Wafer-Scale Photonics Delivers Broadband Trapped-Ion Addressing

Scaling trapped-ion processors is partly an optical packaging problem. A useful device must deliver many wavelengths, hit individual ions a few micrometers apart, survive ultra-high-vacuum assembly, and avoid placing large dielectric surfaces where they destabilize the trap. This paper attacks that bottleneck with a two- and three-dimensional photonic integrated circuit: lithographic waveguides and planar lenses route light in the chip plane, while two-photon-polymerized micromirrors redirect and focus it onto ions above a surface trap.

Core Contribution

The central result is not just another on-chip beam coupler. The authors combine a broadband aluminum-oxide waveguide platform with a 3D-printed, aluminum-coated reflector, then integrate the resulting optics with surface-trap electrodes on 6-inch wafers. That combination gives the circuit a much wider useful spectral range than narrowband grating couplers while keeping the beam-delivery geometry compatible with ion-addressing pitches.

Figure 1 shows the architectural choice clearly: the planar lens expands and collimates light from strip waveguides, and the printed micromirror maps different incoming waveguides to different focal spots at the ion plane.

The novelty is the packaging of these pieces into a monolithic trapped-ion device. Prior photonic ion traps have shown waveguides and grating couplers, but the bandwidth problem is severe because ion species need multiple wavelengths for cooling, repumping, state preparation, and qubit operations. Here, the same optical architecture is characterized from 405 nm to 880 nm, covering the relevant bands for calcium and barium demonstrations.

Technical Approach

The guided photonics use a 120 nm amorphous $\mathrm{Al_2O_3}$ core inside $5\,\mu\mathrm{m}$ of $\mathrm{SiO_2}$ cladding. The methods section reports band-specific waveguide widths for single-mode operation across the 375–866 nm design window: 0.55 µm in near ultraviolet, 0.8 µm in visible, and 1.6 µm in near infrared. Low-loss Euler bends were designed for each band, with simulated bend loss below 0.01 dB per 90° turn.

The free-space transform is split across two optical elements. A planar waveguide lens was optimized in 3D FDTD around 729 nm and scaled for other wavelengths; the micromirror was optimized in Zemax as a biconic-aspherical reflector with a roughly $75\,\mu\mathrm{m} \times 2\,\mu\mathrm{m}$ aperture. The mirror surface is printed by two-photon polymerization, then metallized and electrically tied into the trap electrode stack.

Fabrication is also part of the claim. The devices are made on 6-inch silicon wafers, with 50 micromirrors printed per wafer, followed by post-processing that deposits 200 nm of aluminum for both electrodes and reflectors. The reported substrate-to-mirror alignment accuracy is within 1 µm. Packaged devices survived sequential vacuum bakes to 50, 80, 100, and 125 °C with negligible coupling loss, and the ion experiments ran below $1\times10^{-10}$ Torr.

Results and Analysis

The optical characterization supports the broadband claim. Figure 3 reports focal-plane images at a 75 µm ion height for six wavelengths from 405 to 880 nm, plus simultaneous three-spot illumination at 515 nm and 635 nm with a 5 µm pitch. The key addressing number is -27.6 dB average intensity crosstalk between on-axis and 5 µm-offset focal spots. For trapped-ion control, that is a meaningful result because it links the photonic geometry to an ion-scale spacing rather than only reporting free-space beam quality.

The ion experiments make the device more convincing than a passive optics demonstration. The authors trap both $^{138}\mathrm{Ba}^+$ and $^{40}\mathrm{Ca}^+$ above the integrated structures. With barium, they infer 455 nm integrated-beam crosstalk from excitation rates of the $S_{1/2}\rightarrow P_{3/2}$ transition as the ion is moved near the dielectric cutout. With calcium, they demonstrate selective 866 nm repumping of one ion in a two-ion pair spaced by 5 µm. That is a limited operation, not a full gate demonstration, but it directly tests the intended use case: wavelength-specific light delivered to one ion without equally bright illumination of its neighbor.

Figure 4 captures this distinction between optical addressing and ion-level behavior: the same paper reports trapped Ba and Ca ions, measured 455 nm crosstalk, and selective 866 nm repumping in a two-ion Ca pair.

The main caveat is electrostatic, not photonic. The supplementary dielectric study shows that exposed or charged dielectric regions can create large spatially varying fields near the trap. The authors measure axial stray fields reaching tens of kV/m near the dielectric boundary and report axial secular-frequency changes of more than a factor of three over the same spatial range, corresponding to a factor-of-nine change in axial curvature. Their simulation indicates that even a 0.8 V effective dielectric potential can shift axial frequencies by several megahertz. That limits how freely such reflective dielectric structures can be placed near shuttling paths or interaction zones.

The paper therefore provides strong device-level evidence for broadband integrated addressing, but not yet a complete processor-scale solution. The optical architecture solves a real bandwidth and routing problem; the dielectric charging data show why future versions will need active charge control, shielding, or more conservative placement of exposed optical structures.