Optical Phased Array
# Optical Phased Arrays in Silicon Photonics
An optical phased array (OPA) steers a beam of light electronically — with no moving parts — by controlling the relative phase of light emitted from a grid of closely spaced antenna elements on a chip. Each element radiates a weak, mostly-omnidirectional wave; when many elements emit coherently with a carefully chosen phase offset between neighbors, their waves interfere constructively in one direction and destructively almost everywhere else, producing a narrow beam that can be swept across an angle purely by changing phase settings — no gimbal, no micromirror, no mechanical actuator. This is the same principle long used in radio-frequency phased-array radar, now shrunk to the wavelength of light (~1550 nm) and implemented on a silicon chip using the same waveguide and modulator building blocks already established for WDM transceivers (id 974) and silicon modulators (free-carrier plasma dispersion, id 971).
The steering mechanism follows directly from basic interference: if N emitters are spaced a distance *d* apart and the phase of element *n* is set to *n*·Δφ (a linear ramp across the array), the combined far-field pattern peaks at an angle θ where sin(θ) = λΔφ/(2πd) — set the phase increment, and the beam angle follows. The phase shift at each element is implemented the same way silicon modulators already shift phase: either a resistive heater locally changing the waveguide's refractive index through the thermo-optic effect (slow, μs–ms response, but simple and low-loss), or a free-carrier-plasma-dispersion phase shifter (id 971) using carrier injection or depletion (fast, ns-scale response, but with extra optical loss from free-carrier absorption). Steering range and resolution both improve with more elements, but each additional element needs its own driver and routing, so practical silicon OPAs trade element count against control complexity the same way WDM trades channel count against per-channel circuit complexity (id 974).
The most compelling application driving OPA development is solid-state LiDAR: a beam-steering system with no mechanically rotating parts, small enough and reliable enough to put in a car, a robot, or a phone — something a spinning mechanical LiDAR unit (reliable but bulky, power-hungry, and a wear item) cannot match. A silicon OPA chip can, in principle, steer a LiDAR beam across a full field of view purely electronically, at update rates limited only by how fast the phase shifters can be re-driven. The tradeoff today is that OPAs with enough elements to achieve a narrow beam and wide field of view (which generally requires hundreds to thousands of emitters) are hard to fabricate with the phase uniformity needed — manufacturing variation across the chip means every element's phase response needs individual calibration, a problem that grows with array size exactly when you want more elements for a narrower beam.
| Steering method | Mechanism | Speed | Moving parts | Primary tradeoff |
|---|---|---|---|---|
| Mechanical gimbal | Physically rotate the source/mirror | ms-scale (inertia-limited) | Yes | Mature, reliable per-unit cost, but size/wear/vibration sensitivity |
| MEMS mirror | Small silicon mirror tilted electrostatically | μs–ms | Yes (micro-scale) | Faster than gimbal, still has mechanical fatigue limits |
| Optical phased array (thermal) | Heater-based phase shifters per element | μs–ms | No | Simple, low optical loss, but slow and power-hungry per element |
| Optical phased array (electro-optic) | Free-carrier plasma dispersion phase shifters | ns | No | Fast, but phase shifters add optical loss and need more drive complexity |
OPAs sit conceptually next to WDM (id 974) as a second axis of what coherent control over many parallel optical channels can buy a silicon photonic system — WDM parallelizes in wavelength to multiply data throughput, while an OPA parallelizes in phase across physical space to steer a beam without moving anything. Both techniques ultimately depend on the same underlying building block: a reliable, fast, low-loss way to control the phase of light in a silicon waveguide, which is exactly what the free-carrier-plasma-dispersion entry (id 971) and the laser-wavelength-locking entry (id 972) establish as fundamental and genuinely difficult silicon photonics problems.