Constant overhead, depth, and microwave gates (10/01/26)

TL;DR An all-microwave cross-resonance scheme—microwave drives on one qubit at another's frequency that generate an effective $ZX$ interaction—builds single-step $n$-control/$m$-target gates on frequency-tunable transmons without tunable couplers. Numerics and pulse optimization reach $>99.9\%$ three-qubit fidelity in $60\mathrm{ns}$ and extend to a five-qubit CXXXX (one control, four simultaneous target $X$ flips), offering a hardware path to compress stabilizer-type circuit depth.

• Main result: Cross-resonance drives on frequency-tunable transmon networks produce enhanced two-body $ZX$ couplings that realize single-step multiqubit gates with $n$ controls and $m$ targets, without tunable couplers.
• Theory: Quad-frame rotation and Bloch–Brandow block diagonalization of the driven Hamiltonian map out regimes that strengthen target $ZX$ terms while suppressing parasitic higher-order interactions such as $ZXX$.
• Performance: Gradient-based pulse-shape optimization yields three-qubit gates with fidelities exceeding $99.9\%$ in a $60\mathrm{ns}$ gate time.
• Scalability: The same optimization applied to a five-qubit CXXXX architecture—one control and four simultaneous target $X$ flips—supports physical viability of multi-target microwave driving.
• Hardware impact: Because these operations match stabilizer-measurement structure, they can replace sequential two-qubit decompositions and cut circuit depth for stabilizer-type routines in near-term and fault-tolerant cQED.

TL;DR This paper gives the first fault-tolerant scheme that injects bare physical qubits into a constant-rate quantum error-correcting code—and ejects code states back out—with only constant space and time overhead. Injection and ejection remain reliable under circuit-level locally stochastic noise (each circuit location fails independently with bounded rate), closing a gap that otherwise blocks constant-overhead fault-tolerant computation that must touch unprotected qubits.

• Main result: A family of constant-rate codes admits fault-tolerant injection (encoding bare physical qubits into a code block) and ejection (decoding a code state back to bare qubits) with constant space overhead and single-shot (constant-depth) quantum circuits.
• Noise model: The gadgets succeed under circuit-level locally stochastic noise, paying only a small constant per-qubit corruption probability that bare physical qubits cannot avoid.
• Construction: High-dimensional hypergraph products of classical LDPC (low-density parity-check) codes—simplified Spielman linear-time-encodable codes—yield the outer codes; concatenation with inner codes of varying size upgrades non-uniform error resilience to uniform noise.
• Supporting gadgets: The same framework gives fault-tolerant error correction and code-state preparation, all single-shot, with physical qubits linear in the number of logical qubits given polynomial-size noiseless classical side computation.
• Significance: Constant-overhead FTQC that interfaces with unprotected states previously lacked such an injection/ejection primitive; this supplies that missing I/O step.

TL;DR Every $n$-qubit unitary can be approximated to operator-norm error $ϵ$ by a one- and two-qubit circuit of depth $\mathrm{poly}(n,\log 1/ϵ)$ using $2^{O(n)}$ ancillas, so exponential depth is not necessary once massive parallelism is allowed. With unbounded fan-out (a gate that copies one bit onto many wires at once), the same unitaries collapse further to constant depth, via a link from Aaronson–Kuperberg unitary synthesis to locally decodable codes and private information retrieval.

• Main result: Any $n$-qubit unitary admits an $ϵ$-approximation in operator norm by a depth-$\mathrm{poly}(n,\log 1/ϵ)$ circuit of one- and two-qubit gates with $2^{O(n)}$ ancilla qubits.
• Constant-depth lift: Allowing unbounded fan-out gates reduces those circuits from polynomial depth to constant depth.
• Prior barrier: It was open whether general unitaries require $2^{\Omega (n)}$ depth even with unlimited ancillas; this work shows they do not.
• Mechanism: The construction ties the Aaronson–Kuperberg unitary-synthesis problem (implementing an arbitrary unitary from classical advice about its action) to locally decodable codes and private information retrieval from complexity and cryptography.
• Scope: The claim is asymptotic circuit-complexity parallelization with exponential workspace, not a near-term hardware compilation recipe.