Skip to content

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

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

Simulations predict 96% collection efficiency at a numerical aperture of 0.7, with 95% overlap between the emitted far-field pattern and a Gaussian mode. The proposed SnV spin-photon interface reaches a calculated total efficiency of 95.5% under full alignment. Two dielectric grating layers redirect a whispering-gallery mode toward the optical axis and suppress off-axis light. These results describe a designed and simulated device, not a fabricated prototype.

The paper defines total efficiency as η_tot = η_ZPL η_col, combining spectral efficiency with collection efficiency. In the aligned configuration, collection reaches 96% at NA = 0.7. The far-field pattern has 95% overlap with a Gaussian mode, and the authors report η_tot = 95.5%. This total follows the η_ZPL η_col definition and does not include the Gaussian overlap.

Translating the first grating by half a lattice constant while holding the second fixed reduces collection to 82%. The resulting total efficiency is 81.6%, although the intensity remains concentrated near the center.

The paper estimates coupling into an ideal Gaussian mode at about 91% from 0.95 × 0.96. Including spectral weighting lowers that reference to 90.7%. Its cited near-field grating design reached approximately 40% to 50% collection, below the 96% predicted here.

The device combines a diamond microdisk resonator containing an SnV color center with two hexagonal perturbation grating layers. The first layer sits directly above the disk in the near field. It extracts and redirects the resonant mode into a concentric profile around the optical axis. The second layer occupies the intermediate field and acts as a focusing element and near-lossless spatial filter. Destructive interference removes intensity at high numerical aperture and suppresses side lobes.

Both grating layers use silicon oxynitride with refractive index n_Ox = 1.8. A silicon dioxide spacer with n_SiO2 = 1.4 separates them. The analyzed whispering-gallery mode has mode number M = 17, obtained from FDTD simulations.

The first layer produces orthogonal circular components with topological charges L - 1 and L + 1. The selection rule L = 1 produces nonzero intensity near the optical axis. The second layer adds Bessel-function modulation that confines the result toward the axis. The final axial pattern depends mainly on the first layer, provided the second layer places dense scatterers in high-intensity regions.

An out-of-plane color-center dipole component cannot reach the far field without the intermediate perturbation layer. The two-layer structure transfers this component to the far field. This removes the need to control color-center orientation for far-field transfer.

Bayesian optimization targeted collection efficiency and Gaussian overlap within bounded parameter ranges. Independent parameter sweeps checked the optimized design’s robustness.

ElementParameterOptimized value
MicrodiskRadius r_d1.4687 λ_0
MicrodiskThickness h0.9491 λ_0
First gratingLattice constant a_10.5234 λ_0
Second gratingLattice constant a_20.3406 λ_0
First gratingHole height d_10.3561 λ_0
Second gratingHole height d_20.3561 λ_0
First gratingHole radius rh_10.1875 λ_0
Second gratingHole radius rh_20.1562 λ_0

The optimized hole radii are 120 nm and 100 nm at λ_0 = 640 nm. The representative process starts with Ti/Au alignment marks for later overlays. Electron-beam lithography and O2 ICP-RIE pattern the diamond microdisk. PECVD forms the SiO2 spacer and the SiON/SiN grating layers. Electron-beam lithography and CHF3/O2 or CF4-based ICP-RIE pattern the gratings. The dielectric etch leaves the diamond unaffected. The paper presents this as a proposed flow based on established diamond nanophotonics methods.

The process discussion cites modern EBL overlay capability as 20 nm to 30 nm. Rotational alignment is below 0.5° over array fields of a few hundred microns. The dense second grating makes small rotational errors mostly average out in the far field.

Every grating hole is modeled as a dipole with moment p_n = α E_n. This approximation applies when hole sizes are much smaller than the resonance wavelength. The optimized radii of 120 nm and 100 nm are well below the 640 nm design wavelength.

One full-wave FDTD control simulation calibrates the first-order scattering response. Higher-order multipoles are not included. The model computes a far-field pattern in 1 s on one core. A full 3D FDTD simulation takes 7 h on 128 cores of an AMD EPYC 7763 cluster. Accounting for core count gives the reported 3.2 × 10^6 speedup.

The speed makes bounded optimization and tolerance sweeps practical. First-order truncation remains the main model limit when hole sizes approach the resonance wavelength.

Alignment, emitter, and fabrication tolerance

Section titled “Alignment, emitter, and fabrication tolerance”

Across most first-grating alignments, collection efficiency exceeds 50%. The intensity remains centered in every swept alignment. The calculation neglects the very small resonance shift between the two alignments. Translating the second grating has little effect on collection or the far-field profile. The authors attribute this weak sensitivity to the dense scatterer distribution.

For dipole orientations from vertical to in-plane, the profile changes from broad, multilobed emission to a centered, Gaussian-like pattern. The design remains optimized for TE-like coupling, but any arbitrary dipole orientation retains more than 40% collection at NA = 0.7. A vertical dipole primarily excites the TM-like mode and produces multiple high-intensity lobes. Far-field transfer no longer requires orientation control, although collection efficiency still varies with dipole angle.

Hole-radius sweeps from 140 nm to 160 nm give collection above 85% and preserve a Gaussian-like profile. Varying hole height from 220 nm to 300 nm leaves the far-field profile and collection stable. The paper reports radius tolerance of ±10 nm to 20 nm and height tolerance of ±20 nm to 30 nm. Circular, rectangular, and triangular holes retain a centered Gaussian-like profile.

The tolerance result preserves the centered spatial profile more reliably than peak collection. A half-lattice translation reduces aligned collection from 96% to 82% while leaving the beam centered.

Color-center requirements and design limits

Section titled “Color-center requirements and design limits”

To compare SnV, SiV, and NV centers, the authors use 96% collection and 92% zero-phonon-line efficiency for the aligned design. They calculate the Purcell factors needed to exceed 90% total efficiency. SnV requires F_p = 4, based on a ZPL rate ratio of 0.25. SiV requires F_p = 9, based on a ratio of 0.66. NV requires F_p = 412, based on a ratio of 32.3.

The paper identifies SnV and SiV as compatible with low Purcell factors. NV requires a much larger factor, which raises the extraction and quality-factor trade-off. Under the earlier product definition, the stated 96% collection and 92% ZPL efficiency combine to a value below 90%. The paper does not reconcile this result with its stated threshold.

Separately, the SnV alignment analysis reports Purcell factors of 62 for full alignment and 66 for half-lattice misalignment. It reports average spectral efficiency of 99.6%. The alignment analysis therefore uses a different spectral-efficiency assumption from the cross-center calculation.

The modeled 95% Gaussian overlap is compatible with collection into a fiber’s fundamental mode. The authors identify extraction and quality-factor trade-offs and residual higher-order multipole leakage as limits on fabricated-device efficiency.

The interface needs a diamond microdisk resonator carrying SnV centers, with patterned SiON/SiN layers above it. 6CCVD’s diamond materials catalog lists single-crystal and polycrystalline MPCVD diamond plates and wafers up to 125 mm, with thicknesses from 0.1 to 500 micrometers. The paper’s nanoscale overlay requirements should be reviewed through engineering support before defining the custom fabrication flow.

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.