Photonic crystal cavities based on suspended yttrium iron garnet nanobeams

We report the fabrication and optical characterization of an air-suspended photonic crystal nanobeam cavity in yttrium-iron-garnet (YIG) realized by focused-ion-beam milling. YIG’s combination of low optical loss and ferrimagnetism makes it highly attractive for quantum technologies, yet prior work has largely been focused on millimeter-scale spheres and simple microstructures, hindering true on-chip integration. Demonstrating nanometer-scale patterning in a suspended geometry therefore represents an important advance. Finite-element simulations predict that the same structure supports a flapping-type mechanical mode at $Ω/ 2π\approx 1.52 ,\text{GHz}$ and a backward-volume spin-wave mode at $Ω/ 2π= 11.59 ,\text{GHz}$ under an in-plane bias field. Although we measure only the photonic resonance (intrinsic $Q \sim 2 \times 10^{3}$) in this study, the device lays the groundwork for future exploration of coupled photon-phonon-magnon dynamics once higher optical quality factors are achieved.

💡 Research Summary

This work reports the first demonstration of an air‑suspended photonic‑crystal nanobeam cavity fabricated from yttrium‑iron‑garnet (YIG) using focused‑ion‑beam (FIB) milling. YIG is a uniquely attractive material for hybrid quantum systems because it combines ultra‑low magnetic damping (Gilbert α≈10⁻⁵) with very low optical absorption in the near‑infrared. However, its incompatibility with standard silicon‑based micro‑fabrication has limited previous YIG devices to millimeter‑scale spheres or simple planar structures that support only a single type of excitation. By patterning a 840‑nm‑thick (111) YIG film into a 1.25 µm‑wide, 20 µm‑long rectangular nanobeam and drilling a periodic array of elliptical holes, the authors create a one‑dimensional photonic crystal with a 40 THz TE‑like bandgap. Finite‑element simulations predict a single confined optical mode at ω₀/2π≈187 THz (≈1600 nm) deep within this gap, a flapping‑type mechanical mode at Ωₘ/2π≈1.52 GHz, and a backward‑volume spin‑wave (magnon) mode at ωₘ/2π≈11.59 GHz when an in‑plane bias field is applied.

The fabrication flow relies on Xe plasma FIB rather than the more common Ga‑FIB to avoid gallium contamination. A sacrificial 50 nm Al layer is first deposited to act as a heat sink, a redeposition catcher, and a barrier against ion implantation. Milling proceeds in three stages—coarse trenching (1–3 nA, 30 kV), intermediate shaping (100–300 pA), and fine polishing (≈30 pA)—to define the bridge, undercut the beam, and finally drill the sub‑micron elliptical holes. After milling, the Al mask is removed by a 32 % KOH wet etch at 80 °C, followed by critical‑point drying to preserve the suspended geometry. Scanning electron microscopy confirms the intended geometry but also reveals sidewall roughness on the order of tens of nanometers and a lateral offset of the hole row of roughly 100 nm.

Optical characterization is performed by coupling light from a tunable laser (Santec TSL‑570) into a tapered, dimpled fiber that evanescently excites the TE‑like cavity mode. The transmission spectrum shows a resonance at λ = 1634.8 nm. Lorentzian fitting yields an intrinsic quality factor Q_int≈2 × 10³ and an external coupling rate κ_ex≈9 GHz. This measured Q is two orders of magnitude lower than the radiation‑limited value (≈10⁶) predicted by simulations. The authors attribute the discrepancy primarily to three loss mechanisms: (i) surface damage and shallow Xe implantation from the FIB, (ii) sidewall scattering due to nanometer‑scale roughness, and (iii) geometric misalignment of the hole lattice. Simulations that incorporate the measured 100 nm offset and the slight aspect‑ratio error of the ellipses reduce the ideal Q from 10⁶ to ≈10⁴, and further inclusion of realistic roughness brings it close to the observed 2 × 10³, indicating that the lateral misplacement is the dominant factor.

Mechanical and magnetic simulations show that the same nanobeam simultaneously supports a confined flexural mode at 1.52 GHz (with strain concentrated near the beam center) and a backward‑volume magnon mode at 11.59 GHz. The calculated single‑photon‑single‑phonon coupling rate g₀≈50 kHz is modest compared to silicon‑based optomechanical crystals, reflecting YIG’s relatively small photoelastic coefficients. Nevertheless, the co‑localization of optical, mechanical, and magnonic fields in a sub‑micron volume opens a pathway to strong three‑mode interactions once the optical Q is improved.

The paper concludes with a clear roadmap for future improvements: replace the elliptical hole lattice with a straight‑groove design that is less sensitive to lateral placement errors; introduce surface‑treatment steps (thermal oxidation, oxide removal) to bring sidewall roughness down to a few nanometers; and further optimize FIB parameters or explore alternative mask strategies to minimize ion implantation. Achieving an optical Q > 10⁵ would bring the system into the resolved‑sideband regime (κ < Ωₘ), enabling efficient readout of the mechanical mode, coherent photon‑phonon‑magnon conversion, and applications such as ultra‑sensitive torque magnetometry, microwave‑to‑optical quantum transduction, and hybrid quantum information processing.