Dielectric and substrate backup compositions: BeSiN2, HfGeO4, ZrGeO4, YAl3B4O12, and related species

This asset is a deliberately constructed genus-broadening backup set within the PFAS-free dielectric and process fluids portfolio, covering ten chemically distinct candidate compositions across nitride, germanate, fluoroaluminate, and aluminoborate material classes. Its strategic purpose is not to be the portfolio's headline claim, but to widen the claimable genus around the core Family C compositions — ensuring that if competitors or challengers attempt to design around the primary claims by substituting cations or anion frameworks, they encounter a filed, computationally validated wall of adjacent species. This is standard mature patent strategy, and the asset should be read as such: it is necessary, considered, and adds real portfolio breadth without overpromising on any single candidate. The timing of this filing reflects a real dynamic in advanced dielectric materials: the replacement of legacy PFAS-containing process and dielectric fluids is a regulatory-forced transition, not an optional upgrade. The European PFAS restriction dossier, U.S. EPA actions, and downstream pressure from semiconductor and power-electronics OEMs are collapsing the viable window for incumbent materials. That creates demand for a credible, broad IP position in alternative chemistries — and genus-broadening backup filings are precisely the instrument that converts early computational screening into durable defensive coverage before the market hardens around a smaller set of winners. What distinguishes Lattice Graph's approach here from a simple list of candidate materials is the computational rigor applied before filing. Each of the ten species in this set has been subjected to multi-engine machine-learning interatomic potential (MLIP) relaxation, and the four lead-promoted members — BeSiN2, HfGeO4, ZrGeO4, and YAl3B4O12 — have cleared a three-engine phonon stability consensus screen. This is not a paper claim on a formula; it is a claim supported by independent structural and dynamical validation. That distinction matters enormously for buyer due diligence and for any future IPR or validity challenge.

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

The ten compositions span three chemically distinct sub-families. The nitride arm includes NaSi2N3 and BeSiN2, both of which present wide-bandgap characteristics relevant to dielectric and substrate applications. BeSiN2, the lead nitride member, has been advanced to three-engine phonon-stable status, meaning three independent machine-learning potentials — MACE, CHGNet, and a third engine in the multi-engine screen — all converge on a dynamically stable ground-state structure with no imaginary phonon modes at the zone center or elsewhere in the Brillouin zone. NaSi2N3 passed two of three engines, with the third showing a soft mode that is marginal rather than clearly pathological; it remains in the set as a dependent claim member pending resolution. The germanate arm — HfGeO4 and ZrGeO4 — is of particular interest because both members achieved three-engine phonon-stable promotion and represent a structurally coherent pair: hafnium and zirconium are isoelectronic, share similar ionic radii, and both form germanate phases with high dielectric constants and wide gaps. The chemical logic here is strong, and the pair provides natural redundancy: if one member encounters an FTO issue or an unexpected stability problem at the DFT level, the other provides a substantively equivalent fallback within the same claim genus. Ca3HfSi2O9, also in the set, contains hafnium in a silicate rather than germanate coordination environment; it has been relaxation-converged under two MLIPs but has not yet been put through the phonon screen, leaving its dynamic stability unresolved — an honest open gate. The aluminoborate YAl3B4O12 is the fourth member promoted to three-engine stable status. This is a structurally complex oxide with a garnet-like topology, and its phonon stability under three independent potentials is a meaningful result given that garnets in this composition space can exhibit soft modes when cation size mismatch is present. Li2AlBO4 passed two of two engines (MACE and CHGNet) and is categorized as stable by that criterion, though the multi-engine screen has not yet been extended to a third potential. LaScO3, a perovskite-structured rare-earth scandiate, passed two of three engines with one marginal soft mode — it is retained as a dependent claim member but is not promoted to the lead set. The fluoroaluminate sub-family — CaAlF5 and LiBaAlF6 — occupies a deliberately different role. Both compositions were assessed through the multi-engine phonon screen and were found to exhibit phonon softening, meaning they are dynamically unstable as currently described in their nominal stoichiometry and assumed structure. Rather than being dropped, they are retained as dependent "matrix member" claims: the fluoroaluminate motif is chemically distinct from the oxide and nitride arms and provides compositional diversity that broadens the genus even if the specific stoichiometries require structural refinement before they could anchor independent claims. This is an intellectually honest use of backup claims — acknowledging dynamical instability while preserving the option to return to these compositions as structural data improve. All ten species were subjected to geometry relaxation under at least two MLIPs (MACE and CHGNet) in the current campaign, giving confidence that the reported phonon results reflect converged, physically reasonable starting structures rather than artifacts of unconverged geometries.

The addressable market for this backup set sits at the intersection of two industries: advanced dielectric films and coatings for semiconductor and power-electronics applications, and wide-bandgap substrate materials for next-generation device platforms. The estimated addressable opportunity associated with the specific niches covered by this genus-broadening set — alternative high-k dielectrics, non-PFAS process fluids, and fluoride-matrix specialty materials — is in the $200 million to $500 million range, reflecting a meaningful but not dominant slice of the broader dielectric materials market. These are realistic estimates for the specific compositions claimed, not projections of the entire high-k or substrate market. The primary buyer for materials in this set is the advanced semiconductor supply chain: IDMs and foundries seeking alternative high-k gate dielectrics as hafnium oxide scaling limits approach, compound semiconductor manufacturers needing wide-gap substrates compatible with GaN and SiC device epitaxy, and specialty chemical suppliers developing PFAS-replacement formulations for cleaning and planarization processes. The fluoroaluminate members (even in their current phonon-soft state) are of interest to the solid electrolyte and battery electrolyte communities, where fluoride-rich compositions are actively investigated for their electrochemical stability windows. Each of these buyer segments operates on different royalty and licensing logic: semiconductor IDMs typically prefer cross-license or field-of-use exclusive licenses; specialty chemical suppliers prefer royalty-bearing non-exclusive arrangements; substrate material companies may seek assignment of specific composition claims rather than broad licenses. Regulatory pressure is the primary market accelerant. The PFAS restriction wave eliminates incumbent fluorinated dielectric and process fluids on a statutory timeline, not a performance timeline, which means the substitution window is forced and relatively predictable. Materials that are computationally pre-validated and IP-protected before the regulatory deadline hits have a structural advantage over late entrants, regardless of whether any single composition in this backup set ultimately becomes the commercial winner. The genus-broadening function of this asset is precisely to capture that advantage across a wide chemical space rather than betting entirely on a single formula.

genus-broadening defensive backup coverage across Family C lanes

dependent / genus-broadening / method-of-use form pending full-claim FTO + dielectric/phonon validation

The most natural acquirers or licensees for this asset are companies that already hold or are actively building positions in the core Family C dielectric compositions, since the genus-broadening claims in this set are most valuable as an extension of an existing portfolio rather than as a standalone holding. This points primarily to advanced semiconductor material suppliers (Merck KGaA's electronics materials division, Entegris, Air Products, and similar specialty chemical companies with ALD precursor and high-k dielectric programs), and to IDMs or foundries with in-house materials IP programs (Intel, TSMC, Samsung, and GlobalFoundries all have active high-k dielectric IP portfolios). For any of these buyers, acquiring this backup set in conjunction with the core Family C assets provides portfolio depth that complicates competitor design-around efforts. A secondary buyer category is the power electronics and wide-gap substrate community: companies developing AlN, Ga2O3, or diamond substrates who need IP coverage in the adjacent dielectric and passivation layer space. Wolfspeed, Infineon, and STMicroelectronics are all expanding their wide-gap semiconductor supply chains, and a pre-validated IP position in wide-gap nitride and oxide dielectrics could complement their device IP. Licensing rather than acquisition is the more likely transaction form here, given the early stage of the computational validation and the open FTO gates; a staged license with milestone payments tied to completion of the dielectric tensor calculations and full FTO clearance would be a commercially reasonable structure.

The most significant risk in this asset is the open validation pipeline: four of the ten compositions have achieved three-engine phonon-stable status, but none have been subjected to DFPT dielectric tensor calculation, which means the property claims most directly relevant to the commercial application (dielectric constant, dielectric loss) are not yet computationally substantiated. A buyer or licensor who relies on these claims before that work is completed takes on the risk that one or more promoted compositions yields unexpectedly low dielectric constants or high loss tangents, which would substantially weaken the commercial case for that member. The two fluoroaluminate members carry additional risk given their phonon-soft status, which would need to be resolved (either by finding a stable polymorph or by narrowing the claim to a specific coordination environment) before those compositions could anchor independent claims. Ca3HfSi2O9 has no phonon data at all and should be treated as speculative until screened. The roadmap to de-risk this asset is well-defined and manageable. Priority actions are: completing DFPT dielectric tensor calculations for BeSiN2, HfGeO4, ZrGeO4, and YAl3B4O12 (the four promoted members); commissioning full-claim FTO analysis for those same four compositions; running the phonon screen for Ca3HfSi2O9; and investigating whether a stable polymorph exists for CaAlF5 and LiBaAlF6 that passes the multi-engine phonon screen. None of these are open-ended research problems — they are defined computational tasks with deterministic outputs, executable on a timescale of weeks to a few months within an established pipeline. A buyer with access to that pipeline, or who can contract for it, can convert the current title-level backup position into a substantiated, FTO-cleared genus claim set with manageable additional investment.