End-fire silicon optical phased array with half-wavelength spacing

A multitude of applications depend on the ability to quickly and accurately steer a laser beam, ranging from fiber endoscopes in biological imaging to light detection and ranging (LiDAR) scanning in large-scale landscape surveying and autonomous automobile navigation. In addition, high-speed optical beam steering has the potential to impact a host of new technologies from free space optical communications to next-generation projectors. Mechanical approaches to beam steering require physical movement of some part of the system, which can range from as much as an entire laser module to as little as a single tilt mirror. Non-mechanical beam steering systems, such as optical phased arrays (OPAs), lack physical inertia and can thus outperform their mechanical counterparts.

Many notable demonstrations of OPAs make use of integrated silicon photonic platforms. The majority use emission structures such as grating couplers, scattering elements, or nano-antennas that emit light vertically out of the surface of the chip (known as the broadside geometry), including arrays consisting of up to 4096 passive elements¹ and apertures as large as 4 mm × 4 mm.² A variety of active devices capable of one-dimensional^3–5 and two-dimensional^6–12 steering have also been shown with beam angular divergence as small as 0.14° and up to 500 resolvable spots.¹² An alternative geometry for OPAs uses an edge-emitting (end-fire) array,^3,4,11,13 where light emits into the free space in the same plane of the chip via multimode interference couplers or waveguides terminated at the chip edge. This end-fire design can simplify OPA design and has the potential to reduce loss. Losses from reflections occurring at the interface between the waveguide and free space are typically <1 dB and can be further reduced through the use of anti-reflection coatings.

Existing OPAs in both the end-fire and broadside geometries have a major drawback in that they do not achieve emitting element spacing of λ/2 due to the size of each individual emitter. Vertical grating couplers used for broadside emission are fundamentally larger than the wavelength of light due to their periodic nature. In end-fire approaches, the limitation in sub-λ spacing arises from waveguides having mode sizes larger than the wavelength of light. Such large elements in existing arrays cannot physically be spaced at λ/2. Notably, if an OPA has an element spacing larger than λ/2, grating lobes will form in the beam pattern. Grating lobes are copies of the central lobe that appear at regular angular intervals and have deleterious effects on beam steering systems. First, grating lobes cause aliasing due to the presence of more than one beam thereby limiting the angular steering range of the system to ±1/2 the angular interval at which the first grating lobe appears. While a diverging lens can expand this steering range, it will in turn increase the angular divergence of the beam. The presence of grating lobes also indicates a reduction in power in the central lobe. Lastly, grating lobes can have harmful effects in applications such as free space optical communications in which they can facilitate eavesdropping.¹⁴ 

Certain design choices can mitigate some of the negative effects of grating lobes. For example, non-uniform spacing of elements in an OPA can reduce the peak amplitude of grating lobes by spreading their power over a wider angular area thereby enabling aliasing-free steering.^6,12 Second, using low-loss waveguides in materials such as silicon nitride can greatly improve the power capabilities of the array,^10,15 which can offset the power lost to grating lobes. However, the lower refractive index of silicon nitride waveguides results in larger mode profiles, thereby increasing emitter spacing and further narrowing angular steering range. Here we demonstrate an array with λ/2 emitter spacing that can fundamentally avoid grating lobes and their deleterious effects while simultaneously maximizing steering range.

In order to realize an OPA with λ/2 element spacing, we investigate a one-dimensional edge-emitting array of silicon waveguides clad in silicon dioxide. The high index contrast between the silicon core and silicon dioxide cladding of the waveguides results in light being highly confined to the core of the waveguide, thereby allowing dense waveguide spacing at the emission output of the chip. Beam steering can be achieved through the integration of active phase shifting in each waveguide. An overview of this scheme is illustrated in Fig. 1. Such a geometry is promising as a platform for high-speed and high-power beam steering.¹⁶ A one-dimensional array whose beam forms a stripe rather than a spot and is able to be steered in one dimension can prove useful in applications such as light-sheet microscopy¹⁷ and one-dimensional imaging applications such as bar-code scanning and flow microscopy.¹⁸ 

As an initial investigation, we pursue beam forming using a passive array of end-fire waveguides¹⁶ operating at a wavelength of 1550 nm. We then image the OPA’s far-field interference patterns and calculate the angular central lobe FWHM. A comparison of our simulated and measured data, as well as examples of the imaged far-field patterns, is shown in supplementary material. The largest array, with 16 emitters spaced at 900 nm pitch forms a beam with a measured central lobe of ∼7° at FWHM. An external heat source is applied to the passive array to generate a phase shift between elements via the thermo-optic effect, and the resulting steering is observed. For the smallest device with only two waveguides, this yields 17° of steering.¹⁶ 

Though these devices are successful in establishing the capability of end-fire OPAs composed of high-confinement waveguides, they are passive thereby limiting their steering capability and in addition, their element spacing was 16% larger than λ/2. This latter point means that grating lobes would still be present in their far-field beam patterns, which would impose a fundamental limitation on the steering range of an active version of the device.

As a result, we investigate beam steering with a one-dimensional array yielding five active waveguide elements¹⁹ as shown in Fig. 2. The fabrication details are discussed in the supplementary material. Each element of the array takes the form of a silicon waveguide clad in silicon dioxide, with a width of 500 nm, a height of 220 nm, and emitter pitch of 775 nm (corresponding to λ/2 when λ = 1550 nm). Similar to our passive array,¹⁶ the waveguides in this array emit light into the free space in the end-fire configuration from terminations at the edge of the chip.¹⁹ We calculate the effective refractive index of the TE-like mode of the waveguide to be 2.445, which results in a reflectance of 0.176 for normal incidence on the interface between waveguide and free space (where the waveguide terminates at the edge of the chip) or a reflection loss of 0.84 dB. We simulate the cross talk between these waveguides to be −13 dB (More details are given in the supplementary material). To the best of our knowledge, our active 1D array represents the first demonstrated OPA with λ/2 element spacing.

The experimental setup is shown in Fig. 3(a). The size, focal length, and numerical aperture of the objective lens in this setup limit the viewable angular range to approximately 60°. Tilting the lens up to 15° off-center while still focusing on the device output enables the extension of the viewing range. By splicing multiple images captured at different angles together, we reach a total measurable range of 90°. Further details on the experimental setup can be found in the supplementary material. We initially image the output of a single waveguide to acquire the element factor and compare this to the simulated element factor in Fig. 4. We see that the simulated element factor has a total width at half maximum of 82° compared to the approximately 70° shown by the measured element factor. The discrepancy between simulation and measured data is likely due to imperfections or roughness on the waveguide output facet. This gives the device an expected total angular steering range of roughly 70°, after which the beam magnitude would fall below half of its value when centered. Our simulated vertical element factor corresponds to a FWHM of 98°. Cross talk between adjacent elements is experimentally measured to be −11 dB, consistent with our simulated cross talk as discussed in Sec. II and in the supplementary material. With all five waveguides emitting light, the resulting beam has an angular divergence with FWHM of 17°. Angular beam divergence depends on the relationship between the wavelength and physical array size, with larger arrays forming narrower beams.

To measure active steering, first we externally equalize the power in each element. We set each heater to a starting voltage and then increase and decrease the applied voltage on either side of the array center to create an unwrapped phase ramp at the device output. We define the steering range for this device to be the angular range beyond which the peak intensity of the steered central lobe falls to less than half its maximum value (maximum occurs at the center, consistent with the element factor). This results in a maximum measured steering of ±32° for a total angular range of 64°, as shown in Fig. 3(b). The steering range is thereby limited by the element factor of the array since the intensity of a steered beam depends on the intensity of light emitted by the individual elements in that direction. Conversely, side lobes which are small when the beam is steered toward the center increase in intensity when the beam is steered, as they shift into areas where the element factor is increased. Other investigations, such as beam quality and temperature stability, are discussed in the supplementary material as well as a comparison of the performance of the OPA presented in this work to that of other recent studies as shown in Table II.

Here we demonstrate a one-dimensional OPA device on an integrated silicon photonic platform capable of edge emission of light in the plane of the chip with apertures spaced at λ/2 and steering up to 64°. This shows the large native steering range enabled by using λ/2 element spacing in end-firing waveguide OPAs. Increased resolution can be achieved in future work by increasing the number of waveguide emitters. Larger array sizes with a corresponding increased aperture size would make the device suitable for receiver as well as transmitter applications. Furthermore, tailoring the waveguide output geometry can improve element factor, thereby further expanding the steering range. This one-dimensional OPA has potential applications in one-dimensional imaging such as barcode reading, flow microscopy, light sheet microscopy, and certain LiDAR applications. Three-dimensional integration of such one-dimensional arrays, either by deposition of multiple layers of amorphous materials or by bonding wafers, can enable full two-dimensional phased arrays using this same basic geometry. This could open many possible applications, particularly in free-space optical communications²⁰ and 3D optical interconnects.²¹ 

See supplementary material for more details on our passive OPA, device fabrication, experimental setup and results, temperature stability, and beam quality.