20-metre multimode fiber synthesises programmable high-dimensional quantum states

A team has demonstrated programmable synthesis of high-dimensional quantum states transmitted through 20 metres of multimode optical fibre. The work, published in Nature, combines photons generated in atoms with long coherence times and control techniques that allow complex states to be prepared and recovered despite the modal dispersion characteristic of this type of fibre. Multimode fibre is an attractive channel for quantum information because it can carry many spatial modes at once, multiplying capacity compared with single-mode fibre. Its drawback is that modes couple and mix as they propagate, deforming the state. Earlier work had recovered individual modes or controlled classical transmission through long multimode fibres, but active, programmable synthesis of high-dimensional states in a quantum channel remained a challenge. The method relies on atomic photons with long coherence, which reduce spectral broadening and ease state control. On that basis, the researchers use a modulation and calibration scheme that compensates the fibre's dispersion and allows the desired state to be written into the output modes. The 20-metre length is an order of magnitude longer than previous demonstrations of this kind while maintaining sufficient fidelity to operate with high-dimensional states. The result opens the door to quantum communication and computing protocols that exploit the internal dimensionality of spatial modes as a resource, rather than limiting themselves to qubits. It also raises questions about scalability: whether control can be maintained in longer fibres, with more modes and in less controlled conditions than in the laboratory. The next phase plans to extend the distance and explore applications in metropolitan quantum networks.

Agentic AI compiles quantum circuits more efficiently with classical verification

A team proposes a new paradigm for compiling more efficient quantum circuits using artificial intelligence agents, provided that the critical subroutines of the circuit exhibit a classically verifiable structure. When these subroutines consist exclusively of reversible classical logic, phasing operations, and measurement-based uncomputation, their execution can be efficiently simulated for classical inputs. This property allows the approximate synthesis of the circuit to be certified through a form of fuzz testing, under suitable assumptions about the input distribution. The pattern was recently exploited to improve quantum cryptanalysis circuits. Now, the authors highlight the flexibility of the approach with three notable applications: the computation of 1/√x, which bottlenecks real-space simulations of the Coulomb operator relevant to modelling inertial confinement fusion reactors; the decoding of Reed-Solomon codes, which hinders the decoded quantum interferometry algorithm when approximately solving the optimal polynomial intersection problem; and quantum look-up tables together with the controlled application of Givens rotations, which constrain the most performant quantum algorithms for simulating chemistry in second quantization. The researchers present reference implementations of these algorithms and explain how they can be constrained and verified to seed a Kaggle competition designed to accelerate AI-based research into the optimization of their compilation.

Neural networks describe anyon gases of up to 20 particles

The macroscopic behavior of a non-interacting gas of Abelian anyons has remained unresolved for decades. Previous numerical methods, based on exact diagonalization in the continuum and on the plane, were limited to very small systems of at most N = 4 anyons. A new study explores neural network wavefunction technology to build variational ansätze suited to continuum models of the Abelian anyon gas. The proposed architecture, dubbed "AnyNet", exactly enforces the non-analytic short-distance behavior induced by anyonic statistics. The method reproduces exact diagonalization results with excellent agreement for N = 4 anyons, reaches accurate results up to N = 16 for intermediate statistics, and extends to N = 20 for almost-fermionic anyons, with a 0.4% discrepancy relative to perturbative calculations. Using this methodology, the authors study anyons in a harmonic trap and anyonic atoms as a model for anyon binding to impurities. They analyze the structure of anyonic wavefunctions as the statistics parameter is varied, focusing on the single-particle density, the pair density, the angular momentum, and the scaling of the wavefunction as multiple anyons coalesce. The work opens the door to understanding the macroscopic behavior of anyons, for which rich phases have been predicted theoretically. Next steps will involve extending the method to larger system sizes and exploring those predicted phase transitions.

A balanced parametric amplifier doubles its useful bandwidth without interference

A team has experimentally demonstrated a balanced architecture for superconducting travelling-wave parametric amplifiers (TWPAs) that doubles the instantaneous useful bandwidth while rejecting the image frequency. TWPAs offer broadband gain, low added noise and high dynamic range, qualities that make them attractive for amplifying multiplexed signals spread across a wide frequency range, such as qubit resonator arrays, or inherently wideband signals, such as those from spectrometers and radiometers. In many cases, however, the useful bandwidth of a TWPA is limited by the production of an idler tone at an image frequency. In TWPAs based on three-wave mixing (3WM), an input signal at frequency fₛ is indistinguishable from one at fₚ − fₛ, where fₚ is the pump frequency. This sideband degeneracy introduces spectral ambiguity when the input spectrum is broadband or not known in advance, forcing the use of complex filtering to clean the output. Resolving this ambiguity without sacrificing bandwidth has been one of the practical brakes on using these amplifiers in wideband applications. The proposed architecture spatially separates the amplified signal from the output pump and the generated idler, resolving the degeneracy. It consists of two amplifiers connected between a pair of 180° hybrid couplers, with the signal and pump injected and output through the sum and difference ports, respectively. Using two kinetic-inductance TWPAs, the authors measured approximately 20 dB of signal gain across nearly an octave of bandwidth, with pump rejection below −20 dB and an average image rejection of about −15 dB over the operating band. The results establish the balanced TWPA as a practical route to broadband parametric amplification, with a larger fraction of contamination-free bandwidth and no need for complex filtering. The improvement is especially relevant for reading out superconducting qubits and for sensitive microwave instrumentation, where spectral cleanliness directly at the amplification stage reduces downstream hardware and simplifies the detection chain.

Localized constraints fragment Hilbert space without slowing transport

An analysis of several many-body models shows that spatially localized conserved quantities uniformly fragment the Hilbert or configuration space, yet leave energy transport properties unchanged. The study covers the quantum Volterra model, its nonintegrable classical limit, and two cellular automata derived from the latter. Although such constraints are often associated with lattice gauge theories, the models considered here admit no such interpretation. In the quantum Volterra model, the authors derive an exact mapping of the dynamics within each Hilbert-space fragment onto a quantum chiral Potts chain. With this tool in hand, they show that energy transport is ballistic and that the transport coefficients are identical across all fragments: the fragmentation does not create sectors with slower or faster dynamics. In the nonintegrable classical limit, energy transport becomes diffusive, but the conclusion is analogous: the diffusion constant turns out to be independent of whether the constraints are fixed or resolved. The cellular automata further confirm that, while transport remains intact, localized constraints lead to a strong breakdown of local ergodicity. The upshot is a picture in which fragmentation and transport decouple: the presence of many local conserved quantities suffices to prevent ergodic exploration of configuration space, but not to modify how energy propagates. Whether this decoupling survives in systems with continuous degrees of freedom or in the presence of long-range interactions, where fragmentation might cease to be uniform, remains an open question.

IBM suppresses logical errors in a surface-code memory on the cloud

A team has demonstrated logical error suppression in a rotated surface-code memory implemented on a cloud-accessible IBM Nighthawk r2 superconducting processor. Quantum error correction can reduce the logical error rate exponentially with code distance when physical noise is below threshold, but turning that promise into hardware requires coordinating gates, mid-circuit measurements and qubit coherence in the face of device variation, asymmetric noise and spatiotemporal correlations. The authors optimized the syndrome extraction circuit by combining short readout, fused measurement and reset, and an XX echo on data qubits during ancilla readout. Compared with standard readout without reset or echo, the recipe cuts the median per-round logical error rate by a factor of 3.8 in the X basis and 1.4 in the Z basis. Performance also varies across the tested qubit placements on the processor, which they attribute to differences in qubit quality. In memory experiments of up to 150 rounds, the advantage associated with code distance weakens and long-lived correlations appear that may reflect a persistent error component. Under the applied optimization, increasing the protected distance from 3 to 6 — while holding the other fixed — lowers the per-round logical error rate from 0.53 % to 0.24 % in Z memory and from 0.52 % to 0.22 % in X memory, with maximum-likelihood-decoding error-suppression factors of 1.68 and 1.85, respectively. The results link circuit compilation and optimization to surface-code memory performance — and its limitations — and point toward quantum-processor and error-correction-circuit codesign as a route to fault-tolerant quantum computing primitives on current cloud-accessible quantum hardware.