T Center in Si: Telecom Spin–Photon Interface

• T center in Si is a telecom-band color center formed by a hydrogen–carbon defect, providing a native spin–photon interface in crystalline silicon.
• It exhibits a paramagnetic S=1/2 ground state with long spin lifetimes and narrow optical linewidths, critical for quantum networking applications.
• Advanced fabrication methods, including ion implantation and ultra-low-temperature epitaxy, enable scalable integration of T centers in silicon devices.

The T center in silicon is a telecom-band color center in crystalline Si whose principal optical transition lies near $1326\,\mathrm{nm}$ ($\sim 935\,\mathrm{meV}$) and whose spin-selective optical structure makes it a silicon-native spin–photon interface for quantum networking, nanophotonics, and integrated quantum devices. Across the literature represented here, it is consistently treated as a carbon–hydrogen-related point defect, but its microscopic assignment is not uniform: several works describe a $(\mathrm{C}-\mathrm{C}-\mathrm{H})_{\mathrm{Si}}$ or $\mathrm{C}_2\mathrm{H}$ complex, whereas one recent epitaxial-growth study explicitly refers to “hydrogen-(H)$_2$C point defects in Si”; all nevertheless identify the same telecom optical signature and quantum-device relevance (Bergeron et al., 2020, Dhaliah et al., 2022, Marböck et al., 8 Apr 2026).

1. Microscopic identity and defect models

The dominant microscopic model in the T-center literature is a carbon–hydrogen complex occupying a silicon lattice site. In one widely used description, two carbon atoms share a silicon substitutional site and a hydrogen atom is bonded to one of the carbons; one carbon is then trigonal and carries the unpaired electron in a dangling bond. In a separate recent epitaxial study, the center is instead treated as a hydrogen–carbon complex denoted H$_2$C. The exact atomistic configuration is therefore not presented uniformly across the current literature, and this remains one of the few persistent points of nomenclature and structural interpretation (Brunelle et al., 18 Dec 2025, Day et al., 21 Jan 2025, Marböck et al., 8 Apr 2026).

First-principles calculations give a more specific electronic picture for the $(\mathrm{C}-\mathrm{C}-\mathrm{H})_{\mathrm{Si}}$ model. In that description, the neutral defect is stable over almost the entire Si band gap, with the $(0/-1)$ charge transition level at $E_{\mathrm{VBM}} + 1.07\,\mathrm{eV}$. The ground state contains an unpaired electron in a localized defect state of strong carbon $p$ character, and the excited optical state is a defect-bound exciton formed by a localized defect electron and a weakly bound hole (Dhaliah et al., 2022).

Historically, the T center belongs to the family of radiation-damage-related luminescence centers in Si, but its present importance is not merely spectroscopic classification. It is the combination of a telecom optical transition, a paramagnetic ground state, and compatibility with isotopically purified silicon and silicon photonics that distinguishes it from other Si color centers discussed in the same materials platform (Bergeron et al., 2020, Bergeron et al., 2020).

2. Electronic, spin, and hyperfine structure

The operative optical transition is between a paramagnetic ground state and a bound-exciton state. In the ground state, the T center hosts an electron spin $\sim 935\,\mathrm{meV}$0; in the bound-exciton state TX$\sim 935\,\mathrm{meV}$1, the two electrons form a singlet and the remaining unpaired degree of freedom is a hole. Earlier $\sim 935\,\mathrm{meV}$2Si measurements established that the T center provides electron and nuclear spin lifetimes beyond a millisecond and second respectively, with a TX$\sim 935\,\mathrm{meV}$3 optical lifetime of $\sim 935\,\mathrm{meV}$4 and a Debye–Waller factor of $\sim 935\,\mathrm{meV}$5 (Bergeron et al., 2020).

In the hyperfine-resolved description, the most common isotopic variant is a four-dimensional ground-state manifold consisting of an electron spin $\sim 935\,\mathrm{meV}$6 and a hydrogen nuclear spin $\sim 935\,\mathrm{meV}$7. The ground-state Hamiltonian is written as

$\sim 935\,\mathrm{meV}$8

with anisotropic hyperfine interaction

$\sim 935\,\mathrm{meV}$9

For the $(\mathrm{C}-\mathrm{C}-\mathrm{H})_{\mathrm{Si}}$0 orientation, the experimentally determined principal values are

$(\mathrm{C}-\mathrm{C}-\mathrm{H})_{\mathrm{Si}}$1

with Euler angles $(\mathrm{C}-\mathrm{C}-\mathrm{H})_{\mathrm{Si}}$2. At zero field, the four eigenstates yield three transitions at $(\mathrm{C}-\mathrm{C}-\mathrm{H})_{\mathrm{Si}}$3, $(\mathrm{C}-\mathrm{C}-\mathrm{H})_{\mathrm{Si}}$4, and $(\mathrm{C}-\mathrm{C}-\mathrm{H})_{\mathrm{Si}}$5, with average splitting $(\mathrm{C}-\mathrm{C}-\mathrm{H})_{\mathrm{Si}}$6 (Brunelle et al., 18 Dec 2025).

Magneto-optical studies in $(\mathrm{C}-\mathrm{C}-\mathrm{H})_{\mathrm{Si}}$7Si also resolve the orientational structure imposed by the defect’s low symmetry. Under magnetic field, the TX$(\mathrm{C}-\mathrm{C}-\mathrm{H})_{\mathrm{Si}}$8 zero-phonon line splits into multiple orientational subsets; eleven optical subsets were directly resolved, and magnetic resonance established the expected total of twelve. This orientational multiplicity is one reason the T center is simultaneously rich and experimentally demanding as a spin–photon interface (Bergeron et al., 2020).

3. Formation pathways in bulk Si and silicon-on-insulator

The earliest modern route to T-center creation combined high-energy electron irradiation with hydrogen incorporation and annealing in isotopically enriched bulk $(\mathrm{C}-\mathrm{C}-\mathrm{H})_{\mathrm{Si}}$9Si. That route was sufficient to establish narrow zero-phonon lines, long spin lifetimes, and the defect’s bound-exciton spectrum in bulk material (Bergeron et al., 2020, Bergeron et al., 2020).

For device-compatible silicon-on-insulator, the dominant conventional route has been carbon and hydrogen implantation followed by annealing. One single-emitter nanobeam study used $\mathrm{C}_2\mathrm{H}$0C implantation at $\mathrm{C}_2\mathrm{H}$1 with fluence $\mathrm{C}_2\mathrm{H}$2, rapid thermal annealing at $\mathrm{C}_2\mathrm{H}$3 for $\mathrm{C}_2\mathrm{H}$4, hydrogen implantation at $\mathrm{C}_2\mathrm{H}$5 with the same fluence, boiling in deionized water for 1 hour, and a second anneal at $\mathrm{C}_2\mathrm{H}$6 for 3 min in $\mathrm{C}_2\mathrm{H}$7. That process created T centers at a depth of $\mathrm{C}_2\mathrm{H}$8 within the $\mathrm{C}_2\mathrm{H}$9 SOI device layer (Lee et al., 2023).

A major recent shift is the use of epitaxial formation instead of implantation. Ultra-low-temperature molecular beam epitaxy can self-assemble T centers during kinetically limited growth, avoiding vertical ion straggle and collateral lattice damage. In one ULT-MBE study, the active layer for T-center formation was a $_2$0 Si layer grown at $_2$1, with the centers confined to that $_2$2 region; a reference structure grown at $_2$3 in deep ultra-high vacuum showed no T-center emission, indicating that the defect did not form under those cleaner, higher-temperature conditions. The same study showed that growth chemistry matters strongly: for two samples at similar total pressure but different partial pressures of C, CH$_2$4, and CO$_2$5, the T-center ZPL intensity differed by a factor of 23, and the ensemble ZPL full width at half maximum narrowed from $_2$6 in HV growth to $_2$7 in D-UHV growth (Marböck et al., 8 Apr 2026).

An SOI-compatible epitaxial route has now been demonstrated for single emitters. In that approach, the top Si device layer was thinned to $_2$8, followed by $_2$9 Si buffer growth at $_2$0, a $_2$1 Si:C layer grown at $_2$2, a $_2$3 Si cap at $_2$4, and an ex-situ anneal at $_2$5 for 3 minutes in $_2$6. This yielded single, waveguide-coupled epitaxial T centers in SOI (Christiansen et al., 7 Jul 2026).

4. Optical spectroscopy, coherence, and spectral diffusion

Representative reported values span bulk $_2$7Si, implanted SOI devices, and epitaxial SOI. In $_2$8Si, the TX$_2$9 ZPL energy at $(\mathrm{C}-\mathrm{C}-\mathrm{H})_{\mathrm{Si}}$0 is $(\mathrm{C}-\mathrm{C}-\mathrm{H})_{\mathrm{Si}}$1, the ensemble linewidth can be as narrow as $(\mathrm{C}-\mathrm{C}-\mathrm{H})_{\mathrm{Si}}$2, and the Debye–Waller factor is $(\mathrm{C}-\mathrm{C}-\mathrm{H})_{\mathrm{Si}}$3. Bulk $(\mathrm{C}-\mathrm{C}-\mathrm{H})_{\mathrm{Si}}$4Si hole-burning measurements later pushed the instantaneous homogeneous linewidth to $(\mathrm{C}-\mathrm{C}-\mathrm{H})_{\mathrm{Si}}$5. In epitaxial SOI, spectral hole burning yielded $(\mathrm{C}-\mathrm{C}-\mathrm{H})_{\mathrm{Si}}$6 for a single T center, compared with $(\mathrm{C}-\mathrm{C}-\mathrm{H})_{\mathrm{Si}}$7 for an implanted SOI reference under comparable conditions. A single nanobeam-coupled implanted T center showed a ZPL at $(\mathrm{C}-\mathrm{C}-\mathrm{H})_{\mathrm{Si}}$8, linewidth $(\mathrm{C}-\mathrm{C}-\mathrm{H})_{\mathrm{Si}}$9, and excited-state lifetime $(0/-1)$0 (Bergeron et al., 2020, Bergeron et al., 2020, DeAbreu et al., 2022, Lee et al., 2023, Christiansen et al., 7 Jul 2026).

Quantity	Reported value	Context
TX$(0/-1)$1 ZPL energy	$(0/-1)$2	$(0/-1)$3Si, $(0/-1)$4
Debye–Waller factor	$(0/-1)$5	ensemble $(0/-1)$6Si
Homogeneous linewidth	$(0/-1)$7; $(0/-1)$8	bulk $(0/-1)$9Si; epitaxial SOI
Excited-state lifetime	$E_{\mathrm{VBM}} + 1.07\,\mathrm{eV}$0; $E_{\mathrm{VBM}} + 1.07\,\mathrm{eV}$1	ensemble $E_{\mathrm{VBM}} + 1.07\,\mathrm{eV}$2Si; single nanobeam emitter

The optical spectrum is not limited only by intrinsic radiative physics. In nanophotonic devices, spectral diffusion is a central issue. A recent cavity study found no spectral broadening on short timescales from $E_{\mathrm{VBM}} + 1.07\,\mathrm{eV}$3 to $E_{\mathrm{VBM}} + 1.07\,\mathrm{eV}$4. Using a herald-and-probe protocol, the optical resonance frequency was stable up to $E_{\mathrm{VBM}} + 1.07\,\mathrm{eV}$5 in the dark, whereas laser pulses below the silicon band gap applied during the wait time induced linewidth broadening. That work identified laser-induced processes as the dominant spectral-diffusion mechanism for T centers in devices (Zhang et al., 11 Apr 2025).

Epitaxial SOI changes the balance between broadening mechanisms rather than eliminating all of them. In one comparison, the inhomogeneous linewidths of epitaxial and implanted SOI ensembles were similar, $E_{\mathrm{VBM}} + 1.07\,\mathrm{eV}$6 and $E_{\mathrm{VBM}} + 1.07\,\mathrm{eV}$7, suggesting a common strain contribution from the SOI geometry and buried oxide. By contrast, the homogeneous linewidth improved by about an order of magnitude in the epitaxial sample. A plausible implication is that epitaxy primarily removes bulk-defect-related fast noise, while slow spectral diffusion remains tied to surfaces and interfaces (Christiansen et al., 7 Jul 2026).

5. Nanophotonic, cavity, and electrical integration

Waveguide integration on SOI established that T-center ensembles can be addressed inside single-mode silicon waveguides while retaining long spin relaxation. In one such study, the waveguide-integrated electron-spin relaxation time satisfied $E_{\mathrm{VBM}} + 1.07\,\mathrm{eV}$8, and the measured optical linewidths were already low enough that remote spin-entangling protocols were projected to require only modest cavity Purcell enhancement. The same work reported nearly lifetime-limited homogeneous linewidths in isotopically pure bulk crystals, strengthening the case for $E_{\mathrm{VBM}} + 1.07\,\mathrm{eV}$9Si-on-insulator as a future materials target (DeAbreu et al., 2022).

At the single-emitter level, a tapered silicon nanobeam with a photonic crystal mirror enabled efficient fiber coupling. Simulations predicted $p$0 coupling to the nanobeam mode and $p$1 mode transfer into a lensed fiber; experimentally, the nanobeam-to-fiber coupling efficiency was $p$2, the background-corrected autocorrelation reached $p$3 at low power, and the corrected ZPL single-photon count rate reached $p$4 (Lee et al., 2023).

Cavity coupling addresses the T center’s intrinsically long lifetime and modest ZPL branching ratio. One photonic-crystal-cavity implementation reported a fluorescence decay-rate enhancement factor $p$5 and an average photon outcoupling rate of $p$6 at the zero-phonon line. Another nanophotonic cavity study reported a two-orders-of-magnitude increase in ZPL brightness relative to waveguide-coupled emitters, $p$7 collection efficiency from emitter to fiber, overall ZPL emission efficiency of $p$8, lifetime enhancement of 5, and a Purcell factor exceeding 18 after correcting for emission into the phonon sideband (Johnston et al., 2023, Islam et al., 2023).

The platform has also advanced from single cavities to multiplexed architectures. In arrays of bus-coupled silicon photonic crystal cavities, two T centers in separate cavities were operated in a frequency-multiplexed fashion, with $p$9 and $\sim 935\,\mathrm{meV}$00. In the same platform, cavity-enhanced lifetimes down to $\sim 935\,\mathrm{meV}$01 and Purcell factors around 61 were reported for the strongest emitters, and hybridized modes between physically distant cavities were used to modify a T center’s radiative decay (Komza et al., 28 Jan 2025).

Electrical integration is newer but no longer speculative. A buried lateral P-I-N diode containing a T-center ensemble showed the center’s optical response under static and dynamic electric fields, including electroluminescence and the use of T-center emission to probe negative differential resistance. In that device, reverse bias up to high field produced no meaningful changes in ensemble brightness, center wavelength, or linewidth before breakdown-induced heating quenched the optical signal (Day et al., 21 Jan 2025).

6. Role in silicon quantum technology, comparisons, and open issues

Within silicon color-center photonics, the T center is unusual because it combines telecom-band emission with a spinful ground state. The G center is often brighter, but one single-emitter nanobeam study explicitly notes that the G center lacks an unpaired electron and therefore does not form a stable spin qubit. In the broader ULT-MBE color-center study, T centers were described as more “environment-demanding” than W, G, and G′ centers, yet they alone were emphasized there for “favorable spin properties like long coherence times, and optically addressable spin states,” enabling a spin–photon interface (Lee et al., 2023, Marböck et al., 8 Apr 2026).

Two technical issues remain central. The first is microscopic identification: the literature represented here does not yet use a single structural model, with $\sim 935\,\mathrm{meV}$02, $\sim 935\,\mathrm{meV}$03, and H$\sim 935\,\mathrm{meV}$04C-type descriptions all appearing. The second is process sensitivity, especially to hydrogen. First-principles work concluded that the T center is stable against decomposition into simpler defects when keeping stoichiometry fixed, but also that it is “easily prone to (de)hydrogenation” and therefore requires very precise annealing conditions. That specific hydrogen sensitivity has already motivated proposals for hydrogen-free alternatives, including a CN complex as an isoelectronic analogue of the T center (Dhaliah et al., 2022, Nangoi et al., 2 Nov 2025).

The search for alternatives has expanded beyond one substitute defect. High-throughput screening identified a family of group-III–carbon complexes $\sim 935\,\mathrm{meV}$05, with $\sim 935\,\mathrm{meV}$06, as “T center-like” quantum defects in silicon. Those defects were predicted to emit in the telecom and, in some cases, to improve on the T center in computed radiative lifetime or emission efficiency, while also having higher symmetry than the T center and thus being easier to align with magnetic fields. This suggests that T-center research is broadening from the optimization of one defect toward a wider defect-family design strategy for silicon quantum photonics (Xiong et al., 2024).

Despite those open issues, the central trajectory is clear. The T center has progressed from a radiation-damage luminescence center to a quantitatively characterized telecom spin–photon interface, with demonstrated single-photon emission, cavity enhancement, multiplexed cavity arrays, electrical integration, and epitaxial formation in SOI. The outstanding tasks are now concentrated in microscopic unification, interface-noise reduction, and reproducible materials engineering rather than in establishing the defect’s relevance to silicon quantum technology (Christiansen et al., 7 Jul 2026, Zhang et al., 11 Apr 2025, Brunelle et al., 18 Dec 2025).