**2. Materials and Methods**

First 1 µm thick Al1−*x*Sc*x*N layers with Sc concentrations *x* = 0, 0.14, 0.17, 0.23, 0.32, and 0.41 were deposited on single-sided, polished, 100 mm diameter Al2O3(0001) substrates at chuck temperatures of 300–400 ◦C by reactive pulsed DC magnetron co-sputtering (Evatec cluster sputter tool). In addition 99.9995% pure Al and 99.99% pure Sc targets were sputtered in a pure N<sup>2</sup> atmosphere; a constant total power applied to the Al and Sc targets to achieve different Sc concentrations; all other growth parameters were kept constant. The deposition parameters are described in detail in [9,20]. The Sc content with an accuracy of *x* ± 0.02 was determined by energy dispersive X-ray (EDX) spectroscopy (Bruker Quantax, Bruker Corporation, Billerica, MA, USA) in a scanning electron microscope (Zeiss Auriga Crossbeam FIB-SEM, Carl Zeiss, Oberkochen, Germany). The compositional analysis of the samples is described elsewhere [21]. The *θ*/2*θ* scans were obtained using a X-ray diffractometer equipped with a 4-bounce Ge 220 monochromator, a parallel beam X-ray mirror on the incident side, and a PiXcel3D detector (X'Pert<sup>3</sup> MRD XL, PANalytical, Almelo, The Netherlands). An atomic force microscope (NX20, Park Systems, Suwon, Korea) with the AC160TS tip (radius: 2 nm) was used to investigate the surface morphology and determine the roughness using Gwyddion software [22]. Raman spectra were collected using the micro-Raman spectrometer (inVia Qontor, Renishaw, London, UK). The excitation wavelength of the DPSS laser (Renishaw RL532) was 532 nm when used in combination with the 2400 L/mm grating, yielding the spectral resolution of about 0.1 cm−<sup>1</sup> . The power density of about 10 mW·µm−<sup>2</sup> estimated for the 100× objective (NA = 0.75) was used to avoid the heating of samples. The near-excitation tunable (NExT, Renishaw) filter was used, enabling the collection of the anti-Stokes spectra. The baseline stemming from the photoluminescence was subtracted from all obtained spectra using a polynomial function, describing the background signal increasing towards longer wavelengths. The temperaturedependent Raman measurements were recorded using a thermostat, a hot plate-like stage (T96-P, Linkam Scientific Instruments, Tadworth, UK). Due to the heating of the samples from backside of the sapphire substrate, it was important to realise that the heating of the AlScN films corresponded to the temperature setpoints. Provided the outstanding thermal conductivity of sapphire, the temperature in the films was assumed to be equal to that of the sapphire substrate. The calibration and tracking of the sample's temperature was performed using the A1*<sup>g</sup>* Raman-active mode of the sapphire substrate (417.4 cm−<sup>1</sup> [23]). The thermal coefficient of the Raman band was found to be <sup>−</sup>0.015 K·cm−<sup>1</sup> , which is in an excellent agreement with previous studies [24].
