Microcomb-driven silicon photonic systems | Nature


Design and fabrication of the gadgets

The ring waveguides of the AlGaAsOI resonators have been designed to work throughout the regular dispersion regime within the C band, with dimensions of 400 nm × 1,000 nm. The width of the bus waveguide on the aspect was designed to be 200 nm for environment friendly chip-to-fibre coupling. The fabrication of the AlGaAs microresonators was based mostly on heterogeneous wafer bonding expertise. The method is presently realized on the 100-mm-wafer scale with none strict fabrication processes similar to chemical-mechanical sharpening or high-temperature annealing that aren’t appropriate with the CMOS course of. It will possibly due to this fact be instantly adopted by present III–V/Si photonic foundries51. A Q issue >2 million could be obtained within the AlGaAsOI resonator, comparable to a waveguide lack of <0.3 dB cm−1. The fraction of aluminium is 0.2, which corresponds to a two-photon absorption wavelength of round 1,480 nm. The epitaxial wafer progress was achieved utilizing molecular-beam epitaxy. A 248-nm deep-ultraviolet stepper was used for the lithography. A photoresist reflow course of and an optimized dry etch course of have been utilized in waveguide patterning to attenuate waveguide scattering loss. Extra fabrication particulars could be present in refs. 52,53. The SiPh PIC, together with its Si modulators and Si–Ge PDs, was fabricated on a 200-mm SOI wafer with a Si-layer thickness of 220 nm and a buried oxide layer thickness of two μm utilizing CMOS-compatible processes at CompoundTek Pte in a one-to-one 200-mm-wafer run with its commonplace 90-nm lithography SOI course of. The waveguide loss on this SiPh platform is roughly 1.2 dB cm−1 within the C band. In our experiment, lensed fibres with totally different mode discipline diameters have been chosen for the AlGaAsOI and SOI chips; the coupling loss is about 3–5 dB per aspect for AlGaAsOI waveguides and about 2–3 dB per aspect for Si waveguides.

Characterizations of the building-block items

The FSR of the 144-μm-radius rings utilized on this examine is about 90 GHz. The microcomb reveals advances each in start-up and stabilization. Through the dark-pulse era, a a lot smaller abrupt energy change happens when the comb transits from continuous-wave states to dark-pulse states, indicating the elimination of the well-known triggering drawback in brilliant soliton era. In contrast with basic brilliant solitons, the darkish pulse is inherently tolerant to thermal results that normally make brilliant soliton states troublesome to entry54. For long-term stability measurement, the comb spectra and comb line energy of a free-running dark-pulse comb are recorded by a high-resolution optical spectrum analyser (OSA) each 5 min.

Extra particulars are offered right here for the SiPh gadgets proven in Fig. 2. The opto-electrical BW of the depletion-mode Si-based MZMs was measured by a vector community analyser (Keysight N524), with the everyday outcomes of >30 GHz. The on-chip section compensation items are MZI-based titanium nitride (TiN) microheaters. The resistance is roughly 200 Ω. The TiN steel layer is about 1 μm above the Si layer, making certain a heating effectivity of about 20 mW π−1. In the meantime, a deep trench course of is utilized to isolate every microheater to decrease thermal cross-talk. For the on-chip true-time delay line, we adopted a 2-μm-wide multimode Si waveguide for low-loss transmission. Euler curves have been used within the spiral waveguide for adiabatic bending. For a 60-ps Si delay line, the full loss is <0.5 dB, with a delay-time variation of <3% amongst 8 examined gadgets. For the vertical epitaxial Ge PD, the responsivity declines with the rising on-chip energy. A saturated level of about 20 mW may very well be reached when the ability is additional elevated. Microring filters employed for WDM may very well be tuned by microheaters, with which a 180-GHz channel spacing could be obtained below 20-mW energy dissipation. The CMOS drivers for sign amplification earlier than injection into the Si MZM (not used within the high-bit-rate (>50 Gbps) sign transmission experiment) present a 3-dB achieve BW of about 24 GHz.

The efficiency of different building-block gadgets is offered in Prolonged Knowledge Fig. 1. The linewidth of the DFB laser used because the pump is measured by a delayed self-heterodyne technique55. The measurement and Lorentzian becoming outcome are proven in Prolonged Knowledge Fig. 1a, exhibiting a laser linewidth of about 150 kHz. For the SiPh gadgets, the 3-dB BW of the Si–Ge photodiodes is proven in Prolonged Knowledge Fig. 1b, indicating an roughly 30-GHz S21 parameter. Such a non-optimized BW accounts for the penalty in Fig. 3c. Construction design for a decrease resistor-capacitor time fixed might additional improve the operation BW. For on-chip monitoring, the uneven MMI-based 10:90 energy splitter56 is employed within the system, as proven in Prolonged Knowledge Fig. 1c. The symmetry of the multimode area is damaged by eradicating the nook of the MMI (marked with a pink dashed rectangle), which causes a dramatic redistribution of the optical discipline, thus resulting in an uneven energy splitting by altering the width of the cut-off nook. We randomly selected 4 equivalent 1:9 MMIs and examined the ability splitting ratios. The outcomes have been discovered to be near the design goal (dashed line), exhibiting good consistency, as proven within the backside panel of Prolonged Knowledge Fig. 1c. Furthermore, the grating couplers used on this work (Prolonged Knowledge Fig. 1d) present a roughly 2-dB coupling effectivity distinction throughout the operation band (1,535–1,565 nm).

Turnkey dark-pulse microcomb era

The turnkey microcomb era check setup is proven in Prolonged Knowledge Fig. 2a, with both an ECL or a DFB laser because the pump. Sluggish laser-frequency detuning is sufficient for microcomb era, which could be realized by adjusting the cavity size through tuning the lead zirconate titanate voltage of the business ECL or altering the laser present of DFB, respectively. After the comb era, the spectra are recorded; in the meantime, the full energy of the generated comb strains is measured in actual time. A pre-calibration course of is required to make sure the laser frequency will find on the comb accessing vary in the end. For the ECL-pumped dark-pulse comb (Prolonged Knowledge Fig. 2b), a 1-Hz sq. wave is used because the set off sign, which tunes the pump wavelength about 0.3 nm away from or into the resonance. For the DFB-pumped experiments (Prolonged Knowledge Fig. 2c), when a laser is turned on, there may be all the time an automated frequency ramping-up course of owing to the injected provider and the warming of the cavity, which may instantly provoke the microcomb era so long as the lasing frequency of the ultimate steady state lies throughout the vary of the entry window of the coherent state. In our experiment, the laser present is switched between two values with a interval of 6 s (1 s for the ‘off’ state and 5 s for the ‘on’ state). Each outcomes present quick on–off behaviours of microcomb era together with the low-speed management sign. It’s famous that there’s some energy ripple of the DFB-pumped comb within the first few seconds, which is as a result of temperature vibration brought on by thermoelectric cooler, after which the comb state is stabilized. The comb is reproducible in a number of consecutive switching assessments, with nice robustness.

Particulars of knowledge transmission experiments

In our experiment, the microcomb is first pumped by a business tunable laser (Toptica CTL 1550), then by a DFB laser chip for the next diploma of integration, the place an optical isolator is deployed between the DFB laser and the AlGaAsOI microresonator to remove the reflection. When tuning the pumping wavelength from the blue aspect to a sure detuned worth at round 1,552.5 nm, each configurations generate darkish pulses with 2-FSR comb spacing. The detailed experimental setup for knowledge transmission is proven in Prolonged Knowledge Fig. 3a. For the comb spectrum with giant energy fluctuations, a further amplification course of is required owing to the inadequate achieve of these low-power channels, which introduces additional system complexity and energy consumption on the transmitting aspect. On this work, owing to the sturdy thermal impact, the averted mode-crossing (AMX) energy of the AlGaAs microresonator could be thermally pre-set to acquire a coherent microcomb with a much less disparate energy distribution throughout the operation band. Thus, solely a notch filter is required to attenuate the central three comb strains for the next equalized comb amplification. The comb is amplified by an EDFA after which break up into odd and even check bands39,57,58 by a wavelength-selective change (Finisar Waveshaper 4000s). A Si modulator and a lithium niobate (LN) modulator (EOspace, 35-GHz BW) are deployed on the odd and even bands, respectively. Ten comb strains in every check band are concurrently modulated. The modulators are pushed at a 32-Gbaud or 50-Gbaud image charge. The differential PAM-4 sign is generated by a business pulse sample generator (Anritsu PAM4 PPG MU196020A). The insertion lack of the SiPh (LN) modulator is 13(8) dB. The SiPh modulator undergoes a comparatively excessive loss (together with the sting coupling lack of about 2 dB per aspect), which ends up in an influence distinction between the 2 check bands. The modulated check bands are then mixed by a 50:50 energy coupler and launched into one other WSS for comb energy equalization. On the receiving aspect, every WDM channel encoded by the Si modulator is sequentially filtered out and measured. Eye diagrams are produced by a sampling oscilloscope (Anritsu MP 2110A) with a 13-tap transmitter and dispersion eye closure quaternary (TDECQ) equalizer (accumulation time, 8 s). The BERs are measured on-line by an error detector (Anritsu PAM4 ED MU196040B) with 1-dB low-frequency equalization and a decision-feedback equalization. Prolonged Knowledge Fig. 3b reveals the 100-Gbps PAM4 eye diagrams for every of the 20 channels.

It’s value noting that the efficiency is underestimated. In our proof-of-concept check configuration, ten channels in every check band are modulated on the similar time. Contemplating two-photon absorption in Si waveguides, the utmost enter energy for the Si modulator is about 13 dBm, which ends up in solely 3-dBm optical energy per single lane. Furthermore, contemplating the additional penalty launched by the WSS for energy equalization, pointless in real-word transmission eventualities, the OSNR for every channel could be at the very least 10 dB greater. Thus, a greater transmission result’s attainable.

Noise evaluation of various pump schemes

The noise flooring of the DFB and the ECL are roughly characterised in an OSA, as proven in Prolonged Knowledge Fig. 4a. The laser spectra point out that the noise of the DFB is evidently greater than that of the ECL. The combs in our experiments are pumped by the free-running DFB laser and the ECL individually, as proven in Prolonged Knowledge Fig. 4b, c. With the virtually similar pumping energy of about 10 mW, the DFB chip holds a 10-dB-higher noise flooring in contrast with the ECL, comparable to an equal OSNR discount in every comb line. Furthermore, the amplification after the comb era would additionally lead to OSNR degradation, which may very well be a possible drawback when changing the present EDFA with built-in SOAs (about 4–5-dB-noise-floor increment in a business EDFA and about 7 dB in business on-chip SOAs).The OSNR of the DFB-pumped microcomb could be additional improved by using an on-chip optical filter for comb distillation59,60 or introducing optical injection locking between the microcomb and slave lasers for low-noise amplification61. Additionally, rising the pump energy will result in the next common OSNR and extra steady long-term behaviour, which is a bonus over the injection-locking-based dark-pulse era21,62.

Setup of the dispersive delay-line MPF scheme

Because the non-uniformity of delays owing to the inevitable fabrication errors will degrade the filtering efficiency, the second TDL-MPF method can be carried out to additional decide the optimum filtering efficiency: a spool of single-mode fibre (SMF) is used as an alternative of the on-chip spiral delay strains to supply dispersive delay. Prolonged Knowledge Fig. 5 reveals the experimental setup of the reconfigurable MPF carried out in a dispersive delay-line configuration. In contrast with Fig. 4a, a lot of the MPF system stays unchanged and has one major distinction, which is that the on-chip true-time spiral delay strains are faraway from the SiPh sign processor. The processed comb strains will propagate by means of a spool of 5-km SMF (as a dispersive component) to acquire a stable delay unit between adjoining faucets, which could be expressed as T= δλDL (ignoring the high-order dispersion of SMF), the place δλ represents the comb line spacing, D is the dispersion coefficient of SMF and the L is the size of SMF. On this scheme, the essential delay T amongst comb strains is generated by a single dispersive component, which could be stored as uniform worth and never influenced by fabrication errors. Apart from, this method is extra versatile; as an example, the centre frequency of the filtering passband could be adjusted by merely change the size or dispersion coefficient of SMF.

Particulars of RF filter experiments

The DFB-driven dark-pulse Kerr comb displays 2-FSR (180-GHz) comb spacing. The preliminary comb supply is amplified by an EDFA, and eight comb strains within the vary of 1,547–1,560 nm are chosen utilizing an optical bandpass filter earlier than injection right into a SiPh sign processor chip. The enter and output coupling are achieved through grating couplers of about 40% coupling effectivity. Frequency-swept RF indicators with 9-dBm energy from a vector community analyzer are utilized to the Si MZM in double-sideband format. The faucet weighting coefficients are set by adjusting the relative detuning among the many comb strains and their corresponding resonance wavelengths within the Si MRA with TiN microheaters positioned on the waveguides. The output mild of the Si chip is break up by a ten:90 optical energy coupler: 10% of the sunshine is shipped into an optical spectrum analyser (Yokogawa AQ6370C) for spectral monitoring, whereas the opposite 90% of the sunshine propagates by means of the follow-up optical hyperlink. Within the dispersive delay scheme, a spool of 5-km SMF is used to accumulate the dispersive delay between adjoining comb strains (faucets). Lastly, the processed comb strains are beat in a 50-GHz PD (Finisar 2150R) to transform the optical sign into electrical area. A low-noise EDFA is positioned earlier than the PD to compensate for the hyperlink insertion loss and coupling loss.

For the sensible demonstrations of RF sign filtering, a 50 Gsamples s−1 arbitrary waveform generator (AWG, Tektronix AWG70001) is used to supply the specified RF enter indicators. To validate the BW reconfigurability of this filter, an ultrawideband RF sign is generated, spanning from 5.5 GHz to 9 GHz. To validate the FSR reconfigurability of this filter, a fancy RF sign is produced that comprises a 50-Mb-s−1 QPSK spectrum modulated at 3.6 GHz and a 50-Mb-s−1 QPSK spectrum modulated at 7.2 GHz. The RF outputs from the AWG are amplified by a linear electrical driver (SHF 807C) earlier than routing to the Si MZM. The filtered RF indicators are detected by a sign analyser (Keysight N9010B) for spectrum measurement. An identical FSR multiplication of the MPF has been reported beforehand and defined by temporal Talbot results63. Nonetheless, the essential Talbot processor utilized in these MPF techniques relies on extra complicated discrete gadgets, which is able to improve the ability dissipation and make the system much less steady.

In contrast to the standard waveshaper based mostly on cumbersome liquid-crystal spatial mild modulators64, one of many outstanding benefits of the chip-scale add-drop microring resonator (MRR) array utilized in our work is the fast reconfiguration of RF filtering responses. The reconfiguration operation on filtering spectra is realized by adjusting the shaping profiles of comb strains, by means of the TiN microheater positioned on the waveguides. To discover the utmost reconfiguration velocity, a typical electrical square-wave waveform is generated by a perform waveform generator (RIGOL, DG2102) to drive a single MRR channel. The output of the MRR is acquired by a photodetector (Thorlabs DET08CFC/M), after which recorded by a digital oscilloscope (RIGOL, DS7014 10 GSa s−1). Prolonged Knowledge Fig. 6 reveals the measured switching temporal response. As seen in Prolonged Knowledge Fig. 6b, c, the 90/10 rise and fall occasions are 15 μs and 53 μs, respectively. Due to this fact, the quickest response velocity for the reconfiguration operation is roughly 19 kHz.

Leave a Reply