Noise-resilient nanophotonic gyroscope with sub-prad phase resolution

Abstract

Optical gyroscopes based on the Sagnac effect are the cornerstone of precision orientation and navigation. However, their bulky form factors prevent deployment in emerging mobile and autonomous systems. On nanophotonic platforms, the Sagnac signal plummets under aggressive miniaturization. Consequently, the signal is easily swamped by refractive-index fluctuations, rendering navigation-grade sensitivity within just a few square millimeters a notoriously elusive goal. Here, we demonstrate a noise-resilient nanophotonic optical gyroscope by exploiting a two-chain decoupling architecture to effectively isolate the rotation signal from channel noise. Implemented on a 3 mm^2 passive silicon nitride chip, the proof-of-concept device achieves a bias instability of 1.42 deg/h and an angle random walk of 0.001 deg/\sqrt{h}, representing improvements of 4 and 6 orders of magnitude, respectively, over the representative nanophotonic gyroscope of similar footprint (ref. 27). In the broader context of integrated optical gyroscopes, our approach bridges the long-standing size-performance gap by two to three orders of magnitude, moving chip-scale devices into a previously inaccessible regime and pointing toward navigation-relevant precision for monolithic microsystems. This architecture further enables sub-prad phase resolution with general applicability, establishing a foundational framework for the next generation of robust, monolithically integrated photonic sensing systems.

Figures (7)

• Figure 1: Architecture and readout principle of the noise-decoupled nanophotonic gyroscope. (a) Schematic of the proposed architecture. The silicon nitride (Si$_3$N$_4$) waveguides are in blue. The phase modulators are in yellow; the arm modulator provides the symmetry-breaking bias, while the resonator pair is tuned to align their initial resonance condition. (d) Conceptual comparison between conventional frequency-tracking readout and the proposed intensity-based readout. In the presented architecture, rotation information is mapped onto intensity variations, fundamentally relaxing the requirements on frequency tracking and source coherence. (c) Photographs of the fabricated chip after dicing, including a macroscopic image with a fingertip for intuitive size reference and a zoomed-in view showing the on-chip photonic structures.
• Figure 2: The system achieves dynamic response and high sensitivity through weak-measurement-based readout. Representative 100s time-domain signals recorded without WVA (a) and with WVA (b) under a sinusoidal rotation ($0.1~\mathrm{Hz}$, $2.51^\circ/h$ amplitude). Orange dashed lines indicate the applied angular velocity. (c) Frequency-domain spectra corresponding to (a) and (b). The spectral component at 0.1 Hz represents the driven rotation signal, while the remaining components correspond to noise. The two dashed lines indicate the broadband noise floor levels. (d) The normalized FFT-derived voltage amplitude at the signal frequency is plotted as a function of the corresponding angular velocity. The dashed line indicates a linear fit. The slope enhancement factor (with WVA / without WVA) is 10.6.
• Figure 3: Static performance and noise characterization. Allan deviation of the gyroscope output measured under static conditions. The bias instability (BI) and angular random walk (ARW) are explicitly marked for each configuration. The black dashed line denotes the predicted thermo-refractive noise (TRN) limit, which closely matches the short-term Allan deviation obtained with weak-measurement readout.
• Figure 4: Performance benchmarking versus effective Sagnac area. (a) Comparison of angular random walk (ARW). (b) Comparison of bias instability (BI). The Sagnac enclosed area accounts for the accumulated rotation-sensitive optical path and enables a consistent comparison across different gyroscope implementations.
• Figure 5: Extended Data Fig. 1. Experimental intensity-based spectral response to different non-reciprocal phase bias. A sequence of small phase biases is applied by the phase modulator on one ring, producing a non-reciprocal phase bias between the two arms of this architecture, and the output spectra are measured at the (a) bar and (b) cross ports. Curves in different colors correspond to different applied phase biases. Unlike conventional resonator readout, where a phase perturbation mainly manifests as a resonance frequency shift, the present architecture maps non-reciprocal phase predominantly into intensity change of the resonance peak.