Nanofiber Cavity QED Systems Overview

• Nanofiber cavity QED systems are defined by tight transverse confinement in subwavelength fibers that couple evanescently to quantum emitters.
• They utilize diverse architectures—including fiber Fabry–Perot, composite photonic crystal, and ring resonators—to reach both Purcell and strong coupling regimes.
• These platforms are pivotal for quantum networking, offering precise spectral control and seamless integration with standard fiber-optic infrastructures.

Nanofiber cavity QED systems are cavity quantum electrodynamics platforms in which the fundamental guided mode of a subwavelength optical fiber, or of a closely related hollow or composite fiber structure, is confined longitudinally by distributed reflectors or resonant feedback and coupled to emitters through the nanofiber’s evanescent field. Their defining combination is tight transverse confinement, direct interfacing to standard optical fibers, and a mode geometry that supports both cavity-enhanced spontaneous emission and coherent strong coupling. Realizations span fiber Bragg grating Fabry–Perot resonators, composite photonic crystal cavities formed by external nanogratings, nanofiber-segment ring resonators, and fiber-linked multi-node networks, with operating regimes extending from the Purcell regime to single-atom and collective strong coupling, and from discrete-mode cavity QED to delayed-feedback dynamics reminiscent of waveguide QED (Wuttke et al., 2012, Kato et al., 2015, Shillito et al., 2019, Lechner et al., 2023).

1. Physical architectures and material platforms

The canonical nanofiber cavity QED architecture is the all-fiber Fabry–Perot cavity in which a subwavelength nanofiber waist is sandwiched between two fiber Bragg gratings. In the 2015 single-atom realization, the nanofiber diameter was 400 nm, the waist length 1 mm, and the cavity length approximately 33 cm; the cavity operated as a one-sided device with a high reflector and a temperature-tunable output coupler, while a state-insensitive two-color trap localized a single Cs atom about 170 nm from the nanofiber surface in the evanescent field (Kato et al., 2015). A related tapered-optical-fiber microresonator used a 500 nm diameter, 5 mm waist inside a 2 cm Fabry–Perot cavity near the Cs D2 line and already fulfilled the parameter regime required for strong-coupling cavity QED and low-power nonlinear optics (Wuttke et al., 2012).

A second architectural family uses composite photonic crystal cavities. Here the nanofiber itself is not directly etched into mirrors; instead, an external nanofabricated grating is brought into optical contact with the fiber and imposes a longitudinal index modulation on the guided mode. A composite photonic crystal cavity formed by a tapered optical nanofiber and a silica grating with a central defect of width $3\Lambda_g/2 = 480$ nm operated near $785$–$789$ nm and coupled single colloidal quantum dots deposited on the nanofiber surface (Yalla et al., 2014). Later variants introduced asymmetric defect-mode gratings to produce one-sided ONF cavities whose coupling can be tuned between under-, critical-, and over-coupling regimes, and symmetric defect-mode gratings on optical nanocapillary fibers to place the emitter inside a hollow core rather than on an external surface (Yalla et al., 2022, Gadde et al., 8 Jul 2025).

Ring geometries constitute a third class. In a nanofiber-segment ring resonator, a short subwavelength waist is inserted into a much longer loop of standard single-mode fiber closed by a fused fiber coupler. A proof-of-principle warm-vapor device used a nanofiber diameter of about 320 nm, a 6 mm waist, and a loop length of about 2.8 m; a later cold-atom realization used a 400 nm waist of length about 2 mm in a 1.4 m ring and observed collective vacuum Rabi splitting from an ensemble of Cs atoms in a MOT overlapping the evanescent field (Jones et al., 2016, Ruddell et al., 2017). Extending the ring length from meters to tens of meters while retaining a nanofiber interaction region enabled a controlled transition from conventional cavity-QED dynamics to non-Markovian behavior characteristic of waveguide QED (Lechner et al., 2023).

Within the broader nanofiber-coupled landscape, related fiber-integrated visible-wavelength systems use suspended photonic crystal nanobeams rather than cylindrical nanofibers as the resonant element, but retain the central objective of combining high-$Q$, small-$V$ cavity QED with efficient fiber interfacing. A free-standing Si$_3$N$_4$ nanobeam photonic crystal cavity designed for the GeV zero-phonon line at $\lambda \approx 602$ nm achieved measured $Q \approx 24{,}000$ and fiber–waveguide coupling efficiency $\mu_c \approx 96\% \pm 2\%$ through an etched single-mode fiber tip (Alajlan et al., 2020).

2. Electromagnetic structure and cavity-QED description

The guided mode relevant to nanofiber cavity QED is the fundamental hybrid $785$0 mode. In subwavelength fibers its field has a substantial evanescent component outside the silica, so emitters trapped or deposited tens to hundreds of nanometers from the surface experience a large single-photon field. This geometry is the basis of the cavity coupling rate

$785$1

or, with explicit position dependence,

$785$2

with the usual dependence on dipole moment, mode volume, and local field overlap (Wuttke et al., 2012, Ruddell et al., 2020).

Across the literature, the standard figures of merit are the cavity linewidth $785$3, the spontaneous-emission or polarization-decay rate $785$4, the cooperativity, and the Purcell factor. Frequently used definitions are

$785$5

or $785$6 depending on convention, and

$785$7

or equivalently

$785$8

In weak coupling, $785$9 quantifies cavity-enhanced emission into the guided mode; in strong coupling, the condition $789$0 or $789$1 yields resolvable vacuum Rabi splitting (Shillito et al., 2019, Kato et al., 2015).

For coupled-node nanofiber networks, a minimal single-excitation Hamiltonian includes cavity modes $789$2, atomic lowering operators $789$3, and an explicit fiber mode $789$4,

$789$5

with losses incorporated through a non-Hermitian effective Hamiltonian. In the short-fiber reduced model, the fiber mediates an effective cavity–cavity hopping

$789$6

This description is central to two-node all-fiber cavity-QED links in the single-excitation regime (Shillito et al., 2019).

Input–output theory remains the basic measurement framework. For a cavity port,

$789$7

and in the no-drive, single-excitation regime the photon flux reduces to $789$8. In transfer-matrix treatments of larger fiber-connected networks, mirrors, propagation segments, beam splitters, and atoms are represented as concatenated scattering elements, allowing multimode linear spectra to be computed without truncating to a single standing-wave mode (Shillito et al., 2019, Német et al., 2019).

3. Purcell enhancement, strong coupling, and representative performance

Nanofiber cavity QED systems do not occupy a single operating regime. Some are explicitly Purcell-regime devices, while others reach single-emitter or collective strong coupling.

Platform	Regime	Representative performance
Composite photonic crystal cavity with single QDs (Yalla et al., 2014)	Purcell regime	Measured $789$9, $Q$0, $Q$1
TOF Fabry–Perot microresonator (Wuttke et al., 2012)	Strong-coupling capable	$Q$2, $Q$3, $Q$4 MHz for a Cs atom at the surface
Trapped single atom in all-fiber cavity (Kato et al., 2015)	Single-atom strong coupling	$Q$5 MHz, $Q$6 MHz, $Q$7 MHz, $Q$8
Ultra-low-loss Fabry–Perot cavity (Ruddell et al., 2020)	High-cooperativity design point	$Q$9, $V$0, maximum expected internal cooperativity $V$1 for a Cs atom on the surface
Defect-free nanofiber PhC resonator (Tanaka et al., 27 Jan 2026)	Ultrahigh-$V$2 nanofiber resonator	Loaded $V$3, intrinsic $V$4, $V$5 MHz
Cryo-compatible nanofiber microresonator (Hütner et al., 2020)	Cryogenic cavity QED / Purcell enhancement	$V$6 at $V$7 K, Purcell factor up to $V$8

The Purcell-regime side of the field was established early by composite photonic crystal cavities on nanofibers. In the quantum-dot experiment, the measured corrected enhancement factor reached $V$9 at $_3$0 nm, with simulated $_3$1-polarized channeling efficiency $_3$2 and an analytical Purcell factor $_3$3. The device deliberately operated with $_3$4, so spontaneous emission was enhanced and redirected into the guided modes without vacuum Rabi splitting (Yalla et al., 2014).

Single-emitter strong coupling was realized in the all-fiber Fabry–Perot geometry with a trapped Cs atom. The observed transmission spectra showed vacuum Rabi splitting consistent with the Jaynes–Cummings model, and the measured rates $_3$5 MHz, $_3$6 MHz, and $_3$7 MHz satisfied the strong-coupling condition. Despite a finesse below $_3$8, the tight transverse confinement of the nanofiber made the coherent coupling large enough for $_3$9 and $_4$0 (Kato et al., 2015).

Subsequent work shifted from proof of principle to parameter optimization. An all-fiber cavity with a 207 nm nanofiber waist radius reached an internal round-trip loss of only $_4$1, total finesse $_4$2, and a predicted maximum internal cooperativity of about $_4$3 for a cesium atom on the nanofiber surface. The same study reported single-pass taper transmission of $_4$4 at the optimized radius and identified the optimum as the balance between stronger evanescent coupling at smaller radii and the onset of radiation-mode loss (Ruddell et al., 2020).

A distinct route to very high $_4$5 used single-shot femtosecond laser ablation to fabricate defect-free nanofiber photonic crystal Fabry–Perot resonators on a 500 nm diameter nanofiber waist. The measured resonances at $_4$6 nm and $_4$7 nm had linewidths of $_4$8 MHz and $_4$9 MHz, loaded $\lambda \approx 602$0 values of $\lambda \approx 602$1 and $\lambda \approx 602$2, and intrinsic $\lambda \approx 602$3 values of $\lambda \approx 602$4 and $\lambda \approx 602$5, respectively (Tanaka et al., 27 Jan 2026).

Cryogenic operation extends the same architecture to solid-state emitters that require low temperatures. A fully fiber-integrated nanofiber microresonator retained alignment-free operation at $\lambda \approx 602$6 K, reached $\lambda \approx 602$7, and, together with the small mode volume at the nanofiber surface, supported either the coherent-dynamics or fast-cavity regime, with a reported Purcell factor of up to $\lambda \approx 602$8 (Hütner et al., 2020).

4. Collective dynamics, multimode physics, and non-Hermitian structure

When multiple emitters or multiple cavities are coupled through fibers, nanofiber cavity QED acquires a collective-mode structure that is not captured by single-resonance intuition. In the symmetric two-node model of two nanofiber cavities coupled by a short single-mode fiber, the resonant lossless problem has five collective modes: $\lambda \approx 602$9 with normal-mode splitting

$Q \approx 24{,}000$0

The two fiber-dark modes do not populate the connecting fiber, while the cavity-dark mode does not populate either cavity. For $Q \approx 24{,}000$1, the excitation remains essentially localized in one atom–cavity subsystem; for $Q \approx 24{,}000$2, the fiber mediates rapid exchange and the bright-state doublet dominates the fiber output; and for $Q \approx 24{,}000$3, all five quasi-normal modes contribute and small interference terms generate narrow shoulders, asymmetric peak heights, and cavity-to-cavity contrast reversals. The normal-mode picture captures the dominant splittings, but exact diagonalization of the non-Hermitian Hamiltonian is needed to explain these interference features (Shillito et al., 2019).

A complementary transfer-matrix formulation generalizes this physics to larger all-fiber networks. Rather than assuming a single standing-wave mode per cavity, the transfer-matrix approach composes frequency-dependent propagation phases, FBG reflectivities, beam splitters, and atomic scatterers into an overall linear response. For two cavity-QED nodes connected by a fiber, it reproduces central fiber-dark features and bright side peaks, but also reveals significant deviations from single-mode linearized quantum-optical models when cavity free spectral ranges are small, mirror reflectivities are modest, or multiport driving excites traveling-wave rather than standing-wave superpositions (Német et al., 2019).

Ring resonators with nanofiber sections provide a distinct collective strong-coupling setting. In the cold-atom all-fiber ring cavity, the 1.4 m device had $Q \approx 24{,}000$4 MHz, $Q \approx 24{,}000$5, $Q \approx 24{,}000$6 MHz, an effective single-atom coupling $Q \approx 24{,}000$7 MHz, and an ensemble cooperativity $Q \approx 24{,}000$8 with $Q \approx 24{,}000$9. The resulting collective coupling $\mu_c \approx 96\% \pm 2\%$0 MHz produced well-resolved vacuum Rabi splitting in the weak-driving limit (Ruddell et al., 2017).

If the ring is lengthened so that the round-trip time becomes comparable to or much larger than the atomic lifetime, the discrete-mode cavity-QED picture breaks down. In a fiber-ring resonator containing a 400 nm diameter, 10 mm long nanofiber section, increasing the ring length from $\mu_c \approx 96\% \pm 2\%$1 m to $\mu_c \approx 96\% \pm 2\%$2 m reduced the FSR from $\mu_c \approx 96\% \pm 2\%$3 MHz to $\mu_c \approx 96\% \pm 2\%$4 MHz and increased the round-trip time from $\mu_c \approx 96\% \pm 2\%$5 ns to $\mu_c \approx 96\% \pm 2\%$6 ns. In the long-ring case, vacuum Rabi oscillations disappeared and sharp non-Markovian features separated by the round-trip time emerged, consistent with a delayed-feedback or cascaded-interaction description rather than the Tavis–Cummings model (Lechner et al., 2023).

Non-Hermitian extensions have also been developed inside a single nanofiber cavity. A system of two two-level emitters coupled to a nanofiber cavity under coherent perfect absorption can be described by an effective pseudo-Hermitian Hamiltonian whose eigenvalues are either all real or one real plus a complex-conjugate pair. By tuning the ratios $\mu_c \approx 96\% \pm 2\%$7 and $\mu_c \approx 96\% \pm 2\%$8, the system exhibits both EP3 and EP2 without parity-time symmetry, and these higher-order exceptional points appear in total output and transmission spectra (Li et al., 2022).

5. Fabrication, interfacing, and spectral control

The performance of nanofiber cavity QED systems is set as much by fabrication and interfacing as by the abstract cavity-QED parameters. For all-fiber Fabry–Perot cavities, the dominant challenge is minimizing taper loss while reaching a waist radius that maximizes evanescent coupling. A major advance was continuous in-situ monitoring of both finesse and radius during taper fabrication. In the optimized 852.3 nm cavity, three auxiliary lasers at 682 nm, 633 nm, and 532 nm tracked higher-order-mode cutoffs in spectrograms, while an 852.3 nm ringdown measurement provided real-time finesse. This approach yielded an internal round-trip loss $\mu_c \approx 96\% \pm 2\%$9, under-coupled finesse $785$00, and single-pass taper transmission $785$01 at $785$02 nm (Ruddell et al., 2020).

Direct nanofiber patterning provides another route. Single-shot femtosecond laser ablation on a 500 nm diameter nanofiber created defect-free photonic crystal mirrors with smoothly varying crater size, producing ultrahigh-$785$03 resonances while avoiding the roughness buildup associated with multi-pulse processing. The same study showed that thermo-optic effects dominate the nonlinear response across the entire cavity bandwidth, even for $785$04 ns pulses, and measured a thermal cutoff frequency of $785$05 kHz corresponding to $785$06 (Tanaka et al., 27 Jan 2026).

Composite cavities are especially attractive when directionality or visible-wavelength operation is required. In the one-sided composite ONF cavity, an asymmetric defect-mode grating with $785$07 was designed so that the desired ONF port dominates the external coupling. Simulations predicted a maximum channeling efficiency $785$08 into the chosen ONF port at $785$09 and $785$10, while experiment demonstrated the expected transition from over-coupling to critical coupling and under-coupling as $785$11 was varied (Yalla et al., 2022). A symmetric composite photonic crystal cavity on an optical nanocapillary fiber extended this concept to a hollow core, where the emitter can be placed inside the capillary; for $785$12, $785$13 nm, and $785$14 nm, the maximum channeling efficiency reached approximately $785$15 for a $785$16-polarized emitter at the cavity antinode (Gadde et al., 8 Jul 2025).

Fiber coupling can also be engineered outside strictly cylindrical nanofiber resonators. In suspended Si$785$17N$785$18 nanobeam cavities, adiabatic index matching between an etched single-mode fiber tip and a suspended waveguide taper produced measured fiber–waveguide coupling efficiencies of $785$19 at the cavity resonance, with saturation once the fiber–waveguide overlap reached about $785$20 (Alajlan et al., 2020). This fiber-chip direction is not a nanofiber cavity in the narrow geometric sense, but it shares the same system-level objective: preserving a low-loss fiber interface while reaching high-$785$21 cavity-QED conditions.

Spectral control spans thermal, mechanical, and geometric tuning. Fiber Bragg gratings can be tuned by temperature or strain, ring resonators by loop length and coupler parameters, and composite grating resonances by diameter and grating period. At cryogenic temperatures, nanofiber microresonators remain tunable by piezo strain, while the central wavelength shifts from $785$22 nm at room temperature to $785$23 nm at $785$24 K (Hütner et al., 2020). In ring cavities, the thermal response of the nanofiber itself can be used for active stabilization: a 780 nm heating beam induced thermal self-locking and stabilized the cavity resonance to the Cs line during repeated experimental cycles (Ruddell et al., 2017).

6. Quantum networking, gate operations, and outstanding limitations

The fiber-native format of nanofiber cavity QED has made it a natural candidate for distributed quantum networks. A particularly direct theoretical construction uses atomic mirrors: a periodic lattice of atoms trapped near a nanofiber at spacing $785$25 behaves as a high-reflectivity Bragg mirror for the guided mode, and two such mirrors define a cavity segment in which an impurity atom experiences a Jaynes–Cummings interaction with a subradiant collective spin-wave mode. The effective coupling is

$785$26

while the cavity decay is set by the mirror atoms’ free-space leakage, $785$27, rather than by ordinary mirror transmission. Under realistic conditions the model predicts vacuum Rabi oscillations and strong coupling at effective cavity finesse $785$28 (Chang et al., 2012).

A more implementation-oriented network proposal considers modules formed by one-sided fiber Bragg grating nanofiber cavities containing multiple Cs atoms coupled through the evanescent field, with modules linked by conventional optical fibers. Photon-mediated local and remote controlled-$785$29 gates are implemented by reflecting long single-photon pulses from the cavity and exploiting the state-dependent phase shift between bright and dark atomic manifolds. Crosstalk from spectator atoms can be suppressed either by local AC Stark shifts or by increasing the atom–fiber distance. For a target-atom coupling $785$30 at $785$31 nm and $785$32, spectator detunings $785$33 MHz, spectator positions $785$34, or combined settings such as $785$35 MHz with $785$36 were sufficient to achieve $785$37. The resulting error channels were found to be strongly biased toward dephasing, with essentially vanishing non-dephasing Pauli error rates (Keller et al., 10 Sep 2025).

Several limitations recur across the field. First, not every nanofiber cavity operates in strong coupling; composite photonic crystal devices with quantum dots and many visible-wavelength composite cavities are explicitly Purcell-regime platforms, with $785$38 even when cooperativity is substantial (Yalla et al., 2014, Gadde et al., 8 Jul 2025). Second, longer fiber links do not remain within single-mode cavity-QED theory once the round-trip time approaches the atomic lifetime; delayed feedback and quasi-continuum mode structure then become essential (Lechner et al., 2023). Third, most compact analytic models assume the rotating-wave approximation, Markovian loss, negligible propagation delay, symmetry between couplings, and restriction to the single-excitation manifold, so they omit multi-photon saturation, superradiant multi-atom effects, technical dephasing, and surface-induced disorder (Shillito et al., 2019).

Fabrication losses, surface contamination, polarization asymmetries, and support-induced leakage remain decisive practical constraints. Representative examples include the $785$39 discrepancy between simulated and measured $785$40-mode transmission in a composite photonic crystal cavity, the dominance of intrinsic over external loss in under-coupled all-fiber ring cavities, the contamination-induced one-way transmission degradation to about $785$41 in a vacuum-installed nanofiber Fabry–Perot cavity, and the experimentally observed scattering-limited insertion losses of $785$42 to $785$43 in one-sided composite ONF cavities (Yalla et al., 2014, Ruddell et al., 2017, Kato et al., 2015, Yalla et al., 2022). These are engineering rather than conceptual limitations, but they directly determine whether a device is best understood as a Purcell-enhanced interface, a coherent cavity-QED node, or a multimode fiber network element.

Nanofiber cavity QED systems therefore occupy a broad but coherent research area: they are unified by the guided $785$44 mode, evanescent light–matter coupling, and native compatibility with fiber networks, yet diversified by geometry, mode structure, and operating regime. Their present state includes single-atom strong coupling, collective vacuum Rabi splitting, ultralow-loss Fabry–Perot cavities, composite cavities with directional extraction, cryogenic high-$785$45 resonators, delayed-feedback ring systems, and many-body gate architectures. The central technical theme throughout is the same one identified in the earliest and most recent work alike: maximizing the useful ratio of coherent coupling to loss while preserving the fiber-native interface that makes nanofiber cavity QED attractive in the first place (Wuttke et al., 2012, Ruddell et al., 2020, Tanaka et al., 27 Jan 2026).