Glass-core packaging stack with aluminum borate liner, tungsten boride barrier, and chlorine-retaining RDL dielectric

Advanced semiconductor packaging is undergoing a generational substrate shift from silicon interposers and organic laminates to glass-core panels. Glass offers lower dielectric loss, tighter coefficient of thermal expansion control, and superior planarity for fine-pitch redistribution layers, but it creates a materials integration problem that the industry's incumbent TaN/TiN barrier and SiO2/SiN dielectric toolchain was not designed to solve. The relevant question is not whether glass-core packaging will displace legacy substrates in AI accelerator and high-bandwidth-memory assemblies — that transition is underway — but which barrier, liner, and dielectric chemistry will be standardized into the glass-core RDL process and locked in by the time volume manufacturing begins. This patent family claims the integrated stack: a glass core with through-glass vias and redistribution layer, a conformal aluminum borate (AlBO3) liner at 1-30 nm, a crystalline tungsten diboride copper diffusion barrier at 1-25 nm, copper seed and fill, and an amorphous retained-chlorine aluminum oxyhalide (AlOxCly) RDL dielectric. The claim strategy is transparent about which elements are novel and which are acknowledged art: AlBO3 as a standalone dielectric and ternary W-B-N barriers are prior art. The novelty resides in the specific integration of these layers in operative order on a glass core, the binary crystalline WB2 barrier window that steps around ternary nitrogen-containing barrier art, the conformal liner role of AlBO3, and the retained-chlorine composition of the amorphous dielectric. That specificity is the asset's commercial strength — and its boundary. The race window is now. Glass-core substrate vendors are qualifying processes for AI and HBM packaging customers within a two-to-three year horizon. The party that holds a validated, patent-protected barrier-liner-dielectric recipe during this standardization period captures a durable royalty position, because packaging process recipes, once qualified in high-volume manufacturing, are rarely substituted.

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 stack architecture addresses two physics constraints that glass-core RDL imposes simultaneously. First, copper must be completely contained within the redistribution layer — copper ion migration into the dielectric is the primary reliability failure mode for RDL, and glass cores lack the self-limiting diffusion properties that silicon-based interposers partially provide. Second, the entire stack must remain mechanically stable through repeated thermal cycles against a glass coefficient of thermal expansion that differs significantly from copper and from silicon-based dielectrics. The AlBO3 conformal liner (1-30 nm) is deposited first on the glass and TGV sidewalls, providing adhesion and a chemically resistant interface layer. Binary crystalline WB2 in the Cmcm space group then forms the copper diffusion barrier at 1-25 nm. The key computed property is the copper diffusion barrier energy of 0.6-0.7 eV — this is the activation energy for copper transport across the WB2 film, and it is the figure that determines barrier lifetime under operating conditions. Higher barrier energy means slower copper migration and longer interconnect reliability. The outermost RDL dielectric is an amorphous AlOxCly film formed by a melt-quench process that deliberately retains chlorine from the precursor chemistry. The retained chlorine modifies the amorphous network in ways that distinguish this material from standard alumina (Al2O3), which is excluded by explicit negative claim limitation, and from conventional SiO2/SiN dielectrics. Beyond its dielectric function, the AlOxCly film must maintain adhesion to the underlying boride barrier through the thermal cycling the stack must survive. The claim family includes a range of alternative barrier materials — tungsten-boron-nitrogen ternary, titanium diboride, zirconium diboride, hafnium diboride, chromium diboride, Mo2B, and AlCr3B4 — preserving design flexibility for process integration while building design-around resistance into the family breadth. The preferred embodiment is the nitrogen-free, nickel-free binary WB2, chosen to navigate around the one identified blocking reference (a ternary W-B-N patent). Thickness ranges for the liner (1-30 nm) and barrier (1-25 nm) are consistent with atomic layer deposition and physical vapor deposition capabilities at current packaging fabs.

The addressable market for advanced semiconductor packaging materials and process IP spans wafer-level and panel-level packaging for AI accelerators, high-bandwidth memory, and advanced logic, with an estimated market value exceeding $10 billion. This estimate encompasses packaging substrate revenue, materials supply, and the associated IP licensing opportunity across glass-core substrate vendors and OSAT assemblers. The AI accelerator buildout and HBM capacity expansion are the primary near-term demand drivers, and both depend on glass-core substrates reaching high-volume manufacturing readiness within the late 2020s. Royalty logic for this asset is naturally per-substrate-area or per-wafer, because the barrier, liner, and dielectric films are consumed in every packaged unit. A small per-area royalty on a multi-billion-unit packaging market with high per-unit substrate cost compounds significantly over the volume ramp. A field-of-use license structure could separately address HBM/AI OSAT applications and broader logic packaging, allowing differentiated royalty rates by end-market value. The substitutable barrier chemistry across the claim family is a commercial feature as well as a technical one: it allows a licensee to match the stack to its existing deposition equipment — whether ALD, sputtering, or CVD — rather than requiring capital investment in new tooling. This lowers the barrier to licensing adoption and expands the practical addressable population of licensees within the packaging industry.

CTE-matched glass-compatible Cu-blockable RDL with substitutable barrier/dielectric chemistries

binary crystalline WB2 window + conformal liner integration + retained-Cl AlOxCly; ternary-N as fallback

The two primary buyer archetypes are glass-core substrate vendors and HBM and AI accelerator OSATs. Substrate vendors building commercial glass-core panel offerings need a qualified barrier, liner, and dielectric recipe to offer copper RDL capability to their packaging customers — licensing this stack as a validated process recipe differentiates their glass-core product against competitors still adapting silicon-interposer chemistry. These buyers would most naturally take an exclusive or field-limited license tied to their substrate platform, and the patent protection provides them a defensible process moat during the volume ramp. HBM and AI accelerator OSATs are the volume consumers of the glass-core substrate output and represent a second licensing tier with high per-unit royalty leverage. An OSAT that qualifies this barrier-liner-dielectric stack into its high-bandwidth-memory or AI package process flow locks in the chemistry for the duration of that product's manufacturing lifetime. For an acquirer evaluating the full packaging IP portfolio, this asset bundles naturally with adjacent glass-core dielectric claims from the same portfolio of integrated packaging, storage, and PFAS-treatment systems, offering a prospective buyer or licensor the ability to present an integrated glass-core IP package rather than individual film-by-film licenses — a structurally stronger negotiating position with large strategic customers.

The primary technical risk is integration reliability, not materials stability. The three-potential phonon consensus establishes that WB2 is a stable phase, but phonon stability is a property of an isolated crystal — it does not validate copper containment under thermal cycling when the barrier is sandwiched between a glass core and a copper RDL under mechanical stress. The ab-initio MD simulation that would have partially addressed this at the atomic scale crashed without converging, so reliability under dynamic conditions remains experimentally unvalidated. The 85°C/85% humidity and thermal-cycle reliability coupon is the essential next spend, and its cost and timeline should be factored into any acquisition or licensing negotiation. The patent risk is 103 obviousness exposure: the individual films (AlBO3, WB2, amorphous alumina) are each practiced in adjacent art, and the claim's novelty depends on the integration and the specific binary WB2 and retained-chlorine dielectric windows. A competitor who reaches a working glass-core stack by a different sequence or a different binary boride could argue the combination was obvious from the acknowledged prior art. The negative limitations — liner not Al2O3, barrier nitrogen-free and nickel-free — narrow the claim scope precisely, which simultaneously strengthens validity and reduces the breadth of coverage. A buyer should model the scenario in which a competitor qualifies TiB2 or ZrB2 as the barrier and assess whether the family's alternative-embodiment claims provide adequate coverage of that design-around path before completing due diligence.