Monolithic Cavity SPDC Source

• Monolithic cavity SPDC sources are integrated devices where the nonlinear medium and resonator combine to enable narrowband, cavity-defined down-conversion with high passive stability.
• They employ facet coatings, whispering-gallery modes, and birefringent engineering to selectively generate single-longitudinal-mode photon pairs for applications like telecom quantum communication.
• The design balances linewidth, escape efficiency, and tunability, making these sources ideal for high-purity entanglement and frequency-multiplexed quantum networks.

A monolithic cavity SPDC source is a spontaneous parametric down-conversion source in which the nonlinear medium and the resonator are integrated into a single mechanically rigid optical element, rather than assembled from a separate crystal and an external free-space cavity. In practice, this has been realized in several architectures: facet-coated bulk crystals that form Fabry–Pérot cavities directly on the crystal faces, whispering-gallery-mode resonators in lithium niobate, and closely related monolithic bulk cavities used for narrowband heralded photons or entangled-photon generation (Li et al., 2019, Gao et al., 2024, Planas et al., 4 Feb 2026, Förtsch et al., 2012). The defining consequence is that SPDC occurs into cavity-selected spectral and spatial modes with high passive stability, reduced alignment complexity, and linewidths set by cavity parameters rather than by the intrinsically broad single-pass SPDC bandwidth.

1. Definition, scope, and architectural variants

In the most literal optical implementation, the crystal itself is the cavity. A representative example is type-II PPKTP with dimensions $$1\,\text{mm} \times 2\,\text{mm} \times 0.85\,\text{mm}$$, where the two crystal faces along the pump direction are coated so that one face is AR at 775 nm and HR at 1550 nm, while the opposite face is AR at 775 nm and partially reflective at 1550 nm; these coated facets form a monolithic Fabry–Pérot cavity resonant for the down-converted photons but not for the pump (Li et al., 2019). Closely related facet-defined cavities were later used in telecom-band two-cavity superposition schemes and in a plano–convex ppKTP heralded single-photon source (Gao et al., 2024, Planas et al., 4 Feb 2026).

The term also covers monolithic resonators in which the cavity is not a linear Fabry–Pérot but a single dielectric body supporting high-$Q$ modes. A lithium-niobate whispering gallery mode resonator, fabricated as a disk-shaped monolithic cavity and operated in a triply resonant SPDC regime, is an explicit example (Förtsch et al., 2012).

By contrast, several cavity-enhanced SPDC platforms that are central to the design literature are not monolithic. Examples include a linear flat–concave cavity containing a separate 5-mm type-II PPKTP crystal and actively locked with a Pound–Drever–Hall scheme (Wan et al., 8 Feb 2025), a semi-hemispherical OPO cavity containing 30 mm PPKTP plus a 15 mm BBO tuning crystal (Moqanaki et al., 2018), and a compact two-crystal Fabry–Pérot source design with discrete mirrors (Pušavec et al., 2024). These sources remain directly relevant because the same cluster engineering, birefringent compensation, and cavity-linewidth arguments carry over to monolithic implementations.

A common misconception is that “monolithic” is synonymous with “on-chip.” The literature shows a broader usage: bulk facet-coated crystals, whispering-gallery resonators, and integrated waveguides all fall within the monolithic cavity logic when the resonator and the nonlinear interaction are structurally unified (Förtsch et al., 2012, Kellner et al., 26 Jun 2025).

2. Cavity physics, linewidth control, and single-longitudinal-mode selection

The basic resonance condition is the Fabry–Pérot relation

$m\,\lambda = 2 n L,$

or equivalently

$$\omega_m = \frac{m\pi c}{nL}, \qquad \Delta\omega = \frac{\pi c}{nL},$$

with cavity length $L$, refractive index $n$, and longitudinal index $m$ (Li et al., 2019). In monolithic type-II cavities, birefringence makes the signal and idler free spectral ranges slightly different: $$\Delta\omega_s = \frac{\pi c}{n_y L}, \qquad \Delta\omega_i = \frac{\pi c}{n_z L}.$$ That small mismatch is the origin of the cluster effect: only certain signal-idler mode pairs are simultaneously resonant and phase matched (Li et al., 2019).

A concrete telecom example illustrates the mechanism. In the submillimeter monolithic PPKTP cavity, the measured FSRs were $93.61\,\text{GHz}$ for one polarization and $89.42\,\text{GHz}$ for the orthogonal polarization, with cavity linewidths $Q$0 and $Q$1, respectively. From these values, the cluster spacing was $Q$2, and the frequency mismatch at the cluster edge was $Q$3. Because

$Q$4

only the central doubly resonant mode pair in each cluster remained simultaneously resonant, which guarantees single-longitudinal-mode operation (Li et al., 2019).

This design logic generalizes. In a non-monolithic but highly relevant 852 nm source, an additional birefringent element was used to tune double resonance independently of phase matching, yielding single-longitudinal-mode operation without external mode filters and a measured bandwidth of $Q$5 (Moqanaki et al., 2018). In the compact two-crystal cavity design study, the same physics was formulated through single-crystal cluster spacing, joint cluster spacing, and the criterion that fidelity exceeds 0.9 when the cluster spacing is more than twice the SPDC bandwidth (Pušavec et al., 2024). This suggests that monolithic composite cavities can use extra birefringent sections or engineered path delays to decouple phase matching from cluster selection.

In the 1540/1560 nm plano–convex ppKTP monolithic cavity, the measured signal-arm FSR was $Q$6 and the linewidth $Q$7, corresponding to a finesse $Q$8. Three TEM$Q$9 longitudinal modes lay within the relevant phase-matching window, and the central mode could be isolated with modest filtering, leading to $m\,\lambda = 2 n L,$0 spectral purity for that mode (Planas et al., 4 Feb 2026).

3. Representative monolithic implementations

The submillimeter telecom device used type-II PPKTP with poling period $m\,\lambda = 2 n L,$1, pumped by a 775 nm CW laser and producing time-energy-entangled photon pairs near 1550 nm. Its monolithic character came from direct mirror coatings on the crystal facets, active temperature stabilization at $m\,\lambda = 2 n L,$2, and the absence of any separate mirrors or active 1550 nm cavity lock (Li et al., 2019).

A second telecom implementation used two separate monolithic PPKTP cavities, each $m\,\lambda = 2 n L,$3, placed in a passively stable beam-displacer interferometer. Each crystal had one face coated HR at 1550 nm and AR at 775 nm, and the other face coated approximately $m\,\lambda = 2 n L,$4 reflective at 1550 nm and AR at 775 nm. Entanglement was generated by coherently superposing the outputs of the two monolithic cavities, rather than by a single crystal alone (Gao et al., 2024).

A third realization targeted pure heralded single photons in the telecom C-band. It used a $m\,\lambda = 2 n L,$5 ppKTP crystal with poling period $m\,\lambda = 2 n L,$6, planar and convex coated facets, and a plano–convex doubly resonant cavity. The planar facet was approximately $m\,\lambda = 2 n L,$7 reflective for signal and idler and almost anti-reflection for the pump, while the convex output coupler had $m\,\lambda = 2 n L,$8 and $m\,\lambda = 2 n L,$9. In that device the pump was non-resonant but double-pass, whereas signal and idler were doubly resonant (Planas et al., 4 Feb 2026).

A distinct but canonical monolithic architecture is the lithium-niobate whispering-gallery resonator. There the cavity is a disk-shaped $$\omega_m = \frac{m\pi c}{nL}, \qquad \Delta\omega = \frac{\pi c}{nL},$$0 MgO-doped LiNbO$$\omega_m = \frac{m\pi c}{nL}, \qquad \Delta\omega = \frac{\pi c}{nL},$$1 resonator with radius $$\omega_m = \frac{m\pi c}{nL}, \qquad \Delta\omega = \frac{\pi c}{nL},$$2, quality factor $$\omega_m = \frac{m\pi c}{nL}, \qquad \Delta\omega = \frac{\pi c}{nL},$$3, and evanescent prism coupling. SPDC was operated in a triply resonant regime and the photon bandwidth could be tuned between $$\omega_m = \frac{m\pi c}{nL}, \qquad \Delta\omega = \frac{\pi c}{nL},$$4 and $$\omega_m = \frac{m\pi c}{nL}, \qquad \Delta\omega = \frac{\pi c}{nL},$$5 by varying the prism-resonator gap (Förtsch et al., 2012).

These implementations already span the main monolithic cavity classes: facet-defined Fabry–Pérot bulk crystals, superposed monolithic subcavities for polarization entanglement, and monolithic dielectric resonators with whispering-gallery modes.

4. Entanglement generation and state engineering

Monolithic cavity SPDC does not, by itself, fix the entanglement degree of freedom. The earliest telecom facet-coated PPKTP device generated single-longitudinal-mode time-energy entanglement and characterized it through cross-correlation, Franson-type interference, and single-photon Michelson interference. The measured Franson visibilities were $$\omega_m = \frac{m\pi c}{nL}, \qquad \Delta\omega = \frac{\pi c}{nL},$$6 raw and $$\omega_m = \frac{m\pi c}{nL}, \qquad \Delta\omega = \frac{\pi c}{nL},$$7 net for one idler phase setting, and $$\omega_m = \frac{m\pi c}{nL}, \qquad \Delta\omega = \frac{\pi c}{nL},$$8 raw and $$\omega_m = \frac{m\pi c}{nL}, \qquad \Delta\omega = \frac{\pi c}{nL},$$9 net for another; the work did not report a CHSH test or full state tomography (Li et al., 2019).

Polarization entanglement in a monolithic setting was demonstrated by the two-cavity superposition architecture. There, one cavity contributed an effective $L$0 amplitude and the other an effective $L$1 amplitude, yielding

$L$2

with $L$3 obtained at $L$4. The measured polarization visibilities were $L$5 in the H/V basis and $L$6 in the $L$7 basis, the CHSH parameter was $L$8, and the reconstructed-state fidelity with $L$9 was $n$0 (Gao et al., 2024).

A useful design clarification comes from a later non-monolithic narrowband OAM-entangled source: “To create the entanglement, at least four cavity resonance modes are required: two of them for the path separation of the down-converted photons and the other two for the entanglement preparation.” Although that work used a bulk linear cavity rather than a monolithic one, the statement functions as a general cavity-mode design rule for directly generated cavity entanglement (Wan et al., 8 Feb 2025).

This helps correct another common misconception: a monolithic cavity source is not automatically a directly outputting entangled-photon source. Some monolithic realizations are optimized instead for heralded pure single photons, as in the 1540 nm source with $n$1 spectral purity and $n$2 heralding efficiency (Planas et al., 4 Feb 2026). Entanglement may arise from cavity-internal mode structure, from coherent superposition of two monolithic cavities, or from external interferometric analysis, depending on the architecture (Li et al., 2019, Gao et al., 2024).

5. Performance landscape and application space

The performance envelope of monolithic cavity SPDC sources is best understood through representative implementations.

Source class	Spectral scale	Representative metrics
Submillimeter monolithic type-II PPKTP	$n$3 and $n$4 cavity linewidths	spectral brightness $n$5, heralding efficiency about $n$6, CAR up to about 1800 (Li et al., 2019)
Superposed telecom monolithic cavities	bandwidth below $n$7	CAR $n$8, $n$9, Bell-state fidelity $m$0 (Gao et al., 2024)
ppKTP monolithic heralded source	$m$1	heralding efficiency $m$2, multi-photon contamination below $m$3, spectral purity $m$4, HOM visibility $m$5 (Planas et al., 4 Feb 2026)
LiNbO$m$6 WGMR monolithic source	$m$7–$m$8	wavelength tuning over $m$9, heralded $$\Delta\omega_s = \frac{\pi c}{n_y L}, \qquad \Delta\omega_i = \frac{\pi c}{n_z L}.$$0 (Förtsch et al., 2012)

The application logic is consistent across these implementations. Telecom-band devices around 1550 nm are motivated by low fiber loss, compatibility with standard telecom infrastructure, and improved suitability for long-distance quantum communication and repeaters (Li et al., 2019, Gao et al., 2024). Narrow linewidths in the sub-GHz or MHz regime are closer to quantum-memory bandwidths than the THz-scale bandwidth of bare SPDC, which is why these sources are repeatedly framed as candidates for memory-compatible networking (Li et al., 2019, Förtsch et al., 2012, Planas et al., 4 Feb 2026).

The linewidth scale also determines coherence time. In the submillimeter telecom monolithic cavity, the measured cross-correlation FWHM was about $$\Delta\omega_s = \frac{\pi c}{n_y L}, \qquad \Delta\omega_i = \frac{\pi c}{n_z L}.$$1 (Li et al., 2019). In the 1540 nm heralded source, linewidths of $$\Delta\omega_s = \frac{\pi c}{n_y L}, \qquad \Delta\omega_i = \frac{\pi c}{n_z L}.$$2 and $$\Delta\omega_s = \frac{\pi c}{n_y L}, \qquad \Delta\omega_i = \frac{\pi c}{n_z L}.$$3 corresponded to coherence times on the order of $$\Delta\omega_s = \frac{\pi c}{n_y L}, \qquad \Delta\omega_i = \frac{\pi c}{n_z L}.$$4–$$\Delta\omega_s = \frac{\pi c}{n_y L}, \qquad \Delta\omega_i = \frac{\pi c}{n_z L}.$$5 (Planas et al., 4 Feb 2026). In the lithium-niobate WGMR platform, the linewidth could be tuned to the $$\Delta\omega_s = \frac{\pi c}{n_y L}, \qquad \Delta\omega_i = \frac{\pi c}{n_z L}.$$6–$$\Delta\omega_s = \frac{\pi c}{n_y L}, \qquad \Delta\omega_i = \frac{\pi c}{n_z L}.$$7 range, which is directly relevant to narrowband light-matter interfaces (Förtsch et al., 2012).

6. Design trade-offs, misconceptions, and current directions

Three trade-offs recur across the literature. The first is linewidth versus escape efficiency. Higher finesse narrows the cavity line, but it also reduces output coupling and makes the source more sensitive to loss. The 1540 nm monolithic heralded source explicitly identifies this compromise: higher finesse would better suppress adjacent longitudinal modes, but at the cost of reduced escape efficiency and brightness (Planas et al., 4 Feb 2026).

The second is simplicity versus entanglement engineering. A single monolithic type-II cavity can generate narrowband time-energy entanglement directly (Li et al., 2019), but high-quality polarization entanglement may require superposition of two monolithic cavities in a passive interferometer (Gao et al., 2024). The two-crystal compact design study reaches a similar conclusion in another form: additional path delay or dispersive engineering may be necessary to obtain both usable joint cluster spacing and single-mode fidelity, even though the cavity itself remains simple (Pušavec et al., 2024).

The third is tunability versus passive stability. Monolithic devices substantially reduce alignment complexity and often eliminate the need for active cavity-length locking of the down-converted field (Li et al., 2019, Planas et al., 4 Feb 2026), but they usually concentrate tuning authority into temperature, and sometimes electro-optic control. That improves robustness while narrowing the practical tuning range. In the telecom submillimeter source, wide tuning would require re-engineering coatings or poling period (Li et al., 2019). In the 1540 nm source, uncontrolled pump-phase effects from the double-pass configuration broadened side clusters, motivating a future single-pass pump design with non-reflective pump coatings (Planas et al., 4 Feb 2026).

Several current directions follow directly from these limitations. One is composite birefringent cavity engineering, as demonstrated in a non-monolithic 852 nm source where an additional tuning crystal allowed independent control of cluster separation and phase matching (Moqanaki et al., 2018). Another is direct integrated implementation of spectrally favorable SPDC geometries: a plausible implication is that combining cavity engineering with the counter-propagating LNOI geometry, which already yields spectrally uncorrelated photon pairs without spectral filtering, could produce monolithic cavity sources that begin from a nearly factorable JSA rather than relying on the cavity to remove correlations (Kellner et al., 26 Jun 2025). A further direction is system-level multiplexing: cavity-enhanced SPDC models show that approximately 100 frequency modes can raise heralding probability to approximately $$\Delta\omega_s = \frac{\pi c}{n_y L}, \qquad \Delta\omega_i = \frac{\pi c}{n_z L}.$$8 for $$\Delta\omega_s = \frac{\pi c}{n_y L}, \qquad \Delta\omega_i = \frac{\pi c}{n_z L}.$$9 and to about $93.61\,\text{GHz}$0 for $93.61\,\text{GHz}$1 while maintaining fidelity $93.61\,\text{GHz}$2 or higher, which suggests a natural role for monolithic multi-mode cavity sources in frequency-multiplexed repeaters (Komatsudaira et al., 1 Apr 2026).

Taken together, these results define monolithic cavity SPDC sources as a family of narrowband, cavity-defined nonlinear light sources in which spectral selection, spatial-mode purity, and long-term passive stability are built into the nonlinear element itself. Their main unresolved tasks are not the existence of monolithic cavity enhancement as such, but the simultaneous optimization of linewidth, escape efficiency, mode purity, and directly generated entanglement for specific network interfaces (Li et al., 2019, Gao et al., 2024, Planas et al., 4 Feb 2026, Förtsch et al., 2012).