Integrated Acousto-Optic Frequency Beamsplitters

• Integrated acousto-optic frequency beamsplitters are chip-scale devices that use RF-driven acoustic waves for frequency-selective light manipulation.
• They combine piezoelectric transducers with optical waveguides on platforms like TFLN, SOI, and AlScN to achieve beam routing, multi-tone modulation, and quantum operations.
• Performance metrics reveal scalability with high beam counts, low insertion loss, and potential for advanced free-space communication, computing, and quantum applications.

Integrated acousto-optic frequency beamsplitters are non-mechanical, chip-scale devices that exploit the interaction between guided optical modes and electrically synthesized acoustic waves to realize frequency- and angle-selective splitting, routing, and modulation of light. This class encompasses both traveling-wave and cavity-based acousto-optic architectures, enabling parallel beam steering, frequency-domain linear operations, and scalable quantum photonic processing in the frequency bin domain. The integration of piezoelectric transducers and acousto-optic waveguide structures on platforms such as thin-film lithium niobate (TFLN), silicon-on-insulator (SOI), and hybrid silicon-AlScN underpins the rapid evolution of this field.

1. Physical Principles of Integrated Acousto-Optic Frequency Beamsplitting

Acousto-optic frequency beamsplitting in integrated photonics leverages Bragg diffraction of guided light by a radio-frequency (RF)-excited, coherently propagating surface acoustic wave (SAW) that forms a temporally and spatially periodic refractive index grating. The fundamental phase-matching relation for first-order (m = ±1) Bragg diffraction in a waveguide is

$k_0 ± k_a = k_m$

where $k_0 = 2\pi n_{\mathrm{eff}}/\lambda$ is the optical wavevector and $k_a = 2\pi f_a / v_a$ is the acoustic wavevector (with $f_a$ the acoustic frequency and $v_a$ the SAW velocity). The resultant free-space beam deflection angle is determined by

$\theta_m \approx m \cdot \frac{\lambda f_a}{v_a}$

for small deflection angles and $n_{\mathrm{eff}} \rightarrow 1$ upon outcoupling. Each acoustic frequency $f_a$ thus uniquely maps to a diffraction angle and corresponding frequency shift of the optical output. In quantum and cavity-based systems, sideband-resolved phase modulation by the acoustic wave induces a synthetic frequency “lattice,” with Hamiltonian-mediated coupling between frequency sites defined by acousto-optic interaction strengths (Lin et al., 2024, Zhao et al., 2021).

The diffraction efficiency is set by

$\eta = \sin^2(\kappa L)$

where $\kappa$ is the coupling coefficient (proportional to the SAW amplitude) and $k_0 = 2\pi n_{\mathrm{eff}}/\lambda$0 is the interaction length. The effective $k_0 = 2\pi n_{\mathrm{eff}}/\lambda$1 includes dependencies on refractive index $k_0 = 2\pi n_{\mathrm{eff}}/\lambda$2, elasto-optic coefficient $k_0 = 2\pi n_{\mathrm{eff}}/\lambda$3, drive power $k_0 = 2\pi n_{\mathrm{eff}}/\lambda$4, acoustic velocity $k_0 = 2\pi n_{\mathrm{eff}}/\lambda$5, and the acousto-optic figure-of-merit $k_0 = 2\pi n_{\mathrm{eff}}/\lambda$6.

2. Device Architectures and Materials Platforms

Integrated acousto-optic frequency beamsplitters are implemented in several material systems:

• Thin-Film Lithium Niobate (TFLN): Structures utilize X-cut TFLN on SiO$k_0 = 2\pi n_{\mathrm{eff}}/\lambda$7/Si substrates, enabling high piezoelectric coupling for efficient SAW excitation and low acoustic attenuation. Typical devices define single-mode ridge waveguides ($k_0 = 2\pi n_{\mathrm{eff}}/\lambda$8 nm) expanding to $k_0 = 2\pi n_{\mathrm{eff}}/\lambda$9 μm slab apertures, with waveguide orientation perpendicular to the SAW propagation to optimize interaction (Lin et al., 2024, Fang et al., 5 May 2026).
• Silicon Photonics with AlScN Piezoelectrics: Foundry-compatible SOI waveguides (e.g., $k_a = 2\pi f_a / v_a$0 nm $k_a = 2\pi f_a / v_a$1 $k_a = 2\pi f_a / v_a$2 nm) are coupled to AlScN interdigitated transducers (IDTs), enabling high-yield acousto-optic integration with standard silicon photonics platforms. AlScN provides 340 nm thick piezoelectric films for RF-to-acoustic conversion (Erdil et al., 2024).
• Cavity-Based SOI with Aluminum Nitride (AlN): Nanophotonic crystal cavities in suspended silicon combine with AlN IDTs to create high-Q optical and mechanical resonators supporting strongly enhanced optomechanical coupling (Zhao et al., 2021).
• Etchless LN Bound-State-in-Continuum (BIC) Waveguides: Polymer-patterned waveguides atop unetched TFLN provide strong field overlap, enabling GHz-frequency shifting and clean single-sideband operation (Yu et al., 2020).

Typical IDT geometries involve linearly chirped electrodes for broadband acoustic excitation, with aperture widths matched to the optical slab ($k_a = 2\pi f_a / v_a$3 μm), and finger pitches selected to resonate at the desired acoustic frequency range (e.g., 1–2 GHz). Acoustic aperture, waveguide slab length, and IDT configuration co-determine the device’s angular and frequency resolution, efficiency, and bandwidth.

3. Multi-Tone and Multi-Order Operation

Acousto-optic frequency beamsplitters can be driven by digitally synthesized, multi-tone RF signals, producing a superposition of SAWs and thus forming multiple, individually controllable optical beams. The device bandwidth $k_a = 2\pi f_a / v_a$4, set by IDT geometry and material response, determines the maximum number of resolvable beams per channel,

$k_a = 2\pi f_a / v_a$5

In (Lin et al., 2024), a $k_a = 2\pi f_a / v_a$6 MHz bandwidth and $k_a = 2\pi f_a / v_a$7 μm interaction length yield $k_a = 2\pi f_a / v_a$8 beams per channel; $k_a = 2\pi f_a / v_a$9 beams were experimentally realized. Each beam may be independently amplitude- and phase-modulated, supporting MIMO communication, per-beam data encoding (e.g., OOK, QAM), or quantum channelization.

For higher optical orders (sidebands of $f_a$0 with $f_a$1), driving the device with high RF power increases the phase modulation index $f_a$2, so that sideband amplitudes become $f_a$3, where $f_a$4 is the $f_a$5th Bessel function. This enables multi-line frequency comb generation, multiplexed communication channels, or high-dimensional quantum frequency-bin operations (Erdil et al., 2024, Zhao et al., 2021, Lukens et al., 11 Jan 2026).

4. Performance Metrics, Trade-Offs, and Scalability

Key metrics for integrated acousto-optic frequency beamsplitters include:

Metric	Representative Value	Determinants
Diffraction efficiency	3.3% (LN, single tone)	$f_a$6, RF power, IDT design
Acoustic bandwidth	450 MHz (LN)	IDT chirp, material loss
Spot count (per channel)	54 (LN, 373 μm)	$f_a$7, $f_a$8, $f_a$9
Insertion loss	$v_a$05 dB–20 dB	Coupling, waveguide, diffraction efficiency
On-off extinction	$v_a$1 dB (16 beams, LN)	Beam separation, crosstalk
Sideband conversion	Up to 0.5 (AlScN/Si, at $v_a$2 dB drive)	RF drive, overlap integral
Success in quantum regime	$v_a$3 (FRODO, $v_a$4)	Phase-matching, interaction length, loss

A fixed total RF power $v_a$5 is divided among $v_a$6 tones: increasing $v_a$7 reduces per-tone efficiency as $v_a$8. There is a trade-off between number of beams and per-beam efficiency (Lin et al., 2024). Crosstalk between adjacent beams is minimized by choosing $v_a$9 well above the angular beam divergence-limited threshold, with measured contrasts up to $\theta_m \approx m \cdot \frac{\lambda f_a}{v_a}$0 dB for adequately spaced channels.

Scalability is achieved by integrating multiple AO channels in parallel: with $\theta_m \approx m \cdot \frac{\lambda f_a}{v_a}$1 channels and $\theta_m \approx m \cdot \frac{\lambda f_a}{v_a}$2 beams/channel, chip-scale spot counts approach $\theta_m \approx m \cdot \frac{\lambda f_a}{v_a}$3 from a device area of a few $\theta_m \approx m \cdot \frac{\lambda f_a}{v_a}$4. Advanced device architectures—for example, utilizing AlN passivation or further reduction of acoustic loss—are projected to enable hundreds of beams per channel and multi-kHz reconfiguration rates.

5. Functional Implementation and Application Domains

Integrated acousto-optic frequency beamsplitters employ various architectural choices for different applications:

• Parallel free-space communication and beamsteering: Devices on TFLN or SOI modulate multiple beams for high-throughput free-space optical communication, MIMO links, or parallel quantum-dot/ion addressing (Lin et al., 2024, Fang et al., 5 May 2026).
• 2D beamscanning: Combining acousto-optic azimuthal control with wavelength-dispersive gratings (for polar coverage) enables two-dimensional, electronically reconfigurable beam steering in a compact footprint. Each comb line is assigned a polar angle, and acoustic frequency sets the azimuth, supporting field-of-view coverage up to $\theta_m \approx m \cdot \frac{\lambda f_a}{v_a}$5 (Fang et al., 5 May 2026).
• Frequency-domain optical computing: In synthetic-frequency dimensions, AO-modulated high-Q cavities enable programmable, phase-coherent frequency conversion and full-rank matrix–vector multiplication over tens of frequency bins, supporting optical neural networks and scalable classical data processing (Zhao et al., 2021).
• Quantum frequency processing: Inter-modal Brillouin-scattering-based AO gates (“FRODOs”, Editor’s term) implement analytically decomposable two-tone frequency beamsplitters with tunable phase ($\theta_m \approx m \cdot \frac{\lambda f_a}{v_a}$6) and mixing angle ($\theta_m \approx m \cdot \frac{\lambda f_a}{v_a}$7), supporting universal quantum operations, discrete Fourier transforms, and high-fidelity gate operation ($\theta_m \approx m \cdot \frac{\lambda f_a}{v_a}$8) (Lukens et al., 11 Jan 2026).
• Microwave photonics and signal processing: Devices with single-sideband extinction up to $\theta_m \approx m \cdot \frac{\lambda f_a}{v_a}$9 dB, GHz-scale acoustic modulation, and amplitude–frequency modulation on-chip expand application to non-reciprocal circuits, true-time-delay lines, and frequency up/down-conversion (Yu et al., 2020, Erdil et al., 2024).

6. Design Guidelines and Optimization Considerations

Effective design of integrated acousto-optic frequency beamsplitters must balance factors including IDT geometry (chirped vs. fixed pitch, finger number, aperture), interaction length ($n_{\mathrm{eff}} \rightarrow 1$0), acoustic–optical mode overlap, propagation loss, and RF drive considerations.

• IDT design sets the bandwidth and frequency resolution. Chirped IDTs extend $n_{\mathrm{eff}} \rightarrow 1$1 for higher spot counts, while matching acoustic aperture to optical mode maximizes $n_{\mathrm{eff}} \rightarrow 1$2.
• Optimization of per-tone RF power, channel multiplexing, and substrate processing yields high extinction, crosstalk suppression, and large spot densities.
• Integrated on-chip filtering and routing architectures (MZIs, ring filters, engineered gratings) separate frequency-shifted outputs.
• Apodization, device packaging, and material improvements drive advances in efficiency, bandwidth, and thermal/radiation robustness (Lin et al., 2024, Erdil et al., 2024, Fang et al., 5 May 2026).

Future directions involve resonance-enhanced AO structures, broadband IDT designs (e.g., split-finger or suspended structures), higher-Q optical cavities, and on-chip integration of dispersive elements for monolithic 2D beamsteering and dense frequency multiplexing.

7. Summary Table: Representative Devices and Capabilities

Architecture	Material	Key Metric	Value/Result	Reference
Multi-tone slab AOBS	LN on SiO$n_{\mathrm{eff}} \rightarrow 1$3	Beams/channel	21 (measured), 54 (max)	(Lin et al., 2024)
Foundry SOI–AlScN AOM	Si, AlScN	BW/$n_{\mathrm{eff}} \rightarrow 1$4	100 MHz/(V·cm)	(Erdil et al., 2024)
Cavity synthetic lattice	Si, AlN	Frequency sites	50 (over 40 GHz)	(Zhao et al., 2021)
FRODO intermodal scattering	Si, SiN, AlN	Q. gate fidelity	$n_{\mathrm{eff}} \rightarrow 1$5 (N=10)	(Lukens et al., 11 Jan 2026)
Etchless LN BIC SSB mod.	LiNbO$n_{\mathrm{eff}} \rightarrow 1$6, PMMA	Extinction ratio	>44 dB (SSB); 2.8%/W	(Yu et al., 2020)
2D AOBS + comb + gratings	LN on SiO$n_{\mathrm{eff}} \rightarrow 1$7	FOV	$n_{\mathrm{eff}} \rightarrow 1$8	(Fang et al., 5 May 2026)

8. Outlook and Scalability Prospects

Integrated acousto-optic frequency beamsplitters have demonstrated high channel counts, MHz–GHz bandwidths, field-programmable angular resolution, and compatibility with industrial foundries. Integration density currently surpasses bulk AODs by an order of magnitude (e.g., 579 spots/mm² vs. 64 spots/mm² (Lin et al., 2024)). Prospects for future scaling include extension to >100 beams/channel, robust multiplexed operation across the C-band and visible, and universal quantum operations in the frequency bin domain (Lukens et al., 11 Jan 2026).

The versatility and programmability of these beamsplitters, coupled with scalable photonic integration, position them as core building blocks for next-generation optical communication, computation, and quantum information systems.