Tiled boron arsenide heat-spreader plate for advanced semiconductor packaging

Boron arsenide (BAs) in the cubic zincblende phase is, by a wide margin, the highest-thermal-conductivity semiconductor known outside diamond, with single-crystal thermal conductivity in the range of 1,000 to 1,300 W/m·K — values that rival or exceed natural diamond depending on isotopic purity and measurement technique. That figure sits roughly 3 to 5 times above silicon carbide, 5 to 10 times above aluminum nitride, and an order of magnitude above copper. For two decades, BAs sat as a theoretical curiosity because growing defect-free single crystals at scale proved extremely difficult. The past several years changed that: three independent laboratory groups synthesized measurable BAs crystals and confirmed the anomalous thermal conductivity experimentally, igniting a race among semiconductor packaging engineers to find a manufacturable form factor for real devices. The timing is acutely strategic. AI accelerators — the GPUs and custom ASICs that underpin large-model training and inference — are now routinely dissipating 500 to 1,000 W per chip, and thermal management has become a first-order design constraint. Diamond heat spreaders can handle this thermal load but remain prohibitively expensive and geometrically constrained at the wafer-panel scales that advanced packaging demands. The invention in this family occupies precisely the whitespace that opens when monolithic single-crystal BAs is blocked by existing IP: a tiled or mosaic-bonded multi-crystal BAs plate, assembled from smaller single-crystal tiles into a larger working area (≥5 mm edge), producing a heat-spreader plate rather than a single-crystal thermal interface member. That distinction is not merely legalistic — tiled mosaic architectures are also the practical near-term path to manufacturable large-area BAs components, since growing defect-free monolithic crystals above a few millimeters has not been demonstrated at commercial yield. The asset sits within the catalysts and energy-conversion materials portfolio as a lead filing. It is a composition-plus-device-use claim, not a pure method patent, which gives it the broadest downstream enforceability against the supply chain of OSATs (outsourced semiconductor assembly and test houses), substrate suppliers, and module integrators who will ultimately incorporate a BAs heat spreader into a packaged device. The forced-substitution dynamic — diamond prices, thermal budgets climbing, and the monolithic BAs route blocked — creates a natural urgency window for a licensee or acquirer to secure this alternative architecture before the art fully matures.

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.

Cubic boron arsenide adopts the zincblende crystal structure (space group F-43m), isostructural with GaAs and InP but with dramatically different phonon physics. The anomalous thermal conductivity arises from a near-perfect cancellation of the three-phonon scattering channels that normally limit conductivity in polar semiconductors: the acoustic phonon branches in BAs are unusually bunched together, suppressing the dominant normal and Umklapp scattering pathways that dominate in, for example, GaN or SiC. This bunching simultaneously opens a phonon-frequency gap that blocks acoustic-optical phonon scattering. The result is that heat conduction in BAs is dominated by high-frequency acoustic modes with anomalously long mean free paths, yielding intrinsic lattice thermal conductivity values of approximately 1,000 to 1,300 W/m·K by phono3py calculations at the level of the CE14 force-constant extraction — a range independently corroborated by two DFT source calculations and confirmed by published experimental measurements on small laboratory crystals. The electronic structure is equally notable. HSE06 hybrid-functional calculations place the bandgap at 2.36 eV, confirming BAs as a wide-bandgap semiconductor (and validating the calculation pipeline against the 6.31 eV gap of cubic boron nitride, which serves as the internal reference). The 2.36 eV gap means BAs is electrically insulating enough for direct-bonded heat-spreader applications in power electronics and optoelectronics, where a conductive spreader would create short-circuit paths. This is a practical advantage over copper, which is an excellent thermal conductor but must be isolated by dielectric layers that add interface resistance. The combination of ultra-high thermal conductivity and moderate wide bandgap is essentially unique among non-diamond materials. Dynamic (phonon) stability of cubic BAs has been validated by two independent machine-learning interatomic potentials: both MACE and CHGNet, evaluated independently on the BAs crystal structure, return positive phonon frequencies across the full Brillouin zone with no imaginary modes. Imaginary modes would signal a dynamically unstable structure — one that would spontaneously distort rather than remain cubic at finite temperature — and their absence in both potentials is a meaningful corroboration of the crystal's physical viability. Cubic boron nitride serves as a backup composition in the same family, and boron phosphide (BP) was separately evaluated and shows agreement at +0.97 THz, providing a hierarchy of backup candidates should BAs prove commercially intractable. The tiled/mosaic plate form factor carries its own engineering rationale beyond the patent carve-out: individual tile thicknesses of 50 to 1,000 microns can be grown by chemical vapor transport at manageable defect densities, diced, inspected for quality, and bonded into a larger mosaic with thin metal or bonding-glass interlayers. The effective thermal conductivity of the composite plate will be somewhat below the single-crystal theoretical value due to interface resistance at tile boundaries — quantifying this phonon boundary scattering is among the open validation gates — but modeling suggests the in-plane conductivity of a well-bonded mosaic remains competitive with diamond for spreader geometries where the dominant heat flow is lateral. Thermal boundary resistance (Kapitza resistance) between the BAs tiles and the bonding interlayer is a critical design variable. At 1,000 W/m·K bulk conductivity, even a modest interface resistance of 1×10⁻⁸ m²·K/W adds meaningful thermal resistance across a thin tile stack. Managing this through surface preparation, bonding material selection, and tile thickness optimization is an engineering problem that is well-scoped and solvable with existing thin-film bonding technology, but it has not yet been demonstrated in hardware at the claimed plate dimensions. The three open validation gates — large-area monolithic crystal growth (included as prophetic), tiled-plate bonding demonstration, and time-domain thermoreflectance (TDTR) kappa measurement on a physical coupon (Prophetic Example 15) — are all achievable with current laboratory infrastructure and represent the natural next phase of development for a licensee or acquirer with crystal-growth capability.

The addressable market for advanced thermal management in semiconductor packaging spans multiple high-growth segments. At the top of the stack, AI accelerator packaging is consuming an increasing fraction of the $30–40 billion advanced packaging market, with per-chip thermal budgets that have roughly doubled every two to three years. Thermal interface materials, heat spreaders, and integrated heat spreaders (IHS) for leading-edge GPU and custom-ASIC packages are currently supplied in diamond (for the highest-performance applications), AlN ceramics, and pyrolytic graphite composites. Industry estimates for the high-performance thermal management sub-segment — including spreaders, vapor chambers, and interface materials for data-center AI chips — put the relevant market at approximately $2 to 5 billion by the late 2020s. This is an estimate, not a certified market study, and the actual figure will depend on the rate at which AI accelerator unit volumes scale and on the degree to which thermal limits constrain packaging choices. The customer base divides into two main segments. First, outsourced semiconductor assembly and test houses (OSATs) — companies like ASE Group, Amkor, and their competitors — that physically package AI accelerators and power devices for hyperscaler customers and are directly responsible for sourcing and integrating heat-spreader components. These companies face growing contractual thermal-performance obligations and would benefit from a manufacturable non-diamond spreader option. Second, power-electronics makers in automotive (SiC and GaN-based inverters for electric vehicles), industrial motor drives, and RF power amplifiers, where junction temperatures in wide-bandgap transistors are rising and copper or AlN substrates are increasingly marginal. A BAs heat spreader in these applications could directly extend device lifetimes or enable higher power densities. The licensing and royalty logic for this asset is most naturally structured as a per-unit royalty on heat-spreader plates incorporating tiled BAs (or the backup compositions in the family), combined with a materials supply arrangement since BAs crystal production is not yet commoditized. A licensee that controls BAs crystal growth would have significant process know-how that complements the patent protection. A per-plate royalty of even $5 to $20 on high-performance AI accelerator packages — where the bill of materials for thermal management already runs into hundreds of dollars — would represent an extremely small friction on the value chain while capturing meaningful royalty streams at volume. Alternatively, a strategic acquirer with an existing advanced packaging or thermal materials business could use the IP as a defensive moat alongside a first-mover manufacturing investment.

highest-kappa non-diamond heat spreader for AI/power thermal budgets

tiled/mosaic-bonded multi-crystal plate >=5 mm/edge; monolithic single-crystal TIM excluded

The most natural acquirers and licensees are companies at the intersection of crystal-growth capability and advanced packaging thermal management. Crystal-growth specialists with existing CVD or chemical vapor transport infrastructure for compound semiconductors — including wide-bandgap materials companies that already produce silicon carbide or gallium nitride substrates — are the most direct technology match, because BAs growth by chemical vapor transport is a natural extension of their process knowledge. An acquirer in this category would gain both the IP protection and the first-mover position in what is likely to be an increasingly competitive BAs supply market. Diamond heat-spreader manufacturers represent a second category: companies currently supplying CVD diamond for RF and military applications would see BAs as either a complementary product line or a hedge against the price erosion that will come if BAs scales, and acquiring this IP defensively would make strategic sense. On the systems integration side, leading OSATs with advanced packaging programs — particularly those developing heterogeneous integration for AI accelerators — are directly exposed to the thermal management constraint this invention addresses. A licensing arrangement with one or two leading OSATs, structured as an exclusive or semi-exclusive field-of-use license for AI accelerator packaging, would be the most straightforward commercialization path without requiring the IP holder to operate a crystal-growth business. Power semiconductor module makers (automotive SiC inverter suppliers, GaN power amplifier companies) represent a secondary customer segment that could justify a separate non-exclusive licensing track. The urgency factor is real: as BAs crystal growth becomes more mature and the thermal management crisis in AI packaging intensifies, the window to secure a differentiated IP position on the mosaic architecture will narrow.

The primary technical risk is that BAs single-crystal growth has not yet been demonstrated at the tile sizes required to assemble a 5 mm mosaic plate with commercially acceptable yield. Chemical vapor transport of BAs is slow, the arsenic chemistry is hazardous and requires specialized containment, and the defect density in as-grown crystals is currently high enough that only small (sub-millimeter to few-millimeter) specimens of adequate quality are routinely reported. If tile yield at the required dimensions remains low, the economics of mosaic assembly could make BAs plates non-competitive with diamond even if the thermal performance is superior. This risk is mitigated in part by the fact that the 50 to 1,000 micron tile thickness range and the mosaic architecture are specifically designed to work with smaller crystals, and that crystal-growth technology for compound semiconductors has historically improved rapidly once industrial investment is applied. The IP risk is that the freedom-to-operate position is characterized as narrow, meaning adjacent bonding, integration, or packaging methods may carry third-party IP exposure that a commercializing partner would need to navigate. The prior-art landscape is moving quickly: with multiple academic groups publishing on BAs and the blocking monolithic-crystal patent already issued, any delay in prosecution or in reducing the prophetic claims to practice increases the risk that the whitespace closes further. The roadmap to de-risk follows naturally from the open validation gates: a development partner should prioritize the tiled-plate bonding demonstration and TDTR coupon measurement first, as these are the claims most at risk of being challenged as prophetic, and then build the manufacturing yield argument around the mosaic economics. Regulatory risk is low — BAs is an inorganic semiconductor, not a novel pharmaceutical — but the arsenic content creates occupational health and supply chain handling requirements that are worth addressing in a partnership structure.