Optical phased array (OPA)

An optical phased array applies the principle of a phased-array radar at optical wavelengths. Light from one laser is split, on chip, into $N$ waveguides, each passes through a phase shifter, and each ends in an emitter, usually a weak grating that radiates upward out of the chip. The emitters sit on a pitch $d$. In the far field their contributions add, and a linear phase step $\Delta\phi$ between neighbors tilts the direction in which they add in phase to

Changing the phase step steers the beam, with no moving parts and at the speed of the phase shifters.

The pitch decides how much of the field can be used. Because phase is only defined modulo $2\pi$, the array also radiates grating lobes, copies of the main beam separated by $\sin\theta = \lambda/d$ as in a diffraction grating. Only a pitch of $\lambda/2$ or less gives a single beam over the whole half-space. At 1550 nm that is 0.775 µm, closer than silicon waveguides can sit without coupling to each other, so practical arrays use a larger pitch and accept a limited field: for $d$ = 2 µm the lobe-free steering range is $\pm\sin^{-1}(\lambda/2d)$ = ±22.8°. The beam width is set by the aperture, about $0.886\,\lambda/(Nd\cos\theta)$; for 512 emitters at 2 µm this is 0.077°, so that field holds about 590 resolvable spots. Aperiodic emitter positions, which smear the grating lobes into a background rather than a copy of the beam, and waveguide designs that tolerate half-wavelength spacing are the two ways published arrays have widened the field.

A one-dimensional array steers in one plane only. The usual way to cover two dimensions exploits the grating emitters themselves, whose emission angle along the waveguide depends on wavelength: phase steers the beam across the array, and tuning a tunable laser steers it along. The alternative, a true two-dimensional array of emitters each with its own phase control, scales the number of phase shifters as $N^2$.

The engineering problems follow from the numbers. Every channel has a fabrication-dependent phase error, so an array must be calibrated, typically by optimizing the phases against a far-field camera, and the calibration must hold over temperature. With thermo-optic phase shifters at milliwatts each, hundreds of channels spend watts holding a beam in place; carrier-based or electro-optic shifters reduce the power at some cost in loss. And the fraction of optical power in the main lobe, after emitter efficiency and side lobes, sets the useful range of a lidar built on the array, which is the main comparison with the switched focal-plane arrays that compete for the steering front end of FMCW lidar.

References: J. Sun et al., Nature 493, 195 (2013); M. J. R. Heck, Nanophotonics 6, 93 (2017); C. V. Poulton et al., Opt. Lett. 42, 4091 (2017).