Silicon nitride (Si3N4) is valued in integrated photonics for its low optical propagation loss, broad transparency range and ability to support nonlinear functions such as frequency-comb generation. Electrically tuning Si3N4 resonators quickly, however, is challenging because the material lacks the Pockels effect. A hybrid approach based on HBAR acousto-optic modulation offers another route: use piezoelectric actuation to generate acoustic waves that alter the optical properties of a resonator through stress.
In a demonstrated platform, aluminum nitride (AlN) actuators excite high-overtone bulk acoustic wave resonances (HBARs) in a stack that also contains Si3N4 waveguides and microring resonators. The work shows microwave-frequency modulation across multiple gigahertz and uses spatially and temporally controlled modulation to demonstrate a Si3N4 optical isolator. For engineers evaluating photonic integration, the key design idea is to separate the acoustic actuation path from the in-plane optical circuit.
Why fast tuning is difficult in silicon nitride
Si3N4 is useful for applications including microcombs, optical filters, gyroscopes and coherent communications. Many of these systems benefit from the ability to tune a resonator rapidly and precisely. For example, microcomb stabilization and modulation-based optical functions can require tuning at rates far beyond those provided by thermal control.
Thermo-optic tuning is established, but its relatively slow response, power demand and potential for thermal cross-talk make it unsuitable for some high-speed or densely integrated designs. Hybrid electro-optic approaches can add functionality, but may bring fabrication, optical-loss or material-integration trade-offs. The HBAR approach instead uses the stress-optic effect: mechanical stress changes the waveguide’s refractive index, which shifts the resonator’s optical response.
How HBAR acousto-optic modulation works
The device places a disk-shaped AlN piezoelectric actuator above a Si3N4 microring resonator. An applied vertical electric field deforms the AlN and launches acoustic waves into the underlying substrate. Reflections between the top and bottom surfaces create a Fabry–Pérot-like acoustic cavity. At resonant frequencies, the waves reinforce one another, producing standing-wave stress that changes the optical properties of the nearby waveguide.
The acoustic field travels vertically through the material stack, while the optical circuit remains in the plane of the chip. This geometry allows the waveguide to sit beneath a substantial oxide cladding layer, helping preserve low optical loss by keeping the optical mode away from the metal electrodes. Because the acoustic confinement is largely vertical and localized beneath each actuator, the actuator’s lateral shape does not need to define the acoustic wavelength in the same way as a surface-acoustic-wave design.
The reported structure uses a 1-micrometer AlN layer, with the Si3N4 waveguides buried in SiO2 cladding. Its acoustic resonances have an average free spectral range of about 17.5 MHz and extend across the measured microwave range, up to 6 GHz. The measured cross-talk between adjacent actuators is around -60 dB, a feature relevant to layouts that place multiple actuation points near one resonator.
Speed, tuning and design considerations
The platform demonstrated sub-nanosecond actuation response in time-domain measurements. It also produced a measurable optical response when driven with a 6 GHz square wave. That result demonstrates high-speed actuation, but it should not be confused with a flat broadband modulation response: acoustic resonances can shape or distort the signal. The researchers note that input conditioning or post-processing may be needed where a broadband digital response is required.
For lower-speed tuning, the same structure provided linear, bidirectional resonance tuning of approximately -0.2 picometers per volt. The reported tuning power efficiency was 5 nanowatts per picometer of resonant-wavelength shift. These figures describe the demonstrated device; they do not by themselves establish performance for other geometries, fabrication processes or operating conditions.
HBAR modes are set by the acoustic properties and thicknesses of the device stack. Coupling between the silicon substrate, oxide and AlN layers affects resonance spacing and response strength. Designers can therefore treat layer thickness and acoustic impedance as parameters for engineering the resonance spectrum. At the same time, the resonant nature of the mechanism means that frequency selection and the relationship between mechanical modes and optical linewidth need to be considered during system design.
Demonstrating a silicon nitride optical isolator
One application used multiple actuators around a single microring to create a circulating refractive-index modulation. Three actuators, positioned at equal intervals, were driven with phase offsets of 0, 120 and 240 degrees. This creates an effective rotating modulation pattern that couples two optical modes differently depending on the direction of light propagation.
The resulting Si3N4 isolator achieved 17 dB isolation at resonance, with a reported 1 GHz bandwidth. The isolation depends on the controlled interaction between the optical modes and the modulation pattern; it is not an intrinsic property of the passive ring. In the demonstrated arrangement, the drive frequency matched the separation of a selected pair of optical modes, and the spatial phase sequence established the direction of the modulation.
Practical implications for photonic system design
HBAR-based actuation adds a way to modulate Si3N4 resonators without relying on a direct Pockels response in the nitride. Its vertically confined acoustic field can support compact, localized actuation while retaining the option of deeply buried, low-loss optical waveguides. Potential uses identified for the platform include microcomb stabilization, microwave photonics, spatiotemporal modulation and microwave-to-optical conversion.
For industrial and system-level decision-makers, the work is best viewed as a device-platform demonstration rather than a drop-in production component. Relevant evaluation questions include the required modulation frequency, acceptable resonance-induced signal distortion, actuator drive conditions, optical loss and integration complexity. The demonstrated isolator and fast tuning show what the architecture can enable; applying it to a specific photonic system will require matching the acoustic and optical design to that system’s operating requirements.
Overall, combining AlN piezoelectric actuation with bulk acoustic resonances offers a distinct control mechanism for silicon nitride photonics. It brings microwave-rate stress-optic tuning and direction-dependent optical coupling into a platform designed to preserve the advantages of low-loss Si3N4 waveguides. The practical opportunity lies in using that combination where fast, localized modulation matters and where the acoustic resonances can be incorporated deliberately into the system design.