Optical Forces Enable Low-Power Silicon Phase Shifting

Programmable silicon photonics still relies mostly on electronics to control light, even when the end application is optical computing, sensing, or quantum photonics. Thermo-optic and electro-optic phase shifters are mature, but they bring local wiring, heat, carrier absorption, or material-index trade-offs. This paper demonstrates a different route: a guided pump beam mechanically deforms a suspended slot-mode waveguide, changing the optical path seen by a co-propagating signal.

Core Contribution

The central claim is that gradient optical forces can provide a broadband all-optical phase shifter in a standard silicon photonic platform. The authors are not reporting a new material nonlinearity. They use geometry: optical power in a narrow slot pulls suspended silicon rails together, shifting the effective index of the guided mode. Because the phase shift comes from mechanical displacement rather than electronic carrier injection or resonant index tuning, the mechanism avoids local electrical contacts at the active element and does not rely on a narrow optical resonance.

The headline result is a π phase shift with 60 μW of on-chip pump power in a 178.6 μm-long device. That is the clearest experimental payoff in the paper: low pump power at a length compatible with integrated photonic circuits.

Technical Approach

The device is a subwavelength-grating slot waveguide, or SWGSW. The slot concentrates the optical field where the gradient force is strongest, while the grating tethers provide mechanical support without simply turning the suspended structure into a high-loss perturbation. Figure 1 shows the physical layout, the slot-mode field concentration, and the force-equilibrium picture that connects pump power, elastic restoring force, and phase shift.

The model is deliberately compact. Optical force per unit length is computed as a function of slot width, and the mechanical response is treated through a lumped stiffness $k$. The operating point is determined by intersections between the optical force and the elastic restoring force; stable and unstable equilibria matter because the same geometry can move from smooth phase tuning into pull-in-like behavior as pump power rises. The simulations in Figure 1 also show why geometry is the main design knob: changing slot width and tether dimensions changes both optical confinement and mechanical compliance.

The experimental readout uses interferometric structures rather than relying only on isolated transmission spectra. In pump-probe measurements, a pump beam actuates the suspended SWGSW and a signal beam reads out the phase shift through an unbalanced Mach-Zehnder interferometer. Figure 4 captures that system-level test: wavelength-dependent probe transmission shifts with pump power, and the authors also measure mechanical response spectra and pump-pulse transients.

Results and Analysis

The abstract result is backed by a broader parameter sweep: the authors report measured pump powers across a range of designs and quantitative agreement with the lumped force-equilibrium model. That matters because this mechanism is sensitive to stiffness, slot width, and fabrication tolerances; a single π-shift demonstration would be less convincing without the geometry-dependent comparison.

The supplementary loss analysis gives the main practical caveat. A cutback experiment on complete phase shifters with $s=100$ nm, $L=400$ μm, and $k=0.564$ N/m extracts an insertion loss of 1.95 ± 0.22 dB per phase shifter at 1550 nm. The authors report approximately 2 dB total insertion loss over a 30 nm bandwidth. That is usable for laboratory demonstrations, but it is not yet a negligible cost for large programmable meshes where phase shifters are cascaded many times.

The loss budget also shows where improvement is plausible. The V-groove rectangular-to-slot coupler measurement gives 0.37 ± 0.01 dB per coupler at 1550 nm, and the authors estimate that replacing it with an initial-gap design at 150 nm could reduce per-coupler loss by about 0.27 dB, improving the device by roughly 0.5 dB. Separate propagation-loss measurements indicate that reducing tether width from 30 nm to 20 nm can lower loss by 5–10 dB/cm for fixed $s_0=100$ nm, and that $s_0<60$ nm with $d=20$ nm could make propagation loss contribute less than 0.2 dB for a 150 μm device.

The evidence supports the paper's narrower claim: optomechanical gradient forces can produce low-power, broadband all-optical phase shifting on silicon. The stronger system claim remains conditional. Mechanical actuation introduces speed limits and stability constraints that electronic phase shifters do not face in the same way, and the present loss is still material for scaled circuits. The most plausible early beneficiaries are cryogenic and optically controlled photonic systems where avoiding local wiring and carrier-based absorption is worth accepting a mechanically mediated response.