Programmable quantum emitter formation in silicon

Silicon (Si)-based quantum emitters are emerging as viable candidates for quantum computing, sensing, networking, and communication owing to bright telecom-band emission, scalability, and ease of integration with electronics and photonics. Prominent Si color centers (W, G, T) involve common elements (H, C) and are typically formed via ion implantation and thermal annealing; additional local activation has been demonstrated using focused ion beams and laser pulses. Laser processing in Si enables local annealing, doping, and defect engineering. Here, the authors address the challenge of selective, programmable, single-center formation in Si by combining forming gas annealing (hydrogen present) with femtosecond (fs) laser processing to write and erase specific emitters. They focus on the lesser-studied C1 center (a split Si–C interstitial pair) alongside G-centers, showing hydrogen-enabled selection of C1 and T centers and passivation of G centers, followed by fs-laser-enabled local programming of center populations. The work aims to enable scalable integration of selected quantum emitters and spin–photon interfaces in Si.

Prior studies established formation of Si color centers (e.g., G, W, T) via ion implantation and rapid thermal annealing in inert ambients, often yielding dominant G-center emission. Local laser-driven formation of color centers has been optimized in wide-bandgap semiconductors (diamond, SiC, hBN), including deterministic single-center formation with in situ feedback. Laser processing in Si has been used for annealing, doping, and defect engineering, and telecom-band single-photon emitters (G, T, W) have been demonstrated in SOI. Electron paramagnetic resonance (Watkins and Brower) revealed optically addressable spin-1/2 charge states of the C1 center, indicating potential as a spin–photon interface, yet its formation pathways and level structure have been comparatively underexplored. Recent works also examine kinetics of carbon pair formation (hindering thermal equilibrium), strain/linewidth effects in SOI, and hydrogen passivation of defects in semiconductors. This study builds on these findings by leveraging forming gas annealing to passivate G centers and enhance C1/T formation, and by using fs pulses for selective, local programming of emitters.

Sample preparation: Commercial SOI wafers (220 nm Si device layer, 2 µm buried SiO2, 10–20 Ω·cm, p-type) were used. Carbon implantation employed 38 keV 13C ions targeting ~115 nm depth (mid-device layer). A typical dose for Fig. 2a was 7×10^13 C/cm^2. Rapid thermal annealing (RTA) was performed at 800 °C for 120 s in forming gas (10% H2, 90% N2). Forming gas RTA repairs implant damage, passivates G centers, and favors bright C1 (and T) formation.

Fs-laser processing: An amplified Ti:sapphire laser (800 nm center wavelength, 90 fs pulse duration FWHM, 250 kHz system operated in single-shot mode) was used. A 20×20 µm^2 spot (measured at sample) irradiated samples with single pulses of 65–195 nJ, corresponding to fluences ~16–48 mJ/cm^2 (well below Si damage threshold >100 mJ/cm^2). Additional experiments explored higher fluences up to ~300 mJ/cm^2 (W-center activation) and ~1000–6000 mJ/cm^2 (above melting threshold). By varying fluence, centers were written (e.g., G formation near ~16 mJ/cm^2), passivated (~30 mJ/cm^2), erased/re-activated (~40–45 mJ/cm^2), and isolated single centers formed at ~8–12 mJ/cm^2.

Optical characterization: Photoluminescence (PL) and time-resolved PL (TR-PL) were measured at 6 K using a scanning confocal microscope. Excitation used a 532 nm CW and pulsed laser through an NA=0.85 objective; emission collected to a spectrometer with an InGaAs camera (900–1620 nm, -80 °C) and a 150 g/mm grating (typical) or 600 g/mm (0.05 nm resolution) for single-emitter spectra. Telecom window (1200–1600 nm) was scanned; a 1250 nm long-pass filter was used to suppress background. Second-order autocorrelation g²(τ) was measured with an HBT setup at 160 µW excitation. Polarization-resolved PL and power dependence were recorded. TR-PL lifetimes were extracted via first-order decay fits.

Theory (DFT): First-principles calculations used VASP on 3×3×3 Si supercells. HSE06 functional for electronic structure (450 eV cutoff, energy convergence 1e-10 eV, force tolerance 0.001 eV/Å), Γ-point sampling. Excited-state energies via constrained occupation method; formation energies with PBE plus finite-size corrections (Spinney; Kumagai–Oba corrections). Real-space wavefunctions via VASPKIT. Considered C1 center (split (Si–C)Si) in multiple charge states and structural variants, including a displaced “B” configuration and H-decorated forms (C1+H Types 1–3 with H on C or Si). Transition dipole moments (TDMs) and zero-phonon lines (ZPLs) were computed to assess optical brightness and compare with experiments.

• Forming gas (10% H2 in N2) RTA of 13C-implanted SOI selectively forms bright C1 (and T) centers while passivating G centers, contrasting with inert-ambient anneals that yield dominant G emission.
• C1 centers exhibit narrow emission linewidths in SOI of ~0.03 nm (~4.2 GHz), limited by spectrometer resolution; hydrogen incorporation during anneal likely mitigates interface-induced strain/linewidth broadening via H clustering and impurity trapping.
• Fs-laser single-pulse irradiation (800 nm, 90 fs) below Si damage threshold enables local, programmable writing and erasing of emitters: writing of G centers and modulation of C1 density at ~16 mJ/cm^2; partial/complete passivation near ~30 mJ/cm^2; re-appearance of G and C1 near ~44.5 mJ/cm^2. The experimental Si damage threshold was >100 mJ/cm^2.
• Isolated single C1 and G emitters are obtained at very low fluences (~8–12 mJ/cm^2). Single-photon emission is confirmed with background-corrected g²(0) well below 0.5 (HBT measurement at 160 µW excitation). Single-center PL shows polarization sensitivity and longer lifetimes than ensemble background.
• TR-PL reveals C1 lifetimes ~3 ns pre-irradiation, varying between ~3–8 ns post-irradiation depending on fluence; G centers show similar ns lifetimes. Small fluence changes modulate lifetimes, enabling quality tuning of emitters.
• Mechanism: For fluences 16–48 mJ/cm^2, Keldysh parameter ~19–628 (photoionization regime). Single-photon absorption (1.55 eV) generates hot carriers (>0.43 eV above band edge) that couple to phonons, exciting Si–H (~0.26 eV) and C–H (~0.36 eV) vibrational modes, driving H migration to/from defects and reconfiguring C1+H and G-center configurations, enabling writing/erasing without lattice damage.
• High fluence above melting (~1000–6000 mJ/cm^2) suppresses C1 and yields PL dominated by G and W centers; this regime is unfavorable for reliable single-emitter formation.
• DFT identifies stable C1 charge states (0, −1, −2, −3; −1 stable in intrinsic Si) and shows bare C1 has very small TDM due to mirror symmetries, implying optical darkness. A displaced “B” configuration (0.66 eV higher energy) and H-decorated variants (C1+H Types 1–3) dramatically increase brightness (TDMs up to order-unity Debye^2), with ZPL shifts comparable to experimentally observed side peaks around the main ~1448 nm (856 meV) line. Hydrogen thus enhances C1 brightness by orders of magnitude and accounts for multiple observed ZPLs via charge/structure variations.
• Forming gas passivates G centers: calculated TDMs for H-trapped G-center B configurations (0.067–0.0845 Debye^2) are much smaller than native G (~5 Debye^2), explaining suppressed G PL after forming gas RTA.
• Practical outcome: fs-laser fluence provides fine control over emitter presence, density, and lifetime, enabling programmable integration of selected quantum emitters in Si.

The study demonstrates that incorporating hydrogen during annealing and subsequently using sub-damage-threshold fs-laser pulses allows deterministic control over the creation and suppression of specific Si color centers, directly addressing the challenge of selective, programmable single-emitter formation. Hydrogen plays a dual role: it passivates G centers (reducing their optical activity) and dramatically enhances C1 brightness when bonded to the defect, as corroborated by DFT. Fs pulses operate primarily by modulating H bonding and migration within defect complexes via photoionization-driven hot-carrier–phonon coupling, effecting reversible transformations between optically active and inactive configurations. The ability to write and erase at will, along with formation of isolated single C1 and G emitters (confirmed by g²(0)<0.5) and tunable lifetimes, positions this approach for scalable quantum photonics in Si, including telecom-band single-photon sources and potential spin–photon interfaces. The observed ZPL multiplicity near 1448 nm is consistent with DFT-predicted charge/structure/H-related variants, supporting the proposed mechanisms. Avoiding the high-fluence melting regime is critical for single-emitter reliability, reinforcing the importance of the identified low-fluence operational window.

By combining 13C implantation, forming gas RTA, and localized fs-laser pulses, the authors realize selective writing and erasing of Si quantum emitters, notably enabling bright, narrow-linewidth C1 centers in SOI while passivating G centers. They demonstrate single-emitter formation at low fluence with verified single-photon emission and tunable lifetimes. First-principles calculations reveal how hydrogen incorporation elevates C1 optical activity by orders of magnitude and explain observed spectral features through charge and structural variations. This programmable, hydrogen-mediated laser control provides a pathway to scalable integration of telecom-band quantum emitters and potential spin–photon interfaces in silicon. Future work includes in situ PL feedback for deterministic arrays, detailed mapping of the C1+H level structure and spin properties, and nonadiabatic simulations (e.g., TD-DFT) to fully resolve fs-laser-driven defect dynamics.

• Mechanistic understanding of fs-laser-driven writing/erasing is incomplete; more advanced nonadiabatic simulations (e.g., TD-DFT) are needed to capture excited-state defect dynamics and H migration pathways.
• DFT ZPL predictions for C1 variants exhibit significant systematic errors due to finite supercell size and constrained occupation; absolute energies deviate by several tenths of an eV, though relative shifts are informative.
• SOI interfaces can broaden emitter linewidths via strain/stress; while forming gas may mitigate this, interface effects persist and vary spatially.
• High-fluence regimes (near/above melting) are unsuitable for reliable single-emitter formation; process windows must be tightly controlled.
• The full level structure and spin properties of H-decorated C1 variants remain to be experimentally mapped and validated.