High-confidence fluoride subset combining lowest optical dielectric constant and greatest phonon stability for qubit dielectrics

The five-member phonon-margin-hardened fluoride subset — Na3AlF6, K2SiF6, LiSrAlF6, LiCaAlF6, and LiYbAlF6 — represents the highest-confidence starting point within the broader metal-fluoride qubit dielectric materials portfolio for first fabrication and measurement. Where the broader genus spans the full fluoride design space consistent with low optical dielectric constants and acceptable phonon stability, this subset applies a stricter dual filter: optical dielectric constant (eps_inf) constrained to 1.85-2.05 and no phonon mode softer than -0.3 cm^-1. Both thresholds are tighter than the genus requirements, and both must be satisfied simultaneously. The result is a curated shortlist of the genus members carrying the largest combined margin on the two properties most directly linked to two-level-system (TLS) loss suppression in superconducting qubits. The strategic importance of this subset derives from where the field is right now. The DARPA Quantum Benchmarking Initiative downselect running through 2025-2026 is forcing hardware builders to identify specific dielectric materials for loss characterization, not broad candidate classes. A defined, prioritized five-member list with clear physical rationale maps directly onto that mandate. The subset is designed to be the first object of a measured loss-tangent campaign — the experiment that converts projected performance into demonstrated performance and validates the broader portfolio thesis in a single, focused effort.

The dual selection gate applied to this subset reflects the two independent physical mechanisms by which low-loss dielectric behavior is expected to arise in crystalline fluorides. First, a lower optical dielectric constant (eps_inf) directly reduces the electric-field energy density stored in the dielectric layer at the qubit junction, attenuating coupling to any charged or dipolar defect at the interface or in the bulk. The 1.85-2.05 window is tighter than the broader genus range, selecting members with consistently larger field-energy suppression margins. Second, constraining the softest phonon mode to be no lower than -0.3 cm^-1 targets soft-mode TLS channels: near-zero-frequency lattice modes provide a low-energy bath to which two-level systems can couple resonantly. By selecting for structural configurations with no dangerously soft modes, the subset eliminates the most direct phonon-mediated loss pathway. The two filters are complementary and each is physically grounded; their joint application selects members where both suppression mechanisms operate near their maximum effectiveness simultaneously. The five retained species span two well-characterized crystal families. Na3AlF6 adopts the cryolite structure. K2SiF6, LiSrAlF6, LiCaAlF6, and LiYbAlF6 belong to the colquiriite and related fluoroelpasolite structural types (space group SG163 for the colquiriite members). Both crystal families share fluorine-dominated anion frameworks and large bandgaps; the bandgap for representative subset members is approximately 7.7 eV, ensuring transparency across the microwave-to-optical range relevant to qubit readout and control. The wide bandgap also indicates a low density of electronic states near the Fermi level, reducing bulk electronic loss contributions independent of the phonon and dielectric mechanisms. All five compositions are compatible with established thin-film deposition routes — atomic layer deposition, molecular beam epitaxy, and physical vapor deposition — at thicknesses relevant to Josephson junction dielectric layers. One important distinction from other assets in the metal-fluoride qubit dielectric materials portfolio: dynamic stability for these five members was validated through the dual-property screening gate applied to DFT-derived property data, rather than through multi-engine phonon cross-validation. The phonon-margin threshold (-0.3 cm^-1) serves as the stability proxy, and the tightness of that margin is what distinguishes the subset from the broader genus. The multi-engine consensus methodology used elsewhere in the portfolio was not the selection instrument here; instead, the DFT-computed phonon spectrum and optical dielectric constant were the direct filters. This is an appropriate approach for a curated shortlist — the stability selection is embedded in the property gate itself.

The addressable market is the dielectric layer in superconducting qubit devices, estimated at $1-2 billion across device fabrication, licensing, and materials supply as the superconducting quantum computing segment scales toward fault-tolerant operation. The qubit dielectric is a small component by bill-of-materials cost but a gating factor for qubit coherence: TLS loss in the junction dielectric is among the dominant decoherence mechanisms in state-of-the-art transmon and fluxonium architectures, and coherence time directly determines the computational overhead needed for error correction. A material that demonstrably reduces TLS loss at the junction commands pricing power disproportionate to its volume, because it reduces the number of physical qubits needed per logical qubit. The named strategic buyers — IBM Quantum and Google Quantum AI — both operate large superconducting qubit programs with internal materials integration teams and active supplier qualification pipelines. Both are participants in the DARPA QBI evaluation window, where demonstrating dielectric loss below incumbent benchmarks is a formal program milestone. Beyond these two, the customer set extends to the broader superconducting hardware ecosystem: Rigetti, IQM, Oxford Quantum Circuits, and government-funded fabrication nodes all require qualified low-loss junction dielectrics. The commercial pathway is a development license funding measured loss characterization on the five-member subset, structured as an option to an exclusive or field-of-use license upon successful measurement. Royalty logic attaches to qubit node count or wafer area using a licensed material, with tiered rates for volume production. Because this subset is a defined shortlist rather than a broad genus, the licensing transaction is structurally straightforward. The licensee funds a single loss-tangent measurement campaign across five specific compositions, obtains the measured performance data, and exercises the license option on the member or members that clear the threshold. The cost of that campaign is low relative to the value of validated intellectual property in a field where measurement access is the principal bottleneck.

highest combined eps_inf + stability margin -> highest-confidence first-RTP set

narrowed eps_inf + phonon-margin window in qubit-dielectric use

IBM Quantum and Google Quantum AI are the named strategic targets, both operating at the scale and timeline where a validated qubit-dielectric alternative would have immediate program impact. The DARPA QBI downselect window (2025-2026) is the forcing function: both organizations face external milestones that require demonstrated dielectric loss performance, not projected performance. For either buyer, funding a loss-tangent campaign across the five-member subset in exchange for a development license is a tractable transaction — the experimental cost is low relative to the coherence-time impact of a successful result, and the IP position covers the validated outcome. The most aligned transaction structure is an option-to-license agreement where the buyer funds measured loss characterization on the five compositions and exercises the license option if any member clears the target loss threshold. An exclusive license on the subset is the natural preference for a buyer who wants to lock the measured result from competitors; a field-of-use license suits a buyer seeking qubit-specific rights while leaving other applications open. Acquisition of the subset as a standalone asset is also plausible if a buyer wants the IP without ongoing licensing administration, though the broader portfolio context argues for bundling. Beyond the two named majors, second-tier superconducting hardware builders seeking differentiated dielectric IP for government contract qualification represent a secondary buyer pool at lower transaction value but potentially faster timeline.

The primary risk is straightforward: these five compositions have not been measured at millikelvin temperatures in thin-film qubit-device geometry. All computational projections are DFT-derived; the phonon-stability gate was not cross-validated against multiple independent machine-learning potentials as is done elsewhere in the portfolio. If measured loss tangent on the subset members does not outperform amorphous SiO2/AlOx at qubit operating conditions, the subset's commercial premise collapses and the broader portfolio thesis is materially weakened, since this subset was selected precisely as the highest-confidence starting point. The subset concentrates technical risk as much as it concentrates promise. Novelty risk is the second concern. Na3AlF6 and K2SiF6 in particular have established optical and laser literature; claim validity rests on the use limitation and the specific dual-property selection surviving examination scrutiny. A well-funded examiner or inter partes challenger could argue that selecting known fluorides for low dielectric constant is an obvious optimization in the qubit-dielectric context, given prior work on low-eps dielectrics in other semiconductor applications. The tighter dual gate and the specific phonon-margin threshold are the non-obviousness arguments, and they are meaningful, but they have not been tested in prosecution. The roadmap to de-risk both concerns runs through the same action: fund and execute the loss-tangent measurement campaign on the five members. A positive measurement result simultaneously validates the technical premise and provides the strongest possible prosecution evidence of non-obvious, unexpected performance in the claimed use.