The method, which the authors call Quantum Feature Selection (QFS), exploits analog quantum simulation with arrays of neutral atoms excited to Rydberg states. In these processors, individual atoms are trapped by optical tweezers and act as two-level systems, with a ground state standing for zero and a highly excited Rydberg state standing for one. Crucially, atoms in Rydberg states interact through van der Waals forces that fall off with the sixth power of distance. That fixed physical law, normally a constraint, becomes the computational resource: by rearranging the atoms in a plane, the researchers can program the interaction structure of an optimization problem directly into the geometry of the array.

The mapping works in two steps. First, the relevance of each feature, quantified by its mutual information with the target variable, is encoded as a site-dependent local detuning, a laser parameter that makes certain atoms energetically more likely to be excited. Features with high relevance receive stronger detunings, biasing their atoms toward the Rydberg state. Second, pairwise redundancy between features, measured as mutual information between the features themselves, is translated into spatial distance. The team applies an inverse-sixth-root transform to the normalized redundancy matrix, so that highly redundant pairs of features are placed close together, where their strong interaction penalizes the configuration in which both are simultaneously selected. Weakly correlated features are pushed apart, where they behave nearly independently.

Because a general redundancy matrix cannot usually be embedded exactly in two dimensions, the researchers use multidimensional scaling (MDS), a classical technique that finds atomic positions best approximating the target distances. The resulting layout is then rescaled so that the closest atom pairs sit near the Rydberg blockade radius, the critical distance at which joint excitation is suppressed. An annealing protocol follows: the global Rabi frequency ramps up and down smoothly, a global detuning initializes the system, and the local detunings encoding relevance switch on in the latter half of the evolution. The dynamics steer the array toward low-energy configurations that balance relevance against redundancy, and repeated projective measurements yield bitstrings, each one a candidate feature subset.

The post-processing stage is deliberately conservative. Only the lowest-energy ten percent of measured bitstrings are retained, and features are ranked by their average Rydberg excitation density across this ensemble. A greedy construction then accepts a feature only if its normalized redundancy with all previously accepted features falls below a threshold, set to 0.7 in the reference experiments, ensuring that the final subset is not merely relevant but internally non-redundant. When the threshold rule cannot fill a requested subset size, the remaining slots are filled by density ranking, a completion step the authors flag explicitly so that large subsets are not misread as strict independent sets.

Evaluated in simulation on Amazon Braket, using a noise-free Schrödinger equation solver and 10,000 measurement shots per run, QFS was tested on three real-world binary classification benchmarks: Adult Income, Bank Marketing, and Telco Churn. The comparison set was demanding, including mutual-information ranking, Random Forest and XGBoost feature importances, L1-regularized logistic selection, and two classical optimizers of exactly the same relevance-redundancy objective, an exhaustive search called Exact-Q and a greedy heuristic called Greedy-Q. Across ten stratified train/test splits with a Random Forest downstream classifier, QFS achieved competitive, dataset-dependent performance. On Adult Income, QFS, Exact-Q, Greedy-Q, and XGBoost importance all landed within roughly three ten-thousandths of a point in mean AUC, while on Telco Churn every method was essentially tied within split-to-split variability.

The authors are candid about where the method falls short. On Bank Marketing, tree-based importance rankings achieved slightly higher mean AUC than the quantum-derived subsets, and QFS sometimes reached its best performance at larger subset sizes, a consequence of the evaluation protocol rather than an intrinsic preference of the sampler for compact sets. The geometric embedding itself introduces moderate distortion: the mean relative reconstruction error of the MDS layouts ranged from about 0.22 to 0.26 across the three datasets. Yet a stability analysis over 50 random embedding seeds showed that this distortion does not produce arbitrary subsets. Bank Marketing was invariant across seeds for subset sizes up to five, and Adult Income stabilized for sizes of three or more, with mean Jaccard similarity above 0.96.

Robustness tests of the classical extraction stage added further reassurance. Varying the retained low-energy fraction from 5 to 30 percent shifted the best AUC only marginally on a reference split, and injecting controlled noise into the measured excitation densities, mimicking readout fluctuations or calibration drift, left predictive performance largely intact even when the exact identity of the selected subsets drifted at small cardinalities. The authors interpret this as evidence that the pipeline’s predictive output is more robust than its exact subset choices, though they stress that the simulations remain noise-free: decoherence, atom loss, finite-temperature atomic motion, readout errors, and detuning miscalibration were all excluded and must be addressed in future device-level studies.

What makes the work notable is less a claim of quantum advantage than the physical interpretability of the encoding. Every term of the statistical objective has a hardware counterpart: relevance lives in local detuning amplitudes, redundancy in the sixth-power interaction law, and the relevance-redundancy trade-off emerges from the relative energy scales of the two. The main contribution, the authors write, is a neutral-atom implementation of a relevance-redundancy feature-selection objective together with a quantitative characterization of its geometric and post-processing approximations. Scaling up will require one atom per feature, site-resolved detuning control, and embedding strategies, such as sparsification or community-aware layouts, that preserve the dominant redundancy structure without forcing every pairwise relation into a single plane.

The study points toward a broader program in which analog quantum processors tackle the combinatorial core of data preprocessing rather than the models themselves. Future work, the team says, should focus on experimental validation on real neutral-atom devices, noisy analog Hamiltonian simulation, improved weighted and graph-aware embeddings, dynamic detuning optimization, and larger benchmarks where the scaling of analog sampling can be measured directly. Extensions to multiclass problems, regression, and unsupervised settings are also on the table. For now, QFS stands as a proof of concept that a laser-driven array of atoms, arranged according to the statistical anatomy of a dataset, can select features as well as established classical heuristics, and that the path from information theory to atomic physics can be traversed without losing the meaning along the way.

Subject of Research: Analog quantum feature selection using neutral-atom Rydberg processors

Article Title: Analog quantum feature selection with neutral-atom quantum processors

Article References: Orquín-Marqués, J. J., Flores-Garrigós, C., Gomez Cadavid, A., Simen, A., Solano, E., Hegade, N. N., Martín-Guerrero, J. D., & Vives-Gilabert, Y. (2026). Analog quantum feature selection with neutral-atom quantum processors. Quantum Machine Intelligence, 8(2), Article 108. https://doi.org/10.1007/s42484-026-00456-8

Image Credits: AI Generated

DOI: 10.1007/s42484-026-00456-8

Keywords: neutral atoms, Rydberg arrays, quantum feature selection, analog quantum simulation, mutual information, quantum optimization, machine learning, multidimensional scaling, Rydberg blockade, QUBO, Amazon Braket, Quantum Machine Intelligence

The significance of the achievement becomes clear when the scaling requirements of quantum computing are considered. Neutral-atom quantum computers, which trap arrays of individual atoms in vacuum and use each atom as a qubit, are expected to operate with arrays of roughly 10,000 atoms. That sounds enormous, but fault-tolerant universal quantum computers, machines capable of correcting their own errors and running genuinely useful algorithms, are expected to require more than one million physical qubits. Error correction is the reason for this overhead: fragile quantum states must be protected by encoding logical qubits across many physical ones. No single processor is likely to reach that scale on its own, so researchers have increasingly looked toward networking, distributing entangled photons between multiple quantum processors so that they can work together. That vision demands a photonic interface capable of linking many qubits in parallel, exactly what the Osaka-led team set out to build.

Earlier attempts at multiplexing, the practice of handling many optical channels simultaneously, relied mainly on bundles of parallel optical fibers and were limited to only a handful of channels. The approach suffered from insufficient integration density, and the wide spacing between atoms that fiber-based schemes required made them poorly matched to conventional neutral-atom quantum computers, in which atoms are typically held just a few micrometers apart. The new interface takes a different route. The researchers incorporated an integrated optical waveguide array, a chip-like structure containing many closely spaced light-guiding channels, and demonstrated parallel photon delivery and detection from a neutral-atom array. Photons emitted from ten atoms spaced at micrometer-scale intervals were coupled into ten parallel channels of a 32-channel waveguide array, sent through optical fibers, and detected in parallel, all while preserving the quantum correlations that make the photons useful for networking.

The experiment went beyond simply counting detected photons. The team confirmed that inter-channel crosstalk, the unwanted leakage of light or signal between neighboring channels, was negligible, a critical property for any system that must handle many independent quantum channels at once. They also verified correlations between the quantum states of the atoms and the polarization states of the photons they emitted. Such atom-photon correlations are the raw material of entanglement between a stationary qubit, the atom, and a flying qubit, the photon, and they underpin schemes for multiplexed atom-photon entanglement and, ultimately, for connecting quantum processors across a network. The researchers report that the approach should be scalable to approximately 100 parallel channels, a figure that hints at the kind of density future networked architectures will need.

Detecting single photons reliably is itself a formidable technical challenge, and the Japanese collaboration brought together complementary expertise to solve it. Photon detection in the experiment was performed using a multi-channel superconducting nanostrip photon detector system. The underlying detector technology was developed by Shigehito Miki, Director of the Superconductive ICT Device Laboratory at NICT’s Kobe Frontier Research Center within the Advanced ICT Research Institute, and the complete research system for this experiment was newly built by Hideki Shimoi, a manager at the Electron Tube Division of Hamamatsu Photonics K.K. Superconducting nanostrip detectors are prized in quantum optics because they combine high detection efficiency with very low noise, registering the arrival of individual photons as tiny voltage pulses when the photons break superconducting current paths chilled to cryogenic temperatures.

The institutional division of labor reflects how modern quantum engineering projects increasingly span academia, national research institutes, and industry. The University of Osaka handled overall coordination and implementation of the research. NICT provided the superconducting nanostrip photon detector technology and carried out part of the device fabrication process, using facilities at its Advanced ICT Device R&D Promotion Center. Hamamatsu Photonics, a company with decades of experience in photodetection, developed the detector system specifically for this experiment. Yamamoto framed the collaboration as a full-stack effort: through research and development spanning from neutral-atom arrays to superconducting nanostrip photon detector systems, the team achieved the first demonstration of a multiplexed optical interface, he said, adding that the group will now scale up the degree of multiplexing and work toward connecting neutral-atom quantum computers, accelerating progress toward a fault-tolerant networked quantum computer.

The funding landscape behind the project is equally telling. The research was conducted as part of the Japan Science and Technology Agency’s Moonshot Research and Development Program under Moonshot Goal 6, which aims for the realization of a fault-tolerant universal quantum computer that will revolutionize economy, industry, and security by 2050. The work falls specifically within the projects on fault-tolerant networked quantum computers and on a quantum cyberspace built from networked quantum computers. Additional support came from the JST Program on Open Innovation Platform for Industry-academia Co-creation through a Quantum Software Research Hub, the JST Adopting Sustainable Partnerships for Innovative Research Ecosystem program, and Ministry of Internal Affairs and Communications R&D projects for priority information and communication technologies. The breadth of public investment underscores how strategically important Japan considers the networking pathway to quantum computing.

For the broader field, the demonstration represents an important step toward networked quantum computers with the scalability needed for fault-tolerant universal quantum computing. The architectural analogy that comes to mind is the modern data center, where many computing modules work together as a single system, connected by high-bandwidth links. In the quantum version, multiple quantum processors would be interconnected through parallel photonic links, with entangled photons shuttling quantum information between modules. A multiplexed interface is essential to that picture because a single channel would throttle the connection between processors just as a single wire would cripple a classical cluster. By showing that ten atoms can each have their photons routed through dedicated waveguide channels with negligible crosstalk and preserved quantum correlations, the team has provided a template that others can now push toward the hundred-channel scale and beyond.

Challenges remain before photonic interconnects can knit together million-qubit machines. The waveguide array demonstrated here contains 32 channels, of which ten were used, and extending the technology to the roughly 100 channels the researchers anticipate will require further advances in fabrication, coupling efficiency, and detector multiplexing. The interface must also ultimately preserve entanglement between atoms and photons across the full collection-and-detection chain, not merely the correlations measured in this experiment. Yet the trajectory is clear: integrated photonics, neutral-atom arrays, and superconducting detectors are converging into a coherent hardware stack for quantum networking. As Yamamoto and his colleagues refine their multiplexed interface, the dream of quantum data centers, in which racks of atomic processors exchange entanglement the way classical servers exchange data packets, moves from whiteboard concept toward laboratory reality, one carefully guided photon at a time.

Subject of Research: Multiplexed quantum photonic interfaces for networking neutral-atom quantum computers

Article Title: A multiplexed quantum photonic interface for neutral-atom quantum computers

Article References: A multiplexed quantum photonic interface for neutral-atom quantum computers. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: quantum computing, neutral atoms, photonic interface, waveguide array, multiplexing, quantum networking, entanglement, superconducting nanostrip detectors, fault tolerance, qubits, University of Osaka, Optica

The heart of the design is a tool that quantum physicists have refined over the past two decades: optical tweezers. Unlike the tiny mechanical tongs used by watchmakers, optical tweezers are better imagined as science fiction tractor beams. A single, tightly focused laser beam creates a spot in space that attracts an atom, holding it in place without any physical contact. As Covey describes it, one beam and one focused spot are enough, and the atom simply falls into the light. Arrays of such beams can hold dozens or hundreds of individual atoms, each one acting as a qubit, the fundamental unit of quantum information. The approach has already produced some of the most impressive results in neutral-atom quantum computing, but the choice of atom matters enormously, and that is where helium comes in.

Trapping an atom with light requires the laser to supply enough energy to promote the atom from its ground state to an excited state, and the amount of energy needed depends on the species of atom. Hydrogen, the lightest element of all, is currently out of reach because the jump to its excited state demands more energy than modern technology can conveniently deliver. Helium is also a high-energy atom, but it possesses a crucial feature that hydrogen lacks: a second electron. That extra electron allows helium to settle into a temporary metastable state, an energy level sitting between the ground state and the excited state. Covey offers a vivid analogy: if reaching the excited state is like leaping onto a table, the metastable state is like pulling up a stepstool first. Atoms with only one electron on their outer shell, including hydrogen and lithium, have no such stepstool and must make the jump in a single bound.

The metastable state of helium is remarkable in its own right. In many other elements, comparable states last only seconds before the atom decays back down. Helium’s metastable state persists for roughly two hours, an extraordinarily long lifetime by atomic standards. That longevity gives experimentalists a generous window in which to laser-cool the atoms, trap them, and manipulate them before they decay. Helium’s well-resolved energy structure also makes it easier to laser-cool than lithium, according to Zheyuan Li, a PhD student in Covey’s lab and a co-first author of the paper. Cooling is an essential prerequisite for trapping, because atoms must be slowed almost to a standstill before optical tweezers can grip them reliably.

The second advantage is speed. Because helium is even lighter than lithium, the third-lightest element and the basis of the first fermionic quantum computing demonstrations earlier this year, quantum tunneling rates in a helium-based machine would be about three times faster at minimum, Li explained. Tunneling — the quantum phenomenon in which particles pass through barriers that classical physics says they cannot cross — underlies how atoms interact and exchange information in these simulators. Faster tunneling means faster transport of atoms and faster gate-like operations between them, which translates directly into a machine that can run more complex calculations before decoherence and other imperfections erode the fragile quantum states. In quantum computing, where coherence times are measured in fleeting fractions of a second, a threefold speedup is not a cosmetic improvement; it is a fundamental upgrade to the platform’s capabilities.

There is a subtlety in the choice of isotope, and it is here that the Chicago design makes its most distinctive move. Ordinary helium, the helium-4 that cools MRI machines and fills party balloons, is a boson, one of the two great families into which quantum particles divide. The other family, fermions, is named for University of Chicago legend Enrico Fermi, while bosons take their name from Indian physicist Satyendra Nath Bose. The two families obey fundamentally different rules: fermions cannot occupy the same quantum state at the same time, a prohibition known as the Pauli exclusion principle, while bosons are free to pile into identical states. That antisocial character of fermions turns out to be a blessing for quantum computing, because it frees fermionic machines from many of the errors that plague computers built from bosonic atoms.

Fermionic quantum computing has been a dream since the 1990s, but it was only cracked in early 2026, when two lithium-based models achieved the milestone. The Chicago team proposes to go one better by using helium-3, which has one fewer neutron than helium-4 and is therefore a fermion. Using the lightest fermionic atom that can be trapped means fermionic quantum computing can be implemented natively, Li said, rather than using bosonic atoms and then trying to simulate fermionic structure on top of them. Co-author Zoe Yan, an assistant professor of physics at UChicago who has previously worked with lithium-6, the next-heaviest fermion after helium-3, helped make the case that the extra effort of helium-3 is worthwhile. Hydrogen-1 is a boson, and while fermionic hydrogen-2, or deuterium, exists in principle, Covey noted that deuterium is far more difficult to work with than helium and is not even much lighter than helium-3, erasing most of the mass advantage that motivates the whole design.

The concept has drawn praise from outside the collaboration. Waseem Bakr, a professor of physics at Princeton University who was not involved in the research, said that by using the lightest trappable atom the work turns low mass into a real advantage, delivering faster tunneling, faster transport, and controllable motional qubits, and he called it a compelling blueprint for the next generation of fermionic quantum simulators. Such endorsements matter in a field where competing platforms — superconducting circuits, trapped ions, photonic chips, and neutral atoms — are racing to demonstrate practical advantage, and where the choice of physical substrate can determine how far a design can ultimately scale.

Of course, a concept paper is not a working machine, and the team’s next step is to build one. In collaboration with Yan, the researchers plan to trap and control individual helium-3 atoms for the first time. The foundation is there, Covey said, and progress is advancing to the point where the team hopes to have these atoms in tweezers for the first time probably within the next year or two. The path will begin, somewhat surprisingly, with the wrong isotope. Helium-3 is very expensive, explained Rupsa De, a PhD student in Covey’s lab and the paper’s other co-first author, so the team will start with helium-4 and then proceed toward helium-3. Because helium-4 is bosonic, the early experiments will validate the trapping and control techniques rather than the full fermionic architecture, but they will de-risk the hardest experimental steps before the costly isotope is committed.

Anyone worried that a quantum computer might worsen the world’s helium shortages can rest easy. Covey pointed out that the early experiments will have no impact on global helium-4 supplies, and the arithmetic is striking: two grams of helium, roughly enough to fill a single party balloon, contains a number of atoms that is a three followed by twenty-three zeroes. What the team ultimately wants is only tens of atoms, and a five-liter tank of helium, Covey noted, can last the lab for many years. The contrast between the scale of the ambition and the modesty of the material requirements is part of what makes the proposal so striking. If the team succeeds in suspending single fermionic helium atoms in beams of light over the next couple of years, the lightest trappable atom in the universe may find itself at the center of one of the most consequential technologies of the century, trading its party-trick reputation for a role in machines that compute with the strange rules of quantum mechanics itself.

Subject of Research: A proposed quantum computing architecture using laser-trapped metastable helium-3 atoms for fermionic quantum simulation

Article Title: Helium lifts new quantum computing concept

Article References: Helium lifts new quantum computing concept. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: helium-3, quantum computing, optical tweezers, fermions, metastable atoms, laser cooling, qubits, quantum tunneling, neutral atoms, PRX Quantum, University of Chicago, quantum simulation