Abstract
Photonic qubits should be controllable onchip and noisetolerant when transmitted over optical networks for practical applications. Furthermore, qubit sources should be programmable and have high brightness to be useful for quantum algorithms and grant resilience to losses. However, widespread encoding schemes only combine at most two of these properties. Here, we overcome this hurdle by demonstrating a programmable silicon nanophotonic chip generating frequencybin entangled photons, an encoding scheme compatible with longrange transmission over optical links. The emitted quantum states can be manipulated using existing telecommunication components, including active devices that can be integrated in silicon photonics. As a demonstration, we show our chip can be programmed to generate the four computational basis states, and the four maximallyentangled Bell states, of a twoqubits system. Our device combines all the key properties of onchip state reconfigurability and dense integration, while ensuring high brightness, fidelity, and purity.
Similar content being viewed by others
Introduction
Photons serve as excellent carriers of quantum information. They have long coherence times at room temperature, and are the inescapable choice for broadcasting quantum information over long distances, either in free space or through the optical fiber network. Quantum state initialization is a particularly important task for photonic qubits, since adjusting entanglement after emission is nontrivial. Initialization strategies depend on the degree of freedom used to encode quantum information, and the most common choice for quantum communication over optical channels is timebin encoding^{1}. Here, the twoqubit levels consist of the photon being in one of two time windows, generally separated by a few nanoseconds. Timebin encoding is extremely resilient to phase fluctuations resulting from thermal noise in optical fibers, with qubits maintaining their coherence even over hundreds of kilometers^{2,3}. However, the control of the state in which timebinentangled photons are generated is challenging, and impractical in emerging nanophotonic platforms. For onchip manipulation of qubit states, dualrail encoding, in which the two states of a qubit correspond to the photon propagating in one of two optical waveguides, is a superior strategy^{4,5}, and is thus a common choice for quantum computing and quantum simulation in integrated platforms. Yet this approach is not easily compatible with longdistance transmission links using either optical fibers or free space channels.
Recently, frequencybin encoding has been proposed, and experimentally demonstrated, as an appealing strategy that can combine the best characteristics of timebin and dualrail encodings^{6,7,8,9,10,11}. In this approach, quantum information is encoded by the photon being in a superposition of different frequency bands. Frequency bins can be manipulated by means of phase modulators, and are resistant to phase noise in longdistance propagation. Pioneering studies have investigated the generation and manipulation of frequencybinentangled photons in integrated resonators. They have considered quantum state tomography of entangled photon pairs^{12}, qudit encoding^{13}, and multiphoton entangled states^{14}. The experimental results have all been achievable thanks to the recent development of highQ integrated resonators in the silicon nitride and silicon oxynitride platforms.
Despite all this progress, there are obstacles that must be overcome in order to exploit the full advantage of photonic integration. In frequencybin encoding today, the generation of photon pairs occurs via spontaneous fourwave mixing in a single ring resonator, with the desired state obtained outside the chip, by means of electrooptical modulators and/or pulse shapers. And since commercial modulators have limited bandwidth, the frequency span separating the photons cannot exceed a few tens of gigahertz, which sets a limit to the maximum free spectral range of the resonator. Finally, because spontaneous fourwave mixing efficiency scales quadratically with the resonator free spectral range^{15}, there is also a significant tradeoff between the generation rate and the number of accessible frequency bins.
In this work, we show that these limitations can be overcome by utilizing the flexibility of light manipulation in a nanophotonic platform and the dense optical integration possible in silicon photonics. Our approach is based on constructing the desired state by direct, onchip control of the interference of biphoton amplitudes generated in multiple ring resonators that are coherently pumped. States can thus be constructed “piecebypiece" in a programmable way, by selecting the relative phase of each source. In addition, since the frequencybin spacing is no longer related to the ring radius, one can work with very highfinesse resonators, reaching megahertz generation rates. These two breakthroughs, namely high emission rates in combination with high values of the free spectral range, together with output state control using onchip components, are only possible using multiple rings: they would not be feasible were the frequency bins encoded on the azimuthal modes of a single resonator.
We demonstrate that with the very same device one can generate all superpositions of the \(\left00\right\rangle\) and \(\left11\right\rangle\) states or, in another configuration with different frequencybin spacing, all superpositions of the \(\left01\right\rangle\) and \(\left10\right\rangle\) states. One needs only to drive the onchip phase shifter and set the pump configuration appropriately. This means that all four fullyseparable states of the computational basis and all four maximally entangled Bell states (\(\left{\Phi }^{\pm }\right\rangle=(\left00\right\rangle \pm \left11\right\rangle )/\sqrt{2}\) and \(\left{\Psi }^{\pm }\right\rangle=(\left01\right\rangle \pm \left10\right\rangle )/\sqrt{2}\)) are accessible. Our high generation rate allows us to perform quantum state tomography of all these states, reaching fidelities up to 97.5% with purities close to 100%.
Results
Device characterization and principle of operation
The device is schematically represented in Fig. 1a. The structure is operated by exploiting the fundamental transverse electric (TE) mode of a silicon waveguide, with a 600 × 220 nm^{2} cross section, buried in silica. Two silicon ring resonators (Ring A and Ring B) in allpass configuration act as sources of photon pairs. Their radii are some 30 μm to ensure high generation rates, and they are not commensurate so that the two free spectral ranges are different: FSR_{A} = 377.2 GHz and FSR_{B} = 373.4 GHz, respectively. The two rings are critically coupled to a bus waveguide and their resonance lines can be tuned independently by means of resistive heaters. The device also contains a tunable MachZehnder interferometer (MZI), whose outputs are connected to the input of two tunable adddrop filters that allow one to control the field intensity and relative phase with which Ring A and Ring B are pumped in the spontaneous fourwave mixing experiment^{16}.
Linear transmission measurements through the bus waveguide are shown in Fig. 1b–g. In a first configuration (Fig. 1b–d), which we will later refer to as “Φ”, two resonances of Ring A and Ring B are spectrally aligned to be later used for pumping, thus only one transmission dip is observed at 194 THz (1545 nm) in Fig. 1c. Since Ring A and Ring B have different free spectral ranges, the other resonances are not aligned, and one observes double dips, with spacing Δ(m) = ∣m∣(FSR_{A} − FSR_{B}), with m being the azimuthal order with respect to the pump resonance. In Fig. 1b and d we plot the transmission double dip corresponding to m = − 5 and m = +5, named “idler” and “signal”, respectively. For both the signal and idler bands the resonances of Ring A and Ring B are separated by Δ = 19 GHz. Later, the two frequencies will be used to encode the two states of the qubits, with signal and idler pairs of frequencies representing the two qubits. For this reason, in Fig. 1b and d, we name \({\left0\right\rangle }_{{{{{{{{\rm{s}}}}}}}},{{{{{{{\rm{i}}}}}}}}}\) the two frequency bins closer to the pump, and \({\left1\right\rangle }_{{{{{{{{\rm{s}}}}}}}},{{{{{{{\rm{i}}}}}}}}}\) the two bins further away from the pump, in line with previous works on frequencybin entanglement^{6}. Our device can also operate in a different configuration, which we will refer to as “Ψ”. Here Ring A and Ring B are thermally tuned so that the resonances corresponding to the states \({\left0\right\rangle }_{{{{{{{{\rm{i}}}}}}}}}\) and \({\left1\right\rangle }_{{{{{{{{\rm{s}}}}}}}}}\) belong to Ring B and those corresponding to \({\left0\right\rangle }_{{{{{{{{\rm{s}}}}}}}}}\) and \({\left1\right\rangle }_{{{{{{{{\rm{i}}}}}}}}}\) belong to Ring A (see Fig. 1e–g). As can be seen from all panels in Fig. 1b–g, the resonances of the two generating rings have quality factors Q ≈ 150, 000 (Full width at half maximum Γ ≈ 1.3 GHz), which guarantee wellseparated frequency bins and high generation rates.
The basic principle of operation of the device is the following: (i) Ring A and Ring B are set in the proper configuration (e.g., Φ) by controlling the thermal tuners; (ii) The pump power is coherently distributed between the two rings with the required relative phase and amplitude set either through the MZI or directly through the bus waveguide; (iii) Photon pairs are collected in the bus waveguide, with the desired state resulting from a coherent superposition of the twophoton states that would be generated by each ring separately.
Spontaneous fourwave mixing
The photon generation efficiency through spontaneous fourwave mixing (SFWM) was assessed for the two rings by setting the device in configuration Ψ, which is convenient to pump each ring individually through the bus waveguide. The two resonators were pumped with an external tunable laser, and the chip output was separated in the signal (194.7–197.2 THz), pump (192.2–194.7 THz), and idler (189.7–192.2 THz) bands by means of a telecomgrade coarse wavelength division multiplexer (see Supplementary Fig. 1). The generated signal and idler photons were then narrowband filtered using tunable fiber Bragg gratings with an 8 GHz stopband, and routed to a pair of superconductive singlephoton detectors. The overall insertion losses from the bus waveguide to the detectors are 6 and 7 dB for signal and idler channels, respectively. The results of the experiment are summarized in Fig. 2. The two rings exhibit similar generation efficiency \(\eta=R/{P}_{{{{{{{{\rm{wg}}}}}}}}}^{2}\), with η_{A} = 57.6 ± 2.1 Hz/μW^{2} for Ring A and η_{B} = 62.4 ± 1.7 Hz/μW^{2} for Ring B^{15}. The internal pair generation rate R can exceed 2 MHz for both ring resonators (Fig. 2a). A high coincidencetoaccidental ratio (CAR) exceeding 10^{2} was obtained for any value of the input power, a necessary condition to ensure a high purity of the generated state (Fig. 2b).
We now turn to the spectral properties of the generated photon pairs and the demonstration of entanglement. We set our device to operate in the Φ configuration, which will later be used to generate the maximally entangled state
where \(\left00\right\rangle={\left0\right\rangle }_{{{{{{{{\rm{s}}}}}}}}}{\left0\right\rangle }_{{{{{{{{\rm{i}}}}}}}}}\), \(\left11\right\rangle={\left1\right\rangle }_{{{{{{{{\rm{s}}}}}}}}}{\left1\right\rangle }_{{{{{{{{\rm{i}}}}}}}}}\), and the phase θ can be adjusted by acting on the thermoelectric phase shifter after the interferometer (see Supplementary Note 1); θ = 0 and θ = π correspond to the wellknown Bell states \(\left{\Phi }^{+}\right\rangle\) and \(\left{\Phi }^{}\right\rangle\), respectively. The corresponding SFWM spectrum of the signal and idler bands is shown in Fig. 3a and b (upper panels); the device was electrically tuned to set θ = 0, with the pump power split equally between Rings A and B by means of the MZI. Here we focus on the azimuthal order m = ±5, with the generated frequency bins clearly distinguishable in the marginal signal and idler spectra.
Twophoton interference
In order to demonstrate entanglement, the demultiplexed signal and idler photons were routed (see Supplementary Fig. 1) to two intensity electrooptic modulators (EOMs), coherently driven at f_{m} = 9.5 GHz, which corresponds to half the frequencybin separation of the selected azimuthal order m = ±5. The modulators operate at the minimum transmission point (i.e., at bias voltage V_{π}) to achieve doublesideband suppressedcarrier amplitude modulation. The amplitude of the modulating RF signal was chosen to maximize the transferred power from the carrier to the firstorder sidebands, with a modulation efficiency of around −4.8 dB, corresponding to a modulation index β ≈ 1.7. These losses can be reduced by integrating the modulators on chip. Furthermore, our approach allows the use of frequencybin spacings potentially much lower than the frequency cutoff of the modulators. This will allow the use of complex wavelength shifting modulation techniques^{17,18} to avoid the generation of double sidebands and the consequent 3 dB in added losses.
The resulting spectrum is shown in the lower panels of Fig. 3a and b, in which one can clearly recognize three peaks. Indeed, given the chosen modulated frequency, the central one results from the overlap of the down and upperconverted original bins. From a quantum optics point of view, this operation achieves quantum interference of the original frequency bins^{12} in a similar fashion to what can be done with time bins in a Franson interferometer^{19,20}. Here the achievable visibility of quantum interference depends on the correct superposition of the spectra of the modes encoding the two frequency bins for the signal and idler photons, respectively, as outlined in Fig. 4a.
For coincidence counting, the modulated signal and idler photons were filtered using narrowband fiber Bragg gratings to select only the central line at the output of the corresponding modulator, and routed to the singlephoton detectors. The results of this experiment are shown in Fig. 4b and c as a function of the modulation frequency. The rapid oscillation of the correlation is due to the different phase acquired by the photons during their propagation from the device to the EOMs. If the resonances share the same Q factor and coupling efficiency, the coincidence rate is proportional to the crosscorrelation function (see Supplementary Note 3):
where δT = t_{i} − t_{s} is the difference between the idler and signal arrival times at the EOMs, and φ_{s(i)} is the signal (idler) modulator driving phase. Figure 4b shows good agreement between the experimental results and curve described by Eq. (2) for φ_{s} − φ_{i} = θ/2 and δT = 8.5 ns, which corresponds to the ~2 m path difference between the idler and signal EOMs in our setup. The curve visibility obtained from a leastsquare fit of the model is V = 98.7 ± 1.2%. The twophoton correlation reaches its maximum value \({G}_{{{{{{{{\rm{s}}}}}}}},{{{{{{{\rm{i}}}}}}}}}^{(2)}({f}_{{{{{{{{\rm{m}}}}}}}}})\,\approx \,2\) when f_{m} = Δ/2, as shown in other works on frequencybin entanglement^{12}. Thanks to the high brightness of the source, coincidence counts on the detectors remain well above the noise level even with the added losses from the modulators, with a CAR level > 50 and detected coincidence rate > 2 kHz, thus implying an interference pattern with a high visibility.
With these results in hand, we set f_{m} = Δ/2 and varied φ_{s} to perform a Belllike experiment. The corresponding quantum interference curves are reported in Supplementary Note 2.
Quantum state tomography
Finally, we show that our device can be operated to generate, directly on chip, frequencybin photon pairs with a controllable output state. For each of the explored configurations we performed quantum state tomography^{21}. First we kept the device in configuration Φ, in which Ring A and Ring B generate photon pairs in the state \({\left0\right\rangle }_{{{{{{{{\rm{s}}}}}}}},{{{{{{{\rm{i}}}}}}}}}\) and \({\left1\right\rangle }_{{{{{{{{\rm{s}}}}}}}},{{{{{{{\rm{i}}}}}}}}}\), respectively. Thus, the two states of the computational basis \(\left00\right\rangle={\left0\right\rangle }_{{{{{{{{\rm{s}}}}}}}}}{\left0\right\rangle }_{{{{{{{{\rm{i}}}}}}}}}\) and \(\left11\right\rangle={\left1\right\rangle }_{{{{{{{{\rm{s}}}}}}}}}{\left1\right\rangle }_{{{{{{{{\rm{i}}}}}}}}}\) can be generated by selectively pumping only the appropriate resonator, as shown in Fig. 5a and b. The states were characterized via quantum state tomography^{12,21,22}, as detailed in the Methods section. In both cases the states are accurately reproduced, with fidelity and purity exceeding 90%.
In a second experiment, the MZI was operated to split the pump power so that the probabilities of generating a photon pair in Ring A and in Ring B are equal. If the pump power is sufficiently low that the probability of emitting two photon pairs is negligible, then the generated frequency bins are in the state \(\left\Phi (\theta )\right\rangle\) described by Eq. (1), where the phase factor θ is controlled by the phase shifter after the MZI. By setting θ = 0 or π, we were able to generate the two Bell states \(\left{\Phi }^{+}\right\rangle\) and \(\left{\Phi }^{}\right\rangle\), respectively (see Fig. 5c and d). The real and imaginary parts of the density matrix are shown in Fig. 5g, h, k, and l. As expected, we found nonzero offdiagonal terms in the real part of the density matrix, which indicate entanglement. In these cases as well the device is capable of outputting the desired state with purity and fidelity exceeding 90%. The entanglement of formation, a figure of merit to quantify the entanglement of the generated pairs^{23}, was extracted from the measured density matrices, yielding values > 80% for the two Bell states, in contrast with values < 20% for the two separable states \(\left00\right\rangle\) and \(\left11\right\rangle\).
Our device can also operate in the Ψ configuration, with the ring resonances arranged as shown in Fig. 1e–g. In this case one is able to generate also the two remaining computational basis states \(\left01\right\rangle\), \(\left10\right\rangle\) and the two remaining Bell states \(\left{\Psi }^{+}\right\rangle\) and \(\left{\Psi }^{}\right\rangle\). Note that in this configuration, the pump resonances for the two ring resonators are not aligned (Fig. 1f).
When generating the two separable states, either Ring A (to generate \(\left01\right\rangle\)) or Ring B (to generate \(\left10\right\rangle\)) was pumped through the bus waveguide by simply tuning the pump to the corresponding resonance (see Fig. 6a and b). To generate the two Bell states, the pump pulse spectrum (which is tuned to be in the middle of the two resonances) is shaped using an external EOM operated at the frequency corresponding to half the difference between the two pump resonances (f_{m,p} = Δ_{p}/2 = 19 GHz) (see Fig. 6c and d and the Methods section). The pumping ratio and the phase between the two rings were adjusted by tailoring the modulation to obtain an equal probability amplitude of generating a singlephoton pair for the states \(\left01\right\rangle\) and \(\left10\right\rangle\) respectively, while still keeping the probability of double pair generation negligible. The relative phase of the superposition can be controlled by adjusting the EOM driving phase to select either \(\left{\Psi }^{+}\right\rangle\) or \(\left{\Psi }^{}\right\rangle\).
The four generated states were characterized via quantum state tomography as in the previous case. However, we stress that here two different values of bin spacing for the signal (Δ_{s} = 19 GHz) and idler (Δ_{i} = 3Δ_{s} = 57 GHz) qubits were used. While this does not constitute a problem for the generation of entanglement, as the Hilbert space of the two qubits is built from the tensor product of Hilbert spaces of two qubits with different values for Δ_{s} and Δ_{i}, it offered us the opportunity to demonstrate, for the first time, frequencybin tomography for uneven spacing. This is done by operating the signal and idler EOMs (see Supplementary Fig. 1) at different frequencies equal to half the frequency spacing of the corresponding resonances.
The experimental results are shown in Fig. 6e–l. All four states were prepared with fidelity close to or exceeding 90%, and purity between 85 and 100%. The entanglement of formation is below 5% for the separable states \(\left01\right\rangle\) and \(\left10\right\rangle\), while above 80% for the Bell states \(\left{\Psi }^{+}\right\rangle\) and \(\left{\Psi }^{}\right\rangle\), as expected. The reconstructed density matrices show increased noise with respect to those reported in Fig. 5, because the modulation efficiency of our idler modulator was significantly reduced at such a high frequency, resulting in additional losses lowering the count rate on the detectors (see the Methods section).
Scalability to higherdimensional states
Our approach can be generalized to frequencybin qudits by scaling the number of coherently excited rings. We give a proof of principle demonstration of this capability by using a different device hosting d = 4 rings and adddrop filters. The four sources, labeled A, B, C, D, have radii R_{j} = R_{0} + jδR (with j = 0, …, d − 1), where R_{0} = 30 μm and δR = 0.1 μm, which leads to a bin spacing of ~9 GHz at 7 FSR from the pump. The spectral response of the device at the output of the bus waveguide, indicated in Fig. 7a, shows the four equidistant bins (labeled 0, 1, 2, 3) associated with the signal and with the idler photons, and the overlapping resonances of the rings at the pump frequency. As in the case of qubits, we used an MZI tree to split the pump into four paths, each feeding a different adddrop ring filter that is used to control the field intensity at the photon pair sources. We focused on the capability to generate the four computational basis states and the twodimensional Bell states formed by adjacent frequency bins pairs. First, the adddrop filters are tuned on resonance one at a time. This selects the computational basis state that is generated. We characterized those states by performing a Zbasis correlation measurement, i.e., by projecting the signal and the idler photon on the Zbasis \(\{{\leftl\right\rangle }_{{{{{{{{\rm{s}}}}}}}}}{\leftm\right\rangle }_{{{{{{{{\rm{i}}}}}}}}}\},\,\,l(m)=0,1\,,\,2,\,3\), in order to measure the uniformity and the crosstalk between the four frequency bins. From the correlation matrices, shown in Fig. 7b–e, it was possible to measure the ratio of the coincidence counts n_{ll} in the frequencycorrelated basis \({\leftl\right\rangle }_{{{{{{{{\rm{s}}}}}}}}}{\leftl\right\rangle }_{{{{{{{{\rm{i}}}}}}}}}\) to that in the uncorrelated basis ∑_{l≠m}n_{lm}, and it is about two orders of magnitude. We could compensate for the slightly different amplitude of the different basis states by acting on the MZI tree at the input. Second, the adddrop filters associated with the adjacent frequencybinpairs 0–1, 1–2, and 2–3 are tuned on resonance one at a time, thus generating the Bell states \({\left{\Phi }^{+}\right\rangle }_{0,1}\), \({\left{\Phi }^{+}\right\rangle }_{1,2}\) and \({\left{\Phi }^{+}\right\rangle }_{2,3}\), being \({\left{\Phi }^{+}\right\rangle }_{l,m}=(\leftll\right\rangle+\leftmm\right\rangle )/\sqrt{2}\). The visibility of quantum interference is assessed by mixing the corresponding frequency bins with the electrooptic modulator. Unlike in the qubit experiment, here we choose a modulation frequency that matches the spectral separation between the bins. We used phase modulators configured to create firstorder sidebands of amplitude equal to that of the baseband, and recorded the coincidences in signal/idler bins 0, 1, 2, and 3. The resulting Bell curves, shown in Fig. 7f, have visibilities V_{0,1} = 0.831(5), V_{1,2} = 0.884(6), and V_{2,3} = 0.81(1), indicating the presence of entanglement between the binpairs in all cases. It is worth noting that, as in the twodimensional case, the relative phase between the three Bell curves in Fig. 7f could be adjusted using onchip phase shifters in order to realize maximally entangled highdimensional Bell states.
Discussion
We demonstrated that a rich variety of separable and maximally entangled states, including any linear superposition of \(\{\left00\right\rangle,\left11\right\rangle \}\) or \(\{\left01\right\rangle,\left10\right\rangle \}\), can be generated using frequencybin encoding in a single programmable nanophotonic device, fabricated with existing silicon photonic technologies compatible with multiproject wafer runs. This guarantees that these devices can be available for widespread use in applications, ranging from quantum communication to quantum computing.
Our approach constitutes an innovative paradigm for the integration of frequencybin devices that goes wellbeyond a miniaturization of bulk strategies. Indeed, unlike previous implementations, the states are all generated inside the device, without relying on offchip manipulation of a single initial state. Controllability of the generated state was shown to be readily accessible onchip, via electrical control of thermooptic actuators in one configuration (Φ), and by tailoring the pump spectral properties in another (Ψ). In a future version of the device the use of more than two rings for the definition of the state will allow the two configurations to have the same frequency spacing for the qubits. As a result, the device will be capable of generating all four Bell states with the same physical characteristics, as recently demonstrated using an external periodically poled lithium niobate crystal^{24}; it will also be used to explore more of the Hilbert space of the two qubits.
Since in our approach the frequencybin spacing is only limited by the resonator linewidth, the requirements for the electrooptic modulators are greatly relaxed with respect to previous implementations. Indeed, as demonstrated in this work, the frequencybin separation is compatible with existing silicon integrated modulators^{25}. Thus, one can foresee a future evolution of our device that will involve modulators integrated onchip. This will further increase its suitability for practical applications, such as quantum key distribution and quantum communications in general. In addition, the ability to independently choose the bin spacing Δ for both qubits, as shown in Fig. 1b–g, demonstrates an additional flexibility in choosing the basis for frequencybin encoding that can be exploited for the engineering of the source.
The approach demonstrated here is scalable, for one can design and implement devices with more than two generating rings by taking advantage of silicon dense integration, opening the possibility of using frequency qudits instead of simple qubits. As demonstrated in a number of theoretical proposals, such an ability will be of pivotal importance for multiple applications in quantum communication, sensing, and computing algorithms^{26}. In addition, our approach could be extended to take advantage of recent progress in alloptical frequency conversion^{27,28} to expand the manipulation bandwidth of the frequency bins, thus allowing one to increase the dimension of the accessible Hilbert space enormously.
Finally, our approach allowed us to overcome the tradeoff between the frequencybin spacing and the generation rate that characterized previous work. This was instrumental in achieving a comprehensive assessment of the properties of the generated states, which could be performed using only telecomgrade fiber components—with the sole exception of singlephoton detection—with an overall low loss (<4 dB) ensured by the allfiber technology. The accuracy and the precision that have been achieved in our measurements are stateoftheart for frequencybin encoding, even considering results obtained with bulk sources. wellbeyond any other reported so far on frequencybin encoding. All these results will usher in the use of frequencybin qubits as a practical choice for photonic qubits, capable of combining easy manipulation and robustness for longhaul transmission.
Methods
Sample fabrication
The device was fabricated at CEALeti (Grenoble), on a 200 mm SilicononInsulator (SOI) substrate with a 220 nm thick top device layer of crystalline silicon on 2 μm thick SiO_{2} buried oxide. The patterning process of the silicon photonics devices and circuits combines deep ultraviolet (DUV) lithography with 120 nm resolution, inductively coupled plasma etching (realized in collaboration with LTM—Laboratoire des Technologies de la Microélectronique) and O_{2} plasma resist stripping. Hydrogen annealing was performed in order to strongly reduce etchinginduced waveguide sidewalls roughness^{29}. After highdensity plasma, lowtemperature oxide (HDPLTO) encapsulation—resulting in a 1125 nm thick SiO_{2} layer—110 nm of titanium nitride (TiN) were deposited and patterned to create the thermal phase shifters, while an aluminium copper layer (AlCu) was used for the electrical pad definition. Finally, a deep etch combining two different steps—C_{4}F_{8}/O_{2}/CO/Ar plasma running through the whole thickness of both silica upper cladding and buried oxide, followed by a Bosch deep reactive ion etching (DRIE) step to remove 150 μm of the 725 μm thick Si substrate—was implemented to separate the subdice, thus ensuring highquality opticalgrade lateral facets for chiptofiber edge coupling.
Linear spectroscopy
The experimental apparatus is schematically represented in Supplementary Fig. 1. The linear characterization of the sample shown in Fig. 1 was realized by scanning the wavelength of a tunable laser (Santec TSL710), with its polarization controlled by a fiber polarization controller (PC). Light was coupled to the sample at the input of the bus waveguide and collected at the output using a pair of lensed fibers (nominal mode field diameter: 3 μm), with an insertion loss lower than 3 dB/facet. The output signal was detected by an amplified InGaAs photodiode and recorded in real time by an oscilloscope. The resonance configuration was adjusted by addressing each ring resonator’s phase shifter with electric probes driven by multichannel power supply.
Nonlinear characterization
The SFWM efficiency for each resonator was assessed through powerscaling experiments (Fig. 2). The flux of generated idler and signal photons was measured by varying the pump power coupled to each microring, while keeping the resonances in place by acting on the thermoelectric phase shifters. The tunable laser source spectrum was filtered by a bandpass (BP) filter in order to reduce the amount of spurious photons at signal and idler frequencies coming from the launching part of the setup, mainly associated with amplified spontaneous emission of the laser diode and Raman fluorescence from the fibers. The collected signal and idler photons were first separated using a coarse wavelength division multiplexer (CWDM), with 2.5 THz (20 nm) nominal channel separation and measured interchannel crosstalk < −80 dB. The frequency bins of interest were then narrowband filtered (3 dBbandwidth: 8 GHz) by a pair of tunable fiber Bragg gratings (FBG): besides selecting the frequency bins with high accuracy, this procedure also suppresses any spurious broadband photon falling outside the bandwidth of the input bandpass filter and not eliminated by the CWDM. The resulting signal and idler photons were routed, using circulators, towards two superconductive singlephoton detectors (SSPDs), where timecorrelated singlephoton counting (TCSPC) was performed with a precision of about 35 ps, mainly determined by the detectors jitter. A coincidence window of τ_{c} = 380 ps was chosen by selecting the average full width at half maximum (FWHM) of the histogram peak. Accidental counts were estimated from the background level; note that this value is not subtracted from the number of coincidences counted, but was used only to estimate the coincidencetoaccidental ratio, according to the formula:
Quantum state tomography
Twophoton interferometry and tomography of the generated quantum states were performed by including a pair of intensity EOMs (iXblue MXLN) at the signal and idler demultiplexer outputs, coherently driven by a multichannel RF generator (AnaPico APMS20G). The sidebands of interest were selected by tuning the central stopband wavelength of the FBGs. The tomography of each quantum state involved 16 individual measurements, each performed in an acquisition time of 15 s. For each measurement, each FBG was tuned to one of the three sideband frequencies obtained from the modulation of the signal (idler) bins, and the EOM’s relative phase was adjusted appropriately. Estimation of the density matrices was performed via maximumlikelihood technique^{21,22}. For the generation of states in the \(\{\left01\right\rangle,\,\left10\right\rangle \}\) basis (Ψ configuration), we added a phase EOM at the input of the setup, coherently driven by the same RF source used for tomography, and we entered the chip at the bus waveguide. The two generation rings were then pumped by the firstorder sidebands, while their relative phase was fixed by the phase of the modulation.
Measurement of qudits
For the Zbasis correlation measurement, a total set of different projectors (for each photon) is used for each basis state. The projector \({\leftl\right\rangle }_{{{{{{{{\rm{s}}}}}}}}}{\leftm\right\rangle }_{{{{{{{{\rm{i}}}}}}}}}\) is implemented by setting the signal(idler) FBG to reflect only the frequencybin l(m). For those combinations carrying negligible counts (corresponding to frequency uncorrelated bins), the central frequency of the two FBGs cannot be determined by simply maximizing the coincidence rate or the flux of singles in each bin. To circumvent this, we coupled a secondary laser beam in the counterpropagating direction with respect to that of the pump, and recorded the back reflected light from the sample. The spectra of the latter are monitored after being transmitted by the FBGs, and simultaneously reveal the spectral location of the stopband of the FBG and the four resonance frequencies of the rings. In this way, the stopband can be overlapped with the desired frequency bin with high precision.
Data availability
The data presented in this study are available at https://doi.org/10.5281/zenodo.7464081. Additional data are available from the corresponding authors upon request.
References
Marcikic, I. et al. Timebin entangled qubits for quantum communication created by femtosecond pulses. Phys. Rev. A 66, 062308 (2002).
Marcikic, I. et al. Distribution of timebin entangled qubits over 50 km of optical fiber. Phys. Rev. Lett. 93, 180502 (2004).
Inagaki, T., Matsuda, N., Tadanaga, O., Asobe, M. & Takesue, H. Entanglement distribution over 300 km of fiber. Opt. Express 21, 23241–23249 (2013).
Kok, P. et al. Linear optical quantum computing with photonic qubits. Rev. Mod. Phys. 79, 135–174 (2007).
Silverstone, J. W. et al. Qubit entanglement between ringresonator photonpair sources on a silicon chip. Nat. Commun. 6, 7948 (2015).
Olislager, L. et al. Frequencybin entangled photons. Phys. Rev. A 82, 013804 (2010).
Kaneda, F., Suzuki, H., Shimizu, R. & Edamatsu, K. Direct generation of frequencybin entangled photons via twoperiod quasiphasematched parametric downconversion. Opt. Express 27, 1416 (2019).
Rieländer, D. et al. Frequencybin entanglement of ultranarrow band nondegenerate photon pairs. Quantum Sci. Technol. 3, 014007 (2017).
Lukens, J. M. & Lougovski, P. Frequencyencoded photonic qubits for scalable quantum information processing. Optica 4, 8–16 (2017).
Lu, H.H. et al. Quantum interference and correlation control of frequencybin qubits. Optica 5, 1455–1460 (2018).
Silverstone, J. W. et al. Onchip quantum interference between silicon photonpair sources. Nat. Photonics 8, 104–108 (2014).
Imany, P. et al. 50GHzspaced comb of highdimensional frequencybin entangled photons from an onchip silicon nitride microresonator. Opt. Express 26, 1825 (2018).
Kues, M. et al. Onchip generation of highdimensional entangled quantum states and their coherent control. Nature 546, 622–626 (2017).
Reimer, C. et al. Generation of multiphoton entangled quantum states by means of integrated frequency combs. Science 351, 1176–1180 (2016).
Azzini, S. et al. From classical fourwave mixing to parametric fluorescence in silicon microring resonators. Opt. Lett. 37, 3807–3809 (2012).
Liscidini, M. & Sipe, J. E. Scalable and efficient source of entangled frequency bins. Opt. Lett. 44, 2625 (2019).
Riaziat, M. L., Virshup, G. F. & Eckstein, J. N. Optical wavelength shifting by travelingwave electrooptic modulation. IEEE Photonics Technol. Lett. 5, 1002–1005 (1993).
Kittlaus, E. A. et al. Electrically driven acoustooptics and broadband nonreciprocity in silicon photonics. Nat. Photonics 15, 43–52 (2021).
Franson, J. D. Bell inequality for position and time. Phys. Rev. Lett. 62, 2205–2208 (1989).
Grassani, D. et al. Micrometerscale integrated silicon source of timeenergy entangled photons. Optica 2, 88–94 (2015).
James, D. F. V., Kwiat, P. G., Munro, W. J. & White, A. G. Measurement of qubits. Phys. Rev. A 64, 052312 (2001).
Takesue, H. & Noguchi, Y. Implementation of quantum state tomography for timebin entangled photon pairs. Opt. Express 17, 10976 (2009).
Wootters, W. K. Entanglement of formation of an arbitrary state of two qubits. Phys. Rev. Lett. 80, 2245–2248 (1998).
Seshadri, S., Lu, H.H., Leaird, D. E., Weiner, A. M. & Lukens, J. M. Complete FrequencyBin Bell Basis Synthesizer. Phys. Rev. Lett. 129, 230505 (2022).
Streshinsky, M. et al. The road to affordable, largescale silicon photonics. Opt. Photon. N. 24, 32–39 (2013).
Wang, Y., Hu, Z., Sanders, B. C. & Kais, S. Qudits and highdimensional quantum computing. Front. Phys. 8, 479 (2020).
Kobayashi, T. et al. Frequencydomain HongOuMandel interference. Nat. Photonics 10, 441–444 (2016).
Li, Q. et al. Tunable quantum beat of single photons enabled by nonlinear nanophotonics. Phys. Rev. Appl. 12, 054054 (2019).
Bellegarde, C. et al. Improvement of sidewall roughness of submicron soi waveguides by hydrogen plasma and annealing. IEEE Photonics Technol. Lett. 30, 591–594 (2018).
Acknowledgements
This work has been supported by Ministero dell’Istruzione, dell’Università e della Ricerca (Dipartimenti di Eccellenza Program (2018–2022)  F11I18000680001). The device has been designed using the open source Nazca design^{TM} framework.
Author information
Authors and Affiliations
Contributions
M.L. and J.E.S. conceived the original idea. F.A.S., M.L., M.G., and D.B. conceived the the device design. F.A.S. and H.E.D. engineered and fabricated the experimental device, under the supervision of C.S. L.Y., C.P., E.P., and C.S. contributed to the engineering and supervised the fabrication. M.C., F.A.S., M.B., and N.T. performed the experimental measurements and data analysis. M.G. and D.B. supervised the the experiments. M.C., F.A.S., M.L., and D.B. developed the theory. N.B. contributed to the analysis of quantum state tomography results. J.E.S. and M.L. supervised the theoretical aspects. M.C., F.A.S., M.B., L.G., J.E.S., M.L., M.G., and D.B. wrote and revised the manuscript. M.L., C.S., M.G., and D.B. coordinated and supervised the project. All authors commented on the manuscript.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing interests.
Peer review
Peer review information
Nature Communications thanks Avik Dutt and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Clementi, M., Sabattoli, F.A., Borghi, M. et al. Programmable frequencybin quantum states in a nanoengineered silicon device. Nat Commun 14, 176 (2023). https://doi.org/10.1038/s41467022357736
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41467022357736
This article is cited by

Deterministic entangling gates with nonlinear quantum photonic interferometers
Communications Physics (2024)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.