Influence of nitrogen vacancies on the decomposition route and age hardening of wurtzite Ti_1−xAl_xN_y thin films

Defect-engineering by introducing controlled amounts of vacancies has been a powerful tool in modifying properties of materials. This is especially true in the case of TiAlN, which is a prototypical model system for hard and protective nitride coatings.^1,2 Tuning its nitrogen stoichiometry (y = N/Al + Ti), for example, can result in films with either nitrogen vacancies (substoichiometry or y < 1)^3–7 or metal vacancies (superstoichiometry or y > 1)⁶ since vacancy formation energies determine the stability of various point defects.⁸ The different effects of these vacancy types on the properties of cubic TiAlN have been documented in recent years.^4,9–11 Specifically, metal vacancies are known to enhance the driving force for separation⁴ and, therefore, lower the temperature limit for age hardening. Nitrogen vacancies, on the other hand, have consistently demonstrated a reduction in the driving force for isostructural decomposition of the cubic TiAlN phase,^9,10,12 shift the onset of formation of the undesirable wurtzite-AlN phases to higher temperatures,⁹ and improve the mechanical properties for high-temperature applications.^7,10

The previous studies mentioned above have mainly examined the cubic phase of TiAlN compositions within the miscibility. Studies of the wurtzite TiAlN phase are traditionally neglected due to its reputation as a soft phase, which is detrimental for hard coatings.¹³ However, wurtzite containing coatings grown by PVD^14,15 and CVD^16,17 have been used in attempts to map the properties of w-TiAlN. One of the earliest attempts of Hörling et al.¹⁴ by cathodic arc deposition resulted in a Ti_0.26Al_0.74N coating with a two-phase nanostructure—nanocrystallites of w-AlN embedded in a cubic c-TiAlN matrix. Shimizu et al.¹⁵ have shown that dual-phase coatings with a highly textured wurtzite Ti_1−xAl_xN phase (x = 0.70−0.73) can be grown at low temperatures by high-power impulse magnetron sputtering (HIPIMS), which exhibit excellent hardness (up to 35 GPa) despite the dominant wurtzite phase present. In addition, single-phase wurtzite coatings have been found to have better oxidation resistance compared to cubic and mixed cubic and wurtzite Ti_1−xAl_xN coatings.^17,18 Interestingly, annealing PVD-deposited wurtzite Al₉₀Cr₁₀N¹⁹ resulted in notable improvements in elasticity, fracture stress, and toughness, attributed to the formation of c-CrN and c-Cr(Al)N precipitates. These reports suggest a potential avenue for leveraging wurtzite phases to enhance the mechanical properties of coatings.

Still, detailed information about TiAlN compositions with high Al contents where the wurtzite structure is more energetically favorable is limited. Even more overlooked is the impact of vacancies in the wurtzite phase. As a consequence, the effect of point defects on wurtzite TiAlN has not been investigated. Considering that industrial TiAlN coatings readily form small amounts of wurtzite during deposition,²⁰ it is important to fully understand this phase.

In this study, we utilized DC magnetron sputtering to epitaxially grow nitrogen-deficient single-phase solid solution wurtzite Ti_1−xAl_xN_y films with an array of high Al contents (x = 0.79–0.98). Epitaxial TiN(111) seed layers on MgO(111) and Al₂O₃(0001) substrates were used as a template to grow epitaxial wurtzite Ti_1−xAl_xN phase films. By growing high quality epitaxial wurtzite Ti_1−xAl_xN_y films, it is possible to study its intrinsic material properties without the influence of grain size and grain boundaries. Upon annealing these films to temperatures higher than the deposition temperature, we observe a substantial increase in the hardness (∼4–6 GPa) that was retained even after anneals up to 1200 °C. High-resolution HAADF-STEM imaging revealed that unlike the c-Ti_1−xAl_xN_y and stoichiometric w-Ti_1−xAl_xN, the decomposition pathway of w-Ti_1−xAl_xN_y is more complex, specifically that wurtzite Ti_1−xAl_xN_y decomposes by forming semicoherent intermediary MAX-phase nanoprecipitates, which then segregate into pure c-TiN and w-AlN phases after high-temperature annealing and intermetallic TiAl nanoprecipitates.

Wurtzite Ti_1−xAl_xN_y films were grown onto 85 nm-thick TiN(111) seed layers on MgO(111) and Al₂O₃(0001) c-cut substrates (10 × 10 × 0.5 mm³ of >99.5% purity). These single crystal TiN seed layers were deposited using a Mantis ultrahigh vacuum DC magnetron sputtering system with a base pressure of less than 8 × 10⁻¹⁰ Torr (1 × 10⁻⁷ Pa).²¹ Prior to the deposition of the TiN seed layers, the substrates were subjected to systematic chemical precleaning with trichloroethylene, acetone, ethanol,²² and a 30 min vacuum anneal at 800 °C in the deposition chamber. The TiN seed layers were deposited afterward by opening the substrate shutter at the same temperature of 800 °C using a single element Ti target (99.9% purity) in a 20 sccm argon and 4 sccm nitrogen gas mixture, resulting in a working pressure of 0.3 Pa and 30 min deposition time. Further details of the TiN deposition are outlined in Ref. 21. The TiN seed layer was deposited as a barrier to diffusion of Al into the MgO substrate to prevent MgAl₂O₄ spinel formation.^23,24

Different TiAl alloyed targets [target composition (Al/Ti at. %): 67/33, 75/25, 85/15, and 95/5] from Plansee GmbH (76.2 mm diameter) were used to sputter-deposit wurtzite Ti_1−xAl_xN films with different Al-content. The w-Ti_1−xAlxN films were grown in the same deposition chamber as the seed layers but at a temperature of 700°C while keeping the substrates at floating potential and rotating them at 10 rpm. This temperature was used to strike a balance between minimizing decomposition and maximizing crystallinity in the wurtzite films, while rotation was employed to enhance film uniformity. The substrate temperature was measured through a calibrated thermocouple placed in the vicinity of the substrates. Before every deposition, the alloyed targets were sputter-cleaned at low power in 80 sccm Ar plasma with closed shutters to protect the substrate. During the Ti_1−xAl_xN deposition, the argon and nitrogen flows were also set to a constant 20 and 4 sccm, respectively, resulting in 0.3 Pa working pressure. The target DC current was set to a constant-current mode of 250 mA. All Ti_1−xAl_xN depositions were carried out for 1 h. Reference binary AlN films were also grown on TiN-coated substrates as a reference using the same sputtering conditions mentioned above but at a deposition temperature of 800 °C to increase crystal quality. In each deposition, the growth temperature was attained through a consistent heating rate of 20 °C/min, regulated by a proportional integral differential controller in tandem with a thermocouple feedback loop. After deposition, the sample was slowly cooled to room temperature before transferring it to a load-lock chamber where the system is vented.

The wurtzite thin films were cut into 2.5 × 5 mm² pieces and then individually isothermally annealed at 850, 950, 1100, or 1200 °C using a custom-made tube furnace with a base pressure of ∼5 × 10⁻⁴ Pa. Annealing was carried out at a rate of 20 °C/min and a dwell time of 2 h. After annealing, the films were cooled to room temperature at a rate of 50 °C/min and then purged with an N₂ gas during venting to attain atmospheric pressure.

The elemental compositions and depth profiling of the thin films were measured by time-of-flight elastic recoil detection analysis (ToF-ERDA) at the Tandem Laboratory of Uppsala University. The measurements were carried out using a 36 MeV ¹²⁷I⁸⁺ beam at a recoil angle of 45° and an incidence angle of 67.5° with respect to the sample surface normal. Data analyses were performed with the potku 2 software.²⁵ 

X-ray diffractograms (XRDs) of the thin films were recorded using a PANalytical Empyrean diffractometer and Cu-K_a radiation. Data for phase analysis were recorded with a Channel-cut 2-bounce Ge(220) monochromator as primary optics, while azimuthal phi-scans were measured using an x-ray lens and a parallel plate in point focus.

A Zeiss Sigma 300 scanning electron microscope (SEM) was used for morphological plan view imaging. An aberration-corrected FEI Titan³ scanning transmission electron microscope (STEM) operated at 300 kV was used for high-resolution imaging. Three detectors were simultaneously used during STEM imaging: bright field (BF) and two high-angle annular dark field (HAADF) detectors. The high-resolution STEM images were obtained with a camera length of 29.5 mm, a 21.4 mrad semiangle probe, and 54 pA beam current. Compositional maps were obtained from an attached Bruker SuperX energy dispersive x-ray (EDX) detector. The TEM lamellae were prepared by an in situ lift-out technique in a FEI Helios Nanolab 650 focused ion beam (FIB)-SEM dual beam system. STEM image analyses and fast Fourier transform (FFT) processing were carried out using gatan digital micrograph dm3 software.

Nanoindentation was carried out using a Hysitron TI950 TriboIndenter equipped with a calibrated Berkovich diamond tip to determine hardness and elastic modulus. Indentation was performed under a fixed load range of 8–12 mN to prevent substrate interference. The Oliver and Pharr method²⁶ was employed to extract the hardness and elastic modulus values from the unloading curve. The reported values represent the average of at least 25 indentations per sample, along with their corresponding standard deviations.

The results of the compositional analysis performed by ERDA for the two sets of series are summarized in Table I. Increasing the Al content of the target increased the amount of Al in the films. No concentration gradients induced during growth could be detected. Instead, the Ti, Al, and N concentrations in the wurtzite films remained constant throughout the film thickness (see Fig. S1⁶⁰). However, the Ti:Al ratios of the targets were not exactly replicated in the resulting films similar to magnetron-sputtered films deposited from TiAl alloy targets.²⁷ The nitrogen composition (y) on the films ranged from 0.82 to 0.86. All the sputter-deposited films contain impurities. Including the impurities on the nitrogen sublattice in the stoichiometry estimation (y1) does not significantly change the trend, except for the dual-phase Ti-rich sample. Annealing of the films from 950 to 1200 °C did not significantly change the metal to N ratio.

Figures 1(a) and 1(c) show the x-ray diffractograms of the as-deposited films at varying aluminum compositions (x) on MgO(111) and Al₂O₃(0001) substrates, respectively. Reference lines are superimposed to indicate the peak positions of relevant binary cubic and wurtzite phases (c-TiN, c-AlN, and w-AlN) as well as MgO and Al₂O₃. For all samples, the peak found at ∼36.6° corresponds to 111 of the TiN seed layers. For films with Al compositions of x = 0.86, 0.94, and 0.98, the peak found at ∼36° corresponds to the wurtzite Ti_1−xAl_xN_y 0002, which is shifted to lower angles as the aluminum content is increased. This slight increase in the out of plane lattice parameter as aluminum content is increased is expected since the films approach a large c/a wurtzite lattice ratio. Regardless of a substrate, the sample with x = 0.79 shows a broad hump between 37° and 38°. This peak is attributed to either a high Al content c-Ti_1−xAl_xN phase with preferred (111) orientation or $( 10 1 ¯ 1)$-oriented w-Ti_1−xAl_xN grains. Subsequent transmission electron microscopy analyses shown in the supplementary material⁶⁰ reveal the presence of $( 10 1 ¯ 1)$-oriented w-Ti_1−xAl_xN crystallites growing on an epitaxially (0001)-oriented w-Ti_1−xAl_xN layer.

Figures 1(b) and 1(d) show representative φ-scans of the wurtzite films and the TiN seed layer with respect to MgO(111) and Al₂O₃(0001)substrates, respectively. Similar scans were obtained for all the other samples, confirming the epitaxial growth on the TiN/substrate, and revealing the epitaxial relationship. For the films deposited on MgO (111), the 002 off axis Bragg peaks of the TiN layer are inclined 54° with respect to the growth direction and shows a threefold symmetry, while the $10 1 ¯1$ off axis peak of the wurtzite TiAlN film is inclined ∼62° from the [0001]-growth direction and exhibits a sixfold symmetry. Thus, the epitaxial relationship on MgO(111) was identified to be^28,29 $w - T i 1 − x A l x N(0001)[ 11 2 ¯ 0] / / c - TiN(111)[ 1 1 ¯ 0] / / c - MgO(111)[ 1 1 ¯ 0]$⁠.

For films grown on c-plane Al₂O₃, a 30-degree shift was observed between the film peaks and the substrate peaks. This suggests that the growth of the TiN and wurtzite TiAlN layer on Al₂O₃ is rotated 30° in-plane. The 30° shift was also reported previously for TiN and other nitrides grown on sapphire,^30–32 which was attributed to the large lattice mismatch between them. Additionally, the 002 off axis φ scans of the TiN layer reveal sixfold symmetry instead of threefold symmetry [see Fig. 1(d)]. This is attributed to the two equivalent ways of placing the epitaxial (111) plane of a cubic structure (with threefold symmetry) onto the (0001) plane of a hexagonal structure (with sixfold symmetry), resulting in a mirror symmetry that is also observed in reference.³³ The epitaxial relationship is as follows: $w - T i 1 − x A l x N(0001)[ 11 2 ¯ 0] / / c - TiN(111)[ 1 1 ¯ 0] or c - TiN(111)[ 1 ¯ 10] / / A l 2 O 3(0001)[ 1 1 ¯ 00]$⁠.

Figure 2 shows the plan view morphology of the as-deposited films deposited on two different substrates. The morphology of the TiN (111) and AlN (0001) films is also included for comparison. Similar morphologies were obtained for the films of the same composition regardless of the substrates used. Smoother surfaces were obtained when the Al content, x, was increased. The Ti_0.21Al_0.79N_y films exhibited the roughest surface among the samples. Through HAADF-STEM imaging, it was revealed that these rough surfaces are due to crystallites of wurtzite Ti_0.21Al_0.79N_y that are oriented in the $10 1 ¯1$ direction (see Fig. S2).

Figure 3 shows the evolution of the XRD patterns of the films with respect to increasing annealing temperature. For the samples with x = 0.79, 0.86, and 0.94, discernible changes in symmetry of the 0002 wurtzite peaks can be observed when compared with the peaks of the as-deposited state; i.e., they become less symmetrical as the films are annealed at higher temperatures. Peak asymmetry (pointed by black arrows) is most evident for the w-Ti_0.14Al_0.86N_y and w-Ti_0.06Al_0.94N_y films annealed at 850–1100 °C.

Figure 4(a) shows overview cross-sectional HAADF-STEM (compositional Z-contrast) micrographs of the w-Ti_1−xAl_xN_y films showing the evolution of the microstructure with increasing annealing temperature. In the as-deposited state, the films exhibit minimal Z-contrast, suggesting homogenous composition. At 950 °C, there is notable appearance of round features, which exhibit a brighter contrast than the rest of the film and were originally not present in the as-deposited state. At 1200 °C, these bright round features increase in number. Figure 4(b) shows the EDX compositional maps of the top regions of the Ti_0.14Al_0.86N_y and w-Ti_0.06Al_0.94N_y films [samples within the box in Fig. 4(a)], which uncovers that the bright round features in the film are locally rich in titanium. The intensity changes in the nitrogen composition map suggest that the titanium-rich areas may be slightly deficient in nitrogen.

Figure 5 compares the high-resolution HAADF-STEM images of the (a)–(c) Ti_0.14Al_0.86N_y and (e)–(g) w-Ti_0.06Al_0.94N_y films annealed at 950 °C viewed from the $[ 11 2 ¯ 0]$ zone axis. Due to Z-contrast, the brighter contrast regions and atoms correspond to regions with relatively higher titanium concentration, while the darker contrast indicates the presence of more aluminum. Figures 5(a)–5(c) and 5(e)–5(g) reveal that the round features observed at low resolution [Fig. 4(a)] exhibit an organized MAX-like structure. Figure 5(d) shows an atomic model of a Ti₂AlN MAX phase viewed from the same projection as an example. Overall, these decomposed microstructures show the formation of nanoscale MAX-like precipitates following the Ti₂AlN stacking (and some with TiAl and Ti₄AlN₃, described more in Figs. 6–8).³⁴ It is also worth noting how the width of the MAX-like precipitates in Ti_0.14Al_0.86N_y is slightly wider in size than the ones in w-Ti_0.06Al_0.94N_y.

Figure 6(a) shows an atomic-resolution image of the w-Ti_0.14Al_0.86N_y sample after annealing to 950 °C showing nanoprecipitates formed in the wurtzite matrix. The corresponding FFTs of the regions enclosed inside the white squares in Fig. 6(a) are labeled with their matching numbers. Regions 1, 3, and 4 exhibit the representative hexagonal pattern viewed from a $[ 11 2 ¯ 0]$ zone axis of the wurtzite matrix. The FFTs of regions 2, 5, and 6 show the appearance of additional spots, indicative of phases different from wurtzite and its other orientations. Figure 6(b) is a magnified view of region 2 where the precipitate was identified to be TiAl viewed along the [100] zone axis. Figure 6(c) magnifies region 5, where a mix of the MAX-phase Ti₂AlN and Ti₄AlN₃ stacking is identified, similar to the intergrown structure observed by Schramm et al.³⁴ The characteristic location of Al atoms between 2- and 4-layer Ti atoms as well as the zigzag stacking sequence in these MAX phases³⁵ are also observed. To guide the reader, atomic models were overlaid on the micrographs.

Figure 7 summarizes the observed decomposition route for w-Ti_1−xAl_xN_y with x = 0.86 and 0.94. The combination of XRD and STEM-EDX analyses shows that nitrogen-deficient wurtzite TiAlN (14%–18% N vacancies) decomposed by a nucleating semicoherent MAX phase and TiAl nanoprecipitates when isothermally annealed in vacuum at 850–1100 °C for 2 h. At 1200 °C, semicoherent c-TiN starts to nucleate. Note that the rates of precipitation are highly dependent on the Al content, as seen in the size differences of the precipitates (see Figs. 4 and 5).

Figure 8 summarizes the different coherency relationships observed between the nanoprecipitates and the surrounding wurtzite matrix. The yellow arrows indicate planar faults or extra lattice planes found from the inverse FFT lattice fringes. Figures 8(a) and 8(e) show the semicoherency between MAX-phase Ti₂AlN and w-TiAlN where the extra planes form misfit dislocations at the boundary of the precipitate. Figures 8(b), 8(c), 8(f), and 8(g) show the relationship between MAX-phase TiAlN with TiAl precipitates. Two distinct projections of TiAl, both of which exhibit semicoherence with the MAX-phase TiAlN, were observed: [100]_TiAl and [111]_TiAl. These two directions are ∼60.29° apart, which is equivalent to the angle between the $[ 2 1 ¯ 1 ¯ 0]$ and $[ 1 ¯ 2 1 ¯ 0]$ directions of the hexagonal MAX phase³⁶ giving us the written orientation relationship in relationship (ii). In Figs. 8(d) and 8(h), we observe that the relationship between c-TiN and the purifying w-AlN matrix at 1200 °C is similar to the reported orientation relationship in c-Ti_1−xAl_xN decomposing and coarsening to c-TiN and w-AlN.³⁷ Note that the c-TiN domain in Fig. 8(d) still contains stacking faults (yellow arrows marked by SFs), which are attributed to the incomplete transition from the ABABAB stacking of the wurtzite lattice to the ABCABC stacking of a cubic lattice.

Figure 9 shows the nanoindentation hardness evolution of the films versus annealing temperature. In the as-deposited state, the hardness of the films generally increases with increasing aluminum content, regardless of the substrate. This trend, however, does not include the pure w-AlN film, which shows lower hardness (∼22–23 GPa) than the single crystal w-Ti_1−xAl_xN_y films. After annealing to temperatures higher than the deposition temperature, the hardness of the films starts to increase. The maximum hardness was reached at a lower temperature (850 °C) for the x = 0.79 and 0.86 samples, irrespective of the substrate. The samples with x = 0.94 and 0.98 reach their highest hardness at 950 °C. For comparison and as a baseline, the w-AlN (x = 1.0) films exhibited minimal changes in hardness with increasing annealing temperatures. It is also worth noting that the polycrystalline wurtzite film with two different crystallographic orientations in the growth direction (x = 0.79) generally exhibited lower hardness values compared to the other films.

Here, we have shown that single crystal nitrogen-deficient w-Ti_1−xAl_xN_y films can be epitaxially stabilized through reactive magnetron sputtering. The use of a Ti_1−xAl_x target with varying high Al contents (x = 0.67), combined with high deposition temperatures, and a TiN(111) seed layer as a template, enabled epitaxial growth of the wurtzite nitride films. Furthermore, the habit planes of the cubic-(111) layers of TiN match the w-(0001) layers allowing them to have a coherent relationship during growth.³⁸ The mirrored symmetry of the TiN seed layer on Al₂O₃ did not appear to affect the wurtzitic growth of Ti_1−xAl_xN_y as well as its mechanical properties in comparison with those grown on MgO. However, the stoichiometries of the resulting films did not exactly match the target compositions. Such differences between the stoichiometries of the compound target and the film have also been observed in similar sputtered Ti-Al-N films²⁷ and are affected by many factors, such as the number and angle of sputtered atoms from the target,³⁹ their path via plasma toward the substrate surface,⁴⁰ and their interaction with the substrate.⁴¹ 

Previous attempts to grow stoichiometric epitaxial metastable w-Ti_1−xAl_xN films by Calamba et al.³⁸ have demonstrated the formation of a thin layer of Al-rich c-Ti_1−xAl_xN(111) during the initial stages of growth. This c-Ti_1−xAl_xN(111) layer acts as a stabilizing agent for the subsequent growth of the wurtzite phase. Surprisingly, this growth mode persisted even at a high Al content of x = 0.77³⁸ and under similar deposition temperatures (700 °C). This growth mode occurred despite a high Al content of x = 0.77³⁸ and a similar deposition temperature (700 °C). In our work, we successfully deposited a pure wurtzite layer by achieving a minimum aluminum content of x = 0.79, accompanied by a 17% nitrogen vacancy. Interestingly, this aluminum content is in close proximity to the x = 0.77 composition reported by Calamba et al., where they observed the initial formation of a cubic layer during growth.³⁸ Despite the slight disparity in aluminum content, our films show an inherent ability to grow the wurtzite phase directly, without requiring the nucleation of a transitional c-Ti_1−xAl_xN layer. This observation strongly implies the notable influence of nitrogen vacancies on the growth mode of the films. It was previously shown that under N-deficient conditions, nitrogen atoms tend to adhere to aluminum,⁵ promoting the formation of wurtzite nuclei and suppressing the formation of nuclei with an NaCl structure. This finding underscores the significant impact of nitrogen vacancies on the growth mode of the films.

The observed decomposition of the solid solution w-Ti_1−xAl_xN_y suggests that there is also a significant driving force for separation even in the wurtzite phase consisting of nitrogen vacancies due to a miscibility gap between c-TiN and w-AlN.^42,43 The decomposition observed herein is, however, very different from our observations for stoichiometric wurtzite Ti_1−xAl_xN films, where isostructural decomposition into AlN-rich and TiN-rich wurtzite domains was observed.^42,43 Nonetheless, segregation can be activated by annealing at high temperatures, which allows for the formation of more thermodynamically stable structures through diffusional processes.

The formation of MAX phases (Ti₂AlN and Ti₄AlN₃) from nitrogen-deficient wurtzite films after annealing was also observed in the study of Schramm et al.,³⁴ where an arc deposited polycrystalline Ti_0.37Al_0.63N_0.46 sample containing a significant fraction of a wurtzite phase mixed with cubic phases resulted in Ti₂AlN formation at 1000–1200 °C. Other thin film growth studies of MAX Ti-Al-N have started with an oversupply of titanium and lower aluminum contents to match the stoichiometry of MAX phases. For example, Zhang et al.⁴⁴ and Joelsson et al.⁴⁵ showed that single crystal Ti₂AlN can be epitaxially grown on MgO(111) substrates by using a 2Ti:Al alloyed target and reactively sputtering at 750 and 830 °C, respectively. Meanwhile, Cabioch et al.⁴⁶ have shown that cubic (Ti + Al)/AlN multilayers can transform to Ti₂AlN during vacuum annealing at a low temperature as 600 °C. They proposed that this transformation occurs by intermixing of the (Ti + Al)/AlN multilayers, followed by formation of a metastable solid solution of (Ti,Al)N exhibiting a hexagonal alpha-Ti structure.⁴⁶ Furthermore, it has been reported that nitrogen-deficient conditions usually yield a mixture of inverse perovskite Ti₃AlN and intermetallic phases of TiAl, Ti₃Al, and Ti₂Al₅ during growth⁴⁷ and then undergo solid-state reactions.^34,35 Such solid-state reactions were observed between substoichiometric TiN_y films and the Al₂O₃ (0001) substrate at a pN₂ of 0.2 mTorr and at least 800 °C deposition temperatures.⁴⁸ The study by Schramm et al.³⁴ yielded MAX Ti₄AlN₃ phases via solid-state reactions after annealing free-standing substoichiometric c-Ti_1−xAl_xN_y films (grown by cathodic arc deposition) to 1400 °C. Most of these reports start with a film or target compositions containing higher Ti content than Al.

MAX Ti-Al-N phases can exist in a wide range of compositions and high-temperature ranges, which could explain their formation and stability under certain conditions. For instance, Ti₂AIN has been found to be stable within a temperature range of 700–1600 °C,⁴⁹ while the temperature range in which Ti₄AlN₃ can exist is more limited (∼1300 °C).⁵⁰ In addition, MAX Ti-Al-N phases are also known to accommodate nitrogen^35,51 and Al⁵² vacancies while retaining their hexagonal structures. Meanwhile, the precipitation of TiAl would be favorable in regions of the film with locally high nitrogen deficiency. It is also expected that the formation of w-AlN is thermodynamically favorable and is known to be a competing phase for Al-containing MAX phases.⁵³ 

During the annealing process when diffusion is promoted, N diffuses the fastest, followed by Al and Ti.³⁴ The diffusion of Al to form wurtzite AlN leaves the titanium-rich regions to cluster together and accommodate N-vacancies. Besides the diffusion of the elements, the existing hexagonal structure of the wurtzite film acts as a precursor template for the similar hexagonal structure of MAX phases to form, which also minimizes the interfacial strains between the MAX-nanoprecipitates and the wurtzite matrix.

To summarize, the formation of transitional phases at the intermediate annealing temperatures (850–1100 °C) is heavily influenced by the hexagonal crystal of the wurtzite film acting as a precursor, precipitation of equally stable phases within these temperatures, as well as indication of nitrogen vacancy clustering all enabled by the diffusion of the elements at high temperatures. Ultimately, this exhibits the significant effect of nitrogen vacancies and its effect on the enhancement of diffusion mobilities.

The final transformation from MAX Ti-Al-N and TiAl precipitates to cubic TiN has been observed previously. It has been reported that starting at 800 °C,^54,55 Ti₂AlN films decompose by an outward diffusion of Al from the MAX phase leaving Ti₂N, which then undergoes detwinning into cubic TiN_x. The possible pathway of diffusion for Al was elucidated by Zhang et al.,⁵² where density functional theory calculations have shown that when Al vacancies are accommodated in Ti₂AlN, Al atoms can easily diffuse along the (0001) plane by a vacancy jump. The above-mentioned study by Cabioch et al.⁴⁶ also observed the complete transformation of the Ti₂AlN to TiN by annealing to 750 °C at longer annealing times, which they attributed to the loss of Al and N. In this work, the cubic TiN phase was also recovered after annealing to 1200 °C [see Fig. 7(d)]. This implies that the continued purification of w-AlN and c-TiN phases is favorable at higher temperatures. Interestingly, the TiN remains semicoherent with the surrounding hexagonal lattice, which could explain why hardening is retained up to this temperature.

From these, we can discuss the observed decomposition route of substoichiometric w-Ti_1−xAl_xN_y with 0.86 < x < 0.98 and 0.82 < y < 0.86, as summarized in Fig. 7:

1. At high temperatures (T ≥ 850 °C), we start to see composition fluctuations caused by spinodal decomposition,⁴³ which act as precursors to the MAX phases and intermetallics, i.e., Ti-rich and Al and N-depleted regions compared to the starting w-Ti_1−xAl_xN_y. The diffusion of Al and Ti results in the formation of regions with locally high Ti and low Al concentrations. Nitrogen also diffuses quickly at these temperatures, allowing for the formation of more stable arrangements of nitrogen vacancies. Aluminum tends to diffuse away from Ti-rich w-Ti_1−xAl_xN_y to form its more stable wurtzite-AlN phase. Nitrogen vacancies are disfavored in AlN, forcing them to be accommodated in the titanium-rich regions instead.

2. The diffusional processes continue to generate locally Ti-rich, N-poor regions and AlN-rich regions. It has been previously found that in Ti-rich, N-poor regions, N prefers to stick to Ti, while Al tends to form TiAl.⁵ The hexagonal template of the original wurtzite phase further promotes the self-organization of the clustered Ti-rich, N-poor regions into hexagonal MAX phases, which are thermodynamically stable and minimize interfacial energies. These nucleated MAX phases can effectively accommodate nitrogen deficiency, with Ti₂AlN and Ti₄AlN₃ phases requiring only 25% and 37.5% nitrogen, respectively, to form. The TiAl intermetallic phase is also stable and can have a semicoherent relationship with the hexagonal lattice,³⁶ without necessitating any nitrogen atoms in its structure.

3. At even higher temperatures (∼1200 °C), the continued outdiffusion of Al from the MAX phases (deintercalation) and TiAl precipitates to form the more stable w-AlN further leaves the precipitates with only Ti atoms and N-vacancies. This process is favorable because TiN_y is known to be metastable in a wide range of N concentrations,⁵⁶ and the barrier for Al diffusion along the MAX basal planes is lower.⁵² The possibility of maintaining the coherency of the forming cubic TiN_y domains with the surrounding w-AlN regions is facilitated by stacking faults [see Fig. 8(d)], similar to how stacking faults facilitate the cubic-to-wurtzite transition.⁵⁷ 

Our results have shown that even in the pure wurtzite phase of TiAlN, nitrogen vacancies affect the decomposition behavior and, therefore, the mechanical properties at high temperatures. As briefly mentioned above, the maintained coherency or semicoherency of the forming precipitates with the surrounding wurtzite lattice produces elastic strains that obstruct dislocation motion⁵⁸ and contribute to hardening. Such pinning effects were also observed for TiAl forming semicoherent interfaces with Ti₂AlC MAX precipitates.³⁶  Figure 8 summarizes the examples of the coherent and semicoherent boundaries forming between the nanoprecipitates and THE surrounding matrix.

Just like in FCC materials, atoms, vacancies, or interstitials can agglomerate into basal planes during diffusion, causing either excess or missing planes in the hexagonal lattice, forming what is known as Frank partial dislocations.⁵⁹ In hexagonal close-packed materials, the formation of these Frank partials typically necessitates the formation of different types of stacking faults.⁵⁹ As seen in the analyses of lattice planes in Figs. 8(a)–8(c), there are extra basal (0001) planes at the interfaces between the different phases. On the other hand, there are also interfaces that are smooth and devoid of defects. Still, the changing lattice parameters of the precipitates with the surrounding wurtzite matrix will produce elastic strains. The combination of elasticity differences between the phases, misfit dislocations, and elastic strains all obstructs a dislocation glide, which explains a significant increase in hardness after annealing. Meanwhile, the observed formation of TiAl twin boundaries [Figs. 8(b) and 8(c)] as well as the stacking faults in c-TiN [Fig. 8(d)] points out possible routes for tailoring ductility and brittleness balance in w-Ti_1−xAl_xN_y through controlled heat treatments.

In summary, we investigated the thermal stability and decomposition of nitrogen-deficient epitaxial wurtzite Ti_1−xAl_xN_y films with Al-contents from x = 0.79 to 0.98 and y = 0.82 to 0.86 grown by reactive magnetron sputtering. The nitrogen-deficient films form a homogenous single-phase solid solution of wurtzite TiAlN and do not show segregation in the as-deposited state. Annealing of the films to temperatures of 850 °C up to 1200 °C resulted in increased hardness by about 4–6 GPa, indicating age hardening even in the wurtzite single-phase form. Subsequent atomic-resolution HAADF-STEM and EDX analyses revealed that during annealing from 850 °C, the films decomposed by forming semicoherent MAX phases and TiAl nanoprecipitates, different from the observed decomposition behavior for stoichiometric wurtzite TiAlN. The nanoprecipitates remained semicoherent with the surrounding wurtzite matrix, which provided good sites for dislocation pinning and hardening. Even up to 1200 °C, the high hardness was maintained and was attributed to the preserved semicoherency between the fully segregated c-TiN and w-AlN domains. Our findings contribute to the knowledge on effects of vacancies on wurtzite TiAlN, a phase that has been relatively understudied until now.

The authors acknowledge the support of ViNNOVA (FunMat-II Project Grant No. 2016–05156), the Swedish Research Council (VR Grant Nos. 2017–03813, 2017–06701, and 2021–00357), and the Swedish government strategic research area grant AFM–SFO MatLiU (No. 2009–00971).

The authors have no conflicts to disclose.

J. Salamania: Conceptualization (lead); Data curation (equal); Formal analysis (lead); Investigation (lead); Methodology (lead); Writing – original draft (lead). A. F. Farhadizadeh: Data curation (supporting); Investigation (supporting); Writing – review & editing (supporting). K. M. Calamba Kwick: Formal analysis (supporting); Investigation (supporting). I. C. Schramm: Formal analysis (supporting); Resources (supporting); Writing – review & editing (supporting). T. W. Hsu: Methodology (supporting); Resources (supporting). L. J. S. Johnson: Resources (supporting); Supervision (supporting); Writing – review & editing (supporting). L. Rogström: Project administration (supporting); Resources (supporting); Supervision (supporting); Writing – review & editing (supporting). M. Odén: Funding acquisition (lead); Project administration (lead); Resources (lead); Supervision (lead); Writing – review & editing (lead).

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