Al 0.44 ) 0.48 N 0.52 /-CAE powders the value is 1.5% upon increasing E k from 19 to 144 eV.These changes in the lattice parameters of the powders are consistent with the lattice parameter evolution of thin films as discussed in Fig. (a).The minute differences between the absolute values of the lattice parameters of the powders in comparison to those of the thin films might be due to the thermal stress contribution of up to +1.0 GPa present in the thin films but not in the powders, as they are considered to be free of macro-stresses.

Thermal stability
To probe the effect of ion irradiation-induced defects on the thermal stability of (Ti,Al)N, the heat response of the powders discussed in Fig.(b) and (c) was evaluated by DSC and the results are presented in Fig. 8.The investigated powders show their first enthalpic signals, consistent with recovery processes [7] at ~500 • C for HPPMS powders, Fig. 8 (a) and (c), and ~650 • C for CAD powders, Fig. 8 (e) and (g).The DSC signal of the HPPMS powder deposited at E k = 25 eV, Fig. 8 (a), exhibits a second and third exothermic peak at 900 and 1150 • C, respectively, which indicate, according to Mayrhofer et al. [7,82], the onset of spinodal decomposition and wurtzite phase formation, respectively.These peaks are also observed for the powder deposited at E k = 155 eV, Fig.
(c), at similar temperatures.Comparable to the HPPMS powders, the average onset temperature of spinodal decomposition for the CAD powders occurs at ~900 • C, independently of E k , Fig. 8(e) and (g), while the onset of wurtzite formation for the CAD powder deposited at E k = eV, Fig. 8 (e), occurs at 1100 • C, similar to the case of the HPPMS powders.However, the third exothermic peak, corresponding to the onset of wurtzite formation, is not detected in the DSC signal of the CAD powder deposited at E k = 144 eV, Fig. 8(g).The XRD patterns of powders after annealing are shown in Fig. 8(i) and (j) for HPPMS and CAD powders, respectively.For all four samples, two phases can be detected, corresponding to the c-TiN and the w-AlN reference lines.This indicates that the metastable solid solution has decomposed into the thermodynamically stable phases, despite the fact that no exothermic peak corresponding to the wurtzite formation has been observed for the CAD powder deposited at E k = 144 eV, Fig. 8(g).This anomalous observation may be rationalized by the contemporaneous compensation of the endothermic wurtzite phase formation [7] by the exothermic defect annihilation [25] in films with very highion irradiation-induceddefect concentration.
The mass as a function of temperature (see thermogravimetric analysis (TGA) signals), Fig. 8(b), (d), (f), and (h), shows a slight reduction at around 450 • C (deposition temperature), which indicates nitrogen loss according to Rovere et al. [83].This mass loss is significantly larger for the samples deposited at higher ion kinetic energies (E k = 155 eV for HPPMS powder and E k = 144 eV for CAD powder), which can be explained by the larger absolute concentration of irradiation-induced point defects present in the growing films, being consistent with [28].The second mass loss starting at >1000 • C is caused by the formation of the wurtzite phase [83].In HPPMS, the In an effort to correlate the changes in thermal stability of the cubic solid solution with the irradiation-induced structural changes, the thermal stability increase based on the TGA results from Fig. 8 are displayed in Fig. 9 with respect to the lattice parameter increase of the powders from this study and the work by Holzapfel et al. [28].The corresponding combined concentration of Frenkel pairs and vacancies causing the lattice parameter increase is calculated based on the thermal spike model, see Figs. 5 and 7, and is shown on the second and third x-axes of Fig. 9, respectively.Within the here employed thermal spike model, formation of vacancies and Frenkel pairs are connected and cannot be separately probed.In Fig. 9, it can be observed that the changes in the lattice parameter due to formation of irradiation-induced point defects scale with the increase in the thermal stability of c-(Ti,Al) N. The irradiation-induced increase in the lattice parameter of 0.5% for the HPPMS synthesized powders in this study is attributed to the presence of 1.5% Frenkel pairs and 0.2% vacancies, inferred to be responsible for the 35 • C increase in the onset temperature of the wurtzite phase formation.The lattice parameter increase of 0.9% obtained due to ion irradiation for CAD synthesized powders by Holzapfel et al. [28] is attributed to 2.7% Frenkel pairs and 0.5% vacancies in the lattice, which was shown to enhance the thermal stability of the cubic solid solution by 83 • C. Furthermore, a much larger lattice parameter increase of 1.5% was obtained in the present work by increasing the kinetic energy of ionized species in the CAD process to 144 eV.This lattice parameter increase is causedconsistent with predictions based on the thermal  Previously, we have shown that the Al mobility is a decisive parameter in defining the thermal stability of c-(Ti,Al)N, while the mobility of N is not required [73].Holzapfel et al. [28] have computed the activation energies for bulk diffusion of single Ti, Al, and N atoms in Ti 0.50 Al 0.50 N structural models with and without a Frenkel pair by means of DFT.The activation energies for bulk diffusion of Ti, Al, and N were calculated to be 3.15, 2.60, and 3.85 eV, respectively.However, the activation energy for bulk diffusion of Al in the vicinity of one Ti, Al, or N interstitial atom increases significantly by 94, 125, and 13%, respectively [28].Therefore, the here reported Frenkel pair concentration-dependent thermal stability can be rationalized based on the increase in the activation energy for bulk diffusion of Al, thereby influencing the decomposition pathways of metastable cubic (Ti,Al)N.

Morphological evolution and local chemical composition
To understand the effect of E k on the morphology of the c-(Ti,Al)N deposited in this study, the thin films were characterized at the nanometer scale by cross-sectional BF-STEM, see Fig. 10.All thin films investigated here show no evidence of porosity within the matrix, independent of the processing route and ion kinetic energy.However, pores around macroparticles were observed for all CAD synthesized thin films, being consistent with [84,85].
For the (Ti 0.50 Al 0.50 ) 0.48 N 0.52 /HPPMS series at low E k = 25 eV, a finegrained columnar microstructure is visible, Fig. 10 (a).The increase in kinetic energy of ionized species up to 205 eV, Fig. 10(b) and (c), results in a reduction of column width without significantly altering the columnar growth of the thin films.Reduction in grain size at higher E k is consistent with the extended structure zone diagram proposed by Anders [86].Similarly, for (Ti 0.56 Al 0.44 ) 0.48 N 0.52 /CAD thin films, a very fine columnar structure is observed at E k ≤ 52 eV, Fig. 10(d) and (e).At E k = 190 eV, Fig. 10(f), a distinct morphological evolution from the substrate towards the surface of the thin film is observed.Here, at the region close to the substrate, a columnar structure is grown with significantly larger column width than for the thin films deposited at E k ≤ 52 eV.Further increase in the film thickness, however, resulted in the termination of the columnar growth with grains having a conical shape.From the cross-sectional lamella of the thin film deposited at 190 eV, Fig. 11(c), the near-substrate and near-surface regions are analyzed separately by electron diffraction.The SAED pattern in Fig. 11(d) of the blue square area in Fig. 11(c) shows the single phase cubic structure of (Ti,Al)N at the near substrate regions with a dominant (111) diffraction, consistent with the XRD data in Fig. 3 (b).Fig. 11(e) and (f) display a higher resolution TEM image and SAED pattern of the regions marked by orange and red squares, respectively.From the image in Fig. 11 (e), very fine nano-crystallites embedded in a disordered structure were found.From the SAED pattern in Fig. 11 (f), it is evident that the columnar (111) structure has evolved towards a more polycrystalline nanostructure.The near surface nanostructure could not be captured by the resolution of XRD as only (111) diffraction was observed in Fig. 3(b).The evolution of a fine-grained nanocrystalline structure under ion irradiation with E k = 190 eV may be rationalized by a significant concentration of ion irradiation-induced defects (mostly Frenkel pairs, Fig. 5) resulting in a larger mobility for renucleation.
Additionally, the near-surface local chemical composition of the thin films was measured by atom probe tomography and the data are shown in Figure S4 of the supplementary material.Based on the atomic positions of Al and the frequency distribution functions, chemical homogeneity is observed for the HPPMS thin films independent of the magnitude of E k .For both HPPMS thin films, the Pearson coefficient µ [87] is comparable and close to a random distribution.For CAD thin films, however, the increase of the E k to 190 eV results in clustering of Al, with the Pearson coefficient for Al distribution reaching µ Al = 0.33.The deviation from the random elemental distribution in the CAD thin film at E k = 190 eV might be correlated to the surface and sub-surface diffusion mediated clustering of atoms in the highly defective structure.

Mechanical properties
The ion irradiation-induced changes in the structure of c-(Ti,Al)N are mirrored in intrinsic stress (σ intrinsic ) changes of the thin films.The results are illustrated in Fig. 12 (a).Independent of the processing route, a steep increase in the compressive stress state with respect to E k is observed.
The maximum compressive σ intrinsic of -5.9 and -7.2 GPa are obtained for HPPMS and CAD, respectively, within similar E k of 100-150 eV.This is accompanied by stress relaxation as E k is increased further.This trend is consistent with the evolution of stress state predicted by the thermal spike model.Ion irradiation-induced stress generation and relaxation are correlated to the generated and annihilated Frenkel pairs and vacancies [27].The larger maximum compressive σ intrinsic observed for the CAD thin films can therefore be rationalized by the larger maximum Frenkel pair concentration on (111)-oriented irradiated surface (Figure 5 (e)) in comparison to the (200)-oriented one (Figure 5 (c)).
To investigate the irradiation-induced changes in elasticity of (Ti,Al) N, the elastic moduli of the thin films was measured and compared to the calculated values from the thermal spike model in Fig. 12  For the polycrystalline HPPMS thin films with no preferred orientation, it is reasonable to expect that the slight influence of (111) orientation on reducing the elastic modulus cannot be resolved experimentally by considering the magnitude of the error bars.

Conclusions
By correlatively employing synthesis experiments, integral as well as spatially resolved characterization, and ab initio calculations, the influence of ion kinetic energy (E k ) on the structure evolution, mechanical properties, and thermal stability of (Ti,Al)N has been studied.Singlephase cubic metastable (Ti,Al)N thin films were synthesized by HPPMS and CAD within the E k range from 19 to 236 eV.While for the HPPMS films increasing E k resulted in irradiation-induced random orientation evolution, the CAD films showed preferred (111) orientation at E k ≥ 144 eV.The substrate ion current density of 3.0 mA/cm 2 measured for CAD as compared to HPPMS with 0.9 mA/cm 2 accounts for a 3.3 times higher energy flux at the growing film surface in the CAD process.Hence, the strong (111) texture evolution in the CAD thin films synthesized at higher E k can be rationalized based on ion flux-and ion energy-induced minimization of strain energy in defective cubic (Ti,Al) N. Additionally, the ion irradiation-induced changes in elastic modulus can be understood based on the ion energy-and orientation-dependent point defect formation from ab initio calculations.For HPPMS thin films with random orientation, the concurrent effects of ion irradiationinduced formation of Frenkel pairs and generation of compressive stresses resulted in an apparent independence of the elastic modulus on E k .In contrast, for CAD thin films with (111) preferred orientation, an ~18% reduction in elastic modulus was obtained.It is shown that the reduction of elastic modulus for CAD thin films is due to the significantly larger concentration of Frenkel pairs formed upon irradiation of the experimentally observed (111) orientation in comparison to the (200) orientation.At E k = 182 eV, irradiation of the (111)-oriented surface leads to ~34% higher Frenkel pair concentration than for the (200)oriented surface irradiation at similar E k values.Moreover, we present evidence of ion-energy-dependent thermal stability enhancement of (Ti, Al)N.DSC data reveal that, independent of the processing route, the ion irradiation-induced increase in Frenkel pair concentration leads to an increase in the onset temperature of wurtzite formation by up to 206 • C. Therefore, it is evident that ion irradiation-induced changes in elastic modulus and thermal stability of c-(Ti,Al)N thin films can be understood based on the ion kinetic energy and orientation-dependent formation of point defects.

Fig. 1 .
Fig. 1.Time-averaged IEDFs of (a) gaseous and (b) metallic species of a Ti-Al HPPMS discharge in Ar/N 2 atmosphere and (c) gaseous and (d) metallic species in a Ti-Al cathodic arc discharge in N 2 atmosphere.
surf and strain energy E (hkl) strain of grains with different orientation.The calculated E (hkl) surf for the metastable cubic Ti 0.50 Al 0.50 N in this study shows the lowest surface energy for the (200) plane, E (200) surf = 1.29 J/m 2 , in comparison to the (220) plane, E (220) surf = 2.64 J/m 2 , and (111) plane (N-terminated), E (111)

Fig. 2 .
Fig. 2. Chemical composition of (a) HPPMS-and (b) CAD-synthesized (Ti,Al)N thin films as determined by ERDA with respect to the ion kinetic energy.The dashed lines indicate the average films composition for each processing route.
4 and 5 are used to determine the ground state properties of c-(Ti,Al)N.Fig. 7 (a) shows the comparison between experimentally measured lattice parameters of the thin films together with calculated lattice parameters at 300 K (near room temperature) as a function of E k .For HPPMS (Ti 0.50 Al 0.50 ) 0.48 N 0.52 thin films (blue squares), the lattice parameter increases by 0.8% from 4.202 to 4.231 Å as E k is increased from 25 to 105 eV.Increasing the ion kinetic energy further to 205 eV, however, results in a 0.4% reduction of the lattice parameter to 4.216 Å.The predicted lattice parameter for Ti 0.50 Al 0.50 N(

Fig. 6 .
Fig. 6.Cohesive energy (E coh ) of the irradiated Ti 0.56 Al 0.44 N for irradiated (200) and (111) surfaces as a function of the ion kinetic energy.

Fig. 7 .
Fig. 7. Measured lattice parameter of (a) thin films, (b) HPPMS powders, and (c) CAD powders of (Ti,Al)N together with the DFT calculated lattice parameters at 300 K with respect to ion kinetic energy.

Fig. 8 .
Fig. 8. DSC and TGA measurements of (Ti 0.50 Al 0.50 ) 0.48 N 0.52 /HPPMS powders deposited at (a) and (b) E k = 25 eV, (c) and (d) E k = 155 eV, and (Ti 0.56 Al 0.44 ) 0.48 N 0.52 / CAD powders deposited at (e) and (f) E k = 19 eV, (g) and (h) E k = 144 eV.Heat flow data of the powders stem from the difference between two heating cycles, to obtain only the powders calorimetric response.(i) and (j) post annealing X-ray diffraction patterns of powders deposited by HPPMS and CAD, respectively.

Fig. 9 .
Fig. 9. Thermal stability increase as a function of lattice parameter increase obtained experimentally for c-(Ti,Al)N from this study and the work of Holzapfel et al. [28].The concentration of Frenkel pairs (ρ Frenkel ) and vacancies (ρ Vacancy ) corresponding to the increase in lattice parameter are calculated based on the DFT thermal spike model and are shown on the second and third x-axes, respectively.

Fig. 10 .
Fig. 10.Cross-sectional BF-STEM images of (Ti,Al)N thin films deposited by (top) HPPMS at (a) E k = 25 eV, (b) E k = 155 eV, and (c) E k = 205 eV and (bottom) CAD at (d) E k = 19 eV, (e) E k = 52 eV, and (f) E k = 190 eV.The dashed orange line in (f) shows the morphology alteration for CAD deposited film at E k = 190 eV.The scale bar in (f) is valid for all the micrographs.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 11 .
Fig. 11.TEM characterization of (Ti,Al)N films deposited by CAD: (a) BF-TEM image of E k = 19 eV, (b) SAED pattern of (a), (c) BF-TEM image of E k = 190 eV, (d) SAED pattern of blue region in (c), (e) High resolution (HR) TEM image of orange region in (c), (f) SAED pattern of the red region in (c).(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Table 1
Average charge state, atomic mass, ion energy, and the substrate ion current density in the HPPMS and CAD discharges.