Cryo DACs, Kerr cats, and query–space bounds (09/11/26)

TL;DR A cryo-compatible current-mode DAC (Digital-to-Analog Converter: maps digital codes to a shaped drive current) in planar 90 nm CMOS delivers digitally tunable amplitude and edge slopes, with external triggers setting pulse width, for flux biasing or readout drive of superconducting circuits. Matching cryogenic measurements to circuit predictions strengthens the case for integrated cryo-CMOS control near qubits rather than room-temperature analog chains alone.

• Main result: Cryogenic tests of a current-based CMOS DAC show pulse amplitude and rise/fall slopes set by digital controls, and external triggers set pulse width, while the DAC drives a superconducting load.
• Device physics: Current-mode output couples flux into a qubit loop or injects controlled current into a readout path, so one cryo-CMOS block can replace parts of a warm analog bias/readout chain.
• Process and validation: The circuit is fabricated in planar 90 nm CMOS, and measured cryogenic behavior closely tracks pre-silicon circuit predictions.
• cQED impact: Wide digital tunability lets one DAC topology serve multiple superconducting circuit types without redesigning the analog frontend for each bias or readout recipe.
• Limit: The paper reports design-and-test agreement and application scope; it does not claim a measured end-to-end qubit fidelity gain from this DAC alone.

TL;DR A $\mathrm{S}/\mathrm{I}/\mathrm{S}/\mathrm{I}/\mathrm{S}$ (SISIS) quantum circuit refrigerator (QCR)—a tunnel junction that dumps resonator excitations into a cold quasiparticle bath—can confine a Kerr-cat qubit (logical states in metastable coherent states of a Kerr parametric oscillator) with engineered two-photon dissipation while suppressing single-photon absorption. Relative to conventional $\mathrm{S}/\mathrm{I}/\mathrm{N}/\mathrm{I}/\mathrm{S}$ (SINIS) QCRs, that regime cuts dephasing-induced leakage without paying as large a QCR-induced phase-flip penalty.

• Problem: Pure dephasing in driven Kerr-cat hardware leaks population out of the cat subspace, so active refrigeration is needed to remove excess photons.
• SINIS limit: A SINIS QCR absorbs single photons and thereby suppresses leakage, but that same single-photon channel enhances QCR-induced phase-flip errors.
• Proposal: Couple a SISIS junction to the Kerr parametric oscillator (KPO)—the pumped nonlinear resonator that hosts the Kerr-cat—and bias into a window where single-photon tunneling is suppressed while two-photon absorption dominates.
• Hardware claim: In that two-photon regime the SISIS QCR strongly suppresses dephasing-induced leakage and substantially reduces the phase-flip increase that accompanies QCR operation.
• cQED takeaway: Leakage mitigation for Kerr-cats can be engineered at the fridge junction by changing the tunnel stack from SINIS to SISIS rather than only by stronger single-photon cooling.

TL;DR A dispersively coupled qubit sets the squeeze axis of a SQUID-terminated resonator via a conditional-squeezing gate—parametric resonance that squeezes along a qubit-state-dependent quadrature—made practical by refocusing that cancels slow Hamiltonian drifts during the drive. The gate encodes an arbitrary qubit into error-detectable superpositions of single- and two-mode squeezed states with fidelity above $0.99$ under optimistic Kerr and decay, limited mainly by those same nonlinearities and dissipation.

• Main result: A conditional-squeezing gate squeezes a SQUID-terminated resonator along a quadrature fixed by a dispersively coupled qubit, generalizing the controlled-squeezing gate that only turns squeezing on or off.
• Hardware mechanism: State-dependent parametric resonance drives the SQUID flux at a qubit-conditioned frequency; refocusing—a mid-gate echo that cancels slowly varying Hamiltonian terms—suppresses the unwanted phase and squeeze-axis drifts that otherwise spoil the resonance.
• Encoding application: The gate maps an arbitrary qubit into a superposition of single- or two-mode squeezed resonator states, a non-Gaussian resource whose photon-number parity detects loss errors.
• Fidelity limits: Refocusing raises encoding fidelity substantially; Kerr nonlinearities and resonator/qubit dissipation set the remaining ceiling, with fidelities above $0.99$ for optimistic experimental rates.
• Outlook for cQED: The same SQUID-resonator architecture extends toward higher-order squeezed-state superpositions—a rotation-symmetric bosonic code family—if the control is upgraded from qubit to qudit.

TL;DR Within label-symmetric algorithms—those that treat a random function's output labels as interchangeable—any collision finder using $T$ queries and $S$ qubits must obey $T=\Omega (N^{1/3})$ and $T^{2}S=\Omega (N\log N)$, matching a space-efficient BHT implementation when the domain size equals $N$. The same tradeoff proves Ambainis's quantum walk optimal for Element Distinctness under this symmetry, closing a long-open query–space gap for this natural algorithm class.

• Main result: For a uniformly random $f:[M]\to [N]$, every label-symmetric collision algorithm with constant success probability satisfies $T=\Omega (N^{1/3})$ and $T^{2}S=\Omega (N\log N)$.
• Matching upper bound: When $M=N$, a space-efficient BHT algorithm—collision search that stores $O(N^{1/3})$ input–output pairs in a quantumly accessible classical table—matches both bounds.
• Element Distinctness: The search version on $f:[n]\to [n^2]$ inherits $T=\Omega (n^{2/3})$ and $T^{2}S=\Omega (n^2\log n)$, matching Ambainis's quantum walk (the optimal walk-based collision search on $n$ elements).
• Method: A space-sensitive compressed oracle—recording what the algorithm has learned as a superposition of databases—plus representation theory shows $S$ qubits retain only $O(S/\log N)$ collision-free database entries, which plugs into the progress measure to yield the tradeoffs.

TL;DR A surface-code algorithm-to-execution stack—building dependency-preserving logical schedules and an executable workload that tracks parallelism and time-resolved non-Clifford demand—shows that physical cost is set by scheduling and layout, not by logical gate counts alone. Circuit-specific organization and non-Clifford implementation choice can cut space-time volume by orders of magnitude versus naive baselines.

• Main result: Physical overheads diverge sharply among circuits with similar logical resource counts once surface-code layout, routing latency, and non-Clifford supply are co-planned.
• Framework: From hierarchical algorithm descriptions the pipeline builds a dependency-preserving logical schedule and an executable workload that records logical interactions, operation parallelism, and time-resolved demand for non-Clifford resources—gates outside the Clifford group that need costly magic-state or rotation machinery.
• Non-Clifford choice: Under their direct-rotation calibration—preferring native rotation implementations where cheaper than T-gate synthesis—selecting the non-Clifford path cuts space-time volume by up to $241.5\times$ versus an all-synthesis baseline on a QAOA (Quantum Approximate Optimization Algorithm) amplitude-amplification workload.
• Layout effect: Circuit-specific surface-code layouts lower routed-latency estimates on all twenty benchmarks across seven algorithm families; thirteen also reduce space-time volume because communication savings outweigh added spatial overhead.
• Scale test: The same workflow is applied to a hierarchically composed elliptic-curve discrete-logarithm application, tying logical computation to surface-code organization and fault-tolerant execution in one traceable path.