Refractory boride copper diffusion barrier family for glass-core packaging

The glass-core advanced-packaging substrates portfolio addresses one of the most consequential unsolved problems in next-generation semiconductor packaging: how to prevent copper from migrating through ultra-thin dielectric layers in glass-core interposers and substrates at the interconnect dimensions required by advanced nodes. Copper diffusion poisons adjacent dielectrics and degrades reliability; the industry currently relies on physical vapor-deposited tantalum nitride or titanium nitride liners that are increasingly difficult to scale as via diameters shrink below 5 µm and aspect ratios climb. This asset covers a broad family of alternative barriers — tungsten boride as the lead composition, flanked by six rigorously validated transition-metal diborides and an amorphous boron nitride backup arm — specifically engineered to offer equivalent or superior diffusion-blocking performance with better compatibility at the glass-dielectric interface. The strategic logic here is compositional breadth. Rather than a single-compound claim vulnerable to routine design-around, this filing sweeps an entire class of refractory metal diborides in the hexagonal layered structure (space groups P6_3/mmc and P6/mmm), then adds an amorphous BN fallback that is chemically orthogonal to all crystalline members. Anyone entering this barrier layer space with boride-class conductors — whether they optimize for resistivity, adhesion, or process temperature — lands inside this family. That breadth is the point: the filing is explicitly a design-around coverage instrument, intended to work in concert with the broader portfolio rather than stand alone. The timing is well-chosen. Glass-core packaging has moved from research curiosity to active qualification at multiple Tier 1 OSAT and IDM fabs between 2023 and 2026, with Intel's EMIB-T, Corning's glass interposer programs, and Samsung's advanced packaging roadmap all naming glass substrates as roadmap items. The barrier metallurgy question is now an engineering decision, not a future one, which means the whitespace this filing occupies is commercially proximate rather than speculative.

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.

Tungsten boride (B2W and the sub-stoichiometric WBx phase space) is the lead material in this family. The compound adopts the hexagonal AlB2-prototype structure (P6/mmm) at the B2W stoichiometry, placing tungsten atoms in layers sandwiched between graphene-like boron planes. This geometry produces a material that is electrically conductive — borides of this class typically show metallic resistivity in the 20–80 µΩ·cm range depending on deposition method — while presenting a dense, chemically inert interface to both copper and glass dielectrics. The boron-rich surface chemistry offers advantages over nitride barriers in terms of adhesion to SiO2-based and borosilicate glass surfaces, where B–O bonding at the interface can suppress interfacial void formation. The six validated diboride comparators — TiB2, ZrB2, HfB2, TaB2, VB2, and AlB2 — all adopt the same hexagonal prototype and were subjected to systematic phonon calculations. Phonon stability is the key computational gate for a crystalline diffusion barrier candidate: a material with imaginary (negative-frequency) phonon modes is mechanically or dynamically unstable and will not form a coherent, defect-free film under realistic processing conditions. The minimum phonon frequencies across the validated set are all positive and substantial: TiB2 reaches +1.95 THz at the zone boundary, ZrB2 hits +1.48 THz, VB2 +1.43 THz, AlB2 +1.16 THz, HfB2 +1.06 THz, and TaB2 +0.91 THz. These are not marginal results; the gaps above zero are large enough that thermal fluctuations at typical PVD or CVD deposition temperatures (150–400°C) are unlikely to drive the structures toward instability. The B2W/WBx lead was assessed separately and shows a positive minimum phonon frequency of approximately +1.06 THz by MACE, confirming its position in the stable set. Two independent computational pathways underpin these stability assessments. The primary sweep used the MACE machine-learning interatomic potential across the entire boride series. MatterSim molecular dynamics simulations were then run independently on the three most commercially prominent members — TiB2, ZrB2, and HfB2 — as an explicit cross-validation step. Both methods agree on stability for these three compounds, providing a consensus result that is substantially more defensible than a single-potential screen. Two DFT literature sources further corroborate the stability assignment for the family. MoB2 and CrB2 were included in the MACE sweep but returned adverse results — both are retained as comparator-tail members, meaning they are useful as prior art anchors and experimental controls precisely because they illustrate where the structural stability degrades, but they are not carried forward as positive claim members. The amorphous boron nitride (a-BN) backup arm is chemically and structurally distinct from all crystalline diboride members and from the explicitly excluded hexagonal BN and cubic BN thermal-liner applications. Amorphous BN is an ultralow-dielectric-constant material (k in the 2.0–3.0 range in thin-film form) that can simultaneously serve as a barrier and as part of the dielectric stack, which is a functionality the crystalline metallic diborides cannot provide. Its inclusion in the family broadens the claim to cover process flows where a conductive metallic liner is architecturally undesirable — for instance, in back-end flows where capacitance loading from a metallic barrier would degrade signal integrity. The specific exclusion of h-BN and c-BN from this arm is a deliberate negative limitation that separates this claim from the thermal-management applications of those phases, maintaining clean differentiation from prior filings in the portfolio.

The addressable market for copper diffusion barrier materials in advanced packaging sits at an estimated $0.5–1 billion annually, spanning both the materials supply chain (PVD targets, CVD precursors) and the process IP licensing layer above it. This estimate reflects the barrier layer's role as a thin but mandatory component in every copper interconnect structure — its economic value is disproportionate to its physical thickness because the barrier determines interconnect reliability lifetime and therefore commands a significant licensing premium relative to material cost alone. The relevant buying entities are foundries, OSATs, and IDMs running copper damascene or semi-additive processes in glass-core or silicon interposer substrates, along with barrier material suppliers seeking freedom to commercialize boride-class depositions without encumbering their customers. The transition to glass-core substrates is structurally forcing a re-evaluation of barrier metallurgy. Glass substrates have different thermal expansion coefficients, surface chemistries, and via aspect ratios than silicon or organic laminate substrates; TaN and TiN processes tuned for silicon are not automatically transferable. This creates a near-term qualification window during which fabs will be selecting barrier chemistries for their glass-core process flows — and compositional patents filed ahead of that qualification window carry meaningful leverage. The royalty logic is straightforward: a per-wafer or per-substrate licensing model applied across high-volume advanced-packaging production, where substrate counts are measured in millions per year at leading OSAT facilities, generates recurring revenue even at low per-unit rates. Alternatively, an outright acquisition of this asset by a barrier materials incumbent seeking freedom to operate across the boride class would be motivated by the need to avoid downstream litigation exposure once glass-core substrates reach volume production.

broad design-around barrier coverage

design-around coverage; nearest W-boride art is neutron-shielding/abandoned (li)

The most natural acquirers or licensees for this asset are barrier materials suppliers and equipment companies with positions in the advanced-packaging deposition space. Entegris, Merck KGaA (through its semiconductor materials division), and Linde's specialty gases business all have direct interests in expanding the process materials palette for glass-core packaging, and any of them would find compositional IP covering boride-class barriers relevant to their customer conversations with leading OSATs and IDMs. On the equipment side, companies like Applied Materials and Lam Research that supply PVD and ALD tools for barrier deposition have a secondary interest in IP that validates new barrier chemistries — it de-risks their customers' qualification decisions and supports their own process development roadmaps. The other natural buyer category is a foundry or OSAT with an active glass-core packaging program seeking freedom to operate across the boride barrier class without downstream litigation exposure. Intel (Foundry Services), Samsung Advanced Packaging, and ASE Group all have public glass-core or panel-level packaging programs that will require barrier qualification decisions within the next two to three years. Acquiring this asset outright would provide both freedom to operate and a defensive position against competitors entering the same space. The $0.5–1 billion addressable market estimate suggests that even a modest royalty stack across volume production generates a return that justifies acquisition at the multiples typical for early-stage materials IP with clean FTO.

The primary technical risk is the gap between computational stability validation and demonstrated barrier performance. Phonon stability confirms that these materials are synthesizable and mechanically viable, but copper diffusion barriers are qualified on reliability metrics — bias-temperature stress lifetime, time-dependent dielectric breakdown, copper penetration depth after annealing — that require physical thin-film deposition and characterization. None of the six diboride members or the a-BN arm have published barrier performance data in glass-core geometries, and the path from stable crystal structure to qualified process module involves significant engineering work. A buyer should budget for experimental validation campaigns covering at minimum: PVD or ALD deposition feasibility for B2W and TiB2/ZrB2, resistivity measurement of deposited films, copper adhesion and barrier lifetime testing at relevant temperature and bias conditions, and interface characterization against representative glass substrate surfaces. The a-BN arm carries additional deposition-process uncertainty because amorphous films are inherently more sensitive to deposition conditions than crystalline ones. The secondary risk is claim scope under prosecution. Composition claims for materials that appear in academic literature — even in different application contexts — can face prior art rejections that narrow the granted claims relative to the filed scope. The adverse-result members MoB2 and CrB2 provide useful prosecution tools (they define where the enabled class ends), but prosecution outcomes in the semiconductor materials space depend heavily on examiner familiarity with the specific application context. A buyer should engage patent counsel early to assess prosecution strategy, particularly for the a-BN arm where the distinction from h-BN/c-BN prior art will require careful claim drafting. The portfolio context — this asset working alongside other barrier and dielectric filings in the glass-core advanced-packaging substrates portfolio — suggests the strategic value is partially dependent on the overall portfolio remaining intact and well-prosecuted.