Temperature driven evolution of thermal, electrical, and optical properties of Ti–Al–N coatings

Abstract

Monolithic single phase cubic (c) Ti_1−xAl_xN thin films are used in various industrial applications due to their high thermal stability, which beneficially effects lifetime and performance of cutting and milling tools, but also find increasing utilization in electronic and optical devices. The present study elucidates the temperature-driven evolution of heat conductivity, electrical resistivity and optical reflectance from room temperature up to 1400 °C and links them to structural and chemical changes in Ti_1−xAl_xN coatings. It is shown that various decomposition phenomena, involving recovery and spinodal decomposition (known to account for the age hardening phenomenon in c-Ti_1−xAl_xN), as well as the cubic to wurtzite phase transformation of spinodally formed AlN-enriched domains, effectively increase the thermal conductivity of the coatings from ∼3.8 W m⁻¹ K⁻¹ by a factor of three, while the electrical resistivity is reduced by one order of magnitude. A change in the coating color from metallic grey after deposition to reddish-golden after annealing to 1400 °C is related to the film structure and discussed in terms of film reflectivity.

1. Introduction

Physically vapor deposited (PVD) solid solutions of Ti_1−xAl_xN are an industrially important class of synthetic materials which find use in a wide field of applications, from protective coatings in mechanical engineering for tooling and cutting operations [1], heaters [2] in electronic components, diffusion barrier coatings in energy devices [3–5], and optical applications in solar selective absorbers [6]. The wide range of applications of Ti_1−xAl_xN results from a combination of the mostly metallic character of cubic (fcc, B1 structure, Fm-3m space group) TiN with the semi-conducting behavior of hexagonal (wurtzite, B4 structure, P6₃m space group) AlN. Since Al is insoluble in TiN and vice versa, deposition conditions far from thermodynamic equilibrium (e.g. deposition temperatures T_dep of ∼0.2–0.3 of the melting temperature T_melt and extremely high cooling rates of ∼10¹³ K s⁻¹ [7], etc.) are required to synthesize supersaturated metastable Ti_1−xAl_xN solid solutions. This synthesis results in the industrially preferred Ti_1−xAl_xN cubic modification, which can be obtained by Al substitution of Ti in the metallic sub-lattice at molar fractions x of ∼0.65–0.75 [8]. For higher Al contents the metastable cubic solubility limit is exceeded, resulting in the development of a dual phase structure of c/w-Ti_1−xAl_xN, before a further increase in x leads to single phase w-Ti_1−xAl_xN.

Within the cubic regime an increasing amount of Al on Ti lattice sites typically results in increasing hardness, oxidation resistance, thermal conductivity, and electrical resistivity [9–11]. At the same time, the driving force for decomposition into the stable binary nitrides (the mixing enthalpy H_mix) of supersaturated c-Ti_1−xAl_xN drastically increases with increasing amounts of Al in the coating [12]. For many applications of c-Ti_1−xAl_xN the operating temperatures are above T_dep, which requires a detailed understanding of the structure–property evolution with temperature.

In the recent past much research activity [13,14] has been directed at exploring the age hardening phenomenon in c-Ti_1−xAl_xN, which accounts for an increase in hardness of the coating material with increasing temperature. It has been shown in detail that thermally activated diffusion processes lead to a series of decomposition phenomena. Such processes involve recovery of deposition-induced defects, followed by spinodal decomposition, implying a hardness increase due to isostructural decomposition into c-TiN- and c-AlN-enriched domains [15,16]. These reactions involve significant lattice strain between the coherent cubic domains and are responsible for the hindering of dislocation motion. Since these effects occur at temperatures up to ∼900–1000 °C the age hardening of c-Ti_1−xAl_xN has huge potential for cutting and milling operations. At higher temperatures, however, recrystallization and the development of a dual phase structure composed of c-TiN and non-coherent w-AlN and further structural coarsening results in deteriorating mechanical properties. While the mechanical properties of c-Ti_1−xAl_xN during age hardening have been explored in detail, this work focuses on correlation of the temperature-driven structural evolution with the thermal, electrical and optical properties, which is based on the increasing need for thermal barrier coatings in the electronics industry and the potential of c-Ti_1−xAl_xN thin films in solar devices operated at high temperatures.

2. Experimental

Cubic Ti_1−xAl_xN thin films with an Al content close to the metastable solubility limit were synthesized. These films were deposited from a metallurgically produced powder Ti–Al target (99.9% purity, Plansee AG, with a Ti:Al ratio of 1:1) on two different substrates (low alloy steel foil with a thickness of 0.2 mm, and polycrystalline Al₂O₃ platelets 20 × 7 × 0.5 mm). Film synthesis was achieved in a laboratory scale sputter deposition plant at a constant bias voltage of −60 V, a T_dep of 500 °C, and a mixed Ar/N₂ plasma discharge at a N₂ partial pressure of 0.12 during deposition. The film thickness was ∼3 μm with a composition of Ti_0.46Al_0.54N, as determined by energy dispersive X-ray spectroscopy and atom probe tomography [15,16]. More detailed information on the deposition conditions can be found in Rachbauer et al. [17]. In order to determine the physical properties of the coatings as a function of temperature individual films on Al₂O₃ substrates were annealed in vacuum for 1 min. The annealing temperature T_a was varied in the range 500–1400 °C in 100 °C steps, applying a heating rate of 20 K min⁻¹ and a cooling rate of 50 K min⁻¹. Similar heating and cooling conditions were applied to the powdered coating material for differential thermal (Linseis DTA) and X-ray diffraction (XRD) analyses (Bruker D8 diffractometer, Cu K_α radiation) after resolving the Ti_0.46Al_0.54N films from the steel foil substrate in 10 mol.% nitric acid. By this means it was possible to relate the observed film structure (from powder XRD) to the electrical, thermal, and optical characteristics as a function of T_a up to 1400 °C, as the physical properties were subsequently determined at room temperature on individual samples (on Al₂O₃ substrate) of the respective annealed state.

The electrical resistivity was measured with a Jandel four point probe according as in Rachbauer et al. [16], while a Perkin Elmer double beam UV/vis/NIR spectrophotometer equipped with an Ulbricht sphere was utilized to determine the reflectance of the Ti_0.46Al_0.54N films. Hemispheric reflectance spectra were recorded at normal incidence in the wavelength region between 250 and 2000 nm (step size 2 nm).

After measuring the electrical and optical film properties the thermal conductivity was determined by a two color time domain thermoreflectance (TDTR) technique. An approximately 80 nm Al overlayer was deposited by sputtering on the surface of the samples to provide a thermal transducer layer for laser pump probe experiments. The TDTR experiment was a modified version of a previously reported system [18], and only details of the two color modifications will be given here. The output of a mode-locked Ti:sapphire laser (λ = 787 nm) is split into a pump and a probe beam. The pump beam is first sent through a pulse compressor (for correction of pulse stretching effects) and then through an electro-optic modulator (EOM), which imposes a square wave pulse train with a frequency of 9.8 MHz. The pump beam is then aligned along a mechanical translation stage to systematically alter the timing between the pump and probe pulses. The probe beam is sent through an optical parametric oscillator for wavelength modification (λ = 700 nm). Both the pump and probe beams also have half-waveplate/polarizer combinations to arbitrarily control the beam intensity. Both beams are then focused to a spot size of ∼50 μm diameter at a 45° angle to the sample. The reflected probe beam is spatially filtered, recollimated, and sent through a 750 nm shortpass optical filter to exclude scattered pump beam light (polarization filtering is not required). Finally, the probe beam is passed through a neutral density filter (optical density = 1.0) and focused onto a silicon (Si) photodiode detector. The output of the detector is sent to the input of a dual phase, radiofrequency (RF) lock-in amplifier which has its reference channel connected to the same electronic signal that drives the EOM. The scans and data acquisition are computer controlled by means of an in-house produced LabVIEW program.

Analysis of the data to extract thermal conductivities was accomplished with a frequency domain model [19] in which the ratio of the in-phase to out of phase lock-in amplifier signals is calculated as a function of time:

$$\frac{V_{\text{in}}}{V_{\text{out}}}=\frac{{\sum }_{-m}^m(\mathrm{\Delta }T(m/\tau +f)+\mathrm{\Delta }T(m/\tau -f))\exp (i2\pi \mathit{mt}/\tau )}{i{\sum }_{-m}^m(\mathrm{\Delta }T(m/\tau +f)-\mathrm{\Delta }T(m/\tau -f))\exp (i2\pi \mathit{mt}/\tau )}$$ (1)

Here m is an integer denoting summation over pump pulses, τ is the time between unmodulated laser pulses (12.5 ns), f is the modulation frequency (9.8 MHz), and t is the time delay between the pump and probe pulses (Eq. (1) is also multiplied by a phase shift e^i2πft). The function ΔT is calculated with the Feldman matrix algorithm as fully explained in Cahill [19]. Data for pump advance times earlier than t = 200 ps were not taken into account since electron–phonon coupling was not equilibrated, which allows the Al film to reach a uniform temperature. Picosecond acoustics also perturb this regime (which facilitates direct measurements of the Al thickness). For the model a four layer system was used comprising two layers for the Al film (as explained in Cahill [19]), an interfacial conductance layer, and a semi-infinite Ti_0.46Al_0.54N sample layer. This four layer system was used as the sample thicknesses (∼3 μm) were much larger than the estimated thermal penetration depth in the TDTR experiment $\sqrt{D/\pi f}\approx 200\text{mm}$. Here D is the thermal diffusivity of the sample layer and f is the TDTR modulation frequency. Volumetric heat capacities of the samples were determined from ab initio calculations. In particular, the generalized gradient approximation, as implemented in the Vienna Ab Initio Simulation Package [20,21], was used to determine the phonon frequencies within the quasi-harmonic approximation. The free energy and subsequently the heat capacities were obtained following the methodology described in Grabowski et al. [22]. For the samples studied the TDTR data were acquired from five locations on each sample surface. The scans were individually modelled, and an average thermal conductivity ± standard deviation value was calculated for each sample.

3. Results and discussion

3.1. Decomposition phenomena and structure

The measured heat flow exhibits an exothermal reaction throughout the investigated temperature regime (400 °C < T < 1500 °C), as shown in Fig. 1a, which can be deconvoluted into five partial reactions by fitting the measured heat flow to Gaussian functions (designated DSC-A, -B, -C, -D, and -E, respectively). This observation is in excellent agreement with literature reports [14,23], where the first four individual reactions are referred to as recovery (DSC-A), spinodal decomposition (DSC-B + DSC-C), and the transformation of c-AlN to w-AlN (DSC-D). The correlation between the structure of the coatings (see Fig. 2) and the heat flow as a function of T_a allows verification of the four individual reactions. After deposition the Ti_0.46Al_0.54N coatings exhibit a single phase cubic structure with columnar morphology, as observed by transmission electron microscopy (not shown here) and XRD (see Fig. 2), with a column grain diameter of ∼25–50 nm. The initial lattice parameter of 4.181 Å in the as-deposited state decreases to a minimum of ∼4.16 Å during the course of annealing at T_a = 1000 °C for 1 min. This decrease is due to recovery effects and isostructural phase separation. Therefore, the partial reaction DSC-A mainly stems from recovery of deposition-induced point and line defects above T_dep (500 °C). The overlapping partial reactions DSC-B (650 °C < T_a < 950 °C) and DSC-C (800 °C < T_a < 1100 °C) cover over a wide temperature window and stem from diffusion of lattice atoms at higher T_a which induces the formation of c-AlN-enriched and c-TiN-enriched domains due to spinodal decomposition [14,16].

Fig. 1.

(a) Heat flow during DSC of Ti_0.46Al_0.54N, including the sum fit of the five exothermic partial reactions during thermal exposure up to 1500 °C. (b) Sheet resistance, (c) thermal conductivity, and (d) reflectance minima of Ti_0.46Al_0.54N as a function of T_a.

Fig. 2.

Structural evolution of Ti_0.46Al_0.54N film powder as a function of the annealing temperatures indicated. Compare the impact of the structural changes, as schematically depicted in Fig. 3 a–e, with the physical properties of the coatings shown in Fig. 1.

Three-dimensional (3-D) atom probe tomography has proved that AlN enrichment (corresponding to DSC-B, along with local TiN-depletion) is faster along high diffusivity paths, such as column or grain boundaries, compared with the grain interior, where higher T_a (DSC-C) are required to enable sufficient lattice diffusion. These two reactions result in the formation of a 3-D interconnected network of c-Ti_1−x−ΔAl_x+ΔN (AlN-rich) and c-Ti_1−x+ΔAl_x−ΔN (TiN-rich) domains [15], which account for the age hardening phenomenon. Fig. 3a and b schematically depicts the local element enrichment involving reactions DSC-A, -B, and -C, as discussed above.

Fig. 3.

Schematic representation of the structural evolution of a c-Ti_1−xAl_xN film with columnar structure as a function of decomposition phenomena. (a) As-deposited state with small chemical fluctuations in the metallic sub-lattice. (b) Formation of c-AlN-rich and c-TiN-rich domains in the grain interior due to spinodal decomposition of c-Ti_1−xAl_xN and enhanced AlN enrichment along high diffusivity paths e.g. the column boundaries. (c) Ongoing isostructural decomposition in the grain interior while the transformation of c-AlN to w-AlN occurs at the high Al containing grain boundaries. (d) Al diffusion out of the c-TiN grains and growth of w-AlN. (e) A reduction in grain boundaries due to coarsening of the dual phase structure (c-TiN and w-AlN).

Further increasing T_a up to ∼1100 °C results in transformation of the c-AlN-rich domains to their stable hexagonal modification of w-AlN (corresponding to DSC-D, 1000 °C < T_a < 1400 °C), which can be seen in the XRD pattern in Fig. 2. The appearance of the w-AlN peaks occurs at their stoichiometric positions, while the peaks corresponding to the AlN-depleted matrix shift to lower angles, indicating continuous diffusion out of Al with increasing T_a. An additional small exothermic reaction (DSC-E) can be identified for T_a > 1250 °C, corresponding to a reduction in grain boundary volume and thus grain growth of the dual phase structure (c-TiN and w-AlN), which is evident from the increasing XRD peak intensities (Fig. 2). The different partial reactions are summarized as a decomposition scheme in Fig. 3a–e.

3.2. Electrical conductivity

The microstructural changes as a function of T_a strongly correlate with significant changes in the physical properties. Local chemical variations in the metallic sub-lattice due to recovery and isostructural decomposition up to T_a ≈ 1000 °C imply lattice densification and the development of an enhanced electrically conductive network of c-TiN-rich domains. The measured sheet resistance (ρ of c-Ti_0.46Al_0.54N thus gradually decreases as a function of T_a due to diffusion of Al out from the c-TiN-enriched domains (from ∼4.6 mΩ cm⁻¹ to the reported sheet resistance value of ∼0.1 mΩ cm⁻¹ [24] for c-TiN thin films, see Fig. 1b). The plateau in the ρ–T_a plot at around ∼1000–1100 °C can be related to the cubic wurtzite transformation of the AlN-rich domains, which causes enhanced electron scattering at the grain boundaries.

3.3. Thermal conductivity

A similar behavior was observed for the thermal conductivity (κ) of c-Ti_0.46Al_0.54N, which was found to slightly increase due to recovery (DSC-A) and with an increasing isostructural decomposition state (DSC-B and -C) owing to the formation of an enhanced thermally conductive c-AlN-rich network of a few nanometers in width [15]. The reported literature values for thin films of κ_c-TiN, κ_c-AlN and κ_w-AlN are ∼2.7 [25], 250–600 [26] and 180–320 [27] W m⁻¹ K⁻¹, respectively. Local AlN enrichment thus results in a slight increase in κ from ∼3.8 W m⁻¹ K⁻¹ after deposition to ∼6.4 W m⁻¹ K⁻¹ at T_a ∼900 °C (see Fig. 1c). A direct interpolation of the literature values for κ_c-TiN and κ_c-AlN would reach ∼150–325 W m⁻¹ K⁻¹ for c-Ti_0.46Al_0.54N. However, this is misleading, as the small domain size of the 3-D interconnected network evolving during annealing significantly lowers the extent of thermal conductivity due to enhanced phonon scattering at the domain interfaces and grain boundaries [28] (compare Fig. 3a and b). This observation is in good agreement with the literature [27], where a sixfold reduction in κ_w-AlN from 320 to 50 W m⁻¹ K⁻¹ is reported for a reduction in grain size from 1 μm to 10 nm. At T_a ⩾ 1000 °C the ongoing c-AlN to w-AlN transformation takes place first by nucleation of nanometer sized w-AlN precipitates (with a ∼2.7 times lower κ than c-AlN) at the grain boundaries, which results in a further increase in phonon scattering centers and, hence, a slight reduction in κ. With increasing T_a the amount of w-AlN phase increases (compare Fig. 3c–e), which implies the formation of a 3-D interconnected network with enhanced thermal conductivity. Hence, a significant increase in κ was observed for T_a ⩾ 1200 °C, at which the reduction in grain boundaries due to coarsening of the dual phase structure reduces the phonon scattering centers.

3.4. Optical reflectance

The thermal evolution of c-Ti_0.46Al_0.54N additionally involves changes in the optical properties, which were explored by reflectivity measurements in this work. The coating reflectivity is also useful as a guide for the eye, as one can directly relate the color of the coating with the Al content. While pure c-TiN coatings exhibit a golden color, which stems from a higher reflectivity in the wavelength region 550–700 nm, increasing the Al content of c-Ti_1−xAl_xN results in a color change towards orange, reddish-brown, and then grey, as reported in the literature [29,30]. Here it is shown that the state of decomposition, and thus the mechanical properties, can be correlated with the color of the coating. While the c-Ti_0.46Al_0.54N films investigated after deposition exhibit a metallic grey color due to low reflectivity values throughout the visible range (380–750 nm), the fully decomposed films exhibit an orange-golden color after annealing to 1400 °C (see Fig. 4). The spectral minima are plotted as a function of T_a in Fig. 1d and corroborate the above discussed findings with respect to recovery, isostructural decomposition, and the impact of the appearance of w-AlN as a second phase. Increasing T_a up to ∼1000 °C induces densification of the lattice and spinodal decomposition, which has almost no effect on the shape of the reflectivity spectrum (although the reflectance minimum shifts slightly to a higher wavelength). A higher T_a results in a shift of the reflectance minimum towards the values reported for c-TiN [29,30]. This shift stems from Al depletion of the TiN-rich phase and the continuous transformation of c-AlN to w-AlN, exhibiting a high transparency of ∼80% between wavelengths of 300 and 1000 nm [30]. In other words, while w-AlN turns transparent, the coating approaches the golden color of c-TiN with increasing decomposition state.

Fig. 4.

Reflectance spectra of Ti_0.46Al_0.54N after annealing at T_a for 1 min. The dual phase above T_a ≈ 1000 °C results in a color change of the coatings from grey to reddish-golden, which stems from the formation of optically highly transparent w-AlN and golden c-TiN. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

4. Conclusions

Monolithic, single phase c-Ti_0.46Al_0.54N thin films are shown to drastically change their physical properties due to a series of exothermic reactions, which are correlated with the actual state of the film structure during vacuum annealing. Recovery and isostructural decomposition up to ∼900–1000 °C, accounting for the age hardening phenomenon in c-Ti_1−xAl_xN, are shown to increase the thermal and electrical conductivities, while the color of the coating remains almost unaffected. However, upon entering the dual phase regime at T_a above ∼1000 °C the induced transformation of c-AlN to w-AlN significantly increases both the electrical and thermal conductivities and further results in a shift of the overall coating reflectance spectrum towards the values of c-TiN. It can thus be concluded that the strong interaction of structure and physical properties requires a detailed knowledge of the decomposition phenomena present in complex nitrides in order to address the needs of application-tailored design of optical, electronic, or mechanical devices.