By combining an ultralow-loss silicon nitride reference cavity with a diode laser, the interrogation of a strontium-ion optical clock is possible with excellent accuracy. The development is a step towards miniature, integrated optical clocks.
Arthur Schawlow, the 1981 Nobel Prize laureate in physics, once advised to “never measure anything but frequency”. Presently, frequency (or time) has become the physical quantity that can be measured with the highest precision. The latest, state-of-the-art optical clocks now provide a fractional frequency uncertainty1 of 10−19 and are redefining the SI international standard of the unit of the time, the second. As a result, measurements of other quantities are often translated into frequency measurements to attain a significantly improved accuracy. Although the current SI second has been defined by microwave caesium fountain clocks, optical clocks — which count the oscillation frequency of optical transitions of atomic ensembles or a single ion in the visible to ultraviolet spectrum — have today outperformed their microwave counterparts by orders of magnitude.
Optical clocks have facilitated the discovery of new fundamental physics and have revolutionized today’s positioning, navigation and geodesy. Parallel to the pursuit of even higher clock precisions, with major advances in integrated photonics in the last decade, current research interest has also been placed in the realization of photonic-chip-based atomic and optical clocks2. Although transportable optical clocks with 10−18 frequency precision have been achieved3, chip-based clocks promise further reductions in size, weight, and power consumption, enabling the ubiquitous deployment of state-of-the-art frequency metrology. In particular, with the blooming space industry, chip-based clocks are scheduled to be installed on thousands of satellites in low-earth orbit (LEO), to enhance the capabilities of Global Navigation Satellite Systems (GNSS).
Now, writing in Nature Photonics4, William Loh and colleagues have leveraged integrated photonics to demonstrate a compact, stable, narrow-linewidth (16.7 Hz) laser system that operates at a wavelength of 1,348 nm. The laser, upon frequency-doubling to 674 nm, can directly interrogate a 88Sr+ optical clock transition (Fig. 1). The clock shows short-term fractional frequency instability averaged down to 3.9 × 10−14/√τ (where τ is the averaging time), which surpasses the performance of state-of-the-art microwave clocks.
Fig. 1: Schematic of the narrow-linewidth diode laser system for clock interrogation.
An external-cavity diode laser operating at 1,348 nm is PDH-locked to a photonic-chip-based, metre-long, ultrahigh-Q silicon nitride resonator. The PDH lock endows the 1,348-nm laser with a 16.7-Hz linewidth. Upon frequency doubling to 674 nm, the laser can directly interrogate a 88Sr+ optical clock transition. The clock has a short-term fractional frequency instability of 3.9 ×10−14/√τ (τ is the averaging time), outperforming the state-of-the-art microwave clocks.
To build such a laser, Loh and colleagues used a chip-based resonator made of silicon nitride (Si3N4) with an ultrahigh Q factor exceeding 2 × 108 and a length of 6.1 m. Light was coupled into the chip via optical fibres, and then into the Si3N4 resonator via a bus waveguide.
The authors locked an external-cavity diode laser operating at 1,348 nm to this resonator using the Pound–Drever–Hall (PDH) technique. This requires the use of extra components, including an electro-optic modulator, a semiconductor optical amplifier (SOA) to compensate insertion loss among different components, a few photodetectors, as well as electronics such as a local RF oscillator and PID controllers. The measured fractional frequency instability of the locked 1,348-nm laser reached ?f/f = 7.5 × 10−14 with 30-ms integration time, corresponding to an integrated linewidth of 16.7 Hz, which is on a par with the best fibre lasers to date. The laser still suffers from long-term drift due to the temperature drift of the Si3N4 resonator, which has a thermo-optic coefficient on the order of 10−5/K.
The most critical part, and also the key novelty of the reported laser, is the ultrahigh-Q integrated Si3N4 resonator. In the past decade, ultralow-loss Si3N4 integrated photonics has made stunning progress. Already extensively used as etch hardmasks, passivation layers and stress layers in CMOS manufacturing of microelectronic circuits, amorphous Si3N4 has become the material of choice for photonic integrated circuits (PICs) of exceptionally low optical loss, which is orders of magnitude lower than its silicon counterpart. In fact, among all integrated material platforms developed so far, Si3N4 is the only one that has created PICs with linear optical loss down to 0.1 dB/m (ref. 5). In addition, Si3N4 offers a wide bandgap of 5 eV, and thus a transparency window from ultraviolet to mid-infrared, as well as weak Raman and Brillouin gains.
Usefully, the group-velocity dispersion (GVD) of Si3N4 PICs can be customized to be either positive or negative, making it possible to realize optical frequency combs, supercontinua, and parametric amplifiers6. As scattering loss has been identified as the dominant loss origin in Si3N4 PICs7, wide and thin waveguides (for example, width >5 μm and thickness <80 nm) can guide light with a large mode area (for example, 29 μm2 in this study) and thus weak interaction with waveguide surface roughness and defects. Consequently, the guided light experiences significantly reduced scattering loss.
When using such integrated resonators as reference cavities for laser frequency stabilization, not only the Q factor but also the resonator length matters. The fundamental limit of the resonator’s frequency stability is determined by its thermo-refractive noise (TRN), as small fluctuations of temperature cause resonance frequency jittering via the thermo-optic coefficient8. Longer integrated resonators suffer weaker TRN owing to the averaging effect. In fact, metre-long integrated resonators can enable laser frequency stabilization to sub-hertz intrinsic linewidth9.
The integrated Si3N4 resonator used by Loh and colleagues had a 40-nm waveguide height (to reduce the number of spatial eigenmodes and thus reduce mode mixing), a 15-μm waveguide width (to minimize optical loss), a minimum bending radius of 5 mm (for negligible radiation loss in bent sections), and a 40-μm waveguide spacing (to avoid optical coupling between two neighbouring sections); the resonator was also entirely buried in a 19-μm-thick SiO2 cladding (to avoid light leakage into the silicon substrate). As the resonator is patterned by a deep ultraviolet stepper lithography on a 200-mm-diameter wafer, the maximum device footprint is 26 × 32 mm2, limited by the exposure field size of the stepper lithography. Taking all these geometry parameters into account, and by carefully and smartly spiralling the resonator, the maximally allowed resonator length is 6.1 m, corresponding to a 33.7-MHz free spectral range (FSR). Note that when compared with common photonic chips, the chip here is enormously large; however, it is still much smaller than the standard Fabry–Pérot cavities of similar FSR. Practically, the resonator can be made even larger via field stitching10.
The locked 1,348-nm laser is further amplified by another SOA and frequency-doubled by a LiNbO3 waveguide, yielding a 107.5-µW output power at 674 nm. This 674-nm laser is used to interrogate the 5S1/2 ↔ 4D5/2 quadrupole transition in a 88Sr+ ion on a microfabricated surface-electrode trap. One major challenge is to overcome the laser’s frequency drift, which degrades the clock’s stability. To address this issue, Loh and colleagues have developed a procedure that quickly and successively locks the laser and holds the lock to the ion, using optical gain to fight against laser frequency drift. This procedure makes it possible to lock the laser to the 0.4-Hz-linewidth 88Sr+ ion transition.
Finally, to quantify the short-term fractional frequency instability of the optical ion clock driven by the novel 674-nm laser, the authors have performed a self-comparison between two interleaved clocks, each operated on the same hardware and ion, but without the knowledge of the other’s ion-stabilized frequency. This self-comparison measurement eliminates the common drift effects in the ion’s transition frequency, such as those induced by ambient fluctuations or ion micromotion. By dividing the measured instability by √2, the estimated short-term fractional frequency instability of a single clock is 3.9 × 10−14/√τ, where τ is the averaging time.
Although the demonstrated clock performance is still far below the best Sr optical clock to date, and the entire system is still bulky, the result shown here nevertheless marks a milestone on the long road to miniaturization of optical clocks using integrated photonics. Together with other parallel endeavours, such as building atomic vapour micro-cells11 and chip-scale lasers emitting specific wavelengths12, the ultimate goal is to make the entire clock system — including the optical, electronic and atomic components — on a monolithic substrate via heterogeneous and hybrid integration. These resulting optical clocks will be compact and portable, and will be manufactured in high volume and at low cost, making frequency metrology ubiquitous on mobile platforms and in space.
