Quantum Sensor-Doped Thin-Film Cavities

• Quantum sensor-doped thin-film cavities are resonant photonic structures integrating quantum dopants with engineered optical films to control spontaneous emission via Purcell enhancement.
• They combine diverse material systems like Er:TiO2, diamond, hBN, and lithium niobate to enable applications in quantum sensing, spin–photon interfacing, memory, and topological light control.
• Optimized cavity geometries and interface engineering, including spectral tuning and coupling design, ensure efficient light–matter interaction while mitigating issues like spectral diffusion and interface disorder.

Quantum sensor-doped thin-film cavities are resonant thin-film photonic or x-ray structures in which a doped layer, defect ensemble, or resonant nuclear film functions as the active quantum element, so that cavity confinement, evanescent overlap, or guided-mode enhancement modifies spontaneous emission, storage, or reflectometric readout. Reported realizations span Er$^{3+}$ in TiO$_2$ on silicon, SnV$^-$ and SiV centers in thin-film diamond, VB$^-$ defects in hBN cavities integrated to optical fiber, fluorescent nanodiamonds and hBN nanoparticles in planar Fabry–Perot cavities, isotopically purified $^{167}\mathrm{Er}^{3+}$ in thin-film lithium niobate microrings, and $^{57}$Fe Mössbauer layers in multilayer x-ray cavities (Ji et al., 2023, Lee et al., 7 Nov 2025, Moon et al., 2024, Tibben et al., 5 Sep 2025, Yang et al., 14 May 2026, Zimmermann et al., 12 Jun 2025).

1. Physical principle and cavity-QED description

The central operation of these systems is cavity-mediated modification of the local density of optical states and of the input–output channel through which a quantum dopant is interrogated. In the weak-coupling description used repeatedly across the reported platforms, the spontaneous-emission enhancement is parameterized by the Purcell factor

$$F_P=\frac{3}{4\pi^2}\left(\frac{\lambda}{n}\right)^3\frac{Q}{V}\,\xi,$$

with $Q$ the cavity quality factor, $V$ the mode volume, and $\xi$ an overlap and polarization factor. The corresponding lifetime engineering is written as $_2$0, with the common approximation $_2$1 when the relevant branching ratio is close to unity; the associated cooperativity is commonly expressed as $_2$2 (Ji et al., 2023, Ding et al., 2024, Lee et al., 7 Nov 2025).

This language covers a wide range of devices but not a single dynamical regime. In Er:TiO$_2$3 nanophotonic cavities, the measured spectral-diffusion linewidths are much narrower than the cavity FWHM, placing the system in the bad-cavity regime where $_2$4 scales with $_2$5 (Ji et al., 2023). In thin-film diamond, the same formalism is used to separate enhancement of multiple zero-phonon-line branches and to extract branching ratios rather than only a single effective lifetime (Lee et al., 7 Nov 2025). In x-ray multilayers with embedded Mössbauer nuclei, the analogous cavity-mediated coherent and dissipative couplings are written through the electromagnetic Green’s function, yielding effective few-level or tight-binding descriptions rather than a single optical-mode picture (Heeg et al., 2013, Zimmermann et al., 12 Jun 2025).

A second unifying feature is that the dopant is rarely the structural resonator itself. Instead, the thin film acts as a host or overlay, while the resonant mode is provided by a silicon photonic crystal, a diamond nanobeam, a circular Bragg grating, a microring, a metallic Fabry–Perot stack, or a grazing-incidence x-ray waveguide. This separation between host and resonator is what makes thin-film integration unusually flexible: the active quantum medium can be deposited, transferred, or embedded after the optical cavity geometry has already been defined (Ji et al., 2023, Moon et al., 2024, Pettit et al., 1 May 2025).

2. Materials platforms and dopant classes

The reported material systems are heterogeneous, but they share a common strategy: a quantum-active dopant is introduced into a thin-film host whose morphology, roughness, and interface chemistry are then engineered to remain compatible with cavity optics.

Platform	Active thin film or dopant	Reported role
TiO$_2$6 on Si	Er$_2$7 in ALD or MBE-grown TiO$_2$8	telecom emitters, memories, foundry-integrated single ions
Thin-film diamond	SnV$_2$9 or SiV centers in diamond membranes	spin–photon interfaces and Purcell-enhanced readout
hBN thin films and nanoparticles	VB$^-$0 defects or hBN nanoparticles	fiber-integrated sensing and cavity-enhanced PL
Thin-film lithium niobate	isotopically purified $^-$1	programmable telecom quantum memory
Multilayer x-ray cavities	$^-$2Fe Mössbauer layers in C/Pt stacks	reflectometric sensing and topological control

In ALD-grown Er:TiO$^-$3 on silicon, plasma-enhanced ALD in thermal mode for TiO$^-$4 and oxygen-plasma-assisted cycles for ErO$^-$5 produced Er concentrations ranging from $^-$6 ppm down to $^-$7 ppm, with an ex situ O$^-$8 anneal at $^-$9 for 30 min converting amorphous films into predominantly anatase TiO$^-$0 while preserving sub-nm roughness (Ji et al., 2023). In the earlier MBE materials study, interface engineering rather than epitaxy was the dominant optical lever: on Si(100), polycrystalline anatase with bottom buffer and top capping layers yielded inhomogeneous linewidths as low as $^-$1 GHz and spectral diffusion of $^-$2 MHz, while polycrystalline films on Si were reported to be comparable to or better than epitaxial films on sapphire or SrTiO$^-$3 (Singh et al., 2022). This directly contradicts the common assumption that extended crystallographic perfection automatically produces the narrowest rare-earth lines.

A related but distinct oxide route is CVD-grown Er:Y$^-$4O$^-$5. Direct liquid injection CVD at approximately $^-$6 produced films with nominal Er$^-$7 concentration of $^-$8 ppm on Si, quartz, sapphire, YSZ, and an Si(111)/MBE-Y$^-$9O$^{167}\mathrm{Er}^{3+}$0 template. The substrate strongly altered morphology and texture: Si(111)/MBE-Y$^{167}\mathrm{Er}^{3+}$1O$^{167}\mathrm{Er}^{3+}$2 gave approximately $^{167}\mathrm{Er}^{3+}$3 [111] texture, while YSZ(001) yielded cube-on-cube epitaxy with rocking-curve FWHM of approximately $^{167}\mathrm{Er}^{3+}$4 but AFM RMS roughness of approximately $^{167}\mathrm{Er}^{3+}$5 nm. Optical emission, however, remained comparable across substrates, while a $^{167}\mathrm{Er}^{3+}$6 post-anneal broadened inhomogeneous lines and split the C2 site (Blin et al., 2024). This suggests that crystallographic alignment and optical homogeneity are not interchangeable metrics.

Thin-film diamond and hBN define a second large class in which the thin film itself is the nanophotonic membrane. The diamond devices use single-crystal membranes of $^{167}\mathrm{Er}^{3+}$7 or $^{167}\mathrm{Er}^{3+}$8 nm thickness for visible photonic-crystal cavities (Ding et al., 2024, Lee et al., 7 Nov 2025). The hBN platform instead uses an exfoliated film of about $^{167}\mathrm{Er}^{3+}$9 nm thickness patterned into a hole-based circular Bragg grating cavity and then transferred onto a commercial optical fiber (Moon et al., 2024). A third class uses low-cost composite films rather than crystalline membranes: fluorescent nanodiamonds in PVP and hBN nanoparticles in PMMA were spin-coated into centimeter-scale metallic Fabry–Perot cavities (Tibben et al., 5 Sep 2025). Finally, in thin-film lithium niobate, the active ions were incorporated during bulk crystal growth of the $^{57}$0 nm X-cut LN layer rather than by post-fabrication implantation, producing $^{57}$1-doped microrings with long-lived hyperfine shelving states (Yang et al., 14 May 2026).

3. Resonator architectures and photonic interfaces

The cavity geometries used in this area are highly diverse, but each is designed to place the doped film at a field antinode or in a large evanescent tail. In ALD Er:TiO$^{57}$2 on silicon, the active layer is not etched into the cavity itself; instead, one-dimensional Si photonic-crystal cavities are fabricated in SOI and then overlaid with TiO$^{57}$3, so that Er ensembles couple evanescently to the Si mode. These cavities use elliptically shaped holes, a 14-hole parabolic taper, two mirror holes per side, and one-sided coupling through a suspended inverse-tapered waveguide matched to a lensed fiber (Ji et al., 2023). A closely related foundry version on the AIM Photonics Quantum Flex 300 mm platform uses silicon nanobeam cavities with a parabolically tapered lattice and a bus waveguide plus integrated Sagnac loop mirror; the Er$^{57}$4:TiO$^{57}$5 layer is then deposited through backend oxide trenches above the optical mode (Pettit et al., 1 May 2025).

Thin-film diamond supports both one- and two-dimensional visible photonic-crystal cavities. One-dimensional nanobeam cavities with quadratic lattice-constant taper reached measured $^{57}$6, and two-dimensional slab line-defect cavities reached $^{57}$7, with a fiber-coupled 1D device showing loaded $^{57}$8 and approximately $^{57}$9 cavity–waveguide coupling efficiency (Ding et al., 2024). The SnV$$F_P=\frac{3}{4\pi^2}\left(\frac{\lambda}{n}\right)^3\frac{Q}{V}\,\xi,$$0 implementation also uses 1D photonic-crystal nanobeams in $$F_P=\frac{3}{4\pi^2}\left(\frac{\lambda}{n}\right)^3\frac{Q}{V}\,\xi,$$1 nm-thick single-crystal diamond membranes, with resonances tuned by controlled argon-gas condensation and laser-induced back-evaporation (Lee et al., 7 Nov 2025).

The hBN and polymer systems emphasize interface engineering rather than ultrahigh $$F_P=\frac{3}{4\pi^2}\left(\frac{\lambda}{n}\right)^3\frac{Q}{V}\,\xi,$$2. The hBN cavity is a hole-based circular Bragg grating with central defect radius approximately $$F_P=\frac{3}{4\pi^2}\left(\frac{\lambda}{n}\right)^3\frac{Q}{V}\,\xi,$$3 nm, radial period approximately $$F_P=\frac{3}{4\pi^2}\left(\frac{\lambda}{n}\right)^3\frac{Q}{V}\,\xi,$$4 nm, tangential spacing approximately $$F_P=\frac{3}{4\pi^2}\left(\frac{\lambda}{n}\right)^3\frac{Q}{V}\,\xi,$$5 nm, and nanohole radius approximately $$F_P=\frac{3}{4\pi^2}\left(\frac{\lambda}{n}\right)^3\frac{Q}{V}\,\xi,$$6 nm, transferred onto the pre-etched core of a $$F_P=\frac{3}{4\pi^2}\left(\frac{\lambda}{n}\right)^3\frac{Q}{V}\,\xi,$$7m-core multimode fiber without index-matching adhesive (Moon et al., 2024). The polymer devices are planar metallic Fabry–Perot cavities built from a $$F_P=\frac{3}{4\pi^2}\left(\frac{\lambda}{n}\right)^3\frac{Q}{V}\,\xi,$$8 nm Ag bottom mirror, a nanoparticle-doped polymer spacer, and a semi-transparent Ag top mirror, with the cavity length set by spin-coated film thickness (Tibben et al., 5 Sep 2025).

Thin-film lithium niobate occupies a distinct integrated regime. The resonator is a racetrack microring of perimeter $$F_P=\frac{3}{4\pi^2}\left(\frac{\lambda}{n}\right)^3\frac{Q}{V}\,\xi,$$9m with integrated gold electrodes along the straight sections, deliberately over-coupled in the empty state so that ion absorption can move the device toward critical coupling for impedance matching (Yang et al., 14 May 2026). By contrast, the x-ray multilayer architecture consists of vertically stacked thin-film cavities, each realized as $Q$0 nm C / $Q$1 nm $Q$2Fe / $Q$3 nm C with Pt mirrors and alternating Pt spacer thicknesses $Q$4 and $Q$5, probed at grazing incidence so that the interlayer couplings implement a non-Hermitian SSH chain (Zimmermann et al., 12 Jun 2025).

A recurring practical theme is spectral tuning. Si–Er devices were redshifted onto resonance by controlled N$Q$6 condensation at $Q$7 K (Ji et al., 2023). SnV$Q$8 and SiV cavities were tuned by argon-gas condensation or gas-condensation tuning near $Q$9–$V$0 K (Lee et al., 7 Nov 2025, Ding et al., 2024). The TFLN microring instead used intrinsic electro-optic tuning with measured efficiency of $V$1 GHz/V (Yang et al., 14 May 2026). The availability of a tuning mechanism is not incidental; it is usually essential because film deposition, implantation, and cavity nanofabrication introduce device-to-device resonance offsets.

4. Reported performance regimes

The strongest rare-earth result in ALD TiO$V$2 is a cavity-mediated shortening of Er lifetimes from $V$3s to $V$4s at $V$5 ppm and from $V$6s to $V$7s at $V$8 ppm, corresponding to ensemble-best Purcell factors of approximately $V$9 and $\xi$0 in cavities with $\xi$1 and $\xi$2, respectively (Ji et al., 2023). The same platform reported spectral-diffusion HWHM of $\xi$3 GHz at $\xi$4 ppm and $\xi$5 GHz in a bare $\xi$6 ppm waveguide, with $\xi$7 GHz after correction for the cavity lineshape in the cavity-coupled ensemble (Ji et al., 2023). In the foundry-integrated single-ion extension, a rutile Er ion showed a cavity-coupled lifetime of $\xi$8s from a $\xi$9 ms uncoupled lifetime, implying Purcell enhancement up to about $_2$00, together with antibunching $_2$01 and a representative Voigt linewidth of $_2$02 MHz (Pettit et al., 1 May 2025).

Thin-film diamond shows two rather different operating points. The SnV$_2$03 nanobeam study measured quality factors up to approximately $_2$04, lifetime reductions up to about $_2$05-fold, and a single-transition Purcell factor as high as $_2$06, while simultaneously extracting the intrinsic C/D branching ratio as $_2$07 (Lee et al., 7 Nov 2025). The higher-$_2$08 visible diamond platform instead emphasized cavity quality and fiber interfacing: it reported $_2$09 for a 1D cavity, $_2$10 for a 2D cavity, fiber-coupled loaded $_2$11, and a SiV zero-phonon-line Purcell factor of approximately $_2$12 inferred from lifetime reduction from about $_2$13 ns to about $_2$14 ns (Ding et al., 2024). Together, these results show that thin-film diamond already spans both high-$_2$15 passive photonics and moderate-$_2$16 cavity-enhanced spin readout.

The hBN and polymer platforms demonstrate that thin-film cavities need not pursue only ultrahigh $_2$17. The fiber-integrated hBN circular Bragg grating showed a simulated Purcell factor of approximately $_2$18, a measured cavity peak at approximately $_2$19 nm, about $_2$20 higher PL than a pristine hBN flake under objective excitation with fiber collection, and ODMR contrast of approximately $_2$21 in all-fiber operation (Moon et al., 2024). The centimeter-scale polymer Fabry–Perot cavities reported $_2$22–$_2$23 for FND devices and $_2$24–$_2$25 for hBN devices, NV decay-rate enhancement up to $_2$26, hBN PL decay enhancement up to $_2$27, and a $_2$28 times improved magnetic-field sensitivity of $_2$29 nm FNDs due to cavity-enhanced ODMR contrast and PL brightness (Tibben et al., 5 Sep 2025). These systems trade extreme $_2$30 for area, simplicity, and direct deployability.

Thin-film lithium niobate extends the topic from emission control to storage. In a $_2$31-doped microring with loaded $_2$32, external coupling rate $_2$33 MHz, intrinsic loss rate $_2$34 MHz, and ion-induced loss $_2$35 MHz, cavity-enhanced atomic-frequency-comb storage reached an on-chip efficiency of $_2$36 for $_2$37-ns storage (Yang et al., 14 May 2026). The same device family showed Hahn-echo optical coherence time $_2$38s, single-component comb lifetime $_2$39 s, dynamic electro-optic routing up to $_2$40 MHz with inter-channel crosstalk below $_2$41, and storage of time-energy-entangled telecom photons with entanglement witness $_2$42 (Yang et al., 14 May 2026). Thin-film cavities in this regime are no longer only brightness enhancers; they are programmable light–matter interfaces.

5. Functional regimes: sensing, memory, topology, and engineered dissipation

Experimentally, the reported functions cover quantum sensing, spin–photon interfacing, quantum memory, and reflectometric identification of collective states. hBN VB$_2$43 cavities enabled remote sensing of a ferromagnetic material and arbitrary magnetic fields through a fiber tip (Moon et al., 2024). Polymer cavities improved wide-field NV magnetometry by increasing ODMR contrast and brightness (Tibben et al., 5 Sep 2025). Er:TiO$_2$44 on silicon was positioned as a platform in which high doping densities can support on-chip amplifiers and lasers, while dilute concentrations can realize single-ion quantum memories (Ji et al., 2023). The TFLN implementation made this memory function explicit by combining persistent AFC preparation, impedance matching, and direct telecom electro-optic control in a single thin-film resonator (Yang et al., 14 May 2026).

The x-ray literature adds a qualitatively different regime in which the dopant layers act simultaneously as resonant absorbers and as spectrally sharp probes of the cavity field. In the non-Hermitian SSH proposal, ten vertically stacked thin-film cavities with embedded $_2$45Fe layers realize alternating couplings $_2$46 and $_2$47, with topological phase emerging for $_2$48 and winding number $_2$49; for the representative choice $_2$50 nm and incidence angle $_2$51 mrad, the transition occurs near $_2$52 (Zimmermann et al., 12 Jun 2025). The topological phase is predicted to appear in reflectivity as a single dominant Lorentzian-like feature dominated by edge states, whereas the trivial phase gives a two-peak structure with an interference dip (Zimmermann et al., 12 Jun 2025). This is a cavity-sensing concept in which topology modifies not only robustness but also the readout morphology itself.

A broader theoretical tradition treats thin-film x-ray cavities as inverse-design targets. For a single resonant Mössbauer layer in a multilayer cavity, macroscopic-QED and transfer-matrix methods are used to optimize collective Lamb shift, superradiant broadening, field enhancement, and visibility, with the effective parameters derived from the cavity Green’s function (Diekmann et al., 2021). Closely related quantum-optical modeling of Pt/C/$_2$53Fe/C/Pt stacks showed how magnetization and polarization selection engineer effective few-level schemes, superradiant broadening, collective Lamb shifts, and spontaneously generated coherences in grazing-incidence reflectivity (Heeg et al., 2013). These studies suggest that, in thin-film cavity sensing, the electromagnetic environment itself can be designed as aggressively as the quantum dopant.

Engineered dissipation is another recurrent theme. Quantum embedded superstates were proposed for cavity-coupled V-type emitters, with an ultranarrow line in the strong-coupling regime when the quenching condition $_2$54 is satisfied (Nefedkin et al., 2020). A different sensing proposal based on cooperative atom–light cat states in cavity QED showed estimated metrological gains of $_2$55–$_2$56 dB below the standard quantum limit for weak field sensing (Lewis-Swan et al., 2019). In yet another direction, a giant-cavity sensor with multiple position-dependent couplings was proposed to achieve shot-noise-level output noise and about one order-of-magnitude enhancement in signal-to-noise ratio per photon (Zhu et al., 2022). These are not all experimentally realized thin-film devices, but they delineate the theoretical envelope into which thin-film cavity platforms are being pushed: topology, non-Hermiticity, time reversal, and structured dissipation are increasingly treated as sensing resources rather than parasitic complications.

6. Design constraints, misconceptions, and recurring engineering rules

Several constraints recur across otherwise unrelated platforms. The first is the density–coherence trade-off. In Er:TiO$_2$57, higher Er density increased brightness but also introduced ion–ion quenching, shorter $_2$58, and broader diffusion, while dilute films improved lifetime and spectral quietness but required cavity enhancement to recover useful photon rates (Ji et al., 2023). In thin-film diamond, multiple emitters in one SnV cavity complicated spectroscopy and lifetime analysis, and post-fabrication linewidths broadened because emitters were brought close to etched surfaces (Lee et al., 7 Nov 2025). In polymer cavities, hBN nanoparticles showed ultrafast decay and high apparent enhancement, but the reported analysis explicitly noted that the largest effective Purcell values likely involve plasmonic coupling, near-field effects, or local geometry perturbations beyond pure cavity LDOS enhancement (Tibben et al., 5 Sep 2025). The common implication is that “higher enhancement” is not a single material number; it depends on how cleanly the cavity picture survives the full local environment.

A second misconception concerns crystallinity. The available data do not support a universal hierarchy in which epitaxial films always outperform polycrystalline ones. Polycrystalline Er:TiO$_2$59 on Si achieved narrower inhomogeneous lines than epitaxial TiO$_2$60 on lattice-matched substrates, and the authors attributed the dominant broadening not to extended defects but to local charged defects and interface chemistry (Singh et al., 2022). Conversely, CVD Er:Y$_2$61O$_2$62 on YSZ did yield epitaxy, but also showed rough $_2$63D columnar morphology and did not produce the longest lifetimes, while high-temperature post-annealing degraded spectral homogeneity (Blin et al., 2024). The engineering lesson is not that epitaxy is unimportant, but that it must be evaluated together with roughness, interdiffusion, interface oxides, and point-defect control.

Across the reported platforms, the most consistent practical design rules are straightforward. The active doped layer should be placed at a field antinode or evanescent maximum; the host should preserve smooth surfaces after crystallization; and the cavity linewidth should be designed with respect to the relevant diffusion or homogeneous linewidth rather than in isolation (Ji et al., 2023, Singh et al., 2022). For absorption-based memories or reflectometric sensing, raw $_2$64 is insufficient unless coupling is also engineered: the TFLN work deliberately chose strong over-coupling in the empty device so that ion absorption could drive the microring toward critical coupling, while the x-ray SSH proposal required Pt spacer thicknesses of at least $_2$65 nm to suppress long-range couplings and preserve the intended effective topology (Yang et al., 14 May 2026, Zimmermann et al., 12 Jun 2025). For visible solid-state emitters, polarization alignment and orientation are equally critical: the SnV study showed that the orthogonal C and D dipoles require cavity-mode polarization control, and the authors estimated that orienting the cavity approximately $_2$66 from the $_2$67 axis could improve the Purcell factor by approximately $_2$68 relative to the reported angled orientation (Lee et al., 7 Nov 2025).

Thin-film cavity integration therefore appears not as a single device class but as a general materials-and-resonator strategy. In the reported literature, the same strategy already supports room-temperature fiber sensors, centimeter-scale magnetic-field imagers, visible and telecom spin–photon interfaces, on-chip programmable quantum memories, and topological x-ray reflectometers (Moon et al., 2024, Tibben et al., 5 Sep 2025, Ding et al., 2024, Ji et al., 2023, Yang et al., 14 May 2026, Zimmermann et al., 12 Jun 2025). The principal unresolved problem is no longer whether thin films can host quantum dopants, but how completely surface chemistry, interface disorder, and cavity loading can be controlled while retaining the fabrication scalability that made the thin-film approach attractive in the first place.