Double-perovskite barium scandium tantalate high-k dielectric for package MIM capacitors

Ba2ScTaO6 is an ordered double-perovskite oxide that occupies a deliberate second position within the glass-core advanced-packaging substrates portfolio's high-k dielectric strategy. Its significance lies not in displacing the dominant hafnium-oxide families but in establishing a separately defensible IP vessel with comparable or superior dielectric performance. The compound crystallizes in the rock-salt-ordered cubic structure (space group Fm-3m), with scandium and tantalum alternating on the B-sites of a perovskite lattice balanced by barium on the A-site. This ordering is not incidental; it is the structural origin of the elevated dielectric constant (~47) and is the distinguishing feature that sits in white space relative to the crowded HfO2 patent landscape. The timing logic for this asset is rooted in the substrate industry's need for embedded MIM (metal-insulator-metal) capacitors capable of higher capacitance density as chip packages shrink and on-package power-delivery requirements intensify. Hafnium oxide-based dielectrics dominate this space today, but their patent coverage is dense and licensing is increasingly contested. Double-perovskite A2BB'O6 oxides represent a structurally distinct chemical family with well-understood physics and no meaningful prior art in the package-integrated MIM context. Ba2ScTaO6 was selected as the lead composition from a multi-member series because it clears both independent machine-learning stability screens, carries the highest dielectric constant among the computationally validated members, and anchors a genus that includes backup members—Sr2ScTaO6, Ca2ScTaO6, and Ba2ScNbO6—each independently validated. The portfolio positions this asset honestly: the bandgap of approximately 3.06 eV (from PBE-level DFT) is a disclosed tradeoff. That figure is narrower than HfO2 (~5.7 eV) and means that leakage current at high fields will be a process-engineering concern rather than a solved problem. What is not in question is the structural stability and the dielectric magnitude, both of which are computationally secured. The value of this asset at this stage is as a credible, patent-clean, high-k candidate that gives device integrators and IP portfolios a second lane—distinct from hafnium oxide, distinct from barium titanate MLCCs, and anchored by reproducible multi-method computation.

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.

Ba2ScTaO6 adopts the ordered double-perovskite structure, space group Fm-3m, where the B-site sublattice is fully segregated between Sc3+ (ionic radius ~0.745 Å in octahedral coordination) and Ta5+ (~0.640 Å). This size and charge ordering drives the rock-salt superstructure and suppresses the octahedral tilting instabilities that commonly degrade dielectric performance in disordered or partially ordered perovskites. The result is a high-symmetry cubic unit cell that sustains cooperative ionic polarizability across both B-site species, and the PBE-DFT calculation reproduces a total dielectric constant of approximately 47—substantially above what single-B-site perovskites in this A-site/B-site composition window typically achieve. The bandgap from the same PBE calculation is 3.06 eV. PBE is well known to underestimate bandgaps, so the physical gap is likely somewhat higher, but even a modest HSE06 correction may not push it into the 4+ eV range typical of hafnium oxide. The portfolio discloses this tradeoff explicitly: leakage at elevated operating fields is an open engineering challenge, not a solved one. The stability of Ba2ScTaO6 was assessed using two independent machine-learning interatomic potentials, CHGNet and MACE, applied in a consensus protocol. Both potentials, trained on distinct DFT datasets and using different graph-neural-network architectures, independently concluded that the Fm-3m structure is dynamically stable—meaning no imaginary phonon modes appear across the Brillouin zone. This is a meaningful bar: many candidate high-k oxides fail this test because low-energy polar soft modes or octahedral-tilt instabilities produce imaginary frequencies that signal a ground state different from the proposed structure. Ba2ScTaO6 passes both independent screens without disagreement between them, which substantially raises confidence in the structural assignment before any physical synthesis. The simulations run include CHGNet and MACE potential energy surface evaluations (labeled CE12 in the internal simulation registry) and a PBE gap calculation. Two DFT source records underpin the property values reported. The broader genus—the A2BB'O6 double-perovskite family claimed alongside the lead—was screened in the same framework, and the outcomes were differentiated. Sr2ScTaO6, Ca2ScTaO6, and Ba2ScNbO6 each passed the two-engine consensus stability check and are carried as backup claims within the family. Ba2YTaO6 and Ba2LaTaO6 failed the MACE evaluation, producing unstable phonon predictions from that potential; these are retained in the record as comparative negative examples rather than being quietly dropped, which strengthens the patent disclosure by demonstrating that not all members of the claimed genus are equivalent. This kind of differentiated, adversarially honest computational screening—preserving the failures alongside the successes—is what distinguishes a defensible patent disclosure from a speculative one. The key open validation gate is synthesis of a B-site-ordered, phase-pure coupon. While the computational case for stability is strong, the rock-salt ordering in double perovskites is sensitive to processing conditions—annealing temperature, atmosphere, and cooling rate all influence whether Sc and Ta fully segregate onto alternating B sites or produce a disordered solid solution. A disordered version of Ba2ScTaO6 would likely have a lower dielectric constant and different phonon behavior. Confirming ordering by X-ray or neutron diffraction refinement (looking for the superstructure reflections characteristic of Fm-3m double perovskites), followed by impedance spectroscopy to measure the actual dielectric constant, is the critical experimental step before this asset can be de-risked from computational to experimental proof of concept.

The addressable market for this asset sits within package-integrated passives—specifically embedded MIM capacitors used in advanced semiconductor packaging, including glass-core and organic substrates, interposers, and embedded die platforms. As heterogeneous integration drives package complexity upward, on-package power-delivery networks increasingly require dense, low-profile capacitance placed physically close to logic and memory dies. The total addressable market for high-k dielectric materials in this segment is estimated at $1–5 billion, which encompasses both the dielectric layer materials themselves and the broader embedded-passive module business where dielectric performance is a key differentiator. These are estimates based on the trajectory of the advanced packaging industry and should be understood as such rather than precise figures. The buyers in this market are not consumer end-users but materials suppliers, integrated device manufacturers, and OSAT (outsourced semiconductor assembly and test) companies with embedded-passive roadmaps. MIM and embedded-passive vendors constitute the primary customer channel. These customers typically engage with dielectric innovations through licensing (acquiring rights to use a patented composition in their process flows), through supply agreements (if a synthesis route is specified), or through acquisition of the IP position to exclude competitors. Royalty structures in this segment are typically calculated on processed wafer area or per-module unit volume rather than on raw material weight, which means even a modest market share of advanced package MIM area can translate into meaningful royalty streams. The second-vessel logic matters for licensing economics. Because HfO2-family dielectrics dominate the space, any new packaging program that wants to diversify supplier risk, avoid contested IP, or achieve higher capacitance density through a structurally different material faces limited options today. Ba2ScTaO6 and its validated double-perovskite relatives fill that gap. A licensee in this space is not choosing between Ba2ScTaO6 and HfO2 on technical grounds alone—they are choosing an IP path. The patent-clean status of the double-perovskite A2BB'O6 family in the MIM context is therefore as commercially relevant as the dielectric constant itself.

eps~47, second independent high-k vessel in Group D

package-integrated MIM use; less crowded than HfO2

The most natural acquirers or licensees for this asset are companies with active MIM capacitor programs for advanced semiconductor packaging—specifically those with embedded-passive roadmaps in glass-core or organic substrates, fan-out wafer-level packaging, or 2.5D/3D interposer platforms. This includes materials and process suppliers (ALD and CVD precursor companies, dielectric stack integrators), OSATs expanding into embedded-passive modules, and IDMs or foundries building proprietary embedded-capacitor IP positions. Given the clean FTO and the compositionally distinct nature of the double-perovskite family, a company locked out of the HfO2 ecosystem by licensing constraints would find this particularly valuable as a design-around path with genuine technical merit rather than a purely defensive maneuver. Secondary interest could come from companies building portfolio breadth in high-k dielectrics for defensive or cross-licensing purposes—particularly those that want a second IP family to bring to cross-license negotiations with HfO2-heavy incumbents. The structure of the claims (composition plus device-use in the MIM context) gives a licensee both a process freedom and an offensive position against competitors who might later develop similar double-perovskite MIM dielectrics. The multi-member genus with differentiated computational validation also gives an acquirer room to optimize the composition for their specific process window—substituting Sr or Ca on the A-site, or Nb for Ta on the B'-site—while remaining within the licensed or acquired claim scope.

The primary technical risk is the bandgap limitation. At approximately 3.06 eV from PBE DFT (and plausibly 3.5–4.0 eV after hybrid-functional correction, though this has not been calculated), Ba2ScTaO6 will exhibit higher leakage current density at operating fields than HfO2-based dielectrics. For package-integrated MIM applications where voltage stress is moderate and film thicknesses can be adjusted, this may be manageable, but it is a real constraint that will require process engineering. Quantifying this tradeoff experimentally—measuring leakage as a function of film thickness and operating voltage on actual MIM test structures—is the critical de-risking experiment beyond the phase-purity gate. The second technical risk is ordering fidelity in thin-film deposition: ALD or PVD processes that deposit Ba2ScTaO6 may produce disordered or partially ordered B-site configurations, which would degrade the dielectric constant below the predicted ~47 and could alter the stability behavior. Post-deposition annealing conditions will need to be optimized to achieve the rock-salt B-site order confirmed computationally. The commercial risk is market timing. The advanced packaging dielectric materials market is at an early stage of differentiation; most programs today still use established dielectrics, and new compositions face long qualification timelines (typically 18–36 months from initial process development to production insertion). There is no identified race window in the current data, meaning there is no immediate competitive pressure that would force a rapid decision, but this also means a licensee or acquirer has less urgency to move quickly. The appropriate de-risking roadmap is: (1) synthesize and characterize the B-site-ordered coupon to open the remaining validation gate; (2) measure dielectric constant and leakage on a simple MIM test structure; (3) file and prosecute the composition and device-use claims on the validated timeline; and (4) approach MIM/embedded-passive vendors once experimental data is in hand to support licensing discussions with measured rather than predicted properties.