Alkali-free perovskite hafnate and zirconate high-k dielectric fallback for MIM capacitors

The glass-core advanced-packaging substrates portfolio includes a carefully constructed fallback layer designed around alkali-free perovskite hafnates and zirconates — specifically CaHfO3, SrHfO3, CaZrO3, and SrZrO3 in their ground-state orthorhombic Pnma phase — as high-k dielectric candidates for package-integrated metal-insulator-metal (MIM) capacitors. This asset exists precisely because real manufacturing programs need a viable alternative path when a primary material candidate faces supply, integration, or yield constraints. The strategic logic is sound: if the portfolio's preferred Ruddlesden-Popper hafnate route is unavailable or impractical for a given fab process, this filing preserves the right to use an alkali-free, structurally well-characterized, moderate-permittivity family that is computationally validated and compositionally distinct from the crowded HfO2 prior art space. The timing of this filing reflects a broader shift in advanced packaging. As chipmakers push die-to-die interconnect density and on-package capacitance requirements climb into the hundreds of nanofarads per square millimeter, the embedded MIM capacitor has moved from a nice-to-have to a critical element of the substrate stack. Dielectrics that are thermally stable, CMOS-process-compatible, free of mobile alkali ions (which compromise reliability in package environments), and manufacturable as thin films are genuinely scarce. The perovskite hafnate family covered here sits in a useful permittivity window — approximately 30-35 — that clears the threshold for meaningful capacitance density improvement over conventional SiO2 or even HfO2, while avoiding the reliability concerns associated with alkali-containing alternatives such as uncapped lithium aluminate. As a backup filing, this asset is not positioned as the spearhead of the portfolio. Its role is to prevent a competitor from narrowing the prosecution space or designing around the primary claims by selecting one of these compositions. For a licensing conversation, it adds breadth and raises the switching cost for any party that wants to use this compositional family without engaging the portfolio. For an acquirer, it is part of the defensive perimeter that makes the overall package more durable.

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.

The four primary compositions — CaHfO3, SrHfO3, CaZrO3, and SrZrO3 — all adopt an orthorhombic Pnma perovskite structure under ambient conditions. This is their true ground-state symmetry. An important refinement in the computational workflow was correcting earlier screening runs that had inadvertently assessed these materials in higher-symmetry (cubic or tetragonal) hypothetical phases rather than their actual equilibrium structure. The current validated data (designated as the third-generation cross-engine calculation set) reflects the Pnma ground state throughout, which matters because phonon stability assessments and dielectric tensor calculations performed on the wrong crystal symmetry yield unreliable results. Correcting this earlier error is itself a mark of rigor: the pipeline detected the discrepancy across multiple independent calculation engines and flagged it for re-evaluation. The dielectric properties of CaHfO3 in the Pnma phase are the most thoroughly characterized within this family. The total calculated dielectric constant (epsilon_total) is approximately 32, combining both ionic and electronic contributions. This sits meaningfully above the approximately 20-25 range typical for monoclinic or cubic HfO2 in thin-film form, and it is achieved without any alkali element in the A-site — a deliberate design constraint for package reliability. The bandgap of CaHfO3 is calculated at 4.57 eV, which is wide enough to support low leakage current in thin-film dielectric applications and is characteristic of the hafnate family's favorable insulating properties compared to narrower-gap candidates. SrHfO3 and the zirconate analogs (CaZrO3, SrZrO3) are included in the composition family as structurally isoform alternatives with slightly varying permittivity and bandgap characteristics, providing manufacturing latitude depending on precursor availability and film-deposition conditions. The computational validation protocol drew on three independent machine-learning interatomic potentials — MACE, CHGNet, and MatterSim — applied in parallel to the Pnma structures. The outcome was a majority-stable verdict: the preponderance of independent potential evaluations agreed that these structures are dynamically stable, with no imaginary phonon modes that would indicate a thermodynamic driving force toward structural collapse. This is not a unanimous consensus result, which should be stated candidly; one potential's assessment diverged, and this is why the verdict is characterized as majority rather than full consensus. The practical implication is that the Pnma phase stability is computationally well-supported but would benefit from thin-film experimental confirmation — which is identified as the primary open validation gate. Two independent DFT source calculations corroborate the structural energetics. The combination of multi-potential phonon screening and dual DFT sources places this asset on a stronger footing than a single-method screen, while being honest that it has not yet reached the full four-potential consensus that the portfolio's most advanced candidates have achieved. The asset also covers two alkali-aluminate compositions — Li5AlO4 and LiAlO2 — but only in a capped configuration where a Li-blocking interlayer prevents lithium ion migration into adjacent dielectric or metal layers. Uncapped lithium aluminate is explicitly excluded from the claim scope as a negative limitation, reflecting a finding in the negative-results dataset that uncapped lithium-aluminate films present an unacceptable mobile-ion risk in package dielectric stacks. BaHfO3 is acknowledged as compositionally adjacent but is treated as background art rather than a claimed composition, due to a crowded prior-art landscape around that specific compound. These structural exclusions and inclusions reflect deliberate claim engineering rather than arbitrary omissions.

The addressable market for package-integrated MIM capacitors is estimated in the range of $0.5 billion to $1.0 billion annually, with the caveat that these figures are estimates based on current advanced-packaging adoption curves. This market is best understood as a subset of the broader embedded passive component space within high-density organic and glass-core substrates. The relevant buyers are not consumer electronics OEMs directly, but rather the tier-one substrate manufacturers, advanced-packaging foundries (OSATs and IDM packaging divisions), and the MIM capacitor material vendors who supply dielectric films into those fabs. As glass-core substrates scale toward production — driven by hyperscaler and AI-accelerator demand for higher interconnect density and lower signal loss — the embedded MIM market will scale with it. The royalty logic for this asset is straightforward. A licensee producing MIM capacitors using any of the covered compositions would be subject to a per-wafer or per-unit license. The moderate permittivity window (~32) is high enough to deliver competitive capacitance density at manufacturable film thicknesses (tens of nanometers), which is where the economic argument for switching from HfO2 lies. If a substrate maker can achieve a 30-40% improvement in capacitance per unit area using CaHfO3 or SrHfO3 relative to their current HfO2 process, that improvement may justify qualification effort, particularly if the primary portfolio asset is unavailable or if the alkali-free formulation offers a reliability advantage in moisture-sensitive glass-core environments. The business model is licensing rather than direct materials production. There is no urgent race window identified for this specific asset, which is consistent with its role as a fallback rather than a first-mover play. The commercial urgency is driven by the primary portfolio position; this asset matters most if and when a competitor or customer seeks an alternative path. Its value in a portfolio sale or licensing bundle is primarily defensive: it closes a gap that would otherwise allow a workaround into this compositional family.

alkali-free moderate-eps fallback

package-integrated MIM; lithium arms require Li-blocking cap

The most directly motivated buyers or licensees are substrate manufacturers and advanced-packaging foundries who are actively qualifying embedded MIM capacitor processes on glass-core or high-density organic substrates. This includes companies with in-house dielectrics R&D programs — notably major OSATs (Amkor, ASE, JCET), IDM packaging divisions (Intel Foundry Services, Samsung Electro-Mechanics), and glass-substrate developers (Corning, AGC, Schott pursuing glass-core interposer programs). For these players, the asset's value is as a freedom-to-operate-clean fallback that can be pulled forward if their primary dielectric candidate encounters process or reliability issues during qualification. A secondary category of buyer is a MIM capacitor materials supplier — companies that sell precursor chemistries, ALD targets, or dielectric films into the substrate supply chain. For such a supplier, acquiring or licensing this patent family provides the ability to offer a differentiated, alkali-free hafnate dielectric product with protected compositional IP. The asset is also a logical complement to a broader dielectrics patent portfolio held by a semiconductor materials company (for example, one with existing positions in HfO2 or ZrO2 ALD precursors) that wants to extend coverage into the perovskite hafnate adjacency. The modest addressable market size means this is a licensing play or a portfolio-completion acquisition rather than a standalone platform company asset, but within that framing the strategic fit is clear for any party building coverage in next-generation substrate dielectrics.

The primary technical risk is phase stability during thin-film deposition. The Pnma phase is the ground-state structure in bulk, but thin-film deposition by ALD or MOCVD may favor amorphous or metastable phases depending on substrate temperature, precursor chemistry, and film thickness. If the Pnma phase cannot be reliably stabilized in thin-film form without exotic process conditions, the practical utility of the composition claims narrows considerably. This risk is addressable through thin-film coupon work — which is already identified as the open validation gate — and through process chemistry optimization that is well within the capability of a major substrate or materials company. The corrected computational screen on the true Pnma ground state (rather than the earlier erroneous higher-symmetry phase) actually reduces this risk by ensuring that the validated phase is the thermodynamically stable one, not a metastable target. The commercial risk is the backup nature of the asset: its value is contingent on the primary Ruddlesden-Popper hafnate position being commercially active. If the primary portfolio position is licensed or generates revenue, this asset rides along as a breadth element; if the primary position is not commercially active, this asset's standalone revenue potential is modest. The majority-stable (rather than full-consensus) computational verdict is a secondary risk in the sense that a sophisticated technical buyer will notice the imperfect convergence across potentials and may discount the computational proof accordingly — though two corroborating DFT sources and a majority of three independent potentials is a defensible standard for a pre-experimental filing. The roadmap to de-risk is clear: thin-film coupon deposition, dielectric constant and loss tangent measurement, and leakage current characterization at relevant film thicknesses would convert this from a computationally supported backup into an experimentally validated fallback with substantially higher licensing confidence.