Attosecond-level synchronisation of chip-integrated oscillators

TL;DR

Attosecond-level synchronization across long fiber links for chip-integrated oscillators addresses the need for scalable precision timing in attosecond science. The authors demonstrate Kerr-nonlinear synchronization of two microcombs on separate silicon nitride chips by distributing two cw lasers over a fiber to define a reference and locking the microcombs via injection locking, with $f_ ext{rep} = | u_ ext{p}- u_ ext{inj}|/N$ and $ abla f_ ext{rep} = 0$. In the synchronized state, relative phase noise is suppressed by more than $80$ dB and the integrated relative timing jitter is below $< 400~ ext{as}$ over the range $1~ ext{kHz}$ to $1~ ext{MHz}$, achieved without active stabilization. This attosecond-level synchronization over $100$ m of fiber enables precision timing at scale for large facilities and emerging technologies such as disaggregated computing and quantum networks, and points toward chip-integrated attosecond photonics.

Abstract

Attosecond science provides a window to the fastest processes in chemistry, materials science, and biology. Accessing this time scale requires precisely synchronised oscillators. In free-electron X-ray lasers, which also provide sub-atomic resolution, synchronisation must be achieved across hundreds of meters. Current approaches to synchronisation based on mode-locked lasers deliver this level of performance but complexity, cost and size hinder their deployment in facility-wide multi-node networks. Here, we demonstrate attosecond-level synchronisation of two chip-integrated photonic oscillators (microcombs) separated by 100 m of fibre. A pair of continuous-wave lasers establishes a time reference that is delivered over fibre, and on-chip Kerr-nonlinear synchronisation results in an integrated relative timing jitter of the microcombs below 400 as (1 kHz to 1 MHz), without any active stabilisation. These results unlock precision timing at scale for large facilities and next-generation technologies such as disaggregated computing and quantum networks, and ultimately may lead to chip-integrated attosecond photonics.

Figures (3)

• Figure 1: Concept. a, Two laser frequencies ($\nu_\mathrm{p}$ and $\nu_\mathrm{inj}$) define an optical timing reference that is distributed over fibre to different nodes in a timing network, where synchronised microcombs provide local precision timing. b, Both laser frequencies $\nu_\mathrm{p}$ and $\nu_\mathrm{inj}$ are part of the microcombs spectrum, whose discrete spectral components are spaced by $f_\mathrm{rep}= |\nu_\mathrm{p}-\nu_\mathrm{inj}|/N$, where N is an integer. b, The two lasers with frequencies $\nu_\mathrm{p}$ and $\nu_\mathrm{inj}$ are combined and can then be distributed to $M$ different microcombs via a $2\times M$ fibre coupler. Both microcombs will emit pulses with the repetition rate $f_\mathrm{rep}$, however, uncorrelated noise in both systems will lead to timing jitter.
• Figure 2: Experimental approach.a, Schematic of the experimental setup. Two cw lasers (pump and injection) are combined on a 3 dB coupler and used to pump two identical photonic crystal resonators (PhCRs). An acousto-optic modulator (AOM) shifts the output of one comb by 80 MHz, after which both combs are recombined at a fibre coupler and amplified using an erbium-doped fibre amplifier (EDFA). A tunable bandpass filter (BPF) is used to select a single pair of comb lines (see inset). PD - photodiode, ESA - electrical spectrum analyser, FBG - fiber Bragg grating. b, Optical spectrum of two independent microcombs in the free-running (top, no injection) and synchronised (bottom, with injection) regimes. c, Single-sideband (SSB) phase-noise power spectral density (PSD) of the beatnotes between pairs of comb lines with index $\mu$, measured in the free-running and synchronised states. The measurement is limited by the phase-noise of the AOM, which is retrieved from the beating between shifted and unshifted pump (black) and the ESA noise floor (gray). d, Phase-noise (PN) PSD at 10 kHz analysis frequency, extracted from the data in c, in the free-running (left) and synchronised (right) states. PN is shown using linear scaling ($\mathrm{mrad^2/Hz}$).
• Figure 3: Phase noise and timing jitter of synchronised microcombs.a, Experimental setup with 50 m of spooled single-mode fibre added in both arms before the chips. b, (top) Relative repetition rate single-sideband (SSB) phase-noise power spectral density in the free-running (dark blue) and synchronised (red) states at 300 GHz carrier. The dashed line indicates an analytic estimatekondratiev2018ThermorefractiveNoiseWhispering of the fundamental thermorefractive noise (TRN) for the 300 GHz microresonator, and the gray trace shows the relative repetition rate noise rescaled to 10 GHz carrier. (bottom) The corresponding integrated residual timing jitter in the synchronised state.