Low-alkali barium borosilicate aluminosilicate glass for glass-core package substrates

Glass-core advanced packaging substrates have moved from academic curiosity to industry inflection point, driven by the insatiable bandwidth demands of AI accelerators and high-bandwidth memory stacks. Silicon and organic laminate substrates are running into physical limits — silicon is expensive and constrained in panel size, organics warp and absorb moisture, and neither can hold the tight dimensional tolerances that chiplet-to-chiplet interconnects at pitches below 5 µm require. Glass, with its atomically smooth surfaces, near-zero moisture uptake, and tunable coefficient of thermal expansion, is positioned to capture a meaningful slice of the advanced packaging bill of materials over the next several years. The forced substitution dynamic is real: as AI training and inference clusters push HBM4 and beyond, the interconnect density and signal-integrity requirements of the substrate become a system-level bottleneck, and loss tangent at millimeter-wave frequencies becomes a first-order specification rather than a footnote. This composition filing — a low-alkali barium borosilicate aluminosilicate (Ba-BSA) glass — addresses the core materials science problem that separates a manufacturable glass-core substrate from a laboratory demonstration. Alkali ions (Na⁺, K⁺, Li⁺) are mobile under electric fields, degrade loss tangent at high frequency, and disrupt the hermetic interface with copper metallization. Suppressing alkali content to ≤1 mol% while simultaneously hitting the thermal and dimensional targets required by panel-scale glass handling — glass transition temperature between 720 and 800 °C and CTE between 3 and 9 ppm/K — requires precise compositional engineering. Barium oxide serves as a network modifier that raises Tg and stabilizes the glass network without introducing mobile alkali species; boron and aluminum oxide components adjust liquidus viscosity for manufacturability and tune CTE into the target window matched to silicon chiplets. The result is a composition space with a credible path to sub-0.005 loss tangent at 10–40 GHz — a specification directly relevant to the RF and high-speed digital interconnects embedded in next-generation AI package substrates. The composition claim is filed within the glass-core advanced-packaging substrates portfolio, which covers the full stack from glass core through metallization to panel-level processing. This particular filing is the lead composition claim for the Ba-BSA family, making it the foundational piece that anchors the chemistry side of the portfolio.

The target composition is a quaternary oxide glass in the SiO₂–B₂O₃–Al₂O₃–BaO system, with alkali content deliberately bounded at or below 1 mol% across the entire claimed composition space. This constraint is not arbitrary: alkali ions are the dominant source of ionic conduction in silicate glasses, and even trace amounts of Na₂O or K₂O increase dielectric loss at gigahertz frequencies through relaxation mechanisms that scale with ion mobility. The ≤1 mol% ceiling is the central composition claim and is what differentiates the Ba-BSA family from conventional borosilicate glasses (such as display-grade alkali-containing formulations), which typically carry 3–10 mol% alkali oxides introduced deliberately to lower melting temperature and aid fining. Barium oxide plays a dual role. As a heavy alkaline earth modifier, Ba²⁺ occupies network-modifying positions without contributing mobile charge carriers, raising the dielectric loss threshold relative to alkali modifiers of comparable modifier fraction. Concurrently, BaO suppresses the liquidus temperature of borosilicate systems, improving the meltability of a composition that, absent any modifier, would be extremely viscous and prone to devitrification. The alumina component (Al₂O₃) operates primarily as a network intermediate: at compositions where Al:Na or Al:alkaline-earth ratios favor tetrahedral coordination, Al enters the glass network as [AlO₄]⁻ tetrahedra charge-compensated by nearby Ba²⁺, tightening the network and increasing Tg toward the 720–800 °C window required for compatibility with subsequent metallization anneal steps and panel-handling thermal cycles. Boron oxide (B₂O₃) serves the complementary role of lowering the softening point to keep the composition processable at temperatures below the crystallization boundary, and contributes to the anomalously low CTE of boron-rich regions of the glass network through the well-established boron anomaly (three-coordinate boron forming rigid planar units). The net effect of balancing these four oxides — SiO₂ for network backbone, B₂O₃ for CTE and viscosity, Al₂O₃ for network rigidity and Tg, BaO as a non-alkali modifier — is a composition space that can simultaneously satisfy the Tg (720–800 °C), CTE (3–9 ppm/K), and loss tangent (<0.005 at 10–40 GHz) property targets. The computational and data-science basis for the claimed property windows rests on two complementary analyses. The primary evidence base is a curated aggregation of approximately 278,000 rows of experimental glass property data (the internal database designated DB8), from which compositions meeting the stated property windows were identified and validated as physically reachable within the Ba-BSA system. This is not a simulation of glass properties from first principles — amorphous glasses do not have a well-defined crystalline unit cell amenable to standard DFT phonon or stability calculations — but rather a large-scale empirical cross-validation against measured data from the literature and proprietary sources. The Tg and CTE anchor values cited in the specification are grounded in this database query, giving the property claims a statistical basis across thousands of real compositions rather than relying on a single measurement or a semi-empirical model. The second analysis is a 484-row stack-compatibility study examining how the glass core interacts with the proposed metallization and dielectric layer stack under thermal cycling, informing the CTE window selection and flagging compositions where mismatch would drive interfacial delamination under package assembly conditions. The phonon-stability framework that Lattice Graph applies to crystalline candidates (multi-potential consensus across MACE, CHGNet, MatterSim, and ORB, with DFT confirmation) is not directly applicable to an amorphous glass composition — glasses are inherently metastable, and their stability is assessed through different criteria: resistance to devitrification (crystallization) above Tg, chemical durability, and measured property consistency. The relevant open validation gate for this composition family is in-house loss-tangent measurement on a sintered or melt-quenched sample of the target composition. Until that measurement is in hand across the 10–40 GHz band, the loss-tangent specification remains a computationally informed target derived from the DB8 correlation between alkali content and dielectric loss, rather than a confirmed experimental result.

The addressable market for glass-core advanced packaging substrates is currently estimated in the range of $1–5 billion over the near-to-medium term, with substantial upside contingent on the rate at which AI accelerator and HBM-stacked memory packages transition from organic laminate to glass cores. This range reflects genuine uncertainty: the market is at an early commercial inflection, with the first high-volume glass-core substrate programs announced by major OSAT and substrate suppliers only in recent years. The buyers are glass substrate vendors — companies that produce and qualify panel-scale glass cores for chipmakers and OSATs — rather than semiconductor fabs or end-device OEMs directly. These vendors require glass compositions that can be sourced reliably, processed on existing display-panel or specialty-glass manufacturing lines with minimal capital retooling, and qualified to the increasingly stringent electrical and mechanical specifications imposed by HBM and AI package designs. The royalty and licensing logic for a composition patent in this space is relatively straightforward. Any glass substrate vendor that manufactures a product falling within the Ba-BSA composition space and property window — alkali ≤1 mol%, Tg 720–800 °C, CTE 3–9 ppm/K, loss tangent <0.005 at 10–40 GHz — would require a license. Because glass composition is a material-level claim, the reach extends across the value chain: it captures not just the glass manufacturer but potentially any downstream substrate integrator who specifies or incorporates the glass core in a product. The incumbents in the relevant manufacturing space are largely display and carrier glass makers (companies with established float or fusion-draw glass manufacturing) who are attempting to pivot their process expertise to the more demanding substrate market. These companies have the process infrastructure but historically have not held deep IP in low-alkali specialty compositions optimized for microelectronics dielectric performance rather than optical clarity or substrate flatness, creating a licensing opportunity for a composition that has been purpose-designed and filed for the packaging application.

low-alkali low-loss core matched to the disclosed layer stack

alkali <=1 mol% + property windows

The natural acquirers and licensees for this asset are glass substrate vendors that are actively building or qualifying glass-core panel lines for the advanced packaging market. This includes large display-glass manufacturers pivoting capacity toward packaging substrates, specialty glass companies with established alkaline-earth aluminosilicate manufacturing expertise, and vertically integrated semiconductor substrate suppliers who are developing glass-core alternatives to organic laminates. For a display-glass company, licensing the Ba-BSA composition claim resolves the IP risk of entering the packaging-grade low-alkali space without independently developing and filing a composition portfolio — particularly attractive if their existing formulations carry alkali content above the 1 mol% threshold and they need to reformulate for the HBM/AI substrate market. Strategic acquirers could also include chipmakers or OSATs who are qualifying glass-core interposers internally and wish to secure IP protection on the substrate glass composition as part of their advanced packaging supply chain strategy. The portfolio's coverage of the full stack — glass core composition, metallization layer compatibility, panel-level processing — makes it more valuable as a bundle than as an isolated composition claim, and the glass-core substrate portfolio as a whole would be most attractive to a buyer who intends to participate in substrate manufacturing or to cross-license with panel-glass suppliers. The relatively clean FTO position strengthens the asset's attractiveness to licensees who need confidence that practicing the technology does not expose them to third-party infringement risk.

The primary technical risk is the open loss-tangent validation gate. The sub-0.005 at 10–40 GHz target is the central commercial differentiator of this composition, and while the DB8 correlation strongly supports its achievability at ≤1 mol% alkali, a synthesized sample that falls short of this target — due to residual alkali contamination in the melt, unforeseen structural relaxation effects at these frequencies, or batch impurities — would require narrowing the claimed property window or revising the composition to reduce alkali further. The path to de-risking this is direct: synthesis of a target-composition glass slug by melt-quench or sol-gel route, followed by vector-network-analyzer measurement of the loss tangent on a polished sample across the 10–40 GHz band. This is a routine characterization experiment for a glass laboratory and does not require specialized equipment beyond what any advanced glass research facility would have in house. The commercial risk is timing and market-adoption rate. The glass-core substrate market is real but early: panel-scale manufacturing of glass cores at high yield is technically demanding, and the industry transition from organic to glass substrates has been slower than some early forecasts suggested. If organic laminate substrates extend their performance envelope through new resin chemistries or additive processing, the forced-substitution dynamic could be delayed, compressing the licensing window. The composition claim has a finite patent term, so prolonged market development timelines reduce the period of exclusivity available during peak market demand. Mitigations include ensuring the broader family claims (low-alkali aluminosilicate, low-alkali borosilicate) are filed and prosecuted to provide coverage that is not dependent on the specific Ba-BSA composition, and pursuing licensing discussions with substrate vendors now rather than waiting for the market to fully develop.