Orthorhombic epsilon/kappa gallium oxide power and radiation-hardened semiconductor devices

The orthorhombic epsilon/kappa phase of gallium oxide (Ga2O3, space group Pna2_1) is a wide-bandgap semiconductor with structural properties that neither the commercially dominant beta phase nor the alpha phase can replicate: a non-centrosymmetric crystal structure that supports spontaneous electric polarization of roughly 20–25 µC/cm², a bandgap near 4.9 eV, and the potential for ferroelectric-gate device architectures that achieve normally-off transistor operation without requiring a p-type semiconductor layer. These attributes, absent in centrosymmetric alpha and beta Ga2O3, represent a genuine architectural departure in how gallium oxide transistors and Schottky barrier diodes can be designed for 600 V to 3.3 kV power conversion and radiation-hardened aerospace electronics. The timing dimension is acute. Kappa-Ga2O3 has attracted substantial academic publication activity, and a 2026 Science Advances disclosure has started a §102(b) grace-period clock that expires approximately in the first quarter of 2027. A 60-day effective filing window for priority claims exists, making this a time-sensitive decision for any acquirer or licensee. The commercial incumbents — Flosfia in alpha-Ga2O3, Mitsubishi/Novel Crystal Technology/Tamura and Kyma Technologies in beta-Ga2O3 — have built their manufacturing and IP around centrosymmetric phases. An assignee who secures well-drafted claims on phase-pure orthorhombic Pna2_1 devices, combined with the specific edge-termination and ferroelectric-gate exclusivity, occupies whitespace that incumbent patent portfolios structurally cannot cover. The asset sits within Lattice Graph's dielectric, ferroelectric and wide-bandgap oxides portfolio and is typed as a lead composition-plus-device filing. It covers three distinct device embodiments: a ferroelectric-gate normally-off MOSFET, a fluorine-edge-terminated radiation-hardened Schottky barrier diode, and the underlying phase-pure orthorhombic Ga2O3 composition claims. The filing is candidly proof-gated: dynamic stability under the specific strain conditions imposed by real epitaxial substrates has not yet cleared the computational validation bar, meaning the value here is high but contingent, and a prudent acquirer would treat this as a high-upside asset requiring a defined experimental milestone before full monetization.

The engines did not fully agree here — the asset carries that uncertainty openly rather than overstating confidence.

Gallium oxide crystallizes in multiple polymorphs, of which only the beta (monoclinic, C2/m) and — to a lesser commercial degree — alpha (rhombohedral, R-3c) phases have been reduced to manufacturable wafers. Both are centrosymmetric, meaning their crystal lattices possess inversion symmetry, which eliminates the possibility of spontaneous electric polarization. The orthorhombic epsilon/kappa phase (Pna2_1) breaks that inversion symmetry. The result is a material with a polar axis, published spontaneous polarization in the 20–25 µC/cm² range, and a wide bandgap of approximately 4.9 eV. The polarization enables a fundamentally different transistor architecture: a ferroelectric-gate field-effect transistor in which a thin-film ferroelectric gate dielectric couples to the polarization of the Ga2O3 channel to establish a normally-off threshold without any p-type doping or a p-type buried layer. This matters because Ga2O3 has no practical p-type dopant, a limitation that has constrained every beta-phase normally-off design to workarounds such as recessed gates, fin structures, or heterojunction stacks — each carrying process complexity and reliability tradeoffs. The computational validation picture is the central technical complexity of this asset. Lattice Graph ran a five-state biaxial-strain phonon ladder on May 23, 2026, sweeping the strain range that the Pna2_1 structure would experience when grown on plausible mismatched substrates. Two independent machine-learning interatomic potentials — MACE and CHGNet — were applied in consensus. Both returned imaginary phonon modes (negative frequencies, specifically –1.05 THz from MACE and –1.26 THz from CHGNet), indicating that at the strain states surveyed, the structure is dynamically unstable. The energy-above-hull from the relaxation scout converged to zero, meaning the phase is on the convex hull thermodynamically — it is not a fictitious or high-energy polymorph, but rather a real phase that is known to exist. The instability is phononic, not thermodynamic: under the specific biaxial-strain boundary conditions of these calculations, soft modes are present. An HSE06 electronic-structure calculation was dispatched but did not return results within the simulation cycle. This is a well-understood challenge in the Ga2O3 orthorhombic literature: epsilon/kappa-Ga2O3 is metastable at bulk equilibrium, and its experimental realization is achieved by strain stabilization during epitaxial growth, typically by metal-organic chemical vapor deposition on substrates such as (100)-oriented GaN or c-plane sapphire (with the key exclusion being mist-CVD on (0001) sapphire, which is explicitly carved out of claims). The instability finding is not a death sentence for the asset but it is the validation gate that governs its prosecution posture. The claims are drafted around a strain-stabilized thin-film posture precisely because the free-standing bulk phase is known to relax to beta-Ga2O3 above approximately 500–700°C. A higher-fidelity density-functional-perturbation-theory phonon calculation — one that incorporates substrate boundary conditions (epitaxial strain pinned to realistic GaN or YSZ lattice parameters) and sweeps larger strain windows beyond ±2.5% — could resolve whether stable modes emerge under experimentally relevant growth conditions. Alternatively, a physical MOCVD coupon on a (100) substrate, characterized by XRD to confirm phase purity and Raman spectroscopy to confirm phonon modes, would constitute direct experimental proof. Either route is a defined, tractable milestone rather than an open-ended scientific question. The device embodiments claimed extend the composition into three commercially specific structures. The ferroelectric-gate MOSFET uses the spontaneous polarization of the Pna2_1 phase to set threshold voltage, explicitly excluding any p-type oxide or semiconductor interlayer — a deliberate negative limitation that carves out a distinct architectural space from heterojunction normally-off designs used in GaN power devices. The fluorine-edge-terminated Schottky barrier diode is targeted at radiation-hardened applications: fluorine passivation at device edges suppresses trap-assisted leakage and improves reverse-bias reliability under ionizing-radiation environments, a known failure mode for unpassivated wide-bandgap Schottky devices in satellite and aerospace power conditioning. The combination of a 4.9 eV bandgap and the ferroelectric-gate architecture positions these devices for operation in dose environments where silicon carbide and GaN devices require additional shielding.

The addressable market for wide-bandgap power semiconductors — spanning 600 V to 3.3 kV discrete devices — is estimated at above $5 billion and growing as electrification of transportation, renewable energy conversion, and defense power systems drives demand for devices that outperform silicon insulated-gate bipolar transistors. Ga2O3 as a class competes with silicon carbide and gallium nitride on the basis of a higher theoretical breakdown field, lower substrate cost, and native oxide quality, though all beta-Ga2O3 commercial products remain pre-mass-market as of 2026. The epsilon/kappa phase does not compete with these beta-phase commercial products in the near term; rather it competes at the design-win level for next-generation device architectures where ferroelectric-gate normally-off operation or radiation hardness justifies a higher process-development investment. The radiation-hardened power electronics segment is smaller and more defensible. Aerospace prime contractors, national laboratories such as AFRL, and satellite power subsystem suppliers collectively represent a market where device performance under total ionizing dose and displacement damage is a procurement criterion rather than an afterthought. Rad-hard silicon carbide Schottky diodes command significant price premiums over commercial-grade parts, and a Ga2O3 alternative with fluorine edge termination and a wider bandgap (4.9 eV versus SiC's 3.26 eV) would theoretically offer superior radiation tolerance. A contemplated memorandum of understanding with Kyma Technologies, which already serves the rad-hard beta-Ga2O3 Schottky diode market, represents a plausible licensing or co-development entry point that does not require the assignee to build a fabrication capability independently. Royalty or licensing logic on a rad-hard device that sells at $50–200 per die would be attractive even at modest unit volumes, given the defense procurement pricing structure. The ferroelectric-gate normally-off MOSFET embodiment addresses a longer-horizon design opportunity. If strain-stabilized Pna2_1 thin-film growth can be demonstrated on large-diameter substrates (4-inch GaN templates are commercially available), the ferroelectric-gate architecture removes the single largest obstacle to Ga2O3 normally-off transistors, which is the absence of a manufacturable p-type layer. Power module designers at inverter and motor-drive manufacturers — spanning industrial automation, electric vehicle traction inverters, and wind turbine converters — constitute the eventual customer base. That market is contested by mature SiC players (Wolfspeed, Onsemi, STMicroelectronics) and GaN players (Infineon, GaN Systems, Transphorm), meaning Ga2O3 ferroelectric-gate devices would need to demonstrate reliability and cost trajectories competitive with these entrenched technologies before capturing share at scale.

spontaneous-polarization normally-off ferroelectric-gate unavailable in centrosymmetric alpha/beta incumbents

phase-pure orthorhombic Pna2_1 + edge-termination + ferroelectric-gate; alpha/beta excluded

The most natural strategic fit for this asset is a party already operating in the wide-bandgap power semiconductor space who needs a defensible IP position in ferroelectric-gate device architectures or radiation-hardened Ga2O3 diodes. Kyma Technologies is explicitly identified as a contemplated memorandum-of-understanding partner, and as a company already commercializing rad-hard beta-Ga2O3 Schottky diodes, they have both the fabrication capability to run a validating MOCVD coupon and the customer relationships (AFRL, aerospace prime contractors) to immediately value the radiation-hardened embodiment. A license or co-development agreement with Kyma would be the fastest path to experimental validation and commercial revenue simultaneously. Defense-focused compound semiconductor fabs, AFRL program offices investing in next-generation power electronics for satellite and aircraft applications, and university research groups with MOCVD access under DARPA or DoD contracts represent secondary strategic acquirers who could absorb the asset and run the validation milestone within existing programs. On the commercial power side, any party investing in next-generation normally-off wide-bandgap transistors — including GaN-ecosystem players who recognize the architectural limitation of p-GaN gate scaling and are exploring alternative normally-off mechanisms — would find the ferroelectric-gate Pna2_1 architecture strategically interesting as a hedge or defensive position. The 60-day grace-period clock makes this a time-sensitive conversation for any acquirer who wants to preserve priority options rather than simply acquiring issued claims after the window closes.

The central risk is that the computational phonon instability finding, while not surprising given the known metastability of epsilon/kappa-Ga2O3, has not yet been resolved by the higher-fidelity calculations needed to confirm strain stabilization under realistic growth conditions. If broader strain-window DFPT calculations or a physical MOCVD coupon show that phonon instability persists at all experimentally accessible strain states, the basis for the strain-stabilized thin-film claims would be materially weakened, and prosecution would need to rely heavily on the published experimental literature as prior-art support. This is a known risk with a defined de-risking path: the DFPT calculation and the MOCVD coupon are both estimable in cost and timeline (weeks to a few months), and the existing academic kappa-Ga2O3 MOCVD literature provides reasonable prior probability of success. The grace-period clock is the second major risk — if priority filing does not occur before approximately Q1 2027, the 2026 Science Advances publication becomes prior art, potentially narrowing or eliminating novelty for composition claims not anchored to pre-publication experimental disclosure. The de-risking roadmap is sequential: first, run the strain-aware DFPT phonon calculation with substrate boundary conditions (lower cost, faster turnaround); if that returns stable modes, strengthen the prosecution posture before the grace-period expires; if it does not, prioritize the physical MOCVD coupon as the definitive experimental anchor. In parallel, any acquirer should conduct a full clearance search on the 2026 Science Advances disclosure to map exactly what it does and does not cover, ensuring that the device-use claims (ferroelectric gate, fluorine edge termination) and the explicit negative limitations survive as novel over that reference.