Quantum hardware has a frequency problem that no single device technology can solve alone. The processors that lead the field today — superconducting circuits — compute with microwaves, in the gigahertz range. The carriers best suited to moving quantum information over distance — photons in optical fiber — live five orders of magnitude higher in energy. Connecting the two requires a transducer: a device that converts a quantum state from one carrier to another without destroying the information it holds. A result published September 14, 2026, in Nature Materials, from a City College of New York–led collaboration, demonstrates a new way to build one. Using the layered magnetic semiconductor chromium sulfide bromide, CrSBr, the team converted a microwave signal into a coherent optical signal — not by bolting a magnet onto a resonator, but by letting the crystal itself do the converting.
The paper, “Microwave-to-optical transduction using magnon–exciton coupling,” is by Pratap Chandra Adak, who led the study as a postdoctoral researcher in Vinod Menon’s group at CCNY, with collaborators at the CUNY Advanced Science Research Center, Columbia University, the University of Chemistry and Technology Prague, the University of Chicago, and RPTU Kaiserslautern-Landau. Its claim deserves to be stated precisely, because the distance between what was demonstrated and what a quantum network needs is the whole story. The team showed coherent microwave-to-optical conversion through magnon–exciton coupling in a bulk crystal, with no optical or microwave resonator. It did not transfer individual quantum states, and it did not establish the efficiency or added-noise performance a quantum transducer requires. What it established is a route, with a measured coupling strength, a measured bandwidth, and a scaling law that says where the improvements should come from.
Why a quantum transducer is a harder problem than a frequency converter
A quantum transducer faces stricter demands than a classical converter, and the paper states them plainly: near-unity conversion efficiency, less than one added noise photon, and bandwidth compatible with qubit coherence. Fail any one of the three and the converted state is either lost, swamped, or too slow to be useful.
Existing approaches each sacrifice something. Electro-optic transducers, based on the Pockels effect, offer a direct and potentially low-noise path but are limited in efficiency by the weakness of the electro-optic interaction, with megahertz-scale bandwidths. Optomechanical transducers, which use a vibrating mechanical mode as the intermediary, have reached internal efficiencies around 47 percent, but their performance is constrained by the thermal occupation of the mechanical mode and by bandwidths often in the kilohertz to 10-kilohertz range in high-efficiency operation. Magnon-based approaches — using collective spin waves as the intermediary — have a different appeal: magnons can be tuned with a static magnetic field, their dynamics are intrinsically broadband, and at gigahertz frequencies their thermal occupation is low, which favors low-noise operation. But until now, magnon-based conversion has mostly relied on off-resonant magneto-optical effects, such as Faraday rotation, where the interaction is intrinsically weak and measurable conversion demands large device volumes or high-finesse cavities.
The CrSBr result changes the interaction, not just the device. Instead of an off-resonant effect, it exploits strong light–matter interaction at an exciton resonance: the crystal’s own optical transitions become the readout for its magnetic dynamics.
The mechanism: a magnon that moves an exciton
CrSBr is a layered, orthorhombic crystal whose layers stack along the crystallographic c-axis. Below its Néel temperature of 132 K it is an A-type antiferromagnet: strong intralayer exchange aligns the chromium spins ferromagnetically along the b-axis, while much weaker interlayer exchange aligns successive layers antiparallel. That weak interlayer coupling is a gift. It lets CrSBr support magnons at gigahertz frequencies even without an applied field, where traditional antiferromagnets typically resonate in the terahertz range — and gigahertz is exactly where superconducting qubits live.
The magnetism–optics link runs through the exciton energies. Interlayer electron hopping in CrSBr depends on the relative spin alignment of neighboring layers: spin-forbidden in the antiferromagnetic configuration, progressively allowed as a field cants the spins toward ferromagnetic alignment. So the exciton energies depend on the canting angle between the two magnetic sublattices. Apply a field along the c-axis and the exciton resonances redshift — by as much as about 120 meV for the high-energy exciton near 1.8 eV, and about 20 meV for the low-energy exciton near 1.4 eV, up to the saturation field of about 1.9 tesla.
Transduction is then a two-step coherent process. First, a microwave signal delivered through a coplanar waveguide resonantly drives a uniform magnon mode: all spins precess, and in the optical mode this precession modulates the instantaneous canting angle at the driving frequency. Second, that time-dependent canting modulates the exciton energies and therefore the refractive index near the excitonic resonances, so a laser slightly detuned from an exciton resonance, reflected from the crystal, acquires sidebands offset by the magnon frequency. The microwave signal is now riding on light.
Two features of this mechanism are signatures, not conveniences. Only the optical magnon mode, in which the two sublattices precess out of phase and the canting angle oscillates, modulates the exciton energy; the acoustic mode, in which they precess in phase, leaves the canting angle unchanged. And because CrSBr’s exciton transition dipoles align along the b-axis, the effect appears for b-polarized light and vanishes into the noise floor for a-polarized light. Both confirm that the signal comes from magnon–exciton coupling rather than generic microwave pickup.
What the team actually measured
The experiment used a bulk crystal — about 300 micrometers thick, 0.5 millimeters wide, and 2 millimeters long — placed on a coplanar waveguide inside a cryostat, b-axis aligned to the waveguide’s center conductor. Microwave transmission showed two magnon branches below the saturation field, with linewidths typically around 50 to 300 megahertz — already a hint of the bandwidth to come — and magnon modes persisting even at zero applied field, thanks to the crystal’s large magnetocrystalline anisotropy. White-light reflectivity showed the two dominant excitons, near 1.8 and 1.4 eV, both redshifting with field as expected.
The transduction test then looked for the intersection of the two. Maps of the microwave-induced reflectivity change showed a pronounced change precisely where the magnon frequency met the exciton energy, at each applied field. The change varied linearly with drive power and remained robust up to about 20 K, and tracking the peak frequency across fields reproduced the magnon dispersion — the optical signal follows the magnet, as it must if magnon–exciton coupling is the mechanism.
Coherence was tested with homodyne detection. A continuous-wave diode laser at 1.8 eV, near the high-energy exciton, was split into signal and local-oscillator arms; the signal beam reflected from the crystal, acquired magnon sidebands, and recombined with the local oscillator for readout by a fast photodiode and RF spectrum analyzer. At 0.5 tesla, with the exciton tuned near resonance with the laser, the team observed a robust transduction signal at zero detuning from the magnon sideband. Sweeping the drive frequency gave the headline bandwidth: the response extends over about 300 megahertz, consistent with the magnon linewidth. Sweeping the field showed the transduction frequency tracking the magnon dispersion — the operating frequency is tunable with a magnet, not fixed by a fabricated resonator.
Converting the measured homodyne sideband into photon fluxes, the team extracted a lower limit on the transduction efficiency of 3.3 × 10⁻¹². Read that number for what it is: twelve orders below unity, and a lower bound rather than a device efficiency, because the power actually absorbed by the crystal cannot be fully disentangled from other dissipation pathways. No added-noise figure was reported. By the paper’s own criteria, this device is not a quantum transducer — it is a demonstration that the conversion mechanism works coherently, with the shortfall quantified rather than hidden.
The paper attributes the low efficiency to geometry: the microwave drive excites magnons over the full crystal volume, while the optical probe samples only a micron-scale region, with no cavity or resonator enhancement. That is simultaneously the result’s weakness and its point — conversion was observed without any resonator at all, which means the coupling is intrinsic to the material rather than engineered into a structure.
The coupling strength, and the scaling law that is the real result
The more consequential number may be the single-particle magnon–exciton coupling rate: g₀/2π ≈ 5.5 kHz for the high-energy exciton in the bulk crystal — approximately three orders of magnitude larger than the magneto-optical coupling typically observed in yttrium iron garnet (YIG), the workhorse material of magnon photonics.
More important than the value is its scaling. The coupling is inversely proportional to the square root of the magnetic volume, because the zero-point fluctuation of the canting angle grows as the volume shrinks. In a bulk crystal each magnon moves the canting angle only slightly; in the few-layer limit, the paper estimates the coupling could reach the megahertz regime — surpassing most current optomechanical and electro-optic converters. The layered nature of CrSBr makes this accessible: it can be thinned toward a few layers while retaining its magnetic and optical properties, with linewidths relatively unchanged.
The cooperativity tells the same story in the language device physicists use. In the bulk crystal, the single-exciton cooperativity is about 10⁻¹⁴ — modest, constrained by broad excitonic linewidths. Using experimentally reported linewidths for low-dimensional CrSBr structures, the paper estimates about 10⁻⁹ in typical flakes and up to 10⁻⁴ in idealized bilayer implementations. Those numbers are still far from the cooperativity of order unity that efficient transduction demands, but the trajectory is visible: volume confinement plus pump enhancement plus cavity integration, each a known engineering step, rather than a new physical effect waiting to be discovered.
Exciton-polaritons offer a second lever on the same problem. CrSBr’s large refractive-index contrast with air forms an effective Fabry-Pérot cavity, producing multiple self-hybridized exciton-polariton resonances near both excitons without any fabricated structure. The team showed that microwave-induced reflectivity changes appear across multiple polariton branches: the hybrid states inherit the magnon coupling. Because a polariton borrows photonic character, its linewidth can be narrower than the bare exciton’s, mitigating the dissipative loss that caps cooperativity, while the coupling survives in proportion to the polariton’s excitonic fraction. Polariton engineering also widens the practical optical window, since conversion need not sit exactly on a bare exciton resonance.
What this does and does not change
Placed in context, the result changes the status of magnon-based transduction. Before it, the magnon route was bottlenecked by weak off-resonant interactions demanding large volumes or high-finesse cavities. After it, there is a demonstrated resonant mechanism, intrinsic to a single layered material, with a coupling rate three orders above the YIG baseline, a 300-megahertz bandwidth — broad by transducer standards — and a field-tunable operating frequency needing no fabricated resonator. A single van der Waals crystal hosts the magnetic, electronic, and optical degrees of freedom simultaneously — the platform claim, and the part most likely to survive contact with future work.
What it does not change is equally clear. It does not demonstrate quantum-state transfer, does not measure added noise, and does not approach the efficiency a quantum network requires — the measured lower bound sits twelve orders below unity. The operating regime is cryogenic: the signal was characterized at 2 K and remained robust only to about 20 K, and the crystal’s magnetic order vanishes above 132 K. Nor does it settle which transduction platform will win; each approach trades efficiency, noise, bandwidth, and integrability differently.
One distinction, because CrSBr appears in more than one headline: this is a transduction demonstration by a CCNY-led team, built on magnon modulation of exciton energies in a bulk crystal — a different result, from a different team, than separate work on magnetic control of a CrSBr polariton condensate. Conflating them would obscure what each team actually showed.
What to watch next
The most informative follow-ups are the ones the paper names. First, thinner crystals: does the coupling climb toward the megahertz regime in few-layer flakes as the volume scaling predicts? Second, resonators: does adding a microwave resonator, an optical cavity, or both raise the cooperativity from the 10⁻¹⁴ bulk value toward unity, and does the efficiency follow? Third, noise: does anyone measure added noise, the figure this paper does not report and the one that decides whether the route is quantum-capable? Fourth, polaritons: do engineered devices deliver narrower linewidth with preserved coupling in a fabricated cavity, rather than in the crystal’s self-hybridized resonances? And fifth, the system test every platform faces: a converted state, characterized at the single-quantum level, with efficiency and added noise quoted together on the same device.
The deeper lesson is familiar: excitons are strongly coupled to light, which makes them fast and optically accessible, and strongly coupled to their environment, which makes them lossy. Progress usually comes not from finding a material without tradeoffs, but from finding one whose tradeoffs can be engineered — here, a magnet whose spin waves move its excitons, in a crystal thin enough to make each magnon count. The CCNY team did not build a quantum transducer. They showed that a layered magnet can translate between microwaves and light coherently, without a resonator, and wrote down, with numbers, what it would take to go further — a mechanism with a roadmap, and in this field that is a meaningful kind of progress.
Sources and further reading
- Pratap Chandra Adak et al., “Microwave-to-optical transduction using magnon–exciton coupling,” Nature Materials (published September 14, 2026). DOI: 10.1038/s41563-026-02748-7.
- Pratap Chandra Adak et al., “Microwave-to-optical transduction using magnon–exciton coupling in a layered antiferromagnet,” arXiv:2604.03441 (submitted April 3, 2026). All coupling rates, bandwidths, crystal dimensions, field values, exciton energies, efficiency bound, and cooperativity estimates quoted above are taken from this paper.
- City College of New York announcement, “A new bridge for quantum networks: CCNY physicists convert microwaves to light using 2D magnets,” as reported October 26, 2026: the CCNY-led collaboration, Adak’s role in Menon’s group, the bulk-crystal-without-resonator observation, and the proposed routes toward a real transducer.
- Prior CrSBr literature cited in the paper on GHz magnons, exciton–magnetic-order coupling, and self-hybridized exciton-polaritons in the ultrastrong coupling regime.
- Note on verification: the 300 MHz bandwidth, 5.5 kHz coupling rate, 3.3 × 10⁻¹² efficiency lower bound, cooperativity estimates, and field and temperature values are taken from the paper itself. No added-noise figure is quoted because the paper reports none; the efficiency figure is a lower bound, not a device efficiency, and should not be quoted as one. Device specifics not stated in the paper — such as the exact microwave drive frequencies used at each field — are deliberately not quoted.