From Microwaves to Light: Linking Quantum Computers

Superconducting quantum processors operate with microwave excitations only a fraction of a degree above absolute zero. Long-distance networks communicate through light. Connecting the two requires a frequency converter that preserves quantum information without filling a refrigerator with heat. In 2026, researchers transmitted a coherent microwave signal between superconducting circuits housed in separate dilution refrigerators across a kilometre of fibre. The accomplishment reveals both the promise of modular quantum computing and the formidable physics still separating coherent signals from true quantum interconnects.

A large superconducting quantum computer is as much a refrigeration and wiring problem as a collection of qubits. Each module needs microwave control, readout, filtering, amplification and heat sinking. Increasing the number of cables burdens a dilution refrigerator's limited space and cooling power. Building several processors, each hosted in its own refrigerator, could relieve those limits if they can exchange quantum states or distribute entanglement efficiently. Copper microwave cables are excellent over laboratory distances, but kilometre-scale transmission at microwave frequencies is problematic. Optical fibre offers extraordinarily low attenuation at telecom wavelengths, provided a compatible cryogenic interface can be constructed. [1][2]

In March 2026, Yiyu Zhou and collaborators reported A 1-km photonic link connecting superconducting circuits in two dilution refrigerators in Nature Photonics. Their system employed two frequency-matched aluminum nitride electro-optic transducers joined by standard optical fibre. It transmitted microwave information across independent cryogenic setups with coherent phase preservation and demonstrated more than 0.1% on-chip conversion efficiency at each transducer. The authors report approximately 80 dB improvement compared with commercial electro-optic modulators. It was a coherent microwave-signal link, not high-fidelity quantum-state transfer between two remote qubits; recognizing this distinction is essential to interpreting the result. [1][3]

Quantum network diagram of superconducting quantum processors in dilution refrigerators joined through optical transducers and fiber — Illustration of a modular superconducting quantum network. Sekine, Murakami and Doi, *npj Nanophotonics* 3, 42 (2026), Figure 1, CC BY 4.0.

The mismatch between two perfectly good technologies

A superconducting circuit typically operates at a few gigahertz, often around $5\,\mathrm{GHz}$ . Telecom optical light near $1550\,\mathrm{nm}$ oscillates at approximately $193\,\mathrm{THz}$ . Photons in these frequency bands differ in energy by nearly five orders of magnitude. Optical photons propagate efficiently through room-temperature silica fibres; microwave quantum states can be extremely vulnerable to warm environmental excitations. The thermal occupation of an electromagnetic mode makes the difference explicit:

\bar n_{\mathrm{th}}(\omega,T)=\frac{1}{\exp(\hbar\omega/k_{\mathrm B}T)-1}.

For a $5\,\mathrm{GHz}$ microwave mode, $\hbar\omega/k_{\mathrm B}$ is about $0.24\,\mathrm K$ . At room temperature the thermal occupation is enormous compared with one photon; at a sufficiently cold and well-thermalized millikelvin stage it can be far below one. At optical frequencies, the same thermal occupation at room temperature is negligible. Yet a quantum network does not become noise-free simply by shifting its carrier into the optical band. Noise is also introduced by frequency conversion, optical pumping, loss and the environment at both endpoints.

How a crystal mixes microwaves and light

Electro-optic frequency conversion exploits the second-order nonlinear optical response of a material. An applied microwave electric field modifies the optical polarization and couples pairs of optical resonances. A strong optical pump supplies the large energy difference between the telecom and microwave bands. After linearizing around that pump and selecting a suitable resonant interaction, the simplest useful Hamiltonian is

H_{\mathrm{int}}/\hbar=G a^\dagger b+G^*ab^\dagger.

Here $a$ and $b$ denote optical signal and microwave annihilation operators, and $G$ is an optical-pump-enhanced coupling. The Hamiltonian exchanges excitations between the modes. It does not generate a copy of an arbitrary unknown quantum state. Real conversion also depends on phase matching, cavity linewidths, detunings, optical polarization and suppression of unwanted Stokes scattering that can produce excess noise. [2][4]

The engineering constraints are severe because the optical pump operates near the same cryogenic stage as delicate superconducting circuitry. Heating from optical absorption, pump leakage, photo-generated quasiparticles, optical fibre packaging and electrical loss can matter as much as the mathematical coupling rate. A transducer with a high internal photon-conversion probability may still have poor performance once optical in-coupling, microwave external coupling and packaging loss are included. That is why efficiency must be reported with a clear definition of the measurement plane.

Principle of microwave-to-optical frequency conversion using optical pumping and a quantum channel with efficiency and noise — Quantum frequency conversion and the efficiency–noise description. Sekine, Murakami and Doi, *npj Nanophotonics* 3, 42 (2026), Figure 2, CC BY 4.0.

What the kilometre demonstration accomplished

The 2026 experiment linked superconducting circuits in two distinct dilution refrigerators using aluminum nitride electro-optic devices. A major fabrication obstacle is that two separately fabricated optical resonators seldom operate at precisely the same frequency. If the frequency emitted by one cannot be accepted by the other, excellent individual converters may form a poor network. The researchers adopted an asymmetric photonic molecule design: coupled optical resonances were engineered to permit frequency matching between the independent devices. [1]

The investigators characterized the microwave–optical–microwave scattering response and used phase-sensitive measurements to verify classical coherence through the full link. This demonstrates that a microwave signal can leave one cryostat, cross a kilometre of glass as optical modulation and be recovered at another cryostat without losing its coherent phase relationship. The experiment does not report the fidelity of a transmitted single-photon qubit. Classical coherence of a strong tone can survive an inefficient channel that would erase nearly every individual microwave excitation.

Two converters multiply their losses. Denote the conversion efficiencies by $\eta_1$ and $\eta_2$ , with fibre power transmission $T_{\mathrm f}$ . Ignoring ancillary coupling and detection losses for a moment, an approximate total efficiency is

\eta_{\mathrm{link}}=\eta_1 T_{\mathrm f}\eta_2.

If both endpoints operate at $0.1\%$ efficiency, the combined conversion is about $10^{-6}$ even before fibre attenuation: around $-60\,\mathrm{dB}$ of photon-number transmission. That can still support a measurable coherent classical signal with suitable amplification and detection. But a qubit state encoded in a single photon cannot be deterministically transferred with high probability through such an inefficient link. The distinction between a substantial improvement over an older benchmark and meeting a quantum-state-transfer target is central to honest evaluation of interconnects.

Why optical fibre changes the architecture

Silica fibre has a well-developed infrastructure, excellent attenuation in its telecom window and flexible routing. It permits processors to be physically separated rather than packaged into a single cooling vessel. However, classical optical amplifiers cannot be dropped into a quantum network without consequence: noiseless deterministic amplification of an unknown quantum state is forbidden by quantum mechanics. Depending on the protocol, distant processors may rely on direct conversion, entanglement swapping, heralding or purification instead of assuming every transmitted photon arrives intact. [2]

Optical fibre strands shown in a close-up photograph — Optical fibre, a low-attenuation carrier for long-distance telecommunications. Global Gamers, Wikimedia Commons, CC BY 2.0. This photograph is not from the 2026 cryogenic-link experiment.

Fibre interconnects could also change the heat budget. A room-temperature optical pathway need not carry the same conductive thermal load as a long bundle of cryogenic coaxial cables, but every optical pump, cold photonic component and absorber creates additional opportunities for heat deposition. Scaling requires accounting for laser power, fibre feedthroughs, phase stabilization, optical filters, cold-stage optical absorption, microwave line heat sinking and electrical packaging. A network architecture that reduces one load only to introduce a larger optical load has not solved the cryogenic problem.

Efficiency, noise and the actual quantum channel

An idealized passive frequency converter can be approximated as a bosonic pure-loss channel,

a_{\mathrm{out}}=\sqrt{\eta}\,b_{\mathrm{in}}+\sqrt{1-\eta}\,v,

where $v$ is a vacuum environmental mode. In a real device, thermal reservoirs, pump-induced photons, imperfect sideband selectivity and internal losses contribute additional channels. The full input-output description must preserve bosonic commutation relations and specify how noise is referred to the input. Quoted efficiencies without added-noise measurements do not determine whether a device can faithfully transmit a fragile quantum state. [2]

Input output quantum transducer showing electromagnetic modes, coupling channels and thermal reservoirs — Input-output theory of microwave, optical and thermal environmental modes. Sekine, Murakami and Doi, *npj Nanophotonics* 3, 42 (2026), Figure 3, CC BY 4.0.

For an unassisted ideal bosonic pure-loss channel, quantum capacity vanishes when transmissivity is at or below $1/2$ . That threshold is not a universal prohibition on quantum communication below 50% transmission. Probabilistic heralded entanglement generation and repeaters can use much lower-efficiency links at the cost of heralding rates, memory demands and classical communication. Quantum transducers therefore need to be assessed relative to a specific distributed-computing protocol. Some protocols require direct high-fidelity state transfer; others can tolerate low success probability if successful events are identified reliably.

Electro-optic, mechanical and other routes

There are several ways to couple the two frequency bands. Electro-optic transducers directly mix microwave and optical modes through a nonlinear susceptibility; they avoid the additional thermal reservoir of a mechanical mediator but often require strong optical pumping. Electro-optomechanical systems transfer excitations through a vibrational mode, exploiting piezoelectric and radiation-pressure interactions. Their performance depends on the mechanical linewidth and the relative strengths of electromechanical and optomechanical couplings. The 2026 review by Sekine, Murakami and Doi compares these approaches alongside magnonic and atomic-ensemble transducers. [2]

For an idealized three-mode electromechanical–optomechanical converter operated on resonance, a frequently used internal efficiency factor is

\eta_{\mathrm{int}}=\frac{4C_{\mathrm{em}}C_{\mathrm{om}}}{(1+C_{\mathrm{em}}+C_{\mathrm{om}})^2},

where the cooperativities measure interaction strengths relative to losses. A strong optical interaction alone cannot compensate indefinitely for a poorly coupled microwave side. Matching interactions, minimizing dissipation and ensuring high external coupling are equally important. This formula is specific to that approximate three-mode model; it is not a universal efficiency law for direct electro-optic chips. [2][4]

The missing experiment: move an unknown quantum state

The decisive milestone for superconducting quantum networking is not simply a long fibre or a coherent transmitted tone. It is preservation of nonclassical information across an end-to-end channel: a state-transfer fidelity, an entangled link between cryogenic nodes, or a properly characterized heralded protocol. Earlier microwave–optical experiments have demonstrated nonclassical correlations and entanglement in shorter or different architectures, while the 2026 two-refrigerator result addresses the difficult engineering of phase-matched distributed hardware. These capabilities must eventually coexist in one scalable platform. [1][6]

That future platform will also need robust fabrication yield. Individual devices can be tuned in a laboratory, but connecting hundreds of cryogenic processors requires repeatable optical resonances, automated stabilization, reliable cold fibre packaging and well-characterized thermal loading. The asymmetric photonic-molecule design addresses precisely one practical source of incompatibility between separately manufactured converters. A useful architecture must combine that manufacturability with a quantum noise budget and network protocols that can sustain error-corrected operations.

The kilometre photonic link marks a meaningful transition from a single converter on a chip to an inter-refrigerator experiment. It demonstrates that microwave coherence can cross an optical carrier and return to a cold superconducting circuit. Whether the same kind of system can transmit quantum states with low enough loss and noise will determine how far modular superconducting quantum computers can be distributed before physics, not fibre infrastructure, becomes the dominant constraint.

References

[1] Zhou, Y., Wu, Y., Li, C. et al., A 1-km photonic link connecting superconducting circuits in two dilution refrigerators, Nature Photonics 20, 579–585 (2026). Original experiment.

[2] Sekine, A., Murakami, R. and Doi, Y., Microwave-to-optical quantum transduction of photons for quantum interconnects, npj Nanophotonics 3, 42 (2026). Open-access review.

[3] Brookhaven National Laboratory, Connecting Superconducting Circuits Over a Kilometer of Optical Fiber (2026). Research briefing.

[4] Han, X. et al., Microwave-optical quantum frequency conversion, Optica 8, 1050–1064 (2021). Review.

[5] Shen, M. et al., Photonic link from single-flux-quantum circuits to room temperature, Nature Photonics 18, 371–378 (2024). Cryogenic optical-interface study.

[6] Meesala, S. et al., Quantum Entanglement between Optical and Microwave Photonic Qubits, Physical Review X 14, 031055 (2024). Quantum benchmark.

Cover photograph: seven-channel optical waveguide package with fibre connections, by Optoscribe. Wikimedia Commons, CC BY-SA 3.0.