Depth-graded tungsten boron nitride copper diffusion barrier for glass-core vias

The glass-core packaging substrate is under genuine adoption pressure. As logic chiplet stacks push bandwidth and power density upward, the industry is migrating away from organic laminates toward glass cores precisely because glass offers lower dielectric loss, tighter via pitch, and more stable coefficient-of-thermal-expansion matching to silicon. That migration, however, creates a materials problem that did not exist at scale before: copper vias must be isolated from an oxide-rich glass interface, and the chemistry at that interface is nothing like the silicon-dioxide or silicon-nitride surfaces the barrier community spent two decades optimizing. A homogeneous refractory nitride or boride deposited across that transition simply cannot satisfy both boundary conditions simultaneously — the glass side demands a more oxide-compatible chemistry while the copper side demands a nitrogen-dense, grain-boundary-blocking microstructure. The depth-graded tungsten boron nitride barrier described here is a direct answer to that gap. The invention's core claim is that the composition gradient itself is the invention, not any single stoichiometry. A boron-enriched face contacts the aluminoborosilicate glass or aluminum borate interface layer; a nitrogen-enriched face contacts the copper seed. This architecture allows the depositor to tune each interface independently with a single physical-vapor-deposition or CVD step by ramping reactive-gas partial pressures during deposition — a process already native to sputter tools used in advanced packaging fabs. The gradient is directly verifiable by secondary-ion mass spectrometry, X-ray photoelectron spectroscopy, electron energy-loss spectroscopy, or atom-probe tomography, making the claimed feature analytically traceable and unambiguous in litigation. This asset sits within the glass-core advanced-packaging substrates portfolio as a dependent, backup filing — it adds depth and claim coverage beneath the primary barrier positions rather than standing alone as a flagship. That is an honest characterization and also a meaningful one: dependent claims that survive invalidity of a parent can anchor continuation strategy, licensing negotiations, and design-around pressure on competitors who might attempt to practice a near-homogeneous variant.

Tungsten boron nitride (W-B-N) in its homogeneous form is already understood as a promising copper-diffusion barrier because the ternary alloy suppresses grain growth relative to pure TaN or TiN, reduces grain-boundary density, and provides a thermodynamic sink against copper diffusion. The innovation here moves beyond homogeneous W-B-N by recognizing that the two interfaces a via barrier must serve impose chemically contradictory demands. On the glass-core side — specifically when the dielectric is aluminoborosilicate glass or when an aluminum borate (AlBO3) adhesion layer is present — the barrier surface needs boron-rich coordination environments to bond robustly to an oxide network without creating an interfacial phase that dewets or cracks during thermal cycling. On the copper seed side, nitrogen enrichment is what disrupts grain-boundary pathways: nitrogen pins grain boundaries, raises the activation energy for copper diffusion, and has been shown repeatedly in the copper interconnect literature to extend electromigration lifetime. A single homogeneous composition cannot optimize both simultaneously. The graded film resolves this by sweeping the B/N ratio monotonically through the barrier thickness during deposition. In physical-vapor deposition, this is accomplished by starting with a boron-rich reactive atmosphere (higher B2H6 or boron target contribution relative to N2 flow) and transitioning toward nitrogen-rich conditions as the deposition proceeds. The resulting film is continuous — there is no discrete junction that could delaminate — and the gradient is verifiable at sub-nanometer depth resolution using SIMS depth profiling or APT reconstruction. The depth profile itself becomes the patent-traceable feature: any competitor practicing a gradient from boron-rich to nitrogen-rich across a W-B-N film in this device geometry is squarely within the claimed space, regardless of the exact intermediate stoichiometry. From a simulation standpoint, the portfolio's prophetic example (designated internally as Example 7B) models the interface stabilization contribution of the gradient architecture. It is important to be transparent: this simulation is prophetic rather than experimentally verified at the time of filing, meaning it is a computational prediction of favorable interface energetics rather than a measured delamination strength. The machine-learning interatomic potential consensus framework used elsewhere in the portfolio (MACE, CHGNet, MatterSim, ORB) was not applied here because the graded film is not a periodic crystal amenable to phonon-stability analysis — it is an amorphous-to-nanocrystalline gradient composition, and conventional MLIP/DFT phonon workflows do not map directly onto it. The relevant validation for this material is therefore experimental: a SIMS coupon demonstrating the boron-to-nitrogen gradient through the deposited film is the primary proof gate, and that measurement has not yet been completed as of filing. The key property to understand is that no fixed stoichiometry needs to be claimed and defended. The invention claims the depth-profiled architecture — boron-rich at the glass-facing side, nitrogen-enriched at the copper-facing side — across the class of W-B-N films. This is strategically significant: it means Lattice Graph does not need to know the optimal intermediate composition, and a licensee is not boxed into a narrow process window. The claim covers the gradient concept as a device-incorporated structure, giving it both process and product-use reach.

The addressable market for this asset sits within the glass-core substrate advanced-packaging segment, which is currently in a transition from development-scale to volume production. Analyst estimates for the glass-core substrate market project revenues in the range of several hundred million dollars by the late 2020s, rising toward multi-billion-dollar scale as Intel, Samsung, and their substrate supply chain complete qualification of glass panels at panel scale. The copper diffusion barrier step is a consumable process input — every via in every substrate requires a barrier deposition — so the licensing lever is per-substrate or per-wafer royalty on barrier deposition processes or barrier-containing device structures. The addressable market for the specific barrier segment is estimated at $200 to $500 million, recognizing that this is a sub-component of the total substrate value rather than the total packaging market itself. Customers for a license or technology transfer are primarily the substrate manufacturers (companies like Corning, AGC, Absolics, or contract packaging houses qualifying glass-core panels), the deposition-equipment vendors who supply the sputter tools and want to offer a differentiated barrier recipe on their systems, and potentially the integrated device manufacturers who specify barrier chemistry in their substrate procurement contracts. Royalty logic could be structured as a process-step license (per deposition run on qualifying glass-core panel lines) or as a materials-incorporation license (per substrate unit shipped containing a depth-graded W-B-N barrier). The timing dynamic matters here. The industry is still early enough in glass-core adoption that barrier chemistry has not been locked into long-term supply agreements. A patent position that covers the gradient architecture can shape which barrier approaches become the de facto standard during the qualification window — after which switching costs make displacement very difficult. That window is likely two to four years wide from today.

no exact-stoichiometry claim needed; gradient is the invention

depth-profiled gradient (not homogeneous W-B-N)

The most strategically natural acquirers or licensees for this asset are substrate manufacturers qualifying glass-core panels for high-bandwidth memory and advanced logic packaging — companies such as Absolics, AGC, and the substrate arms of major OSAT and IDM players. For these companies, holding or licensing a granted patent on the gradient barrier architecture gives them a process-differentiation claim against competitors who are depositing homogeneous W-B-N or conventional TaN barriers. Deposition equipment vendors — particularly those supplying physical-vapor-deposition cluster tools to glass-core packaging lines — are a second natural buyer class: a barrier recipe license bundled with tool qualification creates a meaningful lock-in advantage over competing tool suppliers. Finally, integrated device manufacturers who control barrier chemistry in their substrate procurement specifications could use this asset defensively to ensure freedom to specify gradient W-B-N barriers in supplier contracts without risk of third-party claims. Given the backup and dependent nature of this filing, the most likely transaction structure is a portfolio license that includes this asset alongside the primary W-B-N barrier positions rather than a standalone sale.

The principal risk is the open experimental validation gate. This asset currently rests on a prophetic simulation rather than a measured SIMS gradient profile, and sophisticated buyers will treat it as pre-proof until that coupon data exists. The cost to close this gate is low — a SIMS depth profile is inexpensive relative to the licensing potential — but until the experiment is done, a buyer must either accept the prophetic status or fund the measurement themselves. A second risk is that this is a dependent and backup filing, meaning its enforceability is partly linked to the health of the parent claims in the broader glass-core barrier family; if the parent family is successfully invalidated, this gradient claim would need to stand alone, which would require it to be treated as an independent claim position rather than a backup. The roadmap to de-risk is clear: complete the SIMS gradient coupon to establish experimental proof of the depth profile, then generate basic copper diffusion data to demonstrate barrier performance compared to a homogeneous control film. Both experiments are feasible at a university or contract lab within a few months and would materially strengthen the asset's licensing position.