arXiv:2611.02209
From Fabrication Disorder to Quantum-Protocol Yield in Inverse-Designed Silicon Photonic Beam Splitters
Nail Letty, Tahmeed Ul Islam, Wayesh Qarony
Abstract
Telecom-band silicon spin-photon qubits offer a promising platform for scalable quantum networks and distributed information processing, but their fixed emission wavelengths require compact, low-loss beam splitters with precise splitting and fabrication robustness. Conventional designs can face trade-offs between footprint, optical performance, and wavelength-specific operation. Here, we employ adjoint-based inverse design to optimize photonic geometries beyond conventional layouts, developing 1*2 Y-splitters and 2*2 directional couplers for the silicon T (1326 nm) and Al1 (1482.44 nm) spin-photon qubits. The devices achieve footprints of 2.9 and 26 um^2, respectively, with excess losses of 0.14-0.44 dB and back-reflection below 5*10^-4. We assess fabrication tolerance using 100-realization Monte Carlo simulations incorporating etch bias, line-edge roughness, thickness, and refractive-index variations. Mirror symmetry reduces the splitting-ratio spread of the Y-splitter by a factor of seven to nine relative to the directional coupler. We translate these variations into dual-rail fidelity, ideal splitting-ratio-limited Hong-Ou-Mandel visibility, and random-number min-entropy. All realizations satisfy a protocol-performance threshold of 0.95. At 0.99, the tolerated splitting-ratio deviation shrinks from 0.035 for visibility to 0.0035 for min-entropy, and the simulated yields are 68% (T) and 91% (Al1) for coupler visibility and 90% (T) and 66% (Al1) for Y-splitter min-entropy. The tolerance of the target protocol, rather than the device alone, therefore sets the fabrication requirement for silicon spin-photon beam splitters.
AI Summary
This paper examines how fabrication disorder in compact silicon photonic beam splitters translates into performance and yield for quantum protocols at the fixed emission wavelengths of T ($1326$ nm) and Al1 ($1482.44$ nm) spin–photon qubits. It uses adjoint-based inverse design to develop $1\times2$ Y-splitters and $2\times2$ directional couplers, then maps simulated fabrication-induced splitting-ratio variations to protocol metrics. The reported device footprints are $2.9$ and $26\,\mu\mathrm{m}^2$, respectively, with excess losses of $0.14$–$0.44$ dB and back-reflection below $5\times10^{-4}$. Fabrication tolerance is evaluated using 100-realization Monte Carlo simulations incorporating etch bias, line-edge roughness, thickness variation, and refractive-index variation. Mirror symmetry reduces the Y-splitter splitting-ratio spread by a factor of seven to nine relative to the directional coupler. The protocol analysis considers dual-rail fidelity, ideal splitting-ratio-limited Hong–Ou–Mandel visibility, and random-number min-entropy. All simulated realizations meet a performance threshold of $0.95$. At a threshold of $0.99$, the allowed splitting-ratio deviation is approximately $0.035$ for visibility but $0.0035$ for min-entropy. Simulated coupler visibility yields are $68\%$ for T and $91\%$ for Al1; Y-splitter min-entropy yields are $90\%$ and $66\%$, respectively. The results therefore quantify protocol-dependent fabrication requirements rather than a single device-level tolerance. The evidence described in the abstract is simulation-based, not a fabricated-device demonstration, and the stated Hong–Ou–Mandel metric is ideal and limited to splitting-ratio effects. The full text is unavailable for assessing further assumptions or limitations.
Written by gpt-6.1-sol from the abstract only. Automated: it can be wrong.
- Comments
- 37 pages, 8 figures