Colour Centre Formation in Silicon-On-Insulator for On-Chip Photonic Integration

TL;DR

This work addresses the challenge of scalable on-chip quantum photonics using silicon colour centres by systematically mapping how implantation and annealing, together with nanofabrication steps, govern colour-centre formation in silicon-on-insulator. The authors employ carbon and hydrogen ion implantation followed by activation annealing across a broad temperature/time window and examine integration with photonic nanostructures, using SIMS and cryogenic photoluminescence to track defect formation. Key contributions include identifying an optimal T-centre activation temperature of $525 \\pm 10$°C, revealing coupled formation dynamics among W, G, I, C, and M centres, and discovering a stable CN-like centre (CN*) near $1496.7$ nm with potential spin-photon functionality. The findings provide actionable guidelines for fabricating emitter-rich SOI photonic circuits and introduce CN as a promising new colour centre candidate for quantum technologies.

Abstract

Colour centres in silicon have great potential as single photon sources for quantum technologies. Some of them - like the T centre - also possess optically-active spins that enable spin-photon interfaces for generating entangled photons and multi-spin registers. This paper explores the generation of several types of colour centres in silicon for mass-manufacturable silicon-on-insulator quantum devices. We investigate how different processes in the device development affect the presence of the quantum emitters, including thermal annealing and fabrication steps for optical nanostructures. The study reveals coupled formation dynamics between different colour centres, identifies optimal parameters for annealing processes, and reports on the sensitivity to annealing duration and nanofabrication procedures for photonic integrated circuits. Furthermore, we discern stable optical signals from colour centres in silicon which have not been identified before.

Figures (20)

• Figure 1: Characteristic PL spectra of prominent colour centres demonstrating their presence in the 1200 to 1600 nm range after annealing at different temperatures. The atomic structures for the defects associated with each peak are also shown in the insets filippatos_re-examination_2025, except for the M centre, for which the atomic structure is still undetermined. For each defect, we show the PL spectra with the highest photoluminescence signal amongst those obtained with the different annealing parameters we tested. The associated annealing temperature is reported below each of the peaks.
• Figure 2: Photoluminescence signal from several optically active colour centres as a function of the activation annealing temperature. The intensity of the signal is extracted as the area of a Gaussian on a second order polynomial background fitted to a window around the peak. Error bars denote $\pm 1\sigma$ statistical uncertainty.
• Figure 3: Spectra with an unidentified photoluminesce peak centred around 1496.7 nm, which we call CN* due to association with the centre proposed in Ref. nangoi2025cncomplexalternativet. The signal arises only after annealing at 540, but increases in brightness up to 570. The signal from the material response has been subtracted from the signal for ease of visualisation.
• Figure 4: Photoluminescence signal from different Colour centres for samples annealed at 525, shown for a variable duration of the activation anneal. Error bars denote $\pm 1\sigma$ statistical uncertainty.
• Figure 5: Effects of ashing on colour centre formation. (a) Reference samples only undergoing the activation anneal similar to samples presented in Figure \ref{['fig:Annealed_spectra']}. (b) The effect of ashing on annealed centres indicating a reduction in the T, I, and M centre densities but an increase for the G and C centre. (c) The effect of ashing before annealing, showing similar densities as the reference samples. (d) The effect of ashing with a remote plasma. Despite the longer ashing period, the effect on emitter density is negligible, meaning this presents an alternative to direct ashing in cases where ashing must be done after activating colour centres but where it is undesirable to harm existing centres. Error bars denote $\pm 1\sigma$ statistical uncertainty.