Skip to content

Color-Center-Compatible Freestanding Diamond Directional Couplers for Quantum Photonics

FieldValue
JournalACS Photonics
Published2026-09-02
AuthorsColin Sauerzapf, Tom Jäger, Jonathan Enßlin, Oliver von Berg, Vladislav Bushmakin
DOI10.1021/acsphotonics.6c01594

Five nominally 200 nm-separated freestanding diamond waveguides produced a mean coupling ratio of 46(16)% at room temperature and λ0 ≈ 637 nm. An SnV− center in a separate, nominally identical sample retained optical coherence after fabrication, including a 6.4(5) ns lifetime. The study combines a near-balanced on-chip beam splitter with a solid-state quantum emitter in bulk single-crystal diamond.

The coupling ratio is Cmeas = P2/(P1 + P2), where P1 and P2 are detected powers at output ports p1 and p2. This normalization measures the power split only and excludes optical losses. The 50:50 design target corresponds to C = 50%.

Five devices were fabricated at each designed separation. All were released and optically accessible. Room-temperature transmission used a TIR coupler for input and sequential collection at both outputs.

Designed separationMeasured separationMeasured widthMean coupling at 637 nmSupports per device
200 nm221(9) nm287(9) nm46.36(16.29)%1 to 5
250 nm279(9) nm283(9) nm6.26(4.70)%1 to 5
300 nm335(9) nm294(9) nm7.71(1.83)%1 to 5

Parentheses report standard deviations. All five nominally 200 nm-separated devices transferred appreciable power to p2. Only the 250 and 300 nm design groups showed small average transfers. The observed spread is consistent with fabrication-induced geometry changes within the coupling region.

Eigenmode simulations supplied the effective indices of symmetric and antisymmetric TE supermodes. For a 50:50 split, the coupling length follows LC,50:50 = λ0/(4Δneff), where Δneff is their index difference. The model fixed the outer undercut angle at 42(2)°, based on SEM measurements of reference waveguides.

The selected design used a 300 nm waveguide width and a 20 µm coupling length. It predicted a 280 nm separation at 637 nm and a 250 nm separation at 620 nm. A three-dimensional FDTD simulation at 620 nm returned 51% coupling. The authors attribute the slight overcoupling to residual coupling in the transitions and numerical differences from the eigenmode model.

Smaller separations increase evanescent overlap but raise sensitivity to fabrication errors. Longer coupling lengths increase the device footprint and potential propagation loss. The 20 µm design balanced those constraints.

Electron-beam lithography patterned a positive resist on optical-grade bulk single-crystal diamond. A hard mask of 1 nm chromium and 80 nm titanium followed resist development and lift-off. A surface-normal oxygen RIBE transferred the pattern into the diamond.

The second oxygen RIBE tilted the beam to 45° while the sample rotated at 2 rpm. The supplementary process used 120 min for the normal etch and 330 min for the angled etch. This angled undercut released arbitrarily oriented structures from the bulk material.

After release, the authors removed the mask with a dry SF6, Ar, and O2 ICP-RIE sequence. SF6 selectively removed titanium, Ar removed chromium, and the final bias-free O2 step polished and oxygen-terminated the diamond surface. Avoiding wet processing after release mitigates meniscus-force damage to the suspended waveguides. The freestanding geometry also retains thermal contact with the bulk diamond.

SEM analysis of the nominal 200 nm group measured a 221(9) nm separation and a 287(9) nm width. This geometry did not match the 250 nm separation and 300 nm width assumed for the 620 nm simulation. Most geometry uncertainty came from limited edge definition and charging in SEM images.

The inner coupling-gap sidewall angle was not measured directly. Mutual shadowing during the angled etch may have made these faces steeper than the 42(2)° outer angle. Simulations show that nanometer-scale width or separation changes and degree-scale angle changes can strongly alter coupling. A one-to-one assignment of measured coupling to one simulated geometry therefore remains difficult.

Supports outside the coupling region hold the structure in place. Supports within that region also stabilize waveguide separation, but add optical loss. The 637 nm loss analysis gave 3.7(3) dB residual device loss, 0.6(1) dB per support, and 12.9(2) dB combined TIR input and output coupling loss. Calibration waveguides found the lowest loss at a 1.2 µm TIR width, corresponding to 5.5 dB per coupler and 28% transmission under an equal-efficiency assumption.

Before fabrication, the authors implanted electronic-grade diamond with 120Sn ions at 60 keV. The dose was 2 × 10^10 ions/mm², with a depth distribution centered at approximately 50 nm. A 2 h anneal at approximately 1500°C formed SnV− centers before nanofabrication. The detailed optical measurements used a separate, nominally identical implanted and fabricated sample.

Measurements occurred at 4 K. Resonant excitation came from above, while phonon-sideband fluorescence was collected through the tapered fiber. A charge-resonant check with a 70 kcts/s threshold controlled charge-state changes between PLE lines. The PSB is centered near 650 nm, while the SnV− ZPL is near 620 nm. The effective splitting ratio during PSB collection is therefore expected to differ from the ZPL ratio.

QuantityResultMeasurement condition
Cumulative PLE FWHM34.8(4) MHzCRC between lines at 3 nW laser power before the objective
Mean individual PLE FWHM28.1(4) MHzLorentzian fits to individual lines
Optical lifetime6.4(5) ns2 ns Gaussian resonant excitation pulse
Lifetime-limited linewidth25(2) MHzDerived from the measured T1
Optical Rabi period1.2(2) ns20 ns rectangular pulse at 6 µW before the objective
Rabi damping time4.5(9) nsExponentially damped sine fit

The cumulative PLE linewidth reflects spectral diffusion and residual laser drift. The individual linewidth of 28.1(4) MHz is close to the 25(2) MHz lifetime limit. The 1.2(2) ns Rabi period is comparable to values reported for bulk SnV− samples. No post-fabrication anneal was used.

Device-to-device variation in evanescent coupling remains the main limitation. Reproducibility requires better control of the angled etch, mask-to-device transfer, and support geometry. The transmission result was obtained at 637 nm, so it does not directly establish a fixed 50:50 ratio at the 620 nm ZPL.

The authors propose measuring g2(τ) between the two output arms as the next verification. This would test whether photons from one emitter follow the intended paths. Electrodes for Stark or strain tuning would allow separate SnV− centers to be spectrally aligned. Hong-Ou-Mandel interference would then test whether routed photons are indistinguishable.

The 4.5(9) ns Rabi damping time was shorter than the 6.4(5) ns optical lifetime. The authors suggest that spectral diffusion and drive-induced detuning caused the difference. The present data therefore establish coherent emitter coupling, but not single-photon interference between the arms.

Material selection and device requirements

Section titled “Material selection and device requirements”

Freestanding structures made from bulk diamond avoid membrane transfer and interfaces between diamond emitters and foreign photonic materials. The paper identifies those interfaces as sources of alignment difficulty, integration complexity, and cryogenic strain. Angled etching is also largely independent of the crystal axis, which permits arbitrary in-plane device orientation.

For material selection, defect quality alone does not control coupler performance. The process must also control sub-micrometer separation, sidewall angle, mask transfer, and support placement. The measured sensitivity shows that a nominally identical layout can produce a broad coupling distribution after fabrication. Better support designs and dry-release control are needed before a bank of deterministic 50:50 couplers can be made.

The paper requires bulk single-crystal diamond stock for freestanding SnV− waveguides and their quantum-photonic chip. The diamond materials catalog lists MPCVD-grown single-crystal plates and wafers up to 125 mm, with single-crystal thicknesses from 0.1 to 500 micrometers and substrates up to 10 mm. For process-specific layout and fabrication planning, 6CCVD provides engineering support.

View Original Abstract

Abstract Freestanding all-diamond color-center photonics is a promising platform for optical integration of spin-based quantum defects. Within this geometry, we realize a key building block for quantum-network interconnects: a directional coupler that acts as an on-chip beam splitter. We design and simulate directional couplers with triangular cross sections using eigenmode and finite-difference time-domain simulations and target near-50:50 splitting at visible wavelengths. We fabricate the devices directly from bulk single-crystal diamond by angled oxygen reactive-ion-beam etching followed by a dry postrelease hard-mask removal process. Room-temperature measurements at λ0 ≈ 637 nm yield a mean coupling ratio of Cmeas = 46(16) %. Finally, we integrate tin-vacancy (SnV-) centers into the nanophotonic structures and observe near-lifetime-limited optical line widths and coherent optical Rabi oscillations without postfabrication annealing, identifying the platform as a viable route toward integrated diamond quantum photonics.