Surface passivation for narrowing optical linewidth of silicon T centers in nanophotonic devices

Silicon T centers are promising spin–photon interfaces in solid–state platforms for telecom–compatible, scalable quantum information technologies. A major challenge for T centers in nanophotonics is spectral diffusion, where fluctuations in the local electric–field environment from surface and bulk charge states broaden the optical transition and reduce photon indistinguishability. Strategies that directly suppress spectral diffusion are therefore critical for improving T-center-based quantum photonic devices. Here, we use atomic–layer–deposited Al2O3 to passivate the silicon surface and demonstrate a systematic narrowing of T center optical linewidths. Across our measurements, Al2O3 passivation reduces the T center emission linewidth by up to 57%. Complementary above-bandgap illumination and spectral hole burning measurements show that the remaining linewidth contains a significant spectral-diffusion component caused by adjacent charge traps, while placing an upper bound of approximately 75 MHz on the homogeneous linewidth. This work provides a CMOS–compatible path toward generating indistinguishable photons from silicon T centers for scalable quantum photonic applications.

In the introduction, the paper notes that "Optical linewidth broadening of the T centers poses a major challenge for generating indistinguishable photons. The Fourier-transform-limited linewidth of the silicon T center optical transition is approximately 170 kHz and homogeneous broadening has been reported to be 67 MHz in nanofabricated SOI. In practice, however, T centers integrated in nanophotonic devices exhibit far broader optical linewidths of 1–5 GHz, indicating that spectral diffusion, rather than homogeneous broadening, is the dominant mechanism. The paper explains that Spectral diffusion occurs when slow fluctuations in the local electric-field environment shift the optical transition frequency over time. Such fluctuations can arise from surface charge states including dangling bonds and dopant acceptors, or charge traps in the silicon crystal and are particularly severe in T centers integrated with nanophotonics where the emitters are located close to surfaces. The authors state that Surface passivation has improved the optical performance of several solid-state emitter platforms, including diamond color centers and semiconductor quantum dots, by reducing surface-induced charge noise and enhancing linewidth, spectral stability, or quantum efficiency. However, its impact on silicon T centers remains unexplored."

In the results section, the paper reports: "Using atomic-layer-deposited Al2O3 as a passivation layer, we observe linewidth narrowing in both same-emitter comparisons and statistical measurements across multiple emitters. In the same-emitter measurements, Al2O3 passivation reduces the optical linewidth by 32% on average, with a maximum reduction of 57%. Specifically, For this emitter, the linewidth decreased from 2.41 GHz to 1.04 GHz after passivation, corresponding to a 57% reduction. The paper notes that All four emitters showed linewidth narrowing after passivation, with an average reduction of 32%. The authors attribute the variation to differences in proximity to the surface and in the relative weight of surface charge noise versus charge fluctuations in the bulk silicon."

The paper also investigated the optimal Al2O3 thickness: We studied Al2O3 films of 7, 11, 15, 20, and 26 nm, measuring the optical linewidths of ten emitters at each thickness. The results show that the linewidth reduction saturates near 11 nm, suggesting that additional Al2O3 thickness does not further suppress spectral diffusion from surface charge states. However, the lifetime decreases with increasing Al2O3 thickness, which the authors attribute to strain-related enhancement of non-radiative decay, likely through activation or modification of local defect and trap states. This is supported by a concurrent redshift of the T center emission wavelength with increasing Al2O3 thickness, indicating that thicker Al2O3 films impose greater strain on the silicon nanostructure. Thus, these results identify 11 nm as the optimal Al2O3 thickness in our devices that provides near-saturated linewidth improvement while minimizing strain-induced non-radiative decay.

The paper further investigated residual broadening mechanisms. Under above-bandgap illumination with a 980 nm laser, the linewidth decreased by roughly a factor of two, reaching a minimum of 0.3–0.4 GHz, while applying an additional detuned O-band laser produced no systematic narrowing over the measured power range. The authors interpret this as photogenerated carriers filling nearby charge traps and partially stabilizing the local electric-field environment. Spectral hole burning measurements on a T-center nanobeam passivated with a 15-nm-thick Al2O3 layer gave a hole full width at half maximum (FWHM) of 150.8 ± 22.0 MHz. Since the spectral-hole FWHM in the low-power limit is twice the homogeneous linewidth, this places an upper bound of approximately 75 MHz on the homogeneous linewidth. The paper concludes that the substantially larger PLE linewidth therefore indicates that slow spectral diffusion remains the primary contribution to the residual broadening.

In the discussion, the authors state: Our results show that Al2O3 surface passivation suppresses a substantial component of the spectral diffusion of T centers in nanophotonic devices, but does not eliminate the residual charge-induced broadening. They suggest that further improvement will require optimization of the passivation interface rather than simply increasing the coating thickness, and propose a surface treatment using HF:HCl solution to preterminate the silicon dangling bonds, and post deposition annealing for optimum activation of the surface passivation layer. They also note that the additional narrowing observed under above-bandgap illumination suggests that charge traps not fully stabilized by the Al2O3 treatment also contribute to the remaining spectral diffusion. Complementary approaches such as reducing fabrication-induced surface damage through atomic layer etching or electrically stabilizing the local charge environment using diode structures could be integrated with optimized surface passivation.

In conclusion, the paper states: "We demonstrate that Al2O3 surface passivation provides an effective route to reduce optical linewidth broadening in nanophotonically integrated silicon T centers. By comparing individual emitters before and after passivation, we recorded an average linewidth narrowing of 32%, with actual reductions reaching up to 57%. Systematic thickness variations reveal an optimal passivation layer of 11 nm, beyond which strain-induced non-radiative decay begins to negatively impact the excited-state lifetime without offering further linewidth reduction. Above-bandgap illumination further narrows the PLE linewidth to approximately 400 MHz, while spectral hole burning bounds the homogeneous linewidth at approximately 75 MHz, showing that the slow spectral diffusion remains after passivation. Our results establish Al2O3 surface passivation as a practical, CMOS-compatible strategy for mitigating spectral diffusion in silicon T centers."

Improvements to AI Systems:

1. Predictive Optimization of Surface Passivation Parameters: Train an AI model on the dataset (Al2O3 thickness vs. linewidth, lifetime, and emission redshift) to predict the optimal passivation thickness for arbitrary nanophotonic geometries, accounting for trade-offs between spectral diffusion suppression and strain-induced non-radiative decay. The improved AI can automatically design device-specific passivation recipes (e.g., thickness, annealing conditions) to minimize linewidth while preserving quantum efficiency.

2. Autonomous Identification and Classification of Spectral Diffusion Sources: Develop an AI system that analyzes time-resolved photoluminescence excitation (PLE) spectra and spectral hole burning data to distinguish between surface charge noise, bulk charge traps, and homogeneous broadening contributions. The improved AI can classify the dominant broadening mechanism in real-time, enabling dynamic feedback control (e.g., adjusting illumination power or gate voltages) to stabilize the emitter environment.

3. Active Charge-Trap Mitigation via Reinforcement Learning: Implement a reinforcement learning agent that controls above-bandgap illumination intensity and detuned laser parameters to minimize T-center linewidth in situ. The AI learns the optimal illumination schedule and power to fill charge traps without inducing additional heating or photo-ionization, achieving linewidths closer to the 75 MHz homogeneous bound.

4. Transferable Surface Chemistry Optimization: Use a generative AI model to propose novel surface termination chemistries (e.g., HF:HCl pre-treatment, post-deposition annealing profiles) based on the demonstrated Al2O3 passivation results. The improved AI can simulate and rank candidate passivation protocols for silicon T centers, accelerating discovery of even more effective coatings beyond Al2O3.

5. Predictive Strain-Aware Device Layout Design: Train a neural network on the observed strain-induced redshift and lifetime degradation data to predict how Al2O3 thickness and nanostructure geometry affect emitter performance. The improved AI can co-optimize photonic crystal design and passivation layer thickness to minimize strain while maximizing linewidth narrowing, enabling scalable fabrication of indistinguishable-photon sources.

6. Real-Time Spectral Diffusion Forecasting: Build a time-series AI model that predicts spectral diffusion dynamics from initial PLE linewidth measurements and environmental conditions (e.g., temperature, illumination history). The improved AI can forecast when an emitter will drift out of resonance and proactively trigger corrective actions (e.g., voltage pulses or laser repumping), enhancing photon indistinguishability for quantum repeater applications.

7. Automated Cross-Platform Benchmarking: Develop an AI that aggregates data from multiple solid-state emitter platforms (diamond, quantum dots, T centers) to identify universal passivation strategies. The improved AI can transfer insights from Al2O3 on silicon to other materials, suggesting optimal coatings for emerging quantum photonic devices.

Silicon T centers are promising spin-photon interfaces in solid-state platforms for telecom-compatible, scalable quantum information technologies. A major challenge for T centers in nanophotonics is spectral diffusion, where fluctuations in the local electric-field environment from surface and bulk charge states broaden the optical transition and reduce photon indistinguishability. Strategies that directly suppress spectral diffusion are therefore critical for improving T-center-based quantum photonic devices. Here, we use atomic-layer-deposited Al2O3 to passivate the silicon surface and demonstrate a systematic narrowing of T center optical linewidths. Across our measurements, Al2O3 passivation reduces the T center emission linewidth by up to 57%. Complementary above-bandgap illumination and spectral hole burning measurements show that the remaining linewidth contains a significant spectral-diffusion component caused by adjacent charge traps, while placing an upper bound of approximately 75 MHz on the homogeneous linewidth. This work provides a CMOS-compatible path toward generating indistinguishable photons from silicon T centers for scalable quantum photonic applications.