Aluminum borate (AlBO3) dielectric for through-silicon-via barriers and copper diffusion barriers

Aluminum borate — AlBO3 — occupies a peculiar gap in the advanced-packaging dielectric landscape: it is a wide-bandgap oxide that combines aluminum's well-understood ALD chemistry with boron's tendency to suppress crystallization and reduce leakage pathways, yet it has been essentially invisible in production barrier stacks. The PFAS-free dielectric and process fluids portfolio addresses the urgent replacement mandate in semiconductor manufacturing where traditional barrier and liner chemistries face regulatory and reliability pressure at 3D integration pitches below 10 micrometers. AlBO3, with a computed bandgap in the 5.8–6.1 eV range, offers electrical isolation margins competitive with silicon nitride while the amorphous-favoring boron content can reduce grain-boundary diffusion — the primary failure mode for copper barrier dielectrics in through-silicon-via (TSV) and redistribution-layer (RDL) applications. The commercial timing is driven by a structural transition in the semiconductor packaging industry. The move to chiplet architectures, heterogeneous integration, and high-bandwidth-memory stacks is forcing TSV density and aspect ratios upward, which in turn shrinks the permissible liner/barrier thickness budget without relaxing leakage or copper-diffusion requirements. Incumbent silicon nitride and alumina barrier films are reaching physical limits in conformality and pinhole density at sub-micron TSV diameters. An aluminoborate composition that can be deposited in single-layer, bilayer, or graded-stack form — spanning 5 to 200 nm — and that demonstrably blocks copper diffusion while maintaining high electrical isolation is a candidate for licensing into any major OSATs or IDM process-development roadmap. This asset is best understood as a composition-plus-device-use claim covering three thin-film stack configurations: a single-layer AlBO3 film, an AlBO3/Al2O3 bilayer, and an AlBO3/aluminum-rich graded stack. The patent family's breadth across configurations is its core strength, providing coverage whether a process engineer optimizes for the purest aluminoborate stoichiometry, for interface compatibility with downstream alumina, or for graded diffusion-resistance profiles that improve step coverage in high-aspect-ratio vias.

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.

AlBO3 is a ternary oxide with structural sources documented across both the Materials Project (mp-8110) and the JARVIS database, giving it two independent DFT-derived structural pedigrees — an unusual degree of cross-validation for a material at this development stage. The computed electronic bandgap sits at approximately 5.9 eV (with a range of 5.8–6.1 eV depending on functional and structure variant), placing it firmly in the wide-gap insulator class. For context, silicon dioxide sits near 9 eV and silicon nitride near 5 eV, so AlBO3 resides in a practically useful band — large enough for strong leakage suppression at device-relevant electric fields, not so wide as to create deposition-chemistry challenges or thermal instability. The computational validation pipeline applied to this material includes structural relaxation with three independent machine-learning interatomic potentials, all three agreeing on the relaxed geometry. This three-of-three consensus on structural relaxation is meaningful because divergence among potentials — a MACE prediction differing from a CHGNet or ORB prediction — is a common early warning sign that a structure lies in a poorly characterized region of configuration space where DFT training data is sparse and computed properties are unreliable. The agreement here indicates the structure is in a well-characterized compositional neighborhood. The dynamic (phonon) stability question is explicitly held open: finite-displacement phonon calculations have been identified as the next required gate, meaning confirmation that no imaginary phonon modes exist has not yet been completed. This is an honest limitation — the material relaxes stably and the potentials agree, but the full phonon density of states has not been confirmed clean. A follow-on Cu-barrier coupon test (thin-film deposition and SIMS or TEM diffusion characterization) is the corresponding experimental gate. The three stack configurations covered — single-layer AlBO3, AlBO3/Al2O3 bilayer, and AlBO3/aluminum-rich graded stack — reflect real engineering tradeoffs. Pure AlBO3 offers the highest boron content and the greatest tendency toward amorphous microstructure, which suppresses grain-boundary copper diffusion pathways. The bilayer configuration with Al2O3 allows the process engineer to tune interfacial adhesion and use existing alumina ALD process knowledge while adding the aluminoborate isolation layer. The graded Al-rich stack is the most processing-intensive but offers the best prospect for continuous composition tuning to match thermal expansion coefficients across the TSV stack, reducing stress-induced delamination risk. The 5–200 nm thickness range in the claim spans both ultrathin seed layers and thicker liner applications, giving the claim meaningful device-form-factor breadth. From a materials-science standpoint, the boron in AlBO3 serves two purposes simultaneously: it raises the crystallization temperature relative to pure Al2O3, keeping the film amorphous through standard BEOL thermal budgets (typically capped near 400°C), and it introduces B–O network-former character that can improve film density when deposited by ALD. Denser, more amorphous barrier films consistently outperform their crystalline or porous counterparts in copper ion blocking, which is the fundamental reliability metric for this application class. The dielectric constant of aluminoborate compositions is expected to lie between Al2O3 (approximately 9) and amorphous boron oxide (approximately 3–4), producing a moderate-k film that reduces parasitic capacitance in dense TSV arrays compared to pure alumina — an ancillary benefit that matters for signal integrity in high-bandwidth stacks.

The serviceable market for TSV barrier and copper diffusion barrier dielectrics is a subset of the advanced semiconductor packaging materials segment. Estimates for the addressable portion — thin-film dielectric liner and barrier materials consumed in TSV and RDL fabrication across OSATs, IDMs, and substrate makers — suggest a range of $200–500 million per year in direct material and process-chemical value, with the upper bound reflecting continued growth in chiplet and 3D-IC adoption through the late 2020s. These are estimates derived from industry sizing of the 3D packaging consumables market; actual contract values would depend on deposition-tool ecosystem adoption and royalty-rate negotiations. The royalty logic for a barrier dielectric is typically a per-wafer or per-via license tied to the process step, making the capture mechanism relatively clean compared to materials-only licensing. The buyers of this technology are process engineers and materials procurement groups at high-volume assembly and test (OSAT) companies — including the major Taiwan, South Korea, and Southeast Asia-based players — and at IDMs with internal advanced packaging fabs such as Intel, Samsung, and Micron. The immediate use case is qualifying a new barrier dielectric in the TSV liner process, which is a gate-qualified step in any packaging roadmap and therefore subject to long qualification cycles but also to strong switching-cost lock-in once adopted. RDL dielectric applications follow a similar procurement pattern. A secondary buyer class is deposition-equipment companies and process-chemistry suppliers who would license the composition to co-develop an ALD or CVD precursor system around it, packaging the IP with a process recipe for sale to fabs. Licensing logic for this asset is best structured as a composition-plus-process license bundled with the full thin-film stack family. The three stack configurations (single-layer, bilayer, graded) allow a licensee to pick the configuration most compatible with their existing process platform, while the IP holder retains coverage of all three variants. A per-wafer royalty in the range typical for specialty dielectric barrier materials would be standard; alternatively, a paid-up license at acquisition would be attractive to a strategic buyer seeking to control the IP before a competitor licenses it.

wide-gap aluminoborate TSV/Cu-barrier dielectric

thin-film TSV/Cu-barrier dielectric stack form factor

The primary strategic acquirers or licensees for this asset are advanced packaging materials companies and specialty ALD precursor suppliers seeking to expand their dielectric barrier portfolio ahead of the TSV scaling inflection. Companies such as Entegris, Merck KGaA's semiconductor materials division, and Versum Materials (now part of Merck) routinely license novel dielectric compositions to bundle with ALD precursor chemistry offerings. OSATs with internal materials R&D groups — ASE, Amkor, JCET — are secondary targets, as they have direct process-qualification capability and motivation to differentiate on proprietary barrier liner chemistry. IDMs with advanced packaging fabs (Intel Foundry Services, Samsung Semiconductor, SK Hynix) represent the highest-value single-licensee scenario, where a paid-up exclusive license could be bundled into a broader process-IP acquisition. A defensive acquisition scenario is also plausible: a large alumina or silicon nitride barrier supplier could acquire this asset to prevent a competitor from licensing it and undercutting their incumbent position. In either scenario — offensive licensing or defensive acquisition — the asset's value is amplified by its position within the broader PFAS-free dielectric and process fluids portfolio, which provides a thematic rationale for a buyer seeking to assemble a coherent IP position in next-generation packaging dielectrics rather than acquiring isolated claims.

The most material risk is the open phonon-stability gate. If finite-displacement calculations reveal imaginary phonon modes in the AlBO3 structure, the static relaxation consensus loses much of its probative value, and the material's practical stability under deposition and operating conditions would require a different structural polymorph or composition adjustment. This is a binary risk that can be resolved relatively quickly with standard DFT phonon calculations. The second risk is deposition chemistry: AlBO3 does not have a mature ALD precursor system analogous to the trimethylaluminum/water chemistry used for Al2O3. Identifying or synthesizing suitable aluminum-boron co-precursors that produce stoichiometric aluminoborate films at BEOL-compatible temperatures is a process-chemistry development task that falls outside the computational work completed to date and could require 12–24 months of experimental effort. The Cu-barrier coupon test is the gating experiment for de-risking the core performance claim. The roadmap to de-risk these gaps is well-defined. Finite-displacement phonon calculations on the mp-8110 structure can be completed within weeks using existing DFT infrastructure. Precursor scouting can run in parallel, drawing on published atomic-layer-deposition literature for borate-containing oxides. If both gates clear, the asset transitions from computational candidacy to experimentally supported composition, which would materially increase its licensing value and the confidence of any FTO-based freedom-to-operate opinion sought by a prospective acquirer.