*3.1. Microstructure Investigations of AlN and AlScN Thin Films*

Figure 1 shows the surface structures of AlN and AlScN films grown directly on three different substrates. All AlN samples (Figure 1a–c) show a homogeneous surface with small round grains, which indicates the successful growth of columnar grains with *c*-axis orientation as already reported in multiple studies [11,25,30]. On the other hand, the surface of AlScN on all three substrates (Figure 1d–f) is dominated by crystallites with wedge-shaped structure, implying a poor *c*-axis orientation [25]. In addition, AlScN films seem to be grown slightly better on Si and poly-Si substrates compared to SiO<sup>2</sup> because few areas without misorientated grains can be observed.

To further examine the crystal phase and quality of the samples presented in Figure 1, XRD *θ*-2*θ* scans and *ω* scans are performed and shown in Figure 2. AlN and AlScN 0002 reflections at 2*θ* of around 36◦ [39] are detected. In addition, the reflections of crystalline Si (100) orientation, poly-Si (111) and (220) planes are recorded in samples with corresponding substrates. The full width at half maximum (FWHM) of the AlN 0002 reflection rocking curve for all samples is less than 1.5◦ . This confirms that the AlN films on all investigated substrates are indeed well *c*-axis oriented. In contrast, for all AlScN samples, the measured FWHM values of 0002 reflection are larger than 2.2◦ , which is slightly higher than the reported FWHM of AlScN 0002 reflection (approx. 1.6◦ ) reported in [25,39], which use Ti/Pt and Si as substrates, respectively. This indicates a lower quality of *c*-axis orientation of AlScN. The [0001] crystallographic direction of the out-of-the-plane misoriented grains has been shown to be tilted between 60◦ and 90◦ [40].

By using the given process parameters, the growth of highly *c*-axis oriented AlN directly on smooth surfaces (RMS < 2 nm) of amorphous SiO2, (100) Si and poly-Si wafers can be achieved, despite the different crystallographic texture of the substrate materials, in agreement with several studies [29,33,34,41]. However, a smooth surface alone is not sufficient to grow high quality AlScN films on these substrates as the misaligned grains and high FWHM values are measured. Our preliminary investigations show that the growth of AlScN on Ti/Pt bottom electrode for identical process parameters is stable and highly *c*-axis oriented (FWHM of 1.43◦ ). Consequently, AlScN films are more sensitive to the substrate texture and irregularities. This fits with the observation that the *c*-axis orientation of AlScN even on metallic electrodes decreases significantly with increasing Sc concentration [25].

**Figure 1.** SEM surface view of 1 µm AlN deposited directly on (**a**) SiO2, (**b**) (100) Si, (**c**) poly-Si, and 1 µm AlScN deposited directly on (**d**) SiO2, (**e**) (100) Si, (**f**) poly-Si without a seed layer.without a seed layer.

**Figure 2.** (**a**) *θ*-2*θ* scans of 1 µm AlN and AlScN grown directly on SiO2, Si and poly-Si substrates without a seed layer; (**b**) Results of rocking curve measurements of AlN and AlScN 0002 reflections. The FWHM is determined by fitting a pseudo-Voight profile using the XRD fit module (Python based open source tool for XRD peak fitting [42]).

### *3.2. Microstructure and Piezoelectric Response of AlScN Films with a Thin AlN Seed Layer*

To improve the *c*-axis orientation of AlScN on the investigated substrates, one option is to optimize the deposition parameters. In our previous work [43], we showed that the quality of Al0.73Sc0.27N films on SiO<sup>2</sup> can be significantly improved by increasing the cathode–substrate distance offset. In this work, instead of optimizing the process parameters, an ultrathin AlN seed layer is introduced to improve the growth of *c*-axis oriented AlScN films on these substrates. We consider this approach to be more generally applicable and easier to transfer to different substrates.

The SEM images of 500 nm AlScN with 20 nm AlN seed layer on the investigated substrates are shown in Figure 3. Since the misoriented grains originate close to the substrate surface [25,30], there is no major difference in the number of misoriented grains between 500 nm and 1 µm AlScN films. A homogeneous surface with small grains is observed for all three samples, on which only a small number of misoriented grains is visible. Compared to the samples grown without the seed layer (Figure 1d–f), the *c*-axis orientation of AlScN films is significantly improved. The structural quality of AlN/AlScN films grown on SiO2, Si and poly-Si is characterized using XRD (Figure 4). The FWHM values of AlScN 0002 reflection are slightly below 2 ◦ for all samples, which demonstrates only a moderate improvement in respect to Figure 1d–f. Although there is a small difference in 2*θ* values of AlN and AlScN 0002 orientations [10,11,39], a broadening of the 0002 FWHM due to the reflection from the 20 nm thin AlN seed layer can be expected to be negligible.

**Figure 3.** SEM surface view of 500 nm AlScN grown on (**a**) SiO2, (**b**) (100) Si and (**c**) poly-Si with a 20 nm AlN seed layer.

**Figure 4.** (**a**) *θ*-2*θ* scans of 500 nm AlScN grown on SiO2, Si and poly-Si substrates with the AlN seed layer; (**b**) Results of rocking curve measurements of AlScN 0002 reflections. The FWHM is determined by fitting a pseudo-Voight profile using the XRD fit module (Python-based open source tool for XRD peak fitting [42]).

For further investigation of the local chemical composition and nanostructure at the interfaces, the sample with Si/AlN/AlScN is selected for TEM analysis. The scanning TEM annular bright-field (ABF) micrograph in Figure 5a provides an overview of the film cross-section, showing columnar grain structures of AlN and AlScN layers. The quality of the AlN/AlScN interface is further investigated by high-resolution STEM imaging and elemental analysis. A magnified HRSTEM ABF image of the interface is given in Figure 5b. By using the ABF detector, contrast-rich images with atoms shown by black dots are recorded. The columnar grains with diameters <5 nm growing along the *c*-axis on both sides of the interface are well displayed. However, obtaining a clear image of the interface is limited by the in-plane rotational disorder of the columnar grains and their three dimensional superposition along the finite sample thickness, as well as the patchy contrast spanning 2–3 nm in vertical direction across the interface region. The mosaic tilt along the *c*-axis is additionally visualized in the electron diffraction pattern recorded on the Si/AlN/AlScN multi-layers (see Figure 5c). The displayed intensity distribution can be explained by the superposition of the individual [110] Si, [21¯1¯0] and [11¯00] AlN and AlScN zone axis patterns. The high coherency of the Si lattice causes electrons to scatter into sharp and bright reflections, whereas the different lattice constants of AlN and AlScN, as well as the small out-of-plane mosaic tilts of individual columns and the in-plane rotation of the fiber textured microstructure, result into diffuse and elongated intensities. The chemical composition is examined by elemental maps and profiles of the averaged intensity, as shown in Figure 5. Here, a peak in the oxygen signal is detected directly at the AlN/AlScN interface indicating a partial oxidation of the AlN surface during the vacuum break. Such partially oxidized interface has been reported before on a similar system and could not be avoided even after applying an RF etch cleaning step [44]. However, the oxide interface does not impede high-quality *c*-axis-oriented growth.

**Figure 5.** TEM study of the sample Si/AlN/AlScN. (**a**) STEM ABF overview image showing the columnar grain structures of AlN and AlScN layers on a natively passivated Si substrate; (**b**) HRSTEM ABF image showing structural disorder at the AlN/AlScN interface; (**c**) SAED pattern containing reflections of all layers corresponding to the [110] Si, [21¯1¯0] and [11¯00] zone axes of AlN and AlScN; (**d**) STEM EDS elemental maps with integrated intensity profiles over the region of interest (dashed frame). The O-K map demonstrates the formation of an interfacial oxide layer between AlN and AlScN as well as the native oxide on the Si substrate.

To investigate the effect of 20 nm AlN seed layer on the piezoelectric response, the piezoelectric coefficient d33,*<sup>f</sup>* of 500 nm AlScN on sputtered Ti/Pt without and with the seed layer are measured and shown in Figure 6. The measured average d33,*<sup>f</sup>* of AlScN layer with the seed layer is 8.91 ± 0.03 pm/V, which is slightly lower (4.5%) compared to the one without the seed layer (9.33 ± 0.02 pm/V). However, the homogeneity of the distribution is not affected. The slightly lower piezoelectric coefficient is due to the lower dielectric permittivity of AlN which limits the electric charge storage on electrode plates.

**Figure 6.** Measured *d*33, *<sup>f</sup>* of (**a**) 500 nm AlScN and of (**b**) 500 nm AlScN with 20 nm AlN seed layer on Ti/Ptsput on a wafer level.
