Characterization of SiN Integrated Optical Phased Arrays on a Wafer-Scale Test Station

Summary

Here, we describe the operation of a SiN integrated photonic circuit containing optical phased arrays. The circuits are used to emit low divergence laser beams in the near infrared and steer them in two dimensions.

Abstract

Optical phased arrays (OPAs) can produce low-divergence laser beams and can be used to control the emission angle electronically without the need for moving mechanical parts. This technology is particularly useful for beam steering applications. Here, we focus on OPAs integrated into SiN photonic circuits for a wavelength in the near infrared. A characterization method of such circuits is presented, which allows the output beam of integrated OPAs to be shaped and steered. Furthermore, using a wafer-scale characterization setup, several devices can easily be tested across multiple dies on a wafer. In this way, fabrication variations can be studied, and high-performance devices identified. Typical images of OPA beams are shown, including beams emitted from OPAs with and without a uniform waveguide length, and with varying numbers of channels. In addition, the evolution of output beams during the phase optimization process and beam steering in two dimensions is presented. Finally, a study of the variation in the beam divergence of identical devices is performed with respect to their position on the wafer.

Introduction

Optical phased arrays (OPAs) are advantageous due to their ability to shape and steer optical beams nonmechanically - this is useful in a broad range of technological applications such as light detection and ranging (LIDAR), free space communication and holographic displays¹. The integration of OPAs in photonic circuits is of particular interest, as it provides a low cost solution for their fabrication with a small physical footprint. Integrated OPAs have been successfully demonstrated using a number of different material systems including InP, AlGaAs and silicon^2,3,4. Of these systems, silicon photonics is perhaps the most convenient, due to its high refractive index contrast and compatibility with CMOS⁵. Indeed, OPA circuits have been extensively demonstrated in the silicon-on-insulator platform^6,7,8,9,10; however, the application of these circuits is limited both by the wavelength transparency window of silicon and the high nonlinear losses, which lead to a limit on the available output optical power. We focus instead on OPAs integrated in SiN, a material with similar properties to silicon in terms of CMOS capability and footprint size^11,12. In contrast to silicon however, SiN is expected to be suitable for a larger range of applications since the transparency window is broader, down to at least 500 nm, and thanks to the possibly high optical power thanks to the relatively low nonlinear losses.

The principals of OPA integration have recently been demonstrated using SiN^8,13,14. Here, we will extend these principals to demonstrate a method of characterizing and operating integrated OPAs for two dimensional beam steering. In comparison to previous demonstrations of beam steering in two dimensions that rely on the tuning of the wavelength⁶, our circuit can operate at a single wavelength. We first provide a brief overview of the operating principles behind OPAs. This is followed by an introduction to the circuits used in this work. Finally, the characterization method is described and typical images of OPA output beams presented and discussed.

OPAs are composed of an array of closely spaced emitters that can be addressed individually to control the optical phase. If a linear phase relationship exists across the emitter array, the interference pattern in the far field yields several clearly separated maxima - similar to the principles of multi-slit interference. By controlling the magnitude of the phase difference, the position of the maxima can be adjusted, and hence, beam steering performed. In integrated OPAs, emitters consist of closely spaced diffraction gratings where the light is scattered and emitted out of the chip plane. A schematic illustration of an integrated OPA device is shown in Figure 1A,B. Light is coupled into the chip, in this case via an optical fiber, and is then divided into multiple channels, each containing an integrated phase shifter. At the other end of the optical circuit, the waveguides terminate in gratings and combine to form the OPA. The resulting output beam is comprised of multiple interference maxima, the brightest of which is referred to as the fundamental lobe and is the one most often used in beam steering applications. The emission direction of the fundamental lobe is defined by the two azimuthal angles to the orthogonal projection of the chip plane, φ and θ, perpendicular and parallel to the orientation of the grating respectively. In this document, φ and θ will be referred to as the 'perpendicular' and 'parallel' emission angles, respectively. The perpendicular angle φ is determined by the phase difference between the OPA channels, and the parallel angle θ depends on the period of the output gratings.

Our integrated circuits are fabricated using Si₃N₄ waveguides with a cross section of 600 x 300 nm², a design that was optimized for the fundamental transverse electric polarization mode of light at a wavelength of 905 nm. Underneath the waveguides lies a 2.5 µm SiO₂ buffer layer on top of a silicon wafer. The thermal phase shifters were made from a 10(100) nm thick Ti(TiN) layer used to form 500 µm long and 2 µm wide resistive wires. In our circuits, an electric power of 90 mW is required to achieve a phase shift of π. The OPA output gratings consist of 750 fully etched periods with a nominal filling factor of 0.5 and a grating period between 670 nm and 700 nm. Further information on the platform design and fabrication is given in Tyler et al.^15,16.

In this work, two different types of circuits are characterized, a passive circuit without phase shifting capabilities, and a more complex circuit, designed to perform beam steering in two dimensions. The two dimensional beam steering circuit is shown in Figure 2. Figure 2A contains a schematic of the circuit and Figure 2B shows a microscope image of the fabricated device. The light enters the circuit at the input grating. It then reaches a switching network where it can be routed selectively towards one of four sub-circuits. Each sub-circuit splits the light into four channels using multimode interference devices (MMI). The channels each contain a thermal phase shifter and form an OPA at the end of the circuit. The four OPAs originating from the four sub-circuits each comprise a different grating period between 670 nm and 700 nm. These periods correspond to azimuthal angles parallel to the grating axis, θ, between 7° and 10°. A more detailed description on the circuit can be found in Tyler et al.¹⁶.

The presented characterization setup is based on an automated probing station capable of performing a series of measurements on many circuits across a whole wafer. This enables study of the performance variation relative to the position on the wafer and to select the devices with the optimum properties. However, the use of a prober station implies some physical constraints to the OPA characterization scheme due to the relatively small available space above the wafer. The characterization of optical phased arrays requires imaging the OPA output in the far field, which can be performed in a number of ways. For example, a series of lenses may be used in a Fourier imaging system⁶ or the farfield image formed on a Lambertian surface may be viewed in either reflection or transmission. For our system, we chose what we considered to be the simplest and most compact solution of placing a large surface 35 mm x 28 mm CMOS sensor without lenses placed approximately 50 mm above the wafer surface. Despite the increased cost of such a large CCD sensor, this solution allows a sufficient field of view without the use of lenses.

Protocol

1. Preparations

1. Prepare the following experimental setup (Figure 4).
   1. Use a computer.
   2. Use a continuous wave fiber coupled laser source. Depending on the circuit losses, 1 mW power suffices. In the presented characterization setup, the laser source is at a wavelength of 905 nm.
   3. Use a polarization controller adapted for the laser wavelength.
   4. Use a cleaved input fiber to couple light into the input grating coupler of the optical circuit.
   5. Use an electric probe to connect the electronic control board to the electrical contact of the optical circuit.
   6. Use a system capable of controlling the 20 phase modulators of the two dimensional beam steering circuit is required. In the presented characterization setup, this system is a custom electronic board controlled by an Arduino, which is able to apply individually between 0 and 200 mW of electric power at the phase shifters on the optical circuit. A schematic of the electrical circuit is shown in Figure 3. For each channel, the circuit contain a DAC (Digital to Analog Converter) that will translate the digital command voltage to an analog voltage that controls the gate of a high-power transistor. The heater is connected to a high-power current source. Therefore, by controlling the gate tension, the current flow in the heater can be adjusted.
   7. Use a bare image sensor to image the far field of the optical output. In the presented characterization setup, the camera is a 35 mm CCD sensor.
   8. Use a optical microscope in order to image the chip for alignment purposes.
   9. Use a 3-axis translation stage and mount to fit a 200 mm wafer. In the presented characterization setup, this stage is a reconfigurable probe system for silicon photonics.
2. Equipment assembly
   1. Assemble the equipment according to Figure 4 and mount the wafer. The distance between wafer and the sensor needs to be chosen small enough to ensure a high-resolution image of the output beam, but large enough to fit at least two interference maxima in order to be able to find the relationship between sensor pixels and output angle as will be explained in section 4 of the protocol.
   2. Ensure that the sensor and the wafer are parallel; otherwise, it may falsify the computation of the pixel/output angle computation. In the presented characterization setup, set the wafer-sensor distance to 5 cm. If a double sensor configuration is used (like the one presented here), ensure that the bare sensor can easily be removed to give access to the optical microscope in order to image the near-field for fiber alignment purposes.
   3. Be sure that the electrical probe, camera and optical fiber are not touching each other. Connect required elements to a computer. In the presented setup the probe station, the CCD sensor and the electrical circuit for the phase control are driven via a computer and a Python program in order to automate the measurement process.

2. Optical coupling

1. Fiber alignment
   1. Using the microscope, start by carefully lowering the fiber until it touches the wafer surface (away from the input grating coupler to avoid damaging it), and then move it up about 20 µm.
   2. When this is done, maximize the light intensity at the output gratings. To do so, start sweeping the fiber position over the OPA input grating coupler. If the camera attached to the microscope is responsive to the laser wavelength (if not use the bare image sensor), and if the fiber and grating coupler are well aligned, light exiting at the OPA output gratings should be visible on the image. An example can be seen in Figure 5A.
   3. When light is seen from the OPA antennas, adjust the polarization in order to maximize the light intensity at the output gratings. Be sure to avoid any movement or vibration of the input fiber
2. OPA output imaging
   1. Switch to the far-field imaging sensor and improve the image quality: Adjust both the exposure time of the sensor and the laser power in such a way that the OPA output is clearly visible on the camera and the beam does not saturate the sensor. An example image recorded by the sensor is shown in Figure 5B.
   2. If necessary, cover the setup so that the background light does not interfere with the image from the OPA beam. Generally, the weaker the background light, the lower the laser power that can be set.
   3. Block the reflections by placing a highly reflecting sheet between the reflection and the camera. Sometimes, reflections originating from the wafer surface reach the sensor area and contaminate the image of the OPA output (reflections can happen at the input grating).
   4. Readjust the polarization of the input light to obtain a clear image.

3. Beam optimization and steering

NOTE: This section describes the operation of the circuit shown in Figure 2 and how it can be used to perform beam steering in two dimensions.

1. Preparations
   1. Connect the electric circuit for the phase control to a multi-channel electric probe.
   2. Using the microscope, connect the pins of the electric probe to the metal contact pads of the optical circuit.
   3. Re-optimize the position of the input fiber.
   4. Switch to the far field sensor and image the output.
2. Selection of the parallel emission angle θ using the switching network
   1. Study the ring resonators of the switching network in order to control the emission angle in θ. For this purpose, observe the far field image of the output while varying the voltages applied to the phase shifters at the ring resonators. With the correct voltage applied to each resonator, a different area on the sensor will be illuminated, corresponding to a certain θ value, as shown in Figure 6B.
   2. Find the voltages where the rings are on- and off-resonance. For this purpose, an automated script can be used to sweep the resonator voltages and record the intensities on the different θ areas on the sensor. Use found voltages to access the various sub-circuits and to steer the output beam in θ.
3. Selection of the orthogonal emission angle φ by optimizing the OPA phases
   1. Optimize the OPA phases in order to shape and steer the output beam in φ. For this purpose, select a small pixel area (corresponding to the desired φ angle) that should be illuminated with a focused output beam.
   2. Maximize the brightness inside the chosen area by running the following the optimization routine.
      1. Shift the phase of one of the OPA channels in small increments. After each shift, record the integral of the brightness in the pixel area inside, I_i, and outside, I_o, of the selected area. Calculate the ratio R = I_i / I_o. After a full phase shift cycle between 0 and 2π, apply the phase shift with the highest recorded brightness ratio R.
      2. Repeat this phase optimization process on the next OPA channel. Different optimization algorithms may be used, such as a hill climbing.
      3. Repeat the optimization process by optimizing the phases until the optimization process is saturated and a focused output beam is visible. Example images of the output beam taken during an optimization process is shown in Figure 6A. After 16 optimization rounds, the output beam a focused beam is visible.
         NOTE: If some additional unexpected peaks are present, this may be a result of a temporally unstable coupling into the circuit during the optimization process. This may be due to movement of the input fiber and/or an unstable polarization state.
   3. In order to steer the output beam to a different φ angle, select a new pixel area and repeat the optimization process.

4. Beam divergence measurements and image analysis

1. Image acquisition
   1. Optimize the position of the input fiber. Record the image of the output in the far-field. Make sure that at least two clear interference maxima are visible.
   2. Using the alignment system, move the wafer in order to align the next device to the input fiber. Perform fine alignment by maximizing the output intensity recorded by the camera. Record output image.
   3. Repeat the above step until all devices of interest have been characterized. If the selected optical circuit has the capability of phase adjustment of the OPA channels, perform a phase optimization routine before recording the images.
2. Image analysis
   1. Check the recorded images for false data points arising from defective pixels, such as dead or hot pixels. Erase these data points or replace the values by typical values.
   2. Correlate the CCD pixels to OPA output angles φ and θ as follows.
      1. Calculate the angular distance Δφ between the interference maxima according to the OPA design using Δφ = sin^-1(λ/d) [°], where λ is the wavelength and d is the lateral pitch between the OPA gratings. Fit two Gaussian curves to the two interference maxima and determine the positions of the two centers, P₁ and P₂. Since the distance (in pixels) between the two centers, N = P₂ - P₁, is expected to correspond to Δφ, we obtain a conversion factor c between pixel and angle c = Δφ/N [°/pixel], which can be used to obtain a relative angle relationship between pixels.
      2. Obtain the conversion factor, c, via an accurate measurement of the distance between the wafer surface and the sensor, and the pixel size (5.5*5.5µm for the sensor used here).
      3. Estimate the absolute output angles in φ and θ for one of the CCD pixels. Set the beam center in θ to the expected emission angle according to simulations. In order to choose the absolute value in φ, optimize the beam for several angles in φ by adjusting the OPA phases, and record the intensity of the main lobe for each angle. According to the OPA theory, the main lobe is most intense (and the intensity in the side lobes minimized) when emitting at φ = 0°. Hence, set the pixel in the center of the beam with the maximum recorded beam intensity, to φ = 0°. Use this pixel and the conversion factor to assign absolute angles to all pixels of the image.
      4. In the case of an output beam with significant tilt with respect to the vertical axis, and if the beam divergence and position must be measured very accurately, tilt the camera in order to be perfectly perpendicular to the output beam. Otherwise, it is also possible to apply a correction factor to the measured beam size by computing the projection of the beam on the sensor depending on the angle between the output beam and camera plane.
3. Calculation of the beam divergence
   1. Extract cross sections across the center of the fundamental beam along φ and θ.
   2. Fit two Gaussian curves to the cross sections and extract the full-width-at-half-maxima as a measure for the beam divergence φ_div and θ_div.
   3. Calculate the expected beam width φ_div = λ/Nd [°], where λ is the wavelength and d the lateral distance between the OPA gratings.
   4. Estimate the beam divergence θ_div by performing FDTD simulations of the output gratings.
4. Automatic testing
   1. If the characterization bench (as the one presented here) can perform automated measurements, perform some additional steps. First, obtain the chip dimensions and the coordinates of the measured structures from the circuit layout. Then, input those values to the bench control software. Therefore, once the input fiber have been aligned on the first tested structure (as detailed in section 2.1), the bench can switch automatically from one structure to another via a translation of the wafer.

Discussion

We have presented a method to characterize an integrated OPA. The main advantage of the method is the ability to easily probe multiple dies across a wafer, to look for fabrication variations and to identify high-performance devices. This can be seen in Figure 8B. From the wafer scan, it becomes clear that the lower half of the wafer exhibits devices with lower beam divergences. This could be explained by a higher waveguide quality in that area, which reduces random phase shifts and hence the beam divergence.

Using a large area CCD sensor to image the far field output is a convenient method to image the free space output of integrated circuits, since it can easily be added to most characterization set-ups due to their compact size in comparison to the often used, bulkier, Fourier-imaging systems⁶.

In order to guarantee a high accuracy of the beam angle and divergence measurement, particular care must be taken during the camera - OPA alignment. Moreover, the OPA response is sensitive to phase and polarization instabilities during calibration. Therefore, all sources of perturbation must be controlled: movement/vibration of the injection fiber, laser temperature, incoming light polarization etc.

In summary, a method to characterize integrated OPAs was presented. Details on how to couple light, how to control phase shifters in the circuit and how to image the output in the near and the far field were given. Typical images of the output beams of several OPA circuits were shown, including the results of beam steering in two dimensions at a single wavelength in the near infrared. Furthermore, we show the results of measuring multiple devices with the same design across a wafer in terms of beam divergence. A performance trend with respect to the position on the wafer was found, identifying areas with high-quality fabrication properties.

Materials

Name	Company	Catalog Number	Comments
25 ch electrical Probe	Cascade Microtech	InfinityQuad 25ch
35 mm CCD sensor	Allied Vision	Prosilica GT 6600
Arduino uno	Arduino	A100066
laser	Qphotonics	QFLD-905-10S
optical fibre	Corning	HI780
polarization controller	ThorLabs	FPC023
prober station	Cascade Microtech	Elite 300