Microwave-Optical Quantum Transduction

• Microwave-to-optical quantum transduction is the coherent conversion process that bridges microwave and optical photons using platforms like optomechanical, electro-optic, and atomic systems.
• It employs distinct architectures focused on optimizing conversion efficiency, reducing added noise, and maintaining wide bandwidth for effective quantum state transfer.
• Recent approaches leverage entanglement-mediated and heralded protocols to overcome challenges such as pump-induced heating and thermal noise in superconducting and optical interfaces.

Microwave-to-optical quantum transduction, or quantum frequency conversion, is the coherent conversion of quantum states between microwave and optical photons. It is central to architectures in which superconducting computation nodes operating at microwave frequencies are connected by long-distance transmission lines that transmit photons at optical frequencies, combining local microwave-domain control with optical-fibre or waveguide transmission over distance (Li et al., 4 Dec 2025, Tian, 2014). In current research, the subject spans optomechanical, electro-optic, magneto-optic, atomic-ensemble, and entanglement-mediated interfaces, and is organized chiefly around conversion efficiency, input-referred added noise, and bandwidth (Sekine et al., 30 Sep 2025, Aditto et al., 26 May 2026).

1. Generic theory and performance metrics

A standard description of one-stage transduction uses a microwave cavity mode $a_e$, an intermediate bosonic mode $b$, and an optical cavity mode $a_o$, with linearized interaction

$$H_{\rm int}=\hbar\,g\bigl(a_e\,b^\dagger + a_e^\dagger\,b\bigr) + \hbar\,\zeta\bigl(a_o\,b^\dagger + a_o^\dagger\,b\bigr).$$

Within the input-output formalism, the microwave-to-optical power efficiency is $\eta(\omega)=|S_{o,e}(\omega)|^2$, and on triple resonance it reduces to

$$\eta = \eta_e\,\eta_o\,\frac{4\,C_{em}\,C_{om}}{(1+C_{em}+C_{om})^2},$$

with $\eta_e=\kappa_{e,\rm ext}/\kappa_e$, $\eta_o=\kappa_{o,\rm ext}/\kappa_o$, and cooperativities $C_{em}=4g^2/(\kappa_e\kappa_m)$, $C_{om}=4\zeta^2/(\kappa_o\kappa_m)$. For direct electro-optic or atomic $b$0-type schemes, the same logic yields $b$1 with a direct microwave-optical cooperativity $b$2 (Sekine et al., 30 Sep 2025).

Bandwidth and noise enter on equal footing with efficiency. For one-stage transduction, the small-signal bandwidth is commonly expressed as $b$3, whereas in zero-stage electro-optic conversion $b$4 (Sekine et al., 30 Sep 2025). Cross-platform comparisons have increasingly emphasized the distinction between total efficiency $b$5 and internal efficiency $b$6, together with normalized added-noise measures such as

$b$7

because port coupling and internal loss can dominate quoted performance numbers even when the intrinsic mode-conversion process is strong (Aditto et al., 26 May 2026).

A recurrent source of confusion is therefore not the interaction Hamiltonian but the metric convention. Several recent reviews explicitly separate external efficiency, internal efficiency, added noise, and bandwidth, and note that some reported efficiencies—particularly in atomic and free-space settings—are not immediately comparable without specifying the precise input and output ports (Sekine et al., 30 Sep 2025).

2. Mechanical intermediaries: optomechanical, piezo-optomechanical, and HBAR routes

The canonical optoelectromechanical transducer couples a microwave cavity and an optical cavity to a shared mechanical resonator. In the radiation-pressure picture the full Hamiltonian contains $b$8 and $b$9; after strong red-detuned pumping and linearization, the interaction becomes a pair of beam-splitter couplings,

$a_o$0

In this framework, high-fidelity conversion requires resolved-sideband operation $a_o$1, low thermal occupancy $a_o$2, and impedance matching $a_o$3. The same theory also supports mechanical-dark-mode transfer, where a zero-eigenvalue mode $a_o$4 is decoupled from the mechanical resonator and can be adiabatically transported from microwave to optical sectors (Tian, 2014).

Recent experiments have focused on raising throughput while keeping $a_o$5. A crystalline-silicon platform using a high-impedance NbTiN microwave resonator, a phononic-crystal silicon nanobeam, and an optical cavity demonstrated quantum-enabled continuous conversion with $a_o$6 at input-referred added noise $a_o$7 quanta, and $a_o$8 at $a_o$9 quanta; at the higher-efficiency operating point, $$H_{\rm int}=\hbar\,g\bigl(a_e\,b^\dagger + a_e^\dagger\,b\bigr) + \hbar\,\zeta\bigl(a_o\,b^\dagger + a_o^\dagger\,b\bigr).$$0 and the efficiency-bandwidth product $$H_{\rm int}=\hbar\,g\bigl(a_e\,b^\dagger + a_e^\dagger\,b\bigr) + \hbar\,\zeta\bigl(a_o\,b^\dagger + a_o^\dagger\,b\bigr).$$1 (Zhao et al., 2024). This platform relies on electrostatic coupling to avoid the rf-pump heating typical of piezoelectric schemes and uses electromechanical radiative cooling so that $$H_{\rm int}=\hbar\,g\bigl(a_e\,b^\dagger + a_e^\dagger\,b\bigr) + \hbar\,\zeta\bigl(a_o\,b^\dagger + a_o^\dagger\,b\bigr).$$2 continuously (Zhao et al., 2024).

Equivalent-circuit analyses of mechanically mediated transducers have suggested that the remaining limitations are not fundamental. A mechanical-supermode architecture coupling electrical, piezoelectric, and optomechanical resonators found that simultaneously achieving $$H_{\rm int}=\hbar\,g\bigl(a_e\,b^\dagger + a_e^\dagger\,b\bigr) + \hbar\,\zeta\bigl(a_o\,b^\dagger + a_o^\dagger\,b\bigr).$$3 and $$H_{\rm int}=\hbar\,g\bigl(a_e\,b^\dagger + a_e^\dagger\,b\bigr) + \hbar\,\zeta\bigl(a_o\,b^\dagger + a_o^\dagger\,b\bigr).$$4 should be possible with current technology, provided that the electromechanical and optomechanical cooperativities are matched and the device is impedance-matched to the microwave transmission line (Wu et al., 2019). A related high-overtone bulk acoustic resonance scheme predicted $$H_{\rm int}=\hbar\,g\bigl(a_e\,b^\dagger + a_e^\dagger\,b\bigr) + \hbar\,\zeta\bigl(a_o\,b^\dagger + a_o^\dagger\,b\bigr).$$5 conversion efficiency with $$H_{\rm int}=\hbar\,g\bigl(a_e\,b^\dagger + a_e^\dagger\,b\bigr) + \hbar\,\zeta\bigl(a_o\,b^\dagger + a_o^\dagger\,b\bigr).$$6 using piezoelectric HBAR coupling and integrated $$H_{\rm int}=\hbar\,g\bigl(a_e\,b^\dagger + a_e^\dagger\,b\bigr) + \hbar\,\zeta\bigl(a_o\,b^\dagger + a_o^\dagger\,b\bigr).$$7 photonics (Blésin et al., 2021). Across these mechanical platforms, pump-induced heating, mechanical thermalization, and extraction loss remain the dominant constraints rather than the absence of a viable Hamiltonian.

3. Direct electro-optic conversion and multistage electro-optic variants

Electro-optic transducers use the $$H_{\rm int}=\hbar\,g\bigl(a_e\,b^\dagger + a_e^\dagger\,b\bigr) + \hbar\,\zeta\bigl(a_o\,b^\dagger + a_o^\dagger\,b\bigr).$$8 or Pockels interaction to couple a microwave mode directly to an optical sideband. In a triply resonant cavity electro-optic system, the strong pump mode $$H_{\rm int}=\hbar\,g\bigl(a_e\,b^\dagger + a_e^\dagger\,b\bigr) + \hbar\,\zeta\bigl(a_o\,b^\dagger + a_o^\dagger\,b\bigr).$$9, optical signal mode $\eta(\omega)=|S_{o,e}(\omega)|^2$0, and microwave mode $\eta(\omega)=|S_{o,e}(\omega)|^2$1 obey a linearized beam-splitter Hamiltonian

$\eta(\omega)=|S_{o,e}(\omega)|^2$2

with cooperativity $\eta(\omega)=|S_{o,e}(\omega)|^2$3 and bidirectional photon-conversion efficiency

$\eta(\omega)=|S_{o,e}(\omega)|^2$4

A superconducting radio-frequency cavity integrated with a lithium-niobate whispering-gallery resonator was designed to exploit long microwave coherence and tight electro-optic overlap, and the simulated device reaches $\eta(\omega)=|S_{o,e}(\omega)|^2$5 at pump power $\eta(\omega)=|S_{o,e}(\omega)|^2$6, with microwave linewidth $\eta(\omega)=|S_{o,e}(\omega)|^2$7 (Wang et al., 2022).

A complementary line of work attempts to suppress pump-heating by interposing a higher-frequency intermediate mode. In a two-step terahertz-mediated electro-optic scheme, a microwave mode at $\eta(\omega)=|S_{o,e}(\omega)|^2$8 is first converted to an intermediate THz mode and then to an optical mode at $\eta(\omega)=|S_{o,e}(\omega)|^2$9. Under matched-cooperativity conditions $$\eta = \eta_e\,\eta_o\,\frac{4\,C_{em}\,C_{om}}{(1+C_{em}+C_{om})^2},$$0 and $$\eta = \eta_e\,\eta_o\,\frac{4\,C_{em}\,C_{om}}{(1+C_{em}+C_{om})^2},$$1, the total external efficiency is

$$\eta = \eta_e\,\eta_o\,\frac{4\,C_{em}\,C_{om}}{(1+C_{em}+C_{om})^2},$$2

with a design-point value $$\eta = \eta_e\,\eta_o\,\frac{4\,C_{em}\,C_{om}}{(1+C_{em}+C_{om})^2},$$3, while the added noise is

$$\eta = \eta_e\,\eta_o\,\frac{4\,C_{em}\,C_{om}}{(1+C_{em}+C_{om})^2},$$4

yielding $$\eta = \eta_e\,\eta_o\,\frac{4\,C_{em}\,C_{om}}{(1+C_{em}+C_{om})^2},$$5 at $$\eta = \eta_e\,\eta_o\,\frac{4\,C_{em}\,C_{om}}{(1+C_{em}+C_{om})^2},$$6, $$\eta = \eta_e\,\eta_o\,\frac{4\,C_{em}\,C_{om}}{(1+C_{em}+C_{om})^2},$$7, and $$\eta = \eta_e\,\eta_o\,\frac{4\,C_{em}\,C_{om}}{(1+C_{em}+C_{om})^2},$$8 (Sahbaz et al., 2023). The explicit motivation is to break the positive feedback loop of single-step electro-optic transduction, where more pump power leads to more heating, more loss, and then still more pump power (Sahbaz et al., 2023).

Not all electro-optic platforms presently operate near the quantum regime. A silicon-organic hybrid device in a dilution refrigerator measured a single-photon electro-optic coupling $$\eta = \eta_e\,\eta_o\,\frac{4\,C_{em}\,C_{om}}{(1+C_{em}+C_{om})^2},$$9, a peak single-sideband conversion efficiency $\eta_e=\kappa_{e,\rm ext}/\kappa_e$0, and a full-width-half-max conversion bandwidth $\eta_e=\kappa_{e,\rm ext}/\kappa_e$1, with stray-light-induced quasiparticles identified as the main performance ceiling (Witmer et al., 2019). Taken together, electro-optic work shows a broad spread between early weak-coupling demonstrations, long-coherence cavity proposals, and multistage architectures designed specifically to control pump absorption.

4. Atomic, rare-earth, and spin-based transducers

Atomic and spin-based transducers replace the mechanical intermediary by collective or single-defect resonances that are simultaneously addressable in the microwave and optical domains. A representative rare-earth proposal uses a $\eta_e=\kappa_{e,\rm ext}/\kappa_e$2 system in $\eta_e=\kappa_{e,\rm ext}/\kappa_e$3 at zero magnetic field, with an effective dark-state Hamiltonian

$\eta_e=\kappa_{e,\rm ext}/\kappa_e$4

whose dark mode

$\eta_e=\kappa_{e,\rm ext}/\kappa_e$5

permits adiabatic microwave-to-optical transfer without populating the bright spin mode. Using experimentally realistic values, the analysis yields $\eta_e=\kappa_{e,\rm ext}/\kappa_e$6 and $\eta_e=\kappa_{e,\rm ext}/\kappa_e$7 at $\eta_e=\kappa_{e,\rm ext}/\kappa_e$8 (Asadi et al., 2022).

Characterization studies of erbium-doped crystals have addressed the spectroscopy needed to reach that regime. In $\eta_e=\kappa_{e,\rm ext}/\kappa_e$9, optical inhomogeneous linewidths of $\eta_o=\kappa_{o,\rm ext}/\kappa_o$0 and $\eta_o=\kappa_{o,\rm ext}/\kappa_o$1, an ensemble microwave coupling $\eta_o=\kappa_{o,\rm ext}/\kappa_o$2, and an initial classical microwave-to-optical conversion efficiency $\eta_o=\kappa_{o,\rm ext}/\kappa_o$3 were reported in a loop-gap-resonator geometry at $\eta_o=\kappa_{o,\rm ext}/\kappa_o$4 (Xie et al., 2021). By contrast, an on-chip $\eta_o=\kappa_{o,\rm ext}/\kappa_o$5 device at the single-photon level, using an effective resonant $\eta_o=\kappa_{o,\rm ext}/\kappa_o$6 nonlinearity of $\eta_o=\kappa_{o,\rm ext}/\kappa_o$7 and no engineered optical cavity, achieved chip-level efficiency $\eta_o=\kappa_{o,\rm ext}/\kappa_o$8 with added noise as low as $\eta_o=\kappa_{o,\rm ext}/\kappa_o$9 photons, and demonstrated interference of photons originating from two simultaneously operated transducers (Xie et al., 2024).

Atomic media also support very different operating points from solid-state rare-earth ensembles. A six-level Rydberg-atom transducer based on off-resonant scattering achieved a coherent microwave-to-optics conversion efficiency of $C_{em}=4g^2/(\kappa_e\kappa_m)$0 with bandwidth about $C_{em}=4g^2/(\kappa_e\kappa_m)$1, and maintained high efficiency for microwave pulses containing from thousands to about 50 photons, without requiring cavities or aggressive cooling to quantum ground states (Tu et al., 2022). At the other end of the design space, a memory-assisted transducer in low-doping $C_{em}=4g^2/(\kappa_e\kappa_m)$2 at $C_{em}=4g^2/(\kappa_e\kappa_m)$3 integrated transduction with a spin-echo memory protocol, reaching on-demand retrieval with $C_{em}=4g^2/(\kappa_e\kappa_m)$4 and $C_{em}=4g^2/(\kappa_e\kappa_m)$5 noise photons in the detection window at storage durations of $C_{em}=4g^2/(\kappa_e\kappa_m)$6 and $C_{em}=4g^2/(\kappa_e\kappa_m)$7, respectively, while also demonstrating multimode transduction capacity (Gautam et al., 2 May 2026). These results show that atomic and spin-based transduction is not a single architecture but a family ranging from free-space Rydberg scattering to cavity-assisted dark-state transfer and memory-enabled echo protocols.

5. Entanglement-mediated, heralded, and pump-free transduction

A major conceptual shift in the field is the move from direct state transfer to entanglement generation and teleportation-based conversion. In a piezo-optomechanical setting, direct conversion implements a bosonic thermal-loss channel $C_{em}=4g^2/(\kappa_e\kappa_m)$8, for which the pure-loss threshold $C_{em}=4g^2/(\kappa_e\kappa_m)$9 is necessary for positive quantum capacity. The same work shows that microwave-optical entanglement produced under blue-sideband driving can instead define an entanglement-based transduction channel whose quantum capacity remains positive in parameter regimes where the direct channel has zero quantum capacity (Zhong et al., 2022). In a cavity electro-optic setting, continuous-variable teleportation based on a two-mode-squeezing Hamiltonian $C_{om}=4\zeta^2/(\kappa_o\kappa_m)$0 was further shown to admit a positive transduction rate for all $C_{om}=4\zeta^2/(\kappa_o\kappa_m)$1, with higher fidelity or success probability than direct conversion for cat states and Gottesman-Kitaev-Preskill states (Wu et al., 2021).

Heralded transduction schemes make this logic explicit by generating microwave-optical Bell pairs rather than attempting deterministic in-line conversion. A pump-free protocol based on color centers first prepares optical spin-photon entanglement, then converts the spin state into a time-bin-encoded microwave photon using a Purcell-enhanced resonator with

$C_{om}=4\zeta^2/(\kappa_o\kappa_m)$2

With state-of-the-art parameters, the protocol yields for NV$C_{om}=4\zeta^2/(\kappa_o\kappa_m)$3 an optimal $C_{om}=4\zeta^2/(\kappa_o\kappa_m)$4, $C_{om}=4\zeta^2/(\kappa_o\kappa_m)$5, $C_{om}=4\zeta^2/(\kappa_o\kappa_m)$6, and $C_{om}=4\zeta^2/(\kappa_o\kappa_m)$7; for $C_{om}=4\zeta^2/(\kappa_o\kappa_m)$8, $C_{om}=4\zeta^2/(\kappa_o\kappa_m)$9, $b$00, $b$01, and $b$02 (Li et al., 4 Dec 2025). The stated advantage is the absence of a strong optical pump at the resonator, and therefore no unwanted device heating or quasiparticle generation in the superconducting circuit (Li et al., 4 Dec 2025).

A closely related but not fully pump-free approach uses defect-mediated scattering in diamond at extremely low optical power. In a double-resonant scattering scheme based on a single NV$b$03 center embedded in a diamond optomechanical resonator, coherent conversion reaches $b$04 at pump power $b$05, with conversion bandwidth $b$06, dark-count probability per trial $b$07, and a remote-entanglement single-click fidelity $b$08 at heralding rate $b$09 (Goto et al., 15 May 2026). This suggests a broader trend: rather than forcing all performance goals into a single continuous converter, several of the most recent proposals treat the transducer as an entanglement source or heralded network primitive.

6. Magneto-optic transduction, antiferromagnets, and cross-platform comparison

Magneto-optic transduction uses magnons as the intermediary between microwave and optical fields. Reviews of the area report that optomagnonic platforms presently exhibit the lowest efficiencies, typically $b$10 to $b$11, but also highlight intrinsic non-reciprocity, broadband magnonic operation, and theoretical enhancement routes up to $b$12 based on topological heterostructures and magnon squeezing (Aditto et al., 26 May 2026). This low-efficiency regime is not simply a lack of microwave-magnon coupling: in many YIG-like systems the bottleneck is the very weak optical-magnon coupling set by the Faraday interaction, even when microwave cooperativities are large (Sekine et al., 30 Sep 2025).

Antiferromagnets alter that landscape mainly by changing the magnetic operating condition. A recent theory of antiferromagnetic transduction derives analytical efficiencies for cases with and without an optical cavity and finds, unlike ferromagnetic-magnon schemes, that transduction can occur even in the absence of an external static magnetic field. In the cavity case the efficiency has a peak structure as a function of sample thickness, whereas without an optical cavity it increases monotonically with thickness. Using MnF$b$13-like parameters, the estimated single-mode efficiencies are $b$14 at $b$15 and $b$16 at $b$17 (Sekine et al., 2024). The absence of a dc magnetic bias is especially relevant for proximity to superconducting circuits.

Across all platforms, the literature does not support a single universally dominant architecture. One recent review states explicitly that no single platform today simultaneously meets all desiderata of high efficiency $b$18, sub-quantum added noise $b$19, wide bandwidth $b$20, and relaxed-temperature operation, and instead argues for a heterogeneous architecture combining optomechanical, electro-optic, and magneto-optic elements (Aditto et al., 26 May 2026). Another review organizes current bottlenecks around insufficient cooperativity, intrinsic loss, low single-photon coupling, and pump-induced heating, while identifying heralded entanglement generation and continuous-variable teleportation as viable ways to relax the direct-conversion requirement $b$21 for faithful state transfer (Sekine et al., 30 Sep 2025). In that sense, microwave-to-optical quantum transduction is not one technology but a set of partially overlapping strategies for interfacing superconducting quantum devices with optical quantum networks.