Article https://doi.org/10.

1038/s41467-023-37655-x

Individually addressable and spectrally

programmable artificial atoms in silicon
photonics
Received: 11 April 2022 Mihika Prabhu1,4 , Carlos Errando-Herranz 1,2,4 , Lorenzo De Santis 1,3
,
Ian Christen1, Changchen Chen1, Connor Gerlach 1
& Dirk Englund 1
Accepted: 27 March 2023

A central goal for quantum technologies is to develop platforms for precise

Check for updates and scalable control of individually addressable artificial atoms with efficient
1234567890():,;
1234567890():,;

optical interfaces. Color centers in silicon, such as the recently-isolated car-

bon-related G-center, exhibit emission directly into the telecommunications
O-band and can leverage the maturity of silicon-on-insulator photonics. We
demonstrate the generation, individual addressing, and spectral trimming of
G-center artificial atoms in a silicon-on-insulator photonic integrated circuit
platform. Focusing on the neutral charge state emission at 1278 nm, we
observe waveguide-coupled single photon emission with narrow inhomoge-
neous distribution with standard deviation of 1.1 nm, excited state lifetime of
8.3 ± 0.7 ns, and no degradation after over a month of operation. In addition,
we introduce a technique for optical trimming of spectral transitions up to 300
pm (55 GHz) and local deactivation of single artificial atoms. This non-volatile
spectral programming enables alignment of quantum emitters into 25 GHz
telecommunication grid channels. Our demonstration opens the path to
quantum information processing based on implantable artificial atoms in very
large scale integrated photonics.

Artificial atoms in the solid state are leading candidates for spin-based electron and nuclear spin coherence T2 for donors in silicon exceeding
quantum information processing due to their long spin coherence 2 seconds11 and 39 minutes12, respectively.
times and high-quality spin-photon interfaces1–3. However, traditional An artificial atom platform based on the silicon G-center
platforms based on diamond and silicon carbide face two critical could immediately access a vast manufacturing and science toolkit,
challenges for large-scale quantum information processing: lack of including the world’s most advanced complementary metal-oxide-
monolithic manufacturability and inefficient optical interfacing with semiconductor platforms, which already include carbon defects13 and
the optical fiber telecommunication bands. Efforts towards alleviating large-scale patterning with nanometer-scale resolution14. Moreover,
these limitations include hybrid integration2,4,5 and quantum frequency very large scale integrated silicon photonics can integrate millions of
conversion6, although at the cost of greater optical loss (>7 dB at devices onto a single wafer, and already includes nearly all the neces-
present6). Recently, a number of color centers in silicon7–10, such as the sary components for a full-stack quantum photonic system: low-loss
carbon-based G-center8,10, have emerged as promising qubit candi- passive components such as waveguides, splitters, fiber-to-chip
dates, as they can be integrated natively with existing commercial interfaces15, and ultra-high quality factor cavities16; high-speed and
silicon platforms and their telecom wavelength emission obviates the cryogenically-compatible active modulators17 and phase shifters18,19;
need for frequency conversion. Furthermore, the lack of a nuclear spin superconducting single-photon detectors20, and control
bath in isotopically purified 28Si has allowed demonstrations of 21
electronics (Fig. 1a).

1
Massachusetts Institute of Technology, Cambridge, MA, USA. 2University of Münster, Münster, Germany. 3QuTech, Delft University of Technology, Delft, The
Netherlands. 4These authors contributed equally: Mihika Prabhu, Carlos Errando-Herranz. e-mail: [email protected]; [email protected]; [email protected]

Nature Communications | (2023)14:2380 1

Article https://doi.org/10.1038/s41467-023-37655-x

Fig. 1 | G-centers in silicon photonics as a scalable quantum photonic platform. showing light emission in the O-band. d The G-center atomic structure consists of
a The silicon platform features technologically mature opto-electronics, but lacks two substitutional carbon atoms (blue) bonded to an interstitial silicon and e has
integrated quantum emitters and memories. b Schematic showing our system, two optically addressable energy levels and a third metastable level within the
which integrates artificial atoms (G-centers) in silicon photonic waveguides. silicon bandgap. f A finite-difference eigenmode simulation showing the electric
c Measured G-center photoluminescence (PL) spectrum through our waveguide, field norm for our single-mode waveguide geometry.

Although dense ensembles of silicon artificial atoms have been structures. The waveguides in our sample (see Fig. 2b) start and end on
recently integrated into photonic waveguides22,23, the isolation of sin- a cleaved facet (Fig. 2c) and loop in a 63.5 μm radius bend.
gle silicon artificial atoms in photonic integrated circuits (PIC) remains We characterized the samples using low-temperature spectro-
a central challenge. While the integration of quantum emitters in PICs scopy in a cryostat at a base temperature of 6 K. Using above band-gap
in other material systems has already resulted in a plethora of new laser excitation (wavelength λexc = 532 nm, NA = 0.55, and power Pin
physical breakthroughs and applications24, in silicon, a platform that measured before the objective), we induced artificial atoms to emit
combines individually-addressable G-centers with mature integrated into the waveguide. The waveguide-coupled optical emission from the
photonics could open the door to industrial-scale quantum photonics. emitters was subsequently collected with an edge-coupled lensed fiber
Such a technology has the potential to address the scalability chal- at the cleaved chip facet. We band-pass filtered the fiber-collected
lenges of quantum information processing. emission between 1250 nm and 1550 nm to isolate the zero-phonon
Here we report monolithic integration of single artificial atoms— line (ZPL) and phonon sideband of the G-centers while attenuating
G-centers—in silicon PICs (Fig. 1b) and their optically-induced non- residual pump light and other background. The collected emission was
volatile spectral programming. Low temperature spectroscopy reveals then detected with either a cryogenic superconducting-nanowire sin-
single photon emission in the telecommunications O-band (Fig. 1c) gle-photon detector (SNSPD) system or an infrared spectrometer.
with a narrow inhomogeneous distribution of 1.1 nm and spectral shifts
up to 300 pm (55 GHz). Photoluminescence raster scans
We spatially locate our artificial atoms within the waveguides using
Results photoluminescence (PL) raster scans. We scanned the focused
The silicon G-center consists of a pair of substitutional carbon atoms continuous-wave green excitation along the chip plane and detected
bonded to a silicon self-interstitial (Cs-Sii-Cs) within the silicon crystal the filtered emission through the fiber-coupled waveguide with an
lattice (see Fig. 1d). It exhibits a zero phonon line (ZPL) transition at electronically-gated SNSPD. Figure 2d shows PL signal in spatially-
970 meV (1278 nm), resulting from an electron transition between sp- isolated loci along the waveguide, corresponding to single emitters or
like orbitals localized at the Sii atom (Fig. 1e) and features a spin triplet small ensembles of emitters. A range in brightness of the PL hotspots
metastable state25. While the primary mechanism for above-band can be attributed to the presence of clusters of emitters in close
population of the excited state in G-centers is likely to be proximity, which is confirmed through the observation of multiple
Shockley–Read–Hall recombination, additional non-radiative distinct ZPL peaks in the PL spectra taken from these points (Supple-
mechanisms such as surface recombination at electronically active mentary Note 5 and Supplementary Fig. 5b).
surface states and Auger recombination are also known to affect car-
rier dynamics in silicon26 (a more detailed discussion can be found in Spectral characterization of emitters
Supplementary Note 8). Figure 3a shows that the ZPL distribution of 37 waveguide-coupled G-
The device under study consists of a silicon-on-insulator (SOI) centers matches a Gaussian probability distribution with a standard
waveguide designed for single-mode operation at 1278 nm, the ZPL of deviation of σinh = 1.1 nm, nearly an order of magnitude narrower than
the G-center. Simulation results for the electric field distribution within previous reports8. Our narrower inhomogeneous distribution may be
the waveguide are shown in Fig. 1f. We fabricated our samples by a due to strain relaxation induced by our waveguide geometry (see
combination of commercial carbon implantation, thermal annealing, Supplementary Note 7). The fitted mean ZPL wavelength is 1278.7 nm,
and foundry electron beam lithography and etching (see Fig. 2a and in agreement with prior results in bulk SOI8,27. Using the number and
Methods). Our sample contains several waveguides and photonic width of the ZPL peaks in the PL spectrum as an indication of the

Nature Communications | (2023)14:2380 2

Article https://doi.org/10.1038/s41467-023-37655-x

Fig. 2 | Fabrication and imaging of G-centers in silicon photonics. a Fabrication micrograph (SEM) of a cleaved facet of one of our waveguides. d A PL map taken
process for G-centers in silicon photonic waveguides. b Microscope image of our under 10 μW 532 nm green continuous-wave excitation shows discrete light emit-
sample, showing several silicon photonic structures. c Scanning electron ters coupled into the waveguide. e Emitter 0 under study.

number of individual emitters within each excitation region, we iden- tagged SNSPDs (detection efficiencies of 21% and 24% at 1280 nm). We
tified locations within the waveguide that are likely to contain single fitted the measured histogram of coincidences as a function of time
emitters. We isolated one such area, denoted Emitter 0 (E0, shown in delay between the detection events to the second-order autocorrela-
Fig. 2e), and characterized the photophysics of the artificial atom in tion function of a two-level system emitter, displaying an antibunching
this region. The PL spectrum from E0 is shown in Fig. 1e and exhibits a dip of g(2)(0) = 0.38 ± 0.08 (Fig. 3b, see Methods for additional details).
single resolution-limited peak at 1278.473 ± 0.155 nm. To confirm the The observed antibunching dip with g(2)(0) < 0.5 confirms single-
presence of a single artificial atom at E0, we performed second-order photon emission and individual addressing of a single artificial atom
autocorrelation (Fig. 3b) and power-dependent emission intensity coupled to a silicon photonic waveguide.
measurements (Fig. 3c).
Emitter lifetime
Emitter saturation We subsequently measured the excited-state lifetime statistics of
Focusing continuous-wave excitation at 532 nm on E0, we measured emitter E0 and 13 other G-center ensembles (a total of 14 spots) using
the dependence of the emission count rate on the excitation power. In pulsed 532 nm excitation power of 4.4 μW, slightly above the mea-
order to distinguish the contribution to the count rate generated by sured pulsed saturation power for a single emitter. Resulting decay
the emitter from the linear fluorescence observed in the waveguide, we curves fit well to a mono-exponential function, following clipping to
performed background subtraction of the count rate (see Methods remove laser leakage and background (see Methods and Supplemen-
and Supplementary Note 5). The background-corrected emission fits tary Note 6). The lifetime distribution of the 14 measured G-center
well to the characteristic two-level emitter saturation model27, given by ensembles fits well to a Gaussian distribution with mean lifetime
8.33 ns and a standard deviation of 0.68 ns (Fig. 3d). The lifetime of the
P single G-center at E0 was measured to be 8.21 ± 0.14 ns (Fig. 3c), in close
IðPÞ = I 1 , ð1Þ
P + P sat agreement with the mean lifetimes of the measured ensembles.
Additionally, we measured the lifetime of the single emitter E0 to be
with a saturation power P cwsat = 7:6 ± 0:5 μW, and a saturation intensity constant over an order of magnitude variation in excitation power
under continuous wave (CW) laser excitation of I cw 1 = 4753 ± 122 counts (Fig. 3g). Our results agree with previously reported lifetimes in bulk
per second (Fig. 3e). The extracted saturation power corresponds to an SOI G-center ensembles27, but indicate shorter lifetimes compared to
estimated continuous-wave power density of 4.1 kW cm−2, assuming a prior results on isolated G-centers in unpatterned SOI wafers8. Calcu-
diffraction-limited spot size with objective NA = 0.55. We note, lations of the dipole local density of optical states (LDOS) in our
however, that the true excitation area is bounded in one axis by the waveguide were compared to the LDOS of a dipole emitter in a SOI slab
sub-diffraction waveguide width of 400 nm. These results closely (see Supplementary Note 2) and suggest that the reduced lifetime we
match the saturation power observed for single isolated G-centers in observe is not explained by field enhancement in the patterning alone.
unpatterned SOI wafers8. We performed a similar background- Differences in measured lifetimes in our waveguides could be attrib-
subtracted measurement for 532 nm pulsed excitation (see Methods uted to increased surface recombination in etched surfaces (see Sup-
and Supplementary Note 6 for extended details), resulting in a pulsed plementary Note 8). Other effects affecting the measured lifetimes
saturation average power of P pulsed
sat = 3:1 ± 0:4 μW (Fig. 3f). may be doping and defect density variations between wafers. A
quantitative characterization of these effects would include geometry-
Second-order correlation measurements dependent lifetime measurements, which are out of the scope of
Using continuous-wave excitation power of 10 μW (slightly above this study.
saturation), we performed second-order autocorrelation measure-
ments using Hanbury-Brown-Twiss interferometry. We split the Spectral tuning
resulting fiber-collected emission from our sample with a 50:50 Moreover, we demonstrate non-volatile spectral trimming and deac-
O-band fiber beam splitter, followed by detection with two time- tivation of single color centers in PICs. Our spectral programming

Nature Communications | (2023)14:2380 3

Article https://doi.org/10.1038/s41467-023-37655-x

Fig. 3 | Isolation of single emitters and statistics. a ZPL distribution of 37 center ensembles. Saturation curves under e continuous-wave and f pulsed exci-
waveguide-coupled G-centers. b Second-order autocorrelation measurement for E0 tation fit well to a two-level system. Error bars are denoted by the radii of the circle
shows g(2)(0) < 0.5 and single photon emission for a continuous-wave excitation markers. The excitation power used for measurements (b) and (c) are marked with
power of 10 μW, corresponding to an estimated power density of 5.4 kW cm−2. purple and green stars respectively. g The measured lifetime remains constant
Poissonian error bars are included for each sample point, and error in the g(2)(0) is under an order of magnitude variation in excitation power, with error bars deter-
the error of the fit. c Lifetime of emitter E0 and d statistics for 13 G- mined by the error of the extracted lifetime fit.

process consists of in-situ local irradiation of G-centers with a 532 nm emitter, as the signal-to-noise ratio of the emission decreases at low
CW laser with power above 0.1 mW (estimated power density of excitation powers due to the detector dark counts and also at high
54.4 kW cm−2) during a duration of 15 s in our 6 K cryostat, followed by excitation powers due to the background fluorescence (see Supple-
probing PL measurements near emitter saturation powers. Under mentary Note 5). This low signal to noise ratio is the reason behind our
moderate irradiation powers above 0.1 mW, we observe consistent non-zero g(2)(0). However, the coupling efficiency can be easily
non-volatile spectral shifts of the ZPL for 11 out of our 12 probed improved by up to an order of magnitude using widely available
emitters. We observe controllable shifts of average amplitude 150 pm photonic components such as spot-size converters or grating cou-
(27.5 GHz) and up to 300 pm (55 GHz) as shown in Fig. 4a and b and plers, that can achieve coupling efficiencies in excess of 80%28,29.
Supplementary Fig. 10), large enough to match the 25 GHz tele- Another notable difference between our waveguide-integrated G-
communication bands and to enable spectral alignment of separate centers compared to those measured in unpatterned SOI wafers con-
emitters (Fig. 4c). Supplementary Fig. 11a and 11b show the spectral cerns the position of observed emitters selected by the collection
stability of our programmed emitters. Higher powers in the order of mode. Solid-state emitters often show decreased spectral stability in
1 mW (estimated power density of 544.3 kW cm−2) result in broadening proximity to surfaces due to emitter-surface interactions30. Simula-
and deactivation of the emitter, leaving the waveguide and adjacent tions of dipole collection efficiency as a function of dipole depth within
emitters unaffected (Fig. 4d and e and the subtracted PL map in Sup- our waveguides suggest that we preferentially observe emission from
plementary Fig. 12). In our experiments, the programming and deac- G-centers near the center of the silicon film (Supplementary Note 2,
tivation effects are non-reversible. Given that our simulations rule out Supplementary Fig. 2b). On the other hand, prior results using con-
local annealing as a cause for the trimming and deactivation effects focal collection of G-center emission from silicon slabs select for
(see Supplementary Note 10), we hypothesize that these effects are emitters close to the surface8. This geometrical filtering effect may also
caused by optically-induced variation of surface charges leading to explain the absence in our measurements of other emitters previously
Stark tuning of the emitters, followed by ionization into the dark A observed in silicon31, or those created by etching processes. Our
state25 for high charge densities. For more information on the process, measurements show high emission stability down to a 10 ms timescale
results, and our hypotheses on the physics behind this effect, see for a range of excitation powers, as well as robustness to photo-
Supplementary Note 10. bleaching, enabling emission measurements over time intervals
greater than 1 month (see Supplementary Note 9). A comparative
Discussion study looking into G-center emission at varying depths would require
The count rates in our waveguide measurements were limited by probing the emitter-surface interaction for these artificial atoms, but
preventable loss due to imperfect mode matching between our this falls out of the scope of this work.
cleaved waveguide facet and the lensed collection fiber. A mode Our in-situ non-volatile trimming results show a path towards
overlap simulation between the SOI waveguide and fiber mode pre- post-fabrication fine tuning of artificial atoms. Such an effect is large
dicts an upper bound on the coupling efficiency in our system to be enough to tune emitters into standard 25 GHz telecommunication
8.25% (see Supplementary Note 2). In practice, this excess loss restricts bands, to align emitters to cavities and enhance light-matter interac-
our excitation to a small range near the saturation power of the tion, or to spectrally align separate emitters and achieve quantum

Nature Communications | (2023)14:2380 4

Article https://doi.org/10.1038/s41467-023-37655-x

Fig. 4 | Non-volatile trimming of color centers using light. a Example of non- the error of the Lorentzian fit. c Our trimming technique enables non-volatile
volatile spectral shifting of the ZPL of artificial atoms caused by optical irradiation, spectral alignment of separate silicon color centers. PL maps of waveguide sections
with the data represented in blue and the Lorentzian fit in orange. b Fitted ZPL before (d) and after (e) local deactivation of an emitter (marked with a red circle).
central wavelength under increasing optical irradiation power. Error bars denote

interference or cooperative emission. In addition, the local deactiva- In conclusion, we demonstrated individually addressable artificial
tion of emitters presented here can aid in trimming waveguides and atoms operating at telecommunication wavelengths and featuring
cavities of unwanted artificial atoms. Further work is required to reveal narrow inhomogeneous distributions in a foundry-written silicon
the physics behind the observed effect and potentially increase its photonic circuit, as well as a method to spectrally program them and
range. This may involve performing higher resolution spectroscopy, deactivate them using light. Our results show native and spectrally
measuring samples with different geometries and crystal purity, programmable single-photon emission and pave the way towards spin
experimenting with pulsed lasers and other wavelengths, performing qubits in silicon waveguides, establishing silicon photonics as a pro-
in-situ temperature measurements, measuring its dynamics, or mising platform for large-scale quantum information technologies.
searching for correlations between trimming performance and dipole
orientation. Methods
Efficient integration with PICs is a key requirement for any large- Sample description
scale artificial atom qubit platform. Therefore, the results presented The device under study consists of a background p-doped silicon-on-
here characterizing single G-center emission and trimming in silicon- insulator waveguide with 2 μm silicon dioxide bottom cladding on a
on-insulator waveguides provide a key step forward for quantum silicon substrate. The cross-sectional geometry of the waveguide is
information processing based on color centers in silicon. Moreover, shown in Fig. 1f and is rectangular with 400 nm width and 220 nm
our use of a commercial foundry for the fabrication process makes our height. The waveguide starts and ends on a cleaved chip facet and
results scalable and inherently repeatable by the scientific community, loops in a 63.5 μm radius 180∘ bend.
without the need to re-develop an in-house fabrication process. Pro-
mising future research directions include the Purcell-enhanced emis- Sample fabrication
sion from single G-centers in resonant structures aided by localized We generated silicon G-centers using a fabrication process that
implantation32, and the generation of spectrally indistinguishable sin- follows8. The samples started from a commercial (SOITEC) unclad
gle photons for quantum interference using optical trimming, electric silicon-on-insulator wafer (220 nm Si on 2000 nm SiO2). The wafer
fields33, or mechanical strain2. As demonstrated in other color centers1, was then cleaved into 1 cm2 pieces, implanted with 12C with a dose of
coupling emitters into cavities helps overcome low quantum effi- 1 × 1013 ions/cm2 at 36 keV energy and at an angle of 7∘, and then flash
ciencies and Debye-Waller factors. This may not only aid in the annealed for 20 s at 1000 ∘C. The sample was then electron-beam
development of spin-optic interfaces, but also may enable determi- patterned and etched in a foundry (Applied Nanotools). The
nistic single-photon emission and photon-photon interactions via electron-beam patterning was performed using a JEOL JBX8100FS
cavity quantum electrodynamics34. Investigation of the spin properties with 100 kV accelerating voltage and positive-tone resist. The silicon
of the G-center metastable state25, particularly its spin lifetime, and the etching was performed using inductively coupled plasma reactive
13
C and 29Si nuclear degrees of freedom35 could additionally enable ion etching with SF6-C4F8 mixed-gas, in a process optimized for
optically-active quantum memories hosted in silicon photonics. vertical sidewall etching and low propagation loss. This process
Finally, our results also motivate the study and waveguide integration resulted in silicon waveguides with SiO2 bottom cladding and air as
of other radiation damage centers in silicon, such as the recently iso- top cladding. To enable fiber coupling, the sample was cleaved
lated T-center9 and W-center7. across the waveguides.

Nature Communications | (2023)14:2380 5

Article https://doi.org/10.1038/s41467-023-37655-x

Photophysics characterization • ηfilt: Free space fiber filtering with two fiber collimators was
Optical excitation of the G-centers was performed through the cryo- measured to have 51.3% transmission at 1280 nm.
stat window using a long working distance objective with numerical • ηfiber: Fiber routing transmission to the two SNSPD cryostat
aperture of 0.55. Waveguide-coupled emission was collected from a detectors DET1 and DET2 was measured to be 90.6% and 94.8%,
single waveguide output using a SMF-28 lensed fiber and spectrally respectively.
• ηsplitter: Insertion loss of fiber beamsplitter was measured
filtered in free space to select signal between 1250 and 1550 nm. The
to be 92%.
signal was then detected using either time-tagged SNSPDs or an • ηdet : SNSPD detector efficiencies at 1280 nm for DET1 and DET2
infrared spectrometer with a wavelength resolution of 155 pm. We were characterized to be 24% and 21%, respectively. These
measured the excitation powers for all presented measurements SNSPDs were manufactured to have high efficiency at a target
immediately prior to the microscope objective. wavelength of 1550 nm.
PL raster scans were acquired by scanning the focused excitation
spot over a spatial region of the waveguide sample and gating the
SNSPD integration with the electronic trigger of the scanning mirrors. The total system efficiency, given by
Background-corrected PL spectra were then measured using the
infrared spectrometer, where the background spectrum recorded ηtot = ηec ηfilt ηfiber ηsplitter ηdet ð4Þ
environmental light conditions with the excitation laser blocked.
Saturation curves were measured at a spatial location in the can, therefore, be upper bounded to 0.34% for DET1 and 0.31% for
waveguide that exhibited bright PL intensity and a ZPL peak near the DET2. Using the above estimates for the collection efficiency and an
1280 nm transition of the G-center. Correction of the count rate due to observed saturated count rate of 1382 ± 59 counts/second under
the waveguide background was performed prior to fitting with the 78 MHz pulsed excitation, we estimate a lower bound on the quantum
saturation model in Eq. (1) (see Supplementary Note 5 for details on efficiency of 1%. However, the uncertainty on this estimate is large, due
background measurements and subtraction). to low measured count rates and estimates of ηec and ηdipole−wg being
The second-order correlation (g(2)(τ)) was measured with derived from simulation.
continuous-wave 532 nm excitation near the saturation power. Emis-
sion from the waveguide was split with a 50:50 O-band fiber beams- Data availability
plitter prior to time-tagged detection by DET1 and DET2. A histogram The emitter characterization and spectral tuning data presented in this
of the coincidences as a function of time difference between detector study are available in the Zenodo database under https://doi.org/10.
clicks was acquired with 300 ps binwidth and fitted to the second 5281/zenodo.7679379.
order correlation of a two-level system8:
  References
t
g ð2Þ ðtÞ = b 1  ð1  aÞe τ1 , ð2Þ 1. Bhaskar, M. K. et al. Experimental demonstration of memory-
enhanced quantum communication. Nature 580, 60–64 (2020).
where a denotes the normalized zero-delay anti-bunching dip g(2)(0), b 2. Wan, N. H. et al. Large-scale integration of artificial atoms in hybrid
describes the background correlation at large time delay, and τ1 photonic circuits. Nature 583, 226–231 (2020).
describes the time constant of the zero-delay anti-bunching dip. The 3. Wolfowicz, G. et al. Vanadium spin qubits as telecom quantum
resulting second-order correlation of g(2)(0) = 0.38 ± 0.08 indicates emitters in silicon carbide. Sci. Adv. 6, eaaz1192 (2020).
single-photon emission. Poissonian error bars were applied to the data 4. Kim, J.-H., Aghaeimeibodi, S., Carolan, J., Englund, D. & Waks, E.
points in Fig. 3b. The error bar on g(2)(0) refers to the error of the fit Hybrid integration methods for on-chip quantum photonics. Optica
calculated from the covariance in Python’s scipy.curve_fit() 7, 291–308 (2020).
function. 5. Elshaari, A. W., Pernice, W., Srinivasan, K., Benson, O. & Zwiller, V.
Emitter decay lifetimes were measured using pulsed 532 nm Hybrid integrated quantum photonic circuits. Nat. Photonics 14,
excitation at a repetition rate of 34 MHz. A histogram of the SNSPD 285–298 (2020).
clicks as a function of the delay time between the detection event and 6. Dréau, A., Tchebotareva, A., Mahdaoui, A. E., Bonato, C. & Hanson,
the pulsed laser trigger were first clipped to remove laser leakage and R. Quantum frequency conversion of single photons from a
background (see Supplementary Fig. 6a, b for details). The resulting nitrogen-vacancy center in diamond to telecommunication wave-
decay curves fit well to a mono-exponential function: lengths. Phys. Rev. Appl. 9, 064031 (2018).
7. Baron, Y. et al. Detection of single W-centers in silicon. ACS Pho-
tt 0
f ðtÞ = ce τ ð3Þ tonics 9, 2337–2345 (2022).
8. Redjem, W. et al. Single artificial atoms in silicon emitting at tele-
with an extracted emitter lifetime of τ = 8.21 ± 0.14 ns. Here, c describes com wavelengths. Nat. Electron. 3, 738–743 (2020).
a normalization factor given by the number of detected counts at the 9. Bergeron, L. et al. Silicon-integrated telecommunications photon-
pulse peak, t0 is a time shift of the peak, and the error bar on the spin interface. PRX Quantum 1, 020301 (2020).
lifetime of a single emitter is again calculated from the error of the fit. 10. Hollenbach, M., Berencén, Y., Kentsch, U., Helm, M. & Astakhov, G.
The emitter characterization and spectral tuning data presented in this V. Engineering telecom single-photon emitters in silicon for scal-
study are available in the Zenodo database under https://doi.org/10. able quantum photonics. Opt. Express 28, 26111–26121 (2020).
5281/zenodo.7679379. 11. Tyryshkin, A. M. et al. Electron spin coherence exceeding seconds
in high-purity silicon. Nat. Mater. 11, 143–147 (2012).
System efficiency 12. Saeedi, K. et al. Room-temperature quantum bit storage exceeding
The collection efficiency of our system was characterized as follows: 39 minutes using ionized donors in silicon-28. Science 342,
• ηec: Simulated edge coupling efficiency between the cleaved 830–833 (2013).
waveguide mode and a perfectly aligned lensed fiber was cal- 13. Plummer, J. D., Deal, M. & Griffin, P. B. Silicon VLSI Technology:
culated to be 8.3% at 1280 nm. Fundamentals, Practice and Modeling. (Prentice Hall, Upper Saddle
• ηdipole−wg: Maximum dipole coupling to a uni-directional River, NJ, USA, 2000).
waveguide mode was simulated to be 40% (see Supplementary 14. Yeap, G. et al. 5 nm CMOS production technology platform fea-
Fig. 2a and b for details). turing full-fledged EUV, and high mobility channel FinFETs with

Nature Communications | (2023)14:2380 6

Article https://doi.org/10.1038/s41467-023-37655-x

densest 0.021 μm2 SRAM cells for mobile SoC and high perfor- for contributions to the experimental setup. M.P. acknowledges funding
mance computing applications. In Proc. IEEE International Electron from the National Science Foundation (NSF) Convergence Accelerator
Devices Meeting (IEDM), 36.7.1–36.7.4 (2019). Program under grant No.OIA-2040695 and Harvard MURI under grant
15. Vivien, L. & Pavesi, L. Handbook of Silicon Photonics. (CRC Press, No.W911NF-15-1-0548. C.E.-H. and L.D. acknowledge funding from the
Boca Raton, FL, USA, 2016). European Union’s Horizon 2020 research and innovation program under
16. Panuski, C., Englund, D. & Hamerly, R. Fundamental thermal noise the Marie Sklodowska-Curie grant agreements No. 896401 and 840393.
limits for optical microcavities. Phys. Rev. X 10, 041046 (2020). I.C. acknowledges funding from the National Defense Science and
17. Chakraborty, U. et al. Cryogenic operation of silicon photonic Engineering Graduate (NDSEG) Fellowship Program and NSF award
modulators based on the DC Kerr effect. Optica 7, DMR-1747426. D.E. acknowledges support from the NSF RAISE TAQS
1385–1390 (2020). program. This material is based on research sponsored by the Air Force
18. Edinger, P. et al. Silicon photonic microelectromechanical phase Research Laboratory (AFRL), under agreement number FA8750-20-2-
shifters for scalable programmable photonics. Opt. Lett. 46, 1007. The U.S. Government is authorized to reproduce and distribute
5671–5674 (2021). reprints for Governmental purposes notwithstanding any copyright
19. Gyger, S. et al. Reconfigurable photonics with on-chip single-pho- notation thereon. The views and conclusions contained herein are those
ton detectors. Nat. Commun. 12, 1408 (2021). of the authors and should not be interpreted as necessarily representing
20. Pernice, W. H. P. et al. High-speed and high-efficiency travelling the official policies or endorsements, either expressed or implied, of the
wave single-photon detectors embedded in nanophotonic circuits. Air Force Research Laboratory (AFRL), or the U.S. Government.
Nat. Commun. 3, 1325 (2012).
21. Giewont, K. et al. 300-mm monolithic silicon photonics foundry Author contributions
technology. IEEE J. Sel. Top. Quantum Electron. 25, 1–11 (2019). C.E.-H., L.D., and D.E. conceived the idea. M.P., C.E.-H., and L.D.
22. Tait, A. N. et al. Microring resonator-coupled photoluminescence designed the chip, built the measurement setup, and wrote the auto-
from silicon W centers. J. Phys. Photonics 2, 045001 (2020). mated acquisition routines. M.P., C.E.-H., L.D., and C.G. performed the
23. Buckley, S. M. et al. Optimization of photoluminescence from W photophysics characterization and isolation measurements. M.P. ana-
centers in silicon-on-insulator. Opt. Express 28, lyzed the data with input from the authors. C.E.-H. performed the
16057–16072 (2020). spectral trimming measurements and analyzed the data. I.C. performed
24. Lodahl, P., Mahmoodian, S. & Stobbe, S. Interfacing single photons thermal simulations and assisted with cryostat setup. C.C. assisted with
and single quantum dots with photonic nanostructures. Rev. Mod. the single-photon detector measurement setup. All authors contributed
Phys. 87, 347–400 (2015). to the discussion of results and writing of the manuscript.
25. Udvarhelyi, P., Somogyi, B., Thiering, G. & Gali, A. Identification of a
telecom wavelength single photon emitter in silicon. Phys. Rev. Competing interests
Lett. 127, 196402 (2021). The authors declare no competing interests.
26. Coldren, L. A., Corzine, S. W. & Mashanovitch, M. L. Diode Lasers
and Photonic Integrated Circuits. 2nd edition (Wiley, Hoboken, NJ, Additional information
USA, 2012). Supplementary information The online version contains
27. Beaufils, C. et al. Optical properties of an ensemble of G-centers in supplementary material available at
silicon. Phys. Rev. B 97, 035303 (2018). https://doi.org/10.1038/s41467-023-37655-x.
28. Zhang, X. et al. Low-cost and high-efficiency single-mode-fiber
interfaces to silicon photonic circuits. Opt. Commun. 465, Correspondence and requests for materials should be addressed to
125554 (2020). Mihika Prabhu, Carlos Errando-Herranz or Dirk Englund.
29. Chen, X., Thomson, D. J., Crudginton, L., Khokhar, A. Z. & Reed, G. T.
Dual-etch apodised grating couplers for efficient fibre-chip cou- Peer review information Nature Communications thanks the anon-
pling near 1310 nm wavelength. Opt. Express 25, ymous reviewer(s) for their contribution to the peer review of this work.
17864–17871 (2017).
30. Chakravarthi, S., Pederson, C., Kazi, Z., Ivanov, A. & Fu, K.-M. C. Reprints and permissions information is available at
Impact of surface and laser-induced noise on the spectral stability http://www.nature.com/reprints
of implanted nitrogen-vacancy centers in diamond. Phys. Rev. B
104, 085425 (2021). Publisher’s note Springer Nature remains neutral with regard to jur-
31. Durand, A. et al. Broad diversity of near-infrared single-photon isdictional claims in published maps and institutional affiliations.
emitters in silicon. Phys. Rev. Lett. 126, 083602 (2021).
32. Hollenbach, M. et al. Wafer-scale nanofabrication of telecom Open Access This article is licensed under a Creative Commons
single-photon emitters in silicon. Nat. Commun. 13, 7683 (2022). Attribution 4.0 International License, which permits use, sharing,
33. Anderson, C. P. et al. Electrical and optical control of single spins adaptation, distribution and reproduction in any medium or format, as
integrated in scalable semiconductor devices. Science 366, long as you give appropriate credit to the original author(s) and the
1225–1230 (2019). source, provide a link to the Creative Commons license, and indicate if
34. Janitz, E., Bhaskar, M. K. & Childress, L. Cavity quantum electro- changes were made. The images or other third party material in this
dynamics with color centers in diamond. Optica 7, article are included in the article’s Creative Commons license, unless
1232–1252 (2020). indicated otherwise in a credit line to the material. If material is not
35. Wolfowicz, G. et al. 29Si nuclear spins as a resource for donor spin included in the article’s Creative Commons license and your intended
qubits in silicon. N. J. Phys. 18, 023021 (2016). use is not permitted by statutory regulation or exceeds the permitted
use, you will need to obtain permission directly from the copyright
Acknowledgements holder. To view a copy of this license, visit http://creativecommons.org/
The authors would like to acknowledge Dr. Matt Trusheim and Isaac licenses/by/4.0/.
Harris for useful discussions regarding emission physics and defect
structure. The authors would also like to thank Dr. Dalia Ornelas-Huerta © The Author(s) 2023

Nature Communications | (2023)14:2380 7