Quantum Frequency Conversion (QFC)

• Quantum frequency conversion (QFC) is the coherent translation of quantum light between distinct optical bands while preserving its quantum state, including coherence and entanglement.
• It utilizes nonlinear interactions such as χ(2) three-wave mixing and χ(3) four-wave mixing to bridge spectral mismatches between quantum emitters and telecom networks.
• QFC applications include network interfacing, wavelength routing, and mode-selective conversion, enabling high-dimensional state preservation and integration into scalable quantum systems.

Searching arXiv for recent and foundational work on quantum frequency conversion relevant to the requested encyclopedia entry. Quantum frequency conversion (QFC) is the coherent translation of a quantum light field from one optical frequency band to another while preserving the quantum information carried by the field, including non-classical correlations, coherence, and, in the ideal limit, entanglement and photon statistics. In practice, QFC is realized through nonlinear optical interactions—most commonly $\chi^{(2)}$ three-wave mixing and, in some platforms, $\chi^{(3)}$ four-wave mixing—to bridge the spectral mismatch between quantum emitters or memories operating in the visible or near-visible and the low-loss telecom bands used for long-distance fiber transmission (Brevoord et al., 1 Sep 2025, Zaske et al., 2012).

1. Conceptual scope and historical role

QFC occupies a specific place in quantum photonics: it is not primarily a source, detector, or memory, but an interface. Its function is to connect physical systems whose native optical transitions are incompatible with the transmission or processing layer of a quantum network. This role is explicit in visible-to-telecom demonstrations for quantum dots, color centers, rare-earth memories, and trapped ions, where the emitter or memory remains at its natural wavelength while the photonic channel is translated into the telecom O-, S-, C-, or L-band (Zaske et al., 2012, Brevoord et al., 1 Sep 2025, Wengerowsky et al., 2024, Yang et al., 2 Mar 2026).

A central misconception is to treat QFC as ordinary wavelength translation. In the quantum regime, the relevant criterion is not merely spectral relocation but preservation of the quantum state. For that reason, the literature repeatedly evaluates QFC by metrics such as $g^{(2)}(0)$, first-order coherence, process fidelity, temporal decay preservation, cross-correlation, or entanglement visibility, rather than by optical power conversion alone (Zaske et al., 2012, Brevoord et al., 1 Sep 2025, Yamada et al., 22 Apr 2026).

The field also extends beyond simple point-to-point wavelength translation. Recent work treats QFC as a programmable frequency-domain network primitive: it can retune inhomogeneous solid-state emitters to a common telecom frequency, implement channel-selective routing onto DWDM grids, sort temporal modes, and preserve high-dimensional orbital-angular-momentum or polarization-encoded states across frequency translation (Arizono et al., 2024, Yamada et al., 22 Apr 2026, Manurkar et al., 2016, Liu et al., 2019).

2. Nonlinear-optical foundations

Most practical QFC implementations use three-wave mixing in a $\chi^{(2)}$ medium. For difference-frequency generation, the output frequency obeys

$$\omega_{\text{out}} = \omega_{\text{p}} - \omega_{\text{in}},$$

while for sum-frequency generation it obeys

$$\omega_{\text{out}} = \omega_{\text{in}} + \omega_{\text{p}}.$$

In wavelength form, DFG is commonly written as

$\frac{1}{\lambda_{\text{in}} - \frac{1}{\lambda_{\text{p}} = \frac{1}{\lambda_{\text{out}}},$

with the corresponding momentum-conservation or phase-matching condition

$k_3 = k_1 \pm k_2$

or, in quasi-phase-matched media,

$k_3 = k_1 \pm k_2 + \frac{2\pi}{\Lambda}.$

These relations appear across bulk-crystal, periodically poled waveguide, and microring implementations (Zaske et al., 2012, Arizono et al., 2024, Rütz et al., 2016).

In the undepleted-pump approximation, the quantum description reduces to a beam-splitter-like interaction in frequency space. One representative form is

$$\hat{H}_\text{int} \propto \kappa \left(\hat{a}_{\text{out}}^\dagger \hat{a}_{\text{in}}\, \alpha_p + \text{h.c.}\right),$$

which makes explicit that an input photon is annihilated and a converted photon is created with an amplitude set by the classical pump field (Zaske et al., 2012). A related formulation used for channel-selective DFG writes

$\chi^{(3)}$0

again emphasizing the frequency-domain beamsplitter structure (Arizono et al., 2024).

Efficiency usually follows a sine-squared law with pump power and interaction length. Representative forms include

$\chi^{(3)}$1

or

$\chi^{(3)}$2

with deviations arising from loss, thermal loading, imperfect mode overlap, and phase-matching drift (Zaske et al., 2012, Arizono et al., 2024, Brevoord et al., 1 Sep 2025).

Although most experimental systems are non-resonant or cavity-enhanced $\chi^{(3)}$3 devices, QFC is not limited to that regime. Resonant four-wave-mixing models based on EIT show that, in principle, near-unity conversion can coexist with low vacuum-field noise, with the converted field inheriting the wave function and quadrature variance of the input when conversion efficiency approaches 100% (Cheng et al., 2020). This suggests that the dichotomy between efficient conversion and low-noise operation is platform-dependent rather than fundamental.

3. Materials, device classes, and architectures

The dominant material systems are periodically poled lithium niobate and related $\chi^{(3)}$4 crystals, but the architectural diversity is now substantial. Zn:PPLN ridge waveguides were used for visible-to-telecom conversion of 711 nm quantum-dot photons to 1313 nm, with quasi-phase matching set by poling period and temperature tuning (Zaske et al., 2012). Bulk and waveguide PPKTP devices have enabled interfaces reaching the ultraviolet, including a monolithic telecom-to-369.5 nm converter targeting the Yb$\chi^{(3)}$5 transition (Rütz et al., 2016).

Bulk-crystal cavity-assisted systems remain relevant where high intracavity pump powers are advantageous. A recent SnV implementation combined 619 nm photons with a 1064 nm pump in a pump-resonant bow-tie cavity containing bulk monocrystalline KTA and converted the output to 1480 nm; the single photon itself remained single-pass, while only the pump was cavity-enhanced (Brevoord et al., 1 Sep 2025).

Thin-film lithium niobate has become a central integrated platform. LNOI nanophotonic waveguides demonstrated 1531 nm to 860 nm upconversion with 73% internal efficiency and 900 cps on-chip noise, and also supported an on-chip upconversion detector (Wang et al., 2022). Periodically poled TFLN microrings have pushed the resonant extreme: one hybrid-integrated chip reported 57% on-chip quantum efficiency, 386,000 %/W normalized efficiency, and only 360 $\chi^{(3)}$6W pump power by exploiting a triple-resonant periodically poled microring (Wang et al., 4 Sep 2025).

A different route avoids periodic poling entirely. Unpoled InGaP nanophotonic waveguides use modal phase matching rather than QPM, eliminate poling-induced SPDC noise, and support bidirectional 1550 nm $\chi^{(3)}$7 780 nm conversion with record-low pump power for non-resonant integrated QFC (Hu et al., 19 Oct 2025). At shorter wavelengths, first-order quasi-phase matching on TFLN has been pushed to a 3.07 $\chi^{(3)}$8m poling period for 393 nm to telecom conversion, with explicit fabrication tolerances tied to domain-defect statistics (Yang et al., 2 Mar 2026).

These platforms imply different engineering priorities. Bulk systems emphasize thermal stability, pump enhancement, and low-noise filtering; waveguides emphasize overlap, coupling, and fabrication tolerance; microrings emphasize triple resonance, quality factors, and coupling design; fiber-integrated modules emphasize packaging and deployment (Brevoord et al., 1 Sep 2025, Wang et al., 4 Sep 2025, Liao et al., 25 Apr 2026).

4. Performance metrics and representative benchmarks

The reported performance of QFC depends strongly on which efficiency is being quoted. Internal efficiency refers to the nonlinear interaction itself; external or total efficiency includes coupling and filter losses; on-chip efficiency typically excludes off-chip optics but includes propagation and coupler losses. Noise may be quoted as counts per second, counts per second per picometer, or as a derived SNR proxy such as $\chi^{(3)}$9 (Brevoord et al., 1 Sep 2025, Wang et al., 2022, Wengerowsky et al., 2024).

Representative reported systems illustrate the present operating envelope.

System	Platform and process	Reported performance
SnV 619 nm $g^{(2)}(0)$0 1480 nm	KTA, cavity-enhanced pump DFG	Internal $g^{(2)}(0)$1; external $g^{(2)}(0)$2; $g^{(2)}(0)$3 cts/s/pm; efficiency above 80% of maximum over 70 GHz (Brevoord et al., 1 Sep 2025)
Quantum dot 711 nm $g^{(2)}(0)$4 1313 nm	Zn:PPLN ridge-waveguide DFG	$g^{(2)}(0)$5 external; $g^{(2)}(0)$6; SNR $g^{(2)}(0)$7 (Zaske et al., 2012)
Telecom 1531 nm $g^{(2)}(0)$8 860 nm	LNOI SFG chip	73% internal efficiency; 900 cps on-chip noise; upconversion detector with 8.7% detection efficiency and 300 cps noise (Wang et al., 2022)
Telecom $g^{(2)}(0)$9 visible on chip	Periodically poled TFLN microring SFG	57% on-chip quantum efficiency; 386,000 %/W; 360 $\chi^{(2)}$0W pump; noise $\chi^{(2)}$1k cps (Wang et al., 4 Sep 2025)
606 nm $\chi^{(2)}$2 1552 nm, $\chi^{(2)}$3s-long pulses	MgO:ppLN DFG with 12.5 MHz filtering	Device efficiency about 25%; SNR $\chi^{(2)}$4 for 10 $\chi^{(2)}$5s-long pulses containing one photon on average (Wengerowsky et al., 2024)
393 nm $\chi^{(2)}$6 telecom C-band	First-order-QPM TFLN DFG	28.8% external efficiency; 35 cps noise; $\chi^{(2)}$7 normalized efficiency (Yang et al., 2 Mar 2026)
637.2 nm $\chi^{(2)}$8 1588.3 nm	Fiber-integrated PPLN DFG	Total efficiency approximately 9%; pump-induced noise 154 Hz; expected fidelity exceeding 52% over 100 km (Liao et al., 25 Apr 2026)

These benchmarks show that there is no single “best” QFC metric. Resonant integrated devices minimize pump power; bulk and waveguide systems often achieve broader bandwidths or lower absolute noise; fiber-integrated devices trade peak efficiency for packaging robustness. This suggests that QFC performance must be interpreted in the context of the target application: metropolitan entanglement, quantum-memory interfacing, multiplexed routing, or on-chip processing.

5. Preservation of quantum properties and mode selectivity

The defining requirement of QFC is state preservation. This has been verified at several levels. In the 711 nm to 1313 nm quantum-dot experiment, first-order coherence was preserved within error, with $\chi^{(2)}$9 and $$\omega_{\text{out}} = \omega_{\text{p}} - \omega_{\text{in}},$$0, while antibunching improved from $$\omega_{\text{out}} = \omega_{\text{p}} - \omega_{\text{in}},$$1 to $$\omega_{\text{out}} = \omega_{\text{p}} - \omega_{\text{in}},$$2 (Zaske et al., 2012). In the SnV 619 nm to 1480 nm system, the post-conversion lifetime $$\omega_{\text{out}} = \omega_{\text{p}} - \omega_{\text{in}},$$3 matched the pre-conversion lifetime $$\omega_{\text{out}} = \omega_{\text{p}} - \omega_{\text{in}},$$4 within errors, demonstrating preservation of the temporal mode (Brevoord et al., 1 Sep 2025).

Integrated platforms have pushed this toward full qubit and process characterization. In unpoled InGaP waveguides, time-energy entanglement visibility remained about 83% before conversion and about 80% after conversion, and 780 nm time-bin qubits were converted with an average fidelity of $$\omega_{\text{out}} = \omega_{\text{p}} - \omega_{\text{in}},$$5 (Hu et al., 19 Oct 2025). In a polarization-insensitive DWDM hub, full process tomography across 16 telecom channels yielded a process fidelity $$\omega_{\text{out}} = \omega_{\text{p}} - \omega_{\text{in}},$$6 for one representative channel and a minimum fidelity of 0.75 across all 16 channels, well above the classical limit of 0.5 (Yamada et al., 22 Apr 2026).

QFC is also no longer restricted to qubit-like encodings. Temporal-mode-selective conversion in a four-dimensional Hilbert space achieved conversion efficiency as high as 92% and separability as high as 0.84 for a 20-GHz pulse train, enabling reprogrammable temporal-mode sorting by pump shaping (Manurkar et al., 2016). High-dimensional OAM qudit conversion from infrared to visible reported reconstructed-state fidelities of 98.29%, 97.42%, and 86.75% without dark counts in 2, 3, and 5 dimensions, respectively (Liu et al., 2019). These results establish that QFC can act as a structured mode transformer, not merely a wavelength shifter.

6. Network interfaces, tradeoffs, and current directions

QFC is increasingly being designed as network infrastructure rather than as a single bespoke interface. A channel-selective PPLN system demonstrated a 2.5 THz acceptance bandwidth and the ability to establish a 100-channel DWDM dynamic link from a single 780 nm quantum system, while a subsequent polarization-insensitive hub distributed heralded 780 nm photons into 16 ITU-T DWDM channels with 25 GHz spacing using about 2 THz of pump tuning range (Arizono et al., 2024, Yamada et al., 22 Apr 2026). This suggests a shift from fixed-frequency conversion toward reconfigurable frequency routing and multiplexing.

Another active direction is spectral matching for narrowband memories and long-lived emitters. Difference-frequency conversion of 606 nm photons to 1552 nm with ultra-narrow 12.5 MHz filtering maintained sufficiently high SNR for $$\omega_{\text{out}} = \omega_{\text{p}} - \omega_{\text{in}},$$7s-long weak coherent pulses compatible with Pr$$\omega_{\text{out}} = \omega_{\text{p}} - \omega_{\text{in}},$$8:Y$$\omega_{\text{out}} = \omega_{\text{p}} - \omega_{\text{in}},$$9SiO$$\omega_{\text{out}} = \omega_{\text{in}} + \omega_{\text{p}}.$$0 memories, including pulses up to 13.6 $$\omega_{\text{out}} = \omega_{\text{in}} + \omega_{\text{p}}.$$1s (Wengerowsky et al., 2024). Fiber-integrated 637.2 nm to 1588.3 nm conversion, designed for NV-center networking, combined approximately 9% total efficiency with 154 Hz pump-induced noise and yielded an expected fidelity exceeding 52% after 100 km at the emission rate of an NV center (Liao et al., 25 Apr 2026).

The main technical tradeoff remains efficiency versus noise. Long-wavelength pumping is often chosen to suppress Raman noise, as in 1950 nm-pumped LNOI upconversion to 860 nm and in proposed 1990 nm-pumped TSFG/TDFG interfaces for 606 nm $$\omega_{\text{out}} = \omega_{\text{in}} + \omega_{\text{p}}.$$2 1550 nm conversion (Wang et al., 2022, Lu et al., 2020). However, short-wavelength pumping is not intrinsically unusable: first-order-QPM TFLN conversion from 393 nm to telecom exploited the counter-tuning behavior of DFG and SPDC, together with ultra-narrowband filtering, to achieve 28.8% external efficiency and only 35 cps noise (Yang et al., 2 Mar 2026). The practical controversy, therefore, is not whether one pump regime is universally superior, but which combination of dispersion, phase matching, filtering, and packaging is optimal for a given node wavelength and network architecture.

Current research points toward two partially convergent trajectories. One is highly integrated, low-power, multi-channel QFC on TFLN or III-V nanophotonics, where electrical or thermal programmability is central (Wang et al., 4 Sep 2025, Hu et al., 19 Oct 2025). The other is deployable, fiber-integrated or bulk-assisted conversion modules optimized for real emitters, narrowband filtering, and network robustness (Brevoord et al., 1 Sep 2025, Liao et al., 25 Apr 2026). Together, these developments indicate that QFC has evolved from a specialized nonlinear-optical technique into a general-purpose quantum interconnect spanning emitters, memories, detectors, and multiplexed telecom networks.