**2. Materials and Methods**

The synthesis of the piezoelectric thin film was carried out using a home-built reactive magnetron sputtering system. The target alloy (Sc0.6Al0.4) was placed 45 mm from the substrate. The process was carried out without intentional heating, and the temperature was monitored using a K-type thermocouple located below the substrate holder. The magnetron was powered using a pulsed DC generator (ENI RPG50), with a pulse width set to 1616 ns.

The polycrystalline diamond (PCD) substrates were synthesized by microwave plasma chemical vapor deposition (MPCVD) on a 500-µm thick Si (001) supporting layer. The single crystal diamond (111) (SCD) was purchased (EDP Corporation, Osaka, Japan) for evaluating its influence on the SAW propagation characteristics.

Prior to the ScAlN synthesis process, the substrates were cleaned using the following two-solvent method. First, the substrates were rinsed in acetone at 60 ◦C, followed by a sonication bath in methanol, both for 5 min. Afterwards, the samples were blown dry with nitrogen (N2) and introduced into the load-lock chamber.

The synthesis chamber was conditioned for three minutes in a pure Argon (Ar) atmosphere (30 sccm), with a discharge power of 500 W and a process pressure of 1.33 Pa. Nitrogen was then introduced, the gas admixture ratio (N2/(N<sup>2</sup> + Ar)) was adjusted to 25%, and the process pressure and discharge power were set to 1.33 Pa and 500 W, respectively, for 3 min. Finally, the process pressure was adjusted to 0.40 Pa. After 3 min, the plasma was turned off, and the substrates were transferred from the load-lock chamber. Afterwards, with the shutter closed, the plasma was ignited again using the synthesis conditions (0.40 Pa, 500 W, 25% gas admixture ratio), and after two minutes, the shutter was moved away, beginning the synthesis process.

The degree of c-axis orientation of the synthesized piezoelectric thin film was assessed via 0002 ω-θ scans using X-ray diffraction (XRD, Phillips X-Pert Pro MRD diffractometer), with Cu Kα<sup>1</sup> radiation (λ = 1.54059 Å), 45 kV, and 40 mA.

The interdigital transducers (IDT) were fabricated using a standard lift-off process. They were patterned with an e-beam lithography system (Crestec CABL-9500C, Hachioji, Japan). Due to the insulating behavior of the ScAlN thin film and the underlying polycrystalline diamond substrates, an organic anti-static layer (Espacer 300Z, Showa Denko K.K, Tokyo, Japan) was spun on top of the e-beam resist to avoid charge accumulation.

The fabricated 1-port resonators comprise a 2000-nm Sc0.43Al0.57N thin film and 110-nm thick (700 nm width) Au IDT (λ = 2.8 µm). The filters were fabricated using an 850-nm thick Sc0.43Al0.57N thin film and 300-nm wide Au IDT (λ = 1.2 µm). In this case, the Au thickness of the filters IDT were different: 130 nm and 65 nm thick for the PCD and SCD, respectively. A 5-nm thick Cr layer was employed as an adhesion layer. In all devices, a 0.5 metallization ratio was employed (Figure 1). An SEM inspection of the fabricated SAW filters can be found in (Figure S1) The 1-port resonators had 60 finger pairs and two grounded reflectors with 60 finger pairs with the designed IDT wavelength. On the other hand, the π-type SAW filters comprised 3 1-port resonators.

**Figure 1.** Sc0.43Al0.57N heterostructures of the fabricated devices with polycrystalline and single crystal diamond substrates. (**A**) The 1-port SAW resonators and (**B**) SAW filters.

The electrical characterization of the devices was carried out using a vector network analyzer (Agilent N5230 A, Santa Clara, CA, USA), with standard, 300-µm pitch, groundsource-ground (GSG) probes (Picoprobe 40A; C style adaptor, GGB Industries, Inc, Naples, FL, USA). A standard short, open, load, through (SOLT), 50 Ω, one-port calibration was employed. In the case of the 1-port resonators, the measurement resolution was set to 16,001 points in the 0.50 GHz to 6.00 GHz frequency range, whereas the measurement of the fabricated SAW filters was carried out with a resolution of 16,001 points in the 1.00 GHz to 10 GHz frequency range. In both cases, the output power was set to 0 dBm. Afterwards, the data processing and analysis was carried out using scikit-rf, an open-source Python package [16].

### **3. Results**

The devices presented in this work targeted a ratio between piezoelectric thin film thickness and an IDT wavelength (d/λ) of 0.71, which is reported to provide the maximum *K* 2 *e f f* of the Sezawa mode for diamond-based heterostructures [17]. Additionally, in order to efficiently propagate the Sezawa mode at frequencies above 4 GHz, the designed IDT wavelength of the filters (1200 nm) was reduced compared to that of the resonators (2800 nm). In these devices, the resonance frequency was determined by the IDT wavelength and the propagation velocity of the wave through the layered structure [18]. Therefore, reducing

the thickness of the piezoelectric thin film forces the wave to propagate predominantly through the diamond layer, leveraging its faster propagation velocity.

The single crystal diamond datasheet reported a surface roughness *R<sup>a</sup>* value below 2 nm and the AFM analyses carried out on the polycrystalline diamond substrate reported *RRMS* values below 2 nm. Regardless of the substrate, the full width at half maximum (FWHM) below 3◦ of the XRD 0002 ω-θ scans confirmed that the polycrystalline ScAlN thin films are highly c-axis oriented. The RBS analysis reported an Sc 43% concentration within the thin film (Figure S2).
