Chronological feed of everything captured from Mikhail Lukin.
Quantum error correction (QEC) is critical for scalable quantum computing, but existing decoding algorithms limit the potential of quantum low-density parity-check codes. This research introduces a convolutional neural network (CNN) decoder that leverages geometric code structure. The CNN decoder achieves significantly lower logical error rates and higher throughput, suggesting reduced space-time costs for fault-tolerant quantum computation.
The protocol uses local projective measurements and unitary feedback to prepare unknown ground states of frustration-free gapless quantum systems in polynomial time scaling with system size. For single-particle dynamics, performance is analytically proven; many-body generalization relies on quasiparticle physics, with preparation time linear in the inverse finite-size gap (up to logs) when the dynamical critical exponent z ≥ effective quasiparticle dimension d. Transient cooling reveals universal critical properties, verified numerically on 1D/2D Heisenberg ferromagnets, Fredkin chain, and 2D RVB model. Fully digital, it outperforms adiabatic methods for high-fidelity access in near-term devices.
Researchers demonstrate an analog-digital quantum simulator using Rydberg-hyperfine qubit mapping in neutral atom arrays, enabling programmable state preparation, non-destructive readout, atom reuse, and loss mitigation via reservoirs. They engineer ring-exchange and hopping dynamics via Floquet driving and measure single-excitation spectral functions. On a 271-site kagome lattice, closed-loop optimization realizes an out-of-equilibrium Rokhsar-Kivelson critical quantum spin liquid, evidenced by absent local order, long-range dimer coherences up to 18 sites, and field-theory-consistent correlations.
The method maps fault-tolerant Clifford circuits to subsystem codes via spacetime code formalism, enabling estimation of Pauli noise from syndrome data alone. It provides necessary and sufficient conditions for learnability of physical and logical errors, predicting logical fidelities accurately with polynomial sample complexity. This yields exponential advantages over logical-data-only methods like direct fidelity estimation, even for exponentially suppressed logical error rates, and is validated on synthetic and experimental data from Bluvstein et al.
Phantom codes are quantum error-correcting codes that implement fault-tolerant entangling gates between all logical qubits in a code block solely through physical qubit relabeling during compilation, achieving perfect fidelity without spatial or temporal overhead. The authors enumerate 2.71×10^10 CSS codes up to n=14 qubits, identify further instances up to n=21 using SAT methods, and construct higher-distance families from quantum Reed-Muller and binarized qudit codes. Noisy simulations demonstrate 1-2 orders of magnitude lower logical infidelity than surface codes for GHZ preparation and Trotterized simulations at comparable overhead.
Inverse quantum simulation inverts traditional forward simulation by minimizing a cost function encoding desired material properties on quantum hardware to prepare target many-body states. Hamiltonian learning then reconstructs a low-energy Hamiltonian for which this state is an approximate ground state, providing interpretable models for synthesis. Applications include enhancing d-wave correlations in the fermionic Hubbard model for high-Tc superconductors, stabilizing topological order, and optimizing photochemical and spectroscopic properties.
Experiment demonstrates three quantum interaction regimes between a Rydberg polariton in an atomic ensemble and an adjacent single Rydberg atom: polariton blockade, coherent exchange, and probabilistic hopping, distinguished by transmission spectra. A transition through an exceptional point occurs between blockade and coherent exchange. These interactions enable fast, non-destructive Rydberg atom detection and show promise for nonlinear photonic networks.
Researchers introduce dressed-state qubit encoding under a magnetic field perpendicular to the diamond lattice, enabling direct enhancement of native dipolar interactions in NV center ensembles. This yields a 3.2× improvement in the JT₂ coherence parameter over Floquet methods and 2.6× (8.3 dB) better AC magnetometry sensitivity. The extended coherence facilitates probing spin transport at intermediate-to-late times, offering a versatile tool for interacting spin systems.
Mikhail Lukin, Chief Scientist at QuEra, highlights a critical juncture for neutral atom quantum computing, with recent advancements enabling the integration of all necessary elements for large-scale, utility-grade machines. This progress, built on decades of foundational research, signifies a shift from theoretical concepts to practical implementation. The field now requires a co-design approach, integrating diverse disciplines and pushing scientific frontiers simultaneously to accelerate development.
Quantum many-body systems exhibit local arrows of time that can deviate from the global Hamiltonian-induced time evolution, rendering time flow observer- or subsystem-dependent. The paper defines these local arrows and links them to spacetime quantum entropies. Numerical and analytical examples demonstrate manifestations including quantum thermalization and error correction phenomena.
A programmable neutral-atom quantum simulator with up to 180 qubits explores quench dynamics in spin models, identifying stable dynamical regimes stabilized by Floquet-like prethermal steady states over long timescales via strong dynamical constraints. Sharp resonant peaks in the response arise from structured melting of prethermalization. In 2D, a sharp dynamical response shift, converging with system size and tied to Néel-order defect proliferation, signals a dynamical phase transition without equilibrium counterpart.
High-rate qLDPC codes extend to computation via batched operations that apply identical logical gates across multiple code blocks in parallel, achieving constant space-time overhead for arbitrary CSS codes in single-shot error correction, state preparation, surgeries, code switching, and addressable Clifford gates. Parallel non-Clifford gates follow with low space-time cost, enabling efficient compilation of parallel quantum algorithms like lattice Hamiltonian simulations. Near-term implementation uses self-dual Bivariate-Bicycle codes (encoding rate ~1/10) with transversal Cliffords and global T gates, outperforming surface codes and low-rate qLDPCs in simulation costs.
Researchers demonstrate loading, cooling, and imaging of 87Rb atom arrays in finite magnetic fields using electromagnetically induced transparency (EIT) cooling combined with fluorescence imaging. The technique achieves 99.6(3)% readout fidelity at 98.2(3)% survival probability and up to 68(2)% single-atom stochastic loading. A predictive model for survival probability aligns with this and other atom array experiments, cooling both axial and radial directions to support continuous neutral atom quantum processors.
Modular attention-based neural decoders learn gate-induced error correlations, generalizing from random circuit training to multi-qubit algorithmic workloads with fast inference and logical error rates matching most-likely-error decoders. They incorporate loss-resolving readout for realistic noise, including qubit loss, and simplify design by targeting algorithm-specific observables without accuracy loss. Validated on surface and 2D color codes under circuit-level noise, these decoders provide interpretability via attention mechanisms and enable deep-circuit fault-tolerant quantum algorithms.
Conventional optical imaging suffers from shot-noise-limited SNR due to signal integration and classical post-processing of asynchronously arriving photons. The proposed method encodes photonic amplitude into qubit registers for coherent quantum processing, bypassing these constraints. Applied to unresolved point source imaging for exoplanet detection, it achieves orders-of-magnitude performance gains using small-scale quantum processors under realistic conditions.
New method uses reconfigurable quantum systems with non-local connectivity, mid-circuit measurement, and classical feedforward to generate dynamical fermion-to-qubit mappings, reducing space-time overhead from O(N) to O(log N) for N fermionic modes. For structured circuits like fermionic FFT, overhead drops to O(1). Technique enables efficient simulation of SYK model, periodic materials, and free-fermion state preparation, with orders-of-magnitude gate reductions compatible with fault-tolerant devices.
Tricycle codes extend bicycle codes to three homological dimensions, supporting constant-depth physical circuits for logical CCZ gates across three code blocks via new analytical and numerical methods for 3D homological products. They facilitate single-shot magic state preparation and error correction, yielding high circuit-noise thresholds above 0.5%. Simulations demonstrate logical error rates below 6e-10 for 50-100 qubit blocks with modest post-selection, optimized for reconfigurable neutral atom platforms.
Theoretical and numerical analysis identifies a deconfined quantum critical point (DQCP) in triangular-lattice Rydberg atom arrays between ordered phases at 1/3 and 2/3 excitation densities, previously observed experimentally. Field theory predicts critical exponents for infinite cylinders and a critical conformal field theory with emergent U(1) symmetry, a DQCP hallmark, validated numerically. Results extend to ladder geometries, enabling experimental probes of U(1) symmetry via finite tweezer arrays.
Researchers demonstrate a full universal fault-tolerant quantum architecture using reconfigurable arrays of up to 448 neutral atoms, implementing surface codes with 2.14x below-threshold error suppression via repeated QEC and ML decoding. They achieve logical entanglement via transversal gates and lattice surgery, extend to universal logic with 3D [[15,1,3]] codes and transversal teleportation for arbitrary-angle synthesis at logarithmic overhead, and enable mid-circuit qubit reuse boosting cycle rates by 100x for deep circuits with dozens of logical qubits using [[7,1,3]] and [[16,6,4]] codes while keeping constant entropy. These experiments elucidate principles like balancing quantum logic with entropy removal, physical entanglement in gates, and teleportation for universality and reset in neutral atom systems.
Researchers demonstrate continuous operation of a 3000-qubit neutral atom array using dual optical lattice conveyor belts to transport atom reservoirs into the science region for high-rate extraction into tweezers. The system reloads 300,000 atoms per second, enabling over 30,000 initialized qubits per second and sustaining the array for over two hours. Quantum coherence is preserved during refilling, supporting spin-polarized or superposition states for scalable quantum computing and metrology.
This method enables quantum phase estimation using only local control operations, eliminating the need for global unitaries conditioned on auxiliary qubits. It measures the complex phase of the Loschmidt echo for circuit and Hamiltonian dynamics by tracking phase changes, trading reduced circuit depth for higher sampling and classical postprocessing costs. Applicable to any efficiently preparable state without reference states, it supports spectral property measurements in large many-body systems on current hardware.
Researchers demonstrate coherent collective many-body dynamics in dense, disordered electron spin ensembles in diamond by applying time-dependent magnetic field gradients alongside global coherent control. This approach generates and probes nanometer-scale spin spirals, with Hamiltonian engineering enhancing microscopic interaction symmetry to achieve disorder-robust spin evolution. The technique overcomes dipolar coupling limitations from positional disorder, paving the way for interaction-enhanced quantum metrology and ambient-condition nanoscale imaging of materials and biology.
Researchers propose a hybrid light-matter approach for fault-tolerant blind quantum computation (BQC), leveraging high-fidelity local gates on server matter qubits and photon-based delegated blind rotations. This constructs loss-tolerant delegated gates, enabling efficient algorithm compilation and scalable fault-tolerant logical algorithms. The design improves error-correction thresholds, boosts blind logical circuit speed and depth, and maps to neutral atom arrays and solid-state spin defects.
A low-overhead architecture for reconfigurable neutral atom arrays leverages transversal gate operations and dynamic connectivity to implement fault-tolerant building blocks like magic state factories, arithmetic units, and look-up tables. It achieves runtime speed-up scaling with code distance d by minimizing atom move times and decoding volume. Resource estimates show 2048-bit RSA factoring requires 19 million qubits and 5.6 days at 1 ms QEC cycles, yielding ~50x runtime improvement over prior estimates without added space.
Reformulates decoding of transversal quantum circuits by tracking logical operator products, reducing syndrome extraction rounds by code distance d while resembling single-qubit memory decoding. Enables minimum-weight perfect matching for fault-tolerant surface code decoding with thresholds matching single-qubit memory. Benchmarks show decoding runtime below conventional lattice surgery, extending single-qubit QEC to multi-qubit transversal algorithms.
Researchers demonstrate wideband magnetic correlation measurements using spectrally resolved NV centers in diamond, achieving spatial resolution below the optical diffraction limit and MHz noise sensitivity of 15 nT/√Hz. The technique leverages high-fidelity optical readout and long spin coherence for MHz-range probing, extending to GHz-range via correlated T1 relaxometry. Under external GHz noise, NV correlations reveal coherent and incoherent dynamics akin to superradiance, despite featureless individual relaxations, enabling study of nonlocal condensed-matter phenomena.