Photonic crystal cavities based on suspended yttrium iron garnet nanobeams

TL;DR

This work demonstrates the fabrication and optical characterization of an air-suspended photonic crystal nanobeam cavity made from yttrium iron garnet (YIG) designed to host photons, phonons, and magnons within a shared nanoscale volume. Using focused-ion-beam milling with a sacrificial Al mask on an $840\,\text{nm}$-thick YIG film on GGG, finite-element design targets a single optical mode at $\omega_0/2\pi \approx 187\,\text{THz}$, a phononic mode at $\Omega_m/2\pi \approx 1.52\,\text{GHz}$, and a magnonic mode near $\omega_{mag} = 2\pi \times 11.59\,\text{GHz}$ under $H_{ext}=400\,\text{mT}$. Experiment reveals an optical resonance at $\lambda \approx 1634.8\,\text{nm}$ with an intrinsic quality factor $Q_{int} \approx 2\times10^3$, well below the simulated $\sim10^6$ due to fabrication-related losses, notably lateral hole-row misalignment and aspect-ratio errors. This discrepancy motivates design adaptations (e.g., straight-groove lattices) and process optimizations to reach sideband-resolved operation and enable coupling between photons, phonons, and magnons for quantum transduction and magnomechanics in YIG-based on-chip platforms.

Abstract

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.

Figures (5)

• Figure 1: Design of the YIG optomechanical crystal (OMC). (a)-(c) Amplitude profile of the confined optical mode's electric field, illustrating that most of the field is concentrated in the center of the beam and within the YIG, which is essential for achieving a high phonon-photon-magnon coupling rate. (b) A zoomed view of the cavity area of the waveguide. (c) A cross-sectional view of the center of the waveguide. (d) Optical band diagram of the OMC, where the solid blue lines represent the guided transverse electric (TE-like) modes, and the dashed magenta lines depict the guided transverse magnetic (TM-like) modes of the waveguide. The shaded pink region indicates the optical semi-band gap, intersected by the light line (green solid line), while the continuous leaky modes coupling to air are represented by the shaded blue area above the light line. The confined mode at the edge of the first Brillouin zone is highlighted by the cyan horizontal dashed line.
• Figure 2: (a)-(f) Fabrication process flow of the YIG optomechanical crystal cavity (OMC). (a) The process begins with the deposition of a sacrificial aluminum (Al) layer on the YIG substrate. (b)-(d) The Xe ion beam is then used for milling the trenches in a three-step process: initial coarse trench milling at 1–3 nA and 30 kV, followed by intermediate shaping at 100–300 pA, and finally a polishing step at approximately 30 pA. This sequence minimizes sidewall roughness, precisely shapes the beam, and accurately drills the elliptical holes. (e) The final step involves removing the Al layer using a potassium hydroxide (KOH) wet etch to reveal the completed YIG OMC. (f) The result is a suspended structure ready for characterization. (g) Scanning electron micrograph (SEM) of the prepared device before the Al removal step. (h) SEM of the finished device.
• Figure 3: (a) Schematic of the optical characterization setup. The setup uses a Santec TSL570 tunable laser source, with light from this source passing through a polarization controller to maximize coupling efficiency between the tapered dimpled fiber and the device. The transmitted signal is then detected by a Resolved Instruments DPD80 photodetector and analyzed. (b) The transmission spectrum reveals a resonance at $\lambda = 1634.8 \text{ nm}$. By fitting a Lorentzian function to the data, we extract an internal quality factor of 2000 and an external coupling rate of $\kappa_{ex} = 9 \text{ GHz}$. The red solid line represents the Lorentzian fit, while the blue dots show the experimental data points.
• Figure 4: (a),(b) Displacement field of the mechanical mode (not to scale) at $\Omega_{\,m} / 2\pi \approx \qty{1.52}{GHz}$. (c) Mechanical band diagram of the phononic waveguide. The $x$-symmetric and $x$-antisymmetric modes are represented by solid blue lines and dashed magenta lines, respectively. The bandgap formed by the $x$-symmetric modes is shaded in pink, with the supported confined mode indicated by the red horizontal dashed line.
• Figure 5: Magnonic mode simulations. (a) Spatial profile of the $y$-component of the reduced magnetization ($|m_y|$) at $\omega_{\textrm{mag}} = \qty{11.59}{GHz}$, showing the confined magnonic mode within the structure. (b) Mode spectrum at the edge of the Brillouin zone ($k_z = \pi/a$, where $a = \qty{650}{nm}$ is the lattice constant), indicating the presence of the localized mode. (c) Band diagram of the backward volume waves within the Brillouin zone, illustrating the propagation characteristics of the spin waves in the magnonic crystal. It should be noted that, while the optical and mechanical modes were optimized using COMSOL, the magnonic modes have not undergone a similar optimization process. MuMax3 does not currently support the same parameterization strategies as COMSOL, limiting our ability to refine the geometry for maximum magnonic performance in this iteration.