Method of reducing two-level-system dielectric loss in superconducting qubits using crystalline fluoride layers

Superconducting qubit coherence is fundamentally limited by two-level-system (TLS) loss — quantum noise arising from defect states at dielectric interfaces. The dominant commercial stack today uses amorphous oxide dielectrics, and those amorphous films are structurally disordered by nature: they harbor soft phonon modes, dangling bonds, and a broad distribution of two-level fluctuators that couple to the qubit electric field and drain coherence energy. This method patent claims the engineered exit from that trap. By selecting a crystalline fluoride dielectric on the basis of rigorously computed properties — an optical dielectric constant between 1.7 and 2.15, an electronic bandgap of 6 eV or greater, and a lowest phonon frequency no worse than −5 cm⁻¹ — then depositing it in crystalline form on a superconducting electrode and operating the resulting qubit at millikelvin temperatures, the process simultaneously attacks TLS loss through two independent physical mechanisms. The claim is thus not a material claim but a process right: it captures the entire workflow of how a manufacturer selects, deposits, and operates a fluoride dielectric to achieve reduced TLS loss, regardless of which specific fluoride member they choose. The commercial timing is sharp. DARPA's Quantum Benchmarking Initiative (QBI), running through 2025–2026, is directly incentivizing hardware performers — including university groups, national labs, and tier-one quantum computing companies — to demonstrate measurable improvements in qubit coherence metrics. That creates a forced-substitution window: programs that committed to amorphous-oxide baselines years ago now face competitive pressure to adopt low-loss dielectric processes. A method patent that covers the logical selection-and-deposition workflow for the crystalline fluoride family is positioned precisely at the point where the field must make a process change. The addressable population of licensees includes IBM Quantum, Google Quantum AI, AWS Center for Quantum Computing, and the broader DARPA QBI performer base, each of which is actively engineering the dielectric stack of transmon and related qubit architectures.

The physical mechanism underlying this method is best understood through the participation-ratio framework standard in superconducting qubit engineering. The participation ratio quantifies the fraction of the qubit's total electric-field energy that is stored within a given dielectric volume or interface. A higher participation ratio in a lossy dielectric translates directly into a shorter qubit coherence time T1. The optical dielectric constant ε∞ controls how tightly field lines are confined in a dielectric — a material with ε∞ in the 1.7–2.15 range (as specified in the method) concentrates less field energy than conventional amorphous silicon oxide or silicon nitride layers (ε∞ ~ 3.9–7), thereby lowering the effective participation ratio and reducing the coupling of the qubit to any residual loss channels. The bandgap requirement of 6 eV or greater ensures the dielectric is electronically wide-gap and thus electronically transparent at the photon energies and voltages relevant to qubit operation, avoiding leakage and photon-absorption loss pathways. The second, independent loss-suppression mechanism addresses soft-mode TLS channels directly. In amorphous dielectrics, certain atomic configurations can tunnel between nearby energy minima — these are the classical TLS centers. The tunneling rate and density of states are directly linked to the presence of low-frequency phonon modes, including imaginary or near-zero modes that correspond to nearly-flat energy landscapes in structural configuration space. The method's phonon stability criterion — lowest phonon frequency no worse than −5 cm⁻¹ — enforces a dynamically stiff lattice. A crystalline fluoride satisfying this criterion has a well-defined phonon spectrum with a finite acoustic gap or at minimum no strongly anharmonic soft modes, which structurally suppresses the tunneling channels that manifest as TLS loss at millikelvin. This is categorically different from an amorphous oxide of the same bulk composition: the crystalline order eliminates the configurational disorder reservoir entirely, not merely reducing its density. The computational work behind the fluoride genus that defines the scope of this method was performed using the Lattice Graph materials knowledge graph. Candidate fluorides were screened computationally for the target property window (ε∞, Eg, phonon stability), and dynamic stability was validated using multiple independent machine-learning interatomic potentials from the MACE, CHGNet, MatterSim, and ORB families — with consensus across potentials required before a candidate was considered stable. This multi-potential consensus approach guards against artifacts from any single potential's training set and provides a materially stronger stability assertion than single-point DFT or single-MLIP screening. Dielectric-tensor calculations using density functional perturbation theory (DFPT) were used to compute the optical dielectric constants and confirm the ε∞ 1.7–2.15 window across the fluoride genus. The net result is that the material selection criteria embedded in the method claim are backed by a defensible computational provenance, not merely by theoretical argument. The millikelvin operating condition specified in the method is not a formality — it is mechanistically load-bearing. TLS loss mechanisms in superconducting qubits are temperature-dependent: at millikelvin temperatures (typically 10–30 mK in dilution refrigerators), thermally activated TLS transitions are frozen out and the quantum coherence regime is reached. The method's explicit recitation of millikelvin operation ties the claim to the physically relevant operating regime, distinguishing it from room-temperature deposition processes for unrelated applications. The combination of low-ε∞ participation-ratio reduction and crystalline phonon-stability-based TLS suppression, realized through the specified selection and deposition steps and confirmed at millikelvin, constitutes a two-lever approach to the dominant loss mechanism in present-generation superconducting qubits.

The addressable market for this method is the superconducting quantum computing hardware sector. Conservative estimates for the total dielectric engineering and hardware supply market within superconducting quantum computing — spanning process development, equipment, and licensing — are in the $1–2 billion range as the field matures toward fault-tolerant processors over the 2025–2030 horizon. This figure reflects the capital being deployed by IBM Quantum, Google Quantum AI, Amazon Web Services' quantum hardware group, and a growing ecosystem of DARPA-sponsored university and national-laboratory performers. The qubit dielectric stack is a known limiting factor in every transmon and related superconducting qubit architecture, making dielectric process IP broadly relevant across essentially all superconducting hardware programs, regardless of specific qubit geometry or coupling topology. The licensing logic for a method-of-use claim of this structure is straightforward: any entity that selects a crystalline fluoride from the covered property window, deposits it on a superconducting electrode, and operates the resulting device at millikelvin is practicing the claimed method. This creates a process license opportunity that applies to both internal manufacturing (IBM, Google, AWS) and to foundry-style providers who offer qubit fabrication services. The royalty basis could be per-wafer, per-device, or a running royalty on quantum computing service revenue denominated by qubit count or coherence-time milestone — all of which are becoming standard metrics in quantum computing service agreements. Given the DARPA QBI 2025–2026 window, the most immediate near-term revenue pathway is a research license or co-development agreement with a QBI performer seeking to differentiate their coherence metrics in the benchmarking period. The forced-substitution dynamic is already underway at the process level. The quantum computing industry has been tracking the T1 ceiling imposed by amorphous-oxide loss for roughly five years, and multiple groups have independently begun exploring crystalline and epitaxial dielectric alternatives. A method patent that covers the selection-and-deposition workflow for the fluoride class — with a specific, computed property window — converts that exploratory interest into a licensing conversation. The royalty leverage is greatest during the hardware development phase (now through approximately 2028), before a single process architecture becomes fully entrenched across the major platforms.

claims the loss-reduction method as a defensible process asset over the qubit-dielectric use

method-of-use grounded in physical deposition + mK operation steps; FTO leg 1

The most natural acquirers or licensees for this method patent are the tier-one superconducting quantum computing platforms: IBM Quantum (which runs one of the world's largest transmon qubit programs and is systematically engineering each loss channel in its hardware roadmap), Google Quantum AI (which has demonstrated surface-code error correction and is pursuing the dielectric engineering required for fault-tolerant scaling), and AWS Center for Quantum Computing (which is building its own superconducting hardware program in parallel with its cloud access business). Each of these organizations has active process-engineering teams focused precisely on the dielectric stack problem this method addresses. The DARPA QBI performer community — a broader set of university groups and smaller companies — represents a second tier of potential licensees who may seek research licenses to credibly incorporate fluoride dielectric processes into their benchmarked coherence demonstrations through 2026. Strategic fit is strongest for any acquirer that is both (a) committed to superconducting qubit hardware as a long-term platform and (b) not already locked into a crystalline fluoride process that predates this filing. For an acquirer at that position, acquiring the method claim preemptively removes a potential blocking position against their own future fluoride dielectric development, while simultaneously providing a licensing asset against competitors who move in the same direction. A defensive acquisition by one of the major platforms would also have the effect of clearing the FTO landscape for their own process R&D. The patent family context — part of a broader metal-fluoride qubit dielectric materials portfolio — makes this most naturally a portfolio-level transaction rather than a single-asset acquisition, though the method claim can be structured for standalone licensing.

The most significant risk is the open experimental validation gate: the method's core claim has strong computational and theoretical support but has not yet been demonstrated in a fabricated device with measured T1 or loss-tangent data against an amorphous-oxide control. This limits the claim's leverage in licensing negotiations with technically sophisticated counterparties who will ask for device data before agreeing to royalty terms. The path to de-risking this is a focused fabrication experiment — ALD or MBE deposition of a qualifying fluoride on a standard transmon test structure, followed by millikelvin microwave characterization — which is achievable within a 12–18 month timeline at a university quantum fabrication facility or a national lab with appropriate process capabilities. A second risk is the pace of independent discovery: the crystalline fluoride direction is directionally visible in the academic literature, and a competitor with device-fabrication infrastructure could in principle generate enabling prior art or make their own patent filings before experimental validation is complete. The DARPA QBI 2025–2026 window makes the timing acute. Prioritizing the fabrication experiment and filing continuation claims tied to specific measured device results would substantially harden the IP position and transform this from a computationally grounded method claim into a fully corroborated one.