Wide-bandgap oxide ceramic dielectric stack for advanced semiconductor packaging and radiation-hard electronics

The semiconductor packaging and aerospace electronics industries are simultaneously confronting two pressures that conventional dielectric materials handle poorly in isolation: the density demands of advanced 3D packaging (chiplets, HBM stacks, through-glass vias) that require ultra-low-leakage, high-breakdown dielectrics at tight geometries, and the radiation-tolerance requirements of space, defense, and high-energy-physics electronics that must survive cumulative ionizing dose and displacement damage without dielectric breakdown or charge trapping. The opportunity here is a family of wide-bandgap oxide ceramics — spinels, garnets, phosphates, and silicates — that satisfy both requirements with a single compositional platform. Bandgaps in the 5–8 eV range place the conduction band far above typical trap energies, simultaneously suppressing leakage in thin-film capacitor and via-barrier applications and limiting the creation of long-lived radiation-induced electron-hole pairs that degrade conventional oxides. The ability to address packaging and rad-hard aerospace from the same composition set is the core commercial proposition: a licensee in advanced packaging picks up the rad-hard positioning at no additional R&D cost, and vice versa. The timing is driven by a real substitution dynamic. HfO2 and Al2O3, the incumbent high-k dielectrics, are approaching their practical limits in leakage and equivalent-oxide-thickness scaling; both also have well-characterized radiation-induced charge-trapping problems that constrain their use in harsh environments. SiO2 remains the benchmark for rad-hardness in legacy space designs but cannot meet the dielectric-constant and physical-thickness requirements of next-generation 3D packages. This forced substitution creates a window for oxide ceramics with larger bandgaps, established structural stability, and manufacturable thin-film deposition pathways. The computational campaign described here is designed to de-risk the most expensive and time-consuming part of that substitution: identifying which compositions in the broad oxide-ceramic space are thermodynamically and dynamically stable in the phases actually needed for device integration, and computing the key dielectric properties (static dielectric constant, bandgap, phonon stability) before committing to wafer-level experiments. The portfolio containing this asset — catalysts and energy-conversion materials — is broad, but this particular filing is a genuine lead-technology asset with composition-plus-device-use claims, not a defensive or method-only placeholder. It stands on the strength of named, computationally validated compositions covering multiple crystal-structure families, with a freedom-to-operate profile that was explicitly shaped to avoid the most encumbered prior-art compositions in the oxide-dielectric literature.

Each candidate is validated by multiple independent machine-learning interatomic potentials. A material advances only when the engines agree on phonon (dynamic) stability — disagreement is surfaced, not hidden.

The lead composition is MgAl2O4 (spinel, space group Fd-3m), which carries a computed HSE06 bandgap of approximately 7–8 eV — among the widest in any oxide ceramic that also retains structural compatibility with standard thin-film deposition methods such as ALD and sputtered targets. Beyond the spinel, the composition set spans four distinct crystal-structure families: garnets (YAlO3), phosphates (Ca32, Li3PO4, Sr3P2O8), silicates (Y2SiO5, Mg2SiO4, Ca2SiO4), and a mixed boroaluminate (YAl3B4O12), supplemented by MgNb2O6 and BaCeO3. Each family brings somewhat different dielectric-constant versus bandgap trade-offs, but all occupy the 5–8 eV window that is the defining selection criterion. The breadth across crystal families is intentional: it means the patent family covers structurally diverse embodiments rather than a single lattice-matched system, and it provides a licensee with options for different deposition or sintering routes. Computational validation followed a two-stage protocol. In the first stage, hull-stability was assessed for each composition against the Materials Project convex hull, confirming thermodynamic stability (zero or near-zero distance from the hull) in the target phase. Thermodynamic stability alone is insufficient for thin-film device applications — a compound can be hull-stable but dynamically unstable (imaginary phonon modes indicating a structural instability at the nanoscale), so a second stage applied two independent machine-learning interatomic potentials, MACE and CHGNet, to compute phonon dispersion curves. Both potentials independently agree that the representative compositions — including the lead spinel and BPO4 as an early validated borate sub-genus member — are dynamically stable, with no imaginary phonon modes across the Brillouin zone. That consensus across two architecturally distinct ML potentials is the key validation gate: it eliminates the most common failure mode in computational materials pre-screening, where a single model's errors produce false positives. Electronic-structure calculations were performed at the HSE06 hybrid-functional level, which is the appropriate choice for wide-bandgap oxides where standard GGA functionals substantially underestimate the bandgap. HSE06 calculations have been completed for more than ten named members of the composition set, providing a reliable basis for ranking compositions by dielectric constant and leakage-relevant barrier height. A dedicated 13-composition DFPT campaign in the 2026 computation cycle extends this to the full dielectric tensor — not just the electronic contribution but the ionic (phonon-mediated) contribution that dominates the static dielectric response at packaging-relevant frequencies. DFPT dielectric-tensor data directly feeds the figures-of-merit used by packaging engineers: the capacitance-density in MIM structures and the equivalent-oxide-thickness in gate-stack applications. From an applications standpoint, the wide bandgap suppresses radiation-induced charge generation through two mechanisms: the larger band-to-band excitation threshold means fewer electron-hole pairs per unit ionizing dose, and the large oxygen-vacancy formation energies in these close-packed oxide structures reduce the density of pre-existing trapping sites. For packaging applications, the same large bandgap translates into high breakdown fields (typically correlating with bandgap in oxides) and low gate-leakage current density, which are the critical figures of merit for MIM capacitor dielectrics in HBM interposers and for TGV (through-glass-via) barrier layers where leakage across the via sidewall must be minimized. The co-optimization of these two use cases — packaging and rad-hard — within a single composition family is not accidental; it follows from the physics of wide-bandgap oxides and was the explicit design criterion for the computational screening campaign.

The addressable market spans two distinct but partially overlapping segments. In advanced semiconductor packaging, the relevant buyers are OSATs (outsourced semiconductor assembly and test houses) integrating dielectric layers in chiplet and 2.5D/3D architectures, and HBM and DRAM manufacturers who need MIM capacitor dielectrics and via-barrier layers at the highest memory-bandwidth density nodes. The global advanced packaging market has been estimated by multiple industry analysts at well above $50B in total by the late 2020s, though the dielectric-materials subset addressable through this asset is more narrowly the specialty high-k dielectric segment — conservatively estimated at several billion dollars once HBM4 and CoWoS-S ramp at leading-edge foundries. The total addressable figure stated for this asset is $5B, which should be read as an order-of-magnitude estimate spanning both packaging and aerospace electronics. In radiation-hard electronics, the customer set is space-electronics primes (satellite bus integrators, launch-vehicle avionics suppliers), defense electronics contractors, and nuclear-industry instrumentation suppliers. This market is smaller in unit volume but substantially higher in average selling price per wafer or component, and it has a particularly low tolerance for dielectric reliability failures — a single latch-up or dielectric breakdown event in a satellite can be mission-ending. The rad-hard electronics materials market is historically underserved by computational pre-screening methods because experimental space-qualification programs are slow and expensive; a computationally validated composition set with documented stability data accelerates the qualification timeline and reduces per-program NRE cost, which is a direct value proposition for primes bidding on space programs. The licensing and royalty logic for this asset fits both segments. In packaging, the natural model is a per-wafer royalty or a materials-supply agreement structured around the specific composition used in production, because the thin-film deposition step is typically contracted to specialty materials vendors or ALD chemistry suppliers who serve the OSATs. In aerospace, the model is more likely a one-time technology transfer plus a per-program license, consistent with how space-grade materials certifications are typically structured. A dual-use positioning — one composition family, two market segments, two licensing revenue streams — is the commercial case here.

single composition population favorable on both packaging-dielectric and rad-hardness screens

named composition + phase + dual-use stack; broad oxide-dielectric genus not claimed

The most natural acquirers and licensees fall into three categories. First, specialty dielectric materials vendors and ALD-precursor chemistry suppliers — companies like Entegris, Merck KGaA (through its semiconductor materials division), and Versum Materials (now part of Merck) — who supply the physical and chemical vapor deposition materials consumed by OSATs and foundries. A license to the named-composition set would allow a specialty materials vendor to offer a differentiated product line for next-generation HBM interposer and TGV applications without freedom-to-operate exposure. Second, space-electronics primes — Northrop Grumman, L3Harris, Raytheon, and the major European primes — who face recurring per-program NRE for materials qualification in harsh environments and would benefit from a computationally pre-screened composition set with stability documentation that shortens the qualification timeline. Third, HBM and DRAM manufacturers (SK Hynix, Samsung, Micron) who are actively seeking new MIM dielectric options for HBM4 and next-generation stacked memory architectures, where the current HfO2/Al2O3 incumbent stack is approaching its capacitance-density limits. A dual-use asset naturally invites dual-use acquirers: a large defense contractor with both space and advanced-electronics manufacturing arms could use this portfolio to vertically integrate materials for both application segments. Alternatively, a strategic acquirer in the advanced-packaging value chain — an OSAT like ASE Group or Amkor, or a substrate supplier like Ibiden — could use the composition patents defensively to maintain freedom-to-operate as new dielectric materials enter production qualification at their facilities. The composition-plus-device-use structure makes this asset more attractive to strategic acquirers than to pure financial buyers, because the value is realized through manufacturing adoption, not through broad blocking positions.

The primary technical risk is the gap between computational stability and thin-film device performance. Wide-bandgap oxide ceramics that are thermodynamically and dynamically stable in bulk can exhibit significant stoichiometry deviation, second-phase formation, and interface-trap density in thin-film geometries deposited by ALD or sputtering. The breakdown field — the single most commercially critical figure of merit for a packaging dielectric — is highly sensitive to these film-quality factors and cannot be reliably predicted from bulk HSE06 calculations alone. The MIM breakdown coupon measurements flagged as an open validation gate are not a formality; they are the experiment most likely to produce surprises. The risk-mitigation path is straightforward: prioritize a thin-film deposition and characterization program for the two or three compositions with the best computed dielectric-tensor plus bandgap combination (MgAl2O4 and YAlO3 are the natural first targets), and use those results to down-select the composition set before extensive patent prosecution spend on the broader named set. The secondary risk is manufacturing process compatibility. The phosphate and silicate compositions, while computationally well-characterized, do not have mature ALD precursor chemistries at the scale needed for 300 mm wafer production. Spinel and garnet ALD is more established in research literature but not yet in high-volume manufacturing. This is a timing risk rather than a fundamental barrier — the industry has a track record of developing ALD chemistries for new high-k oxides when the performance case is clear — but it means the time-to-production for licensees in advanced packaging is likely five to eight years from a standing start, not two to three. For the aerospace segment, where thin-film deposition is less standardized and low-volume qualification is acceptable, this timeline is less constraining. Positioning this asset honestly as a five-to-eight-year commercial-readiness play, rather than a near-term drop-in replacement, is the right expectation to set with buyers.