Erbium-Doped Silicon Waveguides

• Erbium-doped silicon waveguides are photonic structures incorporating Er³⁺ ions to provide telecom-band emissions and enable quantum photonic applications.
• Key fabrication techniques include ion implantation, epitaxial growth, and sensitization with silicon nanograins to optimize light–matter coupling and reduce losses.
• Devices exhibit tailored optical gain and narrow linewidths, promising scalable platforms for quantum memories, on-chip lasers, and integrated photonic networks.

Erbium-doped silicon waveguides are photonic structures in which Er³⁺ ions are introduced into silicon-based guided-optical circuits to exploit their sharp intra-4f transitions around 1.54 μm (the telecom C-band). These platforms combine the mature nanofabrication and scalability of silicon photonics with the long optical and spin coherence of rare-earth ions, targeting applications in quantum memories, on-chip optical amplifiers, frequency conversion, and integrated quantum photonic networks. The principal approaches involve ion implantation of erbium into silicon or epitaxial integration of crystalline erbium-doped films, with careful waveguide and host engineering to control inhomogeneous broadening, coherence, light–matter coupling, amplification, and device scalability.

1. Waveguide Architectures and Fabrication Strategies

Erbium-doped silicon waveguides have been realized using multiple architectures, each with distinct fabrication requirements, doping strategies, and modal properties:

• Ion-implanted silicon-on-insulator (SOI) waveguides: Standard 220–500 nm thick Si layers on ~2 μm buried oxide are lithographically patterned and subsequently implanted with Er (or ^170Er) in commercial foundry processes. Typical geometries: widths 500–700 nm, heights 220 nm; supports single TE-like modes (n_Si ≈ 3.48, n_SiO2 ≈ 1.44). Waveguides are fabricated in spiral or straight layouts up to 10 mm, with fiber or on-chip reflectors for spectroscopy. Post-implant annealing at 400–800 K is employed to activate Er and reduce crystal damage (Rinner et al., 2023, Gritsch et al., 2021, Weiss et al., 2020, Holzäpfel et al., 2024).
• Crystalline erbium-doped oxide films on Si: Epitaxy of single-crystal Er:Gd₂O₃ (doped up to 0.5 at.% Er, ~130 nm thick) directly on Si(111) by MBE, with Si₃N₄ top-strips for optical confinement, yields a cubic lattice-matched gain medium. Modal confinement factors in Er:Gd₂O₃ reach Γ ≃ 35%, and propagation losses < 3.5 dB/cm have been realized (Xu et al., 28 Nov 2025).
• Dielectric/oxide host waveguides with silicon nanograin sensitization: Si-rich silica matrices loaded with Si nanograins (Si-ng, ~10¹⁹ cm⁻³) and Er³⁺ (~10²⁰ cm⁻³), supporting energy transfer from Si-ng to Er³⁺ under visible pumping, realized as planar strip-loaded structures with typical layer stacks (Si substrate | SiO₂ 3.5 μm | Si-rich SiO₂ 2 μm | SiO₂ 0.4 μm) (Fafin et al., 2014, Cardin et al., 2015).

2. Structural and Spectroscopic Characterization of Er³⁺ in Silicon Hosts

The optical and spin properties of Er³⁺ in silicon waveguides are governed by crystal-field environment, implantation/annealing conditions, and host purity:

• Crystal field and site identification: Spectroscopic studies resolve distinct substitutional lattice sites (A: C₂ᵥ symmetry and B: Cₛ symmetry) with characteristic zero-phonon lines, Zeeman splitting patterns, and branching ratios. For site A (C₂ᵥ): g-tensor principal values g_g = (8.5, 8.5, 0.58); g_e = (6.94, 6.94, 0.24). The site symmetry constrains sub-site multiplicity (A: 12, B: 24 sub-sites) and informs integration yield and homogeneity (Holzäpfel et al., 2024).
• Optical transition properties: In commercial SOI waveguides, inhomogeneous broadening (Δν_inh) is typically 1–2 GHz (site A: 3.5±0.2 GHz, site B: 1.9±0.1 GHz), with homogeneous linewidths (Γ_h) < 30 kHz at high magnetic fields (B > 4 T, T < 2 K) and upper bounds as low as 9–10 kHz in isotopically pure, low-disorder hosts. Lifetimes (τ) for site A/B in high-quality silicon range from 142 to 186 μs, corresponding to lifetime-limited Γ₁ ≈ 0.8–1.1 kHz (Rinner et al., 2023, Gritsch et al., 2021).
• Spin structure and Zeeman regime: The Er³⁺ I(15/2) ground and I(13/2) excited multiplets split according to anisotropic g-tensors, yielding spin-conserving transitions (Δν_sp ~ |g_g – g_e|μ_B B/h ≈ 1–3 GHz/T) and spin-flip transitions (Δν_sf ~ |g_g + g_e|μ_B B/h ≈ 18–21 GHz/T) along specific crystallographic orientations (Rinner et al., 2023, Holzäpfel et al., 2024).

3. Optical Gain, Amplification, and Emission Dynamics

Several operational regimes have been investigated, depending on the host, device length, and excitation protocol:

• Ion-implanted Si/SOI devices: At doping densities ~10¹⁷ cm⁻³, Er-doped Si waveguides show single-pass absorptions α_abs ≈ 10⁻³ cm⁻¹ (σ_0 ~ 10⁻²¹ m²); overall optical depth for 1 cm length is OD < 0.01 (double-pass with reflector OD ≲ 0.02). Propagation losses remain low (<1 dB/cm) after optimization (Rinner et al., 2023). Amplification is not observed; Purcell enhancement with high-Q cavities or slot-waveguides is required for useful gain (Gritsch et al., 2021, Weiss et al., 2020).
• Crystalline Er:Gd₂O₃/Si waveguides: Demonstrated material gains reach 78.3±2.1 dB/cm (peak, at 2.3 K), net on-chip gain >13 dB in 6-mm devices, and continuous-wave lasing with threshold power P_th ≈ 8.5 mW and side-mode suppression ratio 36.5 dB. Room-temperature net modal gain of 1.06±0.77 dB/cm is possible (Xu et al., 28 Nov 2025).
• SRSO/Si-ng sensitized devices: Despite efficient Si-ng to Er³⁺ transfer (K ~ 1×10⁻¹⁴ cm³/s), maximum gross gain at 1532 nm is 2 dB/cm (at 10⁴ mW/mm² pump), insufficient to exceed typical background losses (α_bg ≈ 3 dB/cm); thus net gain is not realized. The three-level Er³⁺ energy scheme sets a high inversion threshold (Fafin et al., 2014, Cardin et al., 2015).

4. Light–Matter Coupling, Mode Engineering, and Quantum Memory Metrics

Optimization of interaction strengths involves both photonic and atomic engineering:

• Mode overlap and field enhancement: For uniform Er doping in Si waveguides, the normalized overlap η between the guided optical mode and Er ions is ~10⁻⁴–10⁻³. Slot-waveguides or photonic-crystal cavities can enhance η and Purcell factors (F_P) by 10–10³×; state-of-the-art cavities on Si can reach F_P >200, with projections up to 10⁶ (Rinner et al., 2023, Gritsch et al., 2021).
• Ensemble cooperativity and optical depth: For collective coupling, C = 4g²/(κγ) with single-emitter coupling g and decay γ = 1/τ. Optical depth OD ≪ 1 in straight guides; microcavity or slow-light geometries are required for functional ensemble storage. Storage protocols such as atomic frequency comb (AFC) and electromagnetically induced transparency (EIT) can operate over GHz bandwidth set by Δν_inh (Weiss et al., 2020, Rinner et al., 2023).
• Propagation losses and integration: Implanted devices can achieve <1 dB/cm added loss with appropriate annealing. High-Q factor devices can be fabricated on-wafer using commercial foundry flows, enabling co-integration with modulators and detectors (Rinner et al., 2023).

5. Spin Coherence, Magnetic Field Control, and Branching Ratios

• Spin coherence and population relaxation: Spin T₁ > 100 ms (bulk analogs), with T₂ expected >10 ms at high fields. Freezing of paramagnetic impurities above B ≳ 4 T (μ_B B ≫ k_B T, T < 2 K) suppresses dephasing; measured optical (T₂) up to 35–50 μs (Rinner et al., 2023, Gritsch et al., 2021, Holzäpfel et al., 2024).
• Zeeman splitting and selection-rule engineering: Full spin Hamiltonians, with C₂ᵥ and Cₛ g-tensor symmetry, permit orientation-dependent engineering of transition frequencies, branching, and polarization selection rules. Optimizing B-field orientation can minimize (clock transitions) or maximize (conversion protocols) spin-flip admixture, with spin-flip optical branching ratios as low as <5% (Holzäpfel et al., 2024).

6. Quantum Photonic and Amplification Applications

The measured and engineered properties of Er-doped silicon waveguides enable several emergent functionalities:

• Quantum memories: Sub-10 kHz homogeneous linewidth, up to GHz bandwidth from low Δν_inh, and T₁-limited decoherence at cryogenic temperatures are suitable for AFC- or EIT-based memories for the telecom band, scalable on chip. Hyperfine shelving (e.g., ^167Er in ²⁸Si) may yield storage times > 1 s (Rinner et al., 2023, Gritsch et al., 2021).
• On-chip lasers and amplifiers: Epitaxial Er:Gd₂O₃/Si platforms demonstrate CW lasing, net gain, and device miniaturization compatible with silicon photonics foundries, overcoming the limitations of amorphous hosts (Xu et al., 28 Nov 2025).
• Microwave-to-optical photon converters: Er³⁺'s large g-factor anisotropy and cubic-site engineering enable strong coupling to both microwave and optical fields, supporting hybrid transduction interfaces for superconducting circuits (Weiss et al., 2020, Gritsch et al., 2021).
• Scalable photonic integration: Compatibility with foundry-standard SOI processes and wafer-scale fabrication allows parallel device arrays, dense photonic routing, and integration with modulators, detectors, and coupling optics—all leveraging mature CMOS technology (Rinner et al., 2023, Gritsch et al., 2021).

7. Limitations, Future Directions, and Optimization Strategies

Major challenges include:

• Amplification limits in Si host: Population inversion thresholds and unavoidable reabsorption in three-level Er³⁺ systems inhibit net gain in SOI and Si-rich SiO₂ unless both doping and background loss are improved or alternative pumping/sensitization schemes are developed (Fafin et al., 2014, Cardin et al., 2015).
• Inhomogeneous broadening and site yield: Achieving <1 GHz Δν_inh and high-fidelity site occupancy requires isotopically pure Si (²⁸Si), defect suppression, and precise control of implantation/annealing; only ≳1% of Er occupies optimal (A/B) sites under typical protocols (Gritsch et al., 2021, Holzäpfel et al., 2024).
• Surface and charge-induced dephasing: In nanostructures, proximity to etched surfaces or local charge noise can broaden linewidths by orders of magnitude; device passivation and surface control remain central research topics (Holzäpfel et al., 2024).
• Materials innovation: Epitaxial integration of single-crystal Er-doped oxides represents a new pathway, enabling high gain and low threshold devices. A plausible implication is the extension of these methods to other rare-earth ions and wide-bandgap hosts (Xu et al., 28 Nov 2025).

References:

• (Rinner et al., 2023) "Erbium emitters in commercially fabricated nanophotonic silicon waveguides"
• (Gritsch et al., 2021) "Narrow optical transitions in erbium-implanted silicon waveguides"
• (Xu et al., 28 Nov 2025) "High-gain optical amplification and lasing from erbium-doped single-crystal films epitaxially grown on silicon"
• (Holzäpfel et al., 2024) "Characterization of the spin and crystal field Hamiltonian of erbium dopants in silicon"
• (Weiss et al., 2020) "Erbium dopants in silicon nanophotonic waveguides"
• (Fafin et al., 2014) "Theoretical investigation of the more suitable rare earth to achieve high gain in waveguide based on silica containing silicon nanograins doped with either Nd3+ or Er3+ ions"
• (Cardin et al., 2015) "Modeling of optical amplifier waveguide based on silicon nanostructures and rare earth ions doped silica matrix gain media by a finite-difference time-domain method: comparison of achievable gain with Er3+ or Nd3+ ions dopants"