AlScN-on-Sapphire Device Platform

• AlScN-on-sapphire is a III-nitride platform that integrates scandium-alloyed aluminum nitride with a GaN intermediary on c-axis sapphire to achieve high-quality ferroelectric and optical properties.
• It employs lithographically defined periodic poling for quasi-phase matching, enabling second-harmonic generation in the deep UV and visible spectra.
• The platform supports multifunctional applications including low-loss photonic routing, high-frequency acoustic delay lines, and GaN-based ferroelectric electronics.

The AlScN-on-sapphire platform is a III-nitride device platform in which scandium-alloyed aluminum nitride is integrated on sapphire, typically through a GaN intermediary, to combine c-axis-oriented wurtzite crystallinity, switchable out-of-plane polarization, sapphire’s wide-bandgap optical substrate properties, and compatibility with ferroelectric, photonic, acoustic, and GaN-electronic processing. In the canonical epitaxial demonstration, the stack is an Al-polar $\mathrm{Sc}_{0.25}\mathrm{Al}_{0.75}\mathrm{N}$ film of thickness $100\,\mathrm{nm}$ on highly Si-doped GaN on c-axis sapphire, and it supports robust ferroelectric switching together with lithographically defined periodic poling down to $0.4\,\mu\mathrm{m}$ (Yang et al., 2023). Related work broadens the platform concept toward sapphire-supported low-loss photonics (Wang et al., 2024, Zhang et al., 28 Mar 2025), multi-GHz AlScN acoustic delay lines (Chang et al., 15 Sep 2025), monolithically integrated GaN systems employing AlScN-based ferroelectric gate stacks (Joo et al., 8 Jul 2026), and ultrathin all-epitaxial AlScN/Pt heterostructures on GaN/sapphire that remain ferroelectric at $10\,\mathrm{nm}$ thickness (Schönweger et al., 2022).

1. Heterostructure and crystallographic basis

In the epitaxial ferroelectric implementation, the platform consists of c-axis $(0001)$ sapphire, a highly Si-doped GaN buffer, and an Al-polar $\mathrm{Sc}_{0.25}\mathrm{Al}_{0.75}\mathrm{N}$ functional layer with thickness $d=100\,\mathrm{nm}$, grown by molecular beam epitaxy on GaN templates. The stack is written explicitly as

$$\text{Al-polar Sc}_{0.25}\text{Al}_{0.75}\text{N} \,(100\,\text{nm})/\text{Si-doped GaN}/c\text{-sapphire}.$$

At $25\%$ Sc, ScAlN is closer to lattice-match with GaN and exhibits reduced dislocation density. The GaN layer supplies the wurtzite symmetry and closely matched lattice parameters needed for single-crystal ScAlN, provides polarity control between Al-polar and N-polar orientations, and serves as an integrated bottom electrode through Si doping (Yang et al., 2023).

The sapphire substrate contributes a robust, low-loss, wide-bandgap platform that is transparent from UV to infrared, thermally and mechanically stable under high-field ferroelectric operation, and naturally aligned for c-axis-oriented GaN and ScAlN growth. That c-axis orientation is crucial for maximizing out-of-plane polarization, piezoelectric, and $\chi^{(2)}$ coefficients such as $100\,\mathrm{nm}$0 and $100\,\mathrm{nm}$1. Before poling, the ScAlN film is explicitly Al-polar, meaning that the spontaneous polarization points toward the Al-terminated surface; poling reverses selected regions to N-polar domains and thereby creates a controllable epitaxial up/down polarization structure normal to the sapphire surface (Yang et al., 2023).

A second, all-epitaxial variant uses commercially available GaN/sapphire substrates together with epitaxial Pt electrodes and ultrathin $100\,\mathrm{nm}$2 films. In that architecture, the full capacitor stack is

$100\,\mathrm{nm}$3

and structural characterization by X-ray diffraction and transmission electron microscopy reveals established epitaxy and high-quality interfaces (Schönweger et al., 2022). A related GaN-transistor variant employs a recessed-gate $100\,\mathrm{nm}$4/Pt/HfO$100\,\mathrm{nm}$5/AlGaN/GaN MFMIS stack that is directly compatible with commercial GaN-on-sapphire processing, even though the specific substrate is not stated in the main text (Joo et al., 8 Jul 2026).

2. Ferroelectric switching, retention, and thickness scaling

Ferroelectric switching in epitaxial ScAlN-on-sapphire is characterized macroscopically by a Positive-Up-Negative-Down protocol using patterned nickel top electrodes, conductive Si-doped GaN as the bottom electrode, triangular voltage pulses of peak $100\,\mathrm{nm}$6 and pulse length $100\,\mathrm{nm}$7, and the field estimate

$100\,\mathrm{nm}$8

which gives $100\,\mathrm{nm}$9 for $0.4\,\mu\mathrm{m}$0. Current density is written as $0.4\,\mu\mathrm{m}$1, and polarization is extracted as

$0.4\,\mu\mathrm{m}$2

with $0.4\,\mu\mathrm{m}$3 obtained by subtracting non-switching currents from the PUND sequence. The specific subtraction used is “U” from “P” for forward switching and “D” from “N” for backward switching, yielding the ferroelectric switching currents $0.4\,\mu\mathrm{m}$4 and $0.4\,\mu\mathrm{m}$5 (Yang et al., 2023).

The key ferroelectric metrics reported for the epitaxial sapphire/GaN platform are a coercive field of approximately $0.4\,\mu\mathrm{m}$6 at $0.4\,\mu\mathrm{m}$7 thickness, a maximum backward switched polarization of approximately $0.4\,\mu\mathrm{m}$8, and an estimated remanent polarization of approximately $0.4\,\mu\mathrm{m}$9. The work therefore confirms clear hysteretic switching with large $10\,\mathrm{nm}$0, a coercive field below dielectric breakdown, and strong out-of-plane polarization on sapphire/GaN (Yang et al., 2023).

Domain reversal is visualized directly by piezoresponse force microscopy and by SEM after selective HCl etching. After poling with trapezoidal pulses of $10\,\mathrm{nm}$1 and $10\,\mathrm{nm}$2, PFM phase imaging shows clear contrast between Al-polar and N-polar regions in a $10\,\mathrm{nm}$3 poled square, and the reversed “YALE” pattern remains sharply defined after selective etching because N-polar domains are more susceptible to HCl. Retention is assessed by reapplying the same poling pulse after different time delays; the current response is unchanged even after a week, indicating negligible polarization loss (Yang et al., 2023).

The broader AlScN literature places these results within an unusually high-polarization, high-coercive-field ferroelectric regime. Ferroelectric AlScN is reported to combine remnant polarizations of about $10\,\mathrm{nm}$4–$10\,\mathrm{nm}$5 with coercive fields of about $10\,\mathrm{nm}$6–$10\,\mathrm{nm}$7, and ferroelectricity is observed in films with thicknesses below $10\,\mathrm{nm}$8 while the coercive field remains largely independent of thickness between $10\,\mathrm{nm}$9 and $(0001)$0 (Fichtner et al., 2020). In the all-epitaxial GaN/sapphire implementation, the dependence of relative permittivity and coercive field on $(0001)$1 thickness from $(0001)$2 down to $(0001)$3 shows only moderate scaling effects, suggesting that the critical thickness for ferroelectricity is not yet approached (Schönweger et al., 2022). This suggests that the sapphire platform is not restricted to the $(0001)$4 regime used for domain engineering.

3. Periodic poling and domain-engineered nonlinear optics

The central platform advance is lithographically defined periodic domain inversion in epitaxial ScAlN on sapphire. The demonstrated poling sequence is: define nickel electrode patterns by standard photolithography, deposit Ni by evaporation and liftoff, apply trapezoid poling pulses with peak voltage $(0001)$5 and pulse width $(0001)$6, remove the Ni electrodes in HCl, and optionally etch the ScAlN in HCl to expose N-polar domains. Because the Si-doped GaN layer functions as a continuous counter-electrode, the system operates as a vertical capacitor with the form Ni/ScAlN/GaN/sapphire and a well-controlled field distribution (Yang et al., 2023).

Periodic stripe domains are demonstrated for the designed periods

$(0001)$7

All patterns are successfully poled, and SEM after Ni removal and selective HCl etching shows stripe domains with periods matching the lithographic design. The designed electrode duty cycle is $(0001)$8 for all cases, but the measured ferroelectric duty cycles deviate systematically as the period shrinks (Yang et al., 2023).

Period $(0001)$9	Measured ferroelectric duty cycle
$\mathrm{Sc}_{0.25}\mathrm{Al}_{0.75}\mathrm{N}$0	$\mathrm{Sc}_{0.25}\mathrm{Al}_{0.75}\mathrm{N}$1
$\mathrm{Sc}_{0.25}\mathrm{Al}_{0.75}\mathrm{N}$2	$\mathrm{Sc}_{0.25}\mathrm{Al}_{0.75}\mathrm{N}$3
$\mathrm{Sc}_{0.25}\mathrm{Al}_{0.75}\mathrm{N}$4	$\mathrm{Sc}_{0.25}\mathrm{Al}_{0.75}\mathrm{N}$5
$\mathrm{Sc}_{0.25}\mathrm{Al}_{0.75}\mathrm{N}$6	$\mathrm{Sc}_{0.25}\mathrm{Al}_{0.75}\mathrm{N}$7
$\mathrm{Sc}_{0.25}\mathrm{Al}_{0.75}\mathrm{N}$8	$\mathrm{Sc}_{0.25}\mathrm{Al}_{0.75}\mathrm{N}$9

The increase in duty cycle at small periods is attributed to lateral domain spreading and field fringing. As electrode width and period shrink, the electric field penetrates laterally and domains expand under the gaps. The minimum demonstrated period, $d=100\,\mathrm{nm}$0, is already in the submicron regime comparable to state-of-the-art periodically poled LiNbO$d=100\,\mathrm{nm}$1 thin films. The paper explicitly states that periodic poling with arbitrary periods, including sub-micron, is achievable “in principle” across ScAlN’s transparency window by lithographic patterning and appropriate pulse engineering (Yang et al., 2023).

The nonlinear-optical significance of this capability follows from quasi-phase matching. For second-order processes such as SHG, the QPM period is

$d=100\,\mathrm{nm}$2

Because strong dispersion in the UV and visible makes $d=100\,\mathrm{nm}$3 large, the required $d=100\,\mathrm{nm}$4 can fall well below $d=100\,\mathrm{nm}$5, so the demonstrated $d=100\,\mathrm{nm}$6 regime is described as a prerequisite for SHG in deep visible and UV. The same work notes, citing Yoshioka et al., that Sc alloying significantly enhances second-order susceptibility: for $d=100\,\mathrm{nm}$7, $d=100\,\mathrm{nm}$8 is approximately twice that of LiNbO$d=100\,\mathrm{nm}$9, while for $$\text{Al-polar Sc}_{0.25}\text{Al}_{0.75}\text{N} \,(100\,\text{nm})/\text{Si-doped GaN}/c\text{-sapphire}.$$0, $$\text{Al-polar Sc}_{0.25}\text{Al}_{0.75}\text{N} \,(100\,\text{nm})/\text{Si-doped GaN}/c\text{-sapphire}.$$1, roughly $$\text{Al-polar Sc}_{0.25}\text{Al}_{0.75}\text{N} \,(100\,\text{nm})/\text{Si-doped GaN}/c\text{-sapphire}.$$2 intrinsic AlN. On that basis, the platform is positioned for SHG, sum-frequency generation, optical parametric oscillators, Pockels comb generation, cascaded $$\text{Al-polar Sc}_{0.25}\text{Al}_{0.75}\text{N} \,(100\,\text{nm})/\text{Si-doped GaN}/c\text{-sapphire}.$$3 processes, and backward-wave quasi-phase matching for mirrorless optical parametric oscillators (Yang et al., 2023).

4. Sapphire photonics architectures and low-loss routing

Direct low-loss optical waveguide benchmarks for epitaxial AlScN-on-sapphire are not yet the main result of the periodic-poling work. Instead, the existing sapphire photonics literature supplies a closely related architectural template. A sapphire-supported low-loss platform based on a “sapphire sandwich” waveguide uses bulk single-crystal sapphire as the substrate, stoichiometric LPCVD SiN as a thin core, and a bonded top sapphire chip together with index-matched SiON top cladding. In that system, microrings with $$\text{Al-polar Sc}_{0.25}\text{Al}_{0.75}\text{N} \,(100\,\text{nm})/\text{Si-doped GaN}/c\text{-sapphire}.$$4 radius reach an intrinsic quality factor of $$\text{Al-polar Sc}_{0.25}\text{Al}_{0.75}\text{N} \,(100\,\text{nm})/\text{Si-doped GaN}/c\text{-sapphire}.$$5 million near $$\text{Al-polar Sc}_{0.25}\text{Al}_{0.75}\text{N} \,(100\,\text{nm})/\text{Si-doped GaN}/c\text{-sapphire}.$$6–$$\text{Al-polar Sc}_{0.25}\text{Al}_{0.75}\text{N} \,(100\,\text{nm})/\text{Si-doped GaN}/c\text{-sapphire}.$$7, and the lowest propagation loss is $$\text{Al-polar Sc}_{0.25}\text{Al}_{0.75}\text{N} \,(100\,\text{nm})/\text{Si-doped GaN}/c\text{-sapphire}.$$8 at $$\text{Al-polar Sc}_{0.25}\text{Al}_{0.75}\text{N} \,(100\,\text{nm})/\text{Si-doped GaN}/c\text{-sapphire}.$$9. Simulations also show top-cladding overlap factors of approximately $25\%$0 for $25\%$1 SiN thickness and approximately $25\%$2 for $25\%$3 thickness (Wang et al., 2024). This suggests that sapphire can support low-loss, moderately confined visible photonics while maintaining substantial modal overlap with active claddings.

A more directly nitride-based template is the three-dimensional SiN/AlN-on-sapphire platform, in which a single-crystalline AlN layer on sapphire is combined with planarized SiO$25\%$4 and a top LPCVD SiN layer. In that platform, an AlN microring at approximately $25\%$5 reaches loaded and intrinsic quality factors of $25\%$6 and $25\%$7, corresponding to a propagation loss of approximately $25\%$8, while at approximately $25\%$9 the loaded and intrinsic quality factors are $\chi^{(2)}$0 and $\chi^{(2)}$1, corresponding to approximately $\chi^{(2)}$2. The same platform demonstrates normalized SHG efficiency up to $\chi^{(2)}$3 and on-chip SPDC photon generation efficiency of approximately $\chi^{(2)}$4, as well as near-unity vertical AlN-to-SiN coupling at $\chi^{(2)}$5 through an adiabatic vertical taper coupler (Zhang et al., 28 Mar 2025).

These results do not measure AlScN directly, but they delimit a credible sapphire-supported photonic process stack for it. The periodic-poling paper already notes sputtered ScAlN photonic circuits on silicon with $\chi^{(2)}$6 propagation loss at $\chi^{(2)}$7 and microring $\chi^{(2)}$8, and it explicitly suggests substantial room for fabrication optimization toward AlN-level losses (Yang et al., 2023). A plausible implication is that an AlScN-on-sapphire photonic platform would reuse the same sapphire/oxide/nitride integration logic, while re-optimizing thickness, etch chemistry, and dispersion for the higher-index, strongly piezoelectric alloy.

5. Acoustic and GaN-electronic manifestations of the platform

In RF acoustics, AlScN-on-sapphire is already realized as a high-frequency surface acoustic wave platform. A sputtered $\chi^{(2)}$9 $100\,\mathrm{nm}$00 film on $100\,\mathrm{nm}$01 c-plane sapphire, combined with $100\,\mathrm{nm}$02 Al electrodes, is used to implement Sezawa-mode SAW acoustic delay lines with acoustic wavelength $100\,\mathrm{nm}$03 and thickness ratio $100\,\mathrm{nm}$04. In that regime, simulations show that Sezawa-mode $100\,\mathrm{nm}$05 is maximized while Rayleigh-mode $100\,\mathrm{nm}$06 becomes very weak. After conjugate matching, the devices exhibit delay times spanning $100\,\mathrm{nm}$07 to $100\,\mathrm{nm}$08, insertion loss from $100\,\mathrm{nm}$09 to $100\,\mathrm{nm}$10, propagation loss as low as $100\,\mathrm{nm}$11 at $100\,\mathrm{nm}$12, group velocity around $100\,\mathrm{nm}$13–$100\,\mathrm{nm}$14, and an acoustic propagation $100\,\mathrm{nm}$15-factor of approximately $100\,\mathrm{nm}$16 (Chang et al., 15 Sep 2025).

The underlying coupling metric is extracted from simulated open- and short-circuit phase velocities,

$100\,\mathrm{nm}$17

and propagation quality is summarized by

$100\,\mathrm{nm}$18

with the dB-to-Neper conversion factor

$100\,\mathrm{nm}$19

The physical basis is the acoustic velocity mismatch between the slower, strongly piezoelectric AlScN layer and the stiff, high-velocity sapphire substrate, which vertically confines energy near the surface and suppresses substrate leakage (Chang et al., 15 Sep 2025). This establishes AlScN-on-sapphire not only as a ferroelectric or photonic material system, but also as a guided acoustic platform.

The platform also intersects GaN electronics through AlScN-based ferroelectric gate stacks. In recessed-gate GaN HEMTs, a $100\,\mathrm{nm}$20 $100\,\mathrm{nm}$21 ferroelectric layer, Pt intermediate electrode, and $100\,\mathrm{nm}$22 HfO$100\,\mathrm{nm}$23 insulator form an MFMIS structure whose functionality is controlled by the area ratio $100\,\mathrm{nm}$24. Systematic engineering of that ratio yields a record memory window of $100\,\mathrm{nm}$25, a forward subthreshold swing of $100\,\mathrm{nm}$26, $100\,\mathrm{nm}$27-bit multi-level-cell operation in low-ratio devices, and a first GaN-based ferroelectric frequency-to-voltage converter with a linear response across $100\,\mathrm{nm}$28–$100\,\mathrm{nm}$29 and conversion gain of $100\,\mathrm{nm}$30 (Joo et al., 8 Jul 2026). Since the work is described as directly compatible with commercial GaN-on-sapphire processes, it extends the meaning of an AlScN-on-sapphire platform from passive material integration to monolithic system functionality.

6. Limitations, misconceptions, and development trajectory

A persistent limitation in the present ferroelectric domain-engineering results is poling accuracy at small periods. When the designed electrode duty cycle is fixed at $100\,\mathrm{nm}$31, the realized ferroelectric duty cycle reaches approximately $100\,\mathrm{nm}$32 at $100\,\mathrm{nm}$33 because of lateral domain broadening, internal bias, and field fringing. The paper therefore identifies tailored pulse shaping, fine tuning of electrode geometry and spacing, and general pulse engineering as necessary for robust quasi-phase-matched operation at sub-micron scale (Yang et al., 2023).

A second limitation is that direct low-loss optical waveguide performance has not yet been demonstrated for epitaxial AlScN-on-sapphire at the level already achieved in related sapphire photonics platforms. The available low-loss records on sapphire presently come from SiN-on-sapphire and SiN/AlN-on-sapphire rather than from periodically poled AlScN itself (Wang et al., 2024, Zhang et al., 28 Mar 2025). This corrects a common misconception: the platform is established directly for ferroelectric domain control and periodic poling, while its low-loss photonic realization remains a development target rather than a completed result.

Thickness, leakage, and reliability remain coupled design variables. The periodic-poling work uses a $100\,\mathrm{nm}$34 ferroelectric film, but also notes that the relatively low coercive field of approximately $100\,\mathrm{nm}$35 makes it possible to pole thicker ScAlN films, which would be beneficial for optical confinement. At the same time, the broader AlScN literature shows that thinner films reduce operating voltage without catastrophic ferroelectric scaling, yet leakage and interface engineering become increasingly important below about $100\,\mathrm{nm}$36 (Yang et al., 2023, Schönweger et al., 2022). In the GaN-system context, high area-ratio analog devices show retention challenges linked to depolarization and possible conduction into the floating Pt electrode, indicating that interface quality and field partition remain central issues (Joo et al., 8 Jul 2026).

Taken together, these results define the AlScN-on-sapphire platform as a convergent III-nitride substrate technology rather than a single device class. It is already validated for epitaxial ferroelectric switching and sub-micron periodic poling, for high-frequency Sezawa-mode SAW propagation, and for GaN-compatible ferroelectric gate integration; related sapphire-supported photonics work provides a mature fabrication and measurement template for extending it toward low-loss $100\,\mathrm{nm}$37 photonics. This suggests that the next stage of development will be determined less by the existence of the platform than by optimization of duty-cycle control, optical loss, thicker waveguide-compatible AlScN growth, and large-area reproducibility.