Purcell filters, coupler CZ, and vdW qubits (09/15/26)

TL;DR A broadband Purcell filter—circuitry that blocks qubit decay into a lossy readout/reset environment (the Purcell effect)—decouples those paths so each can be optimized alone: Purcell-limited relaxation times $T_1>1\mathrm{ms}$ over a $1.2\mathrm{GHz}$ band, $500\mathrm{ns}$ readout without a Josephson parametric amplifier, and $100\mathrm{ns}$ reset at $99.6\%$ efficiency. A dedicated reset resonator with a deliberate $\kappa$–$\chi$ mismatch (linewidth much larger than dispersive shift) cuts photon-shot-noise dephasing by $\sim 70\times$ versus the readout resonator, offering a scalable cQED route past the usual speed–coherence trade-off.

• Architecture: Two engineered notches (an intrinsic notch plus a bandstop notch) give broadband Purcell protection, while a reset stub places a reset mode below that protected band so reset and readout paths can be tuned independently.
• Per-qubit reset: Each qubit couples to its own reset resonator, enabling fast unconditional reset while suppressing filter-mediated qubit–qubit interactions that a shared reset bus would induce.
• Coherence band: Experimentally, Purcell-limited $T_1$ exceeds $1\mathrm{ms}$ across a $1.2\mathrm{GHz}$ qubit-frequency window—protection wide enough for typical multi-qubit frequency allocation.
• Operation speeds: The same hardware supports $500\mathrm{ns}$ dispersive readout without a Josephson parametric amplifier and $100\mathrm{ns}$ reset with $99.6\%$ efficiency.
• Dephasing mitigation: Designing the reset resonator with a large $\kappa /\chi$ mismatch (decay rate $\kappa$ vs dispersive shift $\chi$) suppresses photon-shot-noise-induced dephasing by a factor of $70$ relative to the readout resonator.

TL;DR A microwave-free controlled-Z (CZ) gate—conditional $\pi$ phase on $|11⟩$—is realized in a hybrid fluxonium–fixed-frequency-transmon pair by flux-pulsing only a tunable transmon coupler, reaching $>99.99\%$ average fidelity in $25\mathrm{ns}$ with leakage below $10^{-5}$. Single-parameter flux control turns residual ZZ on at the gate and off at idle, keeping spectator-limited infidelity below $10^{-4}$ and decoherence-aware fidelity near $99.9\%$.

• Architecture: A fluxonium, a fixed-frequency transmon, and a flux-tunable transmon coupler form the hybrid circuit; only the coupler’s external flux is modulated.
• Mechanism: Coupler-flux steering activates conditional-phase accumulation during the gate and suppresses idle residual ZZ (always-on state-dependent frequency shift), cutting spectator-induced errors.
• Control: A low-dimensional Fourier-cosine flux pulse, optimized with a cost that separately targets conditional-phase error and leakage from the computational subspace, yields a microwave-free CZ in $25\mathrm{ns}$.
• Performance: Numerics give average fidelity $>99.99\%$ and leakage $<10^{-5}$; with experimentally relevant decoherence the fidelity stays near $99.9\%$, and with spectators the CZ infidelity remains $<10^{-4}$.

TL;DR Two-qubit gates built from quantum interference of consecutive Landau-Zener (LZ) transitions—nonadiabatic passages through an avoided crossing—bridge baseband (near-DC) and parametric (RF) control in superconducting circuits, with the drive frequency tunable in situ over a continuous band up to hundreds of MHz and the gate speed set by the full coupling strength rather than a dispersive interaction. Demonstrations on modular coupled generalized flux qubits (flux-tunable circuits with a flexible junction network) and on monolithic transmons mark the LZ gate as a practical tool for multiplexed control and chiplet interconnects.

• Mechanism: Consecutive Landau-Zener (LZ) passages through a two-qubit avoided crossing interfere to produce a coherent entangling unitary, continuously connecting near-DC baseband pulses and microwave parametric drives.
• Speed: Relative to dispersive-coupler gates, the LZ gate runs on the timescale of the bare coupling, so larger coupling yields proportionally faster two-qubit operations.
• Tunability: The control frequency can be retuned in situ across a continuous interval up to hundreds of MHz, enabling frequency-agile two-qubit pulses on shared control wiring.
• Experiments: The gate is shown on a modular chiplet architecture of coupled generalized flux qubits and on a monolithic transmon device, spanning both flux-tunable and fixed-frequency-style cQED platforms.
• Systems role: Combined frequency agility and coupling-limited speed target multiplexed control pulses and inter-chiplet links where fixed parametric tones or purely baseband schemes constrain scheduling and interconnects.

TL;DR An all-van-der-Waals (vdW) merged-element transmon—where a thin $\mathrm{hBN}$ barrier both Josephson-couples and capacitively shunts two ${\mathrm{NbSe}}_2$ islands—reaches average energy-relaxation $T_{1,\mathrm{avg}}=55\mu \mathrm{s}$ and Hahn-echo coherence $T_{2\mathrm{E},\mathrm{avg}}=21\mu \mathrm{s}$ under circuit quantum electrodynamics (cQED) readout. The result puts atomically stacked junctions on a comparable coherence footing to conventional $\mathrm{Al}/{\mathrm{AlO}}_x/\mathrm{Al}$ transmons while shrinking the footprint and cutting stray capacitive coupling.

• Device: A ${\mathrm{NbSe}}_2/\mathrm{hBN}/{\mathrm{NbSe}}_2$ stack forms a merged-element transmon in which the same thin $\mathrm{hBN}$ layer supplies Josephson tunneling and shunt capacitance between two ${\mathrm{NbSe}}_2$ islands.
• Main result: cQED temporal characterization yields $T_{1,\mathrm{avg}}=55\mu \mathrm{s}$, Hahn-echo $T_{2\mathrm{E},\mathrm{avg}}=21\mu \mathrm{s}$, and Ramsey $T_{2\mathrm{R},\mathrm{avg}}=1.9\mu \mathrm{s}$.
• Noise diagnosis: The short Ramsey time tracks enhanced charge-noise sensitivity set by the realized device parameters rather than a materials-intrinsic limit of the vdW stack.
• Hardware implication: Lumped-element vdW qubits can match conventional $\mathrm{Al}/{\mathrm{AlO}}_x/\mathrm{Al}$ coherence while offering a smaller footprint and suppressed stray capacitive coupling—useful for denser cQED layouts.

TL;DR Even adaptively secure, quantum-accessible pseudorandom function-like state generators (PRFSGs; keyed oracles that output states indistinguishable from Haar-random ones on chosen inputs) do not imply non-adaptively secure, forward-only pseudorandom unitaries (PRUs; families of circuits indistinguishable from Haar-random unitaries under forward queries), even with non-unitary implementations and arbitrary ancillas. The proof treats any PRU built from state oracles as a map from oracle states to implemented unitaries and uses that map’s inherently low-rank derivatives to distinguish the output from true Haar randomness.

• Main result: A unitary-oracle separation shows the strongest PRFSG notion fails to yield even the weakest PRU notion: adaptive, quantum-accessible PRFSGs do not imply non-adaptive, forward-only PRUs.
• Scope: The separation survives when the candidate PRU may be non-unitary and may use arbitrarily many ancillary qubits.
• Method: View a PRU construction with state-generation oracle access as a map from underlying oracle states to implemented unitaries, then analyze that map’s derivatives.
• Distinguisher: Those derivatives are inherently low-rank, and that low-rank structure separates the resulting unitaries from truly Haar-random ones.
• Implication: Pseudorandomness for quantum states and for quantum unitaries are fundamentally distinct primitives, not interchangeable via black-box state oracles.