Robust room-temperature ferroelectricity in the wide-bandgap semiconductor Ga₂O₃

Abstract

For decades, the integration of power handling and nonvolatile memory has been fundamentally impeded by the incompatibility between wide-bandgap semiconductors and ferroelectric materials. We resolve this challenge by demonstrating robust room-temperature ferroelectricity in epitaxial metastable κ-Ga₂O₃, grown via industry-compatible metal-organic chemical vapor deposition, creating an intrinsically ferroelectric wide-bandgap semiconductor. Through systematic characterization including piezoresponse force microscopy, polarization hysteresis measurements, and positive up–negative down tests, we provide conclusive evidence of stable ferroelectric switching down to 5-nanometer thickness—exceeding conventional ferroelectric limits—via a unique octahedral-tetrahedral transformation. Ferroelectric tunnel junctions achieve giant tunneling electroresistance exceeding 10⁵. This fundamental discovery in a mainstream semiconductor challenges conventional materials paradigms and enables monolithic integration of power and memory functionalities on a unified platform.

A mainstream wide-bandgap semiconductor achieves ferroelectricity, uniting power handling with memory functionality.

INTRODUCTION

The coexistence of wide-bandgap (WBG) character and ferroelectric behavior in a single material has long been considered fundamentally incompatible—a materials science paradox that has remained a challenge in the field (1–4). This incompatibility stems from opposing structural requirements: WBG semiconductors demand rigid crystal structures for electrical stability, while ferroelectrics require flexible ionic arrangements for polarization switching. Resolving this fundamental conflict would not only advance our understanding of materials physics but also unlock unprecedented multifunctional capabilities. WBG semiconductors, such as silicon carbide, gallium nitride, and gallium oxide (Ga₂O₃), have revolutionized high-power and high-frequency electronics (5–8) through exceptional breakdown fields (9) and thermal stability (10). However, their rigid crystal structures, optimized for electrical performance, inherently suppress the atomic displacements needed for memory storage. The same covalent networks that enable operation at kilovolt levels prevent the ionic switching essential for information storage.

Conversely, conventional ferroelectric materials enable robust polarization switching but suffer from narrow bandgaps, chemical instability, complimentary metal-oxide semiconductor (CMOS) incompatibility, and fundamental scaling limitations. Perovskite ferroelectrics face critical thickness limits below 10 nm due to depolarization fields (11–13), while emerging two-dimensional (2D) ferroelectrics, despite impressive scalability (14), have bandgaps below 3 eV that result in prohibitive leakage currents for high-field applications (15, 16). Even CMOS-compatible HfO₂ requires dopant stabilization and remains limited to ultrathin films (17, 18). This fundamental materials incompatibility has represented a major unsolved challenge in condensed matter physics, limiting our ability to design multifunctional materials that transcend traditional property boundaries.

Ga₂O₃ stands out for its rich polymorphism, with five distinct phases (α, β, γ, δ, and κ) offering a versatile platform to tune structural and electronic properties (19, 20). The metastable orthorhombic κ phase (space group Pna2₁), in particular, has emerged as a tantalizing candidate for ferroelectricity owing to its noncentrosymmetric lattice and low-energy polarization reversal barriers (Fig. 1A) (20–22). This structure features Ga atoms in distinctive tetrahedral (GaO₄) and octahedral (GaO₆) coordination sites that facilitate switchable spontaneous polarization. Yet, experimental verification has remained elusive (23, 24), hampered by the dual challenges of stabilizing phase-pure κ-Ga₂O₃ films and distinguishing intrinsic switching from measurement artifacts.

Fig. 1. Epitaxial stabilization and structural verification of ferroelectric κ-Ga₂O₃.

(A) Crystal structure of κ-Ga₂O₃ showing switchable polarization along the [001] direction. Ga atoms occupy tetrahedral (GaO₄ and Ga_tet) and octahedral (GaO₆ and Ga_oct) coordination sites that enable polarization switching. (B) High-resolution XRD θ-2θ scan of 5-nm κ-Ga₂O₃ film on NSTO (111) substrate. Exclusive (00l) reflections confirm phase-pure epitaxial growth. (C) Cross-sectional STEM image revealing atomically abrupt κ-Ga₂O₃/NSTO interface with no secondary phases. (D) SHG polar plots demonstrating noncentrosymmetric crystal symmetry. Experimental data (dots) and theoretical fits (lines) confirm mm2 point group symmetry consistent with ferroelectric behavior. a.u., arbitrary unit.

Here, we demonstrate robust room-temperature ferroelectricity in epitaxial κ-Ga₂O₃ thin films grown by metal-organic chemical vapor deposition (MOCVD), establishing a mainstream semiconductor with intrinsic ferroelectric properties. Through comprehensive characterization including piezoresponse force microscopy (PFM), polarization-resolved second-harmonic generation (SHG), and polarization hysteresis measurements, we confirm switchable polarization in films as thin as 5 nm. Density functional theory (DFT) reveals the mechanism: Concerted Ga-O displacements enable complete polarization reversal through octahedral-to-tetrahedral–to–inverted octahedral transformations without bond breaking. The use of industry-standard MOCVD growth demonstrates the reproducibility and scalability of this phenomenon. This discovery fundamentally challenges the long-held assumption that WBG character and ferroelectricity are mutually exclusive, establishing paradigms for multifunctional material design and opening unexplored territory at the intersection of semiconductor physics and ferroelectric phenomena.

RESULTS

Epitaxial stabilization of ferroelectric κ-Ga₂O₃

We achieved epitaxial stabilization of metastable κ-Ga₂O₃ on Nb-doped SrTiO₃ (NSTO) substrates using MOCVD, enabling systematic investigation of intrinsic ferroelectricity across thickness regimes from 151 nm down to 5 nm. High-resolution X-ray diffraction (XRD) confirms phase-pure κ-Ga₂O₃ films with exclusive (00l) reflections, indicating strong c-axis texture (Fig. 1B and fig. S1). Pole figure and phi-scan analyses confirm epitaxial alignment: κ-Ga₂O₃ (001) || NSTO (111) with in-plane orientation κ-Ga₂O₃ [100] || NSTO [-1 -1 2] (fig. S2, A to C). The rocking curve full width at half maximum of 0.108° demonstrates exceptional crystalline quality—superior to previously reported Ga₂O₃ epitaxial films (fig. S2D) (25). Cross-sectional scanning transmission electron microscopy (STEM) reveals atomically abrupt interfaces with no detectable secondary phases or structural defects (Fig. 1C), confirming the metastable phase stabilization essential for intrinsic ferroelectric behavior.

SHG measurements provide unambiguous verification of the noncentrosymmetric crystal structure (Fig. 1D). The pronounced SHG response in both s- and p-polarized configurations, with intensity modulation following the expected mm2 point group symmetry, confirms the polar nature of κ-Ga₂O₃ and rules out centrosymmetric phases that could masquerade as ferroelectric through extrinsic mechanisms (note S1). Thickness-dependent SHG measurements on κ-Ga₂O₃ films (fig. S3) further corroborate the intrinsic origin of the nonlinear optical response, demonstrating that the polar structure is a bulk property rather than a surface or interface effect.

Nanoscale ferroelectric domains and switching dynamics

PFM reveals robust ferroelectric behavior persisting down to the 5-nm thickness limit—a remarkable achievement that surpasses the scaling constraints of conventional perovskite ferroelectrics. Local PFM measurements demonstrate characteristic ferroelectric signatures: Hysteretic amplitude loops with clear coercive voltages, 180° phase reversals upon polarization switching, and stable domain retention (Fig. 2, A to D, and fig. S4). The domain switching exhibits exceptional temporal stability, with written domains remaining stable for over 4 hours without measurable decay (Fig. 2D and fig. S5). The retention measurement was initiated immediately after domain writing, with the initial state (t = 0 min) corresponding to the phase image in Fig. 2B. Quantitative analysis of the retention kinetics reveals a power-law decay P(t) ∝ t^-α with α = 0.018, dramatically outperforming ultrathin perovskite ferroelectrics (α = 0.14 for 6-nm thickness BaTiO₃) (26). This superior retention stems from the unique octahedral-tetrahedral transformation mechanism that minimizes depolarization fields and interface-induced degradation. Local off-field PFM hysteresis measurements more than 50 cycles reveal characteristic 180° phase shifts and butterfly-shaped amplitude loops, providing conclusive evidence for reversible polarization switching (Fig. 2E).

Fig. 2. Nanoscale ferroelectric switching in ultrathin κ-Ga₂O₃ films.

(A and B) PFM amplitude and phase images of 5-nm κ-Ga₂O₃ on NSTO substrate after domain writing with ±10 V bias. Scale bars, 2 μm. (C) Phase profile along the line indicated in (B), showing 180° phase contrast between oppositely polarized domains. (D) Domain retention measurement showing exceptional stability over 4 hours with power-law decay exponent α = 0.018. Insets, initial and final phase images, where the initial image (t = 0 min) corresponds to (B), representing the starting point of the retention measurement, and the final image (t = 240 min) corresponds to the final time point in fig. S5. (E) Local PFM hysteresis loops (50 cycles) demonstrating reversible amplitude and 180° phase switching. (F) Frequency-dependent off-field hysteresis confirming intrinsic ferroelectric behavior distinct from electrostatic artifacts.

Systematic thickness-dependent studies reveal a pronounced size effect in coercive fields, scaling from 155.7 kV/cm at 151 nm to 2.54 × 10⁴ kV/cm at 5 nm (Fig. 2E and fig. S4, G to I). The coercive field at 5-nm thickness is nearly an order of magnitude larger than that of AlScN (27, 28) and HfZrO₂ (29), while even the 151-nm-thick film surpasses BaTiO₃ (30) and PbTiO₃ thin film (31) by a similar factor. This trend follows depolarization field theory for ultrathin ferroelectrics, confirming the intrinsic nature of the observed switching rather than artifacts from electrode interactions or substrate clamping (13). Off-field hysteresis measurements under various driving frequencies (0.1 to 1 Hz) (Fig. 2F), driving voltages (1.5 to 5 V), and switching voltages (1 to 10 V) (fig. S6) rule out possible extrinsic contributions such as electrostatic or capacitive contributions, confirming intrinsic ferroelectric behaviors (32). PFM measurements on β-Ga₂O₃ control films showed no detectable domain switching under applied electric fields (fig. S7), confirming the absence of switchable ferroelectric polarization in this thermodynamically stable polymorph and validating that the observed ferroelectric behavior is specific to the κ-phase. Notably, the PFM loops collapse above a critical testing voltage of ~5 V, consistent with the coercive voltage of the sample.

Macroscopic polarization hysteresis and endurance

Polarization-voltage (P-V) measurements across all film thicknesses exhibit well-defined ferroelectric hysteresis loops with saturated polarization switching (Fig. 3 and fig. S8). The 5-nm κ-Ga₂O₃ film achieves a remnant polarization of 3 μC/cm², comparable to early reports of conventional ferroelectrics at similar thicknesses, including 5-nm Hf_0.5Zr_0.5O₂ (10 μC/cm²) (33), 10-nm BiFeO₃ (1 μC/cm²) (34), and 5-nm PbZr_0.2Ti_0.8O₃ (5.5 μC/cm²) (35), while exceeding values reported for emerging 2D ferroelectrics such as 50-nm α-In₂Se₃ (0.92 μC/cm²) (36), twisted BN bilayer (0.68 μC/cm²) (37), and 3R MoS₂ bilayer (0.4 μC/cm²) (table S1) (38). The measured polarization is lower than the DFT-predicted value, which may be attributed to multivariant crystallographic domains, domain-wall pinning, and incomplete switching. Stable ferroelectricity spans 5 nm to hundreds of nanometers, confirming that the ferroelectric behavior originates from the intrinsic polar structure rather than ultrathin confinement; below 5 nm, increased leakage current and interfacial defects prevent reliable characterization. Fatigue testing reveals exceptional durability, with spontaneous polarization remaining stable after more than 10⁷ switching cycles, highlighting suitability for nonvolatile memory applications (Fig. 3C). This remarkable stability exceeds most ferroelectric thin films and positions κ-Ga₂O₃ as a viable candidate for nonvolatile memory applications requiring long-term reliability. Positive-up–negative-down (PUND) measurements eliminate nonferroelectric contributions and confirm intrinsic switching (Fig. 3D). The ferroelectric nature is further validated by the characteristic frequency dependence and voltage scaling of the hysteresis loop, ruling out interfacial charge injection or ionic migration mechanisms.

Fig. 3. Macroscopic polarization and endurance characteristics.

(A) P-V hysteresis loops for varying applied voltages, demonstrating saturated switching behavior. (B) Frequency-dependent P-V loops showing characteristic ferroelectric response. (C) Exceptional fatigue resistance with >95% polarization retention after 10⁷ switching cycles at ±5 V, 100 kHz. (D) PUND measurement of 5-nm film confirming intrinsic switching with remnant polarization P_r = 3 μC/cm², comparable to state-of-the-art ultrathin ferroelectrics.

Atomic-scale mechanism of polarization switching

First-principles DFT calculations elucidate the microscopic origin of ferroelectricity in κ-Ga₂O₃. Starting from the paraelectric Pbcn phase (39), Berry phase analysis (40) predicts a spontaneous polarization of 24 μC/cm² along the [001] direction, arising from cooperative displacements of Ga and O atoms from their centrosymmetric positions (Fig. 4A) (41). The calculated polarization is much larger than the measured value, which is attributed to the 120° domain-wall network in the sample that topologically pins the propagation of polarization domain walls, hindering complete domain reversal and thereby reducing the remanent polarization (42). In the following, we focus only on the atomic-scale mechanism of polarization switching. Schematic diagrams of the ferroelectric phase and paraelectric phase structures are shown in fig. S9. The atomic positions of the ferroelectric and paraelectric phases are listed in table S2, and the corresponding atomic displacements between the two phases are presented in table S3. The calculated switching barrier of 26.4 meV/atom is lower than perovskite ferroelectrics (47.2 meV/atom for PbTiO₃) (43), enabling facile room-temperature switching without approaching dielectric breakdown limits. The polarization reversal mechanism involves a remarkable structural transformation where GaO₄ tetrahedra convert to distorted GaO₆ octahedra through concerted atomic displacements, followed by reformation with inverted polarity (movie S1). This pathway preserves local coordination environments and minimizes bond-breaking, contrasting sharply with conventional ferroelectrics that rely on displacive ionic hopping.

Fig. 4. Atomic-scale mechanism of ferroelectric switching.

(A) DFT calculation of polarization switching pathway showing energy barrier of 26.4 meV/atom. PE, paraelectric; FE, ferroelectric. (B) IDPC-STEM image along [010] zone axis. Simulated structure (blue overlay) matches experimental data, confirming Pna2₁ polar symmetry. Scale bar, 1 nm. (C) Schematic of cooperative octahedral-tetrahedral transformation during polarization reversal, as animated in movie S1. Ga_oct, octahedral Ga (GaO₆); Ga_tet, tetrahedral Ga (GaO₄). (D and E) Enlarged views showing structural distortions in first Ga layer during FE-PE phase transition. The region shown in (D) is identical to that in fig. S14A, where (D) highlights structural distortions while fig. S14 presents quantitative displacement analysis. Scale bar, 2 nm.

Atomic-resolution integrated differential phase contrast STEM (iDPC-STEM) directly visualizes the noncentrosymmetric Pna2₁ lattice polar atomic structure (Fig. 4B) (44). Cross-sectional high-resolution STEM imaging was carried out to compare the atomic structure at the near-surface region and the film interior. In both regions, the atomic structure is fully consistent with the polar Pna2₁ space group of κ-Ga₂O₃, revealing no structural reconstruction at the film surface (fig. S10). These results support the conclusion that the ferroelectricity in our κ-Ga₂O₃ films is an intrinsic property of the polar crystal structure rather than an artifact of surface reconstruction. This ionic polarization, quantified as ~22.6 eÅ for each O and Ga atom, resulting in a total polarization matching the theoretical predictions after removing the DFT-calculated electronic polarization. The experimental images match theoretical predictions with quantitative precision, confirming oxygen the mean absolute displacements of ~1.4 Å relative to O positions in paraelectric phase. Combined high-angle annular dark-field (HAADF)–STEM/iDPC-STEM analysis reveals total Ga/O displacement of ~39.5 Å between ferroelectric and paraelectric phases, in excellent agreement with DFT calculations (table S2 and figs. S11 to S14).

In this context, κ-Ga₂O₃ was modeled as a structure consisting of four GaO layers. Structural modeling (Fig. 4C) identifies polarization origins in polar GaO₄ tetrahedra and distorted GaO₆ octahedra within the first and third GaO layers. Our analysis revealed that the polarization primarily originates from the polar GaO₄ tetrahedra and distorted GaO₆ octahedra in the first and third layers, while the adjacent distorted GaO₆ octahedra in the second and fourth layers do not contribute to net polarization. During the phase transition, the polar GaO₄ tetrahedra in the first and third layers transform into distorted GaO₆ octahedra. These four GaO₆ octahedra in the two layers completely cancel out their polarization effects. Subsequently, two GaO₆ octahedra revert to polar GaO₄ tetrahedra, resulting in the complete switching of the ferroelectric polarization (Fig. 4, D and E). These multiscale insights resolve long-standing ambiguities in prior studies, establishing κ-Ga₂O₃ as an intrinsic WBG semiconductor with room-temperature ferroelectricity. Despite its metastable nature relative to β-Ga₂O₃, κ-phase films stored under ambient conditions for 3 years show no detectable phase transformation and retain robust ferroelectric switching, demonstrating the kinetic stability necessary for device applications.

FTJ performance

To explore the fundamental physics of polarization-controlled transport, we fabricated Au/κ-Ga₂O₃/NSTO ferroelectric tunnel junctions (FTJs) that exploit polarization-dependent band alignment for resistive switching. Kelvin probe force microscopy measurements on 5-nm κ-Ga₂O₃ films confirm ferroelectric domain switching and reveal the correlation between local surface potential and polarization state (fig. S15) (45). The optimized 5-nm barrier thickness balances the trade-off between tunneling probability and polarization-induced barrier modulation. Conductive atomic force microscopy measurements on Pt-tip/5-nm κ-Ga₂O₃/NSTO devices reveal two distinct resistance states switchable by ±8 V bias (Fig. 5A), while Au/κ-Ga₂O₃/NSTO vertical structures measured under vacuum at room temperature achieve a colossal tunneling electroresistance (TER) ratio exceeding 10⁵ (Fig. 5B and fig. S16). This TER value surpasses HfO₂-based FTJs by three orders of magnitude and rivals the best perovskite oxide heterostructures, while maintaining compatibility with semiconductor processing (46). Thickness-dependent measurements demonstrate that the switching ratio decreases by four orders of magnitude in thicker 13-nm samples (fig. S17). The gradual resistance transition excludes filamentary conduction mechanisms, confirming that switching arises from polarization-controlled tunneling barriers rather than defect migration.

Fig. 5. High-performance FTJs.

(A) Current-voltage characteristics in Pt/κ-Ga₂O₃(5-nm)/NSTO device. Insets, device schematic and band alignment diagrams for high- resistance state (HRS) and low-resistance state (LRS). E_F, Fermi level. (B) Reproducible resistive switching under programming voltages of ±6 V demonstrating nonvolatile memory operation with a giant TER ratio. (C) Switching cycling stability with <5% resistance degradation over 150 cycles. Inset, programming pulse sequence. (D) Retention measurement showing stable high and low resistance states over >250 s with projected lifetime exceeding years. Insets, read pulse sequences for both resistance states.

Critically, the FTJ devices exhibit exceptional operational stability. The TER ratio remains constant more than 150 switching cycles with <5% degradation (Fig. 5C), while retention measurements demonstrate stable high- and low-resistance states persisting for >250 s with projected stability exceeding years (Fig. 5D). Current-voltage characteristics obtained with ±8 V switching voltages (fig. S16) confirm reproducible, symmetric switching behavior without evidence of voltage-dependent degradation or asymmetric imprint effects. The absence of imprint effects—common failure modes in perovskite ferroelectrics—highlights the superior interfacial stability of κ-Ga₂O₃/NSTO heterostructures. The ultralow switching energy (<3 pJ per switch) combined with high breakdown resilience (>8 MV/cm) positions κ-Ga₂O₃ FTJs as transformative components for neuromorphic computing and high-power electronics integration. This unique combination of ferroelectric and WBG properties in a single material fundamentally challenges our understanding of structure-property relationships in semiconductors.

DISCUSSION

We have demonstrated that metastable κ-Ga₂O₃ unites robust ferroelectricity with WBG semiconductor properties, resolving an important materials incompatibility that has constrained electronic system design. The discovery of intrinsic ferroelectric behavior in this mainstream semiconductor—with polarization switching stable down to 5-nm thickness and exceptional endurance exceeding 10⁷ cycles—establishes a material paradigm that transcends the traditional boundaries between power electronics and memory technologies. The atomic-scale mechanism reveals a unique structural pathway: Polarization reversal occurs through concerted octahedral-tetrahedral transformations that preserve local bonding environments while enabling complete dipole inversion. This low-barrier switching mechanism (26.4 meV per atom) enables room-temperature operation without approaching breakdown limits, contrasting sharply with conventional ferroelectrics that rely on energy-intensive ionic displacements. The combination of theoretical predictions and atomic-resolution imaging provides unprecedented mechanistic insight into ferroelectric switching in WBG materials.

The technological implications are readily apparent. FTJs exhibit giant TER ratios exceeding 10⁵ while retaining compatibility with standard semiconductor processing. These performance metrics are comparable to those of specialized perovskite oxides, yet the junctions leverage the manufacturability inherent to silicon-based technologies. This convergence facilitates monolithic integration of nonvolatile memory, reconfigurable components, and high-power switching within unified device architectures—addressing critical bottlenecks in energy-efficient computing and power management systems. This property convergence enables monolithic power-memory integration, reconfigurable radio-frequency components, and neuromorphic computing—functionalities inaccessible when ferroelectric and WBG properties reside in separate materials.

Looking forward, the identification of metastable phase engineering as a route to ferroelectric semiconductors opens unexplored materials space beyond the Ga₂O₃ system. The success with industry-standard MOCVD growth processes enables large-scale manufacturing, positions these findings for rapid technological translation, while the fundamental insights into octahedral coordination switching provide design principles for finding additional ferroelectric WBG semiconductors. The broader impact extends beyond individual device performance to system-level transformations in electronics architecture. By demonstrating ferroelectricity in a prototypical WBG semiconductor, this work opens avenues for finding multifunctional materials that transcend traditional property boundaries. This work demonstrates how venturing beyond thermodynamic equilibrium can unlock materials functionalities that seemed fundamentally incompatible, suggesting rich opportunities for finding transformative properties in other metastable systems.

MATERIALS AND METHODS

Epitaxial film growth

Epitaxial κ-Ga₂O₃ thin films were grown on (111)–oriented NSTO substrates via MOCVD. NSTO substrates (10 mm by 10 mm by 0.5 mm) were sequentially cleaned in acetone, deionized water, and ethanol for 10 min each, and then dried under high-purity argon. Triethylgallium (TEGa) and O₂ served as precursors with high-purity Ar as carrier gas. Phase-pure κ-Ga₂O₃ requires a narrow growth window [430° to 480°C, TEGa 300 to 800 standard cubic centimeter per minute (SCCM)], reflecting its metastable nature. Optimized conditions were 450°C, 25 torr, O₂ 6000 SCCM, and TEGa 350 SCCM. Film thickness was precisely controlled through deposition time, yielding samples ranging from 5 to 151 nm with >97% thickness uniformity, consistent phase purity, and reproducible ferroelectric response across the 1 cm–by–1 cm growth area. For phase comparison studies, β-Ga₂O₃ control films were grown using identical precursors and pressure conditions but at elevated temperature based on the thermodynamic phase transition relationships of Ga₂O₃ polymorphs (750°C). Phase composition and crystalline orientation were analyzed by XRD, confirming selective formation of β- and κ-Ga₂O₃ phases depending on growth temperature (fig. S18A). κ-Ga₂O₃ films were epitaxially grown on multiple types of substrates, all displaying sharp diffraction peaks characteristic of κ-Ga₂O₃ (001) (fig. S18B), which underscores the material’s excellent crystallinity and highlights its broad substrate compatibility for scalable integration. Among these, NSTO provides both lattice matching and a conductive electrode for subsequent electrical measurements; while the ferroelectricity is intrinsic to κ-Ga₂O₃ regardless of the growth substrate. Optical transmission measurements revealed a direct bandgap of ~4.8 eV for κ-Ga₂O₃ film (fig. S19), confirming its WBG semiconductor nature.

Structural characterization

Phase identification and epitaxial relationships were determined using high-resolution XRD (Rigaku-TTR/S2, Cu Kα radiation). Crystalline quality was assessed through ω-scan rocking curves, while in-plane alignment was confirmed via pole figures and φ-scans. Cross-sectional specimens for electron microscopy were prepared by focused ion beam milling (Helios 5CX, Thermo Fisher Scientific). Atomic-resolution imaging used STEM (Spectra-300, Thermo Fisher Scientific) at 300 kV, using HAADF and iDPC modes (47). Image acquisition parameters: 512 by 512 pixels, 0.015- to 0.019-nm/pixel resolution, with collection angles of 57 to 200 mrad (HAADF) and 8 to 54 mrad (iDPC).

Ferroelectricity characterization

SHG measurements used a femtosecond laser (1030 nm, 190-fs pulse width, 80 MHz repetition rate) at 45° oblique incidence to verify noncentrosymmetric crystal symmetry (fig. S20) (48). The frequency-doubled signal was analyzed using polarization-resolved detection in both s and p configurations. PFM was performed using Asylum Research MFP-3D infinity and Bruker Dimension Icon systems with Pt/Ti-coated cantilevers (AC240TM). Local switching dynamics were probed under ±10-V direct current bias to distinguish intrinsic ferroelectric behavior from electrostatic artifacts. Macroscopic P-V hysteresis loops and PUND measurements were conducted using a dedicated ferroelectric analyzer (TF-Analyzer 1000, aixACCT) to quantify remnant polarization and eliminate nonferroelectric contributions.

Device fabrication and electrical measurements

FTJs used 5-nm κ-Ga₂O₃ films with NSTO bottom electrodes and circular Au (50 nm) top contacts (10-μm diameter) defined by photolithography and electron-beam evaporation. Current-voltage characteristics and TER measurements were performed at room temperature using a semiconductor parameter analyzer (Keithley 4200A-SCS) in a vacuum probe station environment. Conductive atomic force microscopy provided nanoscale current mapping under controlled atmospheric conditions using an Asylum MFP-3D Infinity system.

First-principles calculations

DFT calculations were performed using the Vienna ab initio simulation package (VASP) (49–51) with the PBEsol exchange-correlation functional (52) and projector-augmented wave pseudopotentials. Plane-wave basis sets with 550 eV energy cutoff and Brillouin zone sampling (KSPACING = 0.15 Å⁻¹) ensured computational convergence. Structural optimizations achieved energy and force convergence criteria of 10⁻⁸ eV and 10⁻⁵ eV/Å, respectively. Spontaneous polarization was calculated using the modern theory of polarization via Berry phases, while structural visualizations used VESTA (53) and OVITO [version 3.11.3, OVITO GmbH, Germany (2024)] software packages.

Supplementary Materials

The PDF file includes:

Supplementary Text

Figs. S1 to S20

Tables S1 to S3

Legend for movie S1

References

Other Supplementary Material for this manuscript includes the following:

Movie S1