Bismuth rare-earth sesquioxide high-permittivity filler for package-integrated capacitors

The bismuth rare-earth sesquioxide family — ScBiO3, YBiO3, and TmBiO3 — represents a genus of dielectric filler materials that has not previously been claimed in the patent literature, despite sitting squarely in a high-value application space: embedded capacitors and heat-spreading composites for advanced semiconductor packaging. The core opportunity is the intersection of genuine dielectric performance (total permittivity from 37 to 53 across the three members, computed via density functional perturbation theory) and confirmed dynamic stability across four independent machine-learning interatomic potentials. That combination — whitespace plus multi-engine stability consensus plus quantified property advantage — is precisely what package-substrate designers need as they push for higher passive density under shrinking die footprints. Timing matters here. The semiconductor packaging industry is in the middle of a forced substitution away from conventional discrete passives toward package-integrated embedded capacitors, driven by the power integrity requirements of AI accelerators and high-bandwidth memory stacks. Filler materials for polymer-based embedded-capacitor laminates must simultaneously deliver high permittivity, thermal stability, and chemical compatibility with standard laminate resins. HfO2 and hafnate-based systems currently dominate the research pipeline, but their permittivity ceilings and processing constraints leave room for a new filler genus — one that can be dispersed at 10 to 50 vol% in a polymer matrix and contribute meaningfully to effective composite permittivity. The RBiO3 genus, with a top member sitting at epsilon of 53, addresses that gap. Within the high-power thermal-interface materials portfolio, this asset plays a complementary role to the separately developed Ruddlesden-Popper hafnate high-k arm. Together they form a genus-and-subgenus coverage strategy: the hafnate arm secures one chemical family, and this bismuth sesquioxide arm secures an independent, previously unclaimed chemical family. Each arm can stand alone commercially, but together they provide layered defensive breadth. This asset is characterized honestly as a complementary filler arm — not a standalone flagship — whose value lies partly in its whitespace position and partly in the credibility that four-engine stability consensus lends to its claim of practical processability.

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 three materials at issue — ScBiO3, YBiO3, and TmBiO3 — belong to the ABO3 perovskite-adjacent bismuth sesquioxide family, where the A-site is occupied by a trivalent rare-earth cation (scandium, yttrium, or thulium) and the B-site by bismuth(III). Bismuth-based oxides are known in the literature to achieve elevated polarizability through the lone-pair activity of Bi3+ 6s electrons, which can strongly couple to the lattice and drive high dielectric response without requiring a ferroelectric phase transition. The RBiO3 compounds explored here leverage that mechanism across three rare-earth partners of differing ionic radius — Sc3+ (0.745 Å) is the smallest, Y3+ (0.900 Å) intermediate, and Tm3+ (0.880 Å) between the two — which systematically modulates the unit-cell volume and the strength of the Bi lone-pair displacement, yielding a permittivity gradient of 37 (Sc), 48 (Y), and 53 (Tm). The dielectric constants were computed using density functional perturbation theory (DFPT), which provides the full dielectric tensor by evaluating the response of the electronic structure and ionic positions to a perturbative electric field. DFPT-derived permittivities at this level — with full ionic relaxation at converged supercell size — are the standard of the field for pre-synthesis screening and are consistently within 10 to 20 percent of experimentally measured values for well-converged oxide systems. Two independent DFT source calculations were used to cross-check the DFPT outputs. The permittivity values of 37 to 53 are meaningfully above the 20 to 25 range typical of HfO2 and comfortably exceed the 30 to 35 range reported for many hafnate variants, making the RBiO3 genus competitive for high-loading filler formulations. Dynamic stability — the question of whether the crystal structure is a true energy minimum rather than a saddle point that would decompose under thermal fluctuation — was adjudicated using four independent machine-learning interatomic potentials: MACE, CHGNet, MatterSim, and ORB. Each potential was trained on distinct datasets and uses a distinct architecture, so agreement across all four constitutes a genuine consensus rather than a correlated artifact. All four potentials assessed these structures as harmonically stable at converged supercell — meaning the phonon dispersion contains no imaginary-frequency modes — at the conditions relevant to processing and device operation. This four-of-four consensus is the strongest stability verdict available in the platform's adjudication workflow and provides high confidence that the structures will not spontaneously distort or decompose under ambient synthesis conditions. No space group is presently assigned, which reflects that the precise low-symmetry ground-state structure has not yet been experimentally pinned, but this does not undermine the stability finding, which was computed at the relaxed DFT geometry in each case. The discovery pathway for this genus was a five-criterion intersection screen (internally designated WE144), which filters the accessible materials space simultaneously by predicted dielectric constant, dynamic stability, synthesizability proxies, absence of toxic or critically supply-constrained elements (within acceptable limits), and freedom-to-operate whitespace. This multi-criterion funnel is designed to surface candidates that are not only computationally attractive but also practically actionable — the screen explicitly excludes materials that would face synthesis barriers or IP encumbrances at the composition level. The RBiO3 family cleared all five criteria, placing it in a small set of genuinely actionable high-k filler candidates. The intended deployment geometry is particle dispersion in a polymer laminate matrix at 10 to 50 vol%, a loading range consistent with standard embedded-capacitor laminate processing.

The addressable market for this asset sits at the intersection of two growing segments: embedded passive components in advanced packaging substrates, and high-permittivity filler materials for polymer dielectric laminates. Package-integrated capacitors — built directly into the substrate layers rather than mounted on the surface — are becoming structurally necessary as AI accelerator and high-bandwidth memory package designs push power delivery closer to the die. Substrate suppliers, assembly houses, and OEMs alike are sourcing filler materials that can increase the effective permittivity of dielectric layers without sacrificing processability or thermal reliability. The total addressable market for embedded capacitor materials and associated high-k filler compounds in advanced packaging has been estimated at $1 to 2 billion annually, reflecting the volume of substrate area deployed in leading-edge server, networking, and consumer electronics platforms. The customers in this space are primarily embedded-passive makers: specialty chemical suppliers who formulate filled polymer composites for laminate manufacturers, and large substrate manufacturers who vertically integrate filler sourcing. A patent covering a novel high-k filler genus can be licensed on a per-kilogram royalty basis, as a portfolio license bundled with process know-how, or as an exclusive supply arrangement in which the IP holder captures value through material sales rather than pure royalty streams. At reasonable market penetrations and royalty rates typical of specialty materials IP, the licensing economics are consistent with the total market estimate. The heat-spreading dimension of the use case — RBiO3 particles contributing to thermal conduction through the filled layer — adds a secondary value argument that could command premium positioning relative to pure dielectric fillers. Secular demand drivers are favorable. The move from 2D to 2.5D and 3D packaging architectures increases the area of active substrate requiring high-permittivity dielectric fill. Power budgets for AI inference chips are rising, which tightens power integrity requirements and makes embedded decoupling capacitors with higher effective capacitance per unit area more valuable. No structural headwind is apparent in the near-to-medium term, though the market remains relatively concentrated among a handful of large substrate suppliers, meaning that adoption decisions are made by a small number of procurement and R&D gatekeepers.

previously-unclaimed high-k filler genus (whitespace) with four-engine-confirmed stability

zero-hit targeted full-claim patent prescreen; previously-unclaimed high-k filler genus complementary to RP-hafnate sub-genus

The most natural acquirers or licensees for this asset are specialty materials companies that supply high-k filler compounds to laminate manufacturers and embedded-passive makers. Companies such as Ferro Corporation, Sakai Chemical, Nippon Chemical Industrial, and similar specialty oxide suppliers have established formulation and particle-processing capabilities that could be applied to the RBiO3 compositions without fundamental process innovation — the synthesis of rare-earth bismuth oxides by solid-state reaction or sol-gel routes is well within the technical repertoire of any competent oxide ceramics supplier. A license to the composition and device-use claims would give such a supplier a protected product line with a quantified dielectric performance advantage over HfO2, supported by computational validation that de-risks the development investment. Larger substrate manufacturers with integrated materials R&D — Ibiden, Shinko Electric, AT&S — would also be logical licensees if they pursue vertical integration into filler supply. A second buyer category is semiconductor packaging companies and IDMs seeking to secure IP position in advanced substrate materials ahead of volume production of next-generation AI or HBM packages. For these buyers, the value is defensive as much as commercial: owning or exclusively licensing the RBiO3 genus prevents a competitor from controlling access to a high-permittivity filler alternative. The complementarity with the hafnate arm makes a bundle license attractive — a single transaction securing both bismuth sesquioxide and hafnate high-k coverage — which argues for presenting this asset alongside the hafnate arm in any licensing conversation rather than as a standalone item.

The primary technical risk is the absence of experimental validation. No phase-pure coupon of ScBiO3, YBiO3, or TmBiO3 has yet been synthesized and characterized, meaning the DFPT permittivity values and the predicted crystal structure have not been confirmed by measurement. Bismuth oxide-based compounds can be prone to phase separation, volatility of Bi2O3 at elevated temperatures, and formation of competing secondary phases (e.g., Bi4Ti3O12-type layered structures), all of which could complicate synthesis of phase-pure RBiO3 samples. The roadmap to de-risk this is straightforward in concept: solid-state synthesis from Bi2O3 and the relevant rare-earth oxide (Sc2O3, Y2O3, Tm2O3), followed by XRD phase identification and impedance spectroscopy, with HSE06 corroboration calculations running in parallel to validate the GGA-DFPT permittivity estimates. Thulium oxide is a commercially available but modestly priced rare-earth material; scandium and yttrium oxides are more accessible and lower cost. None of these synthesis routes represents an exotic capability barrier. The secondary risk is commercial: the embedded-capacitor laminate market, while growing, is concentrated among a small number of large Asian substrate manufacturers who have long-standing relationships with incumbent filler suppliers. Displacing HfO2 or BaTiO3 in a qualified laminate formulation requires process qualification cycles that can take 18 to 36 months even for a chemically compelling alternative. The IP position provides leverage for licensing but does not accelerate the qualification timeline. The mitigation is to pursue co-development arrangements with a substrate manufacturer or specialty filler supplier who has existing qualified product lines and can run qualification in parallel with the experimental validation work — converting the patent from a blocking position into an enabling license that reduces the partner's investment risk.