Skip to content

High-efficiency vertical emission spin-photon interface for scalable quantum memories

FieldValue
JournalQuantum
Published2026-03-20
AuthorsSiavash Mirzaei-Ghormish, Jeddy Bennett, Ryan M. Camacho
DOI10.22331/q-2026-03-20-2035

The modeled dual-grating diamond microdisk reaches 96% collection efficiency within a numerical aperture of 0.7 and a 95% overlap with a Gaussian far-field mode. Using an average spectral efficiency of 99.6%, the paper reports (\eta_{\mathrm{tot}} = 95.5%) under its definition (\eta_{\mathrm{tot}} = \eta_{\mathrm{ZPL}}\eta_{\mathrm{col}}). These results come from simulations of a proposed device, not measurements of a fabricated sample.

The target is vertical emission of spin-polarization-entangled photons from SnV centers. The design combines a diamond microdisk resonator containing an SnV center with two hexagonal perturbation gratings above the disk. The first grating occupies the near field and redirects the resonant whispering-gallery mode around the optical axis. The second occupies the intermediate field and acts as a spatial filter. Destructive interference at high numerical aperture suppresses side lobes and concentrates intensity near the optical axis.

The grating layers use silicon oxynitride with refractive index 1.8. A silicon dioxide spacer with refractive index 1.4 separates them. The analytical solution describes the intermediate field as orthogonal LHCP and RHCP waves with topological charges (L-1) and (L+1). Under the selection rule (L=1), the far-field intensity remains concentrated near the optical axis. The final pattern mainly reflects the first grating, while the second supplies spatial filtering.

A cavity design faces a direct tradeoff between a high Purcell factor and photon extraction. High-Q, small-mode-volume cavities can raise the Purcell factor but retain photons strongly. The paper notes that whispering-gallery modes have quality factors of order (10^4) and mode volumes above (3(\lambda/n_{\mathrm{eff}})^3). These properties produce moderate Purcell factors despite strong field confinement. The separate gratings provide outcoupling without placing perturbation holes directly in the diamond disk surface.

The analysis uses a whispering-gallery mode with mode number (M=17), obtained from FDTD. Bayesian optimization tunes the disk radius, disk thickness, both lattice constants, both hole heights, and both hole radii within bounded ranges.

ParameterOptimized value
Disk radius, (r_d)(1.4687\lambda_0)
Disk thickness, (h)(0.9491\lambda_0)
First grating lattice constant, (a_1)(0.5234\lambda_0)
Second grating lattice constant, (a_2)(0.3406\lambda_0)
First hole height, (d_1)(0.3561\lambda_0)
Second hole height, (d_2)(0.3561\lambda_0)
First hole radius, (r_h1)(0.1875\lambda_0)
Second hole radius, (r_h_2)(0.1562\lambda_0)

All values are normalized to the design wavelength. At (\lambda_0=640) nm, the optimized hole radii are 120 nm in the first layer and 100 nm in the second. These dimensions are outputs of the optical optimization, not reported dimensions for a fabricated device.

The dipole model represents every hole as an electric dipole with (p_n=\alpha E_n). It superposes coherent fields from the dipoles and uses Fresnel propagation in the intermediate field. The model uses Fraunhofer propagation for the far field. Only first-order electric-dipole scattering is retained. The authors calibrate the model against one FDTD control simulation and use independent parameter sweeps to check robustness.

At full alignment, the whispering-gallery mode has Purcell factor 62. Translating the first grating by half a lattice constant produces Purcell factor 66. The authors assign an average spectral efficiency of 99.6% to these configurations and neglect the small resonance shift between them.

ConfigurationPurcell factorCollection at NA 0.7(\eta_{\mathrm{tot}})
Fully aligned6296%95.5%
First grating offset by (a/2)6682%81.6%

The 95% Gaussian overlap applies to the aligned far-field pattern. The reported (\eta_{\mathrm{tot}}) does not include that overlap factor. Earlier extra-layer microdisk designs were limited to approximately 40% to 50% collection in the paper’s comparison.

The paper’s ideal mode reference uses (\eta_{\mathrm{Gauss}}\eta_{\mathrm{col}}=0.95\times0.96\approx91%). Including spectral weighting gives (\eta_{\mathrm{ZPL}}\eta_{\mathrm{Gauss}}\eta_{\mathrm{col}}\approx90.7%). This distinction separates the 95.5% metric from a measured fiber coupling result.

Alignment is parameterized by (s=\Delta/a), where (\Delta) is the lateral offset of the reference scatterer and (a) is the lattice constant. The sweep runs from (s=0) for perfect alignment to (s=1) for full misalignment. Across most first-grating offsets, collection exceeds 50%, and the intensity remains centered. Translating the second grating causes little change because its dense holes compensate local position shifts.

Equal hole radii of 140, 150, and 160 nm in both layers retain collection above 85% and a Gaussian-like profile. Hole heights of 220, 250, and 300 nm also retain centered, symmetric profiles. Reported tolerance bands are ±10 to 20 nm for hole radius and ±20 to 30 nm for hole height. Similar errors in spacer thickness and interlayer spacing cause modest changes. Circular, rectangular, and triangular holes all preserve a symmetric, Gaussian-like far field.

The design targets SnV centers and can transfer emitter information to the far field for arbitrary dipole orientation. An in-plane dipole couples mainly to the TE-like mode and produces a symmetric, Gaussian-like profile. Tilting the dipole toward vertical produces hybrid TE and TM coupling. A vertical dipole couples mainly to the TM-like mode, yielding multiple lobes and a broader angular distribution. Collection remains above 40% at NA 0.7 for every examined orientation.

For the fully aligned 96% collection case, the paper estimates the Purcell factors needed for total efficiency above 90%. The estimates are (F_p=4) for SnV, (F_p=9) for SiV, and (F_p=412) for NV. SnV and SiV therefore need much lower Purcell factors than NV. The authors also suggest placing multiple emitters inside the cavity and selectively exciting those aligned with the desired cavity mode.

A representative fabrication flow uses EBL and O2 ICP-RIE to pattern the diamond microdisk. Ti/Au alignment marks are defined early and reused for every later overlay. PECVD forms the SiO2 spacer and the SiON/SiN grating layers. EBL patterns each grating, followed by CHF3/O2 or CF4-based ICP-RIE transfer. The fluorocarbon step leaves the diamond layer unaffected.

The paper does not fabricate or test this stack. Its process flow is a representative route based on established diamond nanophotonics methods. Reported EBL alignment capability is 20 to 30 nm overlay and less than 0.5° rotation over fields of a few hundred microns. Small rotational errors mainly impart a weak azimuthal phase that averages out in the far field.

On an AMD EPYC 7763 cluster, one dipole-model far-field pattern took 1 s on a single core. One full FDTD simulation took 7 h on 128 cores. The authors convert this to an approximately (3.2\times10^6) core-time speedup.

FDTD and dipole patterns agree qualitatively, but the authors say higher-order multipoles should improve model accuracy. They also identify extraction versus quality-factor trade-offs and residual multipole leakage as practical limits.

The proposed interface needs a patterned diamond microdisk resonator and Ti/Au alignment marks for electron-beam lithography overlays. 6CCVD’s diamond materials catalog lists MPCVD-grown single-crystal diamond plates and wafers with thickness from 0.1 to 500 μm and Au or Ti metallization. Process-specific layout and overlay questions can be directed to engineering support.

View Original Abstract

We present an efficient spin-photon interface for free-space vertical emission coupling. Using a dipole model, we show that our design achieves a far-field collection efficiency of 96% at the numerical aperture of 0.7 with a 95% overlap to a Gaussian mode. Our approach is based on a dual perturbation layer design. The first perturbation layer extracts and redirects the resonant mode of a diamond microdisk resonator around the optical axis. The second perturbation layer suppresses side lobes and concentrates most of the light intensity near the center. This dual-layer design enhances control over the farfield pattern and also reduces alignment sensitivity. Additionally, the implemented dipole model performs calculations <mml:math xmlns:mml=“http://www.w3.org/1998/Math/MathML”> <mml:mn>3.2</mml:mn> </mml:math> <mml:math xmlns:mml=“http://www.w3.org/1998/Math/MathML”> <mml:mrow class=“MJX-TeXAtom-ORD”> <mml:mo>&#x00D7;</mml:mo> </mml:mrow> </mml:math> <mml:math xmlns:mml=“http://www.w3.org/1998/Math/MathML”> <mml:msup> <mml:mn>10</mml:mn> <mml:mn>6</mml:mn> </mml:msup> </mml:math> times faster than full-wave FDTD simulations. These features make the design promising for quantum information applications.