Optimization of a programmable λ/2-pitch optical phased array

Abstract

A challenge in optical phased arrays (OPAs) is to achieve single-lobe emission using densely spaced emitters without incurring inter-waveguide optical crosstalk. Here, we propose to heuristically optimize the amplitude and phase of each grating antenna in an OPA to correct for optical non-idealities, including fabrication variations and inter-waveguide crosstalk. This method was applied to a silicon photonic integrated circuit with 1 mm-long gratings at 775 nm spacing for operation in a wavelength range of 1450-1650 nm. We achieved a wide two-dimensional beam-steering range of 110° × 28°, evaluated over a 127° × 47° field-of-view (FOV). Within this FOV, we measured an average sidelobe suppression of 8.2 dB and focused on average, 34.5 % of the emitted power into the main lobe. We achieved a peak sidelobe suppression of 14.5 dB and 50 % of the power concentrated in the main lobe. The approach is suitable for applications that require alias-free out-of-plane emission.

Keywords: crosstalk; optical phased array; optimization; programmable photonics; silicon photonics.

Figure 1:

OPA circuit design. (a) Schematic of the OPA. (b) Annotated micrograph of the fabricated PIC: (i) array of variable optical attenuators (VOAs), (ii) waveguide fan-in region, and (iii) emitter aperture. (c) Measured transmission of a VOA at λ = 1550 nm as a function of the applied power. (d) Measured far-field of an isolated grating antenna at λ = 1450 nm.

Figure 2:

Experimental setup. (a) Schematic of the measurement setup. A rotating Fourier imaging system captures the far-field emission of the packaged OPA chip. The imaging system consists mainly of an infrared camera and a 20× objective lens (NA = 0.4). The hardware specifications are described in the main text. (b) Photograph of a packaged OPA. The PIC is wire-bonded to a printed circuit board with custom driver circuits.

Figure 3:

Measurement results for 9 steering angles. (a) Superimposed far-field images for different optimized angles in φ-axis. Each beam is normalized to its maximum intensity (b) Top: OPA emission before optimization. Bottom: Elliptical line cuts of the optimized solutions shown in (a). (c) Figures of merit (FOM) at each iteration for forming a beam at 0° without prior knowledge of existing beam solutions. A multi-objective genetic algorithm (GA) with a population size of 50 and a generalized pattern search algorithm (GPS) were used for the optimization. (d) FOM at each iteration for forming a beam at 55° with the benefit of knowing the optimized values for a solution at 50°. Only the pattern search algorithm was used for the optimization. For the various optimized angles measured: (e) FWHM, (f) sidelobe levels (g) % beam power: power in the main beam as a percentage of the total radiated power (h) total VOA loss: total radiated power normalized to the maximum observed radiated power across all measurements.

Figure 4:

Analysis of measurement results. (a) Comparison of beam formed at 0° with diffraction pattern of a uniformly illuminated rectangular aperture. (b) Fifteen consecutive trials in which the OPA was alternately programmed using two sets of optimized coefficients, {a _n } for φ = 0° and φ = 12.5°. (c) Left: Captured near-field emission of the OPA when the beam is steered to 0° in φ. Right: Vertical line cuts of the top image after propagating different distances along the grating axis. The line cuts show a persistent near-Gaussian power distribution.

Figure 5:

Demonstration of 2D beam-steering. Top: Superimposed far-field images for  different optimized angles in φ-axis and  θ-axis. To steer in θ, the wavelength is tuned from 1450 to 1600 nm. Bottom: Line cuts in θ of the beams in the image above.