Hetero-integrated diamond-on-AlGaN/AlN waveguides for optical color center interfacing
Paper details
Section titled “Paper details”| Field | Value |
|---|---|
| Journal | npj Nanophotonics |
| Published | 2026-02-02 |
| Authors | Sinan Gündoğdu, Lea M. Rektorschek, Marco E. Stucki, Maarten H. van der Hoeven, Tim Kolbe |
| DOI | 10.1038/s44310-025-00099-w |
A finite element model predicted 97.3% for ideal single-interface coupling with vertical AlGaN/AlN walls. Devices A and B had inferred single-interface efficiencies of 58(6)% and 53(5)%. Their measured total efficiencies were 34(7)% and 29(6)% and closely matched simulations that included placement errors and the 450 nm etch depth. The fabrication-constrained design predicted 71% for one interface.
Complementary functions in one interface
Section titled “Complementary functions in one interface”Diamond color centers give quantum photonic circuits optically controlled spin registers, quantum memories, and single-photon sources. AlGaN/AlN supports the second-order nonlinear and first-order electro-optic effects used in frequency converters and phase shifters. Diamond supports neither effect because its crystal lattice has inversion symmetry. Combining the materials assigns spin access to diamond and electro-optic functions to the AlGaN/AlN waveguide.
Earlier evanescent fiber coupling was efficient and supported entanglement distribution across a telecom fiber network. However, the number of fibers attachable to one diamond chip limited that interface’s scalability. Prior integration work placed GaP components on diamond and diamond chiplets on silicon nitride or aluminum nitride.
AlGaN/AlN was selected for epitaxial control of composition, thickness, dispersion, refractive-index contrast, and nonlinear response. The authors did not claim it was superior to lithium niobate on insulator or silicon carbide. The paper gives no numerical efficiency comparison with those earlier integration approaches. The authors also cite difficult diamond nanofabrication and limited wafer-scale availability of device-grade single-crystal diamond.
Finite element interface design
Section titled “Finite element interface design”Finite element simulations used JCMsuite to calculate the fundamental TE mode in diamond and inject it into port 1. They recorded the overlap with the fundamental TE mode of the AlGaN/AlN waveguide at port 2. Port 2 mode overlap defined the coupling efficiency. The mode moved from diamond at port 1 to AlGaN/AlN at port 2 through evanescent transfer.
At the diamond-side section, about 83% of the fundamental TE mode power lay inside diamond. At the AlGaN/AlN section, about 98% lay within the AlGaN/AlN layers, with residual power in air. The material table listed n = 2.41 for diamond and n = 1.76 for sapphire. AlN had in-plane indices of 2.07 and an out-of-plane index of 2.11; Al0.7Ga0.3N had in-plane indices of 2.14 and an out-of-plane index of 2.20. The AlGaN and AlN layer indices were measured at 632 nm.
A Bayesian optimizer maximized the simulated coupling efficiency within each design’s fixed and variable geometry constraints.
| Design | Geometry condition | Single-interface efficiency | Result type |
|---|---|---|---|
| Design 1 | Wall and tip angles of 0° | 97.3% | Simulated |
| Design 2 | 15° wall angle, 70° tip angle, 500 nm etch limit | 71% | Simulated |
| Design 3 | 458 nm diamond width, with diamond constraints removed | 92% | Simulated |
Design 1 represents the ideal vertical sidewalls that the current etch did not produce. Design 2 imposed measured fabrication angles and the 500 nm etch limit. The dominant loss mechanism was the 70° tip angle, which scattered the mode toward the sapphire substrate. Effective-index mismatch between diamond and AlGaN/AlN modes caused a further reduction.
Design 3 removed constraints on the diamond waveguide and recovered 92% at a width of 458 nm. A ±30 nm sweep of diamond width in Design 1 produced a mean coupling efficiency of 95.6%. The uncertainty analysis found diamond width had the most significant impact. The authors expect better control of diamond width and AlGaN/AlN wall angles to yield the largest improvement.
Layer growth and waveguide fabrication
Section titled “Layer growth and waveguide fabrication”AlGaN/AlN fabrication began with a 400 nm AlN layer deposited by magnetron sputtering and metal-organic vapor phase epitaxy. A 250 nm Al0.7Ga0.3N waveguide layer was then grown by metal-organic vapor phase epitaxy. The pattern used 200 nm SiO2 hard mask, 10 nm Ti charge layer, and 400 nm ZEP 520A resist. The layer etch used BCl3:Cl2:He inductively coupled reactive-ion plasma.
The fabricated etch depth was 450 nm, with 15° walls and a 70° tip along the taper. Diamond waveguides were released through an isotropic undercut with 30 nm Al2O3 sidewall protection. The diamond process used 200 nm SiNx, 10 nm Ti, and 400 nm ZEP 520A resist layers. Electron-beam lithography defined the pattern, and anisotropic oxygen plasma formed the diamond sidewalls. An isotropic plasma etch at high temperature and pressure undercut the structure before HF removed the hard mask.
Pick-and-place integration
Section titled “Pick-and-place integration”A three-axis piezo micromanipulator with a tungsten tip picked up each diamond waveguide. The protruding diamond tip was aligned to two markers on the AlGaN/AlN surface before lowering. Electrostatic forces aided initial attachment, and van der Waals forces supported the final adhesion. Each diamond waveguide was placed across an interrupted AlGaN/AlN guide, creating two tapered interfaces.
Optical excitation and coupling measurement
Section titled “Optical excitation and coupling measurement”End-fire coupling focused a 635 nm laser into the AlGaN/AlN input facet and collected light through the same objective. This wavelength was chosen to resemble the zero-phonon-line emission of diamond NV− centers. A CMOS camera monitored the output facet, and an output region of interest was integrated for power readings. The simulated coupling efficiency of the empty configuration was 5(1)% and served as the reference. Placed-waveguide output was normalized against the empty reference, and total efficiency included both interface crossings. Single-interface efficiency was then inferred under the assumption that the two interfaces were symmetric.
Under separate end-fire excitation, a 532 nm laser and 650 nm longpass filter revealed red fluorescence from NV− centers. That image confirmed that excitation light crossed both diamond-AlGaN/AlN interfaces and reached the color centers.
Device A had a lateral offset of 1.4(5) μm and an angular misalignment of 0.0(5)°. Its measured total efficiency was 34(7)%, corresponding to a single-interface efficiency of 58(6)%. Device B had a lateral offset of 0.2(5) μm and an angular misalignment of 1.5(5)°. Its measured total efficiency was 29(6)%, corresponding to a single-interface efficiency of 53(5)%.
The model used the measured 450 nm etch depth and swept both placement deviations for each device. Predicted total efficiencies closely agreed with the measurements. This agreement showed that SEM-derived placement data and the model could predict fabricated-taper performance under nonideal configurations.
Material and process implications
Section titled “Material and process implications”The measured values show that placement errors and taper geometry reduced performance below the nominal Design 2 prediction. Device B had the smaller lateral offset but the larger angular misalignment and lower total efficiency. This comparison shows that angular misalignment cannot be ignored even when the lateral offset is smaller.
For diamond selection, the final waveguide width is a key criterion because it must remain reproducible after patterning. The ±30 nm sensitivity of the ideal design makes width control a material and process requirement. The diamond must retain the region used for NV− excitation while supporting the undercut release used to form a suspended guide.
For AlGaN/AlN, vertical sidewalls and a small tip angle are needed to suppress scattering toward sapphire. Reducing wall and tip angles through improved etching should raise the feasible single-interface efficiency. However, the paper presents that improvement as a fabrication target, not a measured result. The AlGaN/AlN platform also supports frequency conversion and phase shifting, although this experiment did not test those functions.
Remaining development needs
Section titled “Remaining development needs”Current process issues constrain further device development. The AlGaN/AlN etch is not anisotropic enough to produce the vertical walls assumed by Design 1. The tip is identified as the primary reason Design 2 falls from 97.3% to 71% in simulation.
Alignment remains difficult because the diamond tips used for placement could not be identified clearly in optical images. Matching alignment markers on both materials are planned to improve placement precision. Repositioning a bonded diamond waveguide risks damage, so controlled alignment is needed before contact.
The authors state that further work is required before the platform can implement quantum network nodes. The reported experiment establishes a working optical interface and model agreement, not a complete network node.
Materials and support
Section titled “Materials and support”6CCVD offers single-crystal and polycrystalline diamond grown by MPCVD through the diamond materials catalog. The paper needs a released diamond waveguide chiplet carrying NV− centers, with precise width control and support for isotropic undercut patterning. For geometry review and custom fabrication support, use engineering support.
View Original Abstract
Abstract Quantum photonic devices require efficient access to solid-state spin qubits serving as quantum memories or single-photon sources and nonlinearities. Additionally, they demand nonlinear and electro-optic components for frequency converters and phase shifters. The serial production of quantum devices needs the on-chip integration of all these photonic components. These requirements motivate the hetero-integration of materials with complementary properties. Diamond color centers provide access to spin registers, long coherence times, and suitability for quantum networking. In contrast, aluminum gallium nitride-on-aluminum nitride (AlGaN/AlN) photonics features second-order nonlinear and first-order electro-optic effects. Here, we demonstrate the accurate integration of diamond waveguides containing nitrogen-vacancy (NV) centers into AlGaN/AlN waveguides using a pick-and-place method. Both waveguides possess tapered sections for adiabatic mode transfer. We model the tapered diamond-AlGaN/AlN interface with finite element simulations, implementing fabrication constraints and placement uncertainties. Measured coupling efficiencies coincide closely with simulation model predictions.
Electronic Grade Intrinsic Diamond Epitaxial Wafer H Terminated Superior 100 Mosfet Ready Substrate
View Specs →
High Concentration Quantum Grade Single Crystal Diamond
View Specs →
Diamond Epitaxial Wafer For Diodes 2
View Specs →References we cover
Section titled “References we cover”- Realization of a multinode quantum network of remote solid-state qubits
- Experimental demonstration of memory-enhanced quantum communication
- Efficient Extraction of Zero-Phonon-Line Photons from Single Nitrogen-Vacancy Centers in an Integrated GaP-on-Diamond Platform