Sodium hexafluoroaluminate (cryolite) as an ultra-low-dielectric crystalline layer for superconducting qubits

Sodium hexafluoroaluminate (Na3AlF6, cryolite) is the standout member of the metal-fluoride qubit dielectric materials portfolio: it carries the lowest optical dielectric constant among the crystalline network-solid fluoride candidates screened, an epsilon-infinity of 1.871. In superconducting qubit physics, the optical dielectric constant is the quantity that directly sets the participation-ratio-weighted field energy stored in the capacitor dielectric — so the lowest epsilon-infinity translates directly into the lowest projected TLS-induced participation-ratio loss for a fixed capacitor geometry. No other crystalline member in the screened set beats it on this metric. The forced-substitution dynamic is concrete: IBM's publicly stated 2026-2029 roadmap identifies dielectric two-level-system loss as the primary barrier to scaling coherence times, and amorphous aluminum oxide — the incumbent inter-electrode dielectric in shunt capacitors — is the named culprit. A crystalline fluoride that eliminates the disordered TLS bath inherent to amorphous oxides and also minimizes the participation ratio addresses both failure modes simultaneously. The patent whitespace in the qubit-dielectric application of cryolite is confirmed clear; the composition is known only in UV-optics contexts, making the device-use application the protectable novelty. The gap between current database-level proof and filing-grade confidence can be closed with a single first-principles DFPT calculation — the cheapest possible next step for a material of this importance.

Na3AlF6 crystallizes in the monoclinic cryolite structure and is the lead member of the crystalline fluoride set because of one decisive number: an optical dielectric constant (epsilon-infinity) of 1.871, computed by density-functional perturbation theory (DFPT) against the JARVIS-DFT database entry JVASP-20880. The optical dielectric constant, not the static one, is the relevant loss lever here — it governs participation-ratio-weighted field energy under microwave excitation in a qubit capacitor. Every other network-solid member in the screened family sits at a higher epsilon-infinity, meaning more field energy deposited in the lossy region for the same capacitor geometry, and more TLS coupling. Cryolite is therefore the theoretical best-case for coherence within this material class. Supporting the dielectric advantage are two additional properties of practical consequence. The electronic bandgap of 7.20 eV places cryolite firmly in the strongly insulating regime, well above any leakage concern at millikelvin operating temperatures. The DFPT calculation also returns the lowest zone-center phonon at -0.15 cm-1 — a value at the near-zero, dynamically stable boundary. This is not a phonon-consensus verdict from multiple ML interatomic potentials, because cryolite is not a new structural prediction: it is a well-characterized, experimentally known crystalline phase, so multi-engine phonon consensus is not required to establish confidence in the structure itself. The DFPT phonon result confirms there is no soft-mode instability that would destabilize the lattice at cryogenic temperatures, and the near-zero value is consistent with suppressed soft-mode TLS channels relative to amorphous alternatives. Bulk modulus of 56 GPa is lower than the harder rutile-structured family members such as MgF2 (~101 GPa), which is a process consideration — cryolite films will tolerate less intrinsic stress — but is not a loss-relevant property. The proposed integration pathway is atomic-layer deposition onto a niobium ground plane to form the inter-electrode dielectric of a shunt capacitor. ALD chemistry for fluoride films on metal substrates has been demonstrated in the literature, giving this deposition route practical credibility. The monoclinic cryolite structure is architecturally well-suited to this role: unlike amorphous AlOx, a crystalline cryolite film eliminates the structural disorder that produces the dense TLS bath responsible for energy relaxation in the state-of-the-art devices.

The addressable market for qubit dielectric layers in superconducting quantum computing is estimated at $1-2 billion, a figure that encompasses the licensing value of a protected dielectric material applied across commercial qubit fabrication lines at IBM Quantum, Google Quantum AI, and their competitive peers. The value-capture mechanism is a field-of-use license tied to the qubit-dielectric application of cryolite, priced per qubit node, per wafer, or as a milestone-structured development agreement. Because the dielectric layer represents a small fraction of total device cost but exerts disproportionate leverage on T1 coherence times — the primary commercial differentiator among qubit platforms — the royalty logic is defensible against cost-of-goods arguments. Cryolite is abundant and commercially produced at scale, primarily as a flux in aluminum smelting. This is a double-edged commercial fact: deposition precursors and raw material are not supply-chain risks, and adoption cost is low, which strengthens the licensing case. A builder adopting Na3AlF6 for a shunt-capacitor dielectric captures a coherence benefit that compounds across every qubit on every chip, making even a modest per-node license economically attractive against the R&D cost of independently discovering and characterizing an equivalent material. The race window is tied to IBM's 2026-2029 scaling roadmap, where dielectric TLS has been publicly named as a coherence ceiling. This timeline creates urgency: a buyer that secures rights to the qubit-dielectric use of cryolite before that window closes can incorporate the material into process development cycles running now. Waiting for a competitor to independently file on cryolite's qubit-dielectric use — a use that carries no existing patent coverage — would mean paying to license from an adversary or engineering around a material that is physically optimal for the application.

lowest eps_inf (1.871) of the network-solid members -> lowest participation-ratio loss

qubit-dielectric use of cryolite; computationally-predicted in this use; composition old for UV-optics use only

IBM Quantum and Google Quantum AI are the named primary buyers, both under pressure from IBM's 2026-2029 coherence roadmap where dielectric TLS is the publicly acknowledged blocker. As the lowest-epsilon-infinity network-solid member in a pre-screened, patent-protected set, Na3AlF6 is well suited to a strategic buyer that wants the best-case single demonstrator for a coherence push and is willing to pay for exclusivity in the qubit-dielectric field of use. An exclusive license or outright acquisition of the qubit-dielectric rights to this member denies the marquee material to competitors during the critical 2026-2029 process development window. A non-exclusive license is also viable for a buyer less concerned with competitive lockout, particularly if they intend to combine cryolite with other members of the metal-fluoride qubit dielectric materials portfolio in a broader dielectric optimization program. The most capital-efficient deal structure for a buyer uncertain about the experimental outcome is a paid option that funds the DFPT tensor confirmation and the ALD coupon measurement, converting to a full license upon positive Q-factor data. This aligns the seller's validation cost with the buyer's diligence requirements and gets both parties to a decision point at minimal committed spend.

The primary technical risk is that all proof is currently computational, not measured. The epsilon-infinity of 1.871 and the Q-factor advantage it implies are projected from DFPT; no cryolite qubit capacitor has been fabricated or measured. The gap between a projected participation-ratio reduction and a confirmed T1 improvement is real, and ALD cryolite process development may surface film-quality, adhesion, or crystallinity challenges not visible in a DFT screen. The modest bulk modulus of 56 GPa also means cryolite films will be more sensitive to intrinsic process stress than harder rutile-structured alternatives, which could require process tuning before reliable thin-film deposition is achieved. The primary legal risk is that validity rests on the use limitation over UV-optics prior art. If an examiner or post-grant challenger reads the UV-optics art broadly, the use distinction will need to be defended with specific claim language and experimental data supporting the qubit-dielectric selection rationale. The DFPT confirmation and the measured Q-factor coupon are therefore both de-risking steps for legal and technical purposes simultaneously. The recommended roadmap is sequential: DFPT tensor confirmation first (low cost, upgrades proof grade), then ALD coupon with Q-factor measurement against an AlOx control (higher cost, closes both the technical and legal validation gap), then file to priority with confirmed data in hand.